Ruby

=begin
This request is in other words, enable capture history.

A-5. Disabled functions by default syntax
+ capture history
(?@...) and (?@...)
ex. /(?@a)*/.match("aaa") ==> [<0-1>, <1-2>, <2-3>]
see sample/listcap.c file.
http://www.geocities.jp/kosako3/oniguruma/doc/RE.txt
=end

=begin

This request is in other words, enable capture history.

I'm not sure I understand. The example shows that the history is already captured. Both the #inspect output and #to_a shows that the data is already stored inside the MatchData object; it's just not accessible with the #[Symbol] accessor.

IOW:

'abc'.match(/(?a)(?b)(?c)/)
=> #<MatchData "abc" a:"a" a:"b" a:"c">

Group 'a' is shown as having the values 'a', 'b', and 'c', but #[:a] only returns 'c':

'abc'.match(/(?a)(?b)(?c)/)[:a]
=> "c"
'abc'.match(/(?a)(?b)(?c)/).captures
=> ["a", "b", "c"]

This doesn't seem like a disabled function to me; just an assumption that each group name will only appear once.
=end

=begin
Yes, of cource, your desiable API needs more implementation after the option enabled.

I said by the comment before, Ruby 1.9's regexp is based on Oniguruma. So if a function is already implemented in Oniguruma, the function is more easy to realize than functions which are not implemented in Oniguruma.
=end

=begin
Yui,

Thank you for your help. I didn't realise it would be difficult. I'd assumed that as:

/(?a)(?b)(?c)/.named_captures
=> {"a"=>[1, 2, 3]}

And:

'abc'.match(/(?a)(?b)(?c)/)[1..3]
=> ["a", "b", "c"]

It would be simply be a matter of mapping one to the other.

How about then if we set this ticket's priority to 'Low', then add a note to the documentation of MatchData#[] that explains this quirk? :-)

(As an aside that Oniguruma document would make a great start for the RDoc of Regexp. Currently the actual syntax of the patterns doesn't seem to appear anywhere in ri.)
=end

=begin
Oh sorry, i misunderstood. I think you wanted to access "b" with /(?\w)+/.match("abc").

Your proposal may be able to implement because those data is near by us. (you taught it by inspect data, sorry!)

Anyway however, changing return value from alwasy String to String or Array is difficult because of compatibility.
If you want access another matched string, suggest to add a new API to access them.
So what is the desirable API is the problem.
=end

=begin

Oh sorry, i misunderstood. I think you wanted to access "b" with /(?\w)+/.match("abc").

That's quite alright. :-)

Anyway however, changing return value from alwasy String to String or Array is difficult because of
compatibility.

Hmmm... That's unfortunate MatchData#[] is a lovely API. I guess one approach is to return an Array when there are multiple matches; a String otherwise. Anybody who currently relies on only the last match being returned is both in the minority and taking advantage of a bug. But I confess not to being overly fond of this solution. :-/ I'm not sure.
=end

=begin

one approach is to return an Array when there are multiple matches; a String otherwise

There are few APIs which returns different types except true/false or obj/nil.
This is because such API disturbs duck typing.

So new API which returns always Array seems the way.
=end

=begin
I think adding another method of this form to MatchData will be confusing. How about overloading MatchData#values_at ? It currently takes one or more integer indices and returns an Array of corresponding values. It could be modified to take a list of Symbols (and Strings, if it must) and return an Array of the matches. This is backward compatible, requires no new methods, and uses the same principle as MatchData#[], which previously only accepted Integer arguments, and now accepts Symbols/Strings, too.

>> 'haystack'.match(/(?<h>ay).+(?<h>ack)/).values_at(:h)
['ay','ack']
>> 'haystack'.match(/(?<h>ay).+(?<h>a(?<seek>ck))/).values_at(:seek)
['ck']
>> 'haystack'.match(/(?<h>ay).+(?<h>a(?<seek>ck))/).values_at(:seek, 'h')
['ck','ay','ack']

=end

=begin
Thanks, Erik. I tried the patch out and it works well. :-)
=end

=begin
Hi,

How about overloading MatchData#values_at ?

Why do you attempt to reuse existing method? :-/
I think it is better to introduce new method like MatchData#all_values.

'haystack'.match(/(?ay).+(?a(?ck))/).values_at(:seek, 'h')
['ck','ay','ack']

I expect values_at returns an array whose length is equal to the number
of arguments.

By the way, I think it is strange (or even a bug) for MatchData#values_at
to reject Symbols.

--
Yusuke ENDOH [email protected]
=end

This is the first time I've seen regular expression groups, so it's interesting.

It occurs to me that with this addition MatchData is both a sort of Array and a sort of Hash. That being so consider md.to_h.