Demonstration of regex and string_view problems

regex.cpp

/*
  Demonstration of regex and string_view problems

  $ g++ -std=c++17 -Wall -Wextra regex.cpp -o test

  $ ./test
  wrong()
  - fixed
  - fi
  - fixed
  correct()
  - kitten
  - 42
  - fixed
*/
#include <regex>
#include <string_view>
#include <iostream>
#include <vector>

std::regex re{"name=([a-z]+), number=([0-9]+), fixed=(fixed)"};
std::string_view input{"name=kitten, number=42, fixed=fixed"};

std::vector<std::string_view> wrong()
{
  std::vector<std::string_view> result;
  std::match_results<std::string_view::const_iterator> match;

  if (std::regex_match(input.cbegin(), input.cend(), match, re)) {
    for (size_t i=1; i < match.size(); i++)
      result.push_back(match[i].str());
  }

  return result;
}

std::vector<std::string_view> correct()
{
  std::vector<std::string_view> result;
  std::match_results<std::string_view::const_iterator> match;

  if (std::regex_match(input.cbegin(), input.cend(), match, re)) {
    for (size_t i=1; i < match.size(); i++) {
      const char* first = match[i].first;
      const char* last = match[i].second;
      result.push_back({first, static_cast<std::size_t>(last - first)});
    }
  }

  return result;
}

void dump(const std::vector<std::string_view>& v) {
  for (const auto& sv: v)
    std::cout << "- " << sv << '\n';
}

int main() {

  auto r1 = wrong();
  std::cout << "wrong()" << '\n';
  dump(r1);
  auto r2 = correct();
  std::cout << "correct()" << '\n';
  dump(r2);
}

Thanks for this useful example! Note that in C++20 you can also use this simpler correct form:

      result.push_back({first, last});

@mmatrosov thanks for the C++20 pointer, I wasn't aware of it.

Glad you found this snippet helpful, I wasted a lot of time trying to figure out why a simple regex produces crazy results. :)

Thanks for the gist, the definitions of first and last are not necessary, also one can use std::vector::emplace_back:

std::for_each(match.begin()+1, match.end(), [&args](const auto& m) {result.emplace_back(m.first, m.second);});

or if you prefer the discrete loop:

for (size_t i = 1; i < match.size(); ++i)
    result.emplace_back(match[i].first, match[i].second);

@wich Thank you for the comment. Your version uses the C++20 constructor of string_view, which accepts iterators. The snippet was written in the good old days of C++17 :)