There are some caveats, but citeproc-rs now passes the CSL test suite.
The caveats are:
This is my fork of the test suite, so it had a few edits made mostly early on making it less prone to failure on a young CSL implementation. E.g. cleaning up JSON, removing variables that don't exist, etc. It also lacks recent changes in the official CSL test repo, but last I checked some of those changes are the same ones I made.
There are 24 test ignored completely, mostly because they rely on citeproc-js-only APIs or test modes. The ignored tests are listed in crates/citeproc/tests/data/ignore.txt, with justifications for each, but these are all absolutely open to question.
Similarly, 15 tests are "snapshotted", meaning citeproc-rs produces different output that is still considered acceptable. Again, the list is in crates/citeproc/tests/data/snapshot.txt with justifications, and again all are open to question. Probably the most iffy ones are position_IbidInText.txt and locator_TrickyEntryForPlurals.txt, but at least neither of them is actually documented in the specification, so there's that at least.
There are also 23 additional tests in crates/citeproc/tests/data/humans/. This one's not really a caveat, but they're all candidates for test suite inclusion, and if you're reading this then chances are that's interesting to you.
A couple of inextricable Rust API changes landed along the way, ie Processor::new(InitOptions { style, ..Default::default() }), but the wasm API is severable, untouched and for a separate PR.
Edit: oh, and another inextricable change, the parsing a style's metadata only. There's a new csl::StyleMeta with a parse method. You can use this to figure out whether a style is dependent, and what its e.g. default-locale is, to apply to the parent style as an override. Or just to get some info from the style, without erroring out if there are parse errors outside the <info> block. (At least I should add a test for that last use case.)
There are some caveats, but citeproc-rs now passes the CSL test suite.
The caveats are:
crates/citeproc/tests/data/ignore.txt, with justifications for each, but these are all absolutely open to question.crates/citeproc/tests/data/snapshot.txtwith justifications, and again all are open to question. Probably the most iffy ones areposition_IbidInText.txtandlocator_TrickyEntryForPlurals.txt, but at least neither of them is actually documented in the specification, so there's that at least.There are also 23 additional tests in
crates/citeproc/tests/data/humans/. This one's not really a caveat, but they're all candidates for test suite inclusion, and if you're reading this then chances are that's interesting to you.A couple of inextricable Rust API changes landed along the way, ie
Processor::new(InitOptions { style, ..Default::default() }), but the wasm API is severable, untouched and for a separate PR.Edit: oh, and another inextricable change, the parsing a style's metadata only. There's a new
csl::StyleMetawith aparsemethod. You can use this to figure out whether a style is dependent, and what its e.g. default-locale is, to apply to the parent style as an override. Or just to get some info from the style, without erroring out if there are parse errors outside the<info>block. (At least I should add a test for that last use case.)