Repository navigation
Faster again on large documents. Each Schematron rule is now evaluated in one Saxon call for all the elements it applies to, instead of one call per element. Saxon compiles every expression it is given, so the old way compiled a rule's expression once per element; on the smed <app> rule that compiling took 9 of its 9.6 seconds.
The single call returns, for each element, its path (for first-match shadowing) and the result of each test; only violations go back to Saxon to fill in their messages. If the single call fails (a test that raises an error, for instance), the rule is evaluated element by element as before, so shadowing and error reporting are unchanged.
On the smed Latin edition (about 26,000 lines), the check went from 11.5 s to 1.9 s, measured one after the other on the same machine. Results are identical on the 39 documents compared.
The first check after updating is slower once per schema: a new version of the script rebuilds its cached rules.
The VS Code extension 0.5.12 bundles py-xmlvalidate.py 1.16. The Emacs package uses the repository's copy, so straight-pull-package (or package-vc-upgrade) brings it in.
Files:
py-xmlvalidate-v.1.16.zip: the scripts, the Emacs package and the documentation.validate-schematron-0.5.12.vsix: the VS Code extension (install with Extensions: Install from VSIX…).