> > What I was doing was refactoring some RSS code that didn't have enough > tests, so I wanted to compare pretty much every element of the > resulting RSS. I ended up just eyeballing it in an aggregator, which > seems to have worked out okay but still wasn't ideal. > > I'd want a pretty granular comparison, and at the API level would be > ideal. I don't mind having to do a little work to format the changes > into readable output. Also, maybe having API-level information would > make it easier for me to filter out certain differences. Hm. Sounds like <ins> and <del> equivalents might be perfect, though really, you want exact, full-tree diffs. That's a simpler task, really. Honestly, it sounds like raw Diff::LCS might be the tool you want -- parse both with REXML, and then hit the trees with Diff::LCS -- the gotcha being the way REXML deals with containers. Steal the proxy class from XHTMLDiff and that should be all you need. Ari