tests: cross-engine regression coverage for mget m k ?? d

[danieljohnmorris]

Summary

Persona reports going back several sessions (lines 699, 705, 926 of ilo_assessment_feedback.md) claim mget m k??0 parses as mget(m, k??0), forcing the two-step r=mget m k;v=r??0 tax on every map lookup with a default. On current main the bug is gone: the arity-aware call parser landed in 06477c5 registers mget with fixed arity 2, so ?? falls out as infix nil-coalesce on the whole (mget m k) result, and PR #169's ?? Optional-unwrap symmetry closes the loop on the value side.

No fix needed; what was missing was permanent coverage. This PR adds it.

Repro before / after

The supposedly-broken patterns all work today:

$ ilo 'f>n;m=mset mmap "k" 5;k="k";mget m k ?? 0' --run-tree f
5
$ ilo 'f>n;m=mmap;k="x";mget m k ?? 7' --run-cranelift f
7

AST confirms the grouping: NilCoalesce { value: Call(mget, [m, k]), default: ... }.

What's in the diff

Two commits, no src/ changes.

1. examples: mget m k ?? d one-line lookup with default — examples/mget-default.ilo with 7 -- run: / -- out: assertions covering literal key, missing key, variable key, missing variable key, text-typed value, text default, and path-access key. Picked up automatically by tests/examples_engines.rs so every engine runs it on every CI build.

2. tests: cross-engine regression for mget m k ?? d — tests/regression_mget_default.rs with 9 cases per engine (tree, VM, Cranelift): literal hit/miss, variable hit/miss (the exact shape from persona line 699), text hit/miss, path-access key, call-result key, parenthesised key.

Adjacent finding (logged separately, not fixed here)

While writing this, I confirmed chained mget m "a" ?? mget m "b" ?? 99 does NOT parse — the right-hand side of ?? re-enters call-arg mode and slurps the next ?? as a third arg to the second mget. Single-mget defaults work; only chains break. Logged as a fresh entry in ilo_assessment_feedback.md; a dedicated PR can tackle it once we agree on the parser strategy.

Test plan

• cargo test --release --features cranelift --test regression_mget_default — 3 passed (tree + VM + Cranelift)
• cargo test --release --features cranelift --test examples_engines — all examples green
• cargo test --release --features cranelift — full suite passes, no regressions
• cargo fmt --all -- --check clean
• cargo clippy --release --features cranelift --tests -- -D warnings clean

Follow-ups

• Chained mget ?? mget ?? d parse fix (logged in ilo_assessment_feedback.md).
• Once this merges, move the originating persona entries (lines 699, 705, 926 in the assessment doc) into ## ✅ Addressed noting 06477c5 + verifier + interpreter + VM + cranelift: allow ! on Optional #169 as the fix history.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ All tests successful. No failed tests found.

📢 Thoughts on this report? Let us know!