EXPRESS frontend scope beyond structural extraction

[GeneralPawz]

Proposal

Decide the target scope of openbim_step::express. Today it is an explicitly partial, structural extractor, which is enough for IFC but not for the AP2xx family. Candidates, roughly in order of consumer value:

1. Declaration completeness. Capture what is parsed today but thrown away, plus what is skipped:
   • full attribute types instead of only the last token: aggregate kind, bounds, UNIQUE, OPTIONAL element types
   • INVERSE attributes
   • UNIQUE rules
   • SUPERTYPE OF expressions (ONEOF / AND / ANDOR), which are needed to decide which complex instances are legal
   • CONSTANT blocks, which Part 21 constant-instance names reference
2. Multi-schema input. USE FROM / REFERENCE FROM interfaces (short-form schemas). All three AP schemas on hand are long-form (*_mim_lf), so this is lower priority.
3. Executable semantics. Evaluating WHERE / DERIVE / FUNCTION / RULE needs a full EXPRESS expression language and interpreter. That is a different product.

Evidence

Real schemas parsed with SchemaGraph::from_express at dab2ae37610e2dda8a4ac4ca059d996144acfb13:

ap203e2_mim_lf.exp : 6.6 ms | entities 1006 types 240 attrs 1151 derived 101 where 869
AP214E3_2010.exp   : 7.4 ms | entities  915 types 192 attrs 1137 derived 114 where 1196
242_mim_lf.exp     : 21.0 ms| entities 2407 types 528 attrs 2936 derived 554 where 2642
IFC4X3_ADD2.exp    : 5.0 ms | entities  876 types 436 attrs 1644 derived  60 where 752

Function/rule/procedure declaration counts: 154 (AP203e2), 386 (AP214e3), 413 (AP242). None of them are modelled.

Correctness bugs in what is modelled are filed separately: #2 (multiple inheritance), #3 (explicit redeclarations), #5 (complex-instance resolution).

Prior art

OCCT has no runtime EXPRESS parser. Its typed per-AP classes were generated offline by ExpToCas, which OCCT 8.0 moved out of the main repository. Its runtime schema knowledge (StepData_ESDescr/ECDescr) covers descriptions only, not rule evaluation. STEPcode's fedex is a complete EXPRESS front end (C, 3-clause BSD per its COPYING).

Recommendation

Accept (1) as incremental tasks when a consumer needs them. Defer (2). Decline (3) in this crate: an EXPRESS rule evaluator belongs in a separate crate on top of a complete AST, if ever.