From: Brian Candler <B.Candler@...>
Date: 2003-05-24T16:51:30+09:00
Subject: Re: why is "does" missing from this sub!-stitution?

On Sat, May 24, 2003 at 07:31:51AM +0900, dblack@superlink.net wrote:
> It's interesting, though, that $2 hasn't been (re)evaluated by the
> time the second argument to #sub is evaluated.  Compare with:
> 
>   irb(main):001:0> "abcdef".sub(s="d",s.upcase)
>   => "abcDef"

I don't see any magic there:

1. The first parameter is evaluated. It evaluates to "d", and has the
   side-effect of setting s.
2. The second parameter is evaluated. It evaluates to "D"
3. The method call is made:
         "abcdef".sub("d","D")

> My first thought by way of explaining the $2 behavior was: there's
> been no assignment to $2, because the regex has been compiled but the
> match hasn't happened yet.

The main reason is expression evaluation sequencing: the arguments to gsub!
are each evaluated individually, and *then* the results of those evaluations
are passed to gsub!

So "<u>#{$1}</u>" is evaluated, which generates a new string object
containing "<u></u>", and a reference to that object is passed in to gsub!
as its second parameter, as part of the method call.

> But then \2 wouldn't be meaningful....

\2 is a magic sequence to gsub. What it does is:
  - match on the regular expression
  - for each match, replace the match with the dest string, but first
    replace \1, \2 etc in the dest string with match[1], match[2] etc.

gsub could have been written to interpret a literal "$1" or "$2" in the same
way, but then if you did actually want to insert a "$" there would be more
escaping rules to be learned.

Regards,

Brian.