Skip to content


Switch branches/tags

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time


Docker Image CI

Search Perl 5 sources by given code example.

You can give it file or string as a pattern. Command line interface is compatible with grep, so in common cases you can replace grep with isocode. Binary is build for Linux and completely static, so no need to install any dependencies.

See --help for all supported options.

Rules of matching:

  • All brackets should have pair in pattern. E.g. you can not search for if (. Sorry :(
  • "abc" == 'abc' == q(abc) == qq(abc). Even "123" == 123
  • m/[a-z]/ == /[a-z]/ == "[a-z]" . But if there are any flags - it should be always a regexp.
  • qw( 1 2 3 ) == qw(1 2 3) == (1,2,3) == ("1", 2, "3") ...
  • Trailing separators ignored: (1,2,3,) == (1,2,3) and { a(); b(); } == { a(); b() }
  • Variable names can be differrent! But must be consistent: $a + $a == $b + $b, but $a + $b != $c + $c
  • Whitespace, newlines and comments are ignored.
  • BETA: Wildcard *** matches anything till the end of the block. You can use it like if (! ***) {die *** }. (dont forget to surround it with spaces, so parser dont get confused) . #11


$ ./isocode -nr -e 'my ( $type, %args ) = @_; my $a = {}; bless( $a, $type);' ~/tmp/otrs/
my ( $Type, %Param ) = @_;

    # allocate new hash for object
    my $Self = {};
    bless( $Self, $Type );
my ( $Type, %Param ) = @_;

    # allocate new hash for object
    my $Self = {};
    bless( $Self, $Type );
my ( $Type, %Param ) = @_;

    my $Self = {};
    bless( $Self, $Type );


my ( $Type, %Param ) = @_;

    # Allocate new hash for object.
    my $Self = {};
    bless( $Self, $Type );
Files in directory: 5381
Scanned 2728 files with 0 errors
Total 372 files matches
Total 372 matches


For fun. Maybe it will help someone.

"isocode" means "isomorphic code".


I just gave up setting up ghc no my machine to build static binary, so its built in docker.

# ./
# ./isocode --help
isocode - search source tree by Perl code sample

Usage: isocode (PATTERN | (-e|--regexp PATTERN) | (-f|--file PATTERN_FILE))
               SEARCH_PATHS [-W|--verbose] [-Y|--debug] 
               [-l|--files-with-matches] [-r|--recursive] [-n|--line-number]
  Searches for code similar to PATTERN_FILE or PATTERN in SEARCH_DIR.

Available options:
  PATTERN                  code sample to search for
  -e,--regexp PATTERN      code sample to search for
  -f,--file PATTERN_FILE   path to code sample to search in
  SEARCH_PATHS             files or dirs to search
  -W,--verbose             (non grep compat) 'WHAT?' spew debug information
  -Y,--debug               (non grep compat) 'WHY?' just parse the pattern and
                           show, do not search for it
  -l,--files-with-matches  Show only filenames of matched files
  -r,--recursive           IGNORED, always on
  -n,--line-number         IGNORED, always on
  -h,--help                Show this help text

How it works

'Only Perl can parse Perl', but we only need to understand perl code at very basic level (is it var? is it string literal?).

  1. Parse example code to simple AST
  2. Given that AST, generate parser which search and parses only similar code
  3. Search directory recursively, applying that parser + do variable names consistency check
  4. Profit ???

Search is done by naive algorithm with simple optimization:

  1. get list of first characters that code can start with (e.g. ['u'] for use or ['"', 'q', '\'' ... ] for "string"
  2. Find next occurrence of one of this characters
  3. Run parser from this position
  4. If fail - go to 2

Its faster than parsing whole source, but stil may be slow for common characters.

TLDR: Basically, its three parser generators (one for command line arguments) stiched together.

What (doesn't) work and TODO

See Issues tab on github. Both already implemented and still TBD should be there.