From: MikkelFJ <mikkelfj-anti-spam@...>
Date: 2003-03-12T05:11:59+09:00
Subject: Re: scanning strings


"MikkelFJ" <mikkelfj-anti-spam@bigfoot.com> wrote in message
news:3e6e3970$0$149$edfadb0f@dtext01.news.tele.dk...


> If not, wouldn't that be a useful feature in Ruby?

I imagined something like

string.scan2(/\s+/) |m| { process_text if m.pre_match; process_space if
m[0] }

Of course, what I would really like to have is:

string.scan([:symbol1, Regexp1], [:symbol2, Regexp2], ...]) do |s, m|
  case s
    when :symbol1
    # ... process m
    when :symbol2
     # ...
    else #  s == :default
     # ...
end

Initially I don't care how inefficient it is, but long term, this would
allow the regex array to be compiled efficiently into a lexer. This would
seriously kick ass.
The match object would also be able to track linenumbers, such that the else
branch could produce a sensible error message.

I noticed Ruby 1.8 had named subexpression in regexp source, perhaps some of
that could be used?

For my original problem I could write.

string.scan([:space, /\s+/]) |s, m|
  case s
    when :space
       process_space m[0]
    else
       process_text m[0]
end

but in this case it would be an overkill.

Mikkel