From: nobu.nokada@...
Date: 2002-12-14T16:02:07+09:00
Subject: Re: [RCR] Global Regexp Match Mechanism (//g)

Hi,

At Sat, 14 Dec 2002 14:27:22 +0900,
Austin Ziegler wrote:
> >> Also, if a MatchData is passed as str, it should implicitly
> >> call MatchData#end for pos unless pos is non-nil. IMO, being
> >> explicit about doing a //g-type match is a Good Thing.
> > MatchData as String? Well, in such case I'd expect matching
> > to $&(equals $~.to_s).
> 
> Actually, I was thinking something like:
> 
>   def gmatch(str, pos = nil)
>     if str.class == MatchData
>       if pos.nil?
>         match(str.string, str.end(0))
>       else
>         match(str.string, pos)
>       end
>     else
>       match(str, pos)
>     end
>   end
>       
> I understand that checking the class is considered a bad thing, so
> this should be viewed as a model, only.

I worry if the behavior may surprise/confuse users.

I got new idea, a new method like this.

  class MatchData
    def next_match(re)
      re.match(self.string, self.end(0))
    end
  end

  md = /\A/.match(foostr)
  while md = md.next_match(/\Gfoo/)
    puts md.to_s
  end

But MatchData#sring returns frozen copy, and the name never
look good.  Rather, StringScanner will be better.

-- 
Nobu Nakada