Re: [boost] regex problem

From: Mihaly Zachar (zmihaly_at_[hidden])
Date: 2008-04-10 09:10:09

John Maddock wrote:
> Mihaly Zachar wrote:
>   
>> I'm using libboost-regex1.34.1 from gutsy.
>>
>> here is a sample for my problem:
>>
>> ------- CUT -------
>> #include <boost/regex.hpp>
>> #include <string>
>> #include <iostream>
>>
>> using namespace std;
>> using namespace boost;
>>
>> int main()
>> {
>>    string input = "1111";
>>    string a = "(.*)";
>>    string b = "0$1";
>>
>>    regex reg (a, regex::perl);
>>
>>    string out = regex_replace(input, reg, b);
>>
>>
>>    cout << out << endl;
>>
>>    return 0;
>> }
>>
>> ------- CUT -------
>>
>>
>> the expected result would be "01111" but it gives "011110"...
>> why does it work like this ?
>>     
>
> By design: Perl does the same thing, after the first match against "1111" 
> there is a second match against the empty string at the end of the text.
>
>   
>> if I use anchors (^ and $) it works well, but the problem is I will
>> not
>> get anchors from the other applications :(
>>
>> the problem is that I get regexes from different perl and php
>> applications...
>>
>> will the upgrade to 1.35 solve my problem ?
>>     
>
> Nope, and as I say, Perl does the same thing.  You could set the match flag 
> match_not_null as the last argument to regex_replace, but that would disable 
> matches against all zero-length strings which may not be what you want.
>
> Was this a real use case, or just a "getting to know the library" test case?
>   

unfortunately this is a real problem...
I'm interworking with an existing perl system and it provides me rexes 
what I sohuld use.

We should have the same result for the same regexes.

mitya_at_stamford:~$ perl -e '$in = "1111"; $in =~ s/(.*)/0$1/; print $in."\n"'
01111
mitya_at_stamford:~$

to set match_not_null is not good for me :(

for now I think I will put an anchor in a front of it in some cases, but 
that would be nice if there would be a "more official" way to do this...

if you have any other ideas, please let me know.

Thanks,
Misi