I needed a simple way to test if a list of keywords (tags) had the required words, but one moment later, conditional operators other than AND were required. I didn't find a public module that satisfied my needs. Some of this was discussed in AND and OR on Regular Expressions.

I took my subs and group them in a module, then changed into objects (my first attempt in OOP) and worked nice.

Examples of use:

# Simple query (same as: +xxx -ppp -jjj):
my $query = "xxx-(ppp,jjj)";

use Keywords;
my $kw = Keywords->new(ignorecase => 1);
$kw->prepare($query);

# Simple test:
print "Match!" if $kw->test($list_of_tags);

# @ids has some keys %table:
@ok = grep {$kw->test($table{$_}[$col])} @ids;

# Same as before:
@ok = $kw->grep_keys(map {$_ => $table{$_}[$col]} @ids);

# Hash only has keywords lists:
@ok = $kw->grep_keys(%tags);

# When ids are full keywords:
@ok = $kw->grep(@ids);

[download]

Now, I want to share my efforts and decided to upload this to CPAN, but I need to define or append to a namespace. I thought on Search::Keywords, but Keywords::Match might be a better one.

Hints for a proper namespace?

Hints for extra features? I'm thinking on set/get parms, reparse, sentences (when tags have more than one word), UTF8 (un)support...

Thanks!

Replies are listed 'Best First'.
Re: RFC: A module to evaluate keywords expressions on tag lists.
by ikegami (Patriarch) on Sep 03, 2009 at 18:33 UTC

    Neither is particularly compelling.

    Having a root namespace "Keywords::" means you can do lots of things with keywords. Is there that much you can do to keywords?

    Sounds like it would be better integrated into "Search::" or some other existing top-level namespace.

    I would actually prefer if the query parser provided a standard interface that supported multiple query languages. (The one you provided, Google's, etc)

    • Search::Query (Abstract interface)
    • Search::Query::Google
    • Search::Query::...

    I also have a feeling the matching functionality shouldn't be too heavily integrated.

[reply]
      Is there that much you can do to keywords?

      I think not, but this module provides a query parser, a regexp builder and some tests methods. Not only the query language is relevant. Also it is the way that a list of keywords is being stored if you want to use some specific options, then a method to clean up the list string is provided. So it's not only a Search thing. That's made me doubt.

      I don't think that "Search::Query" is appropiate, for the same reason as above. May be "Regexp::Keywords"?

      BTW, "Search::Query::Google" made me think on a module to retrieve some results from Google (as "Google::Search")... "Search::Keywords::Google" is not much better, but I got the idea and agree on that.

      I also have a feeling the matching functionality shouldn't be too heavily integrated.

      Could you explain this?

[reply]

        Could you explain this?

        If query building is separate from matching, then you'll be able to match against inputs other than a flat string.

        Specifically, only supporting flat strings prevents it from being extended to support "is:unread" or "title:foo"

        I'm not saying you should implement those features. I'm saying it would be great if the designed allowed someone to do so.

[reply]
Re: RFC: A module to evaluate keywords expressions on tag lists.
by vitoco (Hermit) on Sep 30, 2009 at 13:44 UTC
[reply]

Back to Meditations