← All modules

A grammar for shell wildcard syntax, with an actions class that translates a glob into the source text of a Raku regex.

Version

0.0.2 zef:wayland

Depends

nothing outside the core

License

Artistic-2.0

Its own test suite

1 file, green

Checked

2026-09-15 against Raku++ 3.28.0 and Rakudo 2026.08

Where it lives

raku.land · source

Install it #

$ rakupp install Glob::Grammar

zef install Glob::Grammar writes the same store; either installer leaves the module usable by both engines.

What it is for #

Users know *.txt. They do not know / '.txt' $ /, and asking them to learn it for a filter box or a configuration key is the wrong trade. A program that accepts glob patterns has to turn them into something it can match with, and this distribution is the smallest possible answer: a six-token grammar for the syntax, and an actions class that emits the equivalent regex as a string.

It is two files and thirty-four lines, which is worth knowing before reaching for it — the whole thing is small enough to read in a minute, and its limits follow from that.

Glob in, regex source out #

use Glob::Grammar;
use Glob::ToRegexActions;

my $actions = Glob::ToRegexActions.new;
sub to-rx(Str $glob) {
    my $m = Glob::Grammar.parse($glob, :$actions);
    $m ?? $m.made !! Nil
}

for '*.txt', 'a?c', 'README', 'src/*.rakumod' -> $g {
    say sprintf('%-14s %s', $g, to-rx($g).raku);
}

my $rx = to-rx('*.txt');
say $rx.^name;
for <notes.txt notes.md a.txt txt> -> $name {
    say sprintf('  %-10s %s', $name, so $name ~~ /<$rx>/);
}

*.txt          "^.*\\.txt\$"
a?c            "^a.c\$"
README         "^README\$"
src/*.rakumod  "^src\\/.*\\.rakumod\$"
Str
  notes.txt  True
  notes.md   False
  a.txt      True
  txt        False

* becomes .*, ? becomes ., and every other character is escaped so it matches itself. The result is a Str, not a Regex — you interpolate it with /<$rx>/, and because it is anchored at both ends it always does whole-string matching.

The one thing to know #

A bracket group is not translated as a character class in the shell sense. The members are stringified with spaces between them, so a range is not a range and a negation is not a negation:

use Glob::Grammar;
use Glob::ToRegexActions;

my $actions = Glob::ToRegexActions.new;
sub to-rx(Str $g) {
    my $m = Glob::Grammar.parse($g, :$actions);
    $m ?? $m.made !! Nil
}

say to-rx('[abc]x').raku;
say to-rx('[a-c]x').raku;
say to-rx('[!a]bc').raku;
say to-rx('{one,two}').raku;
say to-rx('a\*b').raku;

"^<[a b c]>x\$"
"^<[a - c]>x\$"
"^<[! a]>bc\$"
"^\\\{one\\,two\\}\$"
"^a\\\\.*b\$"

A plain list of characters works, which covers most real patterns. The second line is a range in every shell and is the literal three-member set a, -, c here — so bx does not match and -x does. The third is shell negation and matches abc while rejecting xbc, exactly inverted. Brace alternation is not glob syntax to this grammar at all, and a backslash does not escape anything, so a literal * cannot be matched.

Within a plain wildcard pattern the translation is sound. Anything with brackets or braces in it needs checking against what you meant.

Where the two engines differ #

The generated string is Raku source, and the two engines do not accept quite the same Raku. <[a - c]> — what a range turns into — is a character class with a literal hyphen under Raku++ and a compile error under Rakudo, which refuses a bare - inside a class. A glob containing a space becomes an escaped space, which Raku++ accepts and Rakudo rejects as an unspace. And \w differs on non-ASCII letters, so an accented character is escaped by one engine and not the other.

The failures surface at match time rather than at translation time, because the illegal construct is inside an interpolated string. If the patterns come from users and the code has to run on both engines, restrict the accepted syntax to *, ? and plain bracket lists, and reject the rest before translating.