Pretty fast XML parser

• Simple event-based XML parser with minimal copying, and minimal functionality, ~800 LOC header-only library.
• Designed for high-speed parsing of very large XML files (like OSM XML files).
• Should be pretty fast, parsing speed is comparable to GNU grep (see below).

Usage

#include "pfxml.h"

[...]

pfxml::file xml("myfile.xml");

while (xml.next()) {
  const auto& cur = xml.get();
  std::cout << cur.name << std::endl;  // .name is the xml tag name
  std::cout << xml.level() << std::endl;  // xml.level() is the current tags tree level
  std::cout << cur.attrs.size() << std::endl;  // .attrs contains the attributes
  std::cout << cur.attr("id") << std::endl;  // .attr returns the attribute
}

String Handling

All strings contained in the current element returned by xml.get() are only valid until xml.next() is called. If you need the strings afterwards, you have to copy them. Furthermore, all strings are const char* pointers. Keep in mind that something like cur.name == "mytag" will not work. You have to compare strings via strcmp().

Errors

In case the XML was malformed, an exception is thrown.

Speed

No thorough performance evaluation yet. Searching switzerland-latest.osm (5.8 GB) for the ID of the first defined <way> object takes roughly 25 seconds when compiled with -O3 on an Intel(R) Core(TM) i5 with 2 GHz and a SSD. For comparison, finding the first <way> object with GNU grep takes 17 seconds on the same machine (and would fail if the string "<way>" is contained in some previous attribute or text element).

TODOs

Performance evaluation, tests. Contributors welcome.