readtags - Man Page

The behavior of reading tags can be controlled using these options:

-t TAGFILE, --tag-file TAGFILE

Use specified tag file (default: "tags"). Giving "-" as TAGFILE indicates reading the tags file content from the standard input. "-" can make the command line simpler. However, it doesn't mean efficient; readtags stores the data to a temorary file and reads that file for taking the ACTION.

-s[0|1|2], --override-sort-detection METHOD

Override sort detection of tag file. METHOD: unsorted|sorted|foldcase

The NAME action will perform binary search on sorted (including "foldcase") tags files, which is much faster then on unsorted tags files.

The behavior of the NAME action can be controlled using these options:

-i,  --icase-match

Perform case-insensitive matching in the NAME action.

-p,  --prefix-match

Perform prefix matching in the NAME action.

By default, the output of readtags contains only the name, input and pattern field. The Output can be tweaked using these options:

-A,  --absolute-input

Do the same as -C option but use only absolute path form.

-C,  --canonicalize-input

Resolve '..' and '.' in input fields of regular tags. This produces a unique representation of the input path. This option works only with tags files having !_TAG_PROC_CWD pseudo tag.

NOTE: The current implementation accepts only !_TAG_PROC_CWD starting with /; a Windows directory name starting with a drive letter like C:\Somewhere is not acceptable.

-d,  --debug

Turn on debugging output.

-E,  --escape-output

Escape characters like tabs in output as described in tags(5).

-e,  --extension-fields

Include extension fields in output.

-n,  --line-number

Also include the line number field when -e option is give.

-P,  --with-pseudo-tags

List pseudo tags as if -D option is specified but continues processing without exiting.

About the -E option: certain characters are escaped in a tags file, to make it machine-readable. e.g., ensuring no tabs character appear in fields other than the pattern field. By default, readtags translates them to make it human-readable, but when utilizing readtags output in a script or a client tool, -E option should be used. See ctags-client-tools(7) for more discussion on this.

About printing input fields ({tagfile} in tags(5)) with -E option: readtags always prints the input field literally (as it is in the tags file), and when ctags writes the tags file, the escaping rules are applied only when TAG_OUTPUT_MODE pseudo tag has "u-ctags" and TAG_OUTPUT_FILESEP has "slash" as values for their input fields, as explained in ctags-client-tools(7).

Further filtering, sorting, and formatting on the tags listed by actions are performed using:

-Q EXP, --filter EXP

Filter the tags listed by ACTION with EXP before printing.

-S EXP, --sorter EXP

Sort the tags listed by ACTION with EXP before printing.

-F EXP, --formatter EXP

Format the tags listed by ACTION with EXP when printing.

These are discussed in the Expression section.

• List all tags in "/path/to/tags":

  $ readtags -t /path/to/tags -l

• List all tags in "tags" that start with "mymethod":

  $ readtags -p - mymethod

• List all tags matching "mymethod", case insensitively:

  $ readtags -i - mymethod

• List all tags start with "myvar", and printing all fields (i.e., the whole line):

  $ readtags -p -ne - myvar

The tag entries that make the filter expression produces true value are printed by readtags.

The basic operators for filtering are eq?, prefix?, suffix?, substr?, and #/PATTERN/. Language common fields can be accessed using variables starting with $, e.g., $language represents the language field. For example:

• List all tags start with "myfunc" in Python code files:

  $ readtags -p -Q '(eq? $language "Python")' - myfunc

downcase or upcase operators can be used to perform case-insensitive matching:

• List all tags containing "my", case insensitively:

  $ readtags -Q '(substr? (downcase $name) "my")' -l

We have logical operators like and, or and not. The value of a missing field is #f, so we could deal with missing fields:

• List all tags containing "impl" in Python code files, but allow the language: field to be missing:

  $ readtags -Q '(and (substr? $name "impl")\
                      (or (not $language)\
                          (eq? $language "Python")))' -l

#/PATTERN/ is for the case when string predicates (prefix?, suffix?, and substr?) are not enough. You can use "Posix extended regular expression" as PATTERN.

• List all tags inherits from the class "A":

  $ readtags -Q '(#/(^|,) ?A(,|$)/ $inherits)' -l

Here $inherits is a comma-separated class list like "A,B,C", "P, A, Q", or just "A". Notice that this filter works on both situations where there's a space after each comma or there's not.

Case-insensitive matching can be performed by #/PATTERN/i:

• List all tags inherits from the class "A" or "a":

  $ readtags -Q '(#/(^|,) ?A(,|$)/i $inherits)' -l

To include "/" in a pattern, prefix \ to the "/".

NOTE: The above regular expression pattern for inspecting inheritances is just an example to show how to use #/PATTERN/ expression. Tags file generators have no consensus about the format of inherits:, e.g., whether there should be a space after a comma. Even parsers in ctags have no consensus. Noticing the format of the inherits: field of specific languages is needed for such queries.

The expressions #/PATTERN/ and #/PATTERN/i are for interactive use. Readtags also offers an alias string->regexp, so #/PATTERN/ is equal to (string->regexp "PATTERN"), and #/PATTERN/i is equal to (string->regexp "PATTERN" :case-fold #t). string->regexp doesn't need to prefix \ for including "/" in a pattern. string->regexp may simplify a client tool building an expression. See also ctags-client-tools(7) for building expressions in your tool.

Let's now consider missing fields. The tags file may have tag entries that has no inherits: field. In that case $inherits is #f, and the regular expression matching raises an error, since string operators only work for strings. To avoid this problem:

• Safely list all tags inherits from the class "A":

  $ readtags -Q '(and $inherits (#/(^|,) ?A(,|$)/ $inherits))' -l

This makes sure $inherits is not missing first, then match it by regexp.

Sometimes you want to keep tags where the field is missing. For example, your want to exclude reference tags, which is marked by the extras: field, then you want to keep tags who doesn't have extras: field since they are also not reference tags. Here's how to do it:

• List all tags but the reference tags:

  $ readtags -Q '(or (not $extras) (#/(^|,) ?reference(,|$)/ $extras))' -l

Notice that (not $extras) produces #t when $extras is missing, so the whole or expression produces #t.

The combination of ctags -o - and readtags -t - is handy for inspecting a source file as far as the source file is enough short.

• List all the large (> 100 lines) functions in a file:

  $ ctags -o - --fields=+neKz input.c \
    | ./readtags -t - -en \
                 -Q '(and (eq? $kind "function") $end $line (> (- $end $line) 100))' \
                 -l

• List all the tags including line 80 in a file:

  $ ctags -o - --fields=+neKz input.c \
    | readtags -t - -ne \
               -Q '(and $line
                        (or (eq? $line 80)
                            (and $end (< $line 80) (< 80 $end))))' \
      -l

Run "readtags -H filter" to know about all valid functions and variables.

When sorting, the sorter expression is evaluated on two tag entries to decide which should sort before the other one, until the order of all tag entries is decided.

In a sorter expression, $ and & are used to access the fields in the two tag entries, and let's call them $-entry and &-entry. The sorter expression should have a value of -1, 0 or 1. The value -1 means the $-entry should be put above the &-entry, 1 means the contrary, and 0 makes their order in the output uncertain.

The core operator of sorting is <>. It's used to compare two strings or two numbers (numbers are for the line: or end: fields). In (<> a b), if a < b, the result is -1; a > b produces 1, and a = b produces 0. Strings are compared using the strcmp function, see strcmp(3).

For example, sort by names, and make those shorter or alphabetically smaller ones appear before the others:

$ readtags -S '(<> $name &name)' -l

This reads "If the tag name in the $-entry is smaller, it goes before the &-entry".

The <or> operator is used to chain multiple expressions until one returns -1 or 1. For example, sort by input file names, then line numbers if in the same file:

$ readtags -S '(<or> (<> $input &input) (<> $line &line))' -l

The *- operator is used to flip the compare result. i.e., (*- (<> a b)) is the same as (<> b a).

Filter expressions can be used in sorter expressions. The technique is use if to produce integers that can be compared based on the filter, like:

(<> (if filter-expr-on-$-entry -1 1)
    (if filter-expr-on-&-entry -1 1))

So if $-entry satisfies the filter, while &-entry doesn't, it's the same as (<> -1 1), which produces -1.

For example, we want to put tags with "file" kind below other tags, then the sorter would look like:

(<> (if (eq? $kind "file") 1 -1)
    (if (eq? &kind "file") 1 -1))

A quick read tells us: If $-entry has "file" kind, and &-entry doesn't, the sorter becomes (<> 1 -1), which produces 1, so the $-entry is put below the &-entry, exactly what we want.

A formatter expression defines how readtags prints tag entries.

A formatter expression may produce a string, a boolean, an integer, or a list. Readtags prints the produced string, and integer as is. Readtags prints nothing for #f, and a newline for #t.

A list could contain any number of strings, booleans, integers, and/or lists. Readtags prints the elements of a list sequentially and recursively.

All the operators for filtering are also available in formatter expressions. In addition to the operators, list is available in formatter expressions. As the name shows, list is for making a list. list makes a list containing arguments passed to the operator. e.g., the following expression makes a list contains 1, #f, and "hello":

(list 1 #f "hello")

NOTE: Unlike real-Lisp, backquote constructs are not available.

To show some examples, the following tags file (output.tags) is assumed as input for readtags:

M    input.c 4;"     macro   file:
N    input.c 3;"     macro   file:
bar  input.c 11;"    f       typeref:typename:void   file:   signature:(char ** argv,int * r)
foo  input.c 6;"     f       typeref:typename:int    file:   signature:(int v)
main input.c 16;"    f       typeref:typename:int    signature:(int argc,char ** argv)

An exapmle for printing only function names:

$ readtags -t output.tags -Q '(eq? $kind "function")' -F '(list $name #t)' -l
bar
foo
main

Doing the same only with a formatter expression:

$ readtags -t output.tags -F '(if (eq? $kind "function") (list $name #t) #f)' -l
bar
foo
main

Generating declarations for the functions:

$ readtags -t output.tags -F \
  '(if (eq? $kind "function")
      (list (if $file "static " #f) $typeref-name " " $name $signature ";" #t)
     #f)' -l
static void bar(char ** argv,int * r);
static int foo(int v);
int main(int argc,char ** argv);

The print operator can be used to print the value of an expression. For example:

$ readtags -Q '(print $name)' -l

prints the name of each tag entry before it. Since the return value of print is not #f, all the tag entries are printed. We could control this using the begin or begin0 operator. begin returns the value of its last argument, and begin0 returns the value of its first argument. For example:

$ readtags -Q '(begin0 #f (print (prefix? "ctags" "ct")))' -l

prints a bunch of "#t" (depending on how many lines are in the tags file), and the actual tag entries are not printed.