gawk - pattern-scanning and processing language
gawk [ option ... ] -f program_file [ -- ] file ...
gawk [ option ... ] [ -- ] program_text file ...
The gawk utility is the GNU Project's implementation of the Awk programming language. It conforms to the definition of the language in the POSIX 1003.2 Command Language and Utilities Standard. This version, in turn, is based on the description in The AWK Programming Language, by Aho, Kernighan, and Weinberger, with the additional features found in the System V Release 4 version of UNIX awk. The gawk utility also provides more recent Bell Labs awk extensions and some GNU-specific extensions.
The command line consists of options to gawk itself, the Awk program text (if not supplied through the -f or --file options), and values to be made available in the ARGC and ARGV predefined Awk variables.
The gawk options can be either the traditional POSIX one-letter options, or the GNU-style long options. POSIX options start with a single "-", while long options start with "--". Long options are provided for both GNU-specific features and for POSIX mandated features.
Following the POSIX standard, gawk-specific options are supplied through arguments to the -W option. Multiple -W options can be supplied. Each -W option has a corresponding long option, as detailed later in this topic. Arguments to long options are either joined with the option by an equal sign (=) with no intervening spaces, or they can be provided in the next command-line argument. Long options can be abbreviated, but the abbreviation must remain unique.
In compatibility mode, any other options are flagged as illegal, but are otherwise ignored. In normal operation, as long as program text has been supplied, unknown options are passed on to the Awk program in the ARGV array for processing. This is particularly useful for running Awk programs through the "#!" executable interpreter mechanism.
An Awk program consists of a sequence of pattern-action statements and optional function definitions.
pattern { action_statement ...}
function name(parameter_list) {
statement ...}
The gawk utility first reads the program source from the program file or files specified as arguments to --source, or as the first non-option argument on the command line. The -f and --source options can be used multiple times on the command line. The gawk utility will read the program text as if all the program files and command-line source texts had been concatenated together. This is useful for building libraries of Awk functions without having to include them in each new Awk program that uses them. It also provides the ability to mix library functions with command-line programs.
The environment variable AWKPATH specifies a search path to use when finding source files named with the -f option. If this variable does not exist, the default path is /usr/local/share/awk. (The actual directory might vary, depending upon how gawk was built and installed.) If a file name given to the -f option contains a "/" character, no path search is performed.
The gawk utility executes Awk programs in the following order. First, all variable assignments specified with the -v option are performed. Next, gawk compiles the program into an internal form. Then, gawk executes the code in the BEGIN blocks (if any), and proceeds to read each file named in the ARGV array. If there are no files named on the command line, gawk reads the standard input.
If a file name on the command line has the form var=val, it is treated as a variable assignment. The variable var will be assigned the value val. (This happens after any BEGIN block has been run.) Command-line variable assignment is most useful for dynamically assigning values to the variables Awk uses to control how input is broken into fields and records. It is also useful for controlling state if multiple passes are needed over a single data file.
If the value of a particular element of ARGV is empty (""), gawk skips over it.
For each record in the input, gawk test it to determine whether it matches any pattern in the Awk program. For each pattern that the record matches, the associated action is executed. The patterns are tested in the order in which they occur in the program.
After all the input is exhausted, gawk executes the code in the END block(s) (if any).
Awk variables are dynamic; they come into existence when they are first used. Their values are either floating-point numbers, strings, or both, depending upon how they are used. Awk also has one-dimensional arrays; arrays with multiple dimensions can be simulated. Several predefined variables are set as a program runs. These will be described where they are most relevant and summarized in a later section of this topic.
Records are usually separated by newline characters. You can control how records are separated by assigning values to the built-in variable RS. If RS is any single character, that character separates records. Otherwise, RS is a regular expression. Text in the input that matches this regular expression will separate the record. In compatibility mode, however, only the first character of its string value is used for separating records. If RS is set to the null string, records are separated by blank lines. When RS is set to the null string, the newline character always acts as a field separator in addition to whatever value FS may have.
As each input record is read, gawk splits the record into fields, using the value of the FS variable as the field separator. If FS is a single character, fields are separated by that character. If FS is the null string, each individual character becomes a separate field. Otherwise, FS is expected to be a full regular expression. In the special case that FS is a single space, fields are separated by runs of spaces and/or tabs and/or newlines. (See also the discussion of --posix, which appears later in this topic). Note that the value of IGNORECASE will also affect how fields are split when FS is a regular expression, and how records are separated when RS is a regular expression.
If the FIELDWIDTHS variable is set to a space-separated list of numbers, each field is expected to have fixed width, and gawk will split up the record using the specified widths. The value of FS is ignored. Assigning a new value to FS overrides the use of FIELDWIDTHS and restores the default behavior.
Each field in the input record may be referenced by its position, $1, $2, and so on. $0 is the whole record. A field can also have an assigned value. Fields need not be referenced by constants. The following prints the fifth field in the input record:
n=5
print $n
The variable NF is set to the total number of fields in the input record. References to nonexistent fields (that is, fields after $NF) produce the null string. However, assigning to a nonexistent field (such as $(NF+2)=5) will increase the value of NF, create any intervening fields with the null string as their value, and cause the value of $0 to be recomputed, with the fields being separated by the value of OFS. References to negative-numbered fields cause a fatal error. Decrementing NF causes the values of fields past the new value to be lost, and the value of $0 to be recomputed, with the fields being separated by the value of OFS.
The gawk utility's built-in variables are as follows:
Arrays are subscripted with an expression between square brackets ([ and ]). If the expression is an expression list (expr, expr ...), the array subscript is a string consisting of the concatenation of the (string) value of each expression, separated by the value of the SUBSEP variable. This facility is used to simulate multiply dimensioned arrays. For example:
i = "A"; j = "B"; k = "C"
x[i, j, k] = "hello, world\n"
assigns the string hello, world\n to the element of the array x that is indexed by the string "A\034B\034C". All arrays in Awk are associative; that is, they are indexed by string values.
The special operator in can be used in an if or while statement to determine whether an array has an index consisting of a particular value.
if (val in array)
print array[val]
If the array has multiple subscripts, use (i, j) in array.
The in construct can also be used in a for loop to iterate over all the elements of an array.
An element can be deleted from an array using the delete statement. The delete statement can also be used to delete the entire contents of an array, just by specifying the array name without a subscript.
Variables and fields can be (floating point) numbers, strings, or both. How the value of a variable is interpreted depends upon its context. If used in a numeric expression, it will be treated as a number; if used as a string, it will be treated as a string.
To force a variable to be treated as a number, add 0 to it. To force it to be treated as a string, concatenate it with the null string.
When a string must be converted to a number, the conversion is accomplished using atof(3). A number is converted to a string by using the value of CONVFMT as a format string for sprintf(3), with the numeric value of the variable as the argument. However, even though all numbers in Awk are floating-point, integral values are always converted as integers. Thus, given the following:
CONVFMT = "%2.2f"
a = 12
b = a ""
the variable b has a string value of 12 and not 12.00.
The gawk utility performs comparisons as follows: If two variables are numeric, they are compared numerically. If one value is numeric and the other has a string value that is a "numeric string," comparisons are also done numerically. Otherwise, the numeric value is converted to a string, and a string comparison is performed. Two strings are compared as strings. According to the POSIX standard, even if two strings are numeric strings, a numeric comparison is performed. However, this is clearly incorrect, and gawk does not do this.
Note that string constants, such as 57, are not numeric strings; they are string constants. The idea of "numeric string" only applies to fields, getline input, FILENAME, ARGV elements, ENVIRON elements and the elements of an array created by split() that are numeric strings. The idea is that user input, and only user input that looks numeric, should be treated that way.
Uninitialized variables have the numeric value 0 and the string value "" (the null, or empty, string).
Awk is a line-oriented language. The pattern comes first, followed by the action. Action statements are enclosed in braces: { and }. Either the pattern can be missing, or the action can be missing, but both cannot be missing. If the pattern is missing, the action will be executed for every single record of input. A missing action is equivalent to:
{ print }
which prints the entire record.
Comments begin with the "#" character, and continue until the end of the line. Blank lines can be used to separate statements. A statement usually ends with a newline; however, this is not the case for lines ending in a ",", {, ?, :, &&, or ||. Lines ending in do or else also have their statements automatically continued on the following line. In other cases, a line can be continued by ending it with a "\", in which case the newline will be ignored.
Multiple statements can be put on one line by separating them with a ";". This applies to both the statements within the action part of a pattern-action pair (the usual case), and to the pattern-action statements themselves.
An Awk pattern can be one of the following:
BEGIN and END are two special kinds of patterns that are not tested against the input. The action parts of all BEGIN patterns are merged as if all the statements had been written in a single BEGIN block. They are executed before any of the input is read. Similarly, all the END blocks are merged, and executed when all the input is exhausted (or when an exit statement is executed). BEGIN and END patterns cannot be combined with other patterns in pattern expressions. BEGIN and END patterns cannot have missing action parts.
For /regular_expression/ patterns, the associated statement is executed for each input record that matches the regular expression. Regular expressions are the same as those in egrep(1), and are summarized later in this topic.
A relational_expression can use any of the operators defined in the section on actions, later in this topic. These generally test whether certain fields match certain regular expressions.
The &&, ||, and ! operators are logical AND, logical OR, and logical NOT, respectively, as in C. They perform short-circuit evaluation, also as in C, and are used for combining more primitive pattern expressions. As in most languages, parentheses can be used to change the order of evaluation.
The ?: operator is like the same operator in C. If the first pattern is true, the pattern used for testing is the second pattern; otherwise, the third is used. Only one of the second and third patterns is evaluated.
The pattern1, pattern2 form of an expression is called a range pattern. It matches all input records starting with a record that matches pattern1, continuing until and including a record that matches pattern2. It does not combine with any other sort of pattern expression.
Regular expressions are the extended kind found in egrep. They are composed of characters as follows:
The escape sequences that are valid in string constants (discussed later in this topic) are also legal in regular expressions.
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 where the actual characters themselves can vary among countries/regions and/or from character set to character set. For example, the notion of what defines an alphabetic character is not the same in the United States and France.
A character class is only valid in a regular expression inside the brackets of a character list. Character classes consist of [:, a keyword denoting the class, and :]. The character classes defined by the POSIX standard are as follows:
Before the POSIX standard, to match alphanumeric characters, you would have had to write /[A-Za-z0-9]/. If your character set included other alphabetic characters, this would not match them. With the POSIX character classes, you can write /[[:alnum:]]/, and this will match all 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, as well as several characters that are equivalent for collating or sorting purposes. For example, in French, the letters e and è are equivalent for sorting purposes.
A collating symbol is a multicharacter collating element enclosed in [. and .]. For example, if ch is a collating element, [[.ch.]] is a regular expression that matches this collating element, while [ch] is a regular expression that matches either c or h.
An equivalence class is a locale-specific name for a list of characters that are equivalent. The name is enclosed in [= and =]. For example, the name e might be used to represent all of the letters e, é, and è. In this case, [[=e=]] is a regular expression that matches any of the following:
These features are very valuable in non-English-speaking locales. 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.
The \y, \B, \<, \>, \w, \W, \`, and \' operators are specific to gawk; they are extensions based on facilities in the GNU regular-expression libraries.
The various command-line options control how gawk interprets characters in regular expressions.
Action statements are enclosed in braces ({}). Action statements consist of the usual assignment, conditional, and looping statements found in most languages. The operators, control statements, and input/output statements available are patterned after those in C.
Following is a list of the operators in Awk, in order of decreasing precedence:
Do not use a constant regular expression (/foo/) on the left-hand side of a ~ or !~. The expression /foo/ ~ exp has the same meaning as (($0 ~ /foo/) ~ exp). This is usually not what was intended.
The control statements are as follows:
The input/output (I/O) statements are as follows:
Earlier versions of gawk used next file as two words. While this usage is still recognized, it generates a warning message and will eventually be removed.
Other I/O redirections are also allowed. For print and printf, >>file appends output to the file, while | command writes on a pipe. In a similar fashion, command | getline pipes into getline. The getline command will return 0 on end of file, and -1 on an error.
The Awk versions of the printf statement and sprintf() function (see discussion later in this topic) accept the following conversion-specification formats:
There are optional, additional parameters that can lie between the % and the control letter:
The dynamic width and precision capabilities of the ANSI C printf(3) routines are supported. A * in place of either the width or precision specifications will cause their values to be taken from the argument list to printf or sprintf().
When performing I/O redirection from either print or printf into a file, or with getline from a file, gawk recognizes certain special file names internally. These file names allow access to open file descriptors inherited from gawk's parent process (usually the shell). Other special file names provide access to information about the running gawk process. The file names are:
These are particularly useful for error messages. For example, you could use the following to display an error message on standard error:
print "Unexpected error" > "/dev/stderr"
Without these special file names, you would have to use the following instead.
print "Unexpected error" | "cat 1>&2"
These file names can also be used on the command line to name data files.
Awk has the following predefined arithmetic functions:
The gawk utility has the following predefined string functions:
One of the primary uses of Awk programs is processing log files that contain time stamp information. The gawk utility provides the following two functions for obtaining time stamps and formatting them.
String constants in Awk are sequences of characters enclosed between double quotes (”). Within strings, certain escape sequences are recognized, as in C. These are:
The escape sequences can also be used inside constant regular expressions (/[ \f\n\r\v]/ matches white-space characters).
In compatibility mode, the characters represented by octal, and hexadecimal escape sequences are treated literally when used in regular expression constants. Thus, /a\52b/ is equivalent to /a\*b/.
Functions in Awk are defined as follows:
function name(parameter list) { statements }
Functions are executed when they are called from within expressions in either patterns or actions. Actual parameters supplied in the function call are used to instantiate the formal parameters declared in the function. Arrays are passed by reference; other variables are passed by value.
Since functions were not originally part of the Awk language, the provision for local variables is rather clumsy; they are declared as extra parameters in the parameter list. The convention is to separate local variables from real parameters by including extra spaces in the parameter list, as shown in the following example:
function f(p, q, a, b) # a & b are local
{
.....
}
/abc/ { ... ; f(1, 2) ; ... }
The left parenthesis in a function call is required to immediately follow the function name, without any intervening white space. This is to prevent a syntactic ambiguity with the concatenation operator. This restriction does not apply to the built-in functions listed earlier.
Functions can call each other and can be recursive. Function parameters used as local variables are initialized to the null string and the number zero upon function invocation.
Use return expr to return a value from a function. The return value is undefined if no value is provided, or if the function returns by "falling off" the end.
If --lint has been provided, gawk will warn about calls to undefined functions at parse time instead of at run time. Calling an undefined function at run time is a fatal error.
The word func can be used in place of function.
Print and sort the log on names of all users:
BEGIN { FS = ":" }
{ print $1 | "sort" }
Count lines in a file:
{ nlines++ }
END { print nlines }
Precede each line by its number in the file:
{ print FNR, $0 }
Concatenate and line number (a variation on a theme):
{ print NR, $0 }
A primary goal for gawk is compatibility with the POSIX standard, as well as with the latest version of UNIX awk. To this end, gawk incorporates the following user-visible features, which are not described in the Awk book, but are part of the Bell Labs version of awk, and are in the POSIX standard.
The -v option for assigning variables before program execution starts is new. The book indicates that command-line variable assignment occurs when awk would otherwise open the argument as a file, which is after the BEGIN block is executed. In earlier implementations, however, when such an assignment appeared before any file names, the assignment would occur before the BEGIN block was run. Applications came to depend on this "feature." When awk was changed to match its documentation, this option was added to accommodate applications that depended upon the old behavior. (This feature was agreed upon by both the AT&T and GNU developers.)
The -W option for implementation-specific features is from the POSIX standard.
When processing arguments, gawk uses the special option "--" to signal the end of arguments. In compatibility mode, it will warn about, but otherwise ignore, undefined options. In normal operation, such arguments are passed on to the Awk program for it to process.
The Awk book does not define the return value of srand(). The POSIX standard has it return the seed it was using to allow keeping track of random number sequences. Therefore, srand() in gawk also returns its current seed.
Other new features are: The use of multiple -f options (from MKS awk); the ENVIRON array; the \a, and \v escape sequences (done originally in gawk and fed back into AT&T's); the tolower() and toupper() built-in functions (from AT&T); and the ANSI C conversion specifications in printf (done first in AT&T's version).
The gawk utility includes a number of extensions to POSIX awk. They are described in this section. All of the extensions described here can be disabled by invoking gawk with the --traditional option.
The following features of gawk are not available in POSIX awk.
The Awk book does not define the return value of the close() function. The gawk utility's close() returns the value from fclose(3), or pclose(3), when closing a file or pipe, respectively.
When gawk is invoked with the --traditional option, if the fs argument to the -F option is t, FS will be set to the tab character. Note that typing gawk -F\t ... simply causes the shell to quote the t, and does not pass \t to the -F option. Since this is a rather ugly special case, it is not the default behavior. This behavior also does not occur if --posix has been specified. To really get a tab character as the field separator, it is best to use quotes: gawk -F'\t' ....
There are two features of historical Awk implementations that gawk supports. First, it is possible to call the length() built-in function not only with no argument, but even without parentheses. Thus, the following:
a = length # No argument here!
is the same as either:
a = length()
a = length($0)
This feature is marked as "deprecated" in the POSIX standard, and gawk will issue a warning about its use if --lint is specified on the command line.
The other feature is the use of either the continue or the break statements outside the body of a while, for, or do loop. Traditional Awk implementations have treated such usage as equivalent to the next statement. The gawk utility will support this usage if --traditional has been specified.
If POSIXLY_CORRECT exists in the environment, then gawk behaves exactly as if --posix had been specified on the command line. If --lint has been specified, gawk will issue a warning message to this effect.
The AWKPATH environment variable can be used to provide a list of directories that gawk will search when looking for files named through the -f and --file options.
The -F option is not necessary given the command-line variable assignment feature; it remains only for backwards compatibility.
If your system actually has support for /dev/fd and the associated /dev/stdin, /dev/stdout, and /dev/stderr files, you might get different output from gawk than you would get on a system without those files. When gawk interprets these files internally, it synchronizes output to the standard output with output to /dev/stdout, while on a system with those files, the output is actually to different open files.
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.
This topic documents gawk, version 3.0.4.
The original version of UNIX awk was designed and implemented by Alfred Aho, Peter Weinberger, and Brian Kernighan of AT&T Bell Labs. Brian Kernighan continues to maintain and enhance it.
Paul Rubin and Jay Fenlason of the Free Software Foundation wrote gawk to be compatible with the original version of awk distributed in Seventh Edition UNIX. John Woods contributed a number of bug fixes. David Trueman, with contributions from Arnold Robbins, made gawk compatible with the new version of UNIX awk. Arnold Robbins is the current maintainer.
The initial DOS port was done by Conrad Kwok and Scott Garfinkle. Scott Deifik is the current DOS maintainer. Pat Rankin did the port to VMS, and Michal Jaegermann did the port to the Atari ST. The port to OS/2 was done by Kai Uwe Rommel, with contributions and help from Darrel Hankerson. Fred Fish supplied support for the Amiga.
If you find a bug in gawk, please send electronic mail to bug-gnu-utils@gnu.org, with a carbon copy to arnold@gnu.org. Please include your operating system and its revision, the version of gawk, which C compiler you used to compile it, and a test program and data that are as small as possible for reproducing the problem.
Before sending a bug report, please do two things. First, verify that you have the latest version of gawk. Many bugs (usually subtle ones) are fixed at each release, and if yours is out of date, the problem may already have been solved. Second, please read this topic, the manual page, and the reference manual carefully to ensure that what you think is a bug really is, instead of just a quirk in the language.
Do not post a bug report in comp.lang.awk. While the gawk developers occasionally read this newsgroup, posting bug reports there is an unreliable way to report bugs. Instead, please use the electronic mail addresses given above.
Brian Kernighan of Bell Labs provided valuable assistance during testing and debugging. We thank him.
Copyright © 1996,97,98,99 Free Software Foundation, Inc.
Permission is granted to make and distribute verbatim copies of this manual page provided the copyright notice and this permission notice are preserved on all copies.
Permission is granted to copy and distribute modified versions of this manual page under the conditions for verbatim copying, provided that the entire resulting derived work is distributed under the terms of a permission notice identical to this one.
Permission is granted to copy and distribute translations of this manual page into another language, under the above conditions for modified versions, except that this permission notice may be stated in a translation approved by the Foundation.