Controlling the Tags Reading Behavior

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

Controlling the NAME Action Behavior

Controlling the Output

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.

Filtering, Sorting, and Formatting

These are discussed in the EXPRESSION section.

Examples

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

$ readtags -p - mymethod

$ readtags -i - mymethod

$ readtags -p -ne - myvar

Filtering

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:

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

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

$ 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:

$ 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.

$ 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:

$ 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:

$ 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:

$ 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.

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

$ 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.

Sorting

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.

Formatting

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);

Inspecting the Behavior of Expressions

$ 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.