ctags [-a][-f tagsfile]
pathname ...
The ctags utility shall be provided on systems that support the User Portability Utilities option, the Software Development Utilities option, and either or both of the C-Language Development Utilities option and FORTRAN Development Utilities option. On other systems, it is optional.
The ctags utility shall write a tagsfile or an index of objects from C-language or FORTRAN source files specified by the pathname operands. The tagsfile shall list the locators of language-specific objects within the source files. A locator consists of a name, pathname, and either a search pattern or a line number that can be used in searching for the object definition. The objects that shall be recognized are specified in the EXTENDED DESCRIPTION section.
The ctags utility shall conform to the Base Definitions volume of IEEE Std 1003.1-2001, Section 12.2, Utility Syntax Guidelines.
The following options shall be supported:
The following pathname operands are supported:
The handling of other files is implementation-defined.
The input files shall be text files containing source code in the language indicated by the operand filename suffixes.
The following environment variables shall affect the execution of ctags:
Determine the order in which output is sorted for the -x option. The POSIX locale determines the order in which the tagsfile is written.
The list of object name information produced by the -x option shall be written to standard output in the following format:
"%s %d %s %s", <object-name>, <line-number>, <filename>, <text>
where <text> is the text of line <line-number> of file <filename>.
The standard error shall be used only for diagnostic messages.
When the -x option is not specified, the format of the output file shall be:
"%s\t%s\t/%s/\n", <identifier>, <filename>, <pattern>
where <pattern> is a search pattern that could be used by an editor to find the defining instance of <identifier> in <filename> (where defining instance is indicated by the declarations listed in the EXTENDED DESCRIPTION).
An optional circumflex ( '^' ) can be added as a prefix to <pattern>,
and an optional dollar sign can be
appended to <pattern> to indicate that the pattern is anchored
to the beginning (end) of a line of text. Any slash or
backslash characters in <pattern> shall be preceded by a backslash
character. The anchoring circumflex, dollar sign,
and escaping backslash characters shall not be considered part of
the search pattern. All other characters in the search pattern
shall be considered literal characters.
An alternative format is:
"%s\t%s\t?%s?\n", <identifier>, <filename>, <pattern>
which is identical to the first format except that slashes in <pattern> shall not be preceded by escaping backslash characters, and question mark characters in <pattern> shall be preceded by backslash characters.
A second alternative format is:
"%s\t%s\t%d\n", <identifier>, <filename>, <lineno>
where <lineno> is a decimal line number that could be used by an editor to find <identifier> in <filename>.
Neither alternative format shall be produced by ctags when it is used as described by IEEE Std 1003.1-2001, but the standard utilities that process tags files shall be able to process those formats as well as the first format.
In any of these formats, the file shall be sorted by identifier, based on the collation sequence in the POSIX locale.
If the operand identifies C-language source, the ctags utility shall attempt to produce an output line for each of the following objects:
It may also produce output for any of the following objects:
Any #if and #ifdef statements shall produce no output. The tag main is treated specially in C programs. The tag formed shall be created by prefixing M to the name of the file, with the trailing .c, and leading pathname components (if any) removed.
On systems that do not support the C-Language Development Utilities option, ctags produces unspecified results for C-language source code files. It should write to standard error a message identifying this condition and cause a non-zero exit status to be produced.
If the operand identifies FORTRAN source, the ctags utility shall produce an output line for each function definition. It may also produce output for any of the following objects:
On systems that do not support the FORTRAN Development Utilities option, ctags produces unspecified results for FORTRAN source code files. It should write to standard error a message identifying this condition and cause a non-zero exit status to be produced.
It is implementation-defined what other objects (including duplicate identifiers) produce output.
The following exit values shall be returned:
Default.
The following sections are informative.
The output with -x is meant to be a simple index that can be written out as an off-line readable function index. If the input files to ctags (such as .c files) were not created using the same locale as that in effect when ctags -x is run, results might not be as expected.
The description of C-language processing says "attempts to" because the C language can be greatly confused, especially through the use of #defines, and this utility would be of no use if the real C preprocessor were run to identify them. The output from ctags may be fooled and incorrect for various constructs.
The option list was significantly reduced from that provided by historical implementations. The -F option was omitted as redundant, since it is the default. The -B option was omitted as being of very limited usefulness. The -t option was omitted since the recognition of typedefs is now required for C source files. The -u option was omitted because the update function was judged to be not only inefficient, but also rarely needed.
An early proposal included a -w option to suppress warning diagnostics. Since the types of such diagnostics could not be described, the option was omitted as being not useful.
The text for LC_CTYPE about compatibility with the C locale acknowledges that the ISO C standard imposes requirements on the locale used to process C source. This could easily be a superset of that known as "the C locale" by way of implementation extensions, or one of a few alternative locales for systems supporting different codesets. No statement is made for FORTRAN because the ANSI X3.9-1978 standard (FORTRAN 77) does not (yet) define a similar locale concept. However, a general rule in this volume of IEEE Std 1003.1-2001 is that any time that locales do not match (preparing a file for one locale and processing it in another), the results are suspect.
The collation sequence of the tags file is not affected by LC_COLLATE because it is typically not used by human readers, but only by programs such as vi to locate the tag within the source files. Using the POSIX locale eliminates some of the problems of coordinating locales between the ctags file creator and the vi file reader.
Historically, the tags file has been used only by ex and vi. However, the format of the tags file has been published to encourage other programs to use the tags in new ways. The format allows either patterns or line numbers to find the identifiers because the historical vi recognizes either. The ctags utility does not produce the format using line numbers because it is not useful following any source file changes that add or delete lines. The documented search patterns match historical practice. It should be noted that literal leading circumflex or trailing dollar-sign characters in the search pattern will only behave correctly if anchored to the beginning of the line or end of the line by an additional circumflex or dollar-sign character.
Historical implementations also understand the objects used by the languages Pascal and sometimes LISP, and they understand the C source output by lex and yacc. The ctags utility is not required to accommodate these languages, although implementors are encouraged to do so.
The following historical option was not specified, as vgrind is not included in this volume of IEEE Std 1003.1-2001:
ctags -v files | sort -f > index vgrind -x index
The special treatment of the tag main makes the use of ctags practical in directories with more than one program.