From: Robert Klemme <shortcutter@...>
Date: 2008-08-15T01:57:21+09:00
Subject: Re: Inconsistent regexp behavior

On 14.08.2008 18:40, Glen Holcomb wrote:
> [Note:  parts of this message were removed to make it a legal post.]
> 
> I have a script with two lines:
> 
> puts "toads suck".gsub(/(\s)/, "\\#{$1}")
> 
> puts " a ".gsub(/(\s), "\\#{$1}")
> 
> 
> I get broken output for the first:
> 
> toads\suck
> 
> \ a\
> 
> Can anyone explain why this is happening.  I'm convinced it shouldn't ever
> happen.

I'm sorry, but you're wrong. :-)  When you pass "\\#{$1}" to gsub the 
string is evaluated *once* and even worse, the value of $1 at time of 
call has nothing to do with this gsub operation.  gsub did not even 
start matching at this point in time.

The second replacement apparently works because it gets $1 from the 
first execution...

You also do not need the grouping because the group covers the whole 
match anyway.  You rather want a special replacement expression:

irb(main):001:0> "toads suck".gsub(/\s/, '\\\\\\&')
=> "toads\\ suck"
irb(main):002:0> puts "toads suck".gsub(/\s/, '\\\\\\&')
toads\ suck
=> nil

Note, the high number of backslashes is necessary because there are two 
levels of escaping: first the string, i.e. you need two backslashes to 
have one backslash in the string.  Then you need to escape the backslash 
you want to insert into the replacement because backslash is a 
metacharacter in the substitution string itself.

Kind regards

	robert