Re: [Boost-users] libboost-regex-1.33.1 problem with regex_merge

From: John Maddock (john_at_[hidden])
Date: 2007-05-09 11:58:04

Bruno.Voigt_at_[hidden] wrote:
> Hi all,
> I have a problem with an application build with the current
> debian/etch libboost-regex lib.
> I am not sure if the problem results from an intended behaviour
> change of the regex lib
> or if its a bug of either the regex lib or the debian package.
>
> The following test app worked flawlessly with several older boost
> regex versions
> and I would like to understand why it doesn't work anymore using the
> lib from debian/etch.
>
> Do I have to adjust the invocation parameters of boost_merge for the
> current version?
>
> I do not unnecessarily want to force the users of my production apps
> to have to update their match patterns over dozens of cfg files from
> "(.*)" to "^(.*)$".
>
> TIA for any hints,

I'm not sure you're going to like the answer: this is a deliberate change 
and a bug fix to make the library do the same thing as Perl etc.  When 
matching "(.*)" against "abc" there are *two* matches found: one matches 
"abc" then a second match is possible against the zero-length-string at the 
end of the text.  You can prevent matching null-strings altogether by 
passing match_not_null to regex_merge:

boost::regex_merge(my_string, my_regex_comp, my_replace, 
boost::regex_constants::match_not_null);

But that prevents all zero-length matches, which your users might not 
appreciate either :-(

I can't think of an easy way to emulate the previous buggy behaviour I'm 
afraid.

Regards, John.