Conditionally replace regex matches in string

Question

I am trying to replace certain patterns in a string with different replacement patters.

Example:

string test = \"test replacing \\\"these characters\         


        

Answer 1

The c++ (0x, 11, tr1) regular expressions do not really work (stackoverflow) in every case (look up the phrase regex on this page for gcc), so it is better to use boost for a while.

You may try if your compiler supports the regular expressions needed:

#include <string>
#include <iostream>
#include <regex>

using namespace std;

int main(int argc, char * argv[]) {
    string test = "test replacing \"these characters\"";
    regex reg("[^\\w]+");
    test = regex_replace(test, reg, "_");
    cout << test << endl;
}

The above works in Visual Studio 2012Rc.

Edit 1: To replace by two different strings in one pass (depending on the match), I'd think this won't work here. In Perl, this could easily be done within evaluated replacement expressions (/e switch).

Therefore, you'll need two passes, as you already suspected:

 ...
 string test = "test replacing \"these characters\"";
 test = regex_replace(test, regex("\\s+"), "_");
 test = regex_replace(test, regex("\\W+"), "");
 ...

Edit 2:

If it would be possible to use a callback function tr() in regex_replace, then you could modify the substitution there, like:

 string output = regex_replace(test, regex("\\s+|\\W+"), tr);

with tr() doing the replacement work:

 string tr(const smatch &m) { return m[0].str()[0] == ' ' ? "_" : ""; }

the problem would have been solved. Unfortunately, there's no such overload in some C++11 regex implementations, but Boost has one. The following would work with boost and use one pass:

...
#include <boost/regex.hpp>
using namespace boost;
...
string tr(const smatch &m) { return m[0].str()[0] == ' ' ? "_" : ""; }
...

string test = "test replacing \"these characters\"";
test = regex_replace(test, regex("\\s+|\\W+"), tr);   // <= works in Boost
...

Maybe some day this will work with C++11 or whatever number comes next.

Regards

rbo