From: Robert Klemme <bob.news@...>
Date: 2005-04-19T04:39:35+09:00
Subject: Re: String.sub strangeness


"mark sparshatt" <msparshatt@yahoo.co.uk> schrieb im Newsbeitrag 
news:4263E8A2.7000301@yahoo.co.uk...
> Belorion wrote:
>> Forgive me if my Monday fog is preventing me from seeing the simple
>> answer to this simple problem...
>>
>> irb-enhanced(main):001:0> str = "123&456"
>> => "123&456"
>> irb-enhanced(main):002:0> str.sub( /\&/, "\\&" )
>> => "123&456"
>> irb-enhanced(main):003:0> str.sub( /\&/, "\\!" )
>> => "123\\!456"
>>
>>
> This is because when the string is created \\& is converted into \&. In a 
> substitution string \& refers to the part of the origional string which 
> matches the regexp.
>
>> Why am I not able to escape the & in the string with an a \ ?  If I do 
>> this:
>>
>> irb-enhanced(main):004:0> str.sub( /\&/, "\\\\&" )
>> => "123\\&456"

That's exactly the right conversion!  Apparently you've been fooled by IRB 
which uses #inspect.  Using puts to print the converted string sometimes 
helps avoid confusion:

>> str = "123&456"
=> "123&456"
>> str.gsub( /&/, '\\\\&' )
=> "123\\&456"
>> puts str.gsub( /&/, '\\\\&' )
123\&456
=> nil

Note also:

>> puts '\\\\&'
\\&
=> nil

You need four backslashes because you want two backslashes in the 
substitution string.  The reason for that is that a backslash is a meta 
character in a substitution string (for example if you use '\\2' for the 
second group and '\\&' for the complete match) and thus you have to escape 
it in order to get it literally.

> Another way that works is to use the block form
>
> str.sub!( /\&/) {"\\&"}

I've noticed that people quite often recommend to use the block form of 
sub/gsub to cope with escaping.  Please don't do that: the block form is 
less performant than the direct form.  It's applicable where the 
substitution string (or rather pattern) has to change for each substitution. 
All other cases should use the argument form.

Kind regards

    robert