From: Michael Fellinger <m.fellinger@...>
Date: 2008-02-26T13:00:51+09:00
Subject: Re: [Q] difference between StringScanner#scan and Regexp#match

On Tue, Feb 26, 2008 at 7:12 AM, Caleb Clausen <vikkous@gmail.com> wrote:
>
> On 2/23/08, makoto kuwata <kwa@kuwata-lab.com> wrote:
>  > Hi,
>  >
>  > I'm planning to implement StringScanner in pure Ruby.
>  > But I found that it is hard to implement StringScanner#scan()
>  > in pure Ruby, because of the difference between Regexp#match()
>  > and StringScanner#scan().
>  >
>  > StringScanner#scan() matches only when pattern matches at the
>  > beginning (or at the current position) of input string.
>  >
>  > require 'strscan'
>  > input = 'foo 123'
>  > scanner = StringScanner.new(input)
>  > p scanner.scan(/\d+/) #=> nil
>  >
>  > But Regexp#match() matches whenever input string contains pattern.
>  >
>  > input = 'foo 123'
>  > m = /\d+/.match(input)
>  > p m[0] if m #=> "123"
>  >
>  >
>  > Is it possible to restrict Regexp#match() to match only when
>  > pattern starts at the beginning of input string?
>  > My idea is to convert /regexp/ into /\A(?:regexp)/ every time,
>  > but it is a litte ugly.
>  > Is there any good idea to emulate StringScanner#scan in pure Ruby?
>
>  I've done this kind of thing before, and rewriting the regex was the
>  best I could come up with. If Michael's suggestion is too slow for
>  you, then regex rewriting is the only game in town. (Actually, if you
>  can determine that your regexes always require some leading substring,
>  you might be able to optimize Michael's way a bit more...)
>
>  If speed is an issue, why not just use the existing StringScanner?
>  Creating regex's at runtime can cost you quite a bit in performance as
>  well... some caching can help here, if the same regexes are likely to
>  be encountered again.

I, for one, started to work on a StringScanner replacement as well
just for fun. But it could be useful for rubinius to have a pure Ruby
implementation that can be augmented with C in some core areas by a
simple require.

>  You might want to take a look at String#index (and it's 2nd parameter)
>  rather than String#match or Regexp#match, as it allows you to start
>  matching wherever you want in the string, rather than just the
>  beginning. That doesn't help with your immediate question, but maybe
>  it'll give you some ideas of different ways to approach it.
>
>  Finally, a moment of self-promotion. My library 'sequence' implements
>  basically what you want (using regex rewriting, which in the full
>  elaboration gets rather involved). Maybe it could save you some
>  effort...

Thanks for the hint: http://sequence.rubyforge.org for anyone who is
too lazy to type it :)

^ manveru