Skip to content
A Raku (née Perl 6) module enabling a fuzzy token in grammars and regexen
Raku
Branch: master
Clone or download

Latest commit

Fetching latest commit…
Cannot retrieve the latest commit at this time.

Files

Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
lib/Regex
resources
t
.gitattributes
.gitignore
LICENSE
META6.json
README.md

README.md

Regex::FuzzyToken for Raku

A Raku module enabling the use of fuzzy tokens in regexen and grammars:

use Regex::FuzzyToken;
my @fruits = <apple banana mango orange kiwi>;
"My favorite fruit is a bnana" ~~ /My favorite fruit is a <fuzzy: @fruits>/;
say $<fuzzy>.fuzz; # bnana
say $<fuzzy>.Str;  # banana

Support for more flexible capturing options is forthcoming. The signature for the fuzzy token are the following:

fuzzy(*@words,
      :$i  = False,
      :$m  = False,
      :$ws = True,
      :$q  = 33,
      :$capture = (@words.tail ~~ Regex
                     ?? @words.tail
                     !! /\w+/ )
)

Basically, you should provide a list of strings to be the “goal” words you want to find, but will accept if they're slightly misspelled. The i, m, and ws options mimic the regex matching behavior, and will make comparisons ignore differences in case (i), marks (m) or white space (ws). The q option is the minimum Q-gram score desired to allow for a match. The default of 33 is fine for most cases, but through testing you may find it necessary to increase or decrease the sensitivity (100 = only match exact, 0 = match everything).

The final option of :$capture allows you to specify the capture regex to use. By default it will only capture a sequence of word characters, but that will cause problems if you need it to match spaces/apostrophes. While you can make things explicit with :capture(/foo/), the signature was designed to allow you to specify the final item as the capture regex, and so the following are equivalent:

<fuzzy: @foo, /bar/>
<fuzzy: @foo, :capture(/bar/)>

To do

It could be interesting to allow for a more complex capture, for example, that matches only as many characters as it needs using .match($capture, :exhaustive) on a substring from the current .pos. That would require some tuning of the Q gram algorithm and in many cases could take exponentially more time, but would be more accurate / usable.

Licenses

All code is and documentation is licensed under the Artistic License 2.0, included with the module. The image used on Github is based on this butterfly which is licensed under CC-0 and modified in accordance with that license and released under CC-BY 4.0

You can’t perform that action at this time.