Oops, I've put the benchmark online here: http://users.pandora.be/guyvdb/mediawiki/testDiff3.htm
I'm only comparing against the PHP implementation, not wikidiff2. I have no clue how it compares to that.
2008/8/15 roan.kattouw@home.nl:
---- Guy Van den Broeck guyvdb@gmail.com schrijft:
r39406 contains a brand new implementation of the diff algorithm (in PHP). It's in the same family as the previous one but has better performance and a better heuristic for large diffs.
This one is actually considerably faster than the old diff. In attachment I included
The attachment seems to have been scrubbed by the mailing list.
benchmark output that I use with some of the biggest articles on wikipedia and some medium sized ones. It simulates the actual diff workload with a line based diff first, followed by a word based diff. The comparison is done between a set of sometimes distant versions for both the old algorithm as the new one. The summary in the end concludes that the new implementation is almost 30 times faster.
Faster than what? Then the old PHP DifferenceEngine or then wikidiff2?
Roan Kattouw (Catrope)