From: John Woods <jqwoods@...>
Date: 2008-02-23T03:01:08+09:00
Subject: Re: gsub("\\", "\\\\") seems unintuitive

Thanks Rob, that's exactly what I was missing -- the second round of 
escaping is necessary to make escaped references to regex groups work.


-----Original Message-----
From: Rob Biedenharn
Sent: 02/22/2008 09:48 AM
> On Feb 22, 2008, at 12:27 PM, John Woods wrote:
>> The following confusing behavior is noted in the pickaxe book (2nd ed) 
>> on page 75:
>>
>>  # I would expect two backslashes in the result
>>  irb> puts "\\".gsub("\\","\\\\")
>>  \
>>
>>  # I would expect four backslashes in the result
>>  irb> puts "\\".gsub("\\","\\\\\\\\")
>>  \\
>>
>> I can certainly work around it, but it seems unintuitive. Is there a 
>> reason why gsub behaves this way? Just curious...
> 
> 
> Notwithstanding the earlier responses...
> 
> Since the replacement string is evaluated 'twice', once as a ruby string 
> literal and then again by gsub to look for group refrences like '\1', 
> you need to provide two levels of escaping for a backslash.
> 
> \ is "\\"
> so two of them is "\\\\"
> and you want gsub to see that so it need to have them escaped:  "\\\\\\\\"
> 
> Whew!  Yeah, it's unfortunate, but backslash is doing double-duty here: 
> introducing a group reference to the regular expression and escaping 
> characters in a string literal (just like "\n", but also itself).
> 
> -Rob
> 
> Rob Biedenharn        http://agileconsultingllc.com
> Rob@AgileConsultingLLC.com
> 
> 
>