Allow arguments to satisfy RBS interface-typed parameters by apiology · Pull Request #1228 · castwide/solargraph

apiology · 2026-07-30T03:48:02Z

🤖 Filed by Claude, not Vince — acting on his behalf via his GitHub credentials.

Problem

As of RBS 4.1.0, Hash#fetch's single-argument overload takes its key as the ::Hash::_Key duck-type interface instead of the generic K (see ruby/rbs core/hash.rbs). Pin::Parameter#compatible_arg? had no way to confirm an argument satisfies an interface without an explicit include declaration — which core gems don't add for every class that happens to satisfy Hash::_Key's #hash/#eql? contract — so the single-arg overload was rejected as a non-match for a plain Symbol argument.

Call#inferred_pins then fell back to merging the return types of all of Hash#fetch's overloads (including the unresolved generic<X> placeholder from the block-taking overload) instead of resolving the one overload that actually matched the call, producing a spurious typecheck error like:

Declared type Class<Foo> does not match inferred type Class<Foo>, generic<X> for variable clazz

on any Hash{Symbol => Class<X>}#fetch(:key) call under rbs >= 4.1.0.

Fix

Pass :allow_unmatched_interface into the conformance check in Pin::Parameter#compatible_arg?, so an argument is treated as compatible with an interface-typed parameter rather than rejecting the overload outright. This mirrors how TypeChecker already treats unmatched interfaces when comparing declared vs. inferred types (type_checker.rb:80).

Testing

Added a regression spec in spec/type_checker/levels/strong_spec.rb reproducing the exact repro from Hash#fetch on Hash{Symbol => Class<X>} miscomputes generics when rbs >= 4.1.0 #1227; confirmed it fails without the fix and passes with it.
bundle exec rspec: 1618 examples, 0 failures (65 pre-existing pending, unrelated to this change).
bundle exec rubocop on changed files: no offenses.
bundle exec rake typecheck: 548 problems (down from 551 on the same base commit without this change) — the remaining problems are pre-existing RBS 4.1.0 drift unrelated to this fix, out of scope here.

As of RBS 4.1.0, Hash#fetch's single-argument overload takes its key as the ::Hash::_Key duck-type interface instead of the generic K. Pin::Parameter#compatible_arg? had no way to confirm an argument satisfies an interface without an explicit `include` declaration (which core gems don't add for every class that happens to satisfy Hash::_Key's #hash/#eql? contract), so the overload was rejected as a non-match. Call#inferred_pins then fell back to merging the return types of all of Hash#fetch's overloads, including the unresolved `generic<X>` placeholder from the block-taking overload, producing a spurious typecheck error on `Hash{Symbol => Class<X>}#fetch`. Pass :allow_unmatched_interface into the conformance check so an argument is treated as compatible with an interface-typed parameter, consistent with how TypeChecker already treats unmatched interfaces when checking declared vs. inferred types. Fixes castwide#1227

apiology · 2026-07-30T04:10:53Z

🤖 Comment from Claude, not Vince — acting on his behalf via his GitHub credentials.

CI status: 28/29 checks pass, including Solargraph / strong (the typecheck job this issue is about).

The one failure, run_solargraph_rspec_specs, is pre-existing on master and unrelated to this change — it is the downstream solargraph-rspec fork's own Array-generics inference test, already failing on the last 5 pushes to master, including the commit this PR is based on. This PR does not touch that code path.

See https://github.com/castwide/solargraph/actions/runs/30486979645 for that same failure on master.

4.0.0 and 4.0.1 exercise essentially the same behavior in this matrix; 4.1.0 is the version that introduced the Hash::_Key regression fixed in this PR and wasn't pinned anywhere in CI, so a future rbs release could silently drop coverage for it once it's no longer "latest" in the unpinned jobs (typecheck.yml, plugins.yml).

4.1.1 is now the latest stable release (superseding 4.1.0), and 4.1.1.pre.1 is the latest available prerelease. Testing the current latest stable/prerelease plus the oldest supported version (3.10.0) gives better forward coverage than pinning to already-superseded patch releases.

No prerelease newer than 4.1.1 exists yet (4.1.1.dev.1 and 4.1.1.pre.1 both predate the 4.1.1 stable release), so pinning to it added no forward coverage over testing 4.1.1 itself.

apiology · 2026-08-02T17:05:21Z

@castwide This one's ready for review

rbs 4.1.0 changed Hash's key/value generic params to _Key/_Value duck-type interfaces, exposing a Solargraph bug where Hash#fetch's inferred return type leaks a generic<X> instead of the declared value type. Reproduces on vanilla solargraph 0.60.2, unrelated to the apiology/solargraph fork. castwide/solargraph#1228 already tracks and fixes this upstream, so rather than pinning rbs below 4.1.0 to avoid the bug, remove the cap and add targeted @sg-ignore comments citing the PR at each affected call site, consistent with how #1229 is already cited elsewhere in this repo. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01B8YdoRihmgAK4MFHjmMECr

CIs clean-environment Solargraph run caught 6 problems that did not reproduce in local runs (gem doc cache showed 192 vs CIs 200 gems, suggesting an environment divergence not worth chasing further): - attachments.rb: URI::Generic#host nilability now resolves correctly under rbs 4.1.1, making the old ignore unneeded; removed it. - selector_classes/task.rb (3 sites) and subtasks.rb: additional Asana::Resources method/block-param return type inference regressed under rbs >= 4.1.0, same class of issue as castwide/solargraph#1228; added @sg-ignore comments citing that PR. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01B8YdoRihmgAK4MFHjmMECr

rbs 4.1.0 changed Hash's key/value generic params to _Key/_Value duck-type interfaces, exposing a Solargraph bug where Hash#fetch's inferred return type leaks a generic<X> instead of the declared value type. Reproduces on vanilla solargraph 0.60.2, unrelated to the apiology/solargraph fork. castwide/solargraph#1228 already tracks and fixes this upstream, so rather than pinning rbs below 4.1.0 to avoid the bug, remove the cap and add targeted @sg-ignore comments citing the PR at each affected call site, consistent with how #1229 is already cited elsewhere in this repo. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01B8YdoRihmgAK4MFHjmMECr

CIs clean-environment Solargraph run caught 6 problems that did not reproduce in local runs (gem doc cache showed 192 vs CIs 200 gems, suggesting an environment divergence not worth chasing further): - attachments.rb: URI::Generic#host nilability now resolves correctly under rbs 4.1.1, making the old ignore unneeded; removed it. - selector_classes/task.rb (3 sites) and subtasks.rb: additional Asana::Resources method/block-param return type inference regressed under rbs >= 4.1.0, same class of issue as castwide/solargraph#1228; added @sg-ignore comments citing that PR. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01B8YdoRihmgAK4MFHjmMECr

The 26 markers tagged upstream-type-annotation:rbs-4-1-regression (citing castwide/solargraph#1228) are all instances of the same Hash#fetch(...)-on-interface-typed-key pattern. #1228 is a narrow, still-unmerged patch for one specific shape (Hash{Symbol => Class<X>}) and doesn't cover these. The real root cause is a genuine solargraph engine gap, not missing/wrong RBS data: Hash::_Key's signature is correct, solargraph just has no structural conformance check for RBS interface-typed parameters generally (Hash::_Key, _ToAry, _Each, _ToS, etc.), falling back to a blanket allow instead. That's tracked generally at castwide/solargraph#1232, which is a better citation and correctly buckets this as tool-limitation rather than upstream-type-annotation. Pure comment relabel; typecheck problem count unchanged (127). Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Kk5dVYpJr49U3YqNtFHwTL

…arameters castwide#1228 fixes the same underlying bug as the already-merged castwide#1266 (issue castwide#1227: RBS 4.1's Hash#fetch takes its key as the Hash::_Key duck-type interface instead of a generic, causing Solargraph to fall back to the unresolved generic<X> from the block-form overload) but via a different, earlier mechanism: a blanket :allow_unmatched_interface bypass in Pin::Parameter#compatible_arg?, rather than castwide#1266's later structural Conformance check. Verified castwide#1228's own regression test already passes unmodified on this branch without its compatible_arg? change (isolated it into a standalone spec file and ran it against HEAD before resolving the conflict) - castwide#1266's structural interface verification already covers this case, making castwide#1228's code change redundant here. Kept HEAD's compatible_arg? as-is (including literal_arg_matches?, from an earlier-merged PR that castwide#1228's branch, based directly on castwide/master, never saw) and dropped castwide#1228's interface-bypass hunk entirely. Conflict in spec/type_checker/levels/strong_spec.rb: kept castwide#1228's new regression test (issue castwide#1227) as a sibling of HEAD's intersection-type test block (from castwide#1231), which castwide#1228's branch also never saw. .github/workflows/rspec.yml auto-merged cleanly, taking castwide#1228's RBS matrix bump (4.0.0/4.0.1/4.0.2 -> 3.10.0/4.0.3/4.1.1) - core to what this PR is actually testing (RBS 4.1's Hash#fetch signature change). Verified: spec/type_checker/levels/strong_spec.rb, spec/pin/parameter_spec.rb (104 examples, 0 failures, 5 pending), and a broader safety net - spec/type_checker, spec/source, spec/source_map/clip_spec.rb, spec/complex_type, spec/complex_type_spec.rb (807 examples, 0 failures, 35 pending) - all passing locally.

…ric-rbs-4.1

Compress the compatible_arg? comment from 7 lines to 3, and drop the changelog-style multi-paragraph comment on the new Hash#fetch strong-typecheck spec, moving its reasoning into the example name instead.

castwide#1266 fixes the same underlying bug as castwide#1228 (castwide#1228) via real structural interface conformance instead of a blanket allow_unmatched_interface flag, making castwide#1228 redundant. These two pieces of castwide#1228 are not duplicated by castwide#1266 and are carried over before castwide#1228 is closed: - CI matrix bump in rspec.yml (tests rbs 4.0.3/4.1.1 instead of the now-superseded 4.0.0-4.0.2 range) - End-to-end regression spec reproducing castwide#1227's original repro at the typecheck level (structural-rbs-interface-conformance only had unit-level specs against ComplexType::Conformance directly)

apiology · 2026-09-01T13:26:12Z

Claude: Superseded by #1266 — same fix, done via real structural conformance instead of a blanket allow.

apiology mentioned this pull request Jul 30, 2026

TEMP: verify CI reproduces #1227 without fix (will close) #1230

Closed

apiology added 3 commits July 30, 2026 07:44

Drop 4.1.1.pre.1 from rbs matrix

20511d9

No prerelease newer than 4.1.1 exists yet (4.1.1.dev.1 and 4.1.1.pre.1 both predate the 4.1.1 stable release), so pinning to it added no forward coverage over testing 4.1.1 itself.

apiology mentioned this pull request Jul 30, 2026

Teach Solargraph to structurally verify RBS interface-typed parameters instead of blanket-allowing them #1232

Open

apiology force-pushed the fix-hash-fetch-generic-rbs-4.1 branch from 17ebf7a to 20511d9 Compare July 31, 2026 11:49

apiology marked this pull request as draft August 2, 2026 14:42

apiology marked this pull request as ready for review August 2, 2026 17:05

This was referenced Aug 8, 2026

Integration test: 2026-08-04 apiology/solargraph#49

Draft

Enumerable#sum probe result differs between macOS and Linux CI (BigDecimal vs Integer) with identical gem versions #1275

Open

apiology added 2 commits August 26, 2026 09:55

Merge remote-tracking branch 'origin/master' into fix-hash-fetch-gene…

2477472

…ric-rbs-4.1

Merge remote-tracking branch 'origin/master' into fix-hash-fetch-gene…

3afa821

…ric-rbs-4.1

apiology marked this pull request as draft August 31, 2026 20:20

Trim comment style per review feedback

2977836

Compress the compatible_arg? comment from 7 lines to 3, and drop the changelog-style multi-paragraph comment on the new Hash#fetch strong-typecheck spec, moving its reasoning into the example name instead.

apiology closed this Sep 1, 2026

Navigation Menu

Uh oh!

Allow arguments to satisfy RBS interface-typed parameters - #1228

Allow arguments to satisfy RBS interface-typed parameters#1228
apiology wants to merge 7 commits into
castwide:masterfrom
apiology:fix-hash-fetch-generic-rbs-4.1

apiology commented Jul 30, 2026

Uh oh!

apiology commented Jul 30, 2026

Uh oh!

apiology commented Aug 2, 2026

Uh oh!

apiology commented Sep 1, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Uh oh!

Conversation

apiology commented Jul 30, 2026

Problem

Fix

Testing

Uh oh!

apiology commented Jul 30, 2026

Uh oh!

apiology commented Aug 2, 2026

Uh oh!

apiology commented Sep 1, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant