Create --curonly dump from full existing dump?

[eseiver]

The wikifarm I use only offers full gzip xml dumps of all revisions. I do a lot of pywikibot scripting, so I prefer to test offline to avoid constantly slamming the API. Because all the revisions are included in the dump, it takes a while to retrieve the current version of a page. While I was able to use the API with this library to generate a --curonly dump (thanks for making it so easy to set up and use!), I wonder if that procedure could be adapted to make a current versions dump from an existing full local dump.

I started trying to script it myself and it wasn't hard to write a little etree-based function to iterate through a single <page> node and remove all <revision> elements but the latest one, determined by revision_id. However XML is pretty persnickety and I've struggled with incrementally writing to a new file while preserving all other xml data outside of <page> elements so that the pywikibot xmlreader can still interpret it.

I wouldn't mind working on a PR for a new arg like --from-local if folks could point me in the right direction in this code base. Relatedly, an ability to make a test dump with e.g. --max-pages=3 would help me work on my code with a validly constructed compressed xml file.

[robkam]

@elsiehupp