Is this C++11 regex error me or the compiler?

Question

OK, this isn't the original program I had this problem in, but I duplicated it in a much smaller one. Very simple problem.

main.cpp:

#include <iostream>
#include <regex>
using namespace std;

int main()
{
    regex r1("S");
    printf("S works.\n");
    regex r2(".");
    printf(". works.\n");
    regex r3(".+");
    printf(".+ works.\n");
    regex r4("[0-9]");
    printf("[0-9] works.\n");
    return 0;
}

Compiled successfully with this command, no error messages:

$ g++ -std=c++0x main.cpp

The last line of g++ -v, by the way, is:

gcc version 4.6.1 (Ubuntu/Linaro 4.6.1-9ubuntu3)

And the result when I try to run it:

$ ./a.out 
S works.
. works.
.+ works.
terminate called after throwing an instance of 'std::regex_error'
  what():  regex_error
Aborted

It happens the same way if I change r4 to \\s, \\w, or [a-z]. Is this a problem with the compiler? I might be able to believe that C++11's regex engine has different ways of saying "whitespace" or "word character," but square brackets not working is a stretch. Is it something that's been fixed in 4.6.2?

EDIT:

Joachim Pileborg has supplied a partial solution, using an extra regex_constants parameter to enable a syntax that supports square brackets, but neither basic, extended, awk, nor ECMAScript seem to support backslash-escaped terms like \\s, \\w, or \\t.

EDIT 2:

Using raw strings (R"(\w)" instead of "\\w") doesn't seem to work either.

Answer 1

Update: <regex> is now implemented and released in GCC 4.9.0

---

^{Old answer:}

ECMAScript syntax accepts [0-9], \s, \w, etc, see ECMA-262 (15.10). Here's an example with boost::regex that also uses the ECMAScript syntax by default:

#include <boost/regex.hpp>

int main(int argc, char* argv[]) {
  using namespace boost;
  regex e("[0-9]");
  return argc > 1 ? !regex_match(argv[1], e) : 2;
}

It works:

$ g++ -std=c++0x *.cc -lboost_regex && ./a.out 1

According to the C++11 standard (28.8.2) basic_regex() uses regex_constants::ECMAScript flag by default so it must understand this syntax.

Is this C++11 regex error me or the compiler?

gcc-4.6.1 doesn't support c++11 regular expressions (28.13).

Answer 2

The error is because creating a regex by default uses ECMAScript syntax for the expression, which doesn't support brackets. You should declare the expression with the basic or extended flag:

std::regex r4("[0-9]", std::regex_constants::basic);

Edit Seems like libstdc++ (part of GCC, and the library that handles all C++ stuff) doesn't fully implement regular expressions yet. In their status document they say that Modified ECMAScript regular expression grammar is not implemented yet.

Answer 3

Regex support improved between gcc 4.8.2 and 4.9.2. For example, the regex =[A-Z]{3} was failing for me with:

Regex error

After upgrading to gcc 4.9.2, it works as expected.