h-dv/code-index

Watch

Fork 0

resolution_by_kind is distorted by construction: receiver heads emitted as type, member_access 100% circular, local rows in denominators #549

New issue

Closed

opened 2026-11-09 08:35:50 +01:00 by buildagent · 2 comments

buildagent commented

2026-11-09 08:35:50 +01:00

Member

Consolidated scope (triage 2026-11-09)

Priority: Medium: the per-kind numbers point at the wrong gaps.

This issue is now the canonical tracker for:

the written verdict on composition vs plugin gap; satisfied by the before/after per-language tables here (from #46, closed and merged here; its full report stays readable there)
a gap miner (shape aggregation and WONTFIX register) to build AFTER this issue, or it ranks the receiver-head artifacts first (from #47, closed and merged here; its full report stays readable there)

Original report

Summary

project_overview.resolution_by_kind and resolution_gaps are what agents and operators use to judge where the index is weak. In several languages those percentages are distorted by construction, so they point at gaps that do not exist and hide the ones that do.

Distortion	Evidence
The head of every receiver chain is emitted as a `type` ref (`ctx` in `ctx.x`, locals, module names)	Python `emit_receiver_type_refs_d` (`python.rs` ~1726), TS `typescript.rs:2027-2045`. In zod, `z` alone accounts for 7,418 `type` rows, about a third of all of them, and dominates `top_unresolved`. flask's unqualified `type` is 4,626 refs at 5.7%. JS `type` is 3,429 at 0%. C# emits namespace segments (`System`, `Data`, about 200) and local receivers as `type`.
`member_access` 100% is circular	Ruby and JS guests never emit that kind; the host relabels a ref as `member_access` only after it resolved (comment at `crates/guest/ruby/src/ruby.rs:2349`). So the row is 100% by construction, and Ruby's `type` % is understated by 760 successes.
Local-scope rows pollute symbols and name fallback	js-express: 1,775 of 1,917 symbols are `static`, and 1,392 are declared local. `top_referenced_symbols` is all `require` aliases. `get_symbol View` lists 5 test-local functions as candidates. `find_references(res)` returns 950 rows of handler parameters, none resolved, even though the walker marks them LOCAL (`parity-deltas.toml` `scope-verdict`).
Std/vendor denominators	django's qualified calls resolve 0.3%, and the top unresolved names are all stdlib. That is a denominator artifact the payload does not separate out.

Proposal

Emit a receiver head as read (or a dedicated receiver role) until it is proven to be a type. C# namespace segments become import-like and leave the denominator.
Report member_access under the kind it was emitted as, with the relabel shown as an outcome, not a kind.
Refs carrying the LOCAL scope verdict are excluded from name fallback and from resolution denominators. visibility: local symbols are hidden by default from file_outline, search_symbols, top_referenced_symbols and ambiguity candidate lists, behind an opt-in flag.
Resolution figures split in-project candidates from external/std names, as internal_resolution_pct already partly does, per kind.

Acceptance

Per-language before/after tables of resolution_by_kind, with a written explanation of every move.
The corpus baseline moves only in the measured denominators, never in binds.
A registry test asserts that no guest emits a receiver head as type without the type-position role.

Related: #46 (honest internal-resolution spread), #522.

Filed from the per-language review on 2026-11-09: 8 agents, one per language plus one for Vue/Svelte/XAML/timeline, run against copies of the pinned corpus repos (tests/corpus/corpus.toml) using installed build 2dcfc615, which differs from master only by XAML/CI commits. The top claim of every issue was spot-checked in the main session. Artifacts live under the review session's scratchpad lang-out/<lang>/ and are available on request.

## Consolidated scope (triage 2026-11-09) **Priority: Medium**: the per-kind numbers point at the wrong gaps. This issue is now the canonical tracker for: - [ ] the written verdict on composition vs plugin gap; satisfied by the before/after per-language tables here (from #46, closed and merged here; its full report stays readable there) - [ ] a gap miner (shape aggregation and WONTFIX register) to build AFTER this issue, or it ranks the receiver-head artifacts first (from #47, closed and merged here; its full report stays readable there) --- ## Original report ## Summary `project_overview.resolution_by_kind` and `resolution_gaps` are what agents and operators use to judge where the index is weak. In several languages those percentages are **distorted by construction**, so they point at gaps that do not exist and hide the ones that do. | Distortion | Evidence | |---|---| | **The head of every receiver chain is emitted as a `type` ref** (`ctx` in `ctx.x`, locals, module names) | Python `emit_receiver_type_refs_d` (`python.rs` ~1726), TS `typescript.rs:2027-2045`. In zod, `z` alone accounts for **7,418** `type` rows, about a third of all of them, and dominates `top_unresolved`. flask's unqualified `type` is 4,626 refs at 5.7%. JS `type` is 3,429 at 0%. C# emits namespace segments (`System`, `Data`, about 200) and local receivers as `type`. | | **`member_access` 100% is circular** | Ruby and JS guests never emit that kind; the host relabels a ref as `member_access` only *after* it resolved (comment at `crates/guest/ruby/src/ruby.rs:2349`). So the row is 100% by construction, and Ruby's `type` % is understated by 760 successes. | | **Local-scope rows pollute symbols and name fallback** | js-express: 1,775 of 1,917 symbols are `static`, and 1,392 are declared local. `top_referenced_symbols` is all `require` aliases. `get_symbol View` lists 5 test-local functions as candidates. `find_references(res)` returns 950 rows of handler parameters, none resolved, even though the walker marks them LOCAL (`parity-deltas.toml` `scope-verdict`). | | **Std/vendor denominators** | django's qualified calls resolve 0.3%, and the top unresolved names are all stdlib. That is a denominator artifact the payload does not separate out. | ## Proposal - Emit a receiver head as `read` (or a dedicated `receiver` role) until it is proven to be a type. C# namespace segments become `import`-like and leave the denominator. - Report `member_access` under the kind it was **emitted** as, with the relabel shown as an outcome, not a kind. - Refs carrying the LOCAL scope verdict are excluded from name fallback and from resolution denominators. `visibility: local` symbols are hidden by default from `file_outline`, `search_symbols`, `top_referenced_symbols` and ambiguity candidate lists, behind an opt-in flag. - Resolution figures split **in-project candidates** from **external/std names**, as `internal_resolution_pct` already partly does, per kind. ## Acceptance - Per-language before/after tables of `resolution_by_kind`, with a written explanation of every move. - The corpus baseline moves only in the measured denominators, never in binds. - A registry test asserts that no guest emits a receiver head as `type` without the type-position role. Related: #46 (honest internal-resolution spread), #522. --- _Filed from the per-language review on 2026-11-09: 8 agents, one per language plus one for Vue/Svelte/XAML/timeline, run against copies of the pinned corpus repos (`tests/corpus/corpus.toml`) using installed build `2dcfc615`, which differs from master only by XAML/CI commits. The top claim of every issue was spot-checked in the main session. Artifacts live under the review session's scratchpad `lang-out/<lang>/` and are available on request._

buildagent added the

labels

2026-11-09 08:35:50 +01:00

buildagent referenced this issue

2026-11-09 08:36:24 +01:00

project_overview is ~95% operator/plugin admin and ~5% orientation — make orientation the default, admin a detail #522

buildagent referenced this issue

2026-11-09 09:21:33 +01:00

investigate: honest internal-resolution spread across builtin and plugin languages #46

buildagent referenced this issue

2026-11-09 09:21:33 +01:00

feat: provenance-aware gap miner for builtin and plugin evidence #47

buildagent referenced this issue

2026-11-09 12:05:31 +01:00

Tool discovery and replies cost extra calls, while review and navigation still have gaps #521

buildagent commented

2026-11-09 12:05:31 +01:00

Author

Member

Dogfood triage 2026-11-09: a function path used as a VALUE is labelled `member_access`

find_references("crates/test-support/src/frozen.rs:123") (fn frozen::dir) returns crates/daemon/tests/precision_gate.rs:297 and crates/test-support/src/shipped.rs:86 as resolved refs of kind member_access. Both are crate::frozen::dir passed as a function value (install(..., crate::frozen::dir)). That is not a member access, and it adds to the "member_access 100% circular" distortion this issue tracks. It should be a value use (read, or a dedicated value kind). It matters for the false "earned zero" family too (#535): a value use is the only evidence that such a fn is live. The find_callers disclosure part is in #521.

## Dogfood triage 2026-11-09: a function path used as a VALUE is labelled `member_access` `find_references("crates/test-support/src/frozen.rs:123")` (fn `frozen::dir`) returns `crates/daemon/tests/precision_gate.rs:297` and `crates/test-support/src/shipped.rs:86` as resolved refs of kind **`member_access`**. Both are `crate::frozen::dir` passed as a function value (`install(..., crate::frozen::dir)`). That is not a member access, and it adds to the "member_access 100% circular" distortion this issue tracks. It should be a value use (`read`, or a dedicated value kind). It matters for the false "earned zero" family too (#535): a value use is the only evidence that such a fn is live. The `find_callers` disclosure part is in #521.

buildagent referenced this issue

2026-11-12 22:22:53 +01:00

Valid references remain unresolved across declaration groups, type relations, scopes and imports #536

buildagent commented

2026-11-12 22:23:08 +01:00

Author

Member

Merged into #536.
No denominator-only migration is verified; Python emit_receiver_type_refs_d still emits receiver heads as type refs.
Remaining: Resolution populations, including the tables, invariant gate and later gap miner.

Merged into #536. No denominator-only migration is verified; Python emit_receiver_type_refs_d still emits receiver heads as type refs. Remaining: Resolution populations, including the tables, invariant gate and later gap miner.

buildagent closed this issue

2026-11-12 22:23:08 +01:00

No milestone

No project

No assignees

1 participant

Notifications

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference

h-dv/code-index#549

No description provided.

Rows
Columns

resolution_by_kind is distorted by construction: receiver heads emitted as type, member_access 100% circular, local rows in denominators #549

Consolidated scope (triage 2026-11-09)

Original report

Summary

Proposal

Acceptance

Dogfood triage 2026-11-09: a function path used as a VALUE is labelled member_access

Dogfood triage 2026-11-09: a function path used as a VALUE is labelled `member_access`