mirror of
https://github.com/abhigyanpatwari/GitNexus.git
synced 2026-10-31 06:52:49 +00:00
feat(type-resolution): consolidate Phases 10–13 into Milestone D — enhance completeness with full integration test coverage across 11 languages
This commit is contained in:
Gergo Magyar
parent
28ec598343
commit
be5ba01dcc
1 changed files with 63 additions and 135 deletions
|
|
@ -51,122 +51,63 @@ The goal is not to build a compiler. The goal is to support high-value static an
|
|||
- Fixpoint iterates until stable (max 10 iterations), enabling chains like `getUser() → .address → .getCity() → city.save()`
|
||||
- Reverse-order copy chains now resolve (`const b = a; const a: User = x` → both resolve)
|
||||
|
||||
### Milestone D — Completeness ✅
|
||||
|
||||
**Shipped in** `feat/type-resolution-milestone-d` (PR #387). Consolidated original Phases 10–13 into 3 balanced phases.
|
||||
|
||||
#### Phase A: Fixpoint Completeness ✅
|
||||
|
||||
- Post-fixpoint for-loop replay (ex-9B): `pendingForLoops` collection + replay after fixpoint resolves iterable types
|
||||
- Object destructuring via `fieldAccess` items (TS/JS `object_pattern`, Rust `struct_pattern`) — no new `destructure` PendingAssignment variant needed
|
||||
- Extracted `resolveFixpointBindings()` helper with exhaustive switch + `classDefCache` memoization
|
||||
|
||||
#### Phase B: Inheritance & Receivers ✅
|
||||
|
||||
- `BuildTypeEnvOptions` interface replacing positional params for `buildTypeEnv`
|
||||
- Heritage pre-pass constructing `parentMap` from tree-sitter query matches (not graph edges — heritage-processor runs in parallel)
|
||||
- MRO-aware `walkParentChain()` (depth 5, cycle-safe BFS) for `resolveFieldType` and `resolveMethodReturnType`
|
||||
- `this`/`self`/`$this`/`Me` receiver substitution via `substituteThisReceiver` hook
|
||||
- Go `inc_statement`/`dec_statement` write-access queries
|
||||
|
||||
#### Phase C: Branch-Sensitive Narrowing ✅
|
||||
|
||||
- Null-check narrowing (`!= null`, `!== undefined`, `is not null`) via position-indexed `patternOverrides`
|
||||
- Supported for TS, Kotlin, C# — renamed `PATTERN_BRANCH_TYPES` → `NARROWING_BRANCH_TYPES`
|
||||
- Bug fix: Kotlin narrowing required 3 fixes in `jvm.ts` (AST node type `equality_expression`, anonymous `null` node, `nullable_type` parameter fallback)
|
||||
|
||||
#### Deferred from Milestone D
|
||||
|
||||
- **Type predicates (13A):** Cross-function analysis for niche TS `x is User` feature — deferred
|
||||
- **Swift parity (11D):** tree-sitter-swift Node 22 issues — all Swift work consolidated to Phase S
|
||||
- **Positional destructuring (12C):** Python/Kotlin/C#/C++ tuple-position-to-field mapping — deferred
|
||||
- **Discriminated union narrowing (13C):** Needs tagged union metadata not in SymbolTable — deferred
|
||||
|
||||
#### Integration Test Coverage
|
||||
|
||||
17 fixture directories, 23 describe blocks, 705 lines of test code covering all 11 languages:
|
||||
- Grandparent MRO (depth-2 C→B→A): TS, JS, Kotlin, C#, C++, Java, PHP, Python, Ruby
|
||||
- Object destructuring: TS, JS
|
||||
- Struct destructuring: Rust
|
||||
- Post-fixpoint for-loop replay: TS, JS
|
||||
- Go inc/dec write access
|
||||
- Null-check narrowing: TS, C#, Kotlin
|
||||
|
||||
---
|
||||
|
||||
## Open Phases
|
||||
|
||||
### Phase 10: Loop-Fixpoint Bridge
|
||||
### Phase S: Swift Parity
|
||||
|
||||
**Supersedes Phase 9B.** For-loop element bindings run during `walk()` before the fixpoint. Variables typed by the fixpoint are invisible to for-loop extraction.
|
||||
**Blocked on** tree-sitter-swift Node 22 compatibility.
|
||||
|
||||
**Problem:**
|
||||
```typescript
|
||||
const users = getUsers(); // fixpoint: users → User[]
|
||||
for (const u of users) { // walk-time: users untyped → u unresolved
|
||||
u.save(); // missed CALLS edge
|
||||
}
|
||||
```
|
||||
|
||||
**Approach:** Post-fixpoint for-loop replay. During `walk()`, store for-loop AST nodes whose iterable is unresolved. After fixpoint completes, replay `extractForLoopBinding` on those nodes using the now-resolved iterable types.
|
||||
|
||||
**Scope:**
|
||||
- Infrastructure: `pendingForLoops` collection + replay (~20 lines in `type-env.ts`)
|
||||
- No extractor changes — reuses existing `extractForLoopBinding`
|
||||
- Swift: add for-loop element binding (currently missing)
|
||||
- Nested for-loops with field-dependent iterables are NOT replayed (handled by call-processor chain resolution)
|
||||
|
||||
**Risks:** AST node lifetime (safe — nodes valid within `buildTypeEnv` lifetime).
|
||||
|
||||
**Impact: High | Effort: Low-Medium**
|
||||
|
||||
---
|
||||
|
||||
### Phase 11: Inheritance & this/self
|
||||
|
||||
Four items sharing infrastructure.
|
||||
|
||||
#### 11A: MRO-aware field and method lookups
|
||||
|
||||
**Problem:** `lookupFieldByOwner` only finds direct fields. Inherited fields (`Admin extends User`, field on `User`) don't resolve.
|
||||
|
||||
**Approach:** Pre-compute `parentMap: Map<nodeId, nodeId[]>` from EXTENDS + IMPLEMENTS edges. Pass to `buildTypeEnv`. Update `resolveFieldType` and `resolveMethodReturnType` to walk parent chain on miss (max depth 5, cycle-safe, first-match-wins for diamond inheritance).
|
||||
|
||||
Includes IMPLEMENTS edges so interface-declared fields/methods resolve (Java, Kotlin, C#).
|
||||
|
||||
#### 11B: this/self in fixpoint
|
||||
|
||||
**Problem:** `this.field` and `this.method()` emit pending items with `receiver: 'this'`, but `scopeEnv.get('this')` returns `undefined`.
|
||||
|
||||
**Approach:** At collection time during `walk()`, resolve the enclosing class name immediately via `findEnclosingClassName()` and substitute it as the receiver. Covers both `fieldAccess` and `methodCallResult`. No fixpoint changes needed. 5-10 lines per extractor.
|
||||
|
||||
#### 11C: Go inc/dec write access
|
||||
|
||||
**Problem:** `obj.field++`/`obj.field--` produce `inc_statement`/`dec_statement` — write-access tracking doesn't see them.
|
||||
|
||||
**Approach:** Add these node types to call-processor write detection (~5 lines).
|
||||
|
||||
#### 11D: Swift assignment chains
|
||||
|
||||
**Problem:** Swift has no `extractPendingAssignment` — copy/callResult/fieldAccess/methodCallResult don't work.
|
||||
|
||||
**Approach:** Implement `extractPendingAssignment` for Swift covering all 4 binding kinds.
|
||||
|
||||
**Risks:** Performance of parent chain walking in fixpoint (bounded by `n_pending × depth × iterations`). Interface method ambiguity (mitigated by checking direct class first, parents on miss).
|
||||
|
||||
**Impact: Medium-High | Effort: Medium**
|
||||
|
||||
---
|
||||
|
||||
### Phase 12: Destructuring
|
||||
|
||||
**Problem:** `const { address, name } = user` produces no bindings — LHS is a pattern node, not an identifier.
|
||||
|
||||
**Approach:** Add `{ kind: 'destructure', source, bindings: [{ varName, fieldName }] }` to `PendingAssignment`. In the fixpoint, when source's type resolves, look up each field via `lookupFieldByOwner` and bind each variable.
|
||||
|
||||
Works without Phase 11A (direct fields only). Phase 11A MRO enhances it to resolve inherited fields too.
|
||||
|
||||
**Phased delivery:**
|
||||
|
||||
| Sub-phase | Scope | Languages |
|
||||
|-----------|-------|-----------|
|
||||
| **12A** | Object destructuring | TS/JS (`object_pattern`) |
|
||||
| **12B** | Struct pattern destructuring | Rust (`struct_pattern`) |
|
||||
| **12C** (deferred) | Positional destructuring | Python, Kotlin, C#, C++ — needs tuple-position-to-field mapping |
|
||||
|
||||
Skip: computed properties, rest elements, nested destructuring (`{ address: { city } }` — deferred).
|
||||
|
||||
**Risks:** Nested destructuring requires recursive resolution (explicitly deferred).
|
||||
- For-loop element binding (from Phase 10)
|
||||
- Assignment chains: copy, callResult, fieldAccess, methodCallResult (from Phase 11D)
|
||||
- `guard let` narrowing (from Phase 13B) — uses scopeEnv path, not `patternOverrides`
|
||||
|
||||
**Impact: Medium | Effort: Medium**
|
||||
|
||||
---
|
||||
|
||||
### Phase 13: Branch-Sensitive Narrowing
|
||||
|
||||
**Design principle:** Targeted narrowing, not general control-flow analysis. Skip anything that requires a control-flow graph.
|
||||
|
||||
**Phased delivery:**
|
||||
|
||||
#### 13A: Type predicate functions (TS only)
|
||||
|
||||
`function isUser(x: unknown): x is User` — detect `type_predicate` return type. When called in an `if` condition, emit pattern binding for the narrowed parameter.
|
||||
|
||||
#### 13B: Nullability narrowing (TS/Kotlin/C#/Swift)
|
||||
|
||||
`if (x != null)` → strip nullable wrapper in truthy branch. Uses existing `patternOverrides` mechanism (position-indexed, scope-aware). Swift `guard let` uses standard scopeEnv (narrowing persists for rest of function). Swift work requires Phase 11D (assignment chains) first.
|
||||
|
||||
#### 13C: Discriminated union narrowing (deferred)
|
||||
|
||||
`if (shape.kind === 'circle')` → needs tagged union metadata not in SymbolTable. Defer.
|
||||
|
||||
**What we skip entirely:** Full control-flow graph, arbitrary conditional narrowing, `typeof` guards, exhaustiveness checking.
|
||||
|
||||
**Risks:** Scope leakage (mitigated by `patternOverrides` position indexing). Swift `guard let` needs scopeEnv path (different from `patternOverrides`).
|
||||
|
||||
**Impact: Medium | Effort: Medium-High**
|
||||
|
||||
---
|
||||
|
||||
### Phase 14: Cross-File Binding Propagation
|
||||
|
||||
**Problem:** `buildTypeEnv` is per-file. Inferred types don't cross file boundaries.
|
||||
|
|
@ -202,19 +143,12 @@ config.validate(); // missed
|
|||
## Dependency Graph
|
||||
|
||||
```
|
||||
Phase 10 (loops) ──────────────────────┐
|
||||
│
|
||||
Phase 11 (MRO + this + Go + Swift) ───┤
|
||||
├──→ Phase 14 (cross-file)
|
||||
Phase 12 (destructuring) ─────────────┤
|
||||
│
|
||||
Phase 13 (branch narrowing) ───────────┘
|
||||
Milestone D (Phases A, B, C) ✅ ──┐
|
||||
├──→ Phase 14 (cross-file)
|
||||
Phase S (Swift parity) ───────────┘
|
||||
|
||||
Phases 10–13 are independent of each other.
|
||||
Exception: Phase 13B Swift (guard let) requires Phase 11D (Swift assignment chains).
|
||||
Exception: Phase 12 benefits from Phase 11A (MRO) but works without it.
|
||||
Phase 14 depends on all of 10–13 being stable.
|
||||
Swift parity threaded through Phases 10–13 incrementally.
|
||||
Phase S is independent of Phase 14.
|
||||
Phase 14 depends on Milestone D being stable.
|
||||
```
|
||||
|
||||
---
|
||||
|
|
@ -222,19 +156,11 @@ Swift parity threaded through Phases 10–13 incrementally.
|
|||
## Language-Specific Gaps (remaining)
|
||||
|
||||
### Swift
|
||||
- For-loop element binding → Phase 10
|
||||
- Assignment chains (copy, callResult, fieldAccess, methodCallResult) → Phase 11D
|
||||
- Pattern binding → Phase 13B (`guard let`)
|
||||
|
||||
### Go
|
||||
- `obj.field++`/`obj.field--` write ACCESSES → Phase 11C
|
||||
|
||||
### Rust
|
||||
- Struct-pattern field destructuring → Phase 12B
|
||||
- For-loop element binding → Phase S
|
||||
- Assignment chains (copy, callResult, fieldAccess, methodCallResult) → Phase S
|
||||
- `guard let` narrowing → Phase S
|
||||
|
||||
### All languages
|
||||
- Inherited field/method resolution → Phase 11A
|
||||
- `this`/`self` in fixpoint → Phase 11B
|
||||
- Cross-file binding propagation → Phase 14
|
||||
|
||||
---
|
||||
|
|
@ -253,14 +179,18 @@ Field/property maps, deep chains, mixed chains, stdlib passthroughs.
|
|||
|
||||
Unified fixpoint loop, call-result binding, field access binding, method-call-result binding, arbitrary-depth chain propagation.
|
||||
|
||||
### Milestone D — Completeness ← **next** (Phases 10–13)
|
||||
### Milestone D — Completeness ✅ (Phases A, B, C)
|
||||
|
||||
Loop-fixpoint bridge, inheritance walking, `this`/`self` resolution, destructuring, branch narrowing, Swift parity.
|
||||
Consolidated Phases 10–13 into 3 balanced phases. Loop-fixpoint bridge, MRO-aware inheritance walking, `this`/`self` resolution, object/struct destructuring, null-check narrowing. Kotlin null-check bug fix. Full 11-language integration test coverage.
|
||||
|
||||
### Milestone E — Cross-Boundary (Phase 14)
|
||||
### Milestone E — Cross-Boundary ← **next** (Phase 14)
|
||||
|
||||
Export-type index, cross-file binding propagation.
|
||||
|
||||
### Milestone S — Swift Parity (Phase S)
|
||||
|
||||
For-loop binding, assignment chains, `guard let` narrowing. Blocked on tree-sitter-swift Node 22.
|
||||
|
||||
---
|
||||
|
||||
## Open Design Questions
|
||||
|
|
@ -294,8 +224,6 @@ That supports: better call graphs, more accurate impact analysis, stronger AI co
|
|||
|
||||
## Summary
|
||||
|
||||
**Complete:** Phases 7, 8, 9, 9C — explicit types, constructor inference, loop inference, field/property resolution, deep chains, mixed chains, stdlib passthroughs, comment-based types, unified fixpoint with 4 binding kinds, arbitrary-depth chain propagation across 11 languages.
|
||||
**Complete:** Phases 7, 8, 9, 9C, Milestone D (A, B, C) — explicit types, constructor inference, loop inference, field/property resolution, deep chains, mixed chains, stdlib passthroughs, comment-based types, unified fixpoint with 4 binding kinds, arbitrary-depth chain propagation, MRO-aware inheritance walking, this/self resolution, object/struct destructuring, null-check narrowing — across 11 languages with full integration test coverage.
|
||||
|
||||
**Next:** Phase 10 (loop-fixpoint bridge) → Phase 11 (MRO + this/self + Go + Swift) → Phase 12 (destructuring) → Phase 13 (branch narrowing) → Phase 14 (cross-file propagation).
|
||||
|
||||
Each phase is independently deliverable (except Phase 14 which depends on 10–13 being stable). Swift parity is threaded incrementally through Phases 10–13.
|
||||
**Next:** Phase 14 (cross-file binding propagation) — the architectural capstone. Phase S (Swift parity) is independent and unblocked once tree-sitter-swift Node 22 is resolved.
|
||||
|
|
|
|||
Loading…
Add table
Reference in a new issue