XSLT transforms take not-insignificant amounts of RAM

[hexylena]

This isn't a "serious" issue, but more a feature that I'd like to track and eventually PR/patch.

I've recently refactored the way my prebuilt chado schemas are being built into a way that I can more easily publicise, and others can use. Part of that process involved moving the build to a VM.

The VM has only 1 Gb of RAM, and during the XSLT processing step of the Gene Ontology OBO file, XSLT chews through that gig and crashes. This is because XSLT requires loading the entire document into memory, and that's not always feasible for these large XML documents.

Best practices seem to suggest switching to a streaming transform like XST.