Re: Less greedy pattern match

"Ed Morton" <morton@[EMAIL PROTECTED]
> wrote in message 
news:48081FEB.60605@[EMAIL PROTECTED]
>
>
> On 4/17/2008 6:20 PM, Rajan wrote:
>>
>> "Ed Morton" <morton@[EMAIL PROTECTED]
> wrote in message
>> news:4807A618.5020501@[EMAIL PROTECTED]
>>
>>>
>>>On 4/17/2008 1:32 PM, pk wrote:
>>>
>>>>On Wednesday 16 April 2008 05:23, Ed Morton wrote:
>>>>
>>>>
>>>>
>>>>>There's no "less greedy" operator (to be honest, I'm not sure what
that
>>>>>really means or why it'd be more useful than a different RE)
>>>>
>>>>
>>>>Well, perl REs have non-greedy match. It can be useful, for example, 
>>>>when
>>>>parsing html or xml. The classical example is something like this 
>>>>(please
>>>>note that I'm not a perl guru - alas -):
>>>>
>>>>$ cat file.html
>>>><tag><section><t1>foo</t1><t2>blah</t2><t1>bar
>>>></t1></section>
>>>><section><t1>baz
>>>></t1><t2>blah</t2><t1>baz</t1></section></tag>
>>>>
>>>>Suppose you want to remove only what's inside <t1> tags (tags
included)
>>>>and
>>>>keep everything else, without apriori knowledge of how the lines are
>>>>formatted.
>>>>
>>>>The simple perl one-liner:
>>>>
>>>>$ perl -p0e 's%<t1>.+?</t1>%%gs' file.html
>>>><tag><section><t2>blah</t2></section>
>>>><section><t2>blah</t2></section></tag>
>>>>
>>>>does what can't be done with sed. The "+?" is the non-greedy notation
>>>>for "one or more", just like "*?" is the non-greedy operator for "zero

>>>>or
>>>>more", etc.
>>>>
>>>>Note that I'm not saying that that cannot be done using other
>>>>methods...that
>>>>was just an example to demonstrate how non-greedy match can be useful.
>>>>
>>>
>>>Yeah, that is a bit simpler than the awk equivalent:
>>>
>>>$ cat file
>>><tag><section><t1>foo</t1><t2>blah</t2><t1>bar</t1></section>
>>><section><t1>baz</t1><t2>blah</t2><t1>baz</t1></section></tag>
>>>
>>>$ awk
'{o="</t1>";n=SUBSEP;gsub(o,n);gsub("<t1>[^"n"]*"n,"");gsub(n,o)}1'
>>>file
>>><tag><section><t2>blah</t2></section>
>>><section><t2>blah</t2></section></tag>
>>>
>>>or something else to chop up and reassemble the record. It is a pity 
>>>there
>>>isn't
>>>a "not" operator for RE elements other than characters within [...].
>>>
>>>Oh well...
>>>
>>>Ed.
>>>
>>
>>
>> Wouldn't this do?
>> gawk -v RS='</*t1>' -v ORS=""  'RT!="</t1>"{print}'
>>
>
> Yes, that's a nice solution. It's obviously gawk specific which may be 
> fine.
> Just a couple of tweaks: the "*" should be a "?"
Agreed!

 and you don't need the
> "{print}"
I knocked my head after clicking send.

and you don't NEED the "" when assigning a null variable:
>
>  gawk -v RS='</?t1>' -v ORS=  'RT!="</t1>"'
>
Fore!
> Fore!
>
> Ed.
>