feat(cddl): add .cbor and .cborseq control operators (#89)

[dprevost-LMI]

fixes #89

RFC 8610 §3.8.4: a .cbor control on a byte string asserts the byte string carries a CBOR-encoded data item matching the given type; .cborseq asserts it carries a sequence of CBOR items matching the given (array) type. This is exactly how COSE (RFC 9052) describes a COSE_Sign1's protected header or payload field. Verified the semantics directly against the RFC text before porting, not just the PR description.

Parser (packages/cddl): no new parsing logic needed - the operator's value is a plain type reference, which parseOperator already handles generically for every other operator (bits, and, within, eq, ne, lt, le, gt, ge). Just the two OPERATORS/OPERATORS_EXPECTING_VALUES table entries plus the OperatorType union member. Also updated docs: examples/commons/operators.cddl (the shared "one example per operator" fixture already covered every other operator, so .cbor/.cborseq were a visible gap by omission) and packages/cddl/docs/operators.md, whose hardcoded OperatorType union, numbered operator list, and worked per-operator examples were all stale in the exact same way - checked every doc file that enumerates operators, not just the obvious one.

Ported from Mearman#1 (stacked on #88, already merged upstream, so applies directly to main). One correction made along the way: that PR's .cbor test asserted Occurrence: { n: 0, m: Infinity } for a ?-marked member, but this repo's parser correctly computes { n: 0, m: 1 } for ? (zero-or-one, matching every other optional-member test in this file) - m: Infinity is *, not ?. Fixed the ported test's expectation rather than carry over what looks like a fork-specific parser divergence or copy-paste error. Also added a .cborseq parser test - the ported PR only covered .cbor.

Generators: checked all five, not just cddl. cddl2ts/cddl2py/ cddl2swift/cddl2kotlin already ignore any operator they don't specifically special-case (only "default" is ever checked), so .cbor/ .cborseq fall through generically to the base bstr type - verified via their actual transform() output, not assumed, and added a test to each.

cddl2java was the exception, and needed real code changes: its parseType() throws "Unknown operator" for any operator whose base type isn't bool/range/group - .cbor/.cborseq (attached to plain bstr) hit that path and crashed. Digging into why also surfaced a second, pre-existing bug: cddl2java had no bstr/bytes -> byte[] mapping at all (bare bstr alone resolved to "Unknown"), unlike every other generator. Latent until now since bstr never appears in the real webdriver-bidi spec cddl2java is tested against - but .cbor/.cborseq are RFC-defined to apply specifically to byte strings, so properly supporting this port in cddl2java requires both fixes, not just catching the crash. Fixed both: added the missing byte[] mapping, and made the operator-fallthrough resolve the base type instead of throwing for any operator it doesn't specifically recognize (the same generic behavior the other four generators already have).

Deliberately did not add .cbor/.cborseq to examples/commons/test.cddl, the realistic BiDi-shaped fixture 4 of the 5 generators snapshot-test against - unlike operators.cddl (a catalog, one example per feature), test.cddl reads as an actual mini protocol spec, and a CBOR/COSE property doesn't fit that narrative. Touching it would also mean regenerating snapshots across 4 packages for coverage the explicit per-generator tests above already give more precisely.

Full checks:all passes (464 tests, coverage thresholds met).

[greptile-apps]

Confidence Score: 5/5

The PR appears safe to merge; no new actionable issues remain.

Summary

Adds support for the RFC 8610 .cbor and .cborseq control operators throughout the CDDL parser and documentation.

• Extends the operator AST and parser tables with both operators.
• Adds parser fixtures and tests using valid single-item and sequence controller types.
• Maps Java byte strings to byte[] and lets unrecognized validation operators fall through to their underlying type.
• Confirms all generators preserve byte-string output when these operators are present.
• The previous .cborseq controller issue is fixed, and its thread was resolved.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart LR
    A[CDDL source] --> B[Parser]
    B --> C[AST with cbor or cborseq operator]
    C --> D{Generator}
    D --> E[TypeScript: Uint8Array]
    D --> F[Python: bytes]
    D --> G[Swift: UInt8 array]
    D --> H[Kotlin: ByteArray]
    D --> I[Java: byte array]

Loading

_{Reviews (3) · Last reviewed commit: "test: give .cborseq its own array-sugges..."}

[dprevost-LMI]

@greptileai, can you review again?