Fix annotation and namespace issue (continued) #44
Conversation
|
Do you want to take over maintenance? |
I'm willing to but there may be quite a bit with the code base, Continuous Integration, and Deployment to PiPy that I'm not familiar with. That and I started learning Python this year while only visiting relaxng off and on. That being said I'm interested, enjoy learning, and the package seems pretty stable. If your comfortable with that and can be available for questions then I'll do it. |
|
I've invited you as a collaborator on this repository. Let's see how this works out? Since you're indicating that you're pretty new to things I'll withhold PyPI permissions for now -- feel free to ask questions. |
Alright! I have some reading to do. Let me know if this PR works for you (If you think its aligned with what @puetzk had in mind). If this PR checks out I'll probably re-base it after #45 lands. |
…ence (these currently do not pass, due to djc#16)
If the user already has namespace a = 'something else', the code for a:documentation (## comments) must include an xmlns:a='...' to point it at http://relaxng.org/ns/compatibility/annotations/1.0 despite that redefinition It should be an an error for an annotation element to use an undefined namespace prefix (otherwise we produce invalid XML) annotations.rnc should define the a: prefix it's using, trang requires this to parse the file at all, and I can find no justification in the compact syntax spec for pre-defining namespace a = 'http://relaxng.org/ns/compatibility/annotations/1.0'. But leave the default in anyway, since it should be harmless now that it's overridden by any conflicting definition; removing it would break backward-compatibility.
Parsing a parenthesized expression should still allow annotations in front This seems to be a regression in 214fc52, when the parens were lowered from particle it should have been 'primary', not-annotated-primary, so the annotated-primary production could add to it Serializing ref or parentRef may need to write child elements if there are any annotations present
Also fix the strlit versions to emit correct rng serialization * the <data> element does *not* contain the <value> element instead, <value type=>...</value> can carry type info * add missing production for TOKEN strlit This is partially a regression in 2a487b6
According to the grammar at https://www.oasis-open.org/committees/relax-ng/compact-20021121.html#nt-documentationLineContent the content of documentation comments strips off all # and *one* space, then takes the rest of the line This allows indentation in documentation blocks to be preserved
ASSIGN may not be the first child of DEFINE if the annotated-primary production prepended documentation or other annotations. Search through the whole child list to see if there's an ASSIGN specifying a combining mode.
Add additional tests to datatypes.rnc to exercise this, and fix several that were incorrect (the built-in "string" and "token" should be datatypeLibrary="", not substituted for xsd:string and xsd:token the first type-reference seen (if any) is promoted to be the value of <grammar datatypeLibrary=...> and thereafter omitted (it will be copied from the ancestor during simplification)
since xsd is magically predefined: https://relaxng.org/compact-20021121.html In the initial environment used for the start symbol [...] and xsd is bound as a datatype prefix to http://www.w3.org/2001/XMLSchema-datatypes; And cannot even be changed: It is an error if the value of datatypePrefix is xsd and the the value of the literal is not http://www.w3.org/2001/XMLSchema-datatypes. A grammar-level datatypeLibrary="http://www.w3.org/2001/XMLSchema-datatypes" gets printed if xsd:* was ever referenced. otherwise nothing is set for on `<grammar datatypeLibrary="...">` Any other datatypes URI (that is not `http://www.w3.org/2001/XMLSchema-datatypes`) is always written explicitly on `<data>` or `<value>` - thus doesn't care what's in its ancestors. ROOT Assert just gets removed - Included grammars can only disagree about whether it needs to get printed (if either wanted to print it, it does), not about what value it can have (which might not be reconcilable) Add ROOT's DATATYPES to self.typelibs - so we can get ns from datatype_library via custom datatype prefix in type_attrs No longer take first typelib used for global default - `data` and `value` is always written explicitly (if not `xsd:*`) (So no global default is required) Remove self.types since its no longer used (or needed)
|
Sorry, I won't have time to thoroughly review this PR -- happy to answer any specific questions you have, though! |
|
No worries, figured it could be rude if I pushed without asking for a review. I'll review it again before merging. |
|
Not rude, I'm considering you the primary maintainer from here on, though I'll still want to check before we publish to PyPI. |
Continue the work of #19 to fix #16
Additional changes:
grammar level datatypeLibrary solution
datatypeLibrary="http://www.w3.org/2001/XMLSchema-datatypes"gets printed ifxsd:*was ever referenced<data>or<value>Changes in addition to discussed solution