ogar-loco: data-first vocabulary tables + statement boundaries (rulings R4/R5)

[AdaWorldAPI]

Implements the lockable subset of the ratified design rulings R1–R9 (now recorded in docs/BLOCK-EDITOR-PLAN.md), grounded per the grain-of-salt directive: no breaking change to the day-old trait — ogar-blockly compiles untouched as the non-breakage proof.

R4 — canonical const data, validated into CheckedVocabulary

• FnSpec (stack_arity / body_refs / min_shape / pushes_result) + VocabularyTable::compose() — the composed 256-slot table, with the shared half stamped from the core, never from the vocabulary: a sibling cannot even express a divergent shared-core opinion in the table consumers read. Strictly stronger than catching drift.
• CheckedVocabulary now stores the composed table and answers every semantic query from it — never by delegation — closing the pathological gap (a vocabulary whose methods answer differently across calls) the delegation form left open. table() exposes the data form for compilers/renderers/fuzzers/oracle schemas.
• New pushes_result column: shared core covered (expressions push; control and BREAK/CONTINUE do not; uncovered stays None); the domain hook defaults to None = undeclared, so nothing guesses. The new column is a new drift channel and is guarded like the other two (can-fire: ADD claiming not to push is caught by name).

R5 — StepMask addresses statements, not calls

statements::statement_bounds — the derived metadata a mask dispatcher consumes (statement ordinal → [first_call, call_count]). A statement closes where stack depth returns to zero after a non-pushing call; a trailing value closes a final expression statement. Refuses uncovered calls, underflow, and dangling operands — masking a raw call out of a stack program desynchronizes every later consumer, so the maskable unit is the whole statement. This dissolves the 64-mask vs 180-call capacity mismatch. Two-sided partial-coverage test: an arity-covered-but-pushes-undeclared domain verb refuses segmentation; declaring the column makes the same body segment.

The ledger

docs/BLOCK-EDITOR-PLAN.md gains: R1–R9 as the arc's canon; the R6 proposed structured-control mints (PARALLEL/JOIN/TRY/CATCH/FINALLY — candidate bytes in the free control range, ids the operator's, none assumed); the grain-of-salt grounding for why R4's trait break was NOT taken (zero real domain authors yet — the const-table authoring surface waits for the first one); the projection-engine reframe with one precision correction (the DTO is 30 × classid(4)+payload(12) lanes carved into ≤180 calls — 256 is codebook cardinality, not array length); and the 144-atom unification staying operator-gated.

Gates

fmt / clippy -D warnings / tests (45 ogar-loco + 7 ogar-blockly) / rustdoc -D warnings — all green, scoped.

🤖 Generated with Claude Code

https://claude.ai/code/session_011DN5odWu4eisdevH9bPWyz

---

Generated by Claude Code

[cursor]

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: serverGenReqId_49c21ab3-0546-4875-887b-dca33bea00f0)