Nice, thank you!

I'll do a few corrections and suggestions later, for example there's no to_h in Ruby's MatchData (they chose named_captures, I think it was a good idea). Then some methods seem to allocate a lot of memory, like @regex.name_table.keys.sort.map {|i| @regex.name_table[i]} (3 intermediate arrays.) But more on that later. For now you'll need to run crystal tool format to make sure the code is properly formatted.

@asterite Odd, I'm using the vim-crystal plugin that always runs format prior to buffer save. I will check into why it didn't work and push up a fixed version of my PR.

Regarding the groups_names method, my intent was to ensure the names are in the order they appear in the regex as there is currently no way to grab them in order (short of what that function does) and I'm trying to prevent callers from having to reaching in to the @regex.name_table. I have another less straight-forward method I will replace it with in my next change.

Also, I'm interested in hearing better names for the methods, though I am partial to the consistancy of to_a and to_h. Due to the libpcre that we use, we don't have the same concerns as Ruby because we don't support reusing the same subpattern name:

icr(0.20.1) > r=/(?<letters>.)(?<letters>.)(?<letters>.)(?<letters>.)(?<letters>.)/
invalid regex: two named subpatterns have the same name at 23

I don't like the method names 'group_names' and matched_group_names'; please let me know if you have an alternate recommendation.

Finally, the #to_s method is the same as the #inspect method, though in ruby, #to_s returns the main match (i.e.: self[0]) As this is a breaking change, I didn't want to alter it to the ruby implementation, but was curious why the rational of #to_s being identical to #inspect.

@asterite Please let me know what the next steps are on getting this accepted.

Thanks,
-Dennis

@dennisjbell We basically need to discuss this with the team, so it might take a while before we merge this. Any additions to the language or API take time :-)

@dennisjbell Sorry, I don't know your pull request. I made pull request #3783 which looks like yours.

Closed in favor of #3783