C++ regex bug! Square bracket expression does not work with icase flag

user_2994459 0
Viewed 255
// regex_replace example
#include <iostream>
#include <string>
#include <regex>
#include <iterator>

int main ()
{
  std::string INPUT = "Replace_All_Characters_With_Anything";
  std::string OUTEXP = "0";
  std::regex expression("[A-Za-z]", std::regex_constants::icase);
  std::cout << std::regex_replace(INPUT, expression, OUTEXP);

  return 0;
}

This works here: http://cpp.sh/6gb5a This works here: https://regexr.com/5bt9d

The problem seems to be down to using icase flag or not. A in All, the C in Characters, the W in With, etc. does not get replaced because of the underscore existing. The bug seems to be that using [] to match things only works if said character does not come after a non match.

There does seem to be a quick fix for this, if brackets are followed by a {1}, then it works.

example: [A-Za-z]{1}

Compiler: Microsoft Visual Studio Community 2019 / Version 16.7.3 / c++17

Also tested in c++14, same bad behavior

expected result:
enter image description here

my result:
enter image description here

visual-c++ c++ regex
1 Answers

Not sure if this is an appropriate use of answering. But this is a known bug and it looks like the bug has been known for a few months. No ETA on a fix as far as I can see.

https://github.com/microsoft/STL/issues/993

Looks like RE2 is a recommended alternative regex library.

https://github.com/google/re2/

Instead of using another library, I will create a function that can be used to intercept and change the regex expression string as a temporary fix. Should work whether or not icase flag is used.

test code: https://rextester.com/LSNW3495

// add '{1}' after square bracket ranges unless there already is a quantifier or alternation such as '?' '*' '+' '{}' 
std::string temporaryBugFix(std::string exp)
{
    enum State
    {
        start,
        skipNext,
        lookForEndBracket,
        foundEndBracket,
    };

    State state = start;
    State prevState = start;

    int p = -1;
    std::vector<int> positionsToFix;

    for (auto c : exp)
    {
        ++p;

        switch (state)
        {
        case start:
            if (c == '\\')
            {
                prevState = state;
                state = skipNext;
            }
            else if (c == '[')
                state = lookForEndBracket;

            continue;

        case skipNext:
            state = prevState;
            continue;

        case lookForEndBracket:
            if (c == '\\')
            {
                prevState = state;
                state = skipNext;
            }
            else if (c == ']')
            {
                state = foundEndBracket;
                if (p + 1 == exp.length())
                    positionsToFix.push_back(p + 1);
            }
            continue;

        case foundEndBracket:
            if (c != '+' && c != '*' && c != '?')
                positionsToFix.push_back(p);
            state = start;
            continue;
        }
    }

    // check for valid curly brackets so we don't add an additional one
    std::string s = exp;
    std::smatch m;
    std::regex e("\\{\\d+,?\\d*?\\}");

    int offset = 0;
    vector<int> validCurlyBracketPositions;
    while (regex_search(s, m, e))
    {
        validCurlyBracketPositions.push_back(m.position(0) + offset);
        offset += m.position(0) + m[0].length();
        s = m.suffix();
    }

    // remove valid curly bracket positions from the fix vector
    for (auto p : validCurlyBracketPositions)
        positionsToFix.erase(std::remove(positionsToFix.begin(), positionsToFix.end(), p), positionsToFix.end());

    // insert the fixes
    for (int i = positionsToFix.size(); i--; )
        exp.insert(positionsToFix[i], "{1}");

    return exp;
}
user_29944590
Related