General Introduction ******************** This file documents `awk', a program that you can use to select particular records in a file and perform operations upon them. Copyright (C) 1989, 1991, 1992, 1993, 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003 Free Software Foundation, Inc. This is Edition 3 of `GAWK: Effective AWK Programming: A User's Guide for GNU Awk', for the 3.1.4 (or later) version of the GNU implementation of AWK. Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation; with the Invariant Sections being "GNU General Public License", the Front-Cover texts being (a) (see below), and with the Back-Cover Texts being (b) (see below). A copy of the license is included in the section entitled "GNU Free Documentation License". a. "A GNU Manual" b. "You have freedom to copy and modify this GNU Manual, like GNU software. Copies published by the Free Software Foundation raise funds for GNU development." To Miriam, for making me complete. To Chana, for the joy you bring us. To Rivka, for the exponential increase. To Nachum, for the added dimension. To Malka, for the new beginning. Short Contents ************** General Introduction Foreword Preface 1 Getting Started with `awk' 2 Regular Expressions 3 Reading Input Files 4 Printing Output 5 Expressions 6 Patterns, Actions, and Variables 7 Arrays in `awk' 8 Functions 9 Internationalization with `gawk' 10 Advanced Features of `gawk' 11 Running `awk' and `gawk' 12 A Library of `awk' Functions 13 Practical `awk' Programs Appendix A The Evolution of the `awk' Language Appendix B Installing `gawk' Appendix C Implementation Notes Appendix D Basic Programming Concepts Glossary GNU General Public License GNU Free Documentation License Index Table of Contents ***************** General Introduction Foreword Preface History of `awk' and `gawk' A Rose by Any Other Name Using This Book Typographical Conventions The GNU Project and This Book How to Contribute Acknowledgments 1 Getting Started with `awk' 1.1 How to Run `awk' Programs 1.1.1 One-Shot Throwaway `awk' Programs 1.1.2 Running `awk' Without Input Files 1.1.3 Running Long Programs 1.1.4 Executable `awk' Programs 1.1.5 Comments in `awk' Programs 1.1.6 Shell-Quoting Issues 1.2 Data Files for the Examples 1.3 Some Simple Examples 1.4 An Example with Two Rules 1.5 A More Complex Example 1.6 `awk' Statements Versus Lines 1.7 Other Features of `awk' 1.8 When to Use `awk' 2 Regular Expressions 2.1 How to Use Regular Expressions 2.2 Escape Sequences 2.3 Regular Expression Operators 2.4 Using Character Lists 2.5 `gawk'-Specific Regexp Operators 2.6 Case Sensitivity in Matching 2.7 How Much Text Matches? 2.8 Using Dynamic Regexps 2.9 Where You Are Makes A Difference 3 Reading Input Files 3.1 How Input Is Split into Records 3.2 Examining Fields 3.3 Nonconstant Field Numbers 3.4 Changing the Contents of a Field 3.5 Specifying How Fields Are Separated 3.5.1 Using Regular Expressions to Separate Fields 3.5.2 Making Each Character a Separate Field 3.5.3 Setting `FS' from the Command Line 3.5.4 Field-Splitting Summary 3.6 Reading Fixed-Width Data 3.7 Multiple-Line Records 3.8 Explicit Input with `getline' 3.8.1 Using `getline' with No Arguments 3.8.2 Using `getline' into a Variable 3.8.3 Using `getline' from a File 3.8.4 Using `getline' into a Variable from a File 3.8.5 Using `getline' from a Pipe 3.8.6 Using `getline' into a Variable from a Pipe 3.8.7 Using `getline' from a Coprocess 3.8.8 Using `getline' into a Variable from a Coprocess 3.8.9 Points to Remember About `getline' 3.8.10 Summary of `getline' Variants 4 Printing Output 4.1 The `print' Statement 4.2 Examples of `print' Statements 4.3 Output Separators 4.4 Controlling Numeric Output with `print' 4.5 Using `printf' Statements for Fancier Printing 4.5.1 Introduction to the `printf' Statement 4.5.2 Format-Control Letters 4.5.3 Modifiers for `printf' Formats 4.5.4 Examples Using `printf' 4.6 Redirecting Output of `print' and `printf' 4.7 Special File Names in `gawk' 4.7.1 Special Files for Standard Descriptors 4.7.2 Special Files for Process-Related Information 4.7.3 Special Files for Network Communications 4.7.4 Special File Name Caveats 4.8 Closing Input and Output Redirections 5 Expressions 5.1 Constant Expressions 5.1.1 Numeric and String Constants 5.1.2 Octal and Hexadecimal Numbers 5.1.3 Regular Expression Constants 5.2 Using Regular Expression Constants 5.3 Variables 5.3.1 Using Variables in a Program 5.3.2 Assigning Variables on the Command Line 5.4 Conversion of Strings and Numbers 5.5 Arithmetic Operators 5.6 String Concatenation 5.7 Assignment Expressions 5.8 Increment and Decrement Operators 5.9 True and False in `awk' 5.10 Variable Typing and Comparison Expressions 5.11 Boolean Expressions 5.12 Conditional Expressions 5.13 Function Calls 5.14 Operator Precedence (How Operators Nest) 6 Patterns, Actions, and Variables 6.1 Pattern Elements 6.1.1 Regular Expressions as Patterns 6.1.2 Expressions as Patterns 6.1.3 Specifying Record Ranges with Patterns 6.1.4 The `BEGIN' and `END' Special Patterns 6.1.4.1 Startup and Cleanup Actions 6.1.4.2 Input/Output from `BEGIN' and `END' Rules 6.1.5 The Empty Pattern 6.2 Using Shell Variables in Programs 6.3 Actions 6.4 Control Statements in Actions 6.4.1 The `if'-`else' Statement 6.4.2 The `while' Statement 6.4.3 The `do'-`while' Statement 6.4.4 The `for' Statement 6.4.5 The `switch' Statement 6.4.6 The `break' Statement 6.4.7 The `continue' Statement 6.4.8 The `next' Statement 6.4.9 Using `gawk''s `nextfile' Statement 6.4.10 The `exit' Statement 6.5 Built-in Variables 6.5.1 Built-in Variables That Control `awk' 6.5.2 Built-in Variables That Convey Information 6.5.3 Using `ARGC' and `ARGV' 7 Arrays in `awk' 7.1 Introduction to Arrays 7.2 Referring to an Array Element 7.3 Assigning Array Elements 7.4 Basic Array Example 7.5 Scanning All Elements of an Array 7.6 The `delete' Statement 7.7 Using Numbers to Subscript Arrays 7.8 Using Uninitialized Variables as Subscripts 7.9 Multidimensional Arrays 7.10 Scanning Multidimensional Arrays 7.11 Sorting Array Values and Indices with `gawk' 8 Functions 8.1 Built-in Functions 8.1.1 Calling Built-in Functions 8.1.2 Numeric Functions 8.1.3 String-Manipulation Functions 8.1.3.1 More About `\' and `&' with `sub', `gsub', and `gensub' 8.1.4 Input/Output Functions 8.1.5 Using `gawk''s Timestamp Functions 8.1.6 Bit-Manipulation Functions of `gawk' 8.1.7 Using `gawk''s String-Translation Functions 8.2 User-Defined Functions 8.2.1 Function Definition Syntax 8.2.2 Function Definition Examples 8.2.3 Calling User-Defined Functions 8.2.4 The `return' Statement 8.2.5 Functions and Their Effects on Variable Typing 9 Internationalization with `gawk' 9.1 Internationalization and Localization 9.2 GNU `gettext' 9.3 Internationalizing `awk' Programs 9.4 Translating `awk' Programs 9.4.1 Extracting Marked Strings 9.4.2 Rearranging `printf' Arguments 9.4.3 `awk' Portability Issues 9.5 A Simple Internationalization Example 9.6 `gawk' Can Speak Your Language 10 Advanced Features of `gawk' 10.1 Allowing Nondecimal Input Data 10.2 Two-Way Communications with Another Process 10.3 Using `gawk' for Network Programming 10.4 Using `gawk' with BSD Portals 10.5 Profiling Your `awk' Programs 11 Running `awk' and `gawk' 11.1 Invoking `awk' 11.2 Command-Line Options 11.3 Other Command-Line Arguments 11.4 The `AWKPATH' Environment Variable 11.5 Obsolete Options and/or Features 11.6 Undocumented Options and Features 11.7 Known Bugs in `gawk' 12 A Library of `awk' Functions 12.1 Naming Library Function Global Variables 12.2 General Programming 12.2.1 Implementing `nextfile' as a Function 12.2.2 Converting Strings To Numbers 12.2.3 Assertions 12.2.4 Rounding Numbers 12.2.5 The Cliff Random Number Generator 12.2.6 Translating Between Characters and Numbers 12.2.7 Merging an Array into a String 12.2.8 Managing the Time of Day 12.3 Data File Management 12.3.1 Noting Data File Boundaries 12.3.2 Rereading the Current File 12.3.3 Checking for Readable Data Files 12.3.4 Checking For Zero-length Files 12.3.5 Treating Assignments as File Names 12.4 Processing Command-Line Options 12.5 Reading the User Database 12.6 Reading the Group Database 13 Practical `awk' Programs 13.1 Running the Example Programs 13.2 Reinventing Wheels for Fun and Profit 13.2.1 Cutting out Fields and Columns 13.2.2 Searching for Regular Expressions in Files 13.2.3 Printing out User Information 13.2.4 Splitting a Large File into Pieces 13.2.5 Duplicating Output into Multiple Files 13.2.6 Printing Nonduplicated Lines of Text 13.2.7 Counting Things 13.3 A Grab Bag of `awk' Programs 13.3.1 Finding Duplicated Words in a Document 13.3.2 An Alarm Clock Program 13.3.3 Transliterating Characters 13.3.4 Printing Mailing Labels 13.3.5 Generating Word-Usage Counts 13.3.6 Removing Duplicates from Unsorted Text 13.3.7 Extracting Programs from Texinfo Source Files 13.3.8 A Simple Stream Editor 13.3.9 An Easy Way to Use Library Functions Appendix A The Evolution of the `awk' Language A.1 Major Changes Between V7 and SVR3.1 A.2 Changes Between SVR3.1 and SVR4 A.3 Changes Between SVR4 and POSIX `awk' A.4 Extensions in the Bell Laboratories `awk' A.5 Extensions in `gawk' Not in POSIX `awk' A.6 Major Contributors to `gawk' Appendix B Installing `gawk' B.1 The `gawk' Distribution B.1.1 Getting the `gawk' Distribution B.1.2 Extracting the Distribution B.1.3 Contents of the `gawk' Distribution B.2 Compiling and Installing `gawk' on Unix B.2.1 Compiling `gawk' for Unix B.2.2 Additional Configuration Options B.2.3 The Configuration Process B.3 Installation on Other Operating Systems B.3.1 Installing `gawk' on an Amiga B.3.2 Installing `gawk' on BeOS B.3.3 Installation on PC Operating Systems B.3.3.1 Installing a Prepared Distribution for PC Systems B.3.3.2 Compiling `gawk' for PC Operating Systems B.3.3.3 Compiling `gawk' For Dynamic Libraries B.3.3.4 Using `gawk' on PC Operating Systems B.3.3.5 Using `gawk' In The Cygwin Environment B.3.4 How to Compile and Install `gawk' on VMS B.3.4.1 Compiling `gawk' on VMS B.3.4.2 Installing `gawk' on VMS B.3.4.3 Running `gawk' on VMS B.3.4.4 Building and Using `gawk' on VMS POSIX B.4 Unsupported Operating System Ports B.4.1 Installing `gawk' on the Atari ST B.4.1.1 Compiling `gawk' on the Atari ST B.4.1.2 Running `gawk' on the Atari ST B.4.2 Installing `gawk' on a Tandem B.5 Reporting Problems and Bugs B.6 Other Freely Available `awk' Implementations Appendix C Implementation Notes C.1 Downward Compatibility and Debugging C.2 Making Additions to `gawk' C.2.1 Adding New Features C.2.2 Porting `gawk' to a New Operating System C.3 Adding New Built-in Functions to `gawk' C.3.1 A Minimal Introduction to `gawk' Internals C.3.2 Directory and File Operation Built-ins C.3.2.1 Using `chdir' and `stat' C.3.2.2 C Code for `chdir' and `stat' C.3.2.3 Integrating the Extensions C.4 Probable Future Extensions Appendix D Basic Programming Concepts D.1 What a Program Does D.2 Data Values in a Computer D.3 Floating-Point Number Caveats Glossary GNU General Public License Preamble How to Apply These Terms to Your New Programs GNU Free Documentation License ADDENDUM: How to use this License for your documents Index Foreword ******** Arnold Robbins and I are good friends. We were introduced 11 years ago by circumstances--and our favorite programming language, AWK. The circumstances started a couple of years earlier. I was working at a new job and noticed an unplugged Unix computer sitting in the corner. No one knew how to use it, and neither did I. However, a couple of days later it was running, and I was `root' and the one-and-only user. That day, I began the transition from statistician to Unix programmer. On one of many trips to the library or bookstore in search of books on Unix, I found the gray AWK book, a.k.a. Aho, Kernighan and Weinberger, `The AWK Programming Language', Addison-Wesley, 1988. AWK's simple programming paradigm--find a pattern in the input and then perform an action--often reduced complex or tedious data manipulations to few lines of code. I was excited to try my hand at programming in AWK. Alas, the `awk' on my computer was a limited version of the language described in the AWK book. I discovered that my computer had "old `awk'" and the AWK book described "new `awk'." I learned that this was typical; the old version refused to step aside or relinquish its name. If a system had a new `awk', it was invariably called `nawk', and few systems had it. The best way to get a new `awk' was to `ftp' the source code for `gawk' from `prep.ai.mit.edu'. `gawk' was a version of new `awk' written by David Trueman and Arnold, and available under the GNU General Public License. (Incidentally, it's no longer difficult to find a new `awk'. `gawk' ships with Linux, and you can download binaries or source code for almost any system; my wife uses `gawk' on her VMS box.) My Unix system started out unplugged from the wall; it certainly was not plugged into a network. So, oblivious to the existence of `gawk' and the Unix community in general, and desiring a new `awk', I wrote my own, called `mawk'. Before I was finished I knew about `gawk', but it was too late to stop, so I eventually posted to a `comp.sources' newsgroup. A few days after my posting, I got a friendly email from Arnold introducing himself. He suggested we share design and algorithms and attached a draft of the POSIX standard so that I could update `mawk' to support language extensions added after publication of the AWK book. Frankly, if our roles had been reversed, I would not have been so open and we probably would have never met. I'm glad we did meet. He is an AWK expert's AWK expert and a genuinely nice person. Arnold contributes significant amounts of his expertise and time to the Free Software Foundation. This book is the `gawk' reference manual, but at its core it is a book about AWK programming that will appeal to a wide audience. It is a definitive reference to the AWK language as defined by the 1987 Bell Labs release and codified in the 1992 POSIX Utilities standard. On the other hand, the novice AWK programmer can study a wealth of practical programs that emphasize the power of AWK's basic idioms: data driven control-flow, pattern matching with regular expressions, and associative arrays. Those looking for something new can try out `gawk''s interface to network protocols via special `/inet' files. The programs in this book make clear that an AWK program is typically much smaller and faster to develop than a counterpart written in C. Consequently, there is often a payoff to prototype an algorithm or design in AWK to get it running quickly and expose problems early. Often, the interpreted performance is adequate and the AWK prototype becomes the product. The new `pgawk' (profiling `gawk'), produces program execution counts. I recently experimented with an algorithm that for n lines of input, exhibited ~ C n^2 performance, while theory predicted ~ C n log n behavior. A few minutes poring over the `awkprof.out' profile pinpointed the problem to a single line of code. `pgawk' is a welcome addition to my programmer's toolbox. Arnold has distilled over a decade of experience writing and using AWK programs, and developing `gawk', into this book. If you use AWK or want to learn how, then read this book. Michael Brennan Author of `mawk' Preface ******* Several kinds of tasks occur repeatedly when working with text files. You might want to extract certain lines and discard the rest. Or you may need to make changes wherever certain patterns appear, but leave the rest of the file alone. Writing single-use programs for these tasks in languages such as C, C++, or Pascal is time-consuming and inconvenient. Such jobs are often easier with `awk'. The `awk' utility interprets a special-purpose programming language that makes it easy to handle simple data-reformatting jobs. The GNU implementation of `awk' is called `gawk'; it is fully compatible with the System V Release 4 version of `awk'. `gawk' is also compatible with the POSIX specification of the `awk' language. This means that all properly written `awk' programs should work with `gawk'. Thus, we usually don't distinguish between `gawk' and other `awk' implementations. Using `awk' allows you to: * Manage small, personal databases * Generate reports * Validate data * Produce indexes and perform other document preparation tasks * Experiment with algorithms that you can adapt later to other computer languages In addition, `gawk' provides facilities that make it easy to: * Extract bits and pieces of data for processing * Sort data * Perform simple network communications This Info file teaches you about the `awk' language and how you can use it effectively. You should already be familiar with basic system commands, such as `cat' and `ls',(1) as well as basic shell facilities, such as input/output (I/O) redirection and pipes. Implementations of the `awk' language are available for many different computing environments. This Info file, while describing the `awk' language in general, also describes the particular implementation of `awk' called `gawk' (which stands for "GNU awk"). `gawk' runs on a broad range of Unix systems, ranging from 80386 PC-based computers up through large-scale systems, such as Crays. `gawk' has also been ported to Mac OS X, MS-DOS, Microsoft Windows (all versions) and OS/2 PCs, Atari and Amiga microcomputers, BeOS, Tandem D20, and VMS. ---------- Footnotes ---------- (1) These commands are available on POSIX-compliant systems, as well as on traditional Unix-based systems. If you are using some other operating system, you still need to be familiar with the ideas of I/O redirection and pipes. History of `awk' and `gawk' =========================== Recipe For A Programming Language 1 part `egrep' 1 part `snobol' 2 parts `ed' 3 parts C Blend all parts well using `lex' and `yacc'. Document minimally and release. After eight years, add another part `egrep' and two more parts C. Document very well and release. The name `awk' comes from the initials of its designers: Alfred V. Aho, Peter J. Weinberger and Brian W. Kernighan. The original version of `awk' was written in 1977 at AT&T Bell Laboratories. In 1985, a new version made the programming language more powerful, introducing user-defined functions, multiple input streams, and computed regular expressions. This new version became widely available with Unix System V Release 3.1 (SVR3.1). The version in SVR4 added some new features and cleaned up the behavior in some of the "dark corners" of the language. The specification for `awk' in the POSIX Command Language and Utilities standard further clarified the language. Both the `gawk' designers and the original Bell Laboratories `awk' designers provided feedback for the POSIX specification. Paul Rubin wrote the GNU implementation, `gawk', in 1986. Jay Fenlason completed it, with advice from Richard Stallman. John Woods contributed parts of the code as well. In 1988 and 1989, David Trueman, with help from me, thoroughly reworked `gawk' for compatibility with the newer `awk'. Circa 1995, I became the primary maintainer. Current development focuses on bug fixes, performance improvements, standards compliance, and occasionally, new features. In May of 1997, Ju"rgen Kahrs felt the need for network access from `awk', and with a little help from me, set about adding features to do this for `gawk'. At that time, he also wrote the bulk of `TCP/IP Internetworking with `gawk'' (a separate document, available as part of the `gawk' distribution). His code finally became part of the main `gawk' distribution with `gawk' version 3.1. *Note Contributors::, for a complete list of those who made important contributions to `gawk'. A Rose by Any Other Name ======================== The `awk' language has evolved over the years. Full details are provided in *Note Language History::. The language described in this Info file is often referred to as "new `awk'" (`nawk'). Because of this, many systems have multiple versions of `awk'. Some systems have an `awk' utility that implements the original version of the `awk' language and a `nawk' utility for the new version. Others have an `oawk' version for the "old `awk'" language and plain `awk' for the new one. Still others only have one version, which is usually the new one.(1) All in all, this makes it difficult for you to know which version of `awk' you should run when writing your programs. The best advice I can give here is to check your local documentation. Look for `awk', `oawk', and `nawk', as well as for `gawk'. It is likely that you already have some version of new `awk' on your system, which is what you should use when running your programs. (Of course, if you're reading this Info file, chances are good that you have `gawk'!) Throughout this Info file, whenever we refer to a language feature that should be available in any complete implementation of POSIX `awk', we simply use the term `awk'. When referring to a feature that is specific to the GNU implementation, we use the term `gawk'. ---------- Footnotes ---------- (1) Often, these systems use `gawk' for their `awk' implementation! Using This Book =============== The term `awk' refers to a particular program as well as to the language you use to tell this program what to do. When we need to be careful, we call the language "the `awk' language," and the program "the `awk' utility." This Info file explains both the `awk' language and how to run the `awk' utility. The term "`awk' program" refers to a program written by you in the `awk' programming language. Primarily, this Info file explains the features of `awk', as defined in the POSIX standard. It does so in the context of the `gawk' implementation. While doing so, it also attempts to describe important differences between `gawk' and other `awk' implementations.(1) Finally, any `gawk' features that are not in the POSIX standard for `awk' are noted. There are subsections labelled as *Advanced Notes* scattered throughout the Info file. They add a more complete explanation of points that are relevant, but not likely to be of interest on first reading. All appear in the index, under the heading "advanced features." Most of the time, the examples use complete `awk' programs. In some of the more advanced sections, only the part of the `awk' program that illustrates the concept currently being described is shown. While this Info file is aimed principally at people who have not been exposed to `awk', there is a lot of information here that even the `awk' expert should find useful. In particular, the description of POSIX `awk' and the example programs in *Note Library Functions::, and in *Note Sample Programs::, should be of interest. *Note Getting Started::, provides the essentials you need to know to begin using `awk'. *Note Regexp::, introduces regular expressions in general, and in particular the flavors supported by POSIX `awk' and `gawk'. *Note Reading Files::, describes how `awk' reads your data. It introduces the concepts of records and fields, as well as the `getline' command. I/O redirection is first described here. *Note Printing::, describes how `awk' programs can produce output with `print' and `printf'. *Note Expressions::, describes expressions, which are the basic building blocks for getting most things done in a program. *Note Patterns and Actions::, describes how to write patterns for matching records, actions for doing something when a record is matched, and the built-in variables `awk' and `gawk' use. *Note Arrays::, covers `awk''s one-and-only data structure: associative arrays. Deleting array elements and whole arrays is also described, as well as sorting arrays in `gawk'. *Note Functions::, describes the built-in functions `awk' and `gawk' provide, as well as how to define your own functions. *Note Internationalization::, describes special features in `gawk' for translating program messages into different languages at runtime. *Note Advanced Features::, describes a number of `gawk'-specific advanced features. Of particular note are the abilities to have two-way communications with another process, perform TCP/IP networking, and profile your `awk' programs. *Note Invoking Gawk::, describes how to run `gawk', the meaning of its command-line options, and how it finds `awk' program source files. *Note Library Functions::, and *Note Sample Programs::, provide many sample `awk' programs. Reading them allows you to see `awk' solving real problems. *Note Language History::, describes how the `awk' language has evolved since first release to present. It also describes how `gawk' has acquired features over time. *Note Installation::, describes how to get `gawk', how to compile it under Unix, and how to compile and use it on different non-Unix systems. It also describes how to report bugs in `gawk' and where to get three other freely available implementations of `awk'. *Note Notes::, describes how to disable `gawk''s extensions, as well as how to contribute new code to `gawk', how to write extension libraries, and some possible future directions for `gawk' development. *Note Basic Concepts::, provides some very cursory background material for those who are completely unfamiliar with computer programming. Also centralized there is a discussion of some of the issues surrounding floating-point numbers. The *Note Glossary::, defines most, if not all, the significant terms used throughout the book. If you find terms that you aren't familiar with, try looking them up here. *Note Copying::, and *Note GNU Free Documentation License::, present the licenses that cover the `gawk' source code and this Info file, respectively. ---------- Footnotes ---------- (1) All such differences appear in the index under the entry "differences in `awk' and `gawk'." Typographical Conventions ========================= This Info file is written using Texinfo, the GNU documentation formatting language. A single Texinfo source file is used to produce both the printed and online versions of the documentation. This minor node briefly documents the typographical conventions used in Texinfo. Examples you would type at the command-line are preceded by the common shell primary and secondary prompts, `$' and `>'. Output from the command is preceded by the glyph "-|". This typically represents the command's standard output. Error messages, and other output on the command's standard error, are preceded by the glyph "error-->". For example: $ echo hi on stdout -| hi on stdout $ echo hello on stderr 1>&2 error--> hello on stderr Characters that you type at the keyboard look `like this'. In particular, there are special characters called "control characters." These are characters that you type by holding down both the `CONTROL' key and another key, at the same time. For example, a `Ctrl-d' is typed by first pressing and holding the `CONTROL' key, next pressing the `d' key and finally releasing both keys. Dark Corners ............ Dark corners are basically fractal -- no matter how much you illuminate, there's always a smaller but darker one. Brian Kernighan Until the POSIX standard (and `The Gawk Manual'), many features of `awk' were either poorly documented or not documented at all. Descriptions of such features (often called "dark corners") are noted in this Info file with "(d.c.)". They also appear in the index under the heading "dark corner." As noted by the opening quote, though, any coverage of dark corners is, by definition, something that is incomplete. The GNU Project and This Book ============================= The Free Software Foundation (FSF) is a nonprofit organization dedicated to the production and distribution of freely distributable software. It was founded by Richard M. Stallman, the author of the original Emacs editor. GNU Emacs is the most widely used version of Emacs today. The GNU(1) Project is an ongoing effort on the part of the Free Software Foundation to create a complete, freely distributable, POSIX-compliant computing environment. The FSF uses the "GNU General Public License" (GPL) to ensure that their software's source code is always available to the end user. A copy of the GPL is included for your reference (*note Copying::). The GPL applies to the C language source code for `gawk'. To find out more about the FSF and the GNU Project online, see the GNU Project's home page (http://www.gnu.org). This Info file may also be read from their web site (http://www.gnu.org/manual/gawk/). A shell, an editor (Emacs), highly portable optimizing C, C++, and Objective-C compilers, a symbolic debugger and dozens of large and small utilities (such as `gawk'), have all been completed and are freely available. The GNU operating system kernel (the HURD), has been released but is still in an early stage of development. Until the GNU operating system is more fully developed, you should consider using GNU/Linux, a freely distributable, Unix-like operating system for Intel 80386, DEC Alpha, Sun SPARC, IBM S/390, and other systems.(2) There are many books on GNU/Linux. One that is freely available is `Linux Installation and Getting Started', by Matt Welsh. Many GNU/Linux distributions are often available in computer stores or bundled on CD-ROMs with books about Linux. (There are three other freely available, Unix-like operating systems for 80386 and other systems: NetBSD, FreeBSD, and OpenBSD. All are based on the 4.4-Lite Berkeley Software Distribution, and they use recent versions of `gawk' for their versions of `awk'.) The Info file itself has gone through a number of previous editions. Paul Rubin wrote the very first draft of `The GAWK Manual'; it was around 40 pages in size. Diane Close and Richard Stallman improved it, yielding a version that was around 90 pages long and barely described the original, "old" version of `awk'. I started working with that version in the fall of 1988. As work on it progressed, the FSF published several preliminary versions (numbered 0.X). In 1996, Edition 1.0 was released with `gawk' 3.0.0. The FSF published the first two editions under the title `The GNU Awk User's Guide'. This edition maintains the basic structure of Edition 1.0, but with significant additional material, reflecting the host of new features in `gawk' version 3.1. Of particular note is *Note Array Sorting::, as well as *Note Bitwise Functions::, *Note Internationalization::, and also *Note Advanced Features::, and *Note Dynamic Extensions::. `GAWK: Effective AWK Programming' will undoubtedly continue to evolve. An electronic version comes with the `gawk' distribution from the FSF. If you find an error in this Info file, please report it! *Note Bugs::, for information on submitting problem reports electronically, or write to me in care of the publisher. ---------- Footnotes ---------- (1) GNU stands for "GNU's not Unix." (2) The terminology "GNU/Linux" is explained in the *Note Glossary::. How to Contribute ================= As the maintainer of GNU `awk', I am starting a collection of publicly available `awk' programs. For more information, see `ftp://ftp.freefriends.org/arnold/Awkstuff'. If you have written an interesting `awk' program, or have written a `gawk' extension that you would like to share with the rest of the world, please contact me (). Making things available on the Internet helps keep the `gawk' distribution down to manageable size. Acknowledgments =============== The initial draft of `The GAWK Manual' had the following acknowledgments: Many people need to be thanked for their assistance in producing this manual. Jay Fenlason contributed many ideas and sample programs. Richard Mlynarik and Robert Chassell gave helpful comments on drafts of this manual. The paper `A Supplemental Document for `awk'' by John W. Pierce of the Chemistry Department at UC San Diego, pinpointed several issues relevant both to `awk' implementation and to this manual, that would otherwise have escaped us. I would like to acknowledge Richard M. Stallman, for his vision of a better world and for his courage in founding the FSF and starting the GNU Project. The following people (in alphabetical order) provided helpful comments on various versions of this book, up to and including this edition. Rick Adams, Nelson H.F. Beebe, Karl Berry, Dr. Michael Brennan, Rich Burridge, Claire Cloutier, Diane Close, Scott Deifik, Christopher ("Topher") Eliot, Jeffrey Friedl, Dr. Darrel Hankerson, Michal Jaegermann, Dr. Richard J. LeBlanc, Michael Lijewski, Pat Rankin, Miriam Robbins, Mary Sheehan, and Chuck Toporek. Robert J. Chassell provided much valuable advice on the use of Texinfo. He also deserves special thanks for convincing me _not_ to title this Info file `How To Gawk Politely'. Karl Berry helped significantly with the TeX part of Texinfo. I would like to thank Marshall and Elaine Hartholz of Seattle and Dr. Bert and Rita Schreiber of Detroit for large amounts of quiet vacation time in their homes, which allowed me to make significant progress on this Info file and on `gawk' itself. Phil Hughes of SSC contributed in a very important way by loaning me his laptop GNU/Linux system, not once, but twice, which allowed me to do a lot of work while away from home. David Trueman deserves special credit; he has done a yeoman job of evolving `gawk' so that it performs well and without bugs. Although he is no longer involved with `gawk', working with him on this project was a significant pleasure. The intrepid members of the GNITS mailing list, and most notably Ulrich Drepper, provided invaluable help and feedback for the design of the internationalization features. Nelson Beebe, Martin Brown, Andreas Buening, Scott Deifik, Darrel Hankerson, Isamu Hasegawa, Michal Jaegermann, Ju"rgen Kahrs, Pat Rankin, Kai Uwe Rommel, and Eli Zaretskii (in alphabetical order) make up the `gawk' "crack portability team." Without their hard work and help, `gawk' would not be nearly the fine program it is today. It has been and continues to be a pleasure working with this team of fine people. David and I would like to thank Brian Kernighan of Bell Laboratories for invaluable assistance during the testing and debugging of `gawk', and for help in clarifying numerous points about the language. We could not have done nearly as good a job on either `gawk' or its documentation without his help. Chuck Toporek, Mary Sheehan, and Claire Coutier of O'Reilly & Associates contributed significant editorial help for this Info file for the 3.1 release of `gawk'. I must thank my wonderful wife, Miriam, for her patience through the many versions of this project, for her proofreading, and for sharing me with the computer. I would like to thank my parents for their love, and for the grace with which they raised and educated me. Finally, I also must acknowledge my gratitude to G-d, for the many opportunities He has sent my way, as well as for the gifts He has given me with which to take advantage of those opportunities. Arnold Robbins Nof Ayalon ISRAEL March, 2001 1 Getting Started with `awk' **************************** The basic function of `awk' is to search files for lines (or other units of text) that contain certain patterns. When a line matches one of the patterns, `awk' performs specified actions on that line. `awk' keeps processing input lines in this way until it reaches the end of the input files. Programs in `awk' are different from programs in most other languages, because `awk' programs are "data-driven"; that is, you describe the data you want to work with and then what to do when you find it. Most other languages are "procedural"; you have to describe, in great detail, every step the program is to take. When working with procedural languages, it is usually much harder to clearly describe the data your program will process. For this reason, `awk' programs are often refreshingly easy to read and write. When you run `awk', you specify an `awk' "program" that tells `awk' what to do. The program consists of a series of "rules". (It may also contain "function definitions", an advanced feature that we will ignore for now. *Note User-defined::.) Each rule specifies one pattern to search for and one action to perform upon finding the pattern. Syntactically, a rule consists of a pattern followed by an action. The action is enclosed in curly braces to separate it from the pattern. Newlines usually separate rules. Therefore, an `awk' program looks like this: PATTERN { ACTION } PATTERN { ACTION } ... 1.1 How to Run `awk' Programs ============================= There are several ways to run an `awk' program. If the program is short, it is easiest to include it in the command that runs `awk', like this: awk 'PROGRAM' INPUT-FILE1 INPUT-FILE2 ... When the program is long, it is usually more convenient to put it in a file and run it with a command like this: awk -f PROGRAM-FILE INPUT-FILE1 INPUT-FILE2 ... This minor node discusses both mechanisms, along with several variations of each. 1.1.1 One-Shot Throwaway `awk' Programs --------------------------------------- Once you are familiar with `awk', you will often type in simple programs the moment you want to use them. Then you can write the program as the first argument of the `awk' command, like this: awk 'PROGRAM' INPUT-FILE1 INPUT-FILE2 ... where PROGRAM consists of a series of PATTERNS and ACTIONS, as described earlier. This command format instructs the "shell", or command interpreter, to start `awk' and use the PROGRAM to process records in the input file(s). There are single quotes around PROGRAM so the shell won't interpret any `awk' characters as special shell characters. The quotes also cause the shell to treat all of PROGRAM as a single argument for `awk', and allow PROGRAM to be more than one line long. This format is also useful for running short or medium-sized `awk' programs from shell scripts, because it avoids the need for a separate file for the `awk' program. A self-contained shell script is more reliable because there are no other files to misplace. *Note Very Simple::, presents several short, self-contained programs. 1.1.2 Running `awk' Without Input Files --------------------------------------- You can also run `awk' without any input files. If you type the following command line: awk 'PROGRAM' `awk' applies the PROGRAM to the "standard input", which usually means whatever you type on the terminal. This continues until you indicate end-of-file by typing `Ctrl-d'. (On other operating systems, the end-of-file character may be different. For example, on OS/2 and MS-DOS, it is `Ctrl-z'.) As an example, the following program prints a friendly piece of advice (from Douglas Adams's `The Hitchhiker's Guide to the Galaxy'), to keep you from worrying about the complexities of computer programming (`BEGIN' is a feature we haven't discussed yet): $ awk "BEGIN { print \"Don't Panic!\" }" -| Don't Panic! This program does not read any input. The `\' before each of the inner double quotes is necessary because of the shell's quoting rules--in particular because it mixes both single quotes and double quotes.(1) This next simple `awk' program emulates the `cat' utility; it copies whatever you type on the keyboard to its standard output (why this works is explained shortly). $ awk '{ print }' Now is the time for all good men -| Now is the time for all good men to come to the aid of their country. -| to come to the aid of their country. Four score and seven years ago, ... -| Four score and seven years ago, ... What, me worry? -| What, me worry? Ctrl-d ---------- Footnotes ---------- (1) Although we generally recommend the use of single quotes around the program text, double quotes are needed here in order to put the single quote into the message. 1.1.3 Running Long Programs --------------------------- Sometimes your `awk' programs can be very long. In this case, it is more convenient to put the program into a separate file. In order to tell `awk' to use that file for its program, you type: awk -f SOURCE-FILE INPUT-FILE1 INPUT-FILE2 ... The `-f' instructs the `awk' utility to get the `awk' program from the file SOURCE-FILE. Any file name can be used for SOURCE-FILE. For example, you could put the program: BEGIN { print "Don't Panic!" } into the file `advice'. Then this command: awk -f advice does the same thing as this one: awk "BEGIN { print \"Don't Panic!\" }" This was explained earlier (*note Read Terminal::). Note that you don't usually need single quotes around the file name that you specify with `-f', because most file names don't contain any of the shell's special characters. Notice that in `advice', the `awk' program did not have single quotes around it. The quotes are only needed for programs that are provided on the `awk' command line. If you want to identify your `awk' program files clearly as such, you can add the extension `.awk' to the file name. This doesn't affect the execution of the `awk' program but it does make "housekeeping" easier. 1.1.4 Executable `awk' Programs ------------------------------- Once you have learned `awk', you may want to write self-contained `awk' scripts, using the `#!' script mechanism. You can do this on many Unix systems(1) as well as on the GNU system. For example, you could update the file `advice' to look like this: #! /bin/awk -f BEGIN { print "Don't Panic!" } After making this file executable (with the `chmod' utility), simply type `advice' at the shell and the system arranges to run `awk'(2) as if you had typed `awk -f advice': $ chmod +x advice $ advice -| Don't Panic! (We assume you have the current directory in your shell's search path variable (typically `$PATH'). If not, you may need to type `./advice' at the shell.) Self-contained `awk' scripts are useful when you want to write a program that users can invoke without their having to know that the program is written in `awk'. Advanced Notes: Portability Issues with `#!' -------------------------------------------- Some systems limit the length of the interpreter name to 32 characters. Often, this can be dealt with by using a symbolic link. You should not put more than one argument on the `#!' line after the path to `awk'. It does not work. The operating system treats the rest of the line as a single argument and passes it to `awk'. Doing this leads to confusing behavior--most likely a usage diagnostic of some sort from `awk'. Finally, the value of `ARGV[0]' (*note Built-in Variables::) varies depending upon your operating system. Some systems put `awk' there, some put the full pathname of `awk' (such as `/bin/awk'), and some put the name of your script (`advice'). Don't rely on the value of `ARGV[0]' to provide your script name. ---------- Footnotes ---------- (1) The `#!' mechanism works on Linux systems, systems derived from the 4.4-Lite Berkeley Software Distribution, and most commercial Unix systems. (2) The line beginning with `#!' lists the full file name of an interpreter to run and an optional initial command-line argument to pass to that interpreter. The operating system then runs the interpreter with the given argument and the full argument list of the executed program. The first argument in the list is the full file name of the `awk' program. The rest of the argument list contains either options to `awk', or data files, or both. 1.1.5 Comments in `awk' Programs -------------------------------- A "comment" is some text that is included in a program for the sake of human readers; it is not really an executable part of the program. Comments can explain what the program does and how it works. Nearly all programming languages have provisions for comments, as programs are typically hard to understand without them. In the `awk' language, a comment starts with the sharp sign character (`#') and continues to the end of the line. The `#' does not have to be the first character on the line. The `awk' language ignores the rest of a line following a sharp sign. For example, we could have put the following into `advice': # This program prints a nice friendly message. It helps # keep novice users from being afraid of the computer. BEGIN { print "Don't Panic!" } You can put comment lines into keyboard-composed throwaway `awk' programs, but this usually isn't very useful; the purpose of a comment is to help you or another person understand the program when reading it at a later time. *Caution:* As mentioned in *Note One-shot::, you can enclose small to medium programs in single quotes, in order to keep your shell scripts self-contained. When doing so, _don't_ put an apostrophe (i.e., a single quote) into a comment (or anywhere else in your program). The shell interprets the quote as the closing quote for the entire program. As a result, usually the shell prints a message about mismatched quotes, and if `awk' actually runs, it will probably print strange messages about syntax errors. For example, look at the following: $ awk '{ print "hello" } # let's be cute' > The shell sees that the first two quotes match, and that a new quoted object begins at the end of the command line. It therefore prompts with the secondary prompt, waiting for more input. With Unix `awk', closing the quoted string produces this result: $ awk '{ print "hello" } # let's be cute' > ' error--> awk: can't open file be error--> source line number 1 Putting a backslash before the single quote in `let's' wouldn't help, since backslashes are not special inside single quotes. The next node describes the shell's quoting rules. 1.1.6 Shell-Quoting Issues -------------------------- For short to medium length `awk' programs, it is most convenient to enter the program on the `awk' command line. This is best done by enclosing the entire program in single quotes. This is true whether you are entering the program interactively at the shell prompt, or writing it as part of a larger shell script: awk 'PROGRAM TEXT' INPUT-FILE1 INPUT-FILE2 ... Once you are working with the shell, it is helpful to have a basic knowledge of shell quoting rules. The following rules apply only to POSIX-compliant, Bourne-style shells (such as `bash', the GNU Bourne-Again Shell). If you use `csh', you're on your own. * Quoted items can be concatenated with nonquoted items as well as with other quoted items. The shell turns everything into one argument for the command. * Preceding any single character with a backslash (`\') quotes that character. The shell removes the backslash and passes the quoted character on to the command. * Single quotes protect everything between the opening and closing quotes. The shell does no interpretation of the quoted text, passing it on verbatim to the command. It is _impossible_ to embed a single quote inside single-quoted text. Refer back to *Note Comments::, for an example of what happens if you try. * Double quotes protect most things between the opening and closing quotes. The shell does at least variable and command substitution on the quoted text. Different shells may do additional kinds of processing on double-quoted text. Since certain characters within double-quoted text are processed by the shell, they must be "escaped" within the text. Of note are the characters `$', ``', `\', and `"', all of which must be preceded by a backslash within double-quoted text if they are to be passed on literally to the program. (The leading backslash is stripped first.) Thus, the example seen in *Note Read Terminal::, is applicable: $ awk "BEGIN { print \"Don't Panic!\" }" -| Don't Panic! Note that the single quote is not special within double quotes. * Null strings are removed when they occur as part of a non-null command-line argument, while explicit non-null objects are kept. For example, to specify that the field separator `FS' should be set to the null string, use: awk -F "" 'PROGRAM' FILES # correct Don't use this: awk -F"" 'PROGRAM' FILES # wrong! In the second case, `awk' will attempt to use the text of the program as the value of `FS', and the first file name as the text of the program! This results in syntax errors at best, and confusing behavior at worst. Mixing single and double quotes is difficult. You have to resort to shell quoting tricks, like this: $ awk 'BEGIN { print "Here is a single quote <'"'"'>" }' -| Here is a single quote <'> This program consists of three concatenated quoted strings. The first and the third are single-quoted, the second is double-quoted. This can be "simplified" to: $ awk 'BEGIN { print "Here is a single quote <'\''>" }' -| Here is a single quote <'> Judge for yourself which of these two is the more readable. Another option is to use double quotes, escaping the embedded, `awk'-level double quotes: $ awk "BEGIN { print \"Here is a single quote <'>\" }" -| Here is a single quote <'> This option is also painful, because double quotes, backslashes, and dollar signs are very common in `awk' programs. A third option is to use the octal escape sequence equivalents for the single- and double-quote characters, like so: $ awk 'BEGIN { print "Here is a single quote <\47>" }' -| Here is a single quote <'> $ awk 'BEGIN { print "Here is a double quote <\42>" }' -| Here is a double quote <"> This works nicely, except that you should comment clearly what the escapes mean. A fourth option is to use command-line variable assignment, like this: $ awk -v sq="'" 'BEGIN { print "Here is a single quote <" sq ">" }' -| Here is a single quote <'> If you really need both single and double quotes in your `awk' program, it is probably best to move it into a separate file, where the shell won't be part of the picture, and you can say what you mean. 1.2 Data Files for the Examples =============================== Many of the examples in this Info file take their input from two sample data files. The first, `BBS-list', represents a list of computer bulletin board systems together with information about those systems. The second data file, called `inventory-shipped', contains information about monthly shipments. In both files, each line is considered to be one "record". In the data file `BBS-list', each record contains the name of a computer bulletin board, its phone number, the board's baud rate(s), and a code for the number of hours it is operational. An `A' in the last column means the board operates 24 hours a day. A `B' in the last column means the board only operates on evening and weekend hours. A `C' means the board operates only on weekends: aardvark 555-5553 1200/300 B alpo-net 555-3412 2400/1200/300 A barfly 555-7685 1200/300 A bites 555-1675 2400/1200/300 A camelot 555-0542 300 C core 555-2912 1200/300 C fooey 555-1234 2400/1200/300 B foot 555-6699 1200/300 B macfoo 555-6480 1200/300 A sdace 555-3430 2400/1200/300 A sabafoo 555-2127 1200/300 C The data file `inventory-shipped' represents information about shipments during the year. Each record contains the month, the number of green crates shipped, the number of red boxes shipped, the number of orange bags shipped, and the number of blue packages shipped, respectively. There are 16 entries, covering the 12 months of last year and the first four months of the current year. Jan 13 25 15 115 Feb 15 32 24 226 Mar 15 24 34 228 Apr 31 52 63 420 May 16 34 29 208 Jun 31 42 75 492 Jul 24 34 67 436 Aug 15 34 47 316 Sep 13 55 37 277 Oct 29 54 68 525 Nov 20 87 82 577 Dec 17 35 61 401 Jan 21 36 64 620 Feb 26 58 80 652 Mar 24 75 70 495 Apr 21 70 74 514 If you are reading this in GNU Emacs using Info, you can copy the regions of text showing these sample files into your own test files. This way you can try out the examples shown in the remainder of this document. You do this by using the command `M-x write-region' to copy text from the Info file into a file for use with `awk' (*Note Miscellaneous File Operations: (emacs)Misc File Ops, for more information). Using this information, create your own `BBS-list' and `inventory-shipped' files and practice what you learn in this Info file. If you are using the stand-alone version of Info, see *Note Extract Program::, for an `awk' program that extracts these data files from `gawk.texi', the Texinfo source file for this Info file. 1.3 Some Simple Examples ======================== The following command runs a simple `awk' program that searches the input file `BBS-list' for the character string `foo' (a grouping of characters is usually called a "string"; the term "string" is based on similar usage in English, such as "a string of pearls," or "a string of cars in a train"): awk '/foo/ { print $0 }' BBS-list When lines containing `foo' are found, they are printed because `print $0' means print the current line. (Just `print' by itself means the same thing, so we could have written that instead.) You will notice that slashes (`/') surround the string `foo' in the `awk' program. The slashes indicate that `foo' is the pattern to search for. This type of pattern is called a "regular expression", which is covered in more detail later (*note Regexp::). The pattern is allowed to match parts of words. There are single quotes around the `awk' program so that the shell won't interpret any of it as special shell characters. Here is what this program prints: $ awk '/foo/ { print $0 }' BBS-list -| fooey 555-1234 2400/1200/300 B -| foot 555-6699 1200/300 B -| macfoo 555-6480 1200/300 A -| sabafoo 555-2127 1200/300 C In an `awk' rule, either the pattern or the action can be omitted, but not both. If the pattern is omitted, then the action is performed for _every_ input line. If the action is omitted, the default action is to print all lines that match the pattern. Thus, we could leave out the action (the `print' statement and the curly braces) in the previous example and the result would be the same: all lines matching the pattern `foo' are printed. By comparison, omitting the `print' statement but retaining the curly braces makes an empty action that does nothing (i.e., no lines are printed). Many practical `awk' programs are just a line or two. Following is a collection of useful, short programs to get you started. Some of these programs contain constructs that haven't been covered yet. (The description of the program will give you a good idea of what is going on, but please read the rest of the Info file to become an `awk' expert!) Most of the examples use a data file named `data'. This is just a placeholder; if you use these programs yourself, substitute your own file names for `data'. For future reference, note that there is often more than one way to do things in `awk'. At some point, you may want to look back at these examples and see if you can come up with different ways to do the same things shown here: * Print the length of the longest input line: awk '{ if (length($0) > max) max = length($0) } END { print max }' data * Print every line that is longer than 80 characters: awk 'length($0) > 80' data The sole rule has a relational expression as its pattern and it has no action--so the default action, printing the record, is used. * Print the length of the longest line in `data': expand data | awk '{ if (x < length()) x = length() } END { print "maximum line length is " x }' The input is processed by the `expand' utility to change tabs into spaces, so the widths compared are actually the right-margin columns. * Print every line that has at least one field: awk 'NF > 0' data This is an easy way to delete blank lines from a file (or rather, to create a new file similar to the old file but from which the blank lines have been removed). * Print seven random numbers from 0 to 100, inclusive: awk 'BEGIN { for (i = 1; i <= 7; i++) print int(101 * rand()) }' * Print the total number of bytes used by FILES: ls -l FILES | awk '{ x += $5 } END { print "total bytes: " x }' * Print the total number of kilobytes used by FILES: ls -l FILES | awk '{ x += $5 } END { print "total K-bytes: " (x + 1023)/1024 }' * Print a sorted list of the login names of all users: awk -F: '{ print $1 }' /etc/passwd | sort * Count the lines in a file: awk 'END { print NR }' data * Print the even-numbered lines in the data file: awk 'NR % 2 == 0' data If you use the expression `NR % 2 == 1' instead, the program would print the odd-numbered lines. 1.4 An Example with Two Rules ============================= The `awk' utility reads the input files one line at a time. For each line, `awk' tries the patterns of each of the rules. If several patterns match, then several actions are run in the order in which they appear in the `awk' program. If no patterns match, then no actions are run. After processing all the rules that match the line (and perhaps there are none), `awk' reads the next line. (However, *note Next Statement::, and also *note Nextfile Statement::). This continues until the program reaches the end of the file. For example, the following `awk' program contains two rules: /12/ { print $0 } /21/ { print $0 } The first rule has the string `12' as the pattern and `print $0' as the action. The second rule has the string `21' as the pattern and also has `print $0' as the action. Each rule's action is enclosed in its own pair of braces. This program prints every line that contains the string `12' _or_ the string `21'. If a line contains both strings, it is printed twice, once by each rule. This is what happens if we run this program on our two sample data files, `BBS-list' and `inventory-shipped': $ awk '/12/ { print $0 } > /21/ { print $0 }' BBS-list inventory-shipped -| aardvark 555-5553 1200/300 B -| alpo-net 555-3412 2400/1200/300 A -| barfly 555-7685 1200/300 A -| bites 555-1675 2400/1200/300 A -| core 555-2912 1200/300 C -| fooey 555-1234 2400/1200/300 B -| foot 555-6699 1200/300 B -| macfoo 555-6480 1200/300 A -| sdace 555-3430 2400/1200/300 A -| sabafoo 555-2127 1200/300 C -| sabafoo 555-2127 1200/300 C -| Jan 21 36 64 620 -| Apr 21 70 74 514 Note how the line beginning with `sabafoo' in `BBS-list' was printed twice, once for each rule. 1.5 A More Complex Example ========================== Now that we've mastered some simple tasks, let's look at what typical `awk' programs do. This example shows how `awk' can be used to summarize, select, and rearrange the output of another utility. It uses features that haven't been covered yet, so don't worry if you don't understand all the details: ls -l | awk '$6 == "Nov" { sum += $5 } END { print sum }' This command prints the total number of bytes in all the files in the current directory that were last modified in November (of any year). (1) The `ls -l' part of this example is a system command that gives you a listing of the files in a directory, including each file's size and the date the file was last modified. Its output looks like this: -rw-r--r-- 1 arnold user 1933 Nov 7 13:05 Makefile -rw-r--r-- 1 arnold user 10809 Nov 7 13:03 awk.h -rw-r--r-- 1 arnold user 983 Apr 13 12:14 awk.tab.h -rw-r--r-- 1 arnold user 31869 Jun 15 12:20 awk.y -rw-r--r-- 1 arnold user 22414 Nov 7 13:03 awk1.c -rw-r--r-- 1 arnold user 37455 Nov 7 13:03 awk2.c -rw-r--r-- 1 arnold user 27511 Dec 9 13:07 awk3.c -rw-r--r-- 1 arnold user 7989 Nov 7 13:03 awk4.c The first field contains read-write permissions, the second field contains the number of links to the file, and the third field identifies the owner of the file. The fourth field identifies the group of the file. The fifth field contains the size of the file in bytes. The sixth, seventh, and eighth fields contain the month, day, and time, respectively, that the file was last modified. Finally, the ninth field contains the name of the file.(2) The `$6 == "Nov"' in our `awk' program is an expression that tests whether the sixth field of the output from `ls -l' matches the string `Nov'. Each time a line has the string `Nov' for its sixth field, the action `sum += $5' is performed. This adds the fifth field (the file's size) to the variable `sum'. As a result, when `awk' has finished reading all the input lines, `sum' is the total of the sizes of the files whose lines matched the pattern. (This works because `awk' variables are automatically initialized to zero.) After the last line of output from `ls' has been processed, the `END' rule executes and prints the value of `sum'. In this example, the value of `sum' is 80600. These more advanced `awk' techniques are covered in later sections (*note Action Overview::). Before you can move on to more advanced `awk' programming, you have to know how `awk' interprets your input and displays your output. By manipulating fields and using `print' statements, you can produce some very useful and impressive-looking reports. ---------- Footnotes ---------- (1) In the C shell (`csh'), you need to type a semicolon and then a backslash at the end of the first line; see *Note Statements/Lines::, for an explanation. In a POSIX-compliant shell, such as the Bourne shell or `bash', you can type the example as shown. If the command `echo $path' produces an empty output line, you are most likely using a POSIX-compliant shell. Otherwise, you are probably using the C shell or a shell derived from it. (2) On some very old systems, you may need to use `ls -lg' to get this output. 1.6 `awk' Statements Versus Lines ================================= Most often, each line in an `awk' program is a separate statement or separate rule, like this: awk '/12/ { print $0 } /21/ { print $0 }' BBS-list inventory-shipped However, `gawk' ignores newlines after any of the following symbols and keywords: , { ? : || && do else A newline at any other point is considered the end of the statement.(1) If you would like to split a single statement into two lines at a point where a newline would terminate it, you can "continue" it by ending the first line with a backslash character (`\'). The backslash must be the final character on the line in order to be recognized as a continuation character. A backslash is allowed anywhere in the statement, even in the middle of a string or regular expression. For example: awk '/This regular expression is too long, so continue it\ on the next line/ { print $1 }' We have generally not used backslash continuation in the sample programs in this Info file. In `gawk', there is no limit on the length of a line, so backslash continuation is never strictly necessary; it just makes programs more readable. For this same reason, as well as for clarity, we have kept most statements short in the sample programs presented throughout the Info file. Backslash continuation is most useful when your `awk' program is in a separate source file instead of entered from the command line. You should also note that many `awk' implementations are more particular about where you may use backslash continuation. For example, they may not allow you to split a string constant using backslash continuation. Thus, for maximum portability of your `awk' programs, it is best not to split your lines in the middle of a regular expression or a string. *Caution:* _Backslash continuation does not work as described with the C shell._ It works for `awk' programs in files and for one-shot programs, _provided_ you are using a POSIX-compliant shell, such as the Unix Bourne shell or `bash'. But the C shell behaves differently! There, you must use two backslashes in a row, followed by a newline. Note also that when using the C shell, _every_ newline in your awk program must be escaped with a backslash. To illustrate: % awk 'BEGIN { \ ? print \\ ? "hello, world" \ ? }' -| hello, world Here, the `%' and `?' are the C shell's primary and secondary prompts, analogous to the standard shell's `$' and `>'. Compare the previous example to how it is done with a POSIX-compliant shell: $ awk 'BEGIN { > print \ > "hello, world" > }' -| hello, world `awk' is a line-oriented language. Each rule's action has to begin on the same line as the pattern. To have the pattern and action on separate lines, you _must_ use backslash continuation; there is no other option. Another thing to keep in mind is that backslash continuation and comments do not mix. As soon as `awk' sees the `#' that starts a comment, it ignores _everything_ on the rest of the line. For example: $ gawk 'BEGIN { print "dont panic" # a friendly \ > BEGIN rule > }' error--> gawk: cmd. line:2: BEGIN rule error--> gawk: cmd. line:2: ^ parse error In this case, it looks like the backslash would continue the comment onto the next line. However, the backslash-newline combination is never even noticed because it is "hidden" inside the comment. Thus, the `BEGIN' is noted as a syntax error. When `awk' statements within one rule are short, you might want to put more than one of them on a line. This is accomplished by separating the statements with a semicolon (`;'). This also applies to the rules themselves. Thus, the program shown at the start of this minor node could also be written this way: /12/ { print $0 } ; /21/ { print $0 } NOTE: The requirement that states that rules on the same line must be separated with a semicolon was not in the original `awk' language; it was added for consistency with the treatment of statements within an action. ---------- Footnotes ---------- (1) The `?' and `:' referred to here is the three-operand conditional expression described in *Note Conditional Exp::. Splitting lines after `?' and `:' is a minor `gawk' extension; if `--posix' is specified (*note Options::), then this extension is disabled. 1.7 Other Features of `awk' =========================== The `awk' language provides a number of predefined, or "built-in", variables that your programs can use to get information from `awk'. There are other variables your program can set as well to control how `awk' processes your data. In addition, `awk' provides a number of built-in functions for doing common computational and string-related operations. `gawk' provides built-in functions for working with timestamps, performing bit manipulation, and for runtime string translation. As we develop our presentation of the `awk' language, we introduce most of the variables and many of the functions. They are defined systematically in *Note Built-in Variables::, and *Note Built-in::. 1.8 When to Use `awk' ===================== Now that you've seen some of what `awk' can do, you might wonder how `awk' could be useful for you. By using utility programs, advanced patterns, field separators, arithmetic statements, and other selection criteria, you can produce much more complex output. The `awk' language is very useful for producing reports from large amounts of raw data, such as summarizing information from the output of other utility programs like `ls'. (*Note More Complex::.) Programs written with `awk' are usually much smaller than they would be in other languages. This makes `awk' programs easy to compose and use. Often, `awk' programs can be quickly composed at your terminal, used once, and thrown away. Because `awk' programs are interpreted, you can avoid the (usually lengthy) compilation part of the typical edit-compile-test-debug cycle of software development. Complex programs have been written in `awk', including a complete retargetable assembler for eight-bit microprocessors (*note Glossary::, for more information), and a microcode assembler for a special-purpose Prolog computer. More recently, `gawk' was used for writing a Wiki clone.(1) While the original `awk''s capabilities were strained by tasks of such complexity, modern versions are more capable. Even the Bell Labs version of `awk' has fewer predefined limits, and those that it has are much larger than they used to be. If you find yourself writing `awk' scripts of more than, say, a few hundred lines, you might consider using a different programming language. Emacs Lisp is a good choice if you need sophisticated string or pattern matching capabilities. The shell is also good at string and pattern matching; in addition, it allows powerful use of the system utilities. More conventional languages, such as C, C++, and Java, offer better facilities for system programming and for managing the complexity of large programs. Programs in these languages may require more lines of source code than the equivalent `awk' programs, but they are easier to maintain and usually run more efficiently. ---------- Footnotes ---------- (1) Yet Another Wiki Clone (http://www.awk-scripting.de/cgi/wiki.cgi/yawk/). 2 Regular Expressions ********************* A "regular expression", or "regexp", is a way of describing a set of strings. Because regular expressions are such a fundamental part of `awk' programming, their format and use deserve a separate major node. A regular expression enclosed in slashes (`/') is an `awk' pattern that matches every input record whose text belongs to that set. The simplest regular expression is a sequence of letters, numbers, or both. Such a regexp matches any string that contains that sequence. Thus, the regexp `foo' matches any string containing `foo'. Therefore, the pattern `/foo/' matches any input record containing the three characters `foo' _anywhere_ in the record. Other kinds of regexps let you specify more complicated classes of strings. 2.1 How to Use Regular Expressions ================================== A regular expression can be used as a pattern by enclosing it in slashes. Then the regular expression is tested against the entire text of each record. (Normally, it only needs to match some part of the text in order to succeed.) For example, the following prints the second field of each record that contains the string `foo' anywhere in it: $ awk '/foo/ { print $2 }' BBS-list -| 555-1234 -| 555-6699 -| 555-6480 -| 555-2127 `~' (tilde), `~' operator Regular expressions can also be used in matching expressions. These expressions allow you to specify the string to match against; it need not be the entire current input record. The two operators `~' and `!~' perform regular expression comparisons. Expressions using these operators can be used as patterns, or in `if', `while', `for', and `do' statements. (*Note Statements::.) For example: EXP ~ /REGEXP/ is true if the expression EXP (taken as a string) matches REGEXP. The following example matches, or selects, all input records with the uppercase letter `J' somewhere in the first field: $ awk '$1 ~ /J/' inventory-shipped -| Jan 13 25 15 115 -| Jun 31 42 75 492 -| Jul 24 34 67 436 -| Jan 21 36 64 620 So does this: awk '{ if ($1 ~ /J/) print }' inventory-shipped This next example is true if the expression EXP (taken as a character string) does _not_ match REGEXP: EXP !~ /REGEXP/ The following example matches, or selects, all input records whose first field _does not_ contain the uppercase letter `J': $ awk '$1 !~ /J/' inventory-shipped -| Feb 15 32 24 226 -| Mar 15 24 34 228 -| Apr 31 52 63 420 -| May 16 34 29 208 ... When a regexp is enclosed in slashes, such as `/foo/', we call it a "regexp constant", much like `5.27' is a numeric constant and `"foo"' is a string constant. 2.2 Escape Sequences ==================== Some characters cannot be included literally in string constants (`"foo"') or regexp constants (`/foo/'). Instead, they should be represented with "escape sequences", which are character sequences beginning with a backslash (`\'). One use of an escape sequence is to include a double-quote character in a string constant. Because a plain double quote ends the string, you must use `\"' to represent an actual double-quote character as a part of the string. For example: $ awk 'BEGIN { print "He said \"hi!\" to her." }' -| He said "hi!" to her. The backslash character itself is another character that cannot be included normally; you must write `\\' to put one backslash in the string or regexp. Thus, the string whose contents are the two characters `"' and `\' must be written `"\"\\"'. Backslash also represents unprintable characters such as TAB or newline. While there is nothing to stop you from entering most unprintable characters directly in a string constant or regexp constant, they may look ugly. The following table lists all the escape sequences used in `awk' and what they represent. Unless noted otherwise, all these escape sequences apply to both string constants and regexp constants: `\\' A literal backslash, `\'. `\a' The "alert" character, `Ctrl-g', ASCII code 7 (BEL). (This usually makes some sort of audible noise.) `\b' Backspace, `Ctrl-h', ASCII code 8 (BS). `\f' Formfeed, `Ctrl-l', ASCII code 12 (FF). `\n' Newline, `Ctrl-j', ASCII code 10 (LF). `\r' Carriage return, `Ctrl-m', ASCII code 13 (CR). `\t' Horizontal TAB, `Ctrl-i', ASCII code 9 (HT). `\v' Vertical tab, `Ctrl-k', ASCII code 11 (VT). `\NNN' The octal value NNN, where NNN stands for 1 to 3 digits between `0' and `7'. For example, the code for the ASCII ESC (escape) character is `\033'. `\xHH...' The hexadecimal value HH, where HH stands for a sequence of hexadecimal digits (`0'-`9', and either `A'-`F' or `a'-`f'). Like the same construct in ISO C, the escape sequence continues until the first nonhexadecimal digit is seen. However, using more than two hexadecimal digits produces undefined results. (The `\x' escape sequence is not allowed in POSIX `awk'.) `\/' A literal slash (necessary for regexp constants only). This expression is used when you want to write a regexp constant that contains a slash. Because the regexp is delimited by slashes, you need to escape the slash that is part of the pattern, in order to tell `awk' to keep processing the rest of the regexp. `\"' A literal double quote (necessary for string constants only). This expression is used when you want to write a string constant that contains a double quote. Because the string is delimited by double quotes, you need to escape the quote that is part of the string, in order to tell `awk' to keep processing the rest of the string. In `gawk', a number of additional two-character sequences that begin with a backslash have special meaning in regexps. *Note GNU Regexp Operators::. In a regexp, a backslash before any character that is not in the previous list and not listed in *Note GNU Regexp Operators::, means that the next character should be taken literally, even if it would normally be a regexp operator. For example, `/a\+b/' matches the three characters `a+b'. For complete portability, do not use a backslash before any character not shown in the previous list. To summarize: * The escape sequences in the table above are always processed first, for both string constants and regexp constants. This happens very early, as soon as `awk' reads your program. * `gawk' processes both regexp constants and dynamic regexps (*note Computed Regexps::), for the special operators listed in *Note GNU Regexp Operators::. * A backslash before any other character means to treat that character literally. Advanced Notes: Backslash Before Regular Characters --------------------------------------------------- If you place a backslash in a string constant before something that is not one of the characters previously listed, POSIX `awk' purposely leaves what happens as undefined. There are two choices: Strip the backslash out This is what Unix `awk' and `gawk' both do. For example, `"a\qc"' is the same as `"aqc"'. (Because this is such an easy bug both to introduce and to miss, `gawk' warns you about it.) Consider `FS = "[ \t]+\|[ \t]+"' to use vertical bars surrounded by whitespace as the field separator. There should be two backslashes in the string `FS = "[ \t]+\\|[ \t]+"'.) Leave the backslash alone Some other `awk' implementations do this. In such implementations, typing `"a\qc"' is the same as typing `"a\\qc"'. Advanced Notes: Escape Sequences for Metacharacters --------------------------------------------------- Suppose you use an octal or hexadecimal escape to represent a regexp metacharacter. (See *Note Regexp Operators::.) Does `awk' treat the character as a literal character or as a regexp operator? Historically, such characters were taken literally. (d.c.) However, the POSIX standard indicates that they should be treated as real metacharacters, which is what `gawk' does. In compatibility mode (*note Options::), `gawk' treats the characters represented by octal and hexadecimal escape sequences literally when used in regexp constants. Thus, `/a\52b/' is equivalent to `/a\*b/'. 2.3 Regular Expression Operators ================================ You can combine regular expressions with special characters, called "regular expression operators" or "metacharacters", to increase the power and versatility of regular expressions. The escape sequences described in *Note Escape Sequences::, are valid inside a regexp. They are introduced by a `\' and are recognized and converted into corresponding real characters as the very first step in processing regexps. Here is a list of metacharacters. All characters that are not escape sequences and that are not listed in the table stand for themselves: `\' This is used to suppress the special meaning of a character when matching. For example, `\$' matches the character `$'. `^' This matches the beginning of a string. For example, `^@chapter' matches `@chapter' at the beginning of a string and can be used to identify chapter beginnings in Texinfo source files. The `^' is known as an "anchor", because it anchors the pattern to match only at the beginning of the string. It is important to realize that `^' does not match the beginning of a line embedded in a string. The condition is not true in the following example: if ("line1\nLINE 2" ~ /^L/) ... `$' This is similar to `^', but it matches only at the end of a string. For example, `p$' matches a record that ends with a `p'. The `$' is an anchor and does not match the end of a line embedded in a string. The condition in the following example is not true: if ("line1\nLINE 2" ~ /1$/) ... `.' This matches any single character, _including_ the newline character. For example, `.P' matches any single character followed by a `P' in a string. Using concatenation, we can make a regular expression such as `U.A', which matches any three-character sequence that begins with `U' and ends with `A'. In strict POSIX mode (*note Options::), `.' does not match the NUL character, which is a character with all bits equal to zero. Otherwise, NUL is just another character. Other versions of `awk' may not be able to match the NUL character. `[...]' This is called a "character list".(1) It matches any _one_ of the characters that are enclosed in the square brackets. For example, `[MVX]' matches any one of the characters `M', `V', or `X' in a string. A full discussion of what can be inside the square brackets of a character list is given in *Note Character Lists::. `[^ ...]' This is a "complemented character list". The first character after the `[' _must_ be a `^'. It matches any characters _except_ those in the square brackets. For example, `[^awk]' matches any character that is not an `a', `w', or `k'. `|' This is the "alternation operator" and it is used to specify alternatives. The `|' has the lowest precedence of all the regular expression operators. For example, `^P|[[:digit:]]' matches any string that matches either `^P' or `[[:digit:]]'. This means it matches any string that starts with `P' or contains a digit. The alternation applies to the largest possible regexps on either side. `(...)' Parentheses are used for grouping in regular expressions, as in arithmetic. They can be used to concatenate regular expressions containing the alternation operator, `|'. For example, `@(samp|code)\{[^}]+\}' matches both `@code{foo}' and `@samp{bar}'. (These are Texinfo formatting control sequences. The `+' is explained further on in this list.) `*' This symbol means that the preceding regular expression should be repeated as many times as necessary to find a match. For example, `ph*' applies the `*' symbol to the preceding `h' and looks for matches of one `p' followed by any number of `h's. This also matches just `p' if no `h's are present. The `*' repeats the _smallest_ possible preceding expression. (Use parentheses if you want to repeat a larger expression.) It finds as many repetitions as possible. For example, `awk '/\(c[ad][ad]*r x\)/ { print }' sample' prints every record in `sample' containing a string of the form `(car x)', `(cdr x)', `(cadr x)', and so on. Notice the escaping of the parentheses by preceding them with backslashes. `+' This symbol is similar to `*', except that the preceding expression must be matched at least once. This means that `wh+y' would match `why' and `whhy', but not `wy', whereas `wh*y' would match all three of these strings. The following is a simpler way of writing the last `*' example: awk '/\(c[ad]+r x\)/ { print }' sample `?' This symbol is similar to `*', except that the preceding expression can be matched either once or not at all. For example, `fe?d' matches `fed' and `fd', but nothing else. `{N}' `{N,}' `{N,M}' One or two numbers inside braces denote an "interval expression". If there is one number in the braces, the preceding regexp is repeated N times. If there are two numbers separated by a comma, the preceding regexp is repeated N to M times. If there is one number followed by a comma, then the preceding regexp is repeated at least N times: `wh{3}y' Matches `whhhy', but not `why' or `whhhhy'. `wh{3,5}y' Matches `whhhy', `whhhhy', or `whhhhhy', only. `wh{2,}y' Matches `whhy' or `whhhy', and so on. Interval expressions were not traditionally available in `awk'. They were added as part of the POSIX standard to make `awk' and `egrep' consistent with each other. However, because old programs may use `{' and `}' in regexp constants, by default `gawk' does _not_ match interval expressions in regexps. If either `--posix' or `--re-interval' are specified (*note Options::), then interval expressions are allowed in regexps. For new programs that use `{' and `}' in regexp constants, it is good practice to always escape them with a backslash. Then the regexp constants are valid and work the way you want them to, using any version of `awk'.(2) In regular expressions, the `*', `+', and `?' operators, as well as the braces `{' and `}', have the highest precedence, followed by concatenation, and finally by `|'. As in arithmetic, parentheses can change how operators are grouped. In POSIX `awk' and `gawk', the `*', `+', and `?' operators stand for themselves when there is nothing in the regexp that precedes them. For example, `/+/' matches a literal plus sign. However, many other versions of `awk' treat such a usage as a syntax error. If `gawk' is in compatibility mode (*note Options::), POSIX character classes and interval expressions are not available in regular expressions. ---------- Footnotes ---------- (1) In other literature, you may see a character list referred to as either a "character set", a "character class", or a "bracket expression". (2) Use two backslashes if you're using a string constant with a regexp operator or function. 2.4 Using Character Lists ========================= Within a character list, a "range expression" consists of two characters separated by a hyphen. It matches any single character that sorts between the two characters, using the locale's collating sequence and character set. For example, in the default C locale, `[a-dx-z]' is equivalent to `[abcdxyz]'. Many locales sort characters in dictionary order, and in these locales, `[a-dx-z]' is typically not equivalent to `[abcdxyz]'; instead it might be equivalent to `[aBbCcDdxXyYz]', for example. To obtain the traditional interpretation of bracket expressions, you can use the C locale by setting the `LC_ALL' environment variable to the value `C'. To include one of the characters `\', `]', `-', or `^' in a character list, put a `\' in front of it. For example: [d\]] matches either `d' or `]'. This treatment of `\' in character lists is compatible with other `awk' implementations and is also mandated by POSIX. The regular expressions in `awk' are a superset of the POSIX specification for Extended Regular Expressions (EREs). POSIX EREs are based on the regular expressions accepted by the traditional `egrep' utility. "Character classes" are a new feature introduced in the POSIX standard. A character class is a special notation for describing lists of characters that have a specific attribute, but the actual characters can vary from country to country and/or from character set to character set. For example, the notion of what is an alphabetic character differs between the United States and France. A character class is only valid in a regexp _inside_ the brackets of a character list. Character classes consist of `[:', a keyword denoting the class, and `:]'. *Note table-char-classes:: lists the character classes defined by the POSIX standard. Class Meaning -------------------------------------------------------------------------- `[:alnum:]' Alphanumeric characters. `[:alpha:]' Alphabetic characters. `[:blank:]' Space and TAB characters. `[:cntrl:]' Control characters. `[:digit:]' Numeric characters. `[:graph:]' Characters that are both printable and visible. (A space is printable but not visible, whereas an `a' is both.) `[:lower:]' Lowercase alphabetic characters. `[:print:]' Printable characters (characters that are not control characters). `[:punct:]' Punctuation characters (characters that are not letters, digits, control characters, or space characters). `[:space:]' Space characters (such as space, TAB, and formfeed, to name a few). `[:upper:]' Uppercase alphabetic characters. `[:xdigit:]'Characters that are hexadecimal digits. Table 2.1: POSIX Character Classes For example, before the POSIX standard, you had to write `/[A-Za-z0-9]/' to match alphanumeric characters. If your character set had other alphabetic characters in it, this would not match them, and if your character set collated differently from ASCII, this might not even match the ASCII alphanumeric characters. With the POSIX character classes, you can write `/[[:alnum:]]/' to match the alphabetic and numeric characters in your character set. Two additional special sequences can appear in character lists. These apply to non-ASCII character sets, which can have single symbols (called "collating elements") that are represented with more than one character. They can also have several characters that are equivalent for "collating", or sorting, purposes. (For example, in French, a plain "e" and a grave-accented "e`" are equivalent.) These sequences are: Collating symbols Multicharacter collating elements enclosed between `[.' and `.]'. For example, if `ch' is a collating element, then `[[.ch.]]' is a regexp that matches this collating element, whereas `[ch]' is a regexp that matches either `c' or `h'. Equivalence classes Locale-specific names for a list of characters that are equal. The name is enclosed between `[=' and `=]'. For example, the name `e' might be used to represent all of "e," "e`," and "e'." In this case, `[[=e=]]' is a regexp that matches any of `e', `e'', or `e`'. These features are very valuable in non-English-speaking locales. *Caution:* The library functions that `gawk' uses for regular expression matching currently recognize only POSIX character classes; they do not recognize collating symbols or equivalence classes. 2.5 `gawk'-Specific Regexp Operators ==================================== GNU software that deals with regular expressions provides a number of additional regexp operators. These operators are described in this minor node and are specific to `gawk'; they are not available in other `awk' implementations. Most of the additional operators deal with word matching. For our purposes, a "word" is a sequence of one or more letters, digits, or underscores (`_'): `\w' Matches any word-constituent character--that is, it matches any letter, digit, or underscore. Think of it as shorthand for `[[:alnum:]_]'. `\W' Matches any character that is not word-constituent. Think of it as shorthand for `[^[:alnum:]_]'. `\<' Matches the empty string at the beginning of a word. For example, `/\' Matches the empty string at the end of a word. For example, `/stow\>/' matches `stow' but not `stowaway'. `\y' Matches the empty string at either the beginning or the end of a word (i.e., the word boundar*y*). For example, `\yballs?\y' matches either `ball' or `balls', as a separate word. `\B' Matches the empty string that occurs between two word-constituent characters. For example, `/\Brat\B/' matches `crate' but it does not match `dirty rat'. `\B' is essentially the opposite of `\y'. There are two other operators that work on buffers. In Emacs, a "buffer" is, naturally, an Emacs buffer. For other programs, `gawk''s regexp library routines consider the entire string to match as the buffer. The operators are: `\`' Matches the empty string at the beginning of a buffer (string). `\'' Matches the empty string at the end of a buffer (string). Because `^' and `$' always work in terms of the beginning and end of strings, these operators don't add any new capabilities for `awk'. They are provided for compatibility with other GNU software. In other GNU software, the word-boundary operator is `\b'. However, that conflicts with the `awk' language's definition of `\b' as backspace, so `gawk' uses a different letter. An alternative method would have been to require two backslashes in the GNU operators, but this was deemed too confusing. The current method of using `\y' for the GNU `\b' appears to be the lesser of two evils. The various command-line options (*note Options::) control how `gawk' interprets characters in regexps: No options In the default case, `gawk' provides all the facilities of POSIX regexps and the GNU regexp operators described in *Note Regexp Operators::. However, interval expressions are not supported. `--posix' Only POSIX regexps are supported; the GNU operators are not special (e.g., `\w' matches a literal `w'). Interval expressions are allowed. `--traditional' Traditional Unix `awk' regexps are matched. The GNU operators are not special, interval expressions are not available, nor are the POSIX character classes (`[[:alnum:]]', etc.). Characters described by octal and hexadecimal escape sequences are treated literally, even if they represent regexp metacharacters. `--re-interval' Allow interval expressions in regexps, even if `--traditional' has been provided. (`--posix' automatically enables interval expressions, so `--re-interval' is redundant when `--posix' is is used.) 2.6 Case Sensitivity in Matching ================================ Case is normally significant in regular expressions, both when matching ordinary characters (i.e., not metacharacters) and inside character sets. Thus, a `w' in a regular expression matches only a lowercase `w' and not an uppercase `W'. The simplest way to do a case-independent match is to use a character list--for example, `[Ww]'. However, this can be cumbersome if you need to use it often, and it can make the regular expressions harder to read. There are two alternatives that you might prefer. One way to perform a case-insensitive match at a particular point in the program is to convert the data to a single case, using the `tolower' or `toupper' built-in string functions (which we haven't discussed yet; *note String Functions::). For example: tolower($1) ~ /foo/ { ... } converts the first field to lowercase before matching against it. This works in any POSIX-compliant `awk'. Another method, specific to `gawk', is to set the variable `IGNORECASE' to a nonzero value (*note Built-in Variables::). When `IGNORECASE' is not zero, _all_ regexp and string operations ignore case. Changing the value of `IGNORECASE' dynamically controls the case-sensitivity of the program as it runs. Case is significant by default because `IGNORECASE' (like most variables) is initialized to zero: x = "aB" if (x ~ /ab/) ... # this test will fail IGNORECASE = 1 if (x ~ /ab/) ... # now it will succeed In general, you cannot use `IGNORECASE' to make certain rules case-insensitive and other rules case-sensitive, because there is no straightforward way to set `IGNORECASE' just for the pattern of a particular rule.(1) To do this, use either character lists or `tolower'. However, one thing you can do with `IGNORECASE' only is dynamically turn case-sensitivity on or off for all the rules at once. `IGNORECASE' can be set on the command line or in a `BEGIN' rule (*note Other Arguments::; also *note Using BEGIN/END::). Setting `IGNORECASE' from the command line is a way to make a program case-insensitive without having to edit it. Prior to `gawk' 3.0, the value of `IGNORECASE' affected regexp operations only. It did not affect string comparison with `==', `!=', and so on. Beginning with version 3.0, both regexp and string comparison operations are also affected by `IGNORECASE'. Beginning with `gawk' 3.0, the equivalences between upper- and lowercase characters are based on the ISO-8859-1 (ISO Latin-1) character set. This character set is a superset of the traditional 128 ASCII characters, which also provides a number of characters suitable for use with European languages. As of `gawk' 3.1.4, the case equivalencies are fully locale-aware. They are based on the C `' facilities, such as `isalpha()' and `toupper()'. The value of `IGNORECASE' has no effect if `gawk' is in compatibility mode (*note Options::). Case is always significant in compatibility mode. ---------- Footnotes ---------- (1) Experienced C and C++ programmers will note that it is possible, using something like `IGNORECASE = 1 && /foObAr/ { ... }' and `IGNORECASE = 0 || /foobar/ { ... }'. However, this is somewhat obscure and we don't recommend it. 2.7 How Much Text Matches? ========================== Consider the following: echo aaaabcd | awk '{ sub(/a+/, ""); print }' This example uses the `sub' function (which we haven't discussed yet; *note String Functions::) to make a change to the input record. Here, the regexp `/a+/' indicates "one or more `a' characters," and the replacement text is `'. The input contains four `a' characters. `awk' (and POSIX) regular expressions always match the leftmost, _longest_ sequence of input characters that can match. Thus, all four `a' characters are replaced with `' in this example: $ echo aaaabcd | awk '{ sub(/a+/, ""); print }' -| bcd For simple match/no-match tests, this is not so important. But when doing text matching and substitutions with the `match', `sub', `gsub', and `gensub' functions, it is very important. *Note String Functions::, for more information on these functions. Understanding this principle is also important for regexp-based record and field splitting (*note Records::, and also *note Field Separators::). 2.8 Using Dynamic Regexps ========================= The righthand side of a `~' or `!~' operator need not be a regexp constant (i.e., a string of characters between slashes). It may be any expression. The expression is evaluated and converted to a string if necessary; the contents of the string are used as the regexp. A regexp that is computed in this way is called a "dynamic regexp": BEGIN { digits_regexp = "[[:digit:]]+" } $0 ~ digits_regexp { print } This sets `digits_regexp' to a regexp that describes one or more digits, and tests whether the input record matches this regexp. *Caution:* When using the `~' and `!~' operators, there is a difference between a regexp constant enclosed in slashes and a string constant enclosed in double quotes. If you are going to use a string constant, you have to understand that the string is, in essence, scanned _twice_: the first time when `awk' reads your program, and the second time when it goes to match the string on the lefthand side of the operator with the pattern on the right. This is true of any string-valued expression (such as `digits_regexp', shown previously), not just string constants. What difference does it make if the string is scanned twice? The answer has to do with escape sequences, and particularly with backslashes. To get a backslash into a regular expression inside a string, you have to type two backslashes. For example, `/\*/' is a regexp constant for a literal `*'. Only one backslash is needed. To do the same thing with a string, you have to type `"\\*"'. The first backslash escapes the second one so that the string actually contains the two characters `\' and `*'. Given that you can use both regexp and string constants to describe regular expressions, which should you use? The answer is "regexp constants," for several reasons: * String constants are more complicated to write and more difficult to read. Using regexp constants makes your programs less error-prone. Not understanding the difference between the two kinds of constants is a common source of errors. * It is more efficient to use regexp constants. `awk' can note that you have supplied a regexp and store it internally in a form that makes pattern matching more efficient. When using a string constant, `awk' must first convert the string into this internal form and then perform the pattern matching. * Using regexp constants is better form; it shows clearly that you intend a regexp match. Advanced Notes: Using `\n' in Character Lists of Dynamic Regexps ---------------------------------------------------------------- Some commercial versions of `awk' do not allow the newline character to be used inside a character list for a dynamic regexp: $ awk '$0 ~ "[ \t\n]"' error--> awk: newline in character class [ error--> ]... error--> source line number 1 error--> context is error--> >>> <<< But a newline in a regexp constant works with no problem: $ awk '$0 ~ /[ \t\n]/' here is a sample line -| here is a sample line Ctrl-d `gawk' does not have this problem, and it isn't likely to occur often in practice, but it's worth noting for future reference. 2.9 Where You Are Makes A Difference ==================================== Modern systems support the notion of "locales": a way to tell the system about the local character set and language. The current locale setting can affect the way regexp matching works, often in surprising ways. In particular, many locales do case-insensitive matching, even when you may have specified characters of only one particular case. The following example uses the `sub' function, which does text replacement (*note String Functions::). Here, the intent is to remove trailing uppercase characters: $ echo something1234abc | gawk '{ sub("[A-Z]*$", ""); print }' -| something1234 This output is unexpected, since the `abc' at the end of `something1234abc' should not normally match `[A-Z]*'. This result is due to the locale setting (and thus you may not see it on your system). There are two fixes. The first is to use the POSIX character class `[[:upper:]]', instead of `[A-Z]'. The second is to change the locale setting in the environment, before running `gawk', by using the shell statements: LANG=C LC_ALL=C export LANG LC_ALL The setting `C' forces `gawk' to behave in the traditional Unix manner, where case distinctions do matter. You may wish to put these statements into your shell startup file, e.g., `$HOME/.profile'. Similar considerations apply to other ranges. For example, `["-/]' is perfectly valid in ASCII, but is not valid in many Unicode locales, such as `en_US.UTF-8'. (In general, such ranges should be avoided; either list the characters individually, or use a POSIX character class such as `[[:punct:]]'.) For the normal case of `RS = "\n"', the locale is largely irrelevant. For other single byte record separators, using `LC_ALL=C' will give you much better performance when reading records. Otherwise, `gawk' has to make several function calls, _per input character_ to find the record terminator. 3 Reading Input Files ********************* In the typical `awk' program, all input is read either from the standard input (by default, this is the keyboard, but often it is a pipe from another command) or from files whose names you specify on the `awk' command line. If you specify input files, `awk' reads them in order, processing all the data from one before going on to the next. The name of the current input file can be found in the built-in variable `FILENAME' (*note Built-in Variables::). The input is read in units called "records", and is processed by the rules of your program one record at a time. By default, each record is one line. Each record is automatically split into chunks called "fields". This makes it more convenient for programs to work on the parts of a record. On rare occasions, you may need to use the `getline' command. The `getline' command is valuable, both because it can do explicit input from any number of files, and because the files used with it do not have to be named on the `awk' command line (*note Getline::). 3.1 How Input Is Split into Records =================================== The `awk' utility divides the input for your `awk' program into records and fields. `awk' keeps track of the number of records that have been read so far from the current input file. This value is stored in a built-in variable called `FNR'. It is reset to zero when a new file is started. Another built-in variable, `NR', is the total number of input records read so far from all data files. It starts at zero, but is never automatically reset to zero. Records are separated by a character called the "record separator". By default, the record separator is the newline character. This is why records are, by default, single lines. A different character can be used for the record separator by assigning the character to the built-in variable `RS'. Like any other variable, the value of `RS' can be changed in the `awk' program with the assignment operator, `=' (*note Assignment Ops::). The new record-separator character should be enclosed in quotation marks, which indicate a string constant. Often the right time to do this is at the beginning of execution, before any input is processed, so that the very first record is read with the proper separator. To do this, use the special `BEGIN' pattern (*note BEGIN/END::). For example: awk 'BEGIN { RS = "/" } { print $0 }' BBS-list changes the value of `RS' to `"/"', before reading any input. This is a string whose first character is a slash; as a result, records are separated by slashes. Then the input file is read, and the second rule in the `awk' program (the action with no pattern) prints each record. Because each `print' statement adds a newline at the end of its output, this `awk' program copies the input with each slash changed to a newline. Here are the results of running the program on `BBS-list': $ awk 'BEGIN { RS = "/" } > { print $0 }' BBS-list -| aardvark 555-5553 1200 -| 300 B -| alpo-net 555-3412 2400 -| 1200 -| 300 A -| barfly 555-7685 1200 -| 300 A -| bites 555-1675 2400 -| 1200 -| 300 A -| camelot 555-0542 300 C -| core 555-2912 1200 -| 300 C -| fooey 555-1234 2400 -| 1200 -| 300 B -| foot 555-6699 1200 -| 300 B -| macfoo 555-6480 1200 -| 300 A -| sdace 555-3430 2400 -| 1200 -| 300 A -| sabafoo 555-2127 1200 -| 300 C -| Note that the entry for the `camelot' BBS is not split. In the original data file (*note Sample Data Files::), the line looks like this: camelot 555-0542 300 C It has one baud rate only, so there are no slashes in the record, unlike the others which have two or more baud rates. In fact, this record is treated as part of the record for the `core' BBS; the newline separating them in the output is the original newline in the data file, not the one added by `awk' when it printed the record! Another way to change the record separator is on the command line, using the variable-assignment feature (*note Other Arguments::): awk '{ print $0 }' RS="/" BBS-list This sets `RS' to `/' before processing `BBS-list'. Using an unusual character such as `/' for the record separator produces correct behavior in the vast majority of cases. However, the following (extreme) pipeline prints a surprising `1': $ echo | awk 'BEGIN { RS = "a" } ; { print NF }' -| 1 There is one field, consisting of a newline. The value of the built-in variable `NF' is the number of fields in the current record. Reaching the end of an input file terminates the current input record, even if the last character in the file is not the character in `RS'. (d.c.) The empty string `""' (a string without any characters) has a special meaning as the value of `RS'. It means that records are separated by one or more blank lines and nothing else. *Note Multiple Line::, for more details. If you change the value of `RS' in the middle of an `awk' run, the new value is used to delimit subsequent records, but the record currently being processed, as well as records already processed, are not affected. After the end of the record has been determined, `gawk' sets the variable `RT' to the text in the input that matched `RS'. When using `gawk', the value of `RS' is not limited to a one-character string. It can be any regular expression (*note Regexp::). In general, each record ends at the next string that matches the regular expression; the next record starts at the end of the matching string. This general rule is actually at work in the usual case, where `RS' contains just a newline: a record ends at the beginning of the next matching string (the next newline in the input), and the following record starts just after the end of this string (at the first character of the following line). The newline, because it matches `RS', is not part of either record. When `RS' is a single character, `RT' contains the same single character. However, when `RS' is a regular expression, `RT' contains the actual input text that matched the regular expression. The following example illustrates both of these features. It sets `RS' equal to a regular expression that matches either a newline or a series of one or more uppercase letters with optional leading and/or trailing whitespace: $ echo record 1 AAAA record 2 BBBB record 3 | > gawk 'BEGIN { RS = "\n|( *[[:upper:]]+ *)" } > { print "Record =", $0, "and RT =", RT }' -| Record = record 1 and RT = AAAA -| Record = record 2 and RT = BBBB -| Record = record 3 and RT = -| The final line of output has an extra blank line. This is because the value of `RT' is a newline, and the `print' statement supplies its own terminating newline. *Note Simple Sed::, for a more useful example of `RS' as a regexp and `RT'. If you set `RS' to a regular expression that allows optional trailing text, such as `RS = "abc(XYZ)?"' it is possible, due to implementation constraints, that `gawk' may match the leading part of the regular expression, but not the trailing part, particularly if the input text that could match the trailing part is fairly long. `gawk' attempts to avoid this problem, but currently, there's no guarantee that this will never happen. The use of `RS' as a regular expression and the `RT' variable are `gawk' extensions; they are not available in compatibility mode (*note Options::). In compatibility mode, only the first character of the value of `RS' is used to determine the end of the record. Advanced Notes: `RS = "\0"' Is Not Portable ------------------------------------------- There are times when you might want to treat an entire data file as a single record. The only way to make this happen is to give `RS' a value that you know doesn't occur in the input file. This is hard to do in a general way, such that a program always works for arbitrary input files. You might think that for text files, the NUL character, which consists of a character with all bits equal to zero, is a good value to use for `RS' in this case: BEGIN { RS = "\0" } # whole file becomes one record? `gawk' in fact accepts this, and uses the NUL character for the record separator. However, this usage is _not_ portable to other `awk' implementations. All other `awk' implementations(1) store strings internally as C-style strings. C strings use the NUL character as the string terminator. In effect, this means that `RS = "\0"' is the same as `RS = ""'. (d.c.) The best way to treat a whole file as a single record is to simply read the file in, one record at a time, concatenating each record onto the end of the previous ones. ---------- Footnotes ---------- (1) At least that we know about. 3.2 Examining Fields ==================== When `awk' reads an input record, the record is automatically "parsed" or separated by the interpreter into chunks called "fields". By default, fields are separated by "whitespace", like words in a line. Whitespace in `awk' means any string of one or more spaces, tabs, or newlines;(1) other characters, such as formfeed, vertical tab, etc. that are considered whitespace by other languages, are _not_ considered whitespace by `awk'. The purpose of fields is to make it more convenient for you to refer to these pieces of the record. You don't have to use them--you can operate on the whole record if you want--but fields are what make simple `awk' programs so powerful. A dollar-sign (`$') is used to refer to a field in an `awk' program, followed by the number of the field you want. Thus, `$1' refers to the first field, `$2' to the second, and so on. (Unlike the Unix shells, the field numbers are not limited to single digits. `$127' is the one hundred twenty-seventh field in the record.) For example, suppose the following is a line of input: This seems like a pretty nice example. Here the first field, or `$1', is `This', the second field, or `$2', is `seems', and so on. Note that the last field, `$7', is `example.'. Because there is no space between the `e' and the `.', the period is considered part of the seventh field. `NF' is a built-in variable whose value is the number of fields in the current record. `awk' automatically updates the value of `NF' each time it reads a record. No matter how many fields there are, the last field in a record can be represented by `$NF'. So, `$NF' is the same as `$7', which is `example.'. If you try to reference a field beyond the last one (such as `$8' when the record has only seven fields), you get the empty string. (If used in a numeric operation, you get zero.) The use of `$0', which looks like a reference to the "zero-th" field, is a special case: it represents the whole input record when you are not interested in specific fields. Here are some more examples: $ awk '$1 ~ /foo/ { print $0 }' BBS-list -| fooey 555-1234 2400/1200/300 B -| foot 555-6699 1200/300 B -| macfoo 555-6480 1200/300 A -| sabafoo 555-2127 1200/300 C This example prints each record in the file `BBS-list' whose first field contains the string `foo'. The operator `~' is called a "matching operator" (*note Regexp Usage::); it tests whether a string (here, the field `$1') matches a given regular expression. By contrast, the following example looks for `foo' in _the entire record_ and prints the first field and the last field for each matching input record: $ awk '/foo/ { print $1, $NF }' BBS-list -| fooey B -| foot B -| macfoo A -| sabafoo C ---------- Footnotes ---------- (1) In POSIX `awk', newlines are not considered whitespace for separating fields. 3.3 Nonconstant Field Numbers ============================= The number of a field does not need to be a constant. Any expression in the `awk' language can be used after a `$' to refer to a field. The value of the expression specifies the field number. If the value is a string, rather than a number, it is converted to a number. Consider this example: awk '{ print $NR }' Recall that `NR' is the number of records read so far: one in the first record, two in the second, etc. So this example prints the first field of the first record, the second field of the second record, and so on. For the twentieth record, field number 20 is printed; most likely, the record has fewer than 20 fields, so this prints a blank line. Here is another example of using expressions as field numbers: awk '{ print $(2*2) }' BBS-list `awk' evaluates the expression `(2*2)' and uses its value as the number of the field to print. The `*' sign represents multiplication, so the expression `2*2' evaluates to four. The parentheses are used so that the multiplication is done before the `$' operation; they are necessary whenever there is a binary operator in the field-number expression. This example, then, prints the hours of operation (the fourth field) for every line of the file `BBS-list'. (All of the `awk' operators are listed, in order of decreasing precedence, in *Note Precedence::.) If the field number you compute is zero, you get the entire record. Thus, `$(2-2)' has the same value as `$0'. Negative field numbers are not allowed; trying to reference one usually terminates the program. (The POSIX standard does not define what happens when you reference a negative field number. `gawk' notices this and terminates your program. Other `awk' implementations may behave differently.) As mentioned in *Note Fields::, `awk' stores the current record's number of fields in the built-in variable `NF' (also *note Built-in Variables::). The expression `$NF' is not a special feature--it is the direct consequence of evaluating `NF' and using its value as a field number. 3.4 Changing the Contents of a Field ==================================== The contents of a field, as seen by `awk', can be changed within an `awk' program; this changes what `awk' perceives as the current input record. (The actual input is untouched; `awk' _never_ modifies the input file.) Consider the following example and its output: $ awk '{ nboxes = $3 ; $3 = $3 - 10 > print nboxes, $3 }' inventory-shipped -| 25 15 -| 32 22 -| 24 14 ... The program first saves the original value of field three in the variable `nboxes'. The `-' sign represents subtraction, so this program reassigns field three, `$3', as the original value of field three minus ten: `$3 - 10'. (*Note Arithmetic Ops::.) Then it prints the original and new values for field three. (Someone in the warehouse made a consistent mistake while inventorying the red boxes.) For this to work, the text in field `$3' must make sense as a number; the string of characters must be converted to a number for the computer to do arithmetic on it. The number resulting from the subtraction is converted back to a string of characters that then becomes field three. *Note Conversion::. When the value of a field is changed (as perceived by `awk'), the text of the input record is recalculated to contain the new field where the old one was. In other words, `$0' changes to reflect the altered field. Thus, this program prints a copy of the input file, with 10 subtracted from the second field of each line: $ awk '{ $2 = $2 - 10; print $0 }' inventory-shipped -| Jan 3 25 15 115 -| Feb 5 32 24 226 -| Mar 5 24 34 228 ... It is also possible to also assign contents to fields that are out of range. For example: $ awk '{ $6 = ($5 + $4 + $3 + $2) > print $6 }' inventory-shipped -| 168 -| 297 -| 301 ... We've just created `$6', whose value is the sum of fields `$2', `$3', `$4', and `$5'. The `+' sign represents addition. For the file `inventory-shipped', `$6' represents the total number of parcels shipped for a particular month. Creating a new field changes `awk''s internal copy of the current input record, which is the value of `$0'. Thus, if you do `print $0' after adding a field, the record printed includes the new field, with the appropriate number of field separators between it and the previously existing fields. This recomputation affects and is affected by `NF' (the number of fields; *note Fields::). For example, the value of `NF' is set to the number of the highest field you create. The exact format of `$0' is also affected by a feature that has not been discussed yet: the "output field separator", `OFS', used to separate the fields (*note Output Separators::). Note, however, that merely _referencing_ an out-of-range field does _not_ change the value of either `$0' or `NF'. Referencing an out-of-range field only produces an empty string. For example: if ($(NF+1) != "") print "can't happen" else print "everything is normal" should print `everything is normal', because `NF+1' is certain to be out of range. (*Note If Statement::, for more information about `awk''s `if-else' statements. *Note Typing and Comparison::, for more information about the `!=' operator.) It is important to note that making an assignment to an existing field changes the value of `$0' but does not change the value of `NF', even when you assign the empty string to a field. For example: $ echo a b c d | awk '{ OFS = ":"; $2 = "" > print $0; print NF }' -| a::c:d -| 4 The field is still there; it just has an empty value, denoted by the two colons between `a' and `c'. This example shows what happens if you create a new field: $ echo a b c d | awk '{ OFS = ":"; $2 = ""; $6 = "new" > print $0; print NF }' -| a::c:d::new -| 6 The intervening field, `$5', is created with an empty value (indicated by the second pair of adjacent colons), and `NF' is updated with the value six. Decrementing `NF' throws away the values of the fields after the new value of `NF' and recomputes `$0'. (d.c.) Here is an example: $ echo a b c d e f | awk '{ print "NF =", NF; > NF = 3; print $0 }' -| NF = 6 -| a b c *Caution:* Some versions of `awk' don't rebuild `$0' when `NF' is decremented. Caveat emptor. Finally, there are times when it is convenient to force `awk' to rebuild the entire record, using the current value of the fields and `OFS'. To do this, use the seemingly innocuous assignment: $1 = $1 # force record to be reconstituted print $0 # or whatever else with $0 This forces `awk' rebuild the record. It does help to add a comment, as we've shown here. There is a flip side to the relationship between `$0' and the fields. Any assignment to `$0' causes the record to be reparsed into fields using the _current_ value of `FS'. This also applies to any built-in function that updates `$0', such as `sub' and `gsub' (*note String Functions::). 3.5 Specifying How Fields Are Separated ======================================= The "field separator", which is either a single character or a regular expression, controls the way `awk' splits an input record into fields. `awk' scans the input record for character sequences that match the separator; the fields themselves are the text between the matches. In the examples that follow, we use the bullet symbol (*) to represent spaces in the output. If the field separator is `oo', then the following line: moo goo gai pan is split into three fields: `m', `*g', and `*gai*pan'. Note the leading spaces in the values of the second and third fields. The field separator is represented by the built-in variable `FS'. Shell programmers take note: `awk' does _not_ use the name `IFS' that is used by the POSIX-compliant shells (such as the Unix Bourne shell, `sh', or `bash'). The value of `FS' can be changed in the `awk' program with the assignment operator, `=' (*note Assignment Ops::). Often the right time to do this is at the beginning of execution before any input has been processed, so that the very first record is read with the proper separator. To do this, use the special `BEGIN' pattern (*note BEGIN/END::). For example, here we set the value of `FS' to the string `","': awk 'BEGIN { FS = "," } ; { print $2 }' Given the input line: John Q. Smith, 29 Oak St., Walamazoo, MI 42139 this `awk' program extracts and prints the string `*29*Oak*St.'. Sometimes the input data contains separator characters that don't separate fields the way you thought they would. For instance, the person's name in the example we just used might have a title or suffix attached, such as: John Q. Smith, LXIX, 29 Oak St., Walamazoo, MI 42139 The same program would extract `*LXIX', instead of `*29*Oak*St.'. If you were expecting the program to print the address, you would be surprised. The moral is to choose your data layout and separator characters carefully to prevent such problems. (If the data is not in a form that is easy to process, perhaps you can massage it first with a separate `awk' program.) Fields are normally separated by whitespace sequences (spaces, tabs, and newlines), not by single spaces. Two spaces in a row do not delimit an empty field. The default value of the field separator `FS' is a string containing a single space, `" "'. If `awk' interpreted this value in the usual way, each space character would separate fields, so two spaces in a row would make an empty field between them. The reason this does not happen is that a single space as the value of `FS' is a special case--it is taken to specify the default manner of delimiting fields. If `FS' is any other single character, such as `","', then each occurrence of that character separates two fields. Two consecutive occurrences delimit an empty field. If the character occurs at the beginning or the end of the line, that too delimits an empty field. The space character is the only single character that does not follow these rules. 3.5.1 Using Regular Expressions to Separate Fields -------------------------------------------------- The previous node discussed the use of single characters or simple strings as the value of `FS'. More generally, the value of `FS' may be a string containing any regular expression. In this case, each match in the record for the regular expression separates fields. For example, the assignment: FS = ", \t" makes every area of an input line that consists of a comma followed by a space and a TAB into a field separator. (`\t' is an "escape sequence" that stands for a TAB; *note Escape Sequences::, for the complete list of similar escape sequences.) For a less trivial example of a regular expression, try using single spaces to separate fields the way single commas are used. `FS' can be set to `"[ ]"' (left bracket, space, right bracket). This regular expression matches a single space and nothing else (*note Regexp::). There is an important difference between the two cases of `FS = " "' (a single space) and `FS = "[ \t\n]+"' (a regular expression matching one or more spaces, tabs, or newlines). For both values of `FS', fields are separated by "runs" (multiple adjacent occurrences) of spaces, tabs, and/or newlines. However, when the value of `FS' is `" "', `awk' first strips leading and trailing whitespace from the record and then decides where the fields are. For example, the following pipeline prints `b': $ echo ' a b c d ' | awk '{ print $2 }' -| b However, this pipeline prints `a' (note the extra spaces around each letter): $ echo ' a b c d ' | awk 'BEGIN { FS = "[ \t\n]+" } > { print $2 }' -| a In this case, the first field is "null" or empty. The stripping of leading and trailing whitespace also comes into play whenever `$0' is recomputed. For instance, study this pipeline: $ echo ' a b c d' | awk '{ print; $2 = $2; print }' -| a b c d -| a b c d The first `print' statement prints the record as it was read, with leading whitespace intact. The assignment to `$2' rebuilds `$0' by concatenating `$1' through `$NF' together, separated by the value of `OFS'. Because the leading whitespace was ignored when finding `$1', it is not part of the new `$0'. Finally, the last `print' statement prints the new `$0'. 3.5.2 Making Each Character a Separate Field -------------------------------------------- There are times when you may want to examine each character of a record separately. This can be done in `gawk' by simply assigning the null string (`""') to `FS'. In this case, each individual character in the record becomes a separate field. For example: $ echo a b | gawk 'BEGIN { FS = "" } > { > for (i = 1; i <= NF; i = i + 1) > print "Field", i, "is", $i > }' -| Field 1 is a -| Field 2 is -| Field 3 is b Traditionally, the behavior of `FS' equal to `""' was not defined. In this case, most versions of Unix `awk' simply treat the entire record as only having one field. (d.c.) In compatibility mode (*note Options::), if `FS' is the null string, then `gawk' also behaves this way. 3.5.3 Setting `FS' from the Command Line ---------------------------------------- `FS' can be set on the command line. Use the `-F' option to do so. For example: awk -F, 'PROGRAM' INPUT-FILES sets `FS' to the `,' character. Notice that the option uses an uppercase `F' instead of a lowercase `f'. The latter option (`-f') specifies a file containing an `awk' program. Case is significant in command-line options: the `-F' and `-f' options have nothing to do with each other. You can use both options at the same time to set the `FS' variable _and_ get an `awk' program from a file. The value used for the argument to `-F' is processed in exactly the same way as assignments to the built-in variable `FS'. Any special characters in the field separator must be escaped appropriately. For example, to use a `\' as the field separator on the command line, you would have to type: # same as FS = "\\" awk -F\\\\ '...' files ... Because `\' is used for quoting in the shell, `awk' sees `-F\\'. Then `awk' processes the `\\' for escape characters (*note Escape Sequences::), finally yielding a single `\' to use for the field separator. As a special case, in compatibility mode (*note Options::), if the argument to `-F' is `t', then `FS' is set to the TAB character. If you type `-F\t' at the shell, without any quotes, the `\' gets deleted, so `awk' figures that you really want your fields to be separated with tabs and not `t's. Use `-v FS="t"' or `-F"[t]"' on the command line if you really do want to separate your fields with `t's. For example, let's use an `awk' program file called `baud.awk' that contains the pattern `/300/' and the action `print $1': /300/ { print $1 } Let's also set `FS' to be the `-' character and run the program on the file `BBS-list'. The following command prints a list of the names of the bulletin boards that operate at 300 baud and the first three digits of their phone numbers: $ awk -F- -f baud.awk BBS-list -| aardvark 555 -| alpo -| barfly 555 -| bites 555 -| camelot 555 -| core 555 -| fooey 555 -| foot 555 -| macfoo 555 -| sdace 555 -| sabafoo 555 Note the second line of output. The second line in the original file looked like this: alpo-net 555-3412 2400/1200/300 A The `-' as part of the system's name was used as the field separator, instead of the `-' in the phone number that was originally intended. This demonstrates why you have to be careful in choosing your field and record separators. Perhaps the most common use of a single character as the field separator occurs when processing the Unix system password file. On many Unix systems, each user has a separate entry in the system password file, one line per user. The information in these lines is separated by colons. The first field is the user's logon name and the second is the user's (encrypted or shadow) password. A password file entry might look like this: arnold:xyzzy:2076:10:Arnold Robbins:/home/arnold:/bin/bash The following program searches the system password file and prints the entries for users who have no password: awk -F: '$2 == ""' /etc/passwd 3.5.4 Field-Splitting Summary ----------------------------- It is important to remember that when you assign a string constant as the value of `FS', it undergoes normal `awk' string processing. For example, with Unix `awk' and `gawk', the assignment `FS = "\.."' assigns the character string `".."' to `FS' (the backslash is stripped). This creates a regexp meaning "fields are separated by occurrences of any two characters." If instead you want fields to be separated by a literal period followed by any single character, use `FS = "\\.."'. The following table summarizes how fields are split, based on the value of `FS' (`==' means "is equal to"): `FS == " "' Fields are separated by runs of whitespace. Leading and trailing whitespace are ignored. This is the default. `FS == ANY OTHER SINGLE CHARACTER' Fields are separated by each occurrence of the character. Multiple successive occurrences delimit empty fields, as do leading and trailing occurrences. The character can even be a regexp metacharacter; it does not need to be escaped. `FS == REGEXP' Fields are separated by occurrences of characters that match REGEXP. Leading and trailing matches of REGEXP delimit empty fields. `FS == ""' Each individual character in the record becomes a separate field. (This is a `gawk' extension; it is not specified by the POSIX standard.) Advanced Notes: Changing `FS' Does Not Affect the Fields -------------------------------------------------------- According to the POSIX standard, `awk' is supposed to behave as if each record is split into fields at the time it is read. In particular, this means that if you change the value of `FS' after a record is read, the value of the fields (i.e., how they were split) should reflect the old value of `FS', not the new one. However, many implementations of `awk' do not work this way. Instead, they defer splitting the fields until a field is actually referenced. The fields are split using the _current_ value of `FS'! (d.c.) This behavior can be difficult to diagnose. The following example illustrates the difference between the two methods. (The `sed'(1) command prints just the first line of `/etc/passwd'.) sed 1q /etc/passwd | awk '{ FS = ":" ; print $1 }' which usually prints: root on an incorrect implementation of `awk', while `gawk' prints something like: root:nSijPlPhZZwgE:0:0:Root:/: Advanced Notes: `FS' and `IGNORECASE' ------------------------------------- The `IGNORECASE' variable (*note User-modified::) affects field splitting _only_ when the value of `FS' is a regexp. It has no effect when `FS' is a single character, even if that character is a letter. Thus, in the following code: FS = "c" IGNORECASE = 1 $0 = "aCa" print $1 The output is `aCa'. If you really want to split fields on an alphabetic character while ignoring case, use a regexp that will do it for you. E.g., `FS = "[c]"'. In this case, `IGNORECASE' will take effect. ---------- Footnotes ---------- (1) The `sed' utility is a "stream editor." Its behavior is also defined by the POSIX standard. 3.6 Reading Fixed-Width Data ============================ (This minor node discusses an advanced feature of `awk'. If you are a novice `awk' user, you might want to skip it on the first reading.) `gawk' version 2.13 introduced a facility for dealing with fixed-width fields with no distinctive field separator. For example, data of this nature arises in the input for old Fortran programs where numbers are run together, or in the output of programs that did not anticipate the use of their output as input for other programs. An example of the latter is a table where all the columns are lined up by the use of a variable number of spaces and _empty fields are just spaces_. Clearly, `awk''s normal field splitting based on `FS' does not work well in this case. Although a portable `awk' program can use a series of `substr' calls on `$0' (*note String Functions::), this is awkward and inefficient for a large number of fields. The splitting of an input record into fixed-width fields is specified by assigning a string containing space-separated numbers to the built-in variable `FIELDWIDTHS'. Each number specifies the width of the field, _including_ columns between fields. If you want to ignore the columns between fields, you can specify the width as a separate field that is subsequently ignored. It is a fatal error to supply a field width that is not a positive number. The following data is the output of the Unix `w' utility. It is useful to illustrate the use of `FIELDWIDTHS': 10:06pm up 21 days, 14:04, 23 users User tty login idle JCPU PCPU what hzuo ttyV0 8:58pm 9 5 vi p24.tex hzang ttyV3 6:37pm 50 -csh eklye ttyV5 9:53pm 7 1 em thes.tex dportein ttyV6 8:17pm 1:47 -csh gierd ttyD3 10:00pm 1 elm dave ttyD4 9:47pm 4 4 w brent ttyp0 26Jun91 4:46 26:46 4:41 bash dave ttyq4 26Jun9115days 46 46 wnewmail The following program takes the above input, converts the idle time to number of seconds, and prints out the first two fields and the calculated idle time: NOTE: This program uses a number of `awk' features that haven't been introduced yet. BEGIN { FIELDWIDTHS = "9 6 10 6 7 7 35" } NR > 2 { idle = $4 sub(/^ */, "", idle) # strip leading spaces if (idle == "") idle = 0 if (idle ~ /:/) { split(idle, t, ":") idle = t[1] * 60 + t[2] } if (idle ~ /days/) idle *= 24 * 60 * 60 print $1, $2, idle } Running the program on the data produces the following results: hzuo ttyV0 0 hzang ttyV3 50 eklye ttyV5 0 dportein ttyV6 107 gierd ttyD3 1 dave ttyD4 0 brent ttyp0 286 dave ttyq4 1296000 Another (possibly more practical) example of fixed-width input data is the input from a deck of balloting cards. In some parts of the United States, voters mark their choices by punching holes in computer cards. These cards are then processed to count the votes for any particular candidate or on any particular issue. Because a voter may choose not to vote on some issue, any column on the card may be empty. An `awk' program for processing such data could use the `FIELDWIDTHS' feature to simplify reading the data. (Of course, getting `gawk' to run on a system with card readers is another story!) Assigning a value to `FS' causes `gawk' to use `FS' for field splitting again. Use `FS = FS' to make this happen, without having to know the current value of `FS'. In order to tell which kind of field splitting is in effect, use `PROCINFO["FS"]' (*note Auto-set::). The value is `"FS"' if regular field splitting is being used, or it is `"FIELDWIDTHS"' if fixed-width field splitting is being used: if (PROCINFO["FS"] == "FS") REGULAR FIELD SPLITTING ... else FIXED-WIDTH FIELD SPLITTING ... This information is useful when writing a function that needs to temporarily change `FS' or `FIELDWIDTHS', read some records, and then restore the original settings (*note Passwd Functions::, for an example of such a function). 3.7 Multiple-Line Records ========================= In some databases, a single line cannot conveniently hold all the information in one entry. In such cases, you can use multiline records. The first step in doing this is to choose your data format. One technique is to use an unusual character or string to separate records. For example, you could use the formfeed character (written `\f' in `awk', as in C) to separate them, making each record a page of the file. To do this, just set the variable `RS' to `"\f"' (a string containing the formfeed character). Any other character could equally well be used, as long as it won't be part of the data in a record. Another technique is to have blank lines separate records. By a special dispensation, an empty string as the value of `RS' indicates that records are separated by one or more blank lines. When `RS' is set to the empty string, each record always ends at the first blank line encountered. The next record doesn't start until the first nonblank line that follows. No matter how many blank lines appear in a row, they all act as one record separator. (Blank lines must be completely empty; lines that contain only whitespace do not count.) You can achieve the same effect as `RS = ""' by assigning the string `"\n\n+"' to `RS'. This regexp matches the newline at the end of the record and one or more blank lines after the record. In addition, a regular expression always matches the longest possible sequence when there is a choice (*note Leftmost Longest::). So the next record doesn't start until the first nonblank line that follows--no matter how many blank lines appear in a row, they are considered one record separator. There is an important difference between `RS = ""' and `RS = "\n\n+"'. In the first case, leading newlines in the input data file are ignored, and if a file ends without extra blank lines after the last record, the final newline is removed from the record. In the second case, this special processing is not done. (d.c.) Now that the input is separated into records, the second step is to separate the fields in the record. One way to do this is to divide each of the lines into fields in the normal manner. This happens by default as the result of a special feature. When `RS' is set to the empty string, _and_ `FS' is a set to a single character, the newline character _always_ acts as a field separator. This is in addition to whatever field separations result from `FS'.(1) The original motivation for this special exception was probably to provide useful behavior in the default case (i.e., `FS' is equal to `" "'). This feature can be a problem if you really don't want the newline character to separate fields, because there is no way to prevent it. However, you can work around this by using the `split' function to break up the record manually (*note String Functions::). If you have a single character field separator, you can work around the special feature in a different way, by making `FS' into a regexp for that single character. For example, if the field separator is a percent character, instead of `FS = "%"', use `FS = "[%]"'. Another way to separate fields is to put each field on a separate line: to do this, just set the variable `FS' to the string `"\n"'. (This single character seperator matches a single newline.) A practical example of a data file organized this way might be a mailing list, where each entry is separated by blank lines. Consider a mailing list in a file named `addresses', which looks like this: Jane Doe 123 Main Street Anywhere, SE 12345-6789 John Smith 456 Tree-lined Avenue Smallville, MW 98765-4321 ... A simple program to process this file is as follows: # addrs.awk --- simple mailing list program # Records are separated by blank lines. # Each line is one field. BEGIN { RS = "" ; FS = "\n" } { print "Name is:", $1 print "Address is:", $2 print "City and State are:", $3 print "" } Running the program produces the following output: $ awk -f addrs.awk addresses -| Name is: Jane Doe -| Address is: 123 Main Street -| City and State are: Anywhere, SE 12345-6789 -| -| Name is: John Smith -| Address is: 456 Tree-lined Avenue -| City and State are: Smallville, MW 98765-4321 -| ... *Note Labels Program::, for a more realistic program that deals with address lists. The following table summarizes how records are split, based on the value of `RS'. (`==' means "is equal to.") `RS == "\n"' Records are separated by the newline character (`\n'). In effect, every line in the data file is a separate record, including blank lines. This is the default. `RS == ANY SINGLE CHARACTER' Records are separated by each occurrence of the character. Multiple successive occurrences delimit empty records. `RS == ""' Records are separated by runs of blank lines. The newline character always serves as a field separator, in addition to whatever value `FS' may have. Leading and trailing newlines in a file are ignored. `RS == REGEXP' Records are separated by occurrences of characters that match REGEXP. Leading and trailing matches of REGEXP delimit empty records. (This is a `gawk' extension; it is not specified by the POSIX standard.) In all cases, `gawk' sets `RT' to the input text that matched the value specified by `RS'. ---------- Footnotes ---------- (1) When `FS' is the null string (`""') or a regexp, this special feature of `RS' does not apply. It does apply to the default field separator of a single space: `FS = " "'. 3.8 Explicit Input with `getline' ================================= So far we have been getting our input data from `awk''s main input stream--either the standard input (usually your terminal, sometimes the output from another program) or from the files specified on the command line. The `awk' language has a special built-in command called `getline' that can be used to read input under your explicit control. The `getline' command is used in several different ways and should _not_ be used by beginners. The examples that follow the explanation of the `getline' command include material that has not been covered yet. Therefore, come back and study the `getline' command _after_ you have reviewed the rest of this Info file and have a good knowledge of how `awk' works. The `getline' command returns one if it finds a record and zero if it encounters the end of the file. If there is some error in getting a record, such as a file that cannot be opened, then `getline' returns -1. In this case, `gawk' sets the variable `ERRNO' to a string describing the error that occurred. In the following examples, COMMAND stands for a string value that represents a shell command. 3.8.1 Using `getline' with No Arguments --------------------------------------- The `getline' command can be used without arguments to read input from the current input file. All it does in this case is read the next input record and split it up into fields. This is useful if you've finished processing the current record, but want to do some special processing on the next record _right now_. For example: { if ((t = index($0, "/*")) != 0) { # value of `tmp' will be "" if t is 1 tmp = substr($0, 1, t - 1) u = index(substr($0, t + 2), "*/") while (u == 0) { if (getline <= 0) { m = "unexpected EOF or error" m = (m ": " ERRNO) print m > "/dev/stderr" exit } t = -1 u = index($0, "*/") } # substr expression will be "" if */ # occurred at end of line $0 = tmp substr($0, u + 2) } print $0 } This `awk' program deletes all C-style comments (`/* ... */') from the input. By replacing the `print $0' with other statements, you could perform more complicated processing on the decommented input, such as searching for matches of a regular expression. (This program has a subtle problem--it does not work if one comment ends and another begins on the same line.) This form of the `getline' command sets `NF', `NR', `FNR', and the value of `$0'. NOTE: The new value of `$0' is used to test the patterns of any subsequent rules. The original value of `$0' that triggered the rule that executed `getline' is lost. By contrast, the `next' statement reads a new record but immediately begins processing it normally, starting with the first rule in the program. *Note Next Statement::. 3.8.2 Using `getline' into a Variable ------------------------------------- You can use `getline VAR' to read the next record from `awk''s input into the variable VAR. No other processing is done. For example, suppose the next line is a comment or a special string, and you want to read it without triggering any rules. This form of `getline' allows you to read that line and store it in a variable so that the main read-a-line-and-check-each-rule loop of `awk' never sees it. The following example swaps every two lines of input: { if ((getline tmp) > 0) { print tmp print $0 } else print $0 } It takes the following list: wan tew free phore and produces these results: tew wan phore free The `getline' command used in this way sets only the variables `NR' and `FNR' (and of course, VAR). The record is not split into fields, so the values of the fields (including `$0') and the value of `NF' do not change. 3.8.3 Using `getline' from a File --------------------------------- Use `getline < FILE' to read the next record from FILE. Here FILE is a string-valued expression that specifies the file name. `< FILE' is called a "redirection" because it directs input to come from a different place. For example, the following program reads its input record from the file `secondary.input' when it encounters a first field with a value equal to 10 in the current input file: { if ($1 == 10) { getline < "secondary.input" print } else print } Because the main input stream is not used, the values of `NR' and `FNR' are not changed. However, the record it reads is split into fields in the normal manner, so the values of `$0' and the other fields are changed, resulting in a new value of `NF'. According to POSIX, `getline < EXPRESSION' is ambiguous if EXPRESSION contains unparenthesized operators other than `$'; for example, `getline < dir "/" file' is ambiguous because the concatenation operator is not parenthesized. You should write it as `getline < (dir "/" file)' if you want your program to be portable to other `awk' implementations. 3.8.4 Using `getline' into a Variable from a File ------------------------------------------------- Use `getline VAR < FILE' to read input from the file FILE, and put it in the variable VAR. As above, FILE is a string-valued expression that specifies the file from which to read. In this version of `getline', none of the built-in variables are changed and the record is not split into fields. The only variable changed is VAR. For example, the following program copies all the input files to the output, except for records that say `@include FILENAME'. Such a record is replaced by the contents of the file FILENAME: { if (NF == 2 && $1 == "@include") { while ((getline line < $2) > 0) print line close($2) } else print } Note here how the name of the extra input file is not built into the program; it is taken directly from the data, specifically from the second field on the `@include' line. The `close' function is called to ensure that if two identical `@include' lines appear in the input, the entire specified file is included twice. *Note Close Files And Pipes::. One deficiency of this program is that it does not process nested `@include' statements (i.e., `@include' statements in included files) the way a true macro preprocessor would. *Note Igawk Program::, for a program that does handle nested `@include' statements. 3.8.5 Using `getline' from a Pipe --------------------------------- The output of a command can also be piped into `getline', using `COMMAND | getline'. In this case, the string COMMAND is run as a shell command and its output is piped into `awk' to be used as input. This form of `getline' reads one record at a time from the pipe. For example, the following program copies its input to its output, except for lines that begin with `@execute', which are replaced by the output produced by running the rest of the line as a shell command: { if ($1 == "@execute") { tmp = substr($0, 10) while ((tmp | getline) > 0) print close(tmp) } else print } The `close' function is called to ensure that if two identical `@execute' lines appear in the input, the command is run for each one. *Note Close Files And Pipes::. Given the input: foo bar baz @execute who bletch the program might produce: foo bar baz arnold ttyv0 Jul 13 14:22 miriam ttyp0 Jul 13 14:23 (murphy:0) bill ttyp1 Jul 13 14:23 (murphy:0) bletch Notice that this program ran the command `who' and printed the previous result. (If you try this program yourself, you will of course get different results, depending upon who is logged in on your system.) This variation of `getline' splits the record into fields, sets the value of `NF', and recomputes the value of `$0'. The values of `NR' and `FNR' are not changed. According to POSIX, `EXPRESSION | getline' is ambiguous if EXPRESSION contains unparenthesized operators other than `$'--for example, `"echo " "date" | getline' is ambiguous because the concatenation operator is not parenthesized. You should write it as `("echo " "date") | getline' if you want your program to be portable to other `awk' implementations. 3.8.6 Using `getline' into a Variable from a Pipe ------------------------------------------------- When you use `COMMAND | getline VAR', the output of COMMAND is sent through a pipe to `getline' and into the variable VAR. For example, the following program reads the current date and time into the variable `current_time', using the `date' utility, and then prints it: BEGIN { "date" | getline current_time close("date") print "Report printed on " current_time } In this version of `getline', none of the built-in variables are changed and the record is not split into fields. According to POSIX, `EXPRESSION | getline VAR' is ambiguous if EXPRESSION contains unparenthesized operators other than `$'; for example, `"echo " "date" | getline VAR' is ambiguous because the concatenation operator is not parenthesized. You should write it as `("echo " "date") | getline VAR' if you want your program to be portable to other `awk' implementations. 3.8.7 Using `getline' from a Coprocess -------------------------------------- Input into `getline' from a pipe is a one-way operation. The command that is started with `COMMAND | getline' only sends data _to_ your `awk' program. On occasion, you might want to send data to another program for processing and then read the results back. `gawk' allows you start a "coprocess", with which two-way communications are possible. This is done with the `|&' operator. Typically, you write data to the coprocess first and then read results back, as shown in the following: print "SOME QUERY" |& "db_server" "db_server" |& getline which sends a query to `db_server' and then reads the results. The values of `NR' and `FNR' are not changed, because the main input stream is not used. However, the record is split into fields in the normal manner, thus changing the values of `$0', of the other fields, and of `NF'. Coprocesses are an advanced feature. They are discussed here only because this is the minor node on `getline'. *Note Two-way I/O::, where coprocesses are discussed in more detail. 3.8.8 Using `getline' into a Variable from a Coprocess ------------------------------------------------------ When you use `COMMAND |& getline VAR', the output from the coprocess COMMAND is sent through a two-way pipe to `getline' and into the variable VAR. In this version of `getline', none of the built-in variables are changed and the record is not split into fields. The only variable changed is VAR. Coprocesses are an advanced feature. They are discussed here only because this is the minor node on `getline'. *Note Two-way I/O::, where coprocesses are discussed in more detail. 3.8.9 Points to Remember About `getline' ---------------------------------------- Here are some miscellaneous points about `getline' that you should bear in mind: * When `getline' changes the value of `$0' and `NF', `awk' does _not_ automatically jump to the start of the program and start testing the new record against every pattern. However, the new record is tested against any subsequent rules. * Many `awk' implementations limit the number of pipelines that an `awk' program may have open to just one. In `gawk', there is no such limit. You can open as many pipelines (and coprocesses) as the underlying operating system permits. * An interesting side effect occurs if you use `getline' without a redirection inside a `BEGIN' rule. Because an unredirected `getline' reads from the command-line data files, the first `getline' command causes `awk' to set the value of `FILENAME'. Normally, `FILENAME' does not have a value inside `BEGIN' rules, because you have not yet started to process the command-line data files. (d.c.) (*Note BEGIN/END::, also *note Auto-set::.) * Using `FILENAME' with `getline' (`getline < FILENAME') is likely to be a source for confusion. `awk' opens a separate input stream from the current input file. However, by not using a variable, `$0' and `NR' are still updated. If you're doing this, it's probably by accident, and you should reconsider what it is you're trying to accomplish. 3.8.10 Summary of `getline' Variants ------------------------------------ *Note table-getline-variants:: summarizes the eight variants of `getline', listing which built-in variables are set by each one. Variant Effect -------------------------------------------------------------------------- `getline' Sets `$0', `NF', `FNR', and `NR' `getline' VAR Sets VAR, `FNR', and `NR' `getline <' FILE Sets `$0' and `NF' `getline VAR < FILE' Sets VAR COMMAND `| getline' Sets `$0' and `NF' COMMAND `| getline' VAR Sets VAR COMMAND `|& getline' Sets `$0' and `NF'. This is a `gawk' extension COMMAND `|& getline' VAR Sets VAR. This is a `gawk' extension Table 3.1: getline Variants and What They Set 4 Printing Output ***************** One of the most common programming actions is to "print", or output, some or all of the input. Use the `print' statement for simple output, and the `printf' statement for fancier formatting. The `print' statement is not limited when computing _which_ values to print. However, with two exceptions, you cannot specify _how_ to print them--how many columns, whether to use exponential notation or not, and so on. (For the exceptions, *note Output Separators::, and *Note OFMT::.) For printing with specifications, you need the `printf' statement (*note Printf::). Besides basic and formatted printing, this major node also covers I/O redirections to files and pipes, introduces the special file names that `gawk' processes internally, and discusses the `close' built-in function. 4.1 The `print' Statement ========================= The `print' statement is used to produce output with simple, standardized formatting. Specify only the strings or numbers to print, in a list separated by commas. They are output, separated by single spaces, followed by a newline. The statement looks like this: print ITEM1, ITEM2, ... The entire list of items may be optionally enclosed in parentheses. The parentheses are necessary if any of the item expressions uses the `>' relational operator; otherwise it could be confused with a redirection (*note Redirection::). The items to print can be constant strings or numbers, fields of the current record (such as `$1'), variables, or any `awk' expression. Numeric values are converted to strings and then printed. The simple statement `print' with no items is equivalent to `print $0': it prints the entire current record. To print a blank line, use `print ""', where `""' is the empty string. To print a fixed piece of text, use a string constant, such as `"Don't Panic"', as one item. If you forget to use the double-quote characters, your text is taken as an `awk' expression, and you will probably get an error. Keep in mind that a space is printed between any two items. 4.2 Examples of `print' Statements ================================== Each `print' statement makes at least one line of output. However, it isn't limited to only one line. If an item value is a string that contains a newline, the newline is output along with the rest of the string. A single `print' statement can make any number of lines this way. The following is an example of printing a string that contains embedded newlines (the `\n' is an escape sequence, used to represent the newline character; *note Escape Sequences::): $ awk 'BEGIN { print "line one\nline two\nline three" }' -| line one -| line two -| line three The next example, which is run on the `inventory-shipped' file, prints the first two fields of each input record, with a space between them: $ awk '{ print $1, $2 }' inventory-shipped -| Jan 13 -| Feb 15 -| Mar 15 ... A common mistake in using the `print' statement is to omit the comma between two items. This often has the effect of making the items run together in the output, with no space. The reason for this is that juxtaposing two string expressions in `awk' means to concatenate them. Here is the same program, without the comma: $ awk '{ print $1 $2 }' inventory-shipped -| Jan13 -| Feb15 -| Mar15 ... To someone unfamiliar with the `inventory-shipped' file, neither example's output makes much sense. A heading line at the beginning would make it clearer. Let's add some headings to our table of months (`$1') and green crates shipped (`$2'). We do this using the `BEGIN' pattern (*note BEGIN/END::) so that the headings are only printed once: awk 'BEGIN { print "Month Crates" print "----- ------" } { print $1, $2 }' inventory-shipped When run, the program prints the following: Month Crates ----- ------ Jan 13 Feb 15 Mar 15 ... The only problem, however, is that the headings and the table data don't line up! We can fix this by printing some spaces between the two fields: awk 'BEGIN { print "Month Crates" print "----- ------" } { print $1, " ", $2 }' inventory-shipped Lining up columns this way can get pretty complicated when there are many columns to fix. Counting spaces for two or three columns is simple, but any more than this can take up a lot of time. This is why the `printf' statement was created (*note Printf::); one of its specialties is lining up columns of data. NOTE: You can continue either a `print' or `printf' statement simply by putting a newline after any comma (*note Statements/Lines::). 4.3 Output Separators ===================== As mentioned previously, a `print' statement contains a list of items separated by commas. In the output, the items are normally separated by single spaces. However, this doesn't need to be the case; a single space is only the default. Any string of characters may be used as the "output field separator" by setting the built-in variable `OFS'. The initial value of this variable is the string `" "'--that is, a single space. The output from an entire `print' statement is called an "output record". Each `print' statement outputs one output record, and then outputs a string called the "output record separator" (or `ORS'). The initial value of `ORS' is the string `"\n"'; i.e., a newline character. Thus, each `print' statement normally makes a separate line. In order to change how output fields and records are separated, assign new values to the variables `OFS' and `ORS'. The usual place to do this is in the `BEGIN' rule (*note BEGIN/END::), so that it happens before any input is processed. It can also be done with assignments on the command line, before the names of the input files, or using the `-v' command-line option (*note Options::). The following example prints the first and second fields of each input record, separated by a semicolon, with a blank line added after each newline: $ awk 'BEGIN { OFS = ";"; ORS = "\n\n" } > { print $1, $2 }' BBS-list -| aardvark;555-5553 -| -| alpo-net;555-3412 -| -| barfly;555-7685 ... If the value of `ORS' does not contain a newline, the program's output is run together on a single line. 4.4 Controlling Numeric Output with `print' =========================================== When the `print' statement is used to print numeric values, `awk' internally converts the number to a string of characters and prints that string. `awk' uses the `sprintf' function to do this conversion (*note String Functions::). For now, it suffices to say that the `sprintf' function accepts a "format specification" that tells it how to format numbers (or strings), and that there are a number of different ways in which numbers can be formatted. The different format specifications are discussed more fully in *Note Control Letters::. The built-in variable `OFMT' contains the default format specification that `print' uses with `sprintf' when it wants to convert a number to a string for printing. The default value of `OFMT' is `"%.6g"'. The way `print' prints numbers can be changed by supplying different format specifications as the value of `OFMT', as shown in the following example: $ awk 'BEGIN { > OFMT = "%.0f" # print numbers as integers (rounds) > print 17.23, 17.54 }' -| 17 18 According to the POSIX standard, `awk''s behavior is undefined if `OFMT' contains anything but a floating-point conversion specification. (d.c.) 4.5 Using `printf' Statements for Fancier Printing ================================================== For more precise control over the output format than what is normally provided by `print', use `printf'. `printf' can be used to specify the width to use for each item, as well as various formatting choices for numbers (such as what output base to use, whether to print an exponent, whether to print a sign, and how many digits to print after the decimal point). This is done by supplying a string, called the "format string", that controls how and where to print the other arguments. 4.5.1 Introduction to the `printf' Statement -------------------------------------------- A simple `printf' statement looks like this: printf FORMAT, ITEM1, ITEM2, ... The entire list of arguments may optionally be enclosed in parentheses. The parentheses are necessary if any of the item expressions use the `>' relational operator; otherwise, it can be confused with a redirection (*note Redirection::). The difference between `printf' and `print' is the FORMAT argument. This is an expression whose value is taken as a string; it specifies how to output each of the other arguments. It is called the "format string". The format string is very similar to that in the ISO C library function `printf'. Most of FORMAT is text to output verbatim. Scattered among this text are "format specifiers"--one per item. Each format specifier says to output the next item in the argument list at that place in the format. The `printf' statement does not automatically append a newline to its output. It outputs only what the format string specifies. So if a newline is needed, you must include one in the format string. The output separator variables `OFS' and `ORS' have no effect on `printf' statements. For example: $ awk 'BEGIN { > ORS = "\nOUCH!\n"; OFS = "+" > msg = "Dont Panic!" > printf "%s\n", msg > }' -| Dont Panic! Here, neither the `+' nor the `OUCH' appear when the message is printed. 4.5.2 Format-Control Letters ---------------------------- A format specifier starts with the character `%' and ends with a "format-control letter"--it tells the `printf' statement how to output one item. The format-control letter specifies what _kind_ of value to print. The rest of the format specifier is made up of optional "modifiers" that control _how_ to print the value, such as the field width. Here is a list of the format-control letters: `%c' This prints a number as an ASCII character; thus, `printf "%c", 65' outputs the letter `A'. (The output for a string value is the first character of the string.) `%d, %i' These are equivalent; they both print a decimal integer. (The `%i' specification is for compatibility with ISO C.) `%e, %E' These print a number in scientific (exponential) notation; for example: printf "%4.3e\n", 1950 prints `1.950e+03', with a total of four significant figures, three of which follow the decimal point. (The `4.3' represents two modifiers, discussed in the next node.) `%E' uses `E' instead of `e' in the output. `%f' This prints a number in floating-point notation. For example: printf "%4.3f", 1950 prints `1950.000', with a total of four significant figures, three of which follow the decimal point. (The `4.3' represents two modifiers, discussed in the next node.) On systems supporting IEEE 754 floating point format, values representing negative infinity are formatted as `-inf' or `-infinity', and positive infinity as `inf' and `-infinity'. The special "not a number" value formats as `-nan' or `nan'. `%F' Like `%f' but the infinity and "not a number" values are spelled using uppercase letters. The `%F' format is a POSIX extension to ISO C; not all systems support. On those that don't, `gawk' uses `%f' instead. `%g, %G' These print a number in either scientific notation or in floating-point notation, whichever uses fewer characters; if the result is printed in scientific notation, `%G' uses `E' instead of `e'. `%o' This prints an unsigned octal integer. `%s' This prints a string. `%u' This prints an unsigned decimal integer. (This format is of marginal use, because all numbers in `awk' are floating-point; it is provided primarily for compatibility with C.) `%x, %X' These print an unsigned hexadecimal integer; `%X' uses the letters `A' through `F' instead of `a' through `f'. `%%' This isn't a format-control letter, but it does have meaning--the sequence `%%' outputs one `%'; it does not consume an argument and it ignores any modifiers. NOTE: When using the integer format-control letters for values that are outside the range of the widest C integer type, `gawk' switches to the the `%g' format specifier. If `--lint' is provided on the command line (*note Options::), `gawk' warns about this. Other versions of `awk' may print invalid values or do something else entirely. (d.c.) 4.5.3 Modifiers for `printf' Formats ------------------------------------ A format specification can also include "modifiers" that can control how much of the item's value is printed, as well as how much space it gets. The modifiers come between the `%' and the format-control letter. We will use the bullet symbol "*" in the following examples to represent spaces in the output. Here are the possible modifiers, in the order in which they may appear: `N$' An integer constant followed by a `$' is a "positional specifier". Normally, format specifications are applied to arguments in the order given in the format string. With a positional specifier, the format specification is applied to a specific argument, instead of what would be the next argument in the list. Positional specifiers begin counting with one. Thus: printf "%s %s\n", "don't", "panic" printf "%2$s %1$s\n", "panic", "don't" prints the famous friendly message twice. At first glance, this feature doesn't seem to be of much use. It is in fact a `gawk' extension, intended for use in translating messages at runtime. *Note Printf Ordering::, which describes how and why to use positional specifiers. For now, we will not use them. `-' The minus sign, used before the width modifier (see later on in this table), says to left-justify the argument within its specified width. Normally, the argument is printed right-justified in the specified width. Thus: printf "%-4s", "foo" prints `foo*'. `SPACE' For numeric conversions, prefix positive values with a space and negative values with a minus sign. `+' The plus sign, used before the width modifier (see later on in this table), says to always supply a sign for numeric conversions, even if the data to format is positive. The `+' overrides the space modifier. `#' Use an "alternate form" for certain control letters. For `%o', supply a leading zero. For `%x' and `%X', supply a leading `0x' or `0X' for a nonzero result. For `%e', `%E', and `%f', the result always contains a decimal point. For `%g' and `%G', trailing zeros are not removed from the result. `0' A leading `0' (zero) acts as a flag that indicates that output should be padded with zeros instead of spaces. This applies even to non-numeric output formats. (d.c.) This flag only has an effect when the field width is wider than the value to print. `'' A single quote or apostrohe character is a POSIX extension to ISO C. It indicates that the integer part of a floating point value, or the entire part of an integer decimal value, should have a thousands-separator character in it. This only works in locales that support such characters. For example: $ cat thousands.awk Show source program -| BEGIN { printf "%'d\n", 1234567 } $ LC_ALL=C gawk -f thousands.awk Run it in "C" locale -| 1234567 $ LC_ALL=en_US.UTF-8 gawk -f thousands.awk Run in US English UTF locale -| 1,234,567 For more information about locales and internationalization issues, *FIXME: see xxxx*. NOTE: The `'' flag is a nice feature, but its use complicates things: it now becomes difficult to use it in command-line programs. For information on appropriate quoting tricks, *FIXME: see XXXX*. `WIDTH' This is a number specifying the desired minimum width of a field. Inserting any number between the `%' sign and the format-control character forces the field to expand to this width. The default way to do this is to pad with spaces on the left. For example: printf "%4s", "foo" prints `*foo'. The value of WIDTH is a minimum width, not a maximum. If the item value requires more than WIDTH characters, it can be as wide as necessary. Thus, the following: printf "%4s", "foobar" prints `foobar'. Preceding the WIDTH with a minus sign causes the output to be padded with spaces on the right, instead of on the left. `.PREC' A period followed by an integer constant specifies the precision to use when printing. The meaning of the precision varies by control letter: `%e', `%E', `%f' Number of digits to the right of the decimal point. `%g', `%G' Maximum number of significant digits. `%d', `%i', `%o', `%u', `%x', `%X' Minimum number of digits to print. `%s' Maximum number of characters from the string that should print. Thus, the following: printf "%.4s", "foobar" prints `foob'. The C library `printf''s dynamic WIDTH and PREC capability (for example, `"%*.*s"') is supported. Instead of supplying explicit WIDTH and/or PREC values in the format string, they are passed in the argument list. For example: w = 5 p = 3 s = "abcdefg" printf "%*.*s\n", w, p, s is exactly equivalent to: s = "abcdefg" printf "%5.3s\n", s Both programs output `**abc'. Earlier versions of `awk' did not support this capability. If you must use such a version, you may simulate this feature by using concatenation to build up the format string, like so: w = 5 p = 3 s = "abcdefg" printf "%" w "." p "s\n", s This is not particularly easy to read but it does work. C programmers may be used to supplying additional `l', `L', and `h' modifiers in `printf' format strings. These are not valid in `awk'. Most `awk' implementations silently ignore these modifiers. If `--lint' is provided on the command line (*note Options::), `gawk' warns about their use. If `--posix' is supplied, their use is a fatal error. 4.5.4 Examples Using `printf' ----------------------------- The following is a simple example of how to use `printf' to make an aligned table: awk '{ printf "%-10s %s\n", $1, $2 }' BBS-list This command prints the names of the bulletin boards (`$1') in the file `BBS-list' as a string of 10 characters that are left-justified. It also prints the phone numbers (`$2') next on the line. This produces an aligned two-column table of names and phone numbers, as shown here: $ awk '{ printf "%-10s %s\n", $1, $2 }' BBS-list -| aardvark 555-5553 -| alpo-net 555-3412 -| barfly 555-7685 -| bites 555-1675 -| camelot 555-0542 -| core 555-2912 -| fooey 555-1234 -| foot 555-6699 -| macfoo 555-6480 -| sdace 555-3430 -| sabafoo 555-2127 In this case, the phone numbers had to be printed as strings because the numbers are separated by a dash. Printing the phone numbers as numbers would have produced just the first three digits: `555'. This would have been pretty confusing. It wasn't necessary to specify a width for the phone numbers because they are last on their lines. They don't need to have spaces after them. The table could be made to look even nicer by adding headings to the tops of the columns. This is done using the `BEGIN' pattern (*note BEGIN/END::) so that the headers are only printed once, at the beginning of the `awk' program: awk 'BEGIN { print "Name Number" print "---- ------" } { printf "%-10s %s\n", $1, $2 }' BBS-list The above example mixed `print' and `printf' statements in the same program. Using just `printf' statements can produce the same results: awk 'BEGIN { printf "%-10s %s\n", "Name", "Number" printf "%-10s %s\n", "----", "------" } { printf "%-10s %s\n", $1, $2 }' BBS-list Printing each column heading with the same format specification used for the column elements ensures that the headings are aligned just like the columns. The fact that the same format specification is used three times can be emphasized by storing it in a variable, like this: awk 'BEGIN { format = "%-10s %s\n" printf format, "Name", "Number" printf format, "----", "------" } { printf format, $1, $2 }' BBS-list At this point, it would be a worthwhile exercise to use the `printf' statement to line up the headings and table data for the `inventory-shipped' example that was covered earlier in the minor node on the `print' statement (*note Print::). 4.6 Redirecting Output of `print' and `printf' ============================================== So far, the output from `print' and `printf' has gone to the standard output, usually the terminal. Both `print' and `printf' can also send their output to other places. This is called "redirection". A redirection appears after the `print' or `printf' statement. Redirections in `awk' are written just like redirections in shell commands, except that they are written inside the `awk' program. There are four forms of output redirection: output to a file, output appended to a file, output through a pipe to another command, and output to a coprocess. They are all shown for the `print' statement, but they work identically for `printf': `print ITEMS > OUTPUT-FILE' This type of redirection prints the items into the output file named OUTPUT-FILE. The file name OUTPUT-FILE can be any expression. Its value is changed to a string and then used as a file name (*note Expressions::). When this type of redirection is used, the OUTPUT-FILE is erased before the first output is written to it. Subsequent writes to the same OUTPUT-FILE do not erase OUTPUT-FILE, but append to it. (This is different from how you use redirections in shell scripts.) If OUTPUT-FILE does not exist, it is created. For example, here is how an `awk' program can write a list of BBS names to one file named `name-list', and a list of phone numbers to another file named `phone-list': $ awk '{ print $2 > "phone-list" > print $1 > "name-list" }' BBS-list $ cat phone-list -| 555-5553 -| 555-3412 ... $ cat name-list -| aardvark -| alpo-net ... Each output file contains one name or number per line. `print ITEMS >> OUTPUT-FILE' This type of redirection prints the items into the pre-existing output file named OUTPUT-FILE. The difference between this and the single-`>' redirection is that the old contents (if any) of OUTPUT-FILE are not erased. Instead, the `awk' output is appended to the file. If OUTPUT-FILE does not exist, then it is created. `print ITEMS | COMMAND' It is also possible to send output to another program through a pipe instead of into a file. This type of redirection opens a pipe to COMMAND, and writes the values of ITEMS through this pipe to another process created to execute COMMAND. The redirection argument COMMAND is actually an `awk' expression. Its value is converted to a string whose contents give the shell command to be run. For example, the following produces two files, one unsorted list of BBS names, and one list sorted in reverse alphabetical order: awk '{ print $1 > "names.unsorted" command = "sort -r > names.sorted" print $1 | command }' BBS-list The unsorted list is written with an ordinary redirection, while the sorted list is written by piping through the `sort' utility. The next example uses redirection to mail a message to the mailing list `bug-system'. This might be useful when trouble is encountered in an `awk' script run periodically for system maintenance: report = "mail bug-system" print "Awk script failed:", $0 | report m = ("at record number " FNR " of " FILENAME) print m | report close(report) The message is built using string concatenation and saved in the variable `m'. It's then sent down the pipeline to the `mail' program. (The parentheses group the items to concatenate--see *Note Concatenation::.) The `close' function is called here because it's a good idea to close the pipe as soon as all the intended output has been sent to it. *Note Close Files And Pipes::, for more information. This example also illustrates the use of a variable to represent a FILE or COMMAND--it is not necessary to always use a string constant. Using a variable is generally a good idea, because `awk' requires that the string value be spelled identically every time. `print ITEMS |& COMMAND' This type of redirection prints the items to the input of COMMAND. The difference between this and the single-`|' redirection is that the output from COMMAND can be read with `getline'. Thus COMMAND is a "coprocess", which works together with, but subsidiary to, the `awk' program. This feature is a `gawk' extension, and is not available in POSIX `awk'. *Note Two-way I/O::, for a more complete discussion. Redirecting output using `>', `>>', `|', or `|&' asks the system to open a file, pipe, or coprocess only if the particular FILE or COMMAND you specify has not already been written to by your program or if it has been closed since it was last written to. It is a common error to use `>' redirection for the first `print' to a file, and then to use `>>' for subsequent output: # clear the file print "Don't panic" > "guide.txt" ... # append print "Avoid improbability generators" >> "guide.txt" This is indeed how redirections must be used from the shell. But in `awk', it isn't necessary. In this kind of case, a program should use `>' for all the `print' statements, since the output file is only opened once. Many `awk' implementations limit the number of pipelines that an `awk' program may have open to just one! In `gawk', there is no such limit. `gawk' allows a program to open as many pipelines as the underlying operating system permits. Advanced Notes: Piping into `sh' -------------------------------- A particularly powerful way to use redirection is to build command lines and pipe them into the shell, `sh'. For example, suppose you have a list of files brought over from a system where all the file names are stored in uppercase, and you wish to rename them to have names in all lowercase. The following program is both simple and efficient: { printf("mv %s %s\n", $0, tolower($0)) | "sh" } END { close("sh") } The `tolower' function returns its argument string with all uppercase characters converted to lowercase (*note String Functions::). The program builds up a list of command lines, using the `mv' utility to rename the files. It then sends the list to the shell for execution. 4.7 Special File Names in `gawk' ================================ `gawk' provides a number of special file names that it interprets internally. These file names provide access to standard file descriptors, process-related information, and TCP/IP networking. 4.7.1 Special Files for Standard Descriptors -------------------------------------------- Running programs conventionally have three input and output streams already available to them for reading and writing. These are known as the "standard input", "standard output", and "standard error output". These streams are, by default, connected to your terminal, but they are often redirected with the shell, via the `<', `<<', `>', `>>', `>&', and `|' operators. Standard error is typically used for writing error messages; the reason there are two separate streams, standard output and standard error, is so that they can be redirected separately. In other implementations of `awk', the only way to write an error message to standard error in an `awk' program is as follows: print "Serious error detected!" | "cat 1>&2" This works by opening a pipeline to a shell command that can access the standard error stream that it inherits from the `awk' process. This is far from elegant, and it is also inefficient, because it requires a separate process. So people writing `awk' programs often don't do this. Instead, they send the error messages to the terminal, like this: print "Serious error detected!" > "/dev/tty" This usually has the same effect but not always: although the standard error stream is usually the terminal, it can be redirected; when that happens, writing to the terminal is not correct. In fact, if `awk' is run from a background job, it may not have a terminal at all. Then opening `/dev/tty' fails. `gawk' provides special file names for accessing the three standard streams, as well as any other inherited open files. If the file name matches one of these special names when `gawk' redirects input or output, then it directly uses the stream that the file name stands for. These special file names work for all operating systems that `gawk' has been ported to, not just those that are POSIX-compliant: `/dev/stdin' The standard input (file descriptor 0). `/dev/stdout' The standard output (file descriptor 1). `/dev/stderr' The standard error output (file descriptor 2). `/dev/fd/N' The file associated with file descriptor N. Such a file must be opened by the program initiating the `awk' execution (typically the shell). Unless special pains are taken in the shell from which `gawk' is invoked, only descriptors 0, 1, and 2 are available. The file names `/dev/stdin', `/dev/stdout', and `/dev/stderr' are aliases for `/dev/fd/0', `/dev/fd/1', and `/dev/fd/2', respectively. However, they are more self-explanatory. The proper way to write an error message in a `gawk' program is to use `/dev/stderr', like this: print "Serious error detected!" > "/dev/stderr" Note the use of quotes around the file name. Like any other redirection, the value must be a string. It is a common error to omit the quotes, which leads to confusing results. 4.7.2 Special Files for Process-Related Information --------------------------------------------------- `gawk' also provides special file names that give access to information about the running `gawk' process. Each of these "files" provides a single record of information. To read them more than once, they must first be closed with the `close' function (*note Close Files And Pipes::). The file names are: `/dev/pid' Reading this file returns the process ID of the current process, in decimal form, terminated with a newline. `/dev/ppid' Reading this file returns the parent process ID of the current process, in decimal form, terminated with a newline. `/dev/pgrpid' Reading this file returns the process group ID of the current process, in decimal form, terminated with a newline. `/dev/user' Reading this file returns a single record terminated with a newline. The fields are separated with spaces. The fields represent the following information: `$1' The return value of the `getuid' system call (the real user ID number). `$2' The return value of the `geteuid' system call (the effective user ID number). `$3' The return value of the `getgid' system call (the real group ID number). `$4' The return value of the `getegid' system call (the effective group ID number). If there are any additional fields, they are the group IDs returned by the `getgroups' system call. (Multiple groups may not be supported on all systems.) These special file names may be used on the command line as data files, as well as for I/O redirections within an `awk' program. They may not be used as source files with the `-f' option. NOTE: The special files that provide process-related information are now considered obsolete and will disappear entirely in the next release of `gawk'. `gawk' prints a warning message every time you use one of these files. To obtain process-related information, use the `PROCINFO' array. *Note Auto-set::. 4.7.3 Special Files for Network Communications ---------------------------------------------- Starting with version 3.1 of `gawk', `awk' programs can open a two-way TCP/IP connection, acting as either a client or a server. This is done using a special file name of the form: `/inet/PROTOCOL/LOCAL-PORT/REMOTE-HOST/REMOTE-PORT' The PROTOCOL is one of `tcp', `udp', or `raw', and the other fields represent the other essential pieces of information for making a networking connection. These file names are used with the `|&' operator for communicating with a coprocess (*note Two-way I/O::). This is an advanced feature, mentioned here only for completeness. Full discussion is delayed until *Note TCP/IP Networking::. 4.7.4 Special File Name Caveats ------------------------------- Here is a list of things to bear in mind when using the special file names that `gawk' provides: * Recognition of these special file names is disabled if `gawk' is in compatibility mode (*note Options::). * The special files that provide process-related information are now considered obsolete and will disappear entirely in the next release of `gawk'. `gawk' prints a warning message every time you use one of these files. To obtain process-related information, use the `PROCINFO' array. *Note Built-in Variables::. * Starting with version 3.1, `gawk' _always_ interprets these special file names.(1) For example, using `/dev/fd/4' for output actually writes on file descriptor 4, and not on a new file descriptor that is `dup''ed from file descriptor 4. Most of the time this does not matter; however, it is important to _not_ close any of the files related to file descriptors 0, 1, and 2. Doing so results in unpredictable behavior. ---------- Footnotes ---------- (1) Older versions of `gawk' would interpret these names internally only if the system did not actually have a `/dev/fd' directory or any of the other special files listed earlier. Usually this didn't make a difference, but sometimes it did; thus, it was decided to make `gawk''s behavior consistent on all systems and to have it always interpret the special file names itself. 4.8 Closing Input and Output Redirections ========================================= If the same file name or the same shell command is used with `getline' more than once during the execution of an `awk' program (*note Getline::), the file is opened (or the command is executed) the first time only. At that time, the first record of input is read from that file or command. The next time the same file or command is used with `getline', another record is read from it, and so on. Similarly, when a file or pipe is opened for output, the file name or command associated with it is remembered by `awk', and subsequent writes to the same file or command are appended to the previous writes. The file or pipe stays open until `awk' exits. This implies that special steps are necessary in order to read the same file again from the beginning, or to rerun a shell command (rather than reading more output from the same command). The `close' function makes these things possible: close(FILENAME) or: close(COMMAND) The argument FILENAME or COMMAND can be any expression. Its value must _exactly_ match the string that was used to open the file or start the command (spaces and other "irrelevant" characters included). For example, if you open a pipe with this: "sort -r names" | getline foo then you must close it with this: close("sort -r names") Once this function call is executed, the next `getline' from that file or command, or the next `print' or `printf' to that file or command, reopens the file or reruns the command. Because the expression that you use to close a file or pipeline must exactly match the expression used to open the file or run the command, it is good practice to use a variable to store the file name or command. The previous example becomes the following: sortcom = "sort -r names" sortcom | getline foo ... close(sortcom) This helps avoid hard-to-find typographical errors in your `awk' programs. Here are some of the reasons for closing an output file: * To write a file and read it back later on in the same `awk' program. Close the file after writing it, then begin reading it with `getline'. * To write numerous files, successively, in the same `awk' program. If the files aren't closed, eventually `awk' may exceed a system limit on the number of open files in one process. It is best to close each one when the program has finished writing it. * To make a command finish. When output is redirected through a pipe, the command reading the pipe normally continues to try to read input as long as the pipe is open. Often this means the command cannot really do its work until the pipe is closed. For example, if output is redirected to the `mail' program, the message is not actually sent until the pipe is closed. * To run the same program a second time, with the same arguments. This is not the same thing as giving more input to the first run! For example, suppose a program pipes output to the `mail' program. If it outputs several lines redirected to this pipe without closing it, they make a single message of several lines. By contrast, if the program closes the pipe after each line of output, then each line makes a separate message. If you use more files than the system allows you to have open, `gawk' attempts to multiplex the available open files among your data files. `gawk''s ability to do this depends upon the facilities of your operating system, so it may not always work. It is therefore both good practice and good portability advice to always use `close' on your files when you are done with them. In fact, if you are using a lot of pipes, it is essential that you close commands when done. For example, consider something like this: { ... command = ("grep " $1 " /some/file | my_prog -q " $3) while ((command | getline) > 0) { PROCESS OUTPUT OF command } # need close(command) here } This example creates a new pipeline based on data in _each_ record. Without the call to `close' indicated in the comment, `awk' creates child processes to run the commands, until it eventually runs out of file descriptors for more pipelines. Even though each command has finished (as indicated by the end-of-file return status from `getline'), the child process is not terminated;(1) more importantly, the file descriptor for the pipe is not closed and released until `close' is called or `awk' exits. `close' will silently do nothing if given an argument that does not represent a file, pipe or coprocess that was opened with a redirection. Note also that `close(FILENAME)' has no "magic" effects on the implicit loop that reads through the files named on the command line. It is, more likely, a close of a file that was never opened, so `awk' silently does nothing. When using the `|&' operator to communicate with a coprocess, it is occasionally useful to be able to close one end of the two-way pipe without closing the other. This is done by supplying a second argument to `close'. As in any other call to `close', the first argument is the name of the command or special file used to start the coprocess. The second argument should be a string, with either of the values `"to"' or `"from"'. Case does not matter. As this is an advanced feature, a more complete discussion is delayed until *Note Two-way I/O::, which discusses it in more detail and gives an example. Advanced Notes: Using `close''s Return Value -------------------------------------------- In many versions of Unix `awk', the `close' function is actually a statement. It is a syntax error to try and use the return value from `close': (d.c.) command = "..." command | getline info retval = close(command) # syntax error in most Unix awks `gawk' treats `close' as a function. The return value is -1 if the argument names something that was never opened with a redirection, or if there is a system problem closing the file or process. In these cases, `gawk' sets the built-in variable `ERRNO' to a string describing the problem. In `gawk', when closing a pipe or coprocess, the return value is the exit status of the command.(2) Otherwise, it is the return value from the system's `close' or `fclose' C functions when closing input or output files, respectively. This value is zero if the close succeeds, or -1 if it fails. The POSIX standard is very vague; it says that `close' returns zero on success and non-zero otherwise. In general, different implementations vary in what they report when closing pipes; thus the return value cannot be used portably. (d.c.) ---------- Footnotes ---------- (1) The technical terminology is rather morbid. The finished child is called a "zombie," and cleaning up after it is referred to as "reaping." (2) This is a full 16-bit value as returned by the `wait' system call. See the system manual pages for information on how to decode this value. 5 Expressions ************* Expressions are the basic building blocks of `awk' patterns and actions. An expression evaluates to a value that you can print, test, or pass to a function. Additionally, an expression can assign a new value to a variable or a field by using an assignment operator. An expression can serve as a pattern or action statement on its own. Most other kinds of statements contain one or more expressions that specify the data on which to operate. As in other languages, expressions in `awk' include variables, array references, constants, and function calls, as well as combinations of these with various operators. 5.1 Constant Expressions ======================== The simplest type of expression is the "constant", which always has the same value. There are three types of constants: numeric, string, and regular expression. Each is used in the appropriate context when you need a data value that isn't going to change. Numeric constants can have different forms, but are stored identically internally. 5.1.1 Numeric and String Constants ---------------------------------- A "numeric constant" stands for a number. This number can be an integer, a decimal fraction, or a number in scientific (exponential) notation.(1) Here are some examples of numeric constants that all have the same value: 105 1.05e+2 1050e-1 A string constant consists of a sequence of characters enclosed in double-quotation marks. For example: "parrot" represents the string whose contents are `parrot'. Strings in `gawk' can be of any length, and they can contain any of the possible eight-bit ASCII characters including ASCII NUL (character code zero). Other `awk' implementations may have difficulty with some character codes. ---------- Footnotes ---------- (1) The internal representation of all numbers, including integers, uses double-precision floating-point numbers. On most modern systems, these are in IEEE 754 standard format. 5.1.2 Octal and Hexadecimal Numbers ----------------------------------- In `awk', all numbers are in decimal; i.e., base 10. Many other programming languages allow you to specify numbers in other bases, often octal (base 8) and hexadecimal (base 16). In octal, the numbers go 0, 1, 2, 3, 4, 5, 6, 7, 10, 11, 12, etc. Just as `11', in decimal, is 1 times 10 plus 1, so `11', in octal, is 1 times 8, plus 1. This equals 9 in decimal. In hexadecimal, there are 16 digits. Since the everyday decimal number system only has ten digits (`0'-`9'), the letters `a' through `f' are used to represent the rest. (Case in the letters is usually irrelevant; hexadecimal `a' and `A' have the same value.) Thus, `11', in hexadecimal, is 1 times 16 plus 1, which equals 17 in decimal. Just by looking at plain `11', you can't tell what base it's in. So, in C, C++, and other languages derived from C, there is a special notation to help signify the base. Octal numbers start with a leading `0', and hexadecimal numbers start with a leading `0x' or `0X': `11' Decimal value 11. `011' Octal 11, decimal value 9. `0x11' Hexadecimal 11, decimal value 17. This example shows the difference: $ gawk 'BEGIN { printf "%d, %d, %d\n", 011, 11, 0x11 }' -| 9, 11, 17 Being able to use octal and hexadecimal constants in your programs is most useful when working with data that cannot be represented conveniently as characters or as regular numbers, such as binary data of various sorts. `gawk' allows the use of octal and hexadecimal constants in your program text. However, such numbers in the input data are not treated differently; doing so by default would break old programs. (If you really need to do this, use the `--non-decimal-data' command-line option; *note Nondecimal Data::.) If you have octal or hexadecimal data, you can use the `strtonum' function (*note String Functions::) to convert the data into a number. Most of the time, you will want to use octal or hexadecimal constants when working with the built-in bit manipulation functions; see *Note Bitwise Functions::, for more information. Unlike some early C implementations, `8' and `9' are not valid in octal constants; e.g., `gawk' treats `018' as decimal 18: $ gawk 'BEGIN { print "021 is", 021 ; print 018 }' -| 021 is 17 -| 18 Octal and hexadecimal source code constants are a `gawk' extension. If `gawk' is in compatibility mode (*note Options::), they are not available. Advanced Notes: A Constant's Base Does Not Affect Its Value ----------------------------------------------------------- Once a numeric constant has been converted internally into a number, `gawk' no longer remembers what the original form of the constant was; the internal value is always used. This has particular consequences for conversion of numbers to strings: $ gawk 'BEGIN { printf "0x11 is <%s>\n", 0x11 }' -| 0x11 is <17> 5.1.3 Regular Expression Constants ---------------------------------- A regexp constant is a regular expression description enclosed in slashes, such as `/^beginning and end$/'. Most regexps used in `awk' programs are constant, but the `~' and `!~' matching operators can also match computed or "dynamic" regexps (which are just ordinary strings or variables that contain a regexp). 5.2 Using Regular Expression Constants ====================================== When used on the righthand side of the `~' or `!~' operators, a regexp constant merely stands for the regexp that is to be matched. However, regexp constants (such as `/foo/') may be used like simple expressions. When a regexp constant appears by itself, it has the same meaning as if it appeared in a pattern, i.e., `($0 ~ /foo/)' (d.c.) *Note Expression Patterns::. This means that the following two code segments: if ($0 ~ /barfly/ || $0 ~ /camelot/) print "found" and: if (/barfly/ || /camelot/) print "found" are exactly equivalent. One rather bizarre consequence of this rule is that the following Boolean expression is valid, but does not do what the user probably intended: # note that /foo/ is on the left of the ~ if (/foo/ ~ $1) print "found foo" This code is "obviously" testing `$1' for a match against the regexp `/foo/'. But in fact, the expression `/foo/ ~ $1' actually means `($0 ~ /foo/) ~ $1'. In other words, first match the input record against the regexp `/foo/'. The result is either zero or one, depending upon the success or failure of the match. That result is then matched against the first field in the record. Because it is unlikely that you would ever really want to make this kind of test, `gawk' issues a warning when it sees this construct in a program. Another consequence of this rule is that the assignment statement: matches = /foo/ assigns either zero or one to the variable `matches', depending upon the contents of the current input record. This feature of the language has never been well documented until the POSIX specification. Constant regular expressions are also used as the first argument for the `gensub', `sub', and `gsub' functions, and as the second argument of the `match' function (*note String Functions::). Modern implementations of `awk', including `gawk', allow the third argument of `split' to be a regexp constant, but some older implementations do not. (d.c.) This can lead to confusion when attempting to use regexp constants as arguments to user-defined functions (*note User-defined::). For example: function mysub(pat, repl, str, global) { if (global) gsub(pat, repl, str) else sub(pat, repl, str) return str } { ... text = "hi! hi yourself!" mysub(/hi/, "howdy", text, 1) ... } In this example, the programmer wants to pass a regexp constant to the user-defined function `mysub', which in turn passes it on to either `sub' or `gsub'. However, what really happens is that the `pat' parameter is either one or zero, depending upon whether or not `$0' matches `/hi/'. `gawk' issues a warning when it sees a regexp constant used as a parameter to a user-defined function, since passing a truth value in this way is probably not what was intended. 5.3 Variables ============= Variables are ways of storing values at one point in your program for use later in another part of your program. They can be manipulated entirely within the program text, and they can also be assigned values on the `awk' command line. 5.3.1 Using Variables in a Program ---------------------------------- Variables let you give names to values and refer to them later. Variables have already been used in many of the examples. The name of a variable must be a sequence of letters, digits, or underscores, and it may not begin with a digit. Case is significant in variable names; `a' and `A' are distinct variables. A variable name is a valid expression by itself; it represents the variable's current value. Variables are given new values with "assignment operators", "increment operators", and "decrement operators". *Note Assignment Ops::. A few variables have special built-in meanings, such as `FS' (the field separator), and `NF' (the number of fields in the current input record). *Note Built-in Variables::, for a list of the built-in variables. These built-in variables can be used and assigned just like all other variables, but their values are also used or changed automatically by `awk'. All built-in variables' names are entirely uppercase. Variables in `awk' can be assigned either numeric or string values. The kind of value a variable holds can change over the life of a program. By default, variables are initialized to the empty string, which is zero if converted to a number. There is no need to "initialize" each variable explicitly in `awk', which is what you would do in C and in most other traditional languages. 5.3.2 Assigning Variables on the Command Line --------------------------------------------- Any `awk' variable can be set by including a "variable assignment" among the arguments on the command line when `awk' is invoked (*note Other Arguments::). Such an assignment has the following form: VARIABLE=TEXT With it, a variable is set either at the beginning of the `awk' run or in between input files. When the assignment is preceded with the `-v' option, as in the following: -v VARIABLE=TEXT the variable is set at the very beginning, even before the `BEGIN' rules are run. The `-v' option and its assignment must precede all the file name arguments, as well as the program text. (*Note Options::, for more information about the `-v' option.) Otherwise, the variable assignment is performed at a time determined by its position among the input file arguments--after the processing of the preceding input file argument. For example: awk '{ print $n }' n=4 inventory-shipped n=2 BBS-list prints the value of field number `n' for all input records. Before the first file is read, the command line sets the variable `n' equal to four. This causes the fourth field to be printed in lines from the file `inventory-shipped'. After the first file has finished, but before the second file is started, `n' is set to two, so that the second field is printed in lines from `BBS-list': $ awk '{ print $n }' n=4 inventory-shipped n=2 BBS-list -| 15 -| 24 ... -| 555-5553 -| 555-3412 ... Command-line arguments are made available for explicit examination by the `awk' program in the `ARGV' array (*note ARGC and ARGV::). `awk' processes the values of command-line assignments for escape sequences (*note Escape Sequences::). (d.c.) 5.4 Conversion of Strings and Numbers ===================================== Strings are converted to numbers and numbers are converted to strings, if the context of the `awk' program demands it. For example, if the value of either `foo' or `bar' in the expression `foo + bar' happens to be a string, it is converted to a number before the addition is performed. If numeric values appear in string concatenation, they are converted to strings. Consider the following: two = 2; three = 3 print (two three) + 4 This prints the (numeric) value 27. The numeric values of the variables `two' and `three' are converted to strings and concatenated together. The resulting string is converted back to the number 23, to which 4 is then added. If, for some reason, you need to force a number to be converted to a string, concatenate the empty string, `""', with that number. To force a string to be converted to a number, add zero to that string. A string is converted to a number by interpreting any numeric prefix of the string as numerals: `"2.5"' converts to 2.5, `"1e3"' converts to 1000, and `"25fix"' has a numeric value of 25. Strings that can't be interpreted as valid numbers convert to zero. The exact manner in which numbers are converted into strings is controlled by the `awk' built-in variable `CONVFMT' (*note Built-in Variables::). Numbers are converted using the `sprintf' function with `CONVFMT' as the format specifier (*note String Functions::). `CONVFMT''s default value is `"%.6g"', which prints a value with at least six significant digits. For some applications, you might want to change it to specify more precision. On most modern machines, 17 digits is enough to capture a floating-point number's value exactly, most of the time.(1) Strange results can occur if you set `CONVFMT' to a string that doesn't tell `sprintf' how to format floating-point numbers in a useful way. For example, if you forget the `%' in the format, `awk' converts all numbers to the same constant string. As a special case, if a number is an integer, then the result of converting it to a string is _always_ an integer, no matter what the value of `CONVFMT' may be. Given the following code fragment: CONVFMT = "%2.2f" a = 12 b = a "" `b' has the value `"12"', not `"12.00"'. (d.c.) Prior to the POSIX standard, `awk' used the value of `OFMT' for converting numbers to strings. `OFMT' specifies the output format to use when printing numbers with `print'. `CONVFMT' was introduced in order to separate the semantics of conversion from the semantics of printing. Both `CONVFMT' and `OFMT' have the same default value: `"%.6g"'. In the vast majority of cases, old `awk' programs do not change their behavior. However, these semantics for `OFMT' are something to keep in mind if you must port your new style program to older implementations of `awk'. We recommend that instead of changing your programs, just port `gawk' itself. *Note Print::, for more information on the `print' statement. Finally, once again, where you are can matter when it comes to converting between numbers and strings. In *Note Locales::, we mentioned that the local character set and language (the locale) can affect how `gawk' matches characters. The locale also affects numeric formats. In particular, for `awk' programs, it affects the decimal point character. The `"C"' locale, and most English-language locales, use the period character (`.') as the decimal point. However, many (if not most) European and non-English locales use the comma (`,') as the decimal point character. The POSIX standard says that `awk' always uses the period as the decimal point when reading the `awk' program source code, and for command-line variable assignments (*note Other Arguments::). However, when interpreting input data, for `print' and `printf' output, and for number to string conversion, the local decimal point character is used. As of version 3.1.3, `gawk' fully complies with this aspect of the standard. Here are some examples indicating the difference in behavior, on a GNU/Linux system: $ gawk 'BEGIN { printf "%g\n", 3.1415927 }' -| 3.14159 $ LC_ALL=en_DK gawk 'BEGIN { printf "%g\n", 3.1415927 }' -| 3,14159 $ echo 4,321 | gawk '{ print $1 + 1 }' -| 5 $ echo 4,321 | LC_ALL=en_DK gawk '{ print $1 + 1 }' -| 5,321 The `en_DK' locale is for English in Denmark, where the comma acts as the decimal point separator. In the normal `"C"' locale, `gawk' treats `4,321' as `4', while in the Danish locale, it's treated as the full number, `4.321'. ---------- Footnotes ---------- (1) Pathological cases can require up to 752 digits (!), but we doubt that you need to worry about this. 5.5 Arithmetic Operators ======================== The `awk' language uses the common arithmetic operators when evaluating expressions. All of these arithmetic operators follow normal precedence rules and work as you would expect them to. The following example uses a file named `grades', which contains a list of student names as well as three test scores per student (it's a small class): Pat 100 97 58 Sandy 84 72 93 Chris 72 92 89 This programs takes the file `grades' and prints the average of the scores: $ awk '{ sum = $2 + $3 + $4 ; avg = sum / 3 > print $1, avg }' grades -| Pat 85 -| Sandy 83 -| Chris 84.3333 The following list provides the arithmetic operators in `awk', in order from the highest precedence to the lowest: `- X' Negation. `+ X' Unary plus; the expression is converted to a number. `X ^ Y' `X ** Y' Exponentiation; X raised to the Y power. `2 ^ 3' has the value eight; the character sequence `**' is equivalent to `^'. `X * Y' Multiplication. `X / Y' Division; because all numbers in `awk' are floating-point numbers, the result is _not_ rounded to an integer--`3 / 4' has the value 0.75. (It is a common mistake, especially for C programmers, to forget that _all_ numbers in `awk' are floating-point, and that division of integer-looking constants produces a real number, not an integer.) `X % Y' Remainder; further discussion is provided in the text, just after this list. `X + Y' Addition. `X - Y' Subtraction. Unary plus and minus have the same precedence, the multiplication operators all have the same precedence, and addition and subtraction have the same precedence. When computing the remainder of `X % Y', the quotient is rounded toward zero to an integer and multiplied by Y. This result is subtracted from X; this operation is sometimes known as "trunc-mod." The following relation always holds: b * int(a / b) + (a % b) == a One possibly undesirable effect of this definition of remainder is that `X % Y' is negative if X is negative. Thus: -17 % 8 = -1 In other `awk' implementations, the signedness of the remainder may be machine-dependent. NOTE: The POSIX standard only specifies the use of `^' for exponentiation. For maximum portability, do not use the `**' operator. 5.6 String Concatenation ======================== It seemed like a good idea at the time. Brian Kernighan There is only one string operation: concatenation. It does not have a specific operator to represent it. Instead, concatenation is performed by writing expressions next to one another, with no operator. For example: $ awk '{ print "Field number one: " $1 }' BBS-list -| Field number one: aardvark -| Field number one: alpo-net ... Without the space in the string constant after the `:', the line runs together. For example: $ awk '{ print "Field number one:" $1 }' BBS-list -| Field number one:aardvark -| Field number one:alpo-net ... Because string concatenation does not have an explicit operator, it is often necessary to insure that it happens at the right time by using parentheses to enclose the items to concatenate. For example, the following code fragment does not concatenate `file' and `name' as you might expect: file = "file" name = "name" print "something meaningful" > file name It is necessary to use the following: print "something meaningful" > (file name) Parentheses should be used around concatenation in all but the most common contexts, such as on the righthand side of `='. Be careful about the kinds of expressions used in string concatenation. In particular, the order of evaluation of expressions used for concatenation is undefined in the `awk' language. Consider this example: BEGIN { a = "don't" print (a " " (a = "panic")) } It is not defined whether the assignment to `a' happens before or after the value of `a' is retrieved for producing the concatenated value. The result could be either `don't panic', or `panic panic'. The precedence of concatenation, when mixed with other operators, is often counter-intuitive. Consider this example: $ awk 'BEGIN { print -12 " " -24 }' -| -12-24 This "obviously" is concatenating -12, a space, and -24. But where did the space disappear to? The answer lies in the combination of operator precedences and `awk''s automatic conversion rules. To get the desired result, write the program in the following manner: $ awk 'BEGIN { print -12 " " (-24) }' -| -12 -24 This forces `awk' to treat the `-' on the `-24' as unary. Otherwise, it's parsed as follows: -12 (`" "' - 24) => -12 (0 - 24) => -12 (-24) => -12-24 As mentioned earlier, when doing concatenation, _parenthesize_. Otherwise, you're never quite sure what you'll get. 5.7 Assignment Expressions ========================== An "assignment" is an expression that stores a (usually different) value into a variable. For example, let's assign the value one to the variable `z': z = 1 After this expression is executed, the variable `z' has the value one. Whatever old value `z' had before the assignment is forgotten. Assignments can also store string values. For example, the following stores the value `"this food is good"' in the variable `message': thing = "food" predicate = "good" message = "this " thing " is " predicate This also illustrates string concatenation. The `=' sign is called an "assignment operator". It is the simplest assignment operator because the value of the righthand operand is stored unchanged. Most operators (addition, concatenation, and so on) have no effect except to compute a value. If the value isn't used, there's no reason to use the operator. An assignment operator is different; it does produce a value, but even if you ignore it, the assignment still makes itself felt through the alteration of the variable. We call this a "side effect". The lefthand operand of an assignment need not be a variable (*note Variables::); it can also be a field (*note Changing Fields::) or an array element (*note Arrays::). These are all called "lvalues", which means they can appear on the lefthand side of an assignment operator. The righthand operand may be any expression; it produces the new value that the assignment stores in the specified variable, field, or array element. (Such values are called "rvalues".) It is important to note that variables do _not_ have permanent types. A variable's type is simply the type of whatever value it happens to hold at the moment. In the following program fragment, the variable `foo' has a numeric value at first, and a string value later on: foo = 1 print foo foo = "bar" print foo When the second assignment gives `foo' a string value, the fact that it previously had a numeric value is forgotten. String values that do not begin with a digit have a numeric value of zero. After executing the following code, the value of `foo' is five: foo = "a string" foo = foo + 5 NOTE: Using a variable as a number and then later as a string can be confusing and is poor programming style. The previous two examples illustrate how `awk' works, _not_ how you should write your programs! An assignment is an expression, so it has a value--the same value that is assigned. Thus, `z = 1' is an expression with the value one. One consequence of this is that you can write multiple assignments together, such as: x = y = z = 5 This example stores the value five in all three variables (`x', `y', and `z'). It does so because the value of `z = 5', which is five, is stored into `y' and then the value of `y = z = 5', which is five, is stored into `x'. Assignments may be used anywhere an expression is called for. For example, it is valid to write `x != (y = 1)' to set `y' to one, and then test whether `x' equals one. But this style tends to make programs hard to read; such nesting of assignments should be avoided, except perhaps in a one-shot program. Aside from `=', there are several other assignment operators that do arithmetic with the old value of the variable. For example, the operator `+=' computes a new value by adding the righthand value to the old value of the variable. Thus, the following assignment adds five to the value of `foo': foo += 5 This is equivalent to the following: foo = foo + 5 Use whichever makes the meaning of your program clearer. There are situations where using `+=' (or any assignment operator) is _not_ the same as simply repeating the lefthand operand in the righthand expression. For example: # Thanks to Pat Rankin for this example BEGIN { foo[rand()] += 5 for (x in foo) print x, foo[x] bar[rand()] = bar[rand()] + 5 for (x in bar) print x, bar[x] } The indices of `bar' are practically guaranteed to be different, because `rand' returns different values each time it is called. (Arrays and the `rand' function haven't been covered yet. *Note Arrays::, and see *Note Numeric Functions::, for more information). This example illustrates an important fact about assignment operators: the lefthand expression is only evaluated _once_. It is up to the implementation as to which expression is evaluated first, the lefthand or the righthand. Consider this example: i = 1 a[i += 2] = i + 1 The value of `a[3]' could be either two or four. *Note table-assign-ops:: lists the arithmetic assignment operators. In each case, the righthand operand is an expression whose value is converted to a number. Operator Effect -------------------------------------------------------------------------- LVALUE `+=' INCREMENT Adds INCREMENT to the value of LVALUE. LVALUE `-=' DECREMENT Subtracts DECREMENT from the value of LVALUE. LVALUE `*=' Multiplies the value of LVALUE by COEFFICIENT. COEFFICIENT LVALUE `/=' DIVISOR Divides the value of LVALUE by DIVISOR. LVALUE `%=' MODULUS Sets LVALUE to its remainder by MODULUS. LVALUE `^=' POWER LVALUE `**=' POWER Raises LVALUE to the power POWER. Table 5.1: Arithmetic Assignment Operators NOTE: Only the `^=' operator is specified by POSIX. For maximum portability, do not use the `**=' operator. Advanced Notes: Syntactic Ambiguities Between `/=' and Regular Expressions -------------------------------------------------------------------------- There is a syntactic ambiguity between the `/=' assignment operator and regexp constants whose first character is an `='. (d.c.) This is most notable in commercial `awk' versions. For example: $ awk /==/ /dev/null error--> awk: syntax error at source line 1 error--> context is error--> >>> /= <<< error--> awk: bailing out at source line 1 A workaround is: awk '/[=]=/' /dev/null `gawk' does not have this problem, nor do the other freely available versions described in *Note Other Versions::. 5.8 Increment and Decrement Operators ===================================== "Increment" and "decrement operators" increase or decrease the value of a variable by one. An assignment operator can do the same thing, so the increment operators add no power to the `awk' language; however, they are convenient abbreviations for very common operations. The operator used for adding one is written `++'. It can be used to increment a variable either before or after taking its value. To pre-increment a variable `v', write `++v'. This adds one to the value of `v'--that new value is also the value of the expression. (The assignment expression `v += 1' is completely equivalent.) Writing the `++' after the variable specifies post-increment. This increments the variable value just the same; the difference is that the value of the increment expression itself is the variable's _old_ value. Thus, if `foo' has the value four, then the expression `foo++' has the value four, but it changes the value of `foo' to five. In other words, the operator returns the old value of the variable, but with the side effect of incrementing it. The post-increment `foo++' is nearly the same as writing `(foo += 1) - 1'. It is not perfectly equivalent because all numbers in `awk' are floating-point--in floating-point, `foo + 1 - 1' does not necessarily equal `foo'. But the difference is minute as long as you stick to numbers that are fairly small (less than 10e12). Fields and array elements are incremented just like variables. (Use `$(i++)' when you want to do a field reference and a variable increment at the same time. The parentheses are necessary because of the precedence of the field reference operator `$'.) The decrement operator `--' works just like `++', except that it subtracts one instead of adding it. As with `++', it can be used before the lvalue to pre-decrement or after it to post-decrement. Following is a summary of increment and decrement expressions: `++LVALUE' This expression increments LVALUE, and the new value becomes the value of the expression. `LVALUE++' This expression increments LVALUE, but the value of the expression is the _old_ value of LVALUE. `--LVALUE' This expression is like `++LVALUE', but instead of adding, it subtracts. It decrements LVALUE and delivers the value that is the result. `LVALUE--' This expression is like `LVALUE++', but instead of adding, it subtracts. It decrements LVALUE. The value of the expression is the _old_ value of LVALUE. Advanced Notes: Operator Evaluation Order ----------------------------------------- Doctor, doctor! It hurts when I do this! So don't do that! Groucho Marx What happens for something like the following? b = 6 print b += b++ Or something even stranger? b = 6 b += ++b + b++ print b In other words, when do the various side effects prescribed by the postfix operators (`b++') take effect? When side effects happen is "implementation defined". In other words, it is up to the particular version of `awk'. The result for the first example may be 12 or 13, and for the second, it may be 22 or 23. In short, doing things like this is not recommended and definitely not anything that you can rely upon for portability. You should avoid such things in your own programs. 5.9 True and False in `awk' =========================== Many programming languages have a special representation for the concepts of "true" and "false." Such languages usually use the special constants `true' and `false', or perhaps their uppercase equivalents. However, `awk' is different. It borrows a very simple concept of true and false from C. In `awk', any nonzero numeric value _or_ any nonempty string value is true. Any other value (zero or the null string `""') is false. The following program prints `A strange truth value' three times: BEGIN { if (3.1415927) print "A strange truth value" if ("Four Score And Seven Years Ago") print "A strange truth value" if (j = 57) print "A strange truth value" } There is a surprising consequence of the "nonzero or non-null" rule: the string constant `"0"' is actually true, because it is non-null. (d.c.) 5.10 Variable Typing and Comparison Expressions =============================================== The Guide is definitive. Reality is frequently inaccurate. The Hitchhiker's Guide to the Galaxy Unlike other programming languages, `awk' variables do not have a fixed type. Instead, they can be either a number or a string, depending upon the value that is assigned to them. The 1992 POSIX standard introduced the concept of a "numeric string", which is simply a string that looks like a number--for example, `" +2"'. This concept is used for determining the type of a variable. The type of the variable is important because the types of two variables determine how they are compared. In `gawk', variable typing follows these rules: * A numeric constant or the result of a numeric operation has the NUMERIC attribute. * A string constant or the result of a string operation has the STRING attribute. * Fields, `getline' input, `FILENAME', `ARGV' elements, `ENVIRON' elements, and the elements of an array created by `split' that are numeric strings have the STRNUM attribute. Otherwise, they have the STRING attribute. Uninitialized variables also have the STRNUM attribute. * Attributes propagate across assignments but are not changed by any use. The last rule is particularly important. In the following program, `a' has numeric type, even though it is later used in a string operation: BEGIN { a = 12.345 b = a " is a cute number" print b } When two operands are compared, either string comparison or numeric comparison may be used. This depends upon the attributes of the operands, according to the following symmetric matrix: +--------------------------------------------- | STRING NUMERIC STRNUM -------+--------------------------------------------- | STRING | string string string | NUMERIC | string numeric numeric | STRNUM | string numeric numeric -------+--------------------------------------------- The basic idea is that user input that looks numeric--and _only_ user input--should be treated as numeric, even though it is actually made of characters and is therefore also a string. Thus, for example, the string constant `" +3.14"' is a string, even though it looks numeric, and is _never_ treated as number for comparison purposes. In short, when one operand is a "pure" string, such as a string constant, then a string comparison is performed. Otherwise, a numeric comparison is performed.(1) "Comparison expressions" compare strings or numbers for relationships such as equality. They are written using "relational operators", which are a superset of those in C. *Note table-relational-ops:: describes them. Expression Result -------------------------------------------------------------------------- X `<' Y True if X is less than Y. X `<=' Y True if X is less than or equal to Y. X `>' Y True if X is greater than Y. X `>=' Y True if X is greater than or equal to Y. X `==' Y True if X is equal to Y. X `!=' Y True if X is not equal to Y. X `~' Y True if the string X matches the regexp denoted by Y. X `!~' Y True if the string X does not match the regexp denoted by Y. SUBSCRIPT `in' True if the array ARRAY has an element with the ARRAY subscript SUBSCRIPT. Table 5.2: Relational Operators Comparison expressions have the value one if true and zero if false. When comparing operands of mixed types, numeric operands are converted to strings using the value of `CONVFMT' (*note Conversion::). Strings are compared by comparing the first character of each, then the second character of each, and so on. Thus, `"10"' is less than `"9"'. If there are two strings where one is a prefix of the other, the shorter string is less than the longer one. Thus, `"abc"' is less than `"abcd"'. It is very easy to accidentally mistype the `==' operator and leave off one of the `=' characters. The result is still valid `awk' code, but the program does not do what is intended: if (a = b) # oops! should be a == b ... else ... Unless `b' happens to be zero or the null string, the `if' part of the test always succeeds. Because the operators are so similar, this kind of error is very difficult to spot when scanning the source code. The following table of expressions illustrates the kind of comparison `gawk' performs, as well as what the result of the comparison is: `1.5 <= 2.0' numeric comparison (true) `"abc" >= "xyz"' string comparison (false) `1.5 != " +2"' string comparison (true) `"1e2" < "3"' string comparison (true) `a = 2; b = "2"' `a == b' string comparison (true) `a = 2; b = " +2"' `a == b' string comparison (false) In the next example: $ echo 1e2 3 | awk '{ print ($1 < $2) ? "true" : "false" }' -| false the result is `false' because both `$1' and `$2' are user input. They are numeric strings--therefore both have the STRNUM attribute, dictating a numeric comparison. The purpose of the comparison rules and the use of numeric strings is to attempt to produce the behavior that is "least surprising," while still "doing the right thing." String comparisons and regular expression comparisons are very different. For example: x == "foo" has the value one, or is true if the variable `x' is precisely `foo'. By contrast: x ~ /foo/ has the value one if `x' contains `foo', such as `"Oh, what a fool am I!"'. The righthand operand of the `~' and `!~' operators may be either a regexp constant (`/.../') or an ordinary expression. In the latter case, the value of the expression as a string is used as a dynamic regexp (*note Regexp Usage::; also *note Computed Regexps::). In modern implementations of `awk', a constant regular expression in slashes by itself is also an expression. The regexp `/REGEXP/' is an abbreviation for the following comparison expression: $0 ~ /REGEXP/ One special place where `/foo/' is _not_ an abbreviation for `$0 ~ /foo/' is when it is the righthand operand of `~' or `!~'. *Note Using Constant Regexps::, where this is discussed in more detail. ---------- Footnotes ---------- (1) The POSIX standard is under revision. The revised standard's rules for typing and comparison are the same as just described for `gawk'. 5.11 Boolean Expressions ======================== A "Boolean expression" is a combination of comparison expressions or matching expressions, using the Boolean operators "or" (`||'), "and" (`&&'), and "not" (`!'), along with parentheses to control nesting. The truth value of the Boolean expression is computed by combining the truth values of the component expressions. Boolean expressions are also referred to as "logical expressions". The terms are equivalent. Boolean expressions can be used wherever comparison and matching expressions can be used. They can be used in `if', `while', `do', and `for' statements (*note Statements::). They have numeric values (one if true, zero if false) that come into play if the result of the Boolean expression is stored in a variable or used in arithmetic. In addition, every Boolean expression is also a valid pattern, so you can use one as a pattern to control the execution of rules. The Boolean operators are: `BOOLEAN1 && BOOLEAN2' True if both BOOLEAN1 and BOOLEAN2 are true. For example, the following statement prints the current input record if it contains both `2400' and `foo': if ($0 ~ /2400/ && $0 ~ /foo/) print The subexpression BOOLEAN2 is evaluated only if BOOLEAN1 is true. This can make a difference when BOOLEAN2 contains expressions that have side effects. In the case of `$0 ~ /foo/ && ($2 == bar++)', the variable `bar' is not incremented if there is no substring `foo' in the record. `BOOLEAN1 || BOOLEAN2' True if at least one of BOOLEAN1 or BOOLEAN2 is true. For example, the following statement prints all records in the input that contain _either_ `2400' or `foo' or both: if ($0 ~ /2400/ || $0 ~ /foo/) print The subexpression BOOLEAN2 is evaluated only if BOOLEAN1 is false. This can make a difference when BOOLEAN2 contains expressions that have side effects. `! BOOLEAN' True if BOOLEAN is false. For example, the following program prints `no home!' in the unusual event that the `HOME' environment variable is not defined: BEGIN { if (! ("HOME" in ENVIRON)) print "no home!" } (The `in' operator is described in *Note Reference to Elements::.) The `&&' and `||' operators are called "short-circuit" operators because of the way they work. Evaluation of the full expression is "short-circuited" if the result can be determined part way through its evaluation. Statements that use `&&' or `||' can be continued simply by putting a newline after them. But you cannot put a newline in front of either of these operators without using backslash continuation (*note Statements/Lines::). The actual value of an expression using the `!' operator is either one or zero, depending upon the truth value of the expression it is applied to. The `!' operator is often useful for changing the sense of a flag variable from false to true and back again. For example, the following program is one way to print lines in between special bracketing lines: $1 == "START" { interested = ! interested; next } interested == 1 { print } $1 == "END" { interested = ! interested; next } The variable `interested', as with all `awk' variables, starts out initialized to zero, which is also false. When a line is seen whose first field is `START', the value of `interested' is toggled to true, using `!'. The next rule prints lines as long as `interested' is true. When a line is seen whose first field is `END', `interested' is toggled back to false. NOTE: The `next' statement is discussed in *Note Next Statement::. `next' tells `awk' to skip the rest of the rules, get the next record, and start processing the rules over again at the top. The reason it's there is to avoid printing the bracketing `START' and `END' lines. 5.12 Conditional Expressions ============================ A "conditional expression" is a special kind of expression that has three operands. It allows you to use one expression's value to select one of two other expressions. The conditional expression is the same as in the C language, as shown here: SELECTOR ? IF-TRUE-EXP : IF-FALSE-EXP There are three subexpressions. The first, SELECTOR, is always computed first. If it is "true" (not zero or not null), then IF-TRUE-EXP is computed next and its value becomes the value of the whole expression. Otherwise, IF-FALSE-EXP is computed next and its value becomes the value of the whole expression. For example, the following expression produces the absolute value of `x': x >= 0 ? x : -x Each time the conditional expression is computed, only one of IF-TRUE-EXP and IF-FALSE-EXP is used; the other is ignored. This is important when the expressions have side effects. For example, this conditional expression examines element `i' of either array `a' or array `b', and increments `i': x == y ? a[i++] : b[i++] This is guaranteed to increment `i' exactly once, because each time only one of the two increment expressions is executed and the other is not. *Note Arrays::, for more information about arrays. As a minor `gawk' extension, a statement that uses `?:' can be continued simply by putting a newline after either character. However, putting a newline in front of either character does not work without using backslash continuation (*note Statements/Lines::). If `--posix' is specified (*note Options::), then this extension is disabled. 5.13 Function Calls =================== A "function" is a name for a particular calculation. This enables you to ask for it by name at any point in the program. For example, the function `sqrt' computes the square root of a number. A fixed set of functions are "built-in", which means they are available in every `awk' program. The `sqrt' function is one of these. *Note Built-in::, for a list of built-in functions and their descriptions. In addition, you can define functions for use in your program. *Note User-defined::, for instructions on how to do this. The way to use a function is with a "function call" expression, which consists of the function name followed immediately by a list of "arguments" in parentheses. The arguments are expressions that provide the raw materials for the function's calculations. When there is more than one argument, they are separated by commas. If there are no arguments, just write `()' after the function name. The following examples show function calls with and without arguments: sqrt(x^2 + y^2) one argument atan2(y, x) two arguments rand() no arguments *Caution:* Do not put any space between the function name and the open-parenthesis! A user-defined function name looks just like the name of a variable--a space would make the expression look like concatenation of a variable with an expression inside parentheses. With built-in functions, space before the parenthesis is harmless, but it is best not to get into the habit of using space to avoid mistakes with user-defined functions. Each function expects a particular number of arguments. For example, the `sqrt' function must be called with a single argument, the number of which to take the square root: sqrt(ARGUMENT) Some of the built-in functions have one or more optional arguments. If those arguments are not supplied, the functions use a reasonable default value. *Note Built-in::, for full details. If arguments are omitted in calls to user-defined functions, then those arguments are treated as local variables and initialized to the empty string (*note User-defined::). Like every other expression, the function call has a value, which is computed by the function based on the arguments you give it. In this example, the value of `sqrt(ARGUMENT)' is the square root of ARGUMENT. A function can also have side effects, such as assigning values to certain variables or doing I/O. The following program reads numbers, one number per line, and prints the square root of each one: $ awk '{ print "The square root of", $1, "is", sqrt($1) }' 1 -| The square root of 1 is 1 3 -| The square root of 3 is 1.73205 5 -| The square root of 5 is 2.23607 Ctrl-d 5.14 Operator Precedence (How Operators Nest) ============================================= "Operator precedence" determines how operators are grouped when different operators appear close by in one expression. For example, `*' has higher precedence than `+'; thus, `a + b * c' means to multiply `b' and `c', and then add `a' to the product (i.e., `a + (b * c)'). The normal precedence of the operators can be overruled by using parentheses. Think of the precedence rules as saying where the parentheses are assumed to be. In fact, it is wise to always use parentheses whenever there is an unusual combination of operators, because other people who read the program may not remember what the precedence is in this case. Even experienced programmers occasionally forget the exact rules, which leads to mistakes. Explicit parentheses help prevent any such mistakes. When operators of equal precedence are used together, the leftmost operator groups first, except for the assignment, conditional, and exponentiation operators, which group in the opposite order. Thus, `a - b + c' groups as `(a - b) + c' and `a = b = c' groups as `a = (b = c)'. The precedence of prefix unary operators does not matter as long as only unary operators are involved, because there is only one way to interpret them: innermost first. Thus, `$++i' means `$(++i)' and `++$x' means `++($x)'. However, when another operator follows the operand, then the precedence of the unary operators can matter. `$x^2' means `($x)^2', but `-x^2' means `-(x^2)', because `-' has lower precedence than `^', whereas `$' has higher precedence. This table presents `awk''s operators, in order of highest to lowest precedence: `(...)' Grouping. `$' Field. `++ --' Increment, decrement. `^ **' Exponentiation. These operators group right-to-left. `+ - !' Unary plus, minus, logical "not." `* / %' Multiplication, division, modulus. `+ -' Addition, subtraction. `String Concatenation' No special symbol is used to indicate concatenation. The operands are simply written side by side (*note Concatenation::). `< <= == !=' `> >= >> | |&' Relational and redirection. The relational operators and the redirections have the same precedence level. Characters such as `>' serve both as relationals and as redirections; the context distinguishes between the two meanings. Note that the I/O redirection operators in `print' and `printf' statements belong to the statement level, not to expressions. The redirection does not produce an expression that could be the operand of another operator. As a result, it does not make sense to use a redirection operator near another operator of lower precedence without parentheses. Such combinations (for example, `print foo > a ? b : c'), result in syntax errors. The correct way to write this statement is `print foo > (a ? b : c)'. `~ !~' Matching, nonmatching. `in' Array membership. `&&' Logical "and". `||' Logical "or". `?:' Conditional. This operator groups right-to-left. `= += -= *=' `/= %= ^= **=' Assignment. These operators group right to left. NOTE: The `|&', `**', and `**=' operators are not specified by POSIX. For maximum portability, do not use them. 6 Patterns, Actions, and Variables ********************************** As you have already seen, each `awk' statement consists of a pattern with an associated action. This major node describes how you build patterns and actions, what kinds of things you can do within actions, and `awk''s built-in variables. The pattern-action rules and the statements available for use within actions form the core of `awk' programming. In a sense, everything covered up to here has been the foundation that programs are built on top of. Now it's time to start building something useful. 6.1 Pattern Elements ==================== Patterns in `awk' control the execution of rules--a rule is executed when its pattern matches the current input record. The following is a summary of the types of `awk' patterns: `/REGULAR EXPRESSION/' A regular expression. It matches when the text of the input record fits the regular expression. (*Note Regexp::.) `EXPRESSION' A single expression. It matches when its value is nonzero (if a number) or non-null (if a string). (*Note Expression Patterns::.) `PAT1, PAT2' A pair of patterns separated by a comma, specifying a range of records. The range includes both the initial record that matches PAT1 and the final record that matches PAT2. (*Note Ranges::.) `BEGIN' `END' Special patterns for you to supply startup or cleanup actions for your `awk' program. (*Note BEGIN/END::.) `EMPTY' The empty pattern matches every input record. (*Note Empty::.) 6.1.1 Regular Expressions as Patterns ------------------------------------- Regular expressions are one of the first kinds of patterns presented in this book. This kind of pattern is simply a regexp constant in the pattern part of a rule. Its meaning is `$0 ~ /PATTERN/'. The pattern matches when the input record matches the regexp. For example: /foo|bar|baz/ { buzzwords++ } END { print buzzwords, "buzzwords seen" } 6.1.2 Expressions as Patterns ----------------------------- Any `awk' expression is valid as an `awk' pattern. The pattern matches if the expression's value is nonzero (if a number) or non-null (if a string). The expression is reevaluated each time the rule is tested against a new input record. If the expression uses fields such as `$1', the value depends directly on the new input record's text; otherwise, it depends on only what has happened so far in the execution of the `awk' program. Comparison expressions, using the comparison operators described in *Note Typing and Comparison::, are a very common kind of pattern. Regexp matching and nonmatching are also very common expressions. The left operand of the `~' and `!~' operators is a string. The right operand is either a constant regular expression enclosed in slashes (`/REGEXP/'), or any expression whose string value is used as a dynamic regular expression (*note Computed Regexps::). The following example prints the second field of each input record whose first field is precisely `foo': $ awk '$1 == "foo" { print $2 }' BBS-list (There is no output, because there is no BBS site with the exact name `foo'.) Contrast this with the following regular expression match, which accepts any record with a first field that contains `foo': $ awk '$1 ~ /foo/ { print $2 }' BBS-list -| 555-1234 -| 555-6699 -| 555-6480 -| 555-2127 A regexp constant as a pattern is also a special case of an expression pattern. The expression `/foo/' has the value one if `foo' appears in the current input record. Thus, as a pattern, `/foo/' matches any record containing `foo'. Boolean expressions are also commonly used as patterns. Whether the pattern matches an input record depends on whether its subexpressions match. For example, the following command prints all the records in `BBS-list' that contain both `2400' and `foo': $ awk '/2400/ && /foo/' BBS-list -| fooey 555-1234 2400/1200/300 B The following command prints all records in `BBS-list' that contain _either_ `2400' or `foo' (or both, of course): $ awk '/2400/ || /foo/' BBS-list -| alpo-net 555-3412 2400/1200/300 A -| bites 555-1675 2400/1200/300 A -| fooey 555-1234 2400/1200/300 B -| foot 555-6699 1200/300 B -| macfoo 555-6480 1200/300 A -| sdace 555-3430 2400/1200/300 A -| sabafoo 555-2127 1200/300 C The following command prints all records in `BBS-list' that do _not_ contain the string `foo': $ awk '! /foo/' BBS-list -| aardvark 555-5553 1200/300 B -| alpo-net 555-3412 2400/1200/300 A -| barfly 555-7685 1200/300 A -| bites 555-1675 2400/1200/300 A -| camelot 555-0542 300 C -| core 555-2912 1200/300 C -| sdace 555-3430 2400/1200/300 A The subexpressions of a Boolean operator in a pattern can be constant regular expressions, comparisons, or any other `awk' expressions. Range patterns are not expressions, so they cannot appear inside Boolean patterns. Likewise, the special patterns `BEGIN' and `END', which never match any input record, are not expressions and cannot appear inside Boolean patterns. 6.1.3 Specifying Record Ranges with Patterns -------------------------------------------- A "range pattern" is made of two patterns separated by a comma, in the form `BEGPAT, ENDPAT'. It is used to match ranges of consecutive input records. The first pattern, BEGPAT, controls where the range begins, while ENDPAT controls where the pattern ends. For example, the following: awk '$1 == "on", $1 == "off"' myfile prints every record in `myfile' between `on'/`off' pairs, inclusive. A range pattern starts out by matching BEGPAT against every input record. When a record matches BEGPAT, the range pattern is "turned on" and the range pattern matches this record as well. As long as the range pattern stays turned on, it automatically matches every input record read. The range pattern also matches ENDPAT against every input record; when this succeeds, the range pattern is turned off again for the following record. Then the range pattern goes back to checking BEGPAT against each record. The record that turns on the range pattern and the one that turns it off both match the range pattern. If you don't want to operate on these records, you can write `if' statements in the rule's action to distinguish them from the records you are interested in. It is possible for a pattern to be turned on and off by the same record. If the record satisfies both conditions, then the action is executed for just that record. For example, suppose there is text between two identical markers (e.g., the `%' symbol), each on its own line, that should be ignored. A first attempt would be to combine a range pattern that describes the delimited text with the `next' statement (not discussed yet, *note Next Statement::). This causes `awk' to skip any further processing of the current record and start over again with the next input record. Such a program looks like this: /^%$/,/^%$/ { next } { print } This program fails because the range pattern is both turned on and turned off by the first line, which just has a `%' on it. To accomplish this task, write the program in the following manner, using a flag: /^%$/ { skip = ! skip; next } skip == 1 { next } # skip lines with `skip' set In a range pattern, the comma (`,') has the lowest precedence of all the operators (i.e., it is evaluated last). Thus, the following program attempts to combine a range pattern with another, simpler test: echo Yes | awk '/1/,/2/ || /Yes/' The intent of this program is `(/1/,/2/) || /Yes/'. However, `awk' interprets this as `/1/, (/2/ || /Yes/)'. This cannot be changed or worked around; range patterns do not combine with other patterns: $ echo Yes | gawk '(/1/,/2/) || /Yes/' error--> gawk: cmd. line:1: (/1/,/2/) || /Yes/ error--> gawk: cmd. line:1: ^ parse error error--> gawk: cmd. line:2: (/1/,/2/) || /Yes/ error--> gawk: cmd. line:2: ^ unexpected newline 6.1.4 The `BEGIN' and `END' Special Patterns -------------------------------------------- All the patterns described so far are for matching input records. The `BEGIN' and `END' special patterns are different. They supply startup and cleanup actions for `awk' programs. `BEGIN' and `END' rules must have actions; there is no default action for these rules because there is no current record when they run. `BEGIN' and `END' rules are often referred to as "`BEGIN' and `END' blocks" by long-time `awk' programmers. 6.1.4.1 Startup and Cleanup Actions ................................... A `BEGIN' rule is executed once only, before the first input record is read. Likewise, an `END' rule is executed once only, after all the input is read. For example: $ awk ' > BEGIN { print "Analysis of \"foo\"" } > /foo/ { ++n } > END { print "\"foo\" appears", n, "times." }' BBS-list -| Analysis of "foo" -| "foo" appears 4 times. This program finds the number of records in the input file `BBS-list' that contain the string `foo'. The `BEGIN' rule prints a title for the report. There is no need to use the `BEGIN' rule to initialize the counter `n' to zero, since `awk' does this automatically (*note Variables::). The second rule increments the variable `n' every time a record containing the pattern `foo' is read. The `END' rule prints the value of `n' at the end of the run. The special patterns `BEGIN' and `END' cannot be used in ranges or with Boolean operators (indeed, they cannot be used with any operators). An `awk' program may have multiple `BEGIN' and/or `END' rules. They are executed in the order in which they appear: all the `BEGIN' rules at startup and all the `END' rules at termination. `BEGIN' and `END' rules may be intermixed with other rules. This feature was added in the 1987 version of `awk' and is included in the POSIX standard. The original (1978) version of `awk' required the `BEGIN' rule to be placed at the beginning of the program, the `END' rule to be placed at the end, and only allowed one of each. This is no longer required, but it is a good idea to follow this template in terms of program organization and readability. Multiple `BEGIN' and `END' rules are useful for writing library functions, because each library file can have its own `BEGIN' and/or `END' rule to do its own initialization and/or cleanup. The order in which library functions are named on the command line controls the order in which their `BEGIN' and `END' rules are executed. Therefore, you have to be careful when writing such rules in library files so that the order in which they are executed doesn't matter. *Note Options::, for more information on using library functions. *Note Library Functions::, for a number of useful library functions. If an `awk' program has only a `BEGIN' rule and no other rules, then the program exits after the `BEGIN' rule is run.(1) However, if an `END' rule exists, then the input is read, even if there are no other rules in the program. This is necessary in case the `END' rule checks the `FNR' and `NR' variables. ---------- Footnotes ---------- (1) The original version of `awk' used to keep reading and ignoring input until the end of the file was seen. 6.1.4.2 Input/Output from `BEGIN' and `END' Rules ................................................. There are several (sometimes subtle) points to remember when doing I/O from a `BEGIN' or `END' rule. The first has to do with the value of `$0' in a `BEGIN' rule. Because `BEGIN' rules are executed before any input is read, there simply is no input record, and therefore no fields, when executing `BEGIN' rules. References to `$0' and the fields yield a null string or zero, depending upon the context. One way to give `$0' a real value is to execute a `getline' command without a variable (*note Getline::). Another way is simply to assign a value to `$0'. The second point is similar to the first but from the other direction. Traditionally, due largely to implementation issues, `$0' and `NF' were _undefined_ inside an `END' rule. The POSIX standard specifies that `NF' is available in an `END' rule. It contains the number of fields from the last input record. Most probably due to an oversight, the standard does not say that `$0' is also preserved, although logically one would think that it should be. In fact, `gawk' does preserve the value of `$0' for use in `END' rules. Be aware, however, that Unix `awk', and possibly other implementations, do not. The third point follows from the first two. The meaning of `print' inside a `BEGIN' or `END' rule is the same as always: `print $0'. If `$0' is the null string, then this prints an empty line. Many long time `awk' programmers use an unadorned `print' in `BEGIN' and `END' rules, to mean `print ""', relying on `$0' being null. Although one might generally get away with this in `BEGIN' rules, it is a very bad idea in `END' rules, at least in `gawk'. It is also poor style, since if an empty line is needed in the output, the program should print one explicitly. Finally, the `next' and `nextfile' statements are not allowed in a `BEGIN' rule, because the implicit read-a-record-and-match-against-the-rules loop has not started yet. Similarly, those statements are not valid in an `END' rule, since all the input has been read. (*Note Next Statement::, and see *Note Nextfile Statement::.) 6.1.5 The Empty Pattern ----------------------- An empty (i.e., nonexistent) pattern is considered to match _every_ input record. For example, the program: awk '{ print $1 }' BBS-list prints the first field of every record. 6.2 Using Shell Variables in Programs ===================================== `awk' programs are often used as components in larger programs written in shell. For example, it is very common to use a shell variable to hold a pattern that the `awk' program searches for. There are two ways to get the value of the shell variable into the body of the `awk' program. The most common method is to use shell quoting to substitute the variable's value into the program inside the script. For example, in the following program: echo -n "Enter search pattern: " read pattern awk "/$pattern/ "'{ nmatches++ } END { print nmatches, "found" }' /path/to/data the `awk' program consists of two pieces of quoted text that are concatenated together to form the program. The first part is double-quoted, which allows substitution of the `pattern' variable inside the quotes. The second part is single-quoted. Variable substitution via quoting works, but can be potentially messy. It requires a good understanding of the shell's quoting rules (*note Quoting::), and it's often difficult to correctly match up the quotes when reading the program. A better method is to use `awk''s variable assignment feature (*note Assignment Options::) to assign the shell variable's value to an `awk' variable's value. Then use dynamic regexps to match the pattern (*note Computed Regexps::). The following shows how to redo the previous example using this technique: echo -n "Enter search pattern: " read pattern awk -v pat="$pattern" '$0 ~ pat { nmatches++ } END { print nmatches, "found" }' /path/to/data Now, the `awk' program is just one single-quoted string. The assignment `-v pat="$pattern"' still requires double quotes, in case there is whitespace in the value of `$pattern'. The `awk' variable `pat' could be named `pattern' too, but that would be more confusing. Using a variable also provides more flexibility, since the variable can be used anywhere inside the program--for printing, as an array subscript, or for any other use--without requiring the quoting tricks at every point in the program. 6.3 Actions =========== An `awk' program or script consists of a series of rules and function definitions interspersed. (Functions are described later. *Note User-defined::.) A rule contains a pattern and an action, either of which (but not both) may be omitted. The purpose of the "action" is to tell `awk' what to do once a match for the pattern is found. Thus, in outline, an `awk' program generally looks like this: [PATTERN] [{ ACTION }] [PATTERN] [{ ACTION }] ... function NAME(ARGS) { ... } ... An action consists of one or more `awk' "statements", enclosed in curly braces (`{...}'). Each statement specifies one thing to do. The statements are separated by newlines or semicolons. The curly braces around an action must be used even if the action contains only one statement, or if it contains no statements at all. However, if you omit the action entirely, omit the curly braces as well. An omitted action is equivalent to `{ print $0 }': /foo/ { } match `foo', do nothing -- empty action /foo/ match `foo', print the record -- omitted action The following types of statements are supported in `awk': Expressions Call functions or assign values to variables (*note Expressions::). Executing this kind of statement simply computes the value of the expression. This is useful when the expression has side effects (*note Assignment Ops::). Control statements Specify the control flow of `awk' programs. The `awk' language gives you C-like constructs (`if', `for', `while', and `do') as well as a few special ones (*note Statements::). Compound statements Consist of one or more statements enclosed in curly braces. A compound statement is used in order to put several statements together in the body of an `if', `while', `do', or `for' statement. Input statements Use the `getline' command (*note Getline::). Also supplied in `awk' are the `next' statement (*note Next Statement::), and the `nextfile' statement (*note Nextfile Statement::). Output statements Such as `print' and `printf'. *Note Printing::. Deletion statements For deleting array elements. *Note Delete::. 6.4 Control Statements in Actions ================================= "Control statements", such as `if', `while', and so on, control the flow of execution in `awk' programs. Most of the control statements in `awk' are patterned on similar statements in C. All the control statements start with special keywords, such as `if' and `while', to distinguish them from simple expressions. Many control statements contain other statements. For example, the `if' statement contains another statement that may or may not be executed. The contained statement is called the "body". To include more than one statement in the body, group them into a single "compound statement" with curly braces, separating them with newlines or semicolons. 6.4.1 The `if'-`else' Statement ------------------------------- The `if'-`else' statement is `awk''s decision-making statement. It looks like this: if (CONDITION) THEN-BODY [else ELSE-BODY] The CONDITION is an expression that controls what the rest of the statement does. If the CONDITION is true, THEN-BODY is executed; otherwise, ELSE-BODY is executed. The `else' part of the statement is optional. The condition is considered false if its value is zero or the null string; otherwise, the condition is true. Refer to the following: if (x % 2 == 0) print "x is even" else print "x is odd" In this example, if the expression `x % 2 == 0' is true (that is, if the value of `x' is evenly divisible by two), then the first `print' statement is executed; otherwise, the second `print' statement is executed. If the `else' keyword appears on the same line as THEN-BODY and THEN-BODY is not a compound statement (i.e., not surrounded by curly braces), then a semicolon must separate THEN-BODY from the `else'. To illustrate this, the previous example can be rewritten as: if (x % 2 == 0) print "x is even"; else print "x is odd" If the `;' is left out, `awk' can't interpret the statement and it produces a syntax error. Don't actually write programs this way, because a human reader might fail to see the `else' if it is not the first thing on its line. 6.4.2 The `while' Statement --------------------------- In programming, a "loop" is a part of a program that can be executed two or more times in succession. The `while' statement is the simplest looping statement in `awk'. It repeatedly executes a statement as long as a condition is true. For example: while (CONDITION) BODY BODY is a statement called the "body" of the loop, and CONDITION is an expression that controls how long the loop keeps running. The first thing the `while' statement does is test the CONDITION. If the CONDITION is true, it executes the statement BODY. (The CONDITION is true when the value is not zero and not a null string.) After BODY has been executed, CONDITION is tested again, and if it is still true, BODY is executed again. This process repeats until the CONDITION is no longer true. If the CONDITION is initially false, the body of the loop is never executed and `awk' continues with the statement following the loop. This example prints the first three fields of each record, one per line: awk '{ i = 1 while (i <= 3) { print $i i++ } }' inventory-shipped The body of this loop is a compound statement enclosed in braces, containing two statements. The loop works in the following manner: first, the value of `i' is set to one. Then, the `while' statement tests whether `i' is less than or equal to three. This is true when `i' equals one, so the `i'-th field is printed. Then the `i++' increments the value of `i' and the loop repeats. The loop terminates when `i' reaches four. A newline is not required between the condition and the body; however using one makes the program clearer unless the body is a compound statement or else is very simple. The newline after the open-brace that begins the compound statement is not required either, but the program is harder to read without it. 6.4.3 The `do'-`while' Statement -------------------------------- The `do' loop is a variation of the `while' looping statement. The `do' loop executes the BODY once and then repeats the BODY as long as the CONDITION is true. It looks like this: do BODY while (CONDITION) Even if the CONDITION is false at the start, the BODY is executed at least once (and only once, unless executing BODY makes CONDITION true). Contrast this with the corresponding `while' statement: while (CONDITION) BODY This statement does not execute BODY even once if the CONDITION is false to begin with. The following is an example of a `do' statement: { i = 1 do { print $0 i++ } while (i <= 10) } This program prints each input record 10 times. However, it isn't a very realistic example, since in this case an ordinary `while' would do just as well. This situation reflects actual experience; only occasionally is there a real use for a `do' statement. 6.4.4 The `for' Statement ------------------------- The `for' statement makes it more convenient to count iterations of a loop. The general form of the `for' statement looks like this: for (INITIALIZATION; CONDITION; INCREMENT) BODY The INITIALIZATION, CONDITION, and INCREMENT parts are arbitrary `awk' expressions, and BODY stands for any `awk' statement. The `for' statement starts by executing INITIALIZATION. Then, as long as the CONDITION is true, it repeatedly executes BODY and then INCREMENT. Typically, INITIALIZATION sets a variable to either zero or one, INCREMENT adds one to it, and CONDITION compares it against the desired number of iterations. For example: awk '{ for (i = 1; i <= 3; i++) print $i }' inventory-shipped This prints the first three fields of each input record, with one field per line. It isn't possible to set more than one variable in the INITIALIZATION part without using a multiple assignment statement such as `x = y = 0'. This makes sense only if all the initial values are equal. (But it is possible to initialize additional variables by writing their assignments as separate statements preceding the `for' loop.) The same is true of the INCREMENT part. Incrementing additional variables requires separate statements at the end of the loop. The C compound expression, using C's comma operator, is useful in this context but it is not supported in `awk'. Most often, INCREMENT is an increment expression, as in the previous example. But this is not required; it can be any expression whatsoever. For example, the following statement prints all the powers of two between 1 and 100: for (i = 1; i <= 100; i *= 2) print i If there is nothing to be done, any of the three expressions in the parentheses following the `for' keyword may be omitted. Thus, `for (; x > 0;)' is equivalent to `while (x > 0)'. If the CONDITION is omitted, it is treated as true, effectively yielding an "infinite loop" (i.e., a loop that never terminates). In most cases, a `for' loop is an abbreviation for a `while' loop, as shown here: INITIALIZATION while (CONDITION) { BODY INCREMENT } The only exception is when the `continue' statement (*note Continue Statement::) is used inside the loop. Changing a `for' statement to a `while' statement in this way can change the effect of the `continue' statement inside the loop. The `awk' language has a `for' statement in addition to a `while' statement because a `for' loop is often both less work to type and more natural to think of. Counting the number of iterations is very common in loops. It can be easier to think of this counting as part of looping rather than as something to do inside the loop. There is an alternate version of the `for' loop, for iterating over all the indices of an array: for (i in array) DO SOMETHING WITH array[i] *Note Scanning an Array::, for more information on this version of the `for' loop. 6.4.5 The `switch' Statement ---------------------------- *NOTE:* This node describes an experimental feature added in `gawk' 3.1.3. It is _not_ enabled by default. To enable it, use the `--enable-switch' option to `configure' when `gawk' is being configured and built. *Note Additional Configuration Options::, for more information. The `switch' statement allows the evaluation of an expression and the execution of statements based on a `case' match. Case statements are checked for a match in the order they are defined. If no suitable `case' is found, the `default' section is executed, if supplied. Each `case' contains a single constant, be it numeric, string, or regexp. The `switch' expression is evaluated, and then each `case''s constant is compared against the result in turn. The type of constant determines the comparison: numeric or string do the usual comparisons. A regexp constant does a regular expression match against the string value of the original expression. The general form of the `switch' statement looks like this: switch (EXPRESSION) { case VALUE OR REGULAR EXPRESSION: CASE-BODY default: DEFAULT-BODY } Control flow in the `switch' statement works as it does in C. Once a match to a given case is made, case statement bodies are executed until a `break', `continue', `next', `nextfile' or `exit' is encountered, or the end of the `switch' statement itself. For example: switch (NR * 2 + 1) { case 3: case "11": print NR - 1 break case /2[[:digit:]]+/: print NR default: print NR + 1 case -1: print NR * -1 } Note that if none of the statements specified above halt execution of a matched `case' statement, execution falls through to the next `case' until execution halts. In the above example, for any case value starting with `2' followed by one or more digits, the `print' statement is executed and then falls through into the `default' section, executing its `print' statement. In turn, the -1 case will also be executed since the `default' does not halt execution. 6.4.6 The `break' Statement --------------------------- The `break' statement jumps out of the innermost `for', `while', or `do' loop that encloses it. The following example finds the smallest divisor of any integer, and also identifies prime numbers: # find smallest divisor of num { num = $1 for (div = 2; div*div <= num; div++) if (num % div == 0) break if (num % div == 0) printf "Smallest divisor of %d is %d\n", num, div else printf "%d is prime\n", num } When the remainder is zero in the first `if' statement, `awk' immediately "breaks out" of the containing `for' loop. This means that `awk' proceeds immediately to the statement following the loop and continues processing. (This is very different from the `exit' statement, which stops the entire `awk' program. *Note Exit Statement::.) Th following program illustrates how the CONDITION of a `for' or `while' statement could be replaced with a `break' inside an `if': # find smallest divisor of num { num = $1 for (div = 2; ; div++) { if (num % div == 0) { printf "Smallest divisor of %d is %d\n", num, div break } if (div*div > num) { printf "%d is prime\n", num break } } } The `break' statement has no meaning when used outside the body of a loop. However, although it was never documented, historical implementations of `awk' treated the `break' statement outside of a loop as if it were a `next' statement (*note Next Statement::). Recent versions of Unix `awk' no longer allow this usage. `gawk' supports this use of `break' only if `--traditional' has been specified on the command line (*note Options::). Otherwise, it is treated as an error, since the POSIX standard specifies that `break' should only be used inside the body of a loop. (d.c.) 6.4.7 The `continue' Statement ------------------------------ As with `break', the `continue' statement is used only inside `for', `while', and `do' loops. It skips over the rest of the loop body, causing the next cycle around the loop to begin immediately. Contrast this with `break', which jumps out of the loop altogether. The `continue' statement in a `for' loop directs `awk' to skip the rest of the body of the loop and resume execution with the increment-expression of the `for' statement. The following program illustrates this fact: BEGIN { for (x = 0; x <= 20; x++) { if (x == 5) continue printf "%d ", x } print "" } This program prints all the numbers from 0 to 20--except for 5, for which the `printf' is skipped. Because the increment `x++' is not skipped, `x' does not remain stuck at 5. Contrast the `for' loop from the previous example with the following `while' loop: BEGIN { x = 0 while (x <= 20) { if (x == 5) continue printf "%d ", x x++ } print "" } This program loops forever once `x' reaches 5. The `continue' statement has no meaning when used outside the body of a loop. Historical versions of `awk' treated a `continue' statement outside a loop the same way they treated a `break' statement outside a loop: as if it were a `next' statement (*note Next Statement::). Recent versions of Unix `awk' no longer work this way, and `gawk' allows it only if `--traditional' is specified on the command line (*note Options::). Just like the `break' statement, the POSIX standard specifies that `continue' should only be used inside the body of a loop. (d.c.) 6.4.8 The `next' Statement -------------------------- The `next' statement forces `awk' to immediately stop processing the current record and go on to the next record. This means that no further rules are executed for the current record, and the rest of the current rule's action isn't executed. Contrast this with the effect of the `getline' function (*note Getline::). That also causes `awk' to read the next record immediately, but it does not alter the flow of control in any way (i.e., the rest of the current action executes with a new input record). At the highest level, `awk' program execution is a loop that reads an input record and then tests each rule's pattern against it. If you think of this loop as a `for' statement whose body contains the rules, then the `next' statement is analogous to a `continue' statement. It skips to the end of the body of this implicit loop and executes the increment (which reads another record). For example, suppose an `awk' program works only on records with four fields, and it shouldn't fail when given bad input. To avoid complicating the rest of the program, write a "weed out" rule near the beginning, in the following manner: NF != 4 { err = sprintf("%s:%d: skipped: NF != 4\n", FILENAME, FNR) print err > "/dev/stderr" next } Because of the `next' statement, the program's subsequent rules won't see the bad record. The error message is redirected to the standard error output stream, as error messages should be. For more detail see *Note Special Files::. According to the POSIX standard, the behavior is undefined if the `next' statement is used in a `BEGIN' or `END' rule. `gawk' treats it as a syntax error. Although POSIX permits it, some other `awk' implementations don't allow the `next' statement inside function bodies (*note User-defined::). Just as with any other `next' statement, a `next' statement inside a function body reads the next record and starts processing it with the first rule in the program. If the `next' statement causes the end of the input to be reached, then the code in any `END' rules is executed. *Note BEGIN/END::. 6.4.9 Using `gawk''s `nextfile' Statement ----------------------------------------- `gawk' provides the `nextfile' statement, which is similar to the `next' statement. However, instead of abandoning processing of the current record, the `nextfile' statement instructs `gawk' to stop processing the current data file. The `nextfile' statement is a `gawk' extension. In most other `awk' implementations, or if `gawk' is in compatibility mode (*note Options::), `nextfile' is not special. Upon execution of the `nextfile' statement, `FILENAME' is updated to the name of the next data file listed on the command line, `FNR' is reset to one, `ARGIND' is incremented, and processing starts over with the first rule in the program. (`ARGIND' hasn't been introduced yet. *Note Built-in Variables::.) If the `nextfile' statement causes the end of the input to be reached, then the code in any `END' rules is executed. *Note BEGIN/END::. The `nextfile' statement is useful when there are many data files to process but it isn't necessary to process every record in every file. Normally, in order to move on to the next data file, a program has to continue scanning the unwanted records. The `nextfile' statement accomplishes this much more efficiently. While one might think that `close(FILENAME)' would accomplish the same as `nextfile', this isn't true. `close' is reserved for closing files, pipes, and coprocesses that are opened with redirections. It is not related to the main processing that `awk' does with the files listed in `ARGV'. If it's necessary to use an `awk' version that doesn't support `nextfile', see *Note Nextfile Function::, for a user-defined function that simulates the `nextfile' statement. The current version of the Bell Laboratories `awk' (*note Other Versions::) also supports `nextfile'. However, it doesn't allow the `nextfile' statement inside function bodies (*note User-defined::). `gawk' does; a `nextfile' inside a function body reads the next record and starts processing it with the first rule in the program, just as any other `nextfile' statement. *Caution:* Versions of `gawk' prior to 3.0 used two words (`next file') for the `nextfile' statement. In version 3.0, this was changed to one word, because the treatment of `file' was inconsistent. When it appeared after `next', `file' was a keyword; otherwise, it was a regular identifier. The old usage is no longer accepted; `next file' generates a syntax error. 6.4.10 The `exit' Statement --------------------------- The `exit' statement causes `awk' to immediately stop executing the current rule and to stop processing input; any remaining input is ignored. The `exit' statement is written as follows: exit [RETURN CODE] When an `exit' statement is executed from a `BEGIN' rule, the program stops processing everything immediately. No input records are read. However, if an `END' rule is present, as part of executing the `exit' statement, the `END' rule is executed (*note BEGIN/END::). If `exit' is used as part of an `END' rule, it causes the program to stop immediately. An `exit' statement that is not part of a `BEGIN' or `END' rule stops the execution of any further automatic rules for the current record, skips reading any remaining input records, and executes the `END' rule if there is one. In such a case, if you don't want the `END' rule to do its job, set a variable to nonzero before the `exit' statement and check that variable in the `END' rule. *Note Assert Function::, for an example that does this. If an argument is supplied to `exit', its value is used as the exit status code for the `awk' process. If no argument is supplied, `exit' returns status zero (success). In the case where an argument is supplied to a first `exit' statement, and then `exit' is called a second time from an `END' rule with no argument, `awk' uses the previously supplied exit value. (d.c.) For example, suppose an error condition occurs that is difficult or impossible to handle. Conventionally, programs report this by exiting with a nonzero status. An `awk' program can do this using an `exit' statement with a nonzero argument, as shown in the following example: BEGIN { if (("date" | getline date_now) <= 0) { print "Can't get system date" > "/dev/stderr" exit 1 } print "current date is", date_now close("date") } 6.5 Built-in Variables ====================== Most `awk' variables are available to use for your own purposes; they never change unless your program assigns values to them, and they never affect anything unless your program examines them. However, a few variables in `awk' have special built-in meanings. `awk' examines some of these automatically, so that they enable you to tell `awk' how to do certain things. Others are set automatically by `awk', so that they carry information from the internal workings of `awk' to your program. This minor node documents all the built-in variables of `gawk', most of which are also documented in the chapters describing their areas of activity. 6.5.1 Built-in Variables That Control `awk' ------------------------------------------- The following is an alphabetical list of variables that you can change to control how `awk' does certain things. The variables that are specific to `gawk' are marked with a pound sign (`#'). `BINMODE #' On non-POSIX systems, this variable specifies use of binary mode for all I/O. Numeric values of one, two, or three specify that input files, output files, or all files, respectively, should use binary I/O. Alternatively, string values of `"r"' or `"w"' specify that input files and output files, respectively, should use binary I/O. A string value of `"rw"' or `"wr"' indicates that all files should use binary I/O. Any other string value is equivalent to `"rw"', but `gawk' generates a warning message. `BINMODE' is described in more detail in *Note PC Using::. This variable is a `gawk' extension. In other `awk' implementations (except `mawk', *note Other Versions::), or if `gawk' is in compatibility mode (*note Options::), it is not special. `CONVFMT' This string controls conversion of numbers to strings (*note Conversion::). It works by being passed, in effect, as the first argument to the `sprintf' function (*note String Functions::). Its default value is `"%.6g"'. `CONVFMT' was introduced by the POSIX standard. `FIELDWIDTHS #' This is a space-separated list of columns that tells `gawk' how to split input with fixed columnar boundaries. Assigning a value to `FIELDWIDTHS' overrides the use of `FS' for field splitting. *Note Constant Size::, for more information. If `gawk' is in compatibility mode (*note Options::), then `FIELDWIDTHS' has no special meaning, and field-splitting operations occur based exclusively on the value of `FS'. `FS' This is the input field separator (*note Field Separators::). The value is a single-character string or a multi-character regular expression that matches the separations between fields in an input record. If the value is the null string (`""'), then each character in the record becomes a separate field. (This behavior is a `gawk' extension. POSIX `awk' does not specify the behavior when `FS' is the null string.) The default value is `" "', a string consisting of a single space. As a special exception, this value means that any sequence of spaces, tabs, and/or newlines is a single separator.(1) It also causes spaces, tabs, and newlines at the beginning and end of a record to be ignored. You can set the value of `FS' on the command line using the `-F' option: awk -F, 'PROGRAM' INPUT-FILES If `gawk' is using `FIELDWIDTHS' for field splitting, assigning a value to `FS' causes `gawk' to return to the normal, `FS'-based field splitting. An easy way to do this is to simply say `FS = FS', perhaps with an explanatory comment. `IGNORECASE #' If `IGNORECASE' is nonzero or non-null, then all string comparisons and all regular expression matching are case independent. Thus, regexp matching with `~' and `!~', as well as the `gensub', `gsub', `index', `match', `split', and `sub' functions, record termination with `RS', and field splitting with `FS', all ignore case when doing their particular regexp operations. However, the value of `IGNORECASE' does _not_ affect array subscripting and it does not affect field splitting when using a single-character field separator. *Note Case-sensitivity::. If `gawk' is in compatibility mode (*note Options::), then `IGNORECASE' has no special meaning. Thus, string and regexp operations are always case-sensitive. `LINT #' When this variable is true (nonzero or non-null), `gawk' behaves as if the `--lint' command-line option is in effect. (*note Options::). With a value of `"fatal"', lint warnings become fatal errors. With a value of `"invalid"', only warnings about things that are actually invalid are issued. (This is not fully implemented yet.) Any other true value prints nonfatal warnings. Assigning a false value to `LINT' turns off the lint warnings. This variable is a `gawk' extension. It is not special in other `awk' implementations. Unlike the other special variables, changing `LINT' does affect the production of lint warnings, even if `gawk' is in compatibility mode. Much as the `--lint' and `--traditional' options independently control different aspects of `gawk''s behavior, the control of lint warnings during program execution is independent of the flavor of `awk' being executed. `OFMT' This string controls conversion of numbers to strings (*note Conversion::) for printing with the `print' statement. It works by being passed as the first argument to the `sprintf' function (*note String Functions::). Its default value is `"%.6g"'. Earlier versions of `awk' also used `OFMT' to specify the format for converting numbers to strings in general expressions; this is now done by `CONVFMT'. `OFS' This is the output field separator (*note Output Separators::). It is output between the fields printed by a `print' statement. Its default value is `" "', a string consisting of a single space. `ORS' This is the output record separator. It is output at the end of every `print' statement. Its default value is `"\n"', the newline character. (*Note Output Separators::.) `RS' This is `awk''s input record separator. Its default value is a string containing a single newline character, which means that an input record consists of a single line of text. It can also be the null string, in which case records are separated by runs of blank lines. If it is a regexp, records are separated by matches of the regexp in the input text. (*Note Records::.) The ability for `RS' to be a regular expression is a `gawk' extension. In most other `awk' implementations, or if `gawk' is in compatibility mode (*note Options::), just the first character of `RS''s value is used. `SUBSEP' This is the subscript separator. It has the default value of `"\034"' and is used to separate the parts of the indices of a multidimensional array. Thus, the expression `foo["A", "B"]' really accesses `foo["A\034B"]' (*note Multi-dimensional::). `TEXTDOMAIN #' This variable is used for internationalization of programs at the `awk' level. It sets the default text domain for specially marked string constants in the source text, as well as for the `dcgettext', `dcngettext' and `bindtextdomain' functions (*note Internationalization::). The default value of `TEXTDOMAIN' is `"messages"'. This variable is a `gawk' extension. In other `awk' implementations, or if `gawk' is in compatibility mode (*note Options::), it is not special. ---------- Footnotes ---------- (1) In POSIX `awk', newline does not count as whitespace. 6.5.2 Built-in Variables That Convey Information ------------------------------------------------ The following is an alphabetical list of variables that `awk' sets automatically on certain occasions in order to provide information to your program. The variables that are specific to `gawk' are marked with a pound sign (`#'). `ARGC, ARGV' The command-line arguments available to `awk' programs are stored in an array called `ARGV'. `ARGC' is the number of command-line arguments present. *Note Other Arguments::. Unlike most `awk' arrays, `ARGV' is indexed from 0 to `ARGC' - 1. In the following example: $ awk 'BEGIN { > for (i = 0; i < ARGC; i++) > print ARGV[i] > }' inventory-shipped BBS-list -| awk -| inventory-shipped -| BBS-list `ARGV[0]' contains `"awk"', `ARGV[1]' contains `"inventory-shipped"', and `ARGV[2]' contains `"BBS-list"'. The value of `ARGC' is three, one more than the index of the last element in `ARGV', because the elements are numbered from zero. The names `ARGC' and `ARGV', as well as the convention of indexing the array from 0 to `ARGC' - 1, are derived from the C language's method of accessing command-line arguments. The value of `ARGV[0]' can vary from system to system. Also, you should note that the program text is _not_ included in `ARGV', nor are any of `awk''s command-line options. *Note ARGC and ARGV::, for information about how `awk' uses these variables. `ARGIND #' The index in `ARGV' of the current file being processed. Every time `gawk' opens a new data file for processing, it sets `ARGIND' to the index in `ARGV' of the file name. When `gawk' is processing the input files, `FILENAME == ARGV[ARGIND]' is always true. This variable is useful in file processing; it allows you to tell how far along you are in the list of data files as well as to distinguish between successive instances of the same file name on the command line. While you can change the value of `ARGIND' within your `awk' program, `gawk' automatically sets it to a new value when the next file is opened. This variable is a `gawk' extension. In other `awk' implementations, or if `gawk' is in compatibility mode (*note Options::), it is not special. `ENVIRON' An associative array that contains the values of the environment. The array indices are the environment variable names; the elements are the values of the particular environment variables. For example, `ENVIRON["HOME"]' might be `/home/arnold'. Changing this array does not affect the environment passed on to any programs that `awk' may spawn via redirection or the `system' function. Some operating systems may not have environment variables. On such systems, the `ENVIRON' array is empty (except for `ENVIRON["AWKPATH"]', *note AWKPATH Variable::). `ERRNO #' If a system error occurs during a redirection for `getline', during a read for `getline', or during a `close' operation, then `ERRNO' contains a string describing the error. `ERRNO' works similarly to the C variable `errno'. In particular `gawk' _never_ clears it (sets it to zero or `""'). Thus, you should only expect its value to be meaningful when an I/O operation returns a failure value, such as `getline' returning -1. You are, of course, free to clear it yourself before doing an I/O operation. This variable is a `gawk' extension. In other `awk' implementations, or if `gawk' is in compatibility mode (*note Options::), it is not special. `FILENAME' The name of the file that `awk' is currently reading. When no data files are listed on the command line, `awk' reads from the standard input and `FILENAME' is set to `"-"'. `FILENAME' is changed each time a new file is read (*note Reading Files::). Inside a `BEGIN' rule, the value of `FILENAME' is `""', since there are no input files being processed yet.(1) (d.c.) Note, though, that using `getline' (*note Getline::) inside a `BEGIN' rule can give `FILENAME' a value. `FNR' The current record number in the current file. `FNR' is incremented each time a new record is read (*note Getline::). It is reinitialized to zero each time a new input file is started. `NF' The number of fields in the current input record. `NF' is set each time a new record is read, when a new field is created or when `$0' changes (*note Fields::). Unlike most of the variables described in this node, assigning a value to `NF' has the potential to affect `awk''s internal workings. In particular, assignments to `NF' can be used to create or remove fields from the current record: *Note Changing Fields::. `NR' The number of input records `awk' has processed since the beginning of the program's execution (*note Records::). `NR' is incremented each time a new record is read. `PROCINFO #' The elements of this array provide access to information about the running `awk' program. The following elements (listed alphabetically) are guaranteed to be available: `PROCINFO["egid"]' The value of the `getegid' system call. `PROCINFO["euid"]' The value of the `geteuid' system call. `PROCINFO["FS"]' This is `"FS"' if field splitting with `FS' is in effect, or it is `"FIELDWIDTHS"' if field splitting with `FIELDWIDTHS' is in effect. `PROCINFO["gid"]' The value of the `getgid' system call. `PROCINFO["pgrpid"]' The process group ID of the current process. `PROCINFO["pid"]' The process ID of the current process. `PROCINFO["ppid"]' The parent process ID of the current process. `PROCINFO["uid"]' The value of the `getuid' system call. `PROCINFO["version"]' The version of `gawk'. This is available from version 3.1.4 and later. On some systems, there may be elements in the array, `"group1"' through `"groupN"' for some N. N is the number of supplementary groups that the process has. Use the `in' operator to test for these elements (*note Reference to Elements::). This array is a `gawk' extension. In other `awk' implementations, or if `gawk' is in compatibility mode (*note Options::), it is not special. `RLENGTH' The length of the substring matched by the `match' function (*note String Functions::). `RLENGTH' is set by invoking the `match' function. Its value is the length of the matched string, or -1 if no match is found. `RSTART' The start-index in characters of the substring that is matched by the `match' function (*note String Functions::). `RSTART' is set by invoking the `match' function. Its value is the position of the string where the matched substring starts, or zero if no match was found. `RT #' This is set each time a record is read. It contains the input text that matched the text denoted by `RS', the record separator. This variable is a `gawk' extension. In other `awk' implementations, or if `gawk' is in compatibility mode (*note Options::), it is not special. Advanced Notes: Changing `NR' and `FNR' --------------------------------------- `awk' increments `NR' and `FNR' each time it reads a record, instead of setting them to the absolute value of the number of records read. This means that a program can change these variables and their new values are incremented for each record. (d.c.) This is demonstrated in the following example: $ echo '1 > 2 > 3 > 4' | awk 'NR == 2 { NR = 17 } > { print NR }' -| 1 -| 17 -| 18 -| 19 Before `FNR' was added to the `awk' language (*note V7/SVR3.1::), many `awk' programs used this feature to track the number of records in a file by resetting `NR' to zero when `FILENAME' changed. ---------- Footnotes ---------- (1) Some early implementations of Unix `awk' initialized `FILENAME' to `"-"', even if there were data files to be processed. This behavior was incorrect and should not be relied upon in your programs. 6.5.3 Using `ARGC' and `ARGV' ----------------------------- *Note Auto-set::, presented the following program describing the information contained in `ARGC' and `ARGV': $ awk 'BEGIN { > for (i = 0; i < ARGC; i++) > print ARGV[i] > }' inventory-shipped BBS-list -| awk -| inventory-shipped -| BBS-list In this example, `ARGV[0]' contains `awk', `ARGV[1]' contains `inventory-shipped', and `ARGV[2]' contains `BBS-list'. Notice that the `awk' program is not entered in `ARGV'. The other special command-line options, with their arguments, are also not entered. This includes variable assignments done with the `-v' option (*note Options::). Normal variable assignments on the command line _are_ treated as arguments and do show up in the `ARGV' array: $ cat showargs.awk -| BEGIN { -| printf "A=%d, B=%d\n", A, B -| for (i = 0; i < ARGC; i++) -| printf "\tARGV[%d] = %s\n", i, ARGV[i] -| } -| END { printf "A=%d, B=%d\n", A, B } $ awk -v A=1 -f showargs.awk B=2 /dev/null -| A=1, B=0 -| ARGV[0] = awk -| ARGV[1] = B=2 -| ARGV[2] = /dev/null -| A=1, B=2 A program can alter `ARGC' and the elements of `ARGV'. Each time `awk' reaches the end of an input file, it uses the next element of `ARGV' as the name of the next input file. By storing a different string there, a program can change which files are read. Use `"-"' to represent the standard input. Storing additional elements and incrementing `ARGC' causes additional files to be read. If the value of `ARGC' is decreased, that eliminates input files from the end of the list. By recording the old value of `ARGC' elsewhere, a program can treat the eliminated arguments as something other than file names. To eliminate a file from the middle of the list, store the null string (`""') into `ARGV' in place of the file's name. As a special feature, `awk' ignores file names that have been replaced with the null string. Another option is to use the `delete' statement to remove elements from `ARGV' (*note Delete::). All of these actions are typically done in the `BEGIN' rule, before actual processing of the input begins. *Note Split Program::, and see *Note Tee Program::, for examples of each way of removing elements from `ARGV'. The following fragment processes `ARGV' in order to examine, and then remove, command-line options: BEGIN { for (i = 1; i < ARGC; i++) { if (ARGV[i] == "-v") verbose = 1 else if (ARGV[i] == "-d") debug = 1 else if (ARGV[i] ~ /^-?/) { e = sprintf("%s: unrecognized option -- %c", ARGV[0], substr(ARGV[i], 1, ,1)) print e > "/dev/stderr" } else break delete ARGV[i] } } To actually get the options into the `awk' program, end the `awk' options with `--' and then supply the `awk' program's options, in the following manner: awk -f myprog -- -v -d file1 file2 ... This is not necessary in `gawk'. Unless `--posix' has been specified, `gawk' silently puts any unrecognized options into `ARGV' for the `awk' program to deal with. As soon as it sees an unknown option, `gawk' stops looking for other options that it might otherwise recognize. The previous example with `gawk' would be: gawk -f myprog -d -v file1 file2 ... Because `-d' is not a valid `gawk' option, it and the following `-v' are passed on to the `awk' program. 7 Arrays in `awk' ***************** An "array" is a table of values called "elements". The elements of an array are distinguished by their indices. "Indices" may be either numbers or strings. This major node describes how arrays work in `awk', how to use array elements, how to scan through every element in an array, and how to remove array elements. It also describes how `awk' simulates multidimensional arrays, as well as some of the less obvious points about array usage. The major node finishes with a discussion of `gawk''s facility for sorting an array based on its indices. `awk' maintains a single set of names that may be used for naming variables, arrays, and functions (*note User-defined::). Thus, you cannot have a variable and an array with the same name in the same `awk' program. 7.1 Introduction to Arrays ========================== The `awk' language provides one-dimensional arrays for storing groups of related strings or numbers. Every `awk' array must have a name. Array names have the same syntax as variable names; any valid variable name would also be a valid array name. But one name cannot be used in both ways (as an array and as a variable) in the same `awk' program. Arrays in `awk' superficially resemble arrays in other programming languages, but there are fundamental differences. In `awk', it isn't necessary to specify the size of an array before starting to use it. Additionally, any number or string in `awk', not just consecutive integers, may be used as an array index. In most other languages, arrays must be "declared" before use, including a specification of how many elements or components they contain. In such languages, the declaration causes a contiguous block of memory to be allocated for that many elements. Usually, an index in the array must be a positive integer. For example, the index zero specifies the first element in the array, which is actually stored at the beginning of the block of memory. Index one specifies the second element, which is stored in memory right after the first element, and so on. It is impossible to add more elements to the array, because it has room only for as many elements as given in the declaration. (Some languages allow arbitrary starting and ending indices--e.g., `15 .. 27'--but the size of the array is still fixed when the array is declared.) A contiguous array of four elements might look like the following example, conceptually, if the element values are 8, `"foo"', `""', and 30: +---------+---------+--------+---------+ | 8 | "foo" | "" | 30 | Value +---------+---------+--------+---------+ 0 1 2 3 Index Only the values are stored; the indices are implicit from the order of the values. Here, 8 is the value at index zero, because 8 appears in the position with zero elements before it. Arrays in `awk' are different--they are "associative". This means that each array is a collection of pairs: an index and its corresponding array element value: Element 3 Value 30 Element 1 Value "foo" Element 0 Value 8 Element 2 Value "" The pairs are shown in jumbled order because their order is irrelevant. One advantage of associative arrays is that new pairs can be added at any time. For example, suppose a tenth element is added to the array whose value is `"number ten"'. The result is: Element 10 Value "number ten" Element 3 Value 30 Element 1 Value "foo" Element 0 Value 8 Element 2 Value "" Now the array is "sparse", which just means some indices are missing. It has elements 0-3 and 10, but doesn't have elements 4, 5, 6, 7, 8, or 9. Another consequence of associative arrays is that the indices don't have to be positive integers. Any number, or even a string, can be an index. For example, the following is an array that translates words from English to French: Element "dog" Value "chien" Element "cat" Value "chat" Element "one" Value "un" Element 1 Value "un" Here we decided to translate the number one in both spelled-out and numeric form--thus illustrating that a single array can have both numbers and strings as indices. In fact, array subscripts are always strings; this is discussed in more detail in *Note Numeric Array Subscripts::. Here, the number `1' isn't double-quoted, since `awk' automatically converts it to a string. The value of `IGNORECASE' has no effect upon array subscripting. The identical string value used to store an array element must be used to retrieve it. When `awk' creates an array (e.g., with the `split' built-in function), that array's indices are consecutive integers starting at one. (*Note String Functions::.) `awk''s arrays are efficient--the time to access an element is independent of the number of elements in the array. 7.2 Referring to an Array Element ================================= The principal way to use an array is to refer to one of its elements. An array reference is an expression as follows: ARRAY[INDEX] Here, ARRAY is the name of an array. The expression INDEX is the index of the desired element of the array. The value of the array reference is the current value of that array element. For example, `foo[4.3]' is an expression for the element of array `foo' at index `4.3'. A reference to an array element that has no recorded value yields a value of `""', the null string. This includes elements that have not been assigned any value as well as elements that have been deleted (*note Delete::). Such a reference automatically creates that array element, with the null string as its value. (In some cases, this is unfortunate, because it might waste memory inside `awk'.) To determine whether an element exists in an array at a certain index, use the following expression: INDEX in ARRAY This expression tests whether the particular index exists, without the side effect of creating that element if it is not present. The expression has the value one (true) if `ARRAY[INDEX]' exists and zero (false) if it does not exist. For example, this statement tests whether the array `frequencies' contains the index `2': if (2 in frequencies) print "Subscript 2 is present." Note that this is _not_ a test of whether the array `frequencies' contains an element whose _value_ is two. There is no way to do that except to scan all the elements. Also, this _does not_ create `frequencies[2]', while the following (incorrect) alternative does: if (frequencies[2] != "") print "Subscript 2 is present." 7.3 Assigning Array Elements ============================ Array elements can be assigned values just like `awk' variables: ARRAY[SUBSCRIPT] = VALUE ARRAY is the name of an array. The expression SUBSCRIPT is the index of the element of the array that is assigned a value. The expression VALUE is the value to assign to that element of the array. 7.4 Basic Array Example ======================= The following program takes a list of lines, each beginning with a line number, and prints them out in order of line number. The line numbers are not in order when they are first read--instead they are scrambled. This program sorts the lines by making an array using the line numbers as subscripts. The program then prints out the lines in sorted order of their numbers. It is a very simple program and gets confused upon encountering repeated numbers, gaps, or lines that don't begin with a number: { if ($1 > max) max = $1 arr[$1] = $0 } END { for (x = 1; x <= max; x++) print arr[x] } The first rule keeps track of the largest line number seen so far; it also stores each line into the array `arr', at an index that is the line's number. The second rule runs after all the input has been read, to print out all the lines. When this program is run with the following input: 5 I am the Five man 2 Who are you? The new number two! 4 . . . And four on the floor 1 Who is number one? 3 I three you. Its output is: 1 Who is number one? 2 Who are you? The new number two! 3 I three you. 4 . . . And four on the floor 5 I am the Five man If a line number is repeated, the last line with a given number overrides the others. Gaps in the line numbers can be handled with an easy improvement to the program's `END' rule, as follows: END { for (x = 1; x <= max; x++) if (x in arr) print arr[x] } 7.5 Scanning All Elements of an Array ===================================== In programs that use arrays, it is often necessary to use a loop that executes once for each element of an array. In other languages, where arrays are contiguous and indices are limited to positive integers, this is easy: all the valid indices can be found by counting from the lowest index up to the highest. This technique won't do the job in `awk', because any number or string can be an array index. So `awk' has a special kind of `for' statement for scanning an array: for (VAR in ARRAY) BODY This loop executes BODY once for each index in ARRAY that the program has previously used, with the variable VAR set to that index. The following program uses this form of the `for' statement. The first rule scans the input records and notes which words appear (at least once) in the input, by storing a one into the array `used' with the word as index. The second rule scans the elements of `used' to find all the distinct words that appear in the input. It prints each word that is more than 10 characters long and also prints the number of such words. *Note String Functions::, for more information on the built-in function `length'. # Record a 1 for each word that is used at least once { for (i = 1; i <= NF; i++) used[$i] = 1 } # Find number of distinct words more than 10 characters long END { for (x in used) if (length(x) > 10) { ++num_long_words print x } print num_long_words, "words longer than 10 characters" } *Note Word Sorting::, for a more detailed example of this type. The order in which elements of the array are accessed by this statement is determined by the internal arrangement of the array elements within `awk' and cannot be controlled or changed. This can lead to problems if new elements are added to ARRAY by statements in the loop body; it is not predictable whether the `for' loop will reach them. Similarly, changing VAR inside the loop may produce strange results. It is best to avoid such things. 7.6 The `delete' Statement ========================== To remove an individual element of an array, use the `delete' statement: delete ARRAY[INDEX] Once an array element has been deleted, any value the element once had is no longer available. It is as if the element had never been referred to or had been given a value. The following is an example of deleting elements in an array: for (i in frequencies) delete frequencies[i] This example removes all the elements from the array `frequencies'. Once an element is deleted, a subsequent `for' statement to scan the array does not report that element and the `in' operator to check for the presence of that element returns zero (i.e., false): delete foo[4] if (4 in foo) print "This will never be printed" It is important to note that deleting an element is _not_ the same as assigning it a null value (the empty string, `""'). For example: foo[4] = "" if (4 in foo) print "This is printed, even though foo[4] is empty" It is not an error to delete an element that does not exist. If `--lint' is provided on the command line (*note Options::), `gawk' issues a warning message when an element that is not in the array is deleted. All the elements of an array may be deleted with a single statement by leaving off the subscript in the `delete' statement, as follows: delete ARRAY This ability is a `gawk' extension; it is not available in compatibility mode (*note Options::). Using this version of the `delete' statement is about three times more efficient than the equivalent loop that deletes each element one at a time. The following statement provides a portable but nonobvious way to clear out an array:(1) split("", array) The `split' function (*note String Functions::) clears out the target array first. This call asks it to split apart the null string. Because there is no data to split out, the function simply clears the array and then returns. *Caution:* Deleting an array does not change its type; you cannot delete an array and then use the array's name as a scalar (i.e., a regular variable). For example, the following does not work: a[1] = 3; delete a; a = 3 ---------- Footnotes ---------- (1) Thanks to Michael Brennan for pointing this out. 7.7 Using Numbers to Subscript Arrays ===================================== An important aspect about arrays to remember is that _array subscripts are always strings_. When a numeric value is used as a subscript, it is converted to a string value before being used for subscripting (*note Conversion::). This means that the value of the built-in variable `CONVFMT' can affect how your program accesses elements of an array. For example: xyz = 12.153 data[xyz] = 1 CONVFMT = "%2.2f" if (xyz in data) printf "%s is in data\n", xyz else printf "%s is not in data\n", xyz This prints `12.15 is not in data'. The first statement gives `xyz' a numeric value. Assigning to `data[xyz]' subscripts `data' with the string value `"12.153"' (using the default conversion value of `CONVFMT', `"%.6g"'). Thus, the array element `data["12.153"]' is assigned the value one. The program then changes the value of `CONVFMT'. The test `(xyz in data)' generates a new string value from `xyz'--this time `"12.15"'--because the value of `CONVFMT' only allows two significant digits. This test fails, since `"12.15"' is a different string from `"12.153"'. According to the rules for conversions (*note Conversion::), integer values are always converted to strings as integers, no matter what the value of `CONVFMT' may happen to be. So the usual case of the following works: for (i = 1; i <= maxsub; i++) do something with array[i] The "integer values always convert to strings as integers" rule has an additional consequence for array indexing. Octal and hexadecimal constants (*note Nondecimal-numbers::) are converted internally into numbers, and their original form is forgotten. This means, for example, that `array[17]', `array[021]', and `array[0x11]' all refer to the same element! As with many things in `awk', the majority of the time things work as one would expect them to. But it is useful to have a precise knowledge of the actual rules which sometimes can have a subtle effect on your programs. 7.8 Using Uninitialized Variables as Subscripts =============================================== Suppose it's necessary to write a program to print the input data in reverse order. A reasonable attempt to do so (with some test data) might look like this: $ echo 'line 1 > line 2 > line 3' | awk '{ l[lines] = $0; ++lines } > END { > for (i = lines-1; i >= 0; --i) > print l[i] > }' -| line 3 -| line 2 Unfortunately, the very first line of input data did not come out in the output! At first glance, this program should have worked. The variable `lines' is uninitialized, and uninitialized variables have the numeric value zero. So, `awk' should have printed the value of `l[0]'. The issue here is that subscripts for `awk' arrays are _always_ strings. Uninitialized variables, when used as strings, have the value `""', not zero. Thus, `line 1' ends up stored in `l[""]'. The following version of the program works correctly: { l[lines++] = $0 } END { for (i = lines - 1; i >= 0; --i) print l[i] } Here, the `++' forces `lines' to be numeric, thus making the "old value" numeric zero. This is then converted to `"0"' as the array subscript. Even though it is somewhat unusual, the null string (`""') is a valid array subscript. (d.c.) `gawk' warns about the use of the null string as a subscript if `--lint' is provided on the command line (*note Options::). 7.9 Multidimensional Arrays =========================== A multidimensional array is an array in which an element is identified by a sequence of indices instead of a single index. For example, a two-dimensional array requires two indices. The usual way (in most languages, including `awk') to refer to an element of a two-dimensional array named `grid' is with `grid[X,Y]'. Multidimensional arrays are supported in `awk' through concatenation of indices into one string. `awk' converts the indices into strings (*note Conversion::) and concatenates them together, with a separator between them. This creates a single string that describes the values of the separate indices. The combined string is used as a single index into an ordinary, one-dimensional array. The separator used is the value of the built-in variable `SUBSEP'. For example, suppose we evaluate the expression `foo[5,12] = "value"' when the value of `SUBSEP' is `"@"'. The numbers 5 and 12 are converted to strings and concatenated with an `@' between them, yielding `"5@12"'; thus, the array element `foo["5@12"]' is set to `"value"'. Once the element's value is stored, `awk' has no record of whether it was stored with a single index or a sequence of indices. The two expressions `foo[5,12]' and `foo[5 SUBSEP 12]' are always equivalent. The default value of `SUBSEP' is the string `"\034"', which contains a nonprinting character that is unlikely to appear in an `awk' program or in most input data. The usefulness of choosing an unlikely character comes from the fact that index values that contain a string matching `SUBSEP' can lead to combined strings that are ambiguous. Suppose that `SUBSEP' is `"@"'; then `foo["a@b", "c"]' and `foo["a", "b@c"]' are indistinguishable because both are actually stored as `foo["a@b@c"]'. To test whether a particular index sequence exists in a multidimensional array, use the same operator (`in') that is used for single dimensional arrays. Write the whole sequence of indices in parentheses, separated by commas, as the left operand: (SUBSCRIPT1, SUBSCRIPT2, ...) in ARRAY The following example treats its input as a two-dimensional array of fields; it rotates this array 90 degrees clockwise and prints the result. It assumes that all lines have the same number of elements: { if (max_nf < NF) max_nf = NF max_nr = NR for (x = 1; x <= NF; x++) vector[x, NR] = $x } END { for (x = 1; x <= max_nf; x++) { for (y = max_nr; y >= 1; --y) printf("%s ", vector[x, y]) printf("\n") } } When given the input: 1 2 3 4 5 6 2 3 4 5 6 1 3 4 5 6 1 2 4 5 6 1 2 3 the program produces the following output: 4 3 2 1 5 4 3 2 6 5 4 3 1 6 5 4 2 1 6 5 3 2 1 6 7.10 Scanning Multidimensional Arrays ===================================== There is no special `for' statement for scanning a "multidimensional" array. There cannot be one, because, in truth, there are no multidimensional arrays or elements--there is only a multidimensional _way of accessing_ an array. However, if your program has an array that is always accessed as multidimensional, you can get the effect of scanning it by combining the scanning `for' statement (*note Scanning an Array::) with the built-in `split' function (*note String Functions::). It works in the following manner: for (combined in array) { split(combined, separate, SUBSEP) ... } This sets the variable `combined' to each concatenated combined index in the array, and splits it into the individual indices by breaking it apart where the value of `SUBSEP' appears. The individual indices then become the elements of the array `separate'. Thus, if a value is previously stored in `array[1, "foo"]'; then an element with index `"1\034foo"' exists in `array'. (Recall that the default value of `SUBSEP' is the character with code 034.) Sooner or later, the `for' statement finds that index and does an iteration with the variable `combined' set to `"1\034foo"'. Then the `split' function is called as follows: split("1\034foo", separate, "\034") The result is to set `separate[1]' to `"1"' and `separate[2]' to `"foo"'. Presto! The original sequence of separate indices is recovered. 7.11 Sorting Array Values and Indices with `gawk' ================================================= The order in which an array is scanned with a `for (i in array)' loop is essentially arbitrary. In most `awk' implementations, sorting an array requires writing a `sort' function. While this can be educational for exploring different sorting algorithms, usually that's not the point of the program. `gawk' provides the built-in `asort' and `asorti' functions (*note String Functions::) for sorting arrays. For example: POPULATE THE ARRAY data n = asort(data) for (i = 1; i <= n; i++) DO SOMETHING WITH data[i] After the call to `asort', the array `data' is indexed from 1 to some number N, the total number of elements in `data'. (This count is `asort''s return value.) `data[1]' <= `data[2]' <= `data[3]', and so on. The comparison of array elements is done using `gawk''s usual comparison rules (*note Typing and Comparison::). An important side effect of calling `asort' is that _the array's original indices are irrevocably lost_. As this isn't always desirable, `asort' accepts a second argument: POPULATE THE ARRAY source n = asort(source, dest) for (i = 1; i <= n; i++) DO SOMETHING WITH dest[i] In this case, `gawk' copies the `source' array into the `dest' array and then sorts `dest', destroying its indices. However, the `source' array is not affected. Often, what's needed is to sort on the values of the _indices_ instead of the values of the elements. To do that, starting with `gawk' 3.1.2, use the `asorti' function. The interface is identical to that of `asort', except that the index values are used for sorting, and become the values of the result array: { source[$0] = some_func($0) } END { n = asorti(source, dest) for (i = 1; i <= n; i++) { DO SOMETHING WITH dest[i] Work with sorted indices directly ... DO SOMETHING WITH source[dest[i]] Access original array via sorted indices } } If your version of `gawk' is 3.1.0 or 3.1.1, you don't have `asorti'. Instead, use a helper array to hold the sorted index values, and then access the original array's elements. It works in the following way: POPULATE THE ARRAY data # copy indices j = 1 for (i in data) { ind[j] = i # index value becomes element value j++ } n = asort(ind) # index values are now sorted for (i = 1; i <= n; i++) { DO SOMETHING WITH ind[i] Work with sorted indices directly ... DO SOMETHING WITH data[ind[i]] Access original array via sorted indices } Sorting the array by replacing the indices provides maximal flexibility. To traverse the elements in decreasing order, use a loop that goes from N down to 1, either over the elements or over the indices. Copying array indices and elements isn't expensive in terms of memory. Internally, `gawk' maintains "reference counts" to data. For example, when `asort' copies the first array to the second one, there is only one copy of the original array elements' data, even though both arrays use the values. Similarly, when copying the indices from `data' to `ind', there is only one copy of the actual index strings. We said previously that comparisons are done using `gawk''s "usual comparison rules." Because `IGNORECASE' affects string comparisons, the value of `IGNORECASE' also affects sorting for both `asort' and `asorti'. Caveat Emptor. 8 Functions *********** This major node describes `awk''s built-in functions, which fall into three categories: numeric, string, and I/O. `gawk' provides additional groups of functions to work with values that represent time, do bit manipulation, and internationalize and localize programs. Besides the built-in functions, `awk' has provisions for writing new functions that the rest of a program can use. The second half of this major node describes these "user-defined" functions. 8.1 Built-in Functions ====================== "Built-in" functions are always available for your `awk' program to call. This minor node defines all the built-in functions in `awk'; some of these are mentioned in other sections but are summarized here for your convenience. 8.1.1 Calling Built-in Functions -------------------------------- To call one of `awk''s built-in functions, write the name of the function followed by arguments in parentheses. For example, `atan2(y + z, 1)' is a call to the function `atan2' and has two arguments. Whitespace is ignored between the built-in function name and the open parenthesis, and it is good practice to avoid using whitespace there. User-defined functions do not permit whitespace in this way, and it is easier to avoid mistakes by following a simple convention that always works--no whitespace after a function name. Each built-in function accepts a certain number of arguments. In some cases, arguments can be omitted. The defaults for omitted arguments vary from function to function and are described under the individual functions. In some `awk' implementations, extra arguments given to built-in functions are ignored. However, in `gawk', it is a fatal error to give extra arguments to a built-in function. When a function is called, expressions that create the function's actual parameters are evaluated completely before the call is performed. For example, in the following code fragment: i = 4 j = sqrt(i++) the variable `i' is incremented to the value five before `sqrt' is called with a value of four for its actual parameter. The order of evaluation of the expressions used for the function's parameters is undefined. Thus, avoid writing programs that assume that parameters are evaluated from left to right or from right to left. For example: i = 5 j = atan2(i++, i *= 2) If the order of evaluation is left to right, then `i' first becomes 6, and then 12, and `atan2' is called with the two arguments 6 and 12. But if the order of evaluation is right to left, `i' first becomes 10, then 11, and `atan2' is called with the two arguments 11 and 10. 8.1.2 Numeric Functions ----------------------- The following list describes all of the built-in functions that work with numbers. Optional parameters are enclosed in square brackets ([ ]): `int(X)' This returns the nearest integer to X, located between X and zero and truncated toward zero. For example, `int(3)' is 3, `int(3.9)' is 3, `int(-3.9)' is -3, and `int(-3)' is -3 as well. `sqrt(X)' This returns the positive square root of X. `gawk' reports an error if X is negative. Thus, `sqrt(4)' is 2. `exp(X)' This returns the exponential of X (`e ^ X') or reports an error if X is out of range. The range of values X can have depends on your machine's floating-point representation. `log(X)' This returns the natural logarithm of X, if X is positive; otherwise, it reports an error. `sin(X)' This returns the sine of X, with X in radians. `cos(X)' This returns the cosine of X, with X in radians. `atan2(Y, X)' This returns the arctangent of `Y / X' in radians. `rand()' This returns a random number. The values of `rand' are uniformly distributed between zero and one. The value could be zero but is never one.(1) Often random integers are needed instead. Following is a user-defined function that can be used to obtain a random non-negative integer less than N: function randint(n) { return int(n * rand()) } The multiplication produces a random number greater than zero and less than `n'. Using `int', this result is made into an integer between zero and `n' - 1, inclusive. The following example uses a similar function to produce random integers between one and N. This program prints a new random number for each input record: # Function to roll a simulated die. function roll(n) { return 1 + int(rand() * n) } # Roll 3 six-sided dice and # print total number of points. { printf("%d points\n", roll(6)+roll(6)+roll(6)) } *Caution:* In most `awk' implementations, including `gawk', `rand' starts generating numbers from the same starting number, or "seed", each time you run `awk'. Thus, a program generates the same results each time you run it. The numbers are random within one `awk' run but predictable from run to run. This is convenient for debugging, but if you want a program to do different things each time it is used, you must change the seed to a value that is different in each run. To do this, use `srand'. `srand([X])' The function `srand' sets the starting point, or seed, for generating random numbers to the value X. Each seed value leads to a particular sequence of random numbers.(2) Thus, if the seed is set to the same value a second time, the same sequence of random numbers is produced again. Different `awk' implementations use different random-number generators internally. Don't expect the same `awk' program to produce the same series of random numbers when executed by different versions of `awk'. If the argument X is omitted, as in `srand()', then the current date and time of day are used for a seed. This is the way to get random numbers that are truly unpredictable. The return value of `srand' is the previous seed. This makes it easy to keep track of the seeds in case you need to consistently reproduce sequences of random numbers. ---------- Footnotes ---------- (1) The C version of `rand' is known to produce fairly poor sequences of random numbers. However, nothing requires that an `awk' implementation use the C `rand' to implement the `awk' version of `rand'. In fact, `gawk' uses the BSD `random' function, which is considerably better than `rand', to produce random numbers. (2) Computer-generated random numbers really are not truly random. They are technically known as "pseudorandom." This means that while the numbers in a sequence appear to be random, you can in fact generate the same sequence of random numbers over and over again. 8.1.3 String-Manipulation Functions ----------------------------------- The functions in this minor node look at or change the text of one or more strings. Optional parameters are enclosed in square brackets ([ ]). Those functions that are specific to `gawk' are marked with a pound sign (`#'): `asort(SOURCE [, DEST]) #' `asort' is a `gawk'-specific extension, returning the number of elements in the array SOURCE. The contents of SOURCE are sorted using `gawk''s normal rules for comparing values (in particular, `IGNORECASE' affects the sorting) and the indices of the sorted values of SOURCE are replaced with sequential integers starting with one. If the optional array DEST is specified, then SOURCE is duplicated into DEST. DEST is then sorted, leaving the indices of SOURCE unchanged. For example, if the contents of `a' are as follows: a["last"] = "de" a["first"] = "sac" a["middle"] = "cul" A call to `asort': asort(a) results in the following contents of `a': a[1] = "cul" a[2] = "de" a[3] = "sac" The `asort' function is described in more detail in *Note Array Sorting::. `asort' is a `gawk' extension; it is not available in compatibility mode (*note Options::). `asorti(SOURCE [, DEST]) #' `asorti' is a `gawk'-specific extension, returning the number of elements in the array SOURCE. It works similarly to `asort', however, the _indices_ are sorted, instead of the values. As array indices are always strings, the comparison performed is always a string comparison. (Here too, `IGNORECASE' affects the sorting.) The `asorti' function is described in more detail in *Note Array Sorting::. It was added in `gawk' 3.1.2. `asorti' is a `gawk' extension; it is not available in compatibility mode (*note Options::). `index(IN, FIND)' This searches the string IN for the first occurrence of the string FIND, and returns the position in characters where that occurrence begins in the string IN. Consider the following example: $ awk 'BEGIN { print index("peanut", "an") }' -| 3 If FIND is not found, `index' returns zero. (Remember that string indices in `awk' start at one.) `length([STRING])' This returns the number of characters in STRING. If STRING is a number, the length of the digit string representing that number is returned. For example, `length("abcde")' is 5. By contrast, `length(15 * 35)' works out to 3. In this example, 15 * 35 = 525, and 525 is then converted to the string `"525"', which has three characters. If no argument is supplied, `length' returns the length of `$0'. NOTE: In older versions of `awk', the `length' function could be called without any parentheses. Doing so is marked as "deprecated" in the POSIX standard. This means that while a program can do this, it is a feature that can eventually be removed from a future version of the standard. Therefore, for programs to be maximally portable, always supply the parentheses. `match(STRING, REGEXP [, ARRAY])' The `match' function searches STRING for the longest, leftmost substring matched by the regular expression, REGEXP. It returns the character position, or "index", at which that substring begins (one, if it starts at the beginning of STRING). If no match is found, it returns zero. The REGEXP argument may be either a regexp constant (`/.../') or a string constant ("..."). In the latter case, the string is treated as a regexp to be matched. *Note Computed Regexps::, for a discussion of the difference between the two forms, and the implications for writing your program correctly. The order of the first two arguments is backwards from most other string functions that work with regular expressions, such as `sub' and `gsub'. It might help to remember that for `match', the order is the same as for the `~' operator: `STRING ~ REGEXP'. The `match' function sets the built-in variable `RSTART' to the index. It also sets the built-in variable `RLENGTH' to the length in characters of the matched substring. If no match is found, `RSTART' is set to zero, and `RLENGTH' to -1. For example: { if ($1 == "FIND") regex = $2 else { where = match($0, regex) if (where != 0) print "Match of", regex, "found at", where, "in", $0 } } This program looks for lines that match the regular expression stored in the variable `regex'. This regular expression can be changed. If the first word on a line is `FIND', `regex' is changed to be the second word on that line. Therefore, if given: FIND ru+n My program runs but not very quickly FIND Melvin JF+KM This line is property of Reality Engineering Co. Melvin was here. `awk' prints: Match of ru+n found at 12 in My program runs Match of Melvin found at 1 in Melvin was here. If ARRAY is present, it is cleared, and then the 0th element of ARRAY is set to the entire portion of STRING matched by REGEXP. If REGEXP contains parentheses, the integer-indexed elements of ARRAY are set to contain the portion of STRING matching the corresponding parenthesized subexpression. For example: $ echo foooobazbarrrrr | > gawk '{ match($0, /(fo+).+(bar*)/, arr) > print arr[1], arr[2] }' -| foooo barrrrr In addition, beginning with `gawk' 3.1.2, multidimensional subscripts are available providing the start index and length of each matched subexpression: $ echo foooobazbarrrrr | > gawk '{ match($0, /(fo+).+(bar*)/, arr) > print arr[1], arr[2] > print arr[1, "start"], arr[1, "length"] > print arr[2, "start"], arr[2, "length"] > }' -| foooo barrrrr -| 1 5 -| 9 7 There may not be subscripts for the start and index for every parenthesized subexpressions, since they may not all have matched text; thus they should be tested for with the `in' operator (*note Reference to Elements::). The ARRAY argument to `match' is a `gawk' extension. In compatibility mode (*note Options::), using a third argument is a fatal error. `split(STRING, ARRAY [, FIELDSEP])' This function divides STRING into pieces separated by FIELDSEP and stores the pieces in ARRAY. The first piece is stored in `ARRAY[1]', the second piece in `ARRAY[2]', and so forth. The string value of the third argument, FIELDSEP, is a regexp describing where to split STRING (much as `FS' can be a regexp describing where to split input records). If FIELDSEP is omitted, the value of `FS' is used. `split' returns the number of elements created. The `split' function splits strings into pieces in a manner similar to the way input lines are split into fields. For example: split("cul-de-sac", a, "-") splits the string `cul-de-sac' into three fields using `-' as the separator. It sets the contents of the array `a' as follows: a[1] = "cul" a[2] = "de" a[3] = "sac" The value returned by this call to `split' is three. As with input field-splitting, when the value of FIELDSEP is `" "', leading and trailing whitespace is ignored, and the elements are separated by runs of whitespace. Also as with input field-splitting, if FIELDSEP is the null string, each individual character in the string is split into its own array element. (This is a `gawk'-specific extension.) Note, however, that `RS' has no effect on the way `split' works. Even though `RS = ""' causes newline to also be an input field separator, this does not affect how `split' splits strings. Modern implementations of `awk', including `gawk', allow the third argument to be a regexp constant (`/abc/') as well as a string. (d.c.) The POSIX standard allows this as well. *Note Computed Regexps::, for a discussion of the difference between using a string constant or a regexp constant, and the implications for writing your program correctly. Before splitting the string, `split' deletes any previously existing elements in the array ARRAY. If STRING is null, the array has no elements. (So this is a portable way to delete an entire array with one statement. *Note Delete::.) If STRING does not match FIELDSEP at all (but is not null), ARRAY has one element only. The value of that element is the original STRING. `sprintf(FORMAT, EXPRESSION1, ...)' This returns (without printing) the string that `printf' would have printed out with the same arguments (*note Printf::). For example: pival = sprintf("pi = %.2f (approx.)", 22/7) assigns the string `"pi = 3.14 (approx.)"' to the variable `pival'. `strtonum(STR) #' Examines STR and returns its numeric value. If STR begins with a leading `0', `strtonum' assumes that STR is an octal number. If STR begins with a leading `0x' or `0X', `strtonum' assumes that STR is a hexadecimal number. For example: $ echo 0x11 | > gawk '{ printf "%d\n", strtonum($1) }' -| 17 Using the `strtonum' function is _not_ the same as adding zero to a string value; the automatic coercion of strings to numbers works only for decimal data, not for octal or hexadecimal.(1) `strtonum' is a `gawk' extension; it is not available in compatibility mode (*note Options::). `sub(REGEXP, REPLACEMENT [, TARGET])' The `sub' function alters the value of TARGET. It searches this value, which is treated as a string, for the leftmost, longest substring matched by the regular expression REGEXP. Then the entire string is changed by replacing the matched text with REPLACEMENT. The modified string becomes the new value of TARGET. The REGEXP argument may be either a regexp constant (`/.../') or a string constant ("..."). In the latter case, the string is treated as a regexp to be matched. *Note Computed Regexps::, for a discussion of the difference between the two forms, and the implications for writing your program correctly. This function is peculiar because TARGET is not simply used to compute a value, and not just any expression will do--it must be a variable, field, or array element so that `sub' can store a modified value there. If this argument is omitted, then the default is to use and alter `$0'.(2) For example: str = "water, water, everywhere" sub(/at/, "ith", str) sets `str' to `"wither, water, everywhere"', by replacing the leftmost longest occurrence of `at' with `ith'. The `sub' function returns the number of substitutions made (either one or zero). If the special character `&' appears in REPLACEMENT, it stands for the precise substring that was matched by REGEXP. (If the regexp can match more than one string, then this precise substring may vary.) For example: { sub(/candidate/, "& and his wife"); print } changes the first occurrence of `candidate' to `candidate and his wife' on each input line. Here is another example: $ awk 'BEGIN { > str = "daabaaa" > sub(/a+/, "C&C", str) > print str > }' -| dCaaCbaaa This shows how `&' can represent a nonconstant string and also illustrates the "leftmost, longest" rule in regexp matching (*note Leftmost Longest::). The effect of this special character (`&') can be turned off by putting a backslash before it in the string. As usual, to insert one backslash in the string, you must write two backslashes. Therefore, write `\\&' in a string constant to include a literal `&' in the replacement. For example, the following shows how to replace the first `|' on each line with an `&': { sub(/\|/, "\\&"); print } As mentioned, the third argument to `sub' must be a variable, field or array reference. Some versions of `awk' allow the third argument to be an expression that is not an lvalue. In such a case, `sub' still searches for the pattern and returns zero or one, but the result of the substitution (if any) is thrown away because there is no place to put it. Such versions of `awk' accept expressions such as the following: sub(/USA/, "United States", "the USA and Canada") For historical compatibility, `gawk' accepts erroneous code, such as in the previous example. However, using any other nonchangeable object as the third parameter causes a fatal error and your program will not run. Finally, if the REGEXP is not a regexp constant, it is converted into a string, and then the value of that string is treated as the regexp to match. `gsub(REGEXP, REPLACEMENT [, TARGET])' This is similar to the `sub' function, except `gsub' replaces _all_ of the longest, leftmost, _nonoverlapping_ matching substrings it can find. The `g' in `gsub' stands for "global," which means replace everywhere. For example: { gsub(/Britain/, "United Kingdom"); print } replaces all occurrences of the string `Britain' with `United Kingdom' for all input records. The `gsub' function returns the number of substitutions made. If the variable to search and alter (TARGET) is omitted, then the entire input record (`$0') is used. As in `sub', the characters `&' and `\' are special, and the third argument must be assignable. `gensub(REGEXP, REPLACEMENT, HOW [, TARGET]) #' `gensub' is a general substitution function. Like `sub' and `gsub', it searches the target string TARGET for matches of the regular expression REGEXP. Unlike `sub' and `gsub', the modified string is returned as the result of the function and the original target string is _not_ changed. If HOW is a string beginning with `g' or `G', then it replaces all matches of REGEXP with REPLACEMENT. Otherwise, HOW is treated as a number that indicates which match of REGEXP to replace. If no TARGET is supplied, `$0' is used. `gensub' provides an additional feature that is not available in `sub' or `gsub': the ability to specify components of a regexp in the replacement text. This is done by using parentheses in the regexp to mark the components and then specifying `\N' in the replacement text, where N is a digit from 1 to 9. For example: $ gawk ' > BEGIN { > a = "abc def" > b = gensub(/(.+) (.+)/, "\\2 \\1", "g", a) > print b > }' -| def abc As with `sub', you must type two backslashes in order to get one into the string. In the replacement text, the sequence `\0' represents the entire matched text, as does the character `&'. The following example shows how you can use the third argument to control which match of the regexp should be changed: $ echo a b c a b c | > gawk '{ print gensub(/a/, "AA", 2) }' -| a b c AA b c In this case, `$0' is used as the default target string. `gensub' returns the new string as its result, which is passed directly to `print' for printing. If the HOW argument is a string that does not begin with `g' or `G', or if it is a number that is less than or equal to zero, only one substitution is performed. If HOW is zero, `gawk' issues a warning message. If REGEXP does not match TARGET, `gensub''s return value is the original unchanged value of TARGET. `gensub' is a `gawk' extension; it is not available in compatibility mode (*note Options::). `substr(STRING, START [, LENGTH])' This returns a LENGTH-character-long substring of STRING, starting at character number START. The first character of a string is character number one.(3) For example, `substr("washington", 5, 3)' returns `"ing"'. If LENGTH is not present, this function returns the whole suffix of STRING that begins at character number START. For example, `substr("washington", 5)' returns `"ington"'. The whole suffix is also returned if LENGTH is greater than the number of characters remaining in the string, counting from character START. If START is less than one, `substr' treats it as if it was one. (POSIX doesn't specify what to do in this case: Unix `awk' acts this way, and therefore `gawk' does too.) If START is greater than the number of characters in the string, `substr' returns the null string. Similarly, if LENGTH is present but less than or equal to zero, the null string is returned. The string returned by `substr' _cannot_ be assigned. Thus, it is a mistake to attempt to change a portion of a string, as shown in the following example: string = "abcdef" # try to get "abCDEf", won't work substr(string, 3, 3) = "CDE" It is also a mistake to use `substr' as the third argument of `sub' or `gsub': gsub(/xyz/, "pdq", substr($0, 5, 20)) # WRONG (Some commercial versions of `awk' do in fact let you use `substr' this way, but doing so is not portable.) If you need to replace bits and pieces of a string, combine `substr' with string concatenation, in the following manner: string = "abcdef" ... string = substr(string, 1, 2) "CDE" substr(string, 6) `tolower(STRING)' This returns a copy of STRING, with each uppercase character in the string replaced with its corresponding lowercase character. Nonalphabetic characters are left unchanged. For example, `tolower("MiXeD cAsE 123")' returns `"mixed case 123"'. `toupper(STRING)' This returns a copy of STRING, with each lowercase character in the string replaced with its corresponding uppercase character. Nonalphabetic characters are left unchanged. For example, `toupper("MiXeD cAsE 123")' returns `"MIXED CASE 123"'. ---------- Footnotes ---------- (1) Unless you use the `--non-decimal-data' option, which isn't recommended. *Note Nondecimal Data::, for more information. (2) Note that this means that the record will first be regenerated using the value of `OFS' if any fields have been changed, and that the fields will be updated after the substituion, even if the operation is a "no-op" such as `sub(/^/, "")'. (3) This is different from C and C++, in which the first character is number zero. 8.1.3.1 More About `\' and `&' with `sub', `gsub', and `gensub' ............................................................... When using `sub', `gsub', or `gensub', and trying to get literal backslashes and ampersands into the replacement text, you need to remember that there are several levels of "escape processing" going on. First, there is the "lexical" level, which is when `awk' reads your program and builds an internal copy of it that can be executed. Then there is the runtime level, which is when `awk' actually scans the replacement string to determine what to generate. At both levels, `awk' looks for a defined set of characters that can come after a backslash. At the lexical level, it looks for the escape sequences listed in *Note Escape Sequences::. Thus, for every `\' that `awk' processes at the runtime level, type two backslashes at the lexical level. When a character that is not valid for an escape sequence follows the `\', Unix `awk' and `gawk' both simply remove the initial `\' and put the next character into the string. Thus, for example, `"a\qb"' is treated as `"aqb"'. At the runtime level, the various functions handle sequences of `\' and `&' differently. The situation is (sadly) somewhat complex. Historically, the `sub' and `gsub' functions treated the two character sequence `\&' specially; this sequence was replaced in the generated text with a single `&'. Any other `\' within the REPLACEMENT string that did not precede an `&' was passed through unchanged. This is illustrated in *Note table-sub-escapes::. You type `sub' sees `sub' generates ------- --------- -------------- `\&' `&' the matched text `\\&' `\&' a literal `&' `\\\&' `\&' a literal `&' `\\\\&' `\\&' a literal `\&' `\\\\\&' `\\&' a literal `\&' `\\\\\\&' `\\\&' a literal `\\&' `\\q' `\q' a literal `\q' Table 8.1: Historical Escape Sequence Processing for sub and gsub This table shows both the lexical-level processing, where an odd number of backslashes becomes an even number at the runtime level, as well as the runtime processing done by `sub'. (For the sake of simplicity, the rest of the following tables only show the case of even numbers of backslashes entered at the lexical level.) The problem with the historical approach is that there is no way to get a literal `\' followed by the matched text. The 1992 POSIX standard attempted to fix this problem. That standard says that `sub' and `gsub' look for either a `\' or an `&' after the `\'. If either one follows a `\', that character is output literally. The interpretation of `\' and `&' then becomes as shown in *Note table-sub-posix-92::. You type `sub' sees `sub' generates ------- --------- -------------- `&' `&' the matched text `\\&' `\&' a literal `&' `\\\\&' `\\&' a literal `\', then the matched text `\\\\\\&' `\\\&' a literal `\&' Table 8.2: 1992 POSIX Rules for sub and gsub Escape Sequence Processing This appears to solve the problem. Unfortunately, the phrasing of the standard is unusual. It says, in effect, that `\' turns off the special meaning of any following character, but for anything other than `\' and `&', such special meaning is undefined. This wording leads to two problems: * Backslashes must now be doubled in the REPLACEMENT string, breaking historical `awk' programs. * To make sure that an `awk' program is portable, _every_ character in the REPLACEMENT string must be preceded with a backslash.(1) Because of the problems just listed, in 1996, the `gawk' maintainer submitted proposed text for a revised standard that reverts to rules that correspond more closely to the original existing practice. The proposed rules have special cases that make it possible to produce a `\' preceding the matched text. This is shown in *Note table-sub-proposed::. You type `sub' sees `sub' generates ------- --------- -------------- `\\\\\\&' `\\\&' a literal `\&' `\\\\&' `\\&' a literal `\', followed by the matched text `\\&' `\&' a literal `&' `\\q' `\q' a literal `\q' `\\\\' `\\' `\\' Table 8.3: Propsosed rules for sub and backslash In a nutshell, at the runtime level, there are now three special sequences of characters (`\\\&', `\\&' and `\&') whereas historically there was only one. However, as in the historical case, any `\' that is not part of one of these three sequences is not special and appears in the output literally. `gawk' 3.0 and 3.1 follow these proposed POSIX rules for `sub' and `gsub'. The POSIX standard took much longer to be revised than was expected in 1996. The 2001 standard does not follow the above rules. Instead, the rules there are somewhat simpler. The results are similar except for one case. The 2001 POSIX rules state that `\&' in the replacement string produces a literal `&', `\\' produces a literal `\', and `\' followed by anything else is not special; the `\' is placed straight into the output. These rules are presented in *Note table-posix-2001-sub::. You type `sub' sees `sub' generates ------- --------- -------------- `\\\\\\&' `\\\&' a literal `\&' `\\\\&' `\\&' a literal `\', followed by the matched text `\\&' `\&' a literal `&' `\\q' `\q' a literal `\q' `\\\\' `\\' `\' Table 8.4: POSIX 2001 rules for sub The only case where the difference is noticeable is the last one: `\\\\' is seen as `\\' and produces `\' instead of `\\'. Starting with version 3.1.4, `gawk' follows the POSIX rules when `--posix' is specified (*note Options::). Otherwise, it continues to follow the 1996 proposed rules, since, as of this writing, that has been its behavior for over seven years. NOTE: At the next major release, `gawk' will switch to using the POSIX 2001 rules by default. The rules for `gensub' are considerably simpler. At the runtime level, whenever `gawk' sees a `\', if the following character is a digit, then the text that matched the corresponding parenthesized subexpression is placed in the generated output. Otherwise, no matter what character follows the `\', it appears in the generated text and the `\' does not, as shown in *Note table-gensub-escapes::. You type `gensub' sees `gensub' generates ------- ------------ ----------------- `&' `&' the matched text `\\&' `\&' a literal `&' `\\\\' `\\' a literal `\' `\\\\&' `\\&' a literal `\', then the matched text `\\\\\\&' `\\\&' a literal `\&' `\\q' `\q' a literal `q' Table 8.5: Escape Sequence Processing for gensub Because of the complexity of the lexical and runtime level processing and the special cases for `sub' and `gsub', we recommend the use of `gawk' and `gensub' when you have to do substitutions. Advanced Notes: Matching the Null String ---------------------------------------- In `awk', the `*' operator can match the null string. This is particularly important for the `sub', `gsub', and `gensub' functions. For example: $ echo abc | awk '{ gsub(/m*/, "X"); print }' -| XaXbXcX Although this makes a certain amount of sense, it can be surprising. ---------- Footnotes ---------- (1) This consequence was certainly unintended. 8.1.4 Input/Output Functions ---------------------------- The following functions relate to input/output (I/O). Optional parameters are enclosed in square brackets ([ ]): `close(FILENAME [, HOW])' Close the file FILENAME for input or output. Alternatively, the argument may be a shell command that was used for creating a coprocess, or for redirecting to or from a pipe; then the coprocess or pipe is closed. *Note Close Files And Pipes::, for more information. When closing a coprocess, it is occasionally useful to first close one end of the two-way pipe and then to close the other. This is done by providing a second argument to `close'. This second argument should be one of the two string values `"to"' or `"from"', indicating which end of the pipe to close. Case in the string does not matter. *Note Two-way I/O::, which discusses this feature in more detail and gives an example. `fflush([FILENAME])' Flush any buffered output associated with FILENAME, which is either a file opened for writing or a shell command for redirecting output to a pipe or coprocess. Many utility programs "buffer" their output; i.e., they save information to write to a disk file or terminal in memory until there is enough for it to be worthwhile to send the data to the output device. This is often more efficient than writing every little bit of information as soon as it is ready. However, sometimes it is necessary to force a program to "flush" its buffers; that is, write the information to its destination, even if a buffer is not full. This is the purpose of the `fflush' function--`gawk' also buffers its output and the `fflush' function forces `gawk' to flush its buffers. `fflush' was added to the Bell Laboratories research version of `awk' in 1994; it is not part of the POSIX standard and is not available if `--posix' has been specified on the command line (*note Options::). `gawk' extends the `fflush' function in two ways. The first is to allow no argument at all. In this case, the buffer for the standard output is flushed. The second is to allow the null string (`""') as the argument. In this case, the buffers for _all_ open output files and pipes are flushed. `fflush' returns zero if the buffer is successfully flushed; otherwise, it returns -1. In the case where all buffers are flushed, the return value is zero only if all buffers were flushed successfully. Otherwise, it is -1, and `gawk' warns about the problem FILENAME. `gawk' also issues a warning message if you attempt to flush a file or pipe that was opened for reading (such as with `getline'), or if FILENAME is not an open file, pipe, or coprocess. In such a case, `fflush' returns -1, as well. `system(COMMAND)' Executes operating-system commands and then returns to the `awk' program. The `system' function executes the command given by the string COMMAND. It returns the status returned by the command that was executed as its value. For example, if the following fragment of code is put in your `awk' program: END { system("date | mail -s 'awk run done' root") } the system administrator is sent mail when the `awk' program finishes processing input and begins its end-of-input processing. Note that redirecting `print' or `printf' into a pipe is often enough to accomplish your task. If you need to run many commands, it is more efficient to simply print them down a pipeline to the shell: while (MORE STUFF TO DO) print COMMAND | "/bin/sh" close("/bin/sh") However, if your `awk' program is interactive, `system' is useful for cranking up large self-contained programs, such as a shell or an editor. Some operating systems cannot implement the `system' function. `system' causes a fatal error if it is not supported. Advanced Notes: Interactive Versus Noninteractive Buffering ----------------------------------------------------------- As a side point, buffering issues can be even more confusing, depending upon whether your program is "interactive", i.e., communicating with a user sitting at a keyboard.(1) Interactive programs generally "line buffer" their output; i.e., they write out every line. Noninteractive programs wait until they have a full buffer, which may be many lines of output. Here is an example of the difference: $ awk '{ print $1 + $2 }' 1 1 -| 2 2 3 -| 5 Ctrl-d Each line of output is printed immediately. Compare that behavior with this example: $ awk '{ print $1 + $2 }' | cat 1 1 2 3 Ctrl-d -| 2 -| 5 Here, no output is printed until after the `Ctrl-d' is typed, because it is all buffered and sent down the pipe to `cat' in one shot. Advanced Notes: Controlling Output Buffering with `system' ---------------------------------------------------------- The `fflush' function provides explicit control over output buffering for individual files and pipes. However, its use is not portable to many other `awk' implementations. An alternative method to flush output buffers is to call `system' with a null string as its argument: system("") # flush output `gawk' treats this use of the `system' function as a special case and is smart enough not to run a shell (or other command interpreter) with the empty command. Therefore, with `gawk', this idiom is not only useful, it is also efficient. While this method should work with other `awk' implementations, it does not necessarily avoid starting an unnecessary shell. (Other implementations may only flush the buffer associated with the standard output and not necessarily all buffered output.) If you think about what a programmer expects, it makes sense that `system' should flush any pending output. The following program: BEGIN { print "first print" system("echo system echo") print "second print" } must print: first print system echo second print and not: system echo first print second print If `awk' did not flush its buffers before calling `system', you would see the latter (undesirable) output. ---------- Footnotes ---------- (1) A program is interactive if the standard output is connected to a terminal device. 8.1.5 Using `gawk''s Timestamp Functions ---------------------------------------- `awk' programs are commonly used to process log files containing timestamp information, indicating when a particular log record was written. Many programs log their timestamp in the form returned by the `time' system call, which is the number of seconds since a particular epoch. On POSIX-compliant systems, it is the number of seconds since 1970-01-01 00:00:00 UTC, not counting leap seconds.(1) All known POSIX-compliant systems support timestamps from 0 through 2^31 - 1, which is sufficient to represent times through 2038-01-19 03:14:07 UTC. Many systems support a wider range of timestamps, including negative timestamps that represent times before the epoch. In order to make it easier to process such log files and to produce useful reports, `gawk' provides the following functions for working with timestamps. They are `gawk' extensions; they are not specified in the POSIX standard, nor are they in any other known version of `awk'.(2) Optional parameters are enclosed in square brackets ([ ]): `systime()' This function returns the current time as the number of seconds since the system epoch. On POSIX systems, this is the number of seconds since 1970-01-01 00:00:00 UTC, not counting leap seconds. It may be a different number on other systems. `mktime(DATESPEC)' This function turns DATESPEC into a timestamp in the same form as is returned by `systime'. It is similar to the function of the same name in ISO C. The argument, DATESPEC, is a string of the form `"YYYY MM DD HH MM SS [DST]"'. The string consists of six or seven numbers representing, respectively, the full year including century, the month from 1 to 12, the day of the month from 1 to 31, the hour of the day from 0 to 23, the minute from 0 to 59, the second from 0 to 60,(3) and an optional daylight-savings flag. The values of these numbers need not be within the ranges specified; for example, an hour of -1 means 1 hour before midnight. The origin-zero Gregorian calendar is assumed, with year 0 preceding year 1 and year -1 preceding year 0. The time is assumed to be in the local timezone. If the daylight-savings flag is positive, the time is assumed to be daylight savings time; if zero, the time is assumed to be standard time; and if negative (the default), `mktime' attempts to determine whether daylight savings time is in effect for the specified time. If DATESPEC does not contain enough elements or if the resulting time is out of range, `mktime' returns -1. `strftime([FORMAT [, TIMESTAMP]])' This function returns a string. It is similar to the function of the same name in ISO C. The time specified by TIMESTAMP is used to produce a string, based on the contents of the FORMAT string. The TIMESTAMP is in the same format as the value returned by the `systime' function. If no TIMESTAMP argument is supplied, `gawk' uses the current time of day as the timestamp. If no FORMAT argument is supplied, `strftime' uses `"%a %b %d %H:%M:%S %Z %Y"'. This format string produces output that is (almost) equivalent to that of the `date' utility. (Versions of `gawk' prior to 3.0 require the FORMAT argument.) The `systime' function allows you to compare a timestamp from a log file with the current time of day. In particular, it is easy to determine how long ago a particular record was logged. It also allows you to produce log records using the "seconds since the epoch" format. The `mktime' function allows you to convert a textual representation of a date and time into a timestamp. This makes it easy to do before/after comparisons of dates and times, particularly when dealing with date and time data coming from an external source, such as a log file. The `strftime' function allows you to easily turn a timestamp into human-readable information. It is similar in nature to the `sprintf' function (*note String Functions::), in that it copies nonformat specification characters verbatim to the returned string, while substituting date and time values for format specifications in the FORMAT string. `strftime' is guaranteed by the 1999 ISO C standard(4) to support the following date format specifications: `%a' The locale's abbreviated weekday name. `%A' The locale's full weekday name. `%b' The locale's abbreviated month name. `%B' The locale's full month name. `%c' The locale's "appropriate" date and time representation. (This is `%A %B %d %T %Y' in the `"C"' locale.) `%C' The century. This is the year divided by 100 and truncated to the next lower integer. `%d' The day of the month as a decimal number (01-31). `%D' Equivalent to specifying `%m/%d/%y'. `%e' The day of the month, padded with a space if it is only one digit. `%F' Equivalent to specifying `%Y-%m-%d'. This is the ISO 8601 date format. `%g' The year modulo 100 of the ISO week number, as a decimal number (00-99). For example, January 1, 1993 is in week 53 of 1992. Thus, the year of its ISO week number is 1992, even though its year is 1993. Similarly, December 31, 1973 is in week 1 of 1974. Thus, the year of its ISO week number is 1974, even though its year is 1973. `%G' The full year of the ISO week number, as a decimal number. `%h' Equivalent to `%b'. `%H' The hour (24-hour clock) as a decimal number (00-23). `%I' The hour (12-hour clock) as a decimal number (01-12). `%j' The day of the year as a decimal number (001-366). `%m' The month as a decimal number (01-12). `%M' The minute as a decimal number (00-59). `%n' A newline character (ASCII LF). `%p' The locale's equivalent of the AM/PM designations associated with a 12-hour clock. `%r' The locale's 12-hour clock time. (This is `%I:%M:%S %p' in the `"C"' locale.) `%R' Equivalent to specifying `%H:%M'. `%S' The second as a decimal number (00-60). `%t' A TAB character. `%T' Equivalent to specifying `%H:%M:%S'. `%u' The weekday as a decimal number (1-7). Monday is day one. `%U' The week number of the year (the first Sunday as the first day of week one) as a decimal number (00-53). `%V' The week number of the year (the first Monday as the first day of week one) as a decimal number (01-53). The method for determining the week number is as specified by ISO 8601. (To wit: if the week containing January 1 has four or more days in the new year, then it is week one; otherwise it is week 53 of the previous year and the next week is week one.) `%w' The weekday as a decimal number (0-6). Sunday is day zero. `%W' The week number of the year (the first Monday as the first day of week one) as a decimal number (00-53). `%x' The locale's "appropriate" date representation. (This is `%A %B %d %Y' in the `"C"' locale.) `%X' The locale's "appropriate" time representation. (This is `%T' in the `"C"' locale.) `%y' The year modulo 100 as a decimal number (00-99). `%Y' The full year as a decimal number (e.g., 1995). `%z' The timezone offset in a +HHMM format (e.g., the format necessary to produce RFC 822/RFC 1036 date headers). `%Z' The time zone name or abbreviation; no characters if no time zone is determinable. `%Ec %EC %Ex %EX %Ey %EY %Od %Oe %OH' `%OI %Om %OM %OS %Ou %OU %OV %Ow %OW %Oy' "Alternate representations" for the specifications that use only the second letter (`%c', `%C', and so on).(5) (These facilitate compliance with the POSIX `date' utility.) `%%' A literal `%'. If a conversion specifier is not one of the above, the behavior is undefined.(6) Informally, a "locale" is the geographic place in which a program is meant to run. For example, a common way to abbreviate the date September 4, 1991 in the United States is "9/4/91." In many countries in Europe, however, it is abbreviated "4.9.91." Thus, the `%x' specification in a `"US"' locale might produce `9/4/91', while in a `"EUROPE"' locale, it might produce `4.9.91'. The ISO C standard defines a default `"C"' locale, which is an environment that is typical of what most C programmers are used to. A public-domain C version of `strftime' is supplied with `gawk' for systems that are not yet fully standards-compliant. It supports all of the just listed format specifications. If that version is used to compile `gawk' (*note Installation::), then the following additional format specifications are available: `%k' The hour (24-hour clock) as a decimal number (0-23). Single-digit numbers are padded with a space. `%l' The hour (12-hour clock) as a decimal number (1-12). Single-digit numbers are padded with a space. `%N' The "Emperor/Era" name. Equivalent to `%C'. `%o' The "Emperor/Era" year. Equivalent to `%y'. `%s' The time as a decimal timestamp in seconds since the epoch. `%v' The date in VMS format (e.g., `20-JUN-1991'). Additionally, the alternate representations are recognized but their normal representations are used. This example is an `awk' implementation of the POSIX `date' utility. Normally, the `date' utility prints the current date and time of day in a well-known format. However, if you provide an argument to it that begins with a `+', `date' copies nonformat specifier characters to the standard output and interprets the current time according to the format specifiers in the string. For example: $ date '+Today is %A, %B %d, %Y.' -| Today is Thursday, September 14, 2000. Here is the `gawk' version of the `date' utility. It has a shell "wrapper" to handle the `-u' option, which requires that `date' run as if the time zone is set to UTC: #! /bin/sh # # date --- approximate the P1003.2 'date' command case $1 in -u) TZ=UTC0 # use UTC export TZ shift ;; esac gawk 'BEGIN { format = "%a %b %d %H:%M:%S %Z %Y" exitval = 0 if (ARGC > 2) exitval = 1 else if (ARGC == 2) { format = ARGV[1] if (format ~ /^\+/) format = substr(format, 2) # remove leading + } print strftime(format) exit exitval }' "$@" ---------- Footnotes ---------- (1) *Note Glossary::, especially the entries "Epoch" and "UTC." (2) The GNU `date' utility can also do many of the things described here. Its use may be preferable for simple time-related operations in shell scripts. (3) Occasionally there are minutes in a year with a leap second, which is why the seconds can go up to 60. (4) As this is a recent standard, not every system's `strftime' necessarily supports all of the conversions listed here. (5) If you don't understand any of this, don't worry about it; these facilities are meant to make it easier to "internationalize" programs. Other internationalization features are described in *Note Internationalization::. (6) This is because ISO C leaves the behavior of the C version of `strftime' undefined and `gawk' uses the system's version of `strftime' if it's there. Typically, the conversion specifier either does not appear in the returned string or appears literally. 8.1.6 Bit-Manipulation Functions of `gawk' ------------------------------------------ I can explain it for you, but I can't understand it for you. Anonymous Many languages provide the ability to perform "bitwise" operations on two integer numbers. In other words, the operation is performed on each successive pair of bits in the operands. Three common operations are bitwise AND, OR, and XOR. The operations are described in *Note table-bitwise-ops::. Bit Operator | AND | OR | XOR |--+--+--+--+--+-- Operands | 0 | 1 | 0 | 1 | 0 | 1 ---------+--+--+--+--+--+-- 0 | 0 0 | 0 1 | 0 1 1 | 0 1 | 1 1 | 1 0 Table 8.6: Bitwise Operations As you can see, the result of an AND operation is 1 only when _both_ bits are 1. The result of an OR operation is 1 if _either_ bit is 1. The result of an XOR operation is 1 if either bit is 1, but not both. The next operation is the "complement"; the complement of 1 is 0 and the complement of 0 is 1. Thus, this operation "flips" all the bits of a given value. Finally, two other common operations are to shift the bits left or right. For example, if you have a bit string `10111001' and you shift it right by three bits, you end up with `00010111'.(1) If you start over again with `10111001' and shift it left by three bits, you end up with `11001000'. `gawk' provides built-in functions that implement the bitwise operations just described. They are: `and(V1, V2)' Returns the bitwise AND of the values provided by V1 and V2. `or(V1, V2)' Returns the bitwise OR of the values provided by V1 and V2. `xor(V1, V2)' Returns the bitwise XOR of the values provided by V1 and V2. `compl(VAL)' Returns the bitwise complement of VAL. `lshift(VAL, COUNT)' Returns the value of VAL, shifted left by COUNT bits. `rshift(VAL, COUNT)' Returns the value of VAL, shifted right by COUNT bits. For all of these functions, first the double-precision floating-point value is converted to the widest C unsigned integer type, then the bitwise operation is performed. If the result cannot be represented exactly as a C `double', leading nonzero bits are removed one by one until it can be represented exactly. The result is then converted back into a C `double'. (If you don't understand this paragraph, don't worry about it.) Here is a user-defined function (*note User-defined::) that illustrates the use of these functions: # bits2str --- turn a byte into readable 1's and 0's function bits2str(bits, data, mask) { if (bits == 0) return "0" mask = 1 for (; bits != 0; bits = rshift(bits, 1)) data = (and(bits, mask) ? "1" : "0") data while ((length(data) % 8) != 0) data = "0" data return data } BEGIN { printf "123 = %s\n", bits2str(123) printf "0123 = %s\n", bits2str(0123) printf "0x99 = %s\n", bits2str(0x99) comp = compl(0x99) printf "compl(0x99) = %#x = %s\n", comp, bits2str(comp) shift = lshift(0x99, 2) printf "lshift(0x99, 2) = %#x = %s\n", shift, bits2str(shift) shift = rshift(0x99, 2) printf "rshift(0x99, 2) = %#x = %s\n", shift, bits2str(shift) } This program produces the following output when run: $ gawk -f testbits.awk -| 123 = 01111011 -| 0123 = 01010011 -| 0x99 = 10011001 -| compl(0x99) = 0xffffff66 = 11111111111111111111111101100110 -| lshift(0x99, 2) = 0x264 = 0000001001100100 -| rshift(0x99, 2) = 0x26 = 00100110 The `bits2str' function turns a binary number into a string. The number `1' represents a binary value where the rightmost bit is set to 1. Using this mask, the function repeatedly checks the rightmost bit. ANDing the mask with the value indicates whether the rightmost bit is 1 or not. If so, a `"1"' is concatenated onto the front of the string. Otherwise, a `"0"' is added. The value is then shifted right by one bit and the loop continues until there are no more 1 bits. If the initial value is zero it returns a simple `"0"'. Otherwise, at the end, it pads the value with zeros to represent multiples of 8-bit quantities. This is typical in modern computers. The main code in the `BEGIN' rule shows the difference between the decimal and octal values for the same numbers (*note Nondecimal-numbers::), and then demonstrates the results of the `compl', `lshift', and `rshift' functions. ---------- Footnotes ---------- (1) This example shows that 0's come in on the left side. For `gawk', this is always true, but in some languages, it's possible to have the left side fill with 1's. Caveat emptor. 8.1.7 Using `gawk''s String-Translation Functions ------------------------------------------------- `gawk' provides facilities for internationalizing `awk' programs. These include the functions described in the following list. The descriptions here are purposely brief. *Note Internationalization::, for the full story. Optional parameters are enclosed in square brackets ([ ]): `dcgettext(STRING [, DOMAIN [, CATEGORY]])' This function returns the translation of STRING in text domain DOMAIN for locale category CATEGORY. The default value for DOMAIN is the current value of `TEXTDOMAIN'. The default value for CATEGORY is `"LC_MESSAGES"'. `dcngettext(STRING1, STRING2, NUMBER [, DOMAIN [, CATEGORY]])' This function returns the plural form used for NUMBER of the translation of STRING1 and STRING2 in text domain DOMAIN for locale category CATEGORY. STRING1 is the English singular variant of a message, and STRING2 the English plural variant of the same message. The default value for DOMAIN is the current value of `TEXTDOMAIN'. The default value for CATEGORY is `"LC_MESSAGES"'. `bindtextdomain(DIRECTORY [, DOMAIN])' This function allows you to specify the directory in which `gawk' will look for message translation files, in case they will not or cannot be placed in the "standard" locations (e.g., during testing). It returns the directory in which DOMAIN is "bound." The default DOMAIN is the value of `TEXTDOMAIN'. If DIRECTORY is the null string (`""'), then `bindtextdomain' returns the current binding for the given DOMAIN. 8.2 User-Defined Functions ========================== Complicated `awk' programs can often be simplified by defining your own functions. User-defined functions can be called just like built-in ones (*note Function Calls::), but it is up to you to define them, i.e., to tell `awk' what they should do. 8.2.1 Function Definition Syntax -------------------------------- Definitions of functions can appear anywhere between the rules of an `awk' program. Thus, the general form of an `awk' program is extended to include sequences of rules _and_ user-defined function definitions. There is no need to put the definition of a function before all uses of the function. This is because `awk' reads the entire program before starting to execute any of it. The definition of a function named NAME looks like this: function NAME(PARAMETER-LIST) { BODY-OF-FUNCTION } NAME is the name of the function to define. A valid function name is like a valid variable name: a sequence of letters, digits, and underscores that doesn't start with a digit. Within a single `awk' program, any particular name can only be used as a variable, array, or function. PARAMETER-LIST is a list of the function's arguments and local variable names, separated by commas. When the function is called, the argument names are used to hold the argument values given in the call. The local variables are initialized to the empty string. A function cannot have two parameters with the same name, nor may it have a parameter with the same name as the function itself. The BODY-OF-FUNCTION consists of `awk' statements. It is the most important part of the definition, because it says what the function should actually _do_. The argument names exist to give the body a way to talk about the arguments; local variables exist to give the body places to keep temporary values. Argument names are not distinguished syntactically from local variable names. Instead, the number of arguments supplied when the function is called determines how many argument variables there are. Thus, if three argument values are given, the first three names in PARAMETER-LIST are arguments and the rest are local variables. It follows that if the number of arguments is not the same in all calls to the function, some of the names in PARAMETER-LIST may be arguments on some occasions and local variables on others. Another way to think of this is that omitted arguments default to the null string. Usually when you write a function, you know how many names you intend to use for arguments and how many you intend to use as local variables. It is conventional to place some extra space between the arguments and the local variables, in order to document how your function is supposed to be used. During execution of the function body, the arguments and local variable values hide, or "shadow", any variables of the same names used in the rest of the program. The shadowed variables are not accessible in the function definition, because there is no way to name them while their names have been taken away for the local variables. All other variables used in the `awk' program can be referenced or set normally in the function's body. The arguments and local variables last only as long as the function body is executing. Once the body finishes, you can once again access the variables that were shadowed while the function was running. The function body can contain expressions that call functions. They can even call this function, either directly or by way of another function. When this happens, we say the function is "recursive". The act of a function calling itself is called "recursion". In many `awk' implementations, including `gawk', the keyword `function' may be abbreviated `func'. However, POSIX only specifies the use of the keyword `function'. This actually has some practical implications. If `gawk' is in POSIX-compatibility mode (*note Options::), then the following statement does _not_ define a function: func foo() { a = sqrt($1) ; print a } Instead it defines a rule that, for each record, concatenates the value of the variable `func' with the return value of the function `foo'. If the resulting string is non-null, the action is executed. This is probably not what is desired. (`awk' accepts this input as syntactically valid, because functions may be used before they are defined in `awk' programs.) To ensure that your `awk' programs are portable, always use the keyword `function' when defining a function. 8.2.2 Function Definition Examples ---------------------------------- Here is an example of a user-defined function, called `myprint', that takes a number and prints it in a specific format: function myprint(num) { printf "%6.3g\n", num } To illustrate, here is an `awk' rule that uses our `myprint' function: $3 > 0 { myprint($3) } This program prints, in our special format, all the third fields that contain a positive number in our input. Therefore, when given the following: 1.2 3.4 5.6 7.8 9.10 11.12 -13.14 15.16 17.18 19.20 21.22 23.24 this program, using our function to format the results, prints: 5.6 21.2 This function deletes all the elements in an array: function delarray(a, i) { for (i in a) delete a[i] } When working with arrays, it is often necessary to delete all the elements in an array and start over with a new list of elements (*note Delete::). Instead of having to repeat this loop everywhere that you need to clear out an array, your program can just call `delarray'. (This guarantees portability. The use of `delete ARRAY' to delete the contents of an entire array is a nonstandard extension.) The following is an example of a recursive function. It takes a string as an input parameter and returns the string in backwards order. Recursive functions must always have a test that stops the recursion. In this case, the recursion terminates when the starting position is zero, i.e., when there are no more characters left in the string. function rev(str, start) { if (start == 0) return "" return (substr(str, start, 1) rev(str, start - 1)) } If this function is in a file named `rev.awk', it can be tested this way: $ echo "Don't Panic!" | > gawk --source '{ print rev($0, length($0)) }' -f rev.awk -| !cinaP t'noD The C `ctime' function takes a timestamp and returns it in a string, formatted in a well-known fashion. The following example uses the built-in `strftime' function (*note Time Functions::) to create an `awk' version of `ctime': # ctime.awk # # awk version of C ctime(3) function function ctime(ts, format) { format = "%a %b %d %H:%M:%S %Z %Y" if (ts == 0) ts = systime() # use current time as default return strftime(format, ts) } 8.2.3 Calling User-Defined Functions ------------------------------------ "Calling a function" means causing the function to run and do its job. A function call is an expression and its value is the value returned by the function. A function call consists of the function name followed by the arguments in parentheses. `awk' expressions are what you write in the call for the arguments. Each time the call is executed, these expressions are evaluated, and the values are the actual arguments. For example, here is a call to `foo' with three arguments (the first being a string concatenation): foo(x y, "lose", 4 * z) *Caution:* Whitespace characters (spaces and tabs) are not allowed between the function name and the open-parenthesis of the argument list. If you write whitespace by mistake, `awk' might think that you mean to concatenate a variable with an expression in parentheses. However, it notices that you used a function name and not a variable name, and reports an error. When a function is called, it is given a _copy_ of the values of its arguments. This is known as "call by value". The caller may use a variable as the expression for the argument, but the called function does not know this--it only knows what value the argument had. For example, if you write the following code: foo = "bar" z = myfunc(foo) then you should not think of the argument to `myfunc' as being "the variable `foo'." Instead, think of the argument as the string value `"bar"'. If the function `myfunc' alters the values of its local variables, this has no effect on any other variables. Thus, if `myfunc' does this: function myfunc(str) { print str str = "zzz" print str } to change its first argument variable `str', it does _not_ change the value of `foo' in the caller. The role of `foo' in calling `myfunc' ended when its value (`"bar"') was computed. If `str' also exists outside of `myfunc', the function body cannot alter this outer value, because it is shadowed during the execution of `myfunc' and cannot be seen or changed from there. However, when arrays are the parameters to functions, they are _not_ copied. Instead, the array itself is made available for direct manipulation by the function. This is usually called "call by reference". Changes made to an array parameter inside the body of a function _are_ visible outside that function. NOTE: Changing an array parameter inside a function can be very dangerous if you do not watch what you are doing. For example: function changeit(array, ind, nvalue) { array[ind] = nvalue } BEGIN { a[1] = 1; a[2] = 2; a[3] = 3 changeit(a, 2, "two") printf "a[1] = %s, a[2] = %s, a[3] = %s\n", a[1], a[2], a[3] } prints `a[1] = 1, a[2] = two, a[3] = 3', because `changeit' stores `"two"' in the second element of `a'. Some `awk' implementations allow you to call a function that has not been defined. They only report a problem at runtime when the program actually tries to call the function. For example: BEGIN { if (0) foo() else bar() } function bar() { ... } # note that `foo' is not defined Because the `if' statement will never be true, it is not really a problem that `foo' has not been defined. Usually, though, it is a problem if a program calls an undefined function. If `--lint' is specified (*note Options::), `gawk' reports calls to undefined functions. Some `awk' implementations generate a runtime error if you use the `next' statement (*note Next Statement::) inside a user-defined function. `gawk' does not have this limitation. 8.2.4 The `return' Statement ---------------------------- The body of a user-defined function can contain a `return' statement. This statement returns control to the calling part of the `awk' program. It can also be used to return a value for use in the rest of the `awk' program. It looks like this: return [EXPRESSION] The EXPRESSION part is optional. If it is omitted, then the returned value is undefined, and therefore, unpredictable. A `return' statement with no value expression is assumed at the end of every function definition. So if control reaches the end of the function body, then the function returns an unpredictable value. `awk' does _not_ warn you if you use the return value of such a function. Sometimes, you want to write a function for what it does, not for what it returns. Such a function corresponds to a `void' function in C or to a `procedure' in Pascal. Thus, it may be appropriate to not return any value; simply bear in mind that if you use the return value of such a function, you do so at your own risk. The following is an example of a user-defined function that returns a value for the largest number among the elements of an array: function maxelt(vec, i, ret) { for (i in vec) { if (ret == "" || vec[i] > ret) ret = vec[i] } return ret } You call `maxelt' with one argument, which is an array name. The local variables `i' and `ret' are not intended to be arguments; while there is nothing to stop you from passing more than one argument to `maxelt', the results would be strange. The extra space before `i' in the function parameter list indicates that `i' and `ret' are not supposed to be arguments. You should follow this convention when defining functions. The following program uses the `maxelt' function. It loads an array, calls `maxelt', and then reports the maximum number in that array: function maxelt(vec, i, ret) { for (i in vec) { if (ret == "" || vec[i] > ret) ret = vec[i] } return ret } # Load all fields of each record into nums. { for(i = 1; i <= NF; i++) nums[NR, i] = $i } END { print maxelt(nums) } Given the following input: 1 5 23 8 16 44 3 5 2 8 26 256 291 1396 2962 100 -6 467 998 1101 99385 11 0 225 the program reports (predictably) that `99385' is the largest number in the array. 8.2.5 Functions and Their Effects on Variable Typing ---------------------------------------------------- `awk' is a very fluid language. It is possible that `awk' can't tell if an identifier represents a regular variable or an array until runtime. Here is an annotated sample program: function foo(a) { a[1] = 1 # parameter is an array } BEGIN { b = 1 foo(b) # invalid: fatal type mismatch foo(x) # x uninitialized, becomes an array dynamically x = 1 # now not allowed, runtime error } Usually, such things aren't a big issue, but it's worth being aware of them. 9 Internationalization with `gawk' ********************************** Once upon a time, computer makers wrote software that worked only in English. Eventually, hardware and software vendors noticed that if their systems worked in the native languages of non-English-speaking countries, they were able to sell more systems. As a result, internationalization and localization of programs and software systems became a common practice. Until recently, the ability to provide internationalization was largely restricted to programs written in C and C++. This major node describes the underlying library `gawk' uses for internationalization, as well as how `gawk' makes internationalization features available at the `awk' program level. Having internationalization available at the `awk' level gives software developers additional flexibility--they are no longer required to write in C when internationalization is a requirement. 9.1 Internationalization and Localization ========================================= "Internationalization" means writing (or modifying) a program once, in such a way that it can use multiple languages without requiring further source-code changes. "Localization" means providing the data necessary for an internationalized program to work in a particular language. Most typically, these terms refer to features such as the language used for printing error messages, the language used to read responses, and information related to how numerical and monetary values are printed and read. 9.2 GNU `gettext' ================= The facilities in GNU `gettext' focus on messages; strings printed by a program, either directly or via formatting with `printf' or `sprintf'.(1) When using GNU `gettext', each application has its own "text domain". This is a unique name, such as `kpilot' or `gawk', that identifies the application. A complete application may have multiple components--programs written in C or C++, as well as scripts written in `sh' or `awk'. All of the components use the same text domain. To make the discussion concrete, assume we're writing an application named `guide'. Internationalization consists of the following steps, in this order: 1. The programmer goes through the source for all of `guide''s components and marks each string that is a candidate for translation. For example, `"`-F': option required"' is a good candidate for translation. A table with strings of option names is not (e.g., `gawk''s `--profile' option should remain the same, no matter what the local language). 2. The programmer indicates the application's text domain (`"guide"') to the `gettext' library, by calling the `textdomain' function. 3. Messages from the application are extracted from the source code and collected into a portable object file (`guide.po'), which lists the strings and their translations. The translations are initially empty. The original (usually English) messages serve as the key for lookup of the translations. 4. For each language with a translator, `guide.po' is copied and translations are created and shipped with the application. 5. Each language's `.po' file is converted into a binary message object (`.mo') file. A message object file contains the original messages and their translations in a binary format that allows fast lookup of translations at runtime. 6. When `guide' is built and installed, the binary translation files are installed in a standard place. 7. For testing and development, it is possible to tell `gettext' to use `.mo' files in a different directory than the standard one by using the `bindtextdomain' function. 8. At runtime, `guide' looks up each string via a call to `gettext'. The returned string is the translated string if available, or the original string if not. 9. If necessary, it is possible to access messages from a different text domain than the one belonging to the application, without having to switch the application's default text domain back and forth. In C (or C++), the string marking and dynamic translation lookup are accomplished by wrapping each string in a call to `gettext': printf(gettext("Don't Panic!\n")); The tools that extract messages from source code pull out all strings enclosed in calls to `gettext'. The GNU `gettext' developers, recognizing that typing `gettext' over and over again is both painful and ugly to look at, use the macro `_' (an underscore) to make things easier: /* In the standard header file: */ #define _(str) gettext(str) /* In the program text: */ printf(_("Don't Panic!\n")); This reduces the typing overhead to just three extra characters per string and is considerably easier to read as well. There are locale "categories" for different types of locale-related information. The defined locale categories that `gettext' knows about are: `LC_MESSAGES' Text messages. This is the default category for `gettext' operations, but it is possible to supply a different one explicitly, if necessary. (It is almost never necessary to supply a different category.) `LC_COLLATE' Text-collation information; i.e., how different characters and/or groups of characters sort in a given language. `LC_CTYPE' Character-type information (alphabetic, digit, upper- or lowercase, and so on). This information is accessed via the POSIX character classes in regular expressions, such as `/[[:alnum:]]/' (*note Regexp Operators::). `LC_MONETARY' Monetary information, such as the currency symbol, and whether the symbol goes before or after a number. `LC_NUMERIC' Numeric information, such as which characters to use for the decimal point and the thousands separator.(2) `LC_RESPONSE' Response information, such as how "yes" and "no" appear in the local language, and possibly other information as well. `LC_TIME' Time- and date-related information, such as 12- or 24-hour clock, month printed before or after day in a date, local month abbreviations, and so on. `LC_ALL' All of the above. (Not too useful in the context of `gettext'.) ---------- Footnotes ---------- (1) For some operating systems, the `gawk' port doesn't support GNU `gettext'. This applies most notably to the PC operating systems. As such, these features are not available if you are using one of those operating systems. Sorry. (2) Americans use a comma every three decimal places and a period for the decimal point, while many Europeans do exactly the opposite: `1,234.56' versus `1.234,56'. 9.3 Internationalizing `awk' Programs ===================================== `gawk' provides the following variables and functions for internationalization: `TEXTDOMAIN' This variable indicates the application's text domain. For compatibility with GNU `gettext', the default value is `"messages"'. `_"your message here"' String constants marked with a leading underscore are candidates for translation at runtime. String constants without a leading underscore are not translated. `dcgettext(STRING [, DOMAIN [, CATEGORY]])' This built-in function returns the translation of STRING in text domain DOMAIN for locale category CATEGORY. The default value for DOMAIN is the current value of `TEXTDOMAIN'. The default value for CATEGORY is `"LC_MESSAGES"'. If you supply a value for CATEGORY, it must be a string equal to one of the known locale categories described in *Note Explaining gettext::. You must also supply a text domain. Use `TEXTDOMAIN' if you want to use the current domain. *Caution:* The order of arguments to the `awk' version of the `dcgettext' function is purposely different from the order for the C version. The `awk' version's order was chosen to be simple and to allow for reasonable `awk'-style default arguments. `dcngettext(STRING1, STRING2, NUMBER [, DOMAIN [, CATEGORY]])' This built-in function returns the plural form used for NUMBER of the translation of STRING1 and STRING2 in text domain DOMAIN for locale category CATEGORY. STRING1 is the English singular variant of a message, and STRING2 the English plural variant of the same message. The default value for DOMAIN is the current value of `TEXTDOMAIN'. The default value for CATEGORY is `"LC_MESSAGES"'. The same remarks as for the `dcgettext' function apply. `bindtextdomain(DIRECTORY [, DOMAIN])' This built-in function allows you to specify the directory in which `gettext' looks for `.mo' files, in case they will not or cannot be placed in the standard locations (e.g., during testing). It returns the directory in which DOMAIN is "bound." The default DOMAIN is the value of `TEXTDOMAIN'. If DIRECTORY is the null string (`""'), then `bindtextdomain' returns the current binding for the given DOMAIN. To use these facilities in your `awk' program, follow the steps outlined in *Note Explaining gettext::, like so: 1. Set the variable `TEXTDOMAIN' to the text domain of your program. This is best done in a `BEGIN' rule (*note BEGIN/END::), or it can also be done via the `-v' command-line option (*note Options::): BEGIN { TEXTDOMAIN = "guide" ... } 2. Mark all translatable strings with a leading underscore (`_') character. It _must_ be adjacent to the opening quote of the string. For example: print _"hello, world" x = _"you goofed" printf(_"Number of users is %d\n", nusers) 3. If you are creating strings dynamically, you can still translate them, using the `dcgettext' built-in function: message = nusers " users logged in" message = dcgettext(message, "adminprog") print message Here, the call to `dcgettext' supplies a different text domain (`"adminprog"') in which to find the message, but it uses the default `"LC_MESSAGES"' category. 4. During development, you might want to put the `.mo' file in a private directory for testing. This is done with the `bindtextdomain' built-in function: BEGIN { TEXTDOMAIN = "guide" # our text domain if (Testing) { # where to find our files bindtextdomain("testdir") # joe is in charge of adminprog bindtextdomain("../joe/testdir", "adminprog") } ... } *Note I18N Example::, for an example program showing the steps to create and use translations from `awk'. 9.4 Translating `awk' Programs ============================== Once a program's translatable strings have been marked, they must be extracted to create the initial `.po' file. As part of translation, it is often helpful to rearrange the order in which arguments to `printf' are output. `gawk''s `--gen-po' command-line option extracts the messages and is discussed next. After that, `printf''s ability to rearrange the order for `printf' arguments at runtime is covered. 9.4.1 Extracting Marked Strings ------------------------------- Once your `awk' program is working, and all the strings have been marked and you've set (and perhaps bound) the text domain, it is time to produce translations. First, use the `--gen-po' command-line option to create the initial `.po' file: $ gawk --gen-po -f guide.awk > guide.po When run with `--gen-po', `gawk' does not execute your program. Instead, it parses it as usual and prints all marked strings to standard output in the format of a GNU `gettext' Portable Object file. Also included in the output are any constant strings that appear as the first argument to `dcgettext' or as the first and second argument to `dcngettext'.(1) *Note I18N Example::, for the full list of steps to go through to create and test translations for `guide'. ---------- Footnotes ---------- (1) Starting with `gettext' version 0.11.5, the `xgettext' utility that comes with GNU `gettext' can handle `.awk' files. 9.4.2 Rearranging `printf' Arguments ------------------------------------ Format strings for `printf' and `sprintf' (*note Printf::) present a special problem for translation. Consider the following:(1) printf(_"String `%s' has %d characters\n", string, length(string))) A possible German translation for this might be: "%d Zeichen lang ist die Zeichenkette `%s'\n" The problem should be obvious: the order of the format specifications is different from the original! Even though `gettext' can return the translated string at runtime, it cannot change the argument order in the call to `printf'. To solve this problem, `printf' format specificiers may have an additional optional element, which we call a "positional specifier". For example: "%2$d Zeichen lang ist die Zeichenkette `%1$s'\n" Here, the positional specifier consists of an integer count, which indicates which argument to use, and a `$'. Counts are one-based, and the format string itself is _not_ included. Thus, in the following example, `string' is the first argument and `length(string)' is the second: $ gawk 'BEGIN { > string = "Dont Panic" > printf _"%2$d characters live in \"%1$s\"\n", > string, length(string) > }' -| 10 characters live in "Dont Panic" If present, positional specifiers come first in the format specification, before the flags, the field width, and/or the precision. Positional specifiers can be used with the dynamic field width and precision capability: $ gawk 'BEGIN { > printf("%*.*s\n", 10, 20, "hello") > printf("%3$*2$.*1$s\n", 20, 10, "hello") > }' -| hello -| hello NOTE: When using `*' with a positional specifier, the `*' comes first, then the integer position, and then the `$'. This is somewhat counterintutive. `gawk' does not allow you to mix regular format specifiers and those with positional specifiers in the same string: $ gawk 'BEGIN { printf _"%d %3$s\n", 1, 2, "hi" }' error--> gawk: cmd. line:1: fatal: must use `count$' on all formats or none NOTE: There are some pathological cases that `gawk' may fail to diagnose. In such cases, the output may not be what you expect. It's still a bad idea to try mixing them, even if `gawk' doesn't detect it. Although positional specifiers can be used directly in `awk' programs, their primary purpose is to help in producing correct translations of format strings into languages different from the one in which the program is first written. ---------- Footnotes ---------- (1) This example is borrowed from the GNU `gettext' manual. 9.4.3 `awk' Portability Issues ------------------------------ `gawk''s internationalization features were purposely chosen to have as little impact as possible on the portability of `awk' programs that use them to other versions of `awk'. Consider this program: BEGIN { TEXTDOMAIN = "guide" if (Test_Guide) # set with -v bindtextdomain("/test/guide/messages") print _"don't panic!" } As written, it won't work on other versions of `awk'. However, it is actually almost portable, requiring very little change: * Assignments to `TEXTDOMAIN' won't have any effect, since `TEXTDOMAIN' is not special in other `awk' implementations. * Non-GNU versions of `awk' treat marked strings as the concatenation of a variable named `_' with the string following it.(1) Typically, the variable `_' has the null string (`""') as its value, leaving the original string constant as the result. * By defining "dummy" functions to replace `dcgettext', `dcngettext' and `bindtextdomain', the `awk' program can be made to run, but all the messages are output in the original language. For example: function bindtextdomain(dir, domain) { return dir } function dcgettext(string, domain, category) { return string } function dcngettext(string1, string2, number, domain, category) { return (number == 1 ? string1 : string2) } * The use of positional specifications in `printf' or `sprintf' is _not_ portable. To support `gettext' at the C level, many systems' C versions of `sprintf' do support positional specifiers. But it works only if enough arguments are supplied in the function call. Many versions of `awk' pass `printf' formats and arguments unchanged to the underlying C library version of `sprintf', but only one format and argument at a time. What happens if a positional specification is used is anybody's guess. However, since the positional specifications are primarily for use in _translated_ format strings, and since non-GNU `awk's never retrieve the translated string, this should not be a problem in practice. ---------- Footnotes ---------- (1) This is good fodder for an "Obfuscated `awk'" contest. 9.5 A Simple Internationalization Example ========================================= Now let's look at a step-by-step example of how to internationalize and localize a simple `awk' program, using `guide.awk' as our original source: BEGIN { TEXTDOMAIN = "guide" bindtextdomain(".") # for testing print _"Don't Panic" print _"The Answer Is", 42 print "Pardon me, Zaphod who?" } Run `gawk --gen-po' to create the `.po' file: $ gawk --gen-po -f guide.awk > guide.po This produces: #: guide.awk:4 msgid "Don't Panic" msgstr "" #: guide.awk:5 msgid "The Answer Is" msgstr "" This original portable object file is saved and reused for each language into which the application is translated. The `msgid' is the original string and the `msgstr' is the translation. NOTE: Strings not marked with a leading underscore do not appear in the `guide.po' file. Next, the messages must be translated. Here is a translation to a hypothetical dialect of English, called "Mellow":(1) $ cp guide.po guide-mellow.po ADD TRANSLATIONS TO guide-mellow.po ... Following are the translations: #: guide.awk:4 msgid "Don't Panic" msgstr "Hey man, relax!" #: guide.awk:5 msgid "The Answer Is" msgstr "Like, the scoop is" The next step is to make the directory to hold the binary message object file and then to create the `guide.mo' file. The directory layout shown here is standard for GNU `gettext' on GNU/Linux systems. Other versions of `gettext' may use a different layout: $ mkdir en_US en_US/LC_MESSAGES The `msgfmt' utility does the conversion from human-readable `.po' file to machine-readable `.mo' file. By default, `msgfmt' creates a file named `messages'. This file must be renamed and placed in the proper directory so that `gawk' can find it: $ msgfmt guide-mellow.po $ mv messages en_US/LC_MESSAGES/guide.mo Finally, we run the program to test it: $ gawk -f guide.awk -| Hey man, relax! -| Like, the scoop is 42 -| Pardon me, Zaphod who? If the three replacement functions for `dcgettext', `dcngettext' and `bindtextdomain' (*note I18N Portability::) are in a file named `libintl.awk', then we can run `guide.awk' unchanged as follows: $ gawk --posix -f guide.awk -f libintl.awk -| Don't Panic -| The Answer Is 42 -| Pardon me, Zaphod who? ---------- Footnotes ---------- (1) Perhaps it would be better if it were called "Hippy." Ah, well. 9.6 `gawk' Can Speak Your Language ================================== As of version 3.1, `gawk' itself has been internationalized using the GNU `gettext' package. (GNU `gettext' is described in complete detail in *Note Top::.) As of this writing, the latest version of GNU `gettext' is version 0.11.5 (ftp://ftp.gnu.org/gnu/gettext/gettext-0.11.5.tar.gz). If a translation of `gawk''s messages exists, then `gawk' produces usage messages, warnings, and fatal errors in the local language. On systems that do not use version 2 (or later) of the GNU C library, you should configure `gawk' with the `--with-included-gettext' option before compiling and installing it. *Note Additional Configuration Options::, for more information. 10 Advanced Features of `gawk' ****************************** Write documentation as if whoever reads it is a violent psychopath who knows where you live. Steve English, as quoted by Peter Langston This major node discusses advanced features in `gawk'. It's a bit of a "grab bag" of items that are otherwise unrelated to each other. First, a command-line option allows `gawk' to recognize nondecimal numbers in input data, not just in `awk' programs. Next, two-way I/O, discussed briefly in earlier parts of this Info file, is described in full detail, along with the basics of TCP/IP networking and BSD portal files. Finally, `gawk' can "profile" an `awk' program, making it possible to tune it for performance. *Note Dynamic Extensions::, discusses the ability to dynamically add new built-in functions to `gawk'. As this feature is still immature and likely to change, its description is relegated to an appendix. 10.1 Allowing Nondecimal Input Data =================================== If you run `gawk' with the `--non-decimal-data' option, you can have nondecimal constants in your input data: $ echo 0123 123 0x123 | > gawk --non-decimal-data '{ printf "%d, %d, %d\n", > $1, $2, $3 }' -| 83, 123, 291 For this feature to work, write your program so that `gawk' treats your data as numeric: $ echo 0123 123 0x123 | gawk '{ print $1, $2, $3 }' -| 0123 123 0x123 The `print' statement treats its expressions as strings. Although the fields can act as numbers when necessary, they are still strings, so `print' does not try to treat them numerically. You may need to add zero to a field to force it to be treated as a number. For example: $ echo 0123 123 0x123 | gawk --non-decimal-data ' > { print $1, $2, $3 > print $1 + 0, $2 + 0, $3 + 0 }' -| 0123 123 0x123 -| 83 123 291 Because it is common to have decimal data with leading zeros, and because using it could lead to surprising results, the default is to leave this facility disabled. If you want it, you must explicitly request it. *Caution:* _Use of this option is not recommended._ It can break old programs very badly. Instead, use the `strtonum' function to convert your data (*note Nondecimal-numbers::). This makes your programs easier to write and easier to read, and leads to less surprising results. 10.2 Two-Way Communications with Another Process ================================================ From: brennan@whidbey.com (Mike Brennan) Newsgroups: comp.lang.awk Subject: Re: Learn the SECRET to Attract Women Easily Date: 4 Aug 1997 17:34:46 GMT Message-ID: <5s53rm$eca@news.whidbey.com> On 3 Aug 1997 13:17:43 GMT, Want More Dates??? wrote: >Learn the SECRET to Attract Women Easily > >The SCENT(tm) Pheromone Sex Attractant For Men to Attract Women The scent of awk programmers is a lot more attractive to women than the scent of perl programmers. -- Mike Brennan It is often useful to be able to send data to a separate program for processing and then read the result. This can always be done with temporary files: # write the data for processing tempfile = ("mydata." PROCINFO["pid"]) while (NOT DONE WITH DATA) print DATA | ("subprogram > " tempfile) close("subprogram > " tempfile) # read the results, remove tempfile when done while ((getline newdata < tempfile) > 0) PROCESS newdata APPROPRIATELY close(tempfile) system("rm " tempfile) This works, but not elegantly. Among other things, it requires that the program be run in a directory that cannot be shared among users; for example, `/tmp' will not do, as another user might happen to be using a temporary file with the same name. Starting with version 3.1 of `gawk', it is possible to open a _two-way_ pipe to another process. The second process is termed a "coprocess", since it runs in parallel with `gawk'. The two-way connection is created using the new `|&' operator (borrowed from the Korn shell, `ksh'):(1) do { print DATA |& "subprogram" "subprogram" |& getline results } while (DATA LEFT TO PROCESS) close("subprogram") The first time an I/O operation is executed using the `|&' operator, `gawk' creates a two-way pipeline to a child process that runs the other program. Output created with `print' or `printf' is written to the program's standard input, and output from the program's standard output can be read by the `gawk' program using `getline'. As is the case with processes started by `|', the subprogram can be any program, or pipeline of programs, that can be started by the shell. There are some cautionary items to be aware of: * As the code inside `gawk' currently stands, the coprocess's standard error goes to the same place that the parent `gawk''s standard error goes. It is not possible to read the child's standard error separately. * I/O buffering may be a problem. `gawk' automatically flushes all output down the pipe to the child process. However, if the coprocess does not flush its output, `gawk' may hang when doing a `getline' in order to read the coprocess's results. This could lead to a situation known as "deadlock", where each process is waiting for the other one to do something. It is possible to close just one end of the two-way pipe to a coprocess, by supplying a second argument to the `close' function of either `"to"' or `"from"' (*note Close Files And Pipes::). These strings tell `gawk' to close the end of the pipe that sends data to the process or the end that reads from it, respectively. This is particularly necessary in order to use the system `sort' utility as part of a coprocess; `sort' must read _all_ of its input data before it can produce any output. The `sort' program does not receive an end-of-file indication until `gawk' closes the write end of the pipe. When you have finished writing data to the `sort' utility, you can close the `"to"' end of the pipe, and then start reading sorted data via `getline'. For example: BEGIN { command = "LC_ALL=C sort" n = split("abcdefghijklmnopqrstuvwxyz", a, "") for (i = n; i > 0; i--) print a[i] |& command close(command, "to") while ((command |& getline line) > 0) print "got", line close(command) } This program writes the letters of the alphabet in reverse order, one per line, down the two-way pipe to `sort'. It then closes the write end of the pipe, so that `sort' receives an end-of-file indication. This causes `sort' to sort the data and write the sorted data back to the `gawk' program. Once all of the data has been read, `gawk' terminates the coprocess and exits. As a side note, the assignment `LC_ALL=C' in the `sort' command ensures traditional Unix (ASCII) sorting from `sort'. Beginning with `gawk' 3.1.2, you may use Pseudo-ttys (ptys) for two-way communication instead of pipes, if your system supports them. This is done on a per-command basis, by setting a special element in the `PROCINFO' array (*note Auto-set::), like so: command = "sort -nr" # command, saved in variable for convenience PROCINFO[command, "pty"] = 1 # update PROCINFO print ... |& command # start two-way pipe ... Using ptys avoids the buffer deadlock issues described earlier, at some loss in performance. If your system does not have ptys, or if all the system's ptys are in use, `gawk' automatically falls back to using regular pipes. ---------- Footnotes ---------- (1) This is very different from the same operator in the C shell, `csh'. 10.3 Using `gawk' for Network Programming ========================================= `EMISTERED': A host is a host from coast to coast, and no-one can talk to host that's close, unless the host that isn't close is busy hung or dead. In addition to being able to open a two-way pipeline to a coprocess on the same system (*note Two-way I/O::), it is possible to make a two-way connection to another process on another system across an IP networking connection. You can think of this as just a _very long_ two-way pipeline to a coprocess. The way `gawk' decides that you want to use TCP/IP networking is by recognizing special file names that begin with `/inet/'. The full syntax of the special file name is `/inet/PROTOCOL/LOCAL-PORT/REMOTE-HOST/REMOTE-PORT'. The components are: PROTOCOL The protocol to use over IP. This must be either `tcp', `udp', or `raw', for a TCP, UDP, or raw IP connection, respectively. The use of TCP is recommended for most applications. *Caution:* The use of raw sockets is not currently supported in version 3.1 of `gawk'. LOCAL-PORT The local TCP or UDP port number to use. Use a port number of `0' when you want the system to pick a port. This is what you should do when writing a TCP or UDP client. You may also use a well-known service name, such as `smtp' or `http', in which case `gawk' attempts to determine the predefined port number using the C `getservbyname' function. REMOTE-HOST The IP address or fully-qualified domain name of the Internet host to which you want to connect. REMOTE-PORT The TCP or UDP port number to use on the given REMOTE-HOST. Again, use `0' if you don't care, or else a well-known service name. Consider the following very simple example: BEGIN { Service = "/inet/tcp/0/localhost/daytime" Service |& getline print $0 close(Service) } This program reads the current date and time from the local system's TCP `daytime' server. It then prints the results and closes the connection. Because this topic is extensive, the use of `gawk' for TCP/IP programming is documented separately. *Note Top::, for a much more complete introduction and discussion, as well as extensive examples. 10.4 Using `gawk' with BSD Portals ================================== Similar to the `/inet' special files, if `gawk' is configured with the `--enable-portals' option (*note Quick Installation::), then `gawk' treats files whose pathnames begin with `/p' as 4.4 BSD-style portals. When used with the `|&' operator, `gawk' opens the file for two-way communications. The operating system's portal mechanism then manages creating the process associated with the portal and the corresponding communications with the portal's process. 10.5 Profiling Your `awk' Programs ================================== Beginning with version 3.1 of `gawk', you may produce execution traces of your `awk' programs. This is done with a specially compiled version of `gawk', called `pgawk' ("profiling `gawk'"). `pgawk' is identical in every way to `gawk', except that when it has finished running, it creates a profile of your program in a file named `awkprof.out'. Because it is profiling, it also executes up to 45% slower than `gawk' normally does. As shown in the following example, the `--profile' option can be used to change the name of the file where `pgawk' will write the profile: $ pgawk --profile=myprog.prof -f myprog.awk data1 data2 In the above example, `pgawk' places the profile in `myprog.prof' instead of in `awkprof.out'. Regular `gawk' also accepts this option. When called with just `--profile', `gawk' "pretty prints" the program into `awkprof.out', without any execution counts. You may supply an option to `--profile' to change the file name. Here is a sample session showing a simple `awk' program, its input data, and the results from running `pgawk'. First, the `awk' program: BEGIN { print "First BEGIN rule" } END { print "First END rule" } /foo/ { print "matched /foo/, gosh" for (i = 1; i <= 3; i++) sing() } { if (/foo/) print "if is true" else print "else is true" } BEGIN { print "Second BEGIN rule" } END { print "Second END rule" } function sing( dummy) { print "I gotta be me!" } Following is the input data: foo bar baz foo junk Here is the `awkprof.out' that results from running `pgawk' on this program and data (this example also illustrates that `awk' programmers sometimes have to work late): # gawk profile, created Sun Aug 13 00:00:15 2000 # BEGIN block(s) BEGIN { 1 print "First BEGIN rule" 1 print "Second BEGIN rule" } # Rule(s) 5 /foo/ { # 2 2 print "matched /foo/, gosh" 6 for (i = 1; i <= 3; i++) { 6 sing() } } 5 { 5 if (/foo/) { # 2 2 print "if is true" 3 } else { 3 print "else is true" } } # END block(s) END { 1 print "First END rule" 1 print "Second END rule" } # Functions, listed alphabetically 6 function sing(dummy) { 6 print "I gotta be me!" } This example illustrates many of the basic rules for profiling output. The rules are as follows: * The program is printed in the order `BEGIN' rule, pattern/action rules, `END' rule and functions, listed alphabetically. Multiple `BEGIN' and `END' rules are merged together. * Pattern-action rules have two counts. The first count, to the left of the rule, shows how many times the rule's pattern was _tested_. The second count, to the right of the rule's opening left brace in a comment, shows how many times the rule's action was _executed_. The difference between the two indicates how many times the rule's pattern evaluated to false. * Similarly, the count for an `if'-`else' statement shows how many times the condition was tested. To the right of the opening left brace for the `if''s body is a count showing how many times the condition was true. The count for the `else' indicates how many times the test failed. * The count for a loop header (such as `for' or `while') shows how many times the loop test was executed. (Because of this, you can't just look at the count on the first statement in a rule to determine how many times the rule was executed. If the first statement is a loop, the count is misleading.) * For user-defined functions, the count next to the `function' keyword indicates how many times the function was called. The counts next to the statements in the body show how many times those statements were executed. * The layout uses "K&R" style with tabs. Braces are used everywhere, even when the body of an `if', `else', or loop is only a single statement. * Parentheses are used only where needed, as indicated by the structure of the program and the precedence rules. For example, `(3 + 5) * 4' means add three plus five, then multiply the total by four. However, `3 + 5 * 4' has no parentheses, and means `3 + (5 * 4)'. * All string concatenations are parenthesized too. (This could be made a bit smarter.) * Parentheses are used around the arguments to `print' and `printf' only when the `print' or `printf' statement is followed by a redirection. Similarly, if the target of a redirection isn't a scalar, it gets parenthesized. * `pgawk' supplies leading comments in front of the `BEGIN' and `END' rules, the pattern/action rules, and the functions. The profiled version of your program may not look exactly like what you typed when you wrote it. This is because `pgawk' creates the profiled version by "pretty printing" its internal representation of the program. The advantage to this is that `pgawk' can produce a standard representation. The disadvantage is that all source-code comments are lost, as are the distinctions among multiple `BEGIN' and `END' rules. Also, things such as: /foo/ come out as: /foo/ { print $0 } which is correct, but possibly surprising. Besides creating profiles when a program has completed, `pgawk' can produce a profile while it is running. This is useful if your `awk' program goes into an infinite loop and you want to see what has been executed. To use this feature, run `pgawk' in the background: $ pgawk -f myprog & [1] 13992 The shell prints a job number and process ID number; in this case, 13992. Use the `kill' command to send the `USR1' signal to `pgawk': $ kill -USR1 13992 As usual, the profiled version of the program is written to `awkprof.out', or to a different file if you use the `--profile' option. Along with the regular profile, as shown earlier, the profile includes a trace of any active functions: # Function Call Stack: # 3. baz # 2. bar # 1. foo # -- main -- You may send `pgawk' the `USR1' signal as many times as you like. Each time, the profile and function call trace are appended to the output profile file. If you use the `HUP' signal instead of the `USR1' signal, `pgawk' produces the profile and the function call trace and then exits. When `pgawk' runs on MS-DOS or MS-Windows, it uses the `INT' and `QUIT' signals for producing the profile and, in the case of the `INT' signal, `pgawk' exits. This is because these systems don't support the `kill' command, so the only signals you can deliver to a program are those generated by the keyboard. The `INT' signal is generated by the `Ctrl-' or `Ctrl-' key, while the `QUIT' signal is generated by the `Ctrl-<\>' key. 11 Running `awk' and `gawk' *************************** This major node covers how to run awk, both POSIX-standard and `gawk'-specific command-line options, and what `awk' and `gawk' do with non-option arguments. It then proceeds to cover how `gawk' searches for source files, obsolete options and/or features, and known bugs in `gawk'. This major node rounds out the discussion of `awk' as a program and as a language. While a number of the options and features described here were discussed in passing earlier in the book, this major node provides the full details. 11.1 Invoking `awk' =================== There are two ways to run `awk'--with an explicit program or with one or more program files. Here are templates for both of them; items enclosed in [...] in these templates are optional: awk [OPTIONS] -f progfile [`--'] FILE ... awk [OPTIONS] [`--'] 'PROGRAM' FILE ... Besides traditional one-letter POSIX-style options, `gawk' also supports GNU long options. It is possible to invoke `awk' with an empty program: awk '' datafile1 datafile2 Doing so makes little sense, though; `awk' exits silently when given an empty program. (d.c.) If `--lint' has been specified on the command line, `gawk' issues a warning that the program is empty. 11.2 Command-Line Options ========================= Options begin with a dash and consist of a single character. GNU-style long options consist of two dashes and a keyword. The keyword can be abbreviated, as long as the abbreviation allows the option to be uniquely identified. If the option takes an argument, then the keyword is either immediately followed by an equals sign (`=') and the argument's value, or the keyword and the argument's value are separated by whitespace. If a particular option with a value is given more than once, it is the last value that counts. Each long option for `gawk' has a corresponding POSIX-style option. The long and short options are interchangeable in all contexts. The options and their meanings are as follows: `-F FS' `--field-separator FS' Sets the `FS' variable to FS (*note Field Separators::). `-f SOURCE-FILE' `--file SOURCE-FILE' Indicates that the `awk' program is to be found in SOURCE-FILE instead of in the first non-option argument. `-v VAR=VAL' `--assign VAR=VAL' Sets the variable VAR to the value VAL _before_ execution of the program begins. Such variable values are available inside the `BEGIN' rule (*note Other Arguments::). The `-v' option can only set one variable, but it can be used more than once, setting another variable each time, like this: `awk -v foo=1 -v bar=2 ...'. *Caution:* Using `-v' to set the values of the built-in variables may lead to surprising results. `awk' will reset the values of those variables as it needs to, possibly ignoring any predefined value you may have given. `-mf N' `-mr N' Sets various memory limits to the value N. The `f' flag sets the maximum number of fields and the `r' flag sets the maximum record size. These two flags and the `-m' option are from the Bell Laboratories research version of Unix `awk'. They are provided for compatibility but otherwise ignored by `gawk', since `gawk' has no predefined limits. (The Bell Laboratories `awk' no longer needs these options; it continues to accept them to avoid breaking old programs.) `-W GAWK-OPT' Following the POSIX standard, implementation-specific options are supplied as arguments to the `-W' option. These options also have corresponding GNU-style long options. Note that the long options may be abbreviated, as long as the abbreviations remain unique. The full list of `gawk'-specific options is provided next. `--' Signals the end of the command-line options. The following arguments are not treated as options even if they begin with `-'. This interpretation of `--' follows the POSIX argument parsing conventions. This is useful if you have file names that start with `-', or in shell scripts, if you have file names that will be specified by the user that could start with `-'. The previous list described options mandated by the POSIX standard, as well as options available in the Bell Laboratories version of `awk'. The following list describes `gawk'-specific options: `-W compat' `-W traditional' `--compat' `--traditional' Specifies "compatibility mode", in which the GNU extensions to the `awk' language are disabled, so that `gawk' behaves just like the Bell Laboratories research version of Unix `awk'. `--traditional' is the preferred form of this option. *Note POSIX/GNU::, which summarizes the extensions. Also see *Note Compatibility Mode::. `-W copyright' `--copyright' Print the short version of the General Public License and then exit. `-W copyleft' `--copyleft' Just like `--copyright'. This option may disappear in a future version of `gawk'. `-W dump-variables[=FILE]' `--dump-variables[=FILE]' Prints a sorted list of global variables, their types, and final values to FILE. If no FILE is provided, `gawk' prints this list to the file named `awkvars.out' in the current directory. Having a list of all global variables is a good way to look for typographical errors in your programs. You would also use this option if you have a large program with a lot of functions, and you want to be sure that your functions don't inadvertently use global variables that you meant to be local. (This is a particularly easy mistake to make with simple variable names like `i', `j', etc.) `-W gen-po' `--gen-po' Analyzes the source program and generates a GNU `gettext' Portable Object file on standard output for all string constants that have been marked for translation. *Note Internationalization::, for information about this option. `-W help' `-W usage' `--help' `--usage' Prints a "usage" message summarizing the short and long style options that `gawk' accepts and then exit. `-W lint[=fatal]' `--lint[=fatal]' Warns about constructs that are dubious or nonportable to other `awk' implementations. Some warnings are issued when `gawk' first reads your program. Others are issued at runtime, as your program executes. With an optional argument of `fatal', lint warnings become fatal errors. This may be drastic, but its use will certainly encourage the development of cleaner `awk' programs. With an optional argument of `invalid', only warnings about things that are actually invalid are issued. (This is not fully implemented yet.) `-W lint-old' `--lint-old' Warns about constructs that are not available in the original version of `awk' from Version 7 Unix (*note V7/SVR3.1::). `-W non-decimal-data' `--non-decimal-data' Enable automatic interpretation of octal and hexadecimal values in input data (*note Nondecimal Data::). *Caution:* This option can severely break old programs. Use with care. `-W posix' `--posix' Operates in strict POSIX mode. This disables all `gawk' extensions (just like `--traditional') and adds the following additional restrictions: * `\x' escape sequences are not recognized (*note Escape Sequences::). * Newlines do not act as whitespace to separate fields when `FS' is equal to a single space (*note Fields::). * Newlines are not allowed after `?' or `:' (*note Conditional Exp::). * The synonym `func' for the keyword `function' is not recognized (*note Definition Syntax::). * The `**' and `**=' operators cannot be used in place of `^' and `^=' (*note Arithmetic Ops::, and also *note Assignment Ops::). * Specifying `-Ft' on the command-line does not set the value of `FS' to be a single TAB character (*note Field Separators::). * The `fflush' built-in function is not supported (*note I/O Functions::). If you supply both `--traditional' and `--posix' on the command line, `--posix' takes precedence. `gawk' also issues a warning if both options are supplied. `-W profile[=FILE]' `--profile[=FILE]' Enable profiling of `awk' programs (*note Profiling::). By default, profiles are created in a file named `awkprof.out'. The optional FILE argument allows you to specify a different file name for the profile file. When run with `gawk', the profile is just a "pretty printed" version of the program. When run with `pgawk', the profile contains execution counts for each statement in the program in the left margin, and function call counts for each function. `-W re-interval' `--re-interval' Allows interval expressions (*note Regexp Operators::) in regexps. Because interval expressions were traditionally not available in `awk', `gawk' does not provide them by default. This prevents old `awk' programs from breaking. `-W source PROGRAM-TEXT' `--source PROGRAM-TEXT' Allows you to mix source code in files with source code that you enter on the command line. Program source code is taken from the PROGRAM-TEXT. This is particularly useful when you have library functions that you want to use from your command-line programs (*note AWKPATH Variable::). `-W version' `--version' Prints version information for this particular copy of `gawk'. This allows you to determine if your copy of `gawk' is up to date with respect to whatever the Free Software Foundation is currently distributing. It is also useful for bug reports (*note Bugs::). As long as program text has been supplied, any other options are flagged as invalid with a warning message but are otherwise ignored. In compatibility mode, as a special case, if the value of FS supplied to the `-F' option is `t', then `FS' is set to the TAB character (`"\t"'). This is true only for `--traditional' and not for `--posix' (*note Field Separators::). The `-f' option may be used more than once on the command line. If it is, `awk' reads its program source from all of the named files, as if they had been concatenated together into one big file. This is useful for creating libraries of `awk' functions. These functions can be written once and then retrieved from a standard place, instead of having to be included into each individual program. (As mentioned in *Note Definition Syntax::, function names must be unique.) Library functions can still be used, even if the program is entered at the terminal, by specifying `-f /dev/tty'. After typing your program, type `Ctrl-d' (the end-of-file character) to terminate it. (You may also use `-f -' to read program source from the standard input but then you will not be able to also use the standard input as a source of data.) Because it is clumsy using the standard `awk' mechanisms to mix source file and command-line `awk' programs, `gawk' provides the `--source' option. This does not require you to pre-empt the standard input for your source code; it allows you to easily mix command-line and library source code (*note AWKPATH Variable::). If no `-f' or `--source' option is specified, then `gawk' uses the first non-option command-line argument as the text of the program source code. If the environment variable `POSIXLY_CORRECT' exists, then `gawk' behaves in strict POSIX mode, exactly as if you had supplied the `--posix' command-line option. Many GNU programs look for this environment variable to turn on strict POSIX mode. If `--lint' is supplied on the command line and `gawk' turns on POSIX mode because of `POSIXLY_CORRECT', then it issues a warning message indicating that POSIX mode is in effect. You would typically set this variable in your shell's startup file. For a Bourne-compatible shell (such as `bash'), you would add these lines to the `.profile' file in your home directory: POSIXLY_CORRECT=true export POSIXLY_CORRECT For a `csh'-compatible shell,(1) you would add this line to the `.login' file in your home directory: setenv POSIXLY_CORRECT true Having `POSIXLY_CORRECT' set is not recommended for daily use, but it is good for testing the portability of your programs to other environments. ---------- Footnotes ---------- (1) Not recommended. 11.3 Other Command-Line Arguments ================================= Any additional arguments on the command line are normally treated as input files to be processed in the order specified. However, an argument that has the form `VAR=VALUE', assigns the value VALUE to the variable VAR--it does not specify a file at all. (This was discussed earlier in *Note Assignment Options::.) All these arguments are made available to your `awk' program in the `ARGV' array (*note Built-in Variables::). Command-line options and the program text (if present) are omitted from `ARGV'. All other arguments, including variable assignments, are included. As each element of `ARGV' is processed, `gawk' sets the variable `ARGIND' to the index in `ARGV' of the current element. The distinction between file name arguments and variable-assignment arguments is made when `awk' is about to open the next input file. At that point in execution, it checks the file name to see whether it is really a variable assignment; if so, `awk' sets the variable instead of reading a file. Therefore, the variables actually receive the given values after all previously specified files have been read. In particular, the values of variables assigned in this fashion are _not_ available inside a `BEGIN' rule (*note BEGIN/END::), because such rules are run before `awk' begins scanning the argument list. The variable values given on the command line are processed for escape sequences (*note Escape Sequences::). (d.c.) In some earlier implementations of `awk', when a variable assignment occurred before any file names, the assignment would happen _before_ the `BEGIN' rule was executed. `awk''s behavior was thus inconsistent; some command-line assignments were available inside the `BEGIN' rule, while others were not. Unfortunately, some applications came to depend upon this "feature." When `awk' was changed to be more consistent, the `-v' option was added to accommodate applications that depended upon the old behavior. The variable assignment feature is most useful for assigning to variables such as `RS', `OFS', and `ORS', which control input and output formats before scanning the data files. It is also useful for controlling state if multiple passes are needed over a data file. For example: awk 'pass == 1 { PASS 1 STUFF } pass == 2 { PASS 2 STUFF }' pass=1 mydata pass=2 mydata Given the variable assignment feature, the `-F' option for setting the value of `FS' is not strictly necessary. It remains for historical compatibility. 11.4 The `AWKPATH' Environment Variable ======================================= The previous minor node described how `awk' program files can be named on the command-line with the `-f' option. In most `awk' implementations, you must supply a precise path name for each program file, unless the file is in the current directory. But in `gawk', if the file name supplied to the `-f' option does not contain a `/', then `gawk' searches a list of directories (called the "search path"), one by one, looking for a file with the specified name. The search path is a string consisting of directory names separated by colons. `gawk' gets its search path from the `AWKPATH' environment variable. If that variable does not exist, `gawk' uses a default path, `.:/usr/local/share/awk'.(1) (Programs written for use by system administrators should use an `AWKPATH' variable that does not include the current directory, `.'.) The search path feature is particularly useful for building libraries of useful `awk' functions. The library files can be placed in a standard directory in the default path and then specified on the command line with a short file name. Otherwise, the full file name would have to be typed for each file. By using both the `--source' and `-f' options, your command-line `awk' programs can use facilities in `awk' library files (*note Library Functions::). Path searching is not done if `gawk' is in compatibility mode. This is true for both `--traditional' and `--posix'. *Note Options::. NOTE: If you want files in the current directory to be found, you must include the current directory in the path, either by including `.' explicitly in the path or by writing a null entry in the path. (A null entry is indicated by starting or ending the path with a colon or by placing two colons next to each other (`::').) If the current directory is not included in the path, then files cannot be found in the current directory. This path search mechanism is identical to the shell's. Starting with version 3.0, if `AWKPATH' is not defined in the environment, `gawk' places its default search path into `ENVIRON["AWKPATH"]'. This makes it easy to determine the actual search path that `gawk' will use from within an `awk' program. While you can change `ENVIRON["AWKPATH"]' within your `awk' program, this has no effect on the running program's behavior. This makes sense: the `AWKPATH' environment variable is used to find the program source files. Once your program is running, all the files have been found, and `gawk' no longer needs to use `AWKPATH'. ---------- Footnotes ---------- (1) Your version of `gawk' may use a different directory; it will depend upon how `gawk' was built and installed. The actual directory is the value of `$(datadir)' generated when `gawk' was configured. You probably don't need to worry about this, though. 11.5 Obsolete Options and/or Features ===================================== This minor node describes features and/or command-line options from previous releases of `gawk' that are either not available in the current version or that are still supported but deprecated (meaning that they will _not_ be in the next release). For version 3.1 of `gawk', there are no deprecated command-line options from the previous version of `gawk'. The use of `next file' (two words) for `nextfile' was deprecated in `gawk' 3.0 but still worked. Starting with version 3.1, the two-word usage is no longer accepted. The process-related special files described in *Note Special Process::, work as described, but are now considered deprecated. `gawk' prints a warning message every time they are used. (Use `PROCINFO' instead; see *Note Auto-set::.) They will be removed from the next release of `gawk'. 11.6 Undocumented Options and Features ====================================== Use the Source, Luke! Obi-Wan This minor node intentionally left blank. 11.7 Known Bugs in `gawk' ========================= * The `-F' option for changing the value of `FS' (*note Options::) is not necessary given the command-line variable assignment feature; it remains only for backward compatibility. * Syntactically invalid single-character programs tend to overflow the parse stack, generating a rather unhelpful message. Such programs are surprisingly difficult to diagnose in the completely general case, and the effort to do so really is not worth it. 12 A Library of `awk' Functions ******************************* *Note User-defined::, describes how to write your own `awk' functions. Writing functions is important, because it allows you to encapsulate algorithms and program tasks in a single place. It simplifies programming, making program development more manageable, and making programs more readable. One valuable way to learn a new programming language is to _read_ programs in that language. To that end, this major node and *Note Sample Programs::, provide a good-sized body of code for you to read, and hopefully, to learn from. This major node presents a library of useful `awk' functions. Many of the sample programs presented later in this Info file use these functions. The functions are presented here in a progression from simple to complex. *Note Extract Program::, presents a program that you can use to extract the source code for these example library functions and programs from the Texinfo source for this Info file. (This has already been done as part of the `gawk' distribution.) If you have written one or more useful, general-purpose `awk' functions and would like to contribute them to the author's collection of `awk' programs, see *Note How To Contribute::, for more information. The programs in this major node and in *Note Sample Programs::, freely use features that are `gawk'-specific. Rewriting these programs for different implementations of awk is pretty straightforward. Diagnostic error messages are sent to `/dev/stderr'. Use `| "cat 1>&2"' instead of `> "/dev/stderr"' if your system does not have a `/dev/stderr', or if you cannot use `gawk'. A number of programs use `nextfile' (*note Nextfile Statement::) to skip any remaining input in the input file. *Note Nextfile Function::, shows you how to write a function that does the same thing. Finally, some of the programs choose to ignore upper- and lowercase distinctions in their input. They do so by assigning one to `IGNORECASE'. You can achieve almost the same effect(1) by adding the following rule to the beginning of the program: # ignore case { $0 = tolower($0) } Also, verify that all regexp and string constants used in comparisons use only lowercase letters. ---------- Footnotes ---------- (1) The effects are not identical. Output of the transformed record will be in all lowercase, while `IGNORECASE' preserves the original contents of the input record. 12.1 Naming Library Function Global Variables ============================================= Due to the way the `awk' language evolved, variables are either "global" (usable by the entire program) or "local" (usable just by a specific function). There is no intermediate state analogous to `static' variables in C. Library functions often need to have global variables that they can use to preserve state information between calls to the function--for example, `getopt''s variable `_opti' (*note Getopt Function::). Such variables are called "private", since the only functions that need to use them are the ones in the library. When writing a library function, you should try to choose names for your private variables that will not conflict with any variables used by either another library function or a user's main program. For example, a name like `i' or `j' is not a good choice, because user programs often use variable names like these for their own purposes. The example programs shown in this major node all start the names of their private variables with an underscore (`_'). Users generally don't use leading underscores in their variable names, so this convention immediately decreases the chances that the variable name will be accidentally shared with the user's program. In addition, several of the library functions use a prefix that helps indicate what function or set of functions use the variables--for example, `_pw_byname' in the user database routines (*note Passwd Functions::). This convention is recommended, since it even further decreases the chance of inadvertent conflict among variable names. Note that this convention is used equally well for variable names and for private function names as well.(1) As a final note on variable naming, if a function makes global variables available for use by a main program, it is a good convention to start that variable's name with a capital letter--for example, `getopt''s `Opterr' and `Optind' variables (*note Getopt Function::). The leading capital letter indicates that it is global, while the fact that the variable name is not all capital letters indicates that the variable is not one of `awk''s built-in variables, such as `FS'. It is also important that _all_ variables in library functions that do not need to save state are, in fact, declared local.(2) If this is not done, the variable could accidentally be used in the user's program, leading to bugs that are very difficult to track down: function lib_func(x, y, l1, l2) { ... USE VARIABLE some_var # some_var should be local ... # but is not by oversight } A different convention, common in the Tcl community, is to use a single associative array to hold the values needed by the library function(s), or "package." This significantly decreases the number of actual global names in use. For example, the functions described in *Note Passwd Functions::, might have used array elements `PW_data["inited"]', `PW_data["total"]', `PW_data["count"]', and `PW_data["awklib"]', instead of `_pw_inited', `_pw_awklib', `_pw_total', and `_pw_count'. The conventions presented in this minor node are exactly that: conventions. You are not required to write your programs this way--we merely recommend that you do so. ---------- Footnotes ---------- (1) While all the library routines could have been rewritten to use this convention, this was not done, in order to show how my own `awk' programming style has evolved and to provide some basis for this discussion. (2) `gawk''s `--dump-variables' command-line option is useful for verifying this. 12.2 General Programming ======================== This minor node presents a number of functions that are of general programming use. 12.2.1 Implementing `nextfile' as a Function -------------------------------------------- The `nextfile' statement, presented in *Note Nextfile Statement::, is a `gawk'-specific extension--it is not available in most other implementations of `awk'. This minor node shows two versions of a `nextfile' function that you can use to simulate `gawk''s `nextfile' statement if you cannot use `gawk'. A first attempt at writing a `nextfile' function is as follows: # nextfile --- skip remaining records in current file # this should be read in before the "main" awk program function nextfile() { _abandon_ = FILENAME; next } _abandon_ == FILENAME { next } Because it supplies a rule that must be executed first, this file should be included before the main program. This rule compares the current data file's name (which is always in the `FILENAME' variable) to a private variable named `_abandon_'. If the file name matches, then the action part of the rule executes a `next' statement to go on to the next record. (The use of `_' in the variable name is a convention. It is discussed more fully in *Note Library Names::.) The use of the `next' statement effectively creates a loop that reads all the records from the current data file. The end of the file is eventually reached and a new data file is opened, changing the value of `FILENAME'. Once this happens, the comparison of `_abandon_' to `FILENAME' fails, and execution continues with the first rule of the "real" program. The `nextfile' function itself simply sets the value of `_abandon_' and then executes a `next' statement to start the loop. This initial version has a subtle problem. If the same data file is listed _twice_ on the commandline, one right after the other or even with just a variable assignment between them, this code skips right through the file a second time, even though it should stop when it gets to the end of the first occurrence. A second version of `nextfile' that remedies this problem is shown here: # nextfile --- skip remaining records in current file # correctly handle successive occurrences of the same file # this should be read in before the "main" awk program function nextfile() { _abandon_ = FILENAME; next } _abandon_ == FILENAME { if (FNR == 1) _abandon_ = "" else next } The `nextfile' function has not changed. It makes `_abandon_' equal to the current file name and then executes a `next' statement. The `next' statement reads the next record and increments `FNR' so that `FNR' is guaranteed to have a value of at least two. However, if `nextfile' is called for the last record in the file, then `awk' closes the current data file and moves on to the next one. Upon doing so, `FILENAME' is set to the name of the new file and `FNR' is reset to one. If this next file is the same as the previous one, `_abandon_' is still equal to `FILENAME'. However, `FNR' is equal to one, telling us that this is a new occurrence of the file and not the one we were reading when the `nextfile' function was executed. In that case, `_abandon_' is reset to the empty string, so that further executions of this rule fail (until the next time that `nextfile' is called). If `FNR' is not one, then we are still in the original data file and the program executes a `next' statement to skip through it. An important question to ask at this point is: given that the functionality of `nextfile' can be provided with a library file, why is it built into `gawk'? Adding features for little reason leads to larger, slower programs that are harder to maintain. The answer is that building `nextfile' into `gawk' provides significant gains in efficiency. If the `nextfile' function is executed at the beginning of a large data file, `awk' still has to scan the entire file, splitting it up into records, just to skip over it. The built-in `nextfile' can simply close the file immediately and proceed to the next one, which saves a lot of time. This is particularly important in `awk', because `awk' programs are generally I/O-bound (i.e., they spend most of their time doing input and output, instead of performing computations). 12.2.2 Converting Strings To Numbers ------------------------------------ The `strtonum' function (*note String Functions::) is a `gawk' extension. The following function provides an implementation for other versions of `awk': # strtonum --- convert string to number function mystrtonum(str, ret, chars, n, i, k, c) { if (str ~ /^0[0-7]*$/) { # octal n = length(str) ret = 0 for (i = 1; i <= n; i++) { c = substr(str, i, 1) if ((k = index("01234567", c)) > 0) k-- # adjust for 1-basing in awk ret = ret * 8 + k } } else if (str ~ /^0[xX][0-9a-fA-f]+/) { # hexadecimal str = substr(str, 3) # lop off leading 0x n = length(str) ret = 0 for (i = 1; i <= n; i++) { c = substr(str, i, 1) c = tolower(c) if ((k = index("0123456789", c)) > 0) k-- # adjust for 1-basing in awk else if ((k = index("abcdef", c)) > 0) k += 9 ret = ret * 16 + k } } else if (str ~ /^[-+]?([0-9]+([.][0-9]*([Ee][0-9]+)?)?|([.][0-9]+([Ee][-+]?[0-9]+)?))$/) { # decimal number, possibly floating point ret = str + 0 } else ret = "NOT-A-NUMBER" return ret } # BEGIN { # gawk test harness # a[1] = "25" # a[2] = ".31" # a[3] = "0123" # a[4] = "0xdeadBEEF" # a[5] = "123.45" # a[6] = "1.e3" # a[7] = "1.32" # a[7] = "1.32E2" # # for (i = 1; i in a; i++) # print a[i], strtonum(a[i]), mystrtonum(a[i]) # } The function first looks for C-style octal numbers (base 8). If the input string matches a regular expression describing octal numbers, then `mystrtonum' loops through each character in the string. It sets `k' to the index in `"01234567"' of the current octal digit. Since the return value is one-based, the `k--' adjusts `k' so it can be used in computing the return value. Similar logic applies to the code that checks for and converts a hexadecimal value, which starts with `0x' or `0X'. The use of `tolower' simplifies the computation for finding the correct numeric value for each hexadecimal digit. Finally, if the string matches the (rather complicated) regex for a regular decimal integer or floating-point numer, the computation `ret = str + 0' lets `awk' convert the value to a number. A commented-out test program is included, so that the function can be tested with `gawk' and the results compared to the built-in `strtonum' function. 12.2.3 Assertions ----------------- When writing large programs, it is often useful to know that a condition or set of conditions is true. Before proceeding with a particular computation, you make a statement about what you believe to be the case. Such a statement is known as an "assertion". The C language provides an `' header file and corresponding `assert' macro that the programmer can use to make assertions. If an assertion fails, the `assert' macro arranges to print a diagnostic message describing the condition that should have been true but was not, and then it kills the program. In C, using `assert' looks this: #include int myfunc(int a, double b) { assert(a <= 5 && b >= 17.1); ... } If the assertion fails, the program prints a message similar to this: prog.c:5: assertion failed: a <= 5 && b >= 17.1 The C language makes it possible to turn the condition into a string for use in printing the diagnostic message. This is not possible in `awk', so this `assert' function also requires a string version of the condition that is being tested. Following is the function: # assert --- assert that a condition is true. Otherwise exit. function assert(condition, string) { if (! condition) { printf("%s:%d: assertion failed: %s\n", FILENAME, FNR, string) > "/dev/stderr" _assert_exit = 1 exit 1 } } END { if (_assert_exit) exit 1 } The `assert' function tests the `condition' parameter. If it is false, it prints a message to standard error, using the `string' parameter to describe the failed condition. It then sets the variable `_assert_exit' to one and executes the `exit' statement. The `exit' statement jumps to the `END' rule. If the `END' rules finds `_assert_exit' to be true, it exits immediately. The purpose of the test in the `END' rule is to keep any other `END' rules from running. When an assertion fails, the program should exit immediately. If no assertions fail, then `_assert_exit' is still false when the `END' rule is run normally, and the rest of the program's `END' rules execute. For all of this to work correctly, `assert.awk' must be the first source file read by `awk'. The function can be used in a program in the following way: function myfunc(a, b) { assert(a <= 5 && b >= 17.1, "a <= 5 && b >= 17.1") ... } If the assertion fails, you see a message similar to the following: mydata:1357: assertion failed: a <= 5 && b >= 17.1 There is a small problem with this version of `assert'. An `END' rule is automatically added to the program calling `assert'. Normally, if a program consists of just a `BEGIN' rule, the input files and/or standard input are not read. However, now that the program has an `END' rule, `awk' attempts to read the input data files or standard input (*note Using BEGIN/END::), most likely causing the program to hang as it waits for input. There is a simple workaround to this: make sure the `BEGIN' rule always ends with an `exit' statement. 12.2.4 Rounding Numbers ----------------------- The way `printf' and `sprintf' (*note Printf::) perform rounding often depends upon the system's C `sprintf' subroutine. On many machines, `sprintf' rounding is "unbiased," which means it doesn't always round a trailing `.5' up, contrary to naive expectations. In unbiased rounding, `.5' rounds to even, rather than always up, so 1.5 rounds to 2 but 4.5 rounds to 4. This means that if you are using a format that does rounding (e.g., `"%.0f"'), you should check what your system does. The following function does traditional rounding; it might be useful if your awk's `printf' does unbiased rounding: # round.awk --- do normal rounding function round(x, ival, aval, fraction) { ival = int(x) # integer part, int() truncates # see if fractional part if (ival == x) # no fraction return x if (x < 0) { aval = -x # absolute value ival = int(aval) fraction = aval - ival if (fraction >= .5) return int(x) - 1 # -2.5 --> -3 else return int(x) # -2.3 --> -2 } else { fraction = x - ival if (fraction >= .5) return ival + 1 else return ival } } # test harness { print $0, round($0) } 12.2.5 The Cliff Random Number Generator ---------------------------------------- The Cliff random number generator(1) is a very simple random number generator that "passes the noise sphere test for randomness by showing no structure." It is easily programmed, in less than 10 lines of `awk' code: # cliff_rand.awk --- generate Cliff random numbers BEGIN { _cliff_seed = 0.1 } function cliff_rand() { _cliff_seed = (100 * log(_cliff_seed)) % 1 if (_cliff_seed < 0) _cliff_seed = - _cliff_seed return _cliff_seed } This algorithm requires an initial "seed" of 0.1. Each new value uses the current seed as input for the calculation. If the built-in `rand' function (*note Numeric Functions::) isn't random enough, you might try using this function instead. ---------- Footnotes ---------- (1) `http://mathworld.wolfram.com/CliffRandomNumberGenerator.hmtl' 12.2.6 Translating Between Characters and Numbers ------------------------------------------------- One commercial implementation of `awk' supplies a built-in function, `ord', which takes a character and returns the numeric value for that character in the machine's character set. If the string passed to `ord' has more than one character, only the first one is used. The inverse of this function is `chr' (from the function of the same name in Pascal), which takes a number and returns the corresponding character. Both functions are written very nicely in `awk'; there is no real reason to build them into the `awk' interpreter: # ord.awk --- do ord and chr # Global identifiers: # _ord_: numerical values indexed by characters # _ord_init: function to initialize _ord_ BEGIN { _ord_init() } function _ord_init( low, high, i, t) { low = sprintf("%c", 7) # BEL is ascii 7 if (low == "\a") { # regular ascii low = 0 high = 127 } else if (sprintf("%c", 128 + 7) == "\a") { # ascii, mark parity low = 128 high = 255 } else { # ebcdic(!) low = 0 high = 255 } for (i = low; i <= high; i++) { t = sprintf("%c", i) _ord_[t] = i } } Some explanation of the numbers used by `chr' is worthwhile. The most prominent character set in use today is ASCII. Although an 8-bit byte can hold 256 distinct values (from 0 to 255), ASCII only defines characters that use the values from 0 to 127.(1) In the now distant past, at least one minicomputer manufacturer used ASCII, but with mark parity, meaning that the leftmost bit in the byte is always 1. This means that on those systems, characters have numeric values from 128 to 255. Finally, large mainframe systems use the EBCDIC character set, which uses all 256 values. While there are other character sets in use on some older systems, they are not really worth worrying about: function ord(str, c) { # only first character is of interest c = substr(str, 1, 1) return _ord_[c] } function chr(c) { # force c to be numeric by adding 0 return sprintf("%c", c + 0) } #### test code #### # BEGIN \ # { # for (;;) { # printf("enter a character: ") # if (getline var <= 0) # break # printf("ord(%s) = %d\n", var, ord(var)) # } # } An obvious improvement to these functions is to move the code for the `_ord_init' function into the body of the `BEGIN' rule. It was written this way initially for ease of development. There is a "test program" in a `BEGIN' rule, to test the function. It is commented out for production use. ---------- Footnotes ---------- (1) ASCII has been extended in many countries to use the values from 128 to 255 for country-specific characters. If your system uses these extensions, you can simplify `_ord_init' to simply loop from 0 to 255. 12.2.7 Merging an Array into a String ------------------------------------- When doing string processing, it is often useful to be able to join all the strings in an array into one long string. The following function, `join', accomplishes this task. It is used later in several of the application programs (*note Sample Programs::). Good function design is important; this function needs to be general but it should also have a reasonable default behavior. It is called with an array as well as the beginning and ending indices of the elements in the array to be merged. This assumes that the array indices are numeric--a reasonable assumption since the array was likely created with `split' (*note String Functions::): # join.awk --- join an array into a string function join(array, start, end, sep, result, i) { if (sep == "") sep = " " else if (sep == SUBSEP) # magic value sep = "" result = array[start] for (i = start + 1; i <= end; i++) result = result sep array[i] return result } An optional additional argument is the separator to use when joining the strings back together. If the caller supplies a nonempty value, `join' uses it; if it is not supplied, it has a null value. In this case, `join' uses a single blank as a default separator for the strings. If the value is equal to `SUBSEP', then `join' joins the strings with no separator between them. `SUBSEP' serves as a "magic" value to indicate that there should be no separation between the component strings.(1) ---------- Footnotes ---------- (1) It would be nice if `awk' had an assignment operator for concatenation. The lack of an explicit operator for concatenation makes string operations more difficult than they really need to be. 12.2.8 Managing the Time of Day ------------------------------- The `systime' and `strftime' functions described in *Note Time Functions::, provide the minimum functionality necessary for dealing with the time of day in human readable form. While `strftime' is extensive, the control formats are not necessarily easy to remember or intuitively obvious when reading a program. The following function, `gettimeofday', populates a user-supplied array with preformatted time information. It returns a string with the current time formatted in the same way as the `date' utility: # gettimeofday.awk --- get the time of day in a usable format # Returns a string in the format of output of date(1) # Populates the array argument time with individual values: # time["second"] -- seconds (0 - 59) # time["minute"] -- minutes (0 - 59) # time["hour"] -- hours (0 - 23) # time["althour"] -- hours (0 - 12) # time["monthday"] -- day of month (1 - 31) # time["month"] -- month of year (1 - 12) # time["monthname"] -- name of the month # time["shortmonth"] -- short name of the month # time["year"] -- year modulo 100 (0 - 99) # time["fullyear"] -- full year # time["weekday"] -- day of week (Sunday = 0) # time["altweekday"] -- day of week (Monday = 0) # time["dayname"] -- name of weekday # time["shortdayname"] -- short name of weekday # time["yearday"] -- day of year (0 - 365) # time["timezone"] -- abbreviation of timezone name # time["ampm"] -- AM or PM designation # time["weeknum"] -- week number, Sunday first day # time["altweeknum"] -- week number, Monday first day function gettimeofday(time, ret, now, i) { # get time once, avoids unnecessary system calls now = systime() # return date(1)-style output ret = strftime("%a %b %d %H:%M:%S %Z %Y", now) # clear out target array delete time # fill in values, force numeric values to be # numeric by adding 0 time["second"] = strftime("%S", now) + 0 time["minute"] = strftime("%M", now) + 0 time["hour"] = strftime("%H", now) + 0 time["althour"] = strftime("%I", now) + 0 time["monthday"] = strftime("%d", now) + 0 time["month"] = strftime("%m", now) + 0 time["monthname"] = strftime("%B", now) time["shortmonth"] = strftime("%b", now) time["year"] = strftime("%y", now) + 0 time["fullyear"] = strftime("%Y", now) + 0 time["weekday"] = strftime("%w", now) + 0 time["altweekday"] = strftime("%u", now) + 0 time["dayname"] = strftime("%A", now) time["shortdayname"] = strftime("%a", now) time["yearday"] = strftime("%j", now) + 0 time["timezone"] = strftime("%Z", now) time["ampm"] = strftime("%p", now) time["weeknum"] = strftime("%U", now) + 0 time["altweeknum"] = strftime("%W", now) + 0 return ret } The string indices are easier to use and read than the various formats required by `strftime'. The `alarm' program presented in *Note Alarm Program::, uses this function. A more general design for the `gettimeofday' function would have allowed the user to supply an optional timestamp value to use instead of the current time. 12.3 Data File Management ========================= This minor node presents functions that are useful for managing command-line data files. 12.3.1 Noting Data File Boundaries ---------------------------------- The `BEGIN' and `END' rules are each executed exactly once at the beginning and end of your `awk' program, respectively (*note BEGIN/END::). We (the `gawk' authors) once had a user who mistakenly thought that the `BEGIN' rule is executed at the beginning of each data file and the `END' rule is executed at the end of each data file. When informed that this was not the case, the user requested that we add new special patterns to `gawk', named `BEGIN_FILE' and `END_FILE', that would have the desired behavior. He even supplied us the code to do so. Adding these special patterns to `gawk' wasn't necessary; the job can be done cleanly in `awk' itself, as illustrated by the following library program. It arranges to call two user-supplied functions, `beginfile' and `endfile', at the beginning and end of each data file. Besides solving the problem in only nine(!) lines of code, it does so _portably_; this works with any implementation of `awk': # transfile.awk # # Give the user a hook for filename transitions # # The user must supply functions beginfile() and endfile() # that each take the name of the file being started or # finished, respectively. FILENAME != _oldfilename \ { if (_oldfilename != "") endfile(_oldfilename) _oldfilename = FILENAME beginfile(FILENAME) } END { endfile(FILENAME) } This file must be loaded before the user's "main" program, so that the rule it supplies is executed first. This rule relies on `awk''s `FILENAME' variable that automatically changes for each new data file. The current file name is saved in a private variable, `_oldfilename'. If `FILENAME' does not equal `_oldfilename', then a new data file is being processed and it is necessary to call `endfile' for the old file. Because `endfile' should only be called if a file has been processed, the program first checks to make sure that `_oldfilename' is not the null string. The program then assigns the current file name to `_oldfilename' and calls `beginfile' for the file. Because, like all `awk' variables, `_oldfilename' is initialized to the null string, this rule executes correctly even for the first data file. The program also supplies an `END' rule to do the final processing for the last file. Because this `END' rule comes before any `END' rules supplied in the "main" program, `endfile' is called first. Once again the value of multiple `BEGIN' and `END' rules should be clear. This version has same problem as the first version of `nextfile' (*note Nextfile Function::). If the same data file occurs twice in a row on the command line, then `endfile' and `beginfile' are not executed at the end of the first pass and at the beginning of the second pass. The following version solves the problem: # ftrans.awk --- handle data file transitions # # user supplies beginfile() and endfile() functions FNR == 1 { if (_filename_ != "") endfile(_filename_) _filename_ = FILENAME beginfile(FILENAME) } END { endfile(_filename_) } *Note Wc Program::, shows how this library function can be used and how it simplifies writing the main program. 12.3.2 Rereading the Current File --------------------------------- Another request for a new built-in function was for a `rewind' function that would make it possible to reread the current file. The requesting user didn't want to have to use `getline' (*note Getline::) inside a loop. However, as long as you are not in the `END' rule, it is quite easy to arrange to immediately close the current input file and then start over with it from the top. For lack of a better name, we'll call it `rewind': # rewind.awk --- rewind the current file and start over function rewind( i) { # shift remaining arguments up for (i = ARGC; i > ARGIND; i--) ARGV[i] = ARGV[i-1] # make sure gawk knows to keep going ARGC++ # make current file next to get done ARGV[ARGIND+1] = FILENAME # do it nextfile } This code relies on the `ARGIND' variable (*note Auto-set::), which is specific to `gawk'. If you are not using `gawk', you can use ideas presented in *Note Filetrans Function::, to either update `ARGIND' on your own or modify this code as appropriate. The `rewind' function also relies on the `nextfile' keyword (*note Nextfile Statement::). *Note Nextfile Function::, for a function version of `nextfile'. 12.3.3 Checking for Readable Data Files --------------------------------------- Normally, if you give `awk' a data file that isn't readable, it stops with a fatal error. There are times when you might want to just ignore such files and keep going. You can do this by prepending the following program to your `awk' program: # readable.awk --- library file to skip over unreadable files BEGIN { for (i = 1; i < ARGC; i++) { if (ARGV[i] ~ /^[A-Za-z_][A-Za-z0-9_]*=.*/ \ || ARGV[i] == "-") continue # assignment or standard input else if ((getline junk < ARGV[i]) < 0) # unreadable delete ARGV[i] else close(ARGV[i]) } } In `gawk', the `getline' won't be fatal (unless `--posix' is in force). Removing the element from `ARGV' with `delete' skips the file (since it's no longer in the list). 12.3.4 Checking For Zero-length Files ------------------------------------- All known `awk' implementations silently skip over zero-length files. This is a by-product of `awk''s implicit read-a-record-and-match-against-the-rules loop: when `awk' tries to read a record from an empty file, it immediately receives an end of file indication, closes the file, and proceeds on to the next command-line data file, _without_ executing any user-level `awk' program code. Using `gawk''s `ARGIND' variable (*note Built-in Variables::), it is possible to detect when an empty data file has been skipped. Similar to the library file presented in *Note Filetrans Function::, the following library file calls a function named `zerofile' that the user must provide. The arguments passed are the file name and the position in `ARGV' where it was found: # zerofile.awk --- library file to process empty input files BEGIN { Argind = 0 } ARGIND > Argind + 1 { for (Argind++; Argind < ARGIND; Argind++) zerofile(ARGV[Argind], Argind) } ARGIND != Argind { Argind = ARGIND } END { if (ARGIND > Argind) for (Argind++; Argind <= ARGIND; Argind++) zerofile(ARGV[Argind], Argind) } The user-level variable `Argind' allows the `awk' program to track its progress through `ARGV'. Whenever the program detects that `ARGIND' is greater than `Argind + 1', it means that one or more empty files were skipped. The action then calls `zerofile' for each such file, incrementing `Argind' along the way. The `Argind != ARGIND' rule simply keeps `Argind' up to date in the normal case. Finally, the `END' rule catches the case of any empty files at the end of the command-line arguments. Note that the test in the condition of the `for' loop uses the `<=' operator, not `<'. As an exercise, you might consider whether this same problem can be solved without relying on `gawk''s `ARGIND' variable. As a second exercise, revise this code to handle the case where an intervening value in `ARGV' is a variable assignment. 12.3.5 Treating Assignments as File Names ----------------------------------------- Occasionally, you might not want `awk' to process command-line variable assignments (*note Assignment Options::). In particular, if you have file names that contain an `=' character, `awk' treats the file name as an assignment, and does not process it. Some users have suggested an additional command-line option for `gawk' to disable command-line assignments. However, some simple programming with a library file does the trick: # noassign.awk --- library file to avoid the need for a # special option that disables command-line assignments function disable_assigns(argc, argv, i) { for (i = 1; i < argc; i++) if (argv[i] ~ /^[A-Za-z_][A-Za-z_0-9]*=.*/) argv[i] = ("./" argv[i]) } BEGIN { if (No_command_assign) disable_assigns(ARGC, ARGV) } You then run your program this way: awk -v No_command_assign=1 -f noassign.awk -f yourprog.awk * The function works by looping through the arguments. It prepends `./' to any argument that matches the form of a variable assignment, turning that argument into a file name. The use of `No_command_assign' allows you to disable command-line assignments at invocation time, by giving the variable a true value. When not set, it is initially zero (i.e., false), so the command-line arguments are left alone. 12.4 Processing Command-Line Options ==================================== Most utilities on POSIX compatible systems take options, or "switches," on the command line that can be used to change the way a program behaves. `awk' is an example of such a program (*note Options::). Often, options take "arguments"; i.e., data that the program needs to correctly obey the command-line option. For example, `awk''s `-F' option requires a string to use as the field separator. The first occurrence on the command line of either `--' or a string that does not begin with `-' ends the options. Modern Unix systems provide a C function named `getopt' for processing command-line arguments. The programmer provides a string describing the one-letter options. If an option requires an argument, it is followed in the string with a colon. `getopt' is also passed the count and values of the command-line arguments and is called in a loop. `getopt' processes the command-line arguments for option letters. Each time around the loop, it returns a single character representing the next option letter that it finds, or `?' if it finds an invalid option. When it returns -1, there are no options left on the command line. When using `getopt', options that do not take arguments can be grouped together. Furthermore, options that take arguments require that the argument is present. The argument can immediately follow the option letter, or it can be a separate command-line argument. Given a hypothetical program that takes three command-line options, `-a', `-b', and `-c', where `-b' requires an argument, all of the following are valid ways of invoking the program: prog -a -b foo -c data1 data2 data3 prog -ac -bfoo -- data1 data2 data3 prog -acbfoo data1 data2 data3 Notice that when the argument is grouped with its option, the rest of the argument is considered to be the option's argument. In this example, `-acbfoo' indicates that all of the `-a', `-b', and `-c' options were supplied, and that `foo' is the argument to the `-b' option. `getopt' provides four external variables that the programmer can use: `optind' The index in the argument value array (`argv') where the first nonoption command-line argument can be found. `optarg' The string value of the argument to an option. `opterr' Usually `getopt' prints an error message when it finds an invalid option. Setting `opterr' to zero disables this feature. (An application might want to print its own error message.) `optopt' The letter representing the command-line option. The following C fragment shows how `getopt' might process command-line arguments for `awk': int main(int argc, char *argv[]) { ... /* print our own message */ opterr = 0; while ((c = getopt(argc, argv, "v:f:F:W:")) != -1) { switch (c) { case 'f': /* file */ ... break; case 'F': /* field separator */ ... break; case 'v': /* variable assignment */ ... break; case 'W': /* extension */ ... break; case '?': default: usage(); break; } } ... } As a side point, `gawk' actually uses the GNU `getopt_long' function to process both normal and GNU-style long options (*note Options::). The abstraction provided by `getopt' is very useful and is quite handy in `awk' programs as well. Following is an `awk' version of `getopt'. This function highlights one of the greatest weaknesses in `awk', which is that it is very poor at manipulating single characters. Repeated calls to `substr' are necessary for accessing individual characters (*note String Functions::).(1) The discussion that follows walks through the code a bit at a time: # getopt.awk --- do C library getopt(3) function in awk # External variables: # Optind -- index in ARGV of first nonoption argument # Optarg -- string value of argument to current option # Opterr -- if nonzero, print our own diagnostic # Optopt -- current option letter # Returns: # -1 at end of options # ? for unrecognized option # a character representing the current option # Private Data: # _opti -- index in multi-flag option, e.g., -abc The function starts out with a list of the global variables it uses, what the return values are, what they mean, and any global variables that are "private" to this library function. Such documentation is essential for any program, and particularly for library functions. The `getopt' function first checks that it was indeed called with a string of options (the `options' parameter). If `options' has a zero length, `getopt' immediately returns -1: function getopt(argc, argv, options, thisopt, i) { if (length(options) == 0) # no options given return -1 if (argv[Optind] == "--") { # all done Optind++ _opti = 0 return -1 } else if (argv[Optind] !~ /^-[^: \t\n\f\r\v\b]/) { _opti = 0 return -1 } The next thing to check for is the end of the options. A `--' ends the command-line options, as does any command-line argument that does not begin with a `-'. `Optind' is used to step through the array of command-line arguments; it retains its value across calls to `getopt', because it is a global variable. The regular expression that is used, `/^-[^: \t\n\f\r\v\b]/', is perhaps a bit of overkill; it checks for a `-' followed by anything that is not whitespace and not a colon. If the current command-line argument does not match this pattern, it is not an option, and it ends option processing: if (_opti == 0) _opti = 2 thisopt = substr(argv[Optind], _opti, 1) Optopt = thisopt i = index(options, thisopt) if (i == 0) { if (Opterr) printf("%c -- invalid option\n", thisopt) > "/dev/stderr" if (_opti >= length(argv[Optind])) { Optind++ _opti = 0 } else _opti++ return "?" } The `_opti' variable tracks the position in the current command-line argument (`argv[Optind]'). If multiple options are grouped together with one `-' (e.g., `-abx'), it is necessary to return them to the user one at a time. If `_opti' is equal to zero, it is set to two, which is the index in the string of the next character to look at (we skip the `-', which is at position one). The variable `thisopt' holds the character, obtained with `substr'. It is saved in `Optopt' for the main program to use. If `thisopt' is not in the `options' string, then it is an invalid option. If `Opterr' is nonzero, `getopt' prints an error message on the standard error that is similar to the message from the C version of `getopt'. Because the option is invalid, it is necessary to skip it and move on to the next option character. If `_opti' is greater than or equal to the length of the current command-line argument, it is necessary to move on to the next argument, so `Optind' is incremented and `_opti' is reset to zero. Otherwise, `Optind' is left alone and `_opti' is merely incremented. In any case, because the option is invalid, `getopt' returns `?'. The main program can examine `Optopt' if it needs to know what the invalid option letter actually is. Continuing on: if (substr(options, i + 1, 1) == ":") { # get option argument if (length(substr(argv[Optind], _opti + 1)) > 0) Optarg = substr(argv[Optind], _opti + 1) else Optarg = argv[++Optind] _opti = 0 } else Optarg = "" If the option requires an argument, the option letter is followed by a colon in the `options' string. If there are remaining characters in the current command-line argument (`argv[Optind]'), then the rest of that string is assigned to `Optarg'. Otherwise, the next command-line argument is used (`-xFOO' versus `-x FOO'). In either case, `_opti' is reset to zero, because there are no more characters left to examine in the current command-line argument. Continuing: if (_opti == 0 || _opti >= length(argv[Optind])) { Optind++ _opti = 0 } else _opti++ return thisopt } Finally, if `_opti' is either zero or greater than the length of the current command-line argument, it means this element in `argv' is through being processed, so `Optind' is incremented to point to the next element in `argv'. If neither condition is true, then only `_opti' is incremented, so that the next option letter can be processed on the next call to `getopt'. The `BEGIN' rule initializes both `Opterr' and `Optind' to one. `Opterr' is set to one, since the default behavior is for `getopt' to print a diagnostic message upon seeing an invalid option. `Optind' is set to one, since there's no reason to look at the program name, which is in `ARGV[0]': BEGIN { Opterr = 1 # default is to diagnose Optind = 1 # skip ARGV[0] # test program if (_getopt_test) { while ((_go_c = getopt(ARGC, ARGV, "ab:cd")) != -1) printf("c = <%c>, optarg = <%s>\n", _go_c, Optarg) printf("non-option arguments:\n") for (; Optind < ARGC; Optind++) printf("\tARGV[%d] = <%s>\n", Optind, ARGV[Optind]) } } The rest of the `BEGIN' rule is a simple test program. Here is the result of two sample runs of the test program: $ awk -f getopt.awk -v _getopt_test=1 -- -a -cbARG bax -x -| c = , optarg = <> -| c = , optarg = <> -| c = , optarg = -| non-option arguments: -| ARGV[3] = -| ARGV[4] = <-x> $ awk -f getopt.awk -v _getopt_test=1 -- -a -x -- xyz abc -| c = , optarg = <> error--> x -- invalid option -| c = , optarg = <> -| non-option arguments: -| ARGV[4] = -| ARGV[5] = In both runs, the first `--' terminates the arguments to `awk', so that it does not try to interpret the `-a', etc., as its own options. Several of the sample programs presented in *Note Sample Programs::, use `getopt' to process their arguments. ---------- Footnotes ---------- (1) This function was written before `gawk' acquired the ability to split strings into single characters using `""' as the separator. We have left it alone, since using `substr' is more portable. 12.5 Reading the User Database ============================== The `PROCINFO' array (*note Built-in Variables::) provides access to the current user's real and effective user and group ID numbers, and if available, the user's supplementary group set. However, because these are numbers, they do not provide very useful information to the average user. There needs to be some way to find the user information associated with the user and group ID numbers. This minor node presents a suite of functions for retrieving information from the user database. *Note Group Functions::, for a similar suite that retrieves information from the group database. The POSIX standard does not define the file where user information is kept. Instead, it provides the `' header file and several C language subroutines for obtaining user information. The primary function is `getpwent', for "get password entry." The "password" comes from the original user database file, `/etc/passwd', which stores user information, along with the encrypted passwords (hence the name). While an `awk' program could simply read `/etc/passwd' directly, this file may not contain complete information about the system's set of users.(1) To be sure you are able to produce a readable and complete version of the user database, it is necessary to write a small C program that calls `getpwent'. `getpwent' is defined as returning a pointer to a `struct passwd'. Each time it is called, it returns the next entry in the database. When there are no more entries, it returns `NULL', the null pointer. When this happens, the C program should call `endpwent' to close the database. Following is `pwcat', a C program that "cats" the password database: /* * pwcat.c * * Generate a printable version of the password database */ #include #include int main(argc, argv) int argc; char **argv; { struct passwd *p; while ((p = getpwent()) != NULL) printf("%s:%s:%ld:%ld:%s:%s:%s\n", p->pw_name, p->pw_passwd, (long) p->pw_uid, (long) p->pw_gid, p->pw_gecos, p->pw_dir, p->pw_shell); endpwent(); return 0; } If you don't understand C, don't worry about it. The output from `pwcat' is the user database, in the traditional `/etc/passwd' format of colon-separated fields. The fields are: Login name The user's login name. Encrypted password The user's encrypted password. This may not be available on some systems. User-ID The user's numeric user ID number. Group-ID The user's numeric group ID number. Full name The user's full name, and perhaps other information associated with the user. Home directory The user's login (or "home") directory (familiar to shell programmers as `$HOME'). Login shell The program that is run when the user logs in. This is usually a shell, such as `bash'. A few lines representative of `pwcat''s output are as follows: $ pwcat -| root:3Ov02d5VaUPB6:0:1:Operator:/:/bin/sh -| nobody:*:65534:65534::/: -| daemon:*:1:1::/: -| sys:*:2:2::/:/bin/csh -| bin:*:3:3::/bin: -| arnold:xyzzy:2076:10:Arnold Robbins:/home/arnold:/bin/sh -| miriam:yxaay:112:10:Miriam Robbins:/home/miriam:/bin/sh -| andy:abcca2:113:10:Andy Jacobs:/home/andy:/bin/sh ... With that introduction, following is a group of functions for getting user information. There are several functions here, corresponding to the C functions of the same names: # passwd.awk --- access password file information BEGIN { # tailor this to suit your system _pw_awklib = "/usr/local/libexec/awk/" } function _pw_init( oldfs, oldrs, olddol0, pwcat, using_fw) { if (_pw_inited) return oldfs = FS oldrs = RS olddol0 = $0 using_fw = (PROCINFO["FS"] == "FIELDWIDTHS") FS = ":" RS = "\n" pwcat = _pw_awklib "pwcat" while ((pwcat | getline) > 0) { _pw_byname[$1] = $0 _pw_byuid[$3] = $0 _pw_bycount[++_pw_total] = $0 } close(pwcat) _pw_count = 0 _pw_inited = 1 FS = oldfs if (using_fw) FIELDWIDTHS = FIELDWIDTHS RS = oldrs $0 = olddol0 } The `BEGIN' rule sets a private variable to the directory where `pwcat' is stored. Because it is used to help out an `awk' library routine, we have chosen to put it in `/usr/local/libexec/awk'; however, you might want it to be in a different directory on your system. The function `_pw_init' keeps three copies of the user information in three associative arrays. The arrays are indexed by username (`_pw_byname'), by user ID number (`_pw_byuid'), and by order of occurrence (`_pw_bycount'). The variable `_pw_inited' is used for efficiency; `_pw_init' needs only to be called once. Because this function uses `getline' to read information from `pwcat', it first saves the values of `FS', `RS', and `$0'. It notes in the variable `using_fw' whether field splitting with `FIELDWIDTHS' is in effect or not. Doing so is necessary, since these functions could be called from anywhere within a user's program, and the u