Ticket #12619: boostbug.cpp

// Boost.Regex partial_match implementation is incorrect
//
// Submitted by Dr. Robert van Engelen, engelen@genivia.com
// Related Ticket: #11776
// Affects: RE/flex project https://sourceforge.net/projects/re-flex/
//
// Problem: Using iterators with partial_match to search incremental input
// (such as interactive input) and searching long patterns in files is
// impossible when greedy repetition patterns are used (*, +, etc).
// Repetitions may behave as lazy repetitions (*?, +?, etc).

#include <iostream>
#include <fstream>
#include <sstream>
#include <string>
#include <cctype>
#include <boost/regex.hpp>

void search(std::istream& is, const boost::regex& e, unsigned n)
{
  // buffer we'll be searching in:
  char buf[4096];
  // current position
  const char *cur = buf;
  // end of filled buffer text
  char *pos = buf;
  // flag to indicate whether there is more input to come:
  bool have_more = true;
  // flags
  boost::match_flag_type flg = boost::match_default | boost::regex_constants::match_not_null | boost::match_partial;
  // iterator end
  boost::cregex_iterator b;

  while (have_more)
  {
    // fill more of the buffer by adding n chars
    is.read(pos, n);
    unsigned read = is.gcount();
    // check to see if we've run out of text:
    have_more = read == n;
    // update the end of filled buffer text
    pos += read;
    // and then iterate:
    boost::cregex_iterator a(
        cur,
        pos,
        e,
        flg);
    // while more matches
    while (a != b)
    {
      if ((*a)[0].matched == false)
      {
        // Partial match
        break;
      }
      else if ((*a)[0].second < pos || !have_more) // as per documentation
      {
        // full match:
        std::string full((*a)[0].first, (*a)[0].second - (*a)[0].first);
        std::cout << "matched: " << full << std::endl;
        size_t cap, groups = (*a).size();
        for (cap = 1; cap < groups && !(*a)[cap].matched; ++cap)
          continue;
        if (cap < groups)
          std::cout << "group = " << cap << std::endl;
        // save next current position
        cur = (*a)[0].second;
      }

      // move to next match:
      ++a;
    }
  }
}

int main()
{
  boost::regex e("a.*c");
  std::stringstream s("abc abc");

  // read blocks of 100 chars at a time, OK because match fits in 100 bytes
  std::cout << "\nCorrectly ";
  search(s, e, 100);

  // read by one char at a time (interactive), pattern behaves as a lazy .*?
  std::cout << "\nIncorrectly ";
  s.clear();
  s.seekg(0);
  search(s, e, 1);

  return 0;
}