refactor(scope-resolution): generic orchestrator + language-agnostic phase

G-Units 6-7 of the emit-pipeline generalization plan, plus the
pipeline-phase generalization (the user's observation that the phase
itself is generic once the orchestrator is).

Changes:

- emit-core/orchestrator.ts — runScopeResolution(input, provider).
  The 180 lines of pipeline glue moved here, parametrized by
  EmitProvider. Provider supplies LanguageProvider, importEdgeReason,
  and the 6 emit-side hooks.
- emit-core/emit-provider.ts — EmitProvider gains languageProvider
  and importEdgeReason fields so the orchestrator needs nothing else.
  resolveImportTarget now takes (targetRaw, fromFile, allFilePaths).
- languages/python/emit/index.ts — pythonEmitProvider + thin
  runPythonScopeResolution wrapper. The first reference impl every
  next-language migration copies.
- emit-providers-registry.ts (NEW) — registry of per-language
  EmitProviders keyed by SupportedLanguages. Adding a language is
  one line here + the provider file.
- pipeline-phases/scope-resolution.ts (NEW) — language-agnostic phase
  iterating EMIT_PROVIDERS ∩ MIGRATED_LANGUAGES. Replaces
  pipeline-phases/python-scope.ts (deleted).
- python-scope-emit.ts deleted.
- pipeline.ts swaps pythonScopePhase → scopeResolutionPhase.

The next language migration is now: implement EmitProvider, register
it, add to MIGRATED_LANGUAGES. No new pipeline phase, no orchestrator
copy-paste. The Python migration's 700+ lines of glue collapse to
~80 lines per future language.

Verification:
- REGISTRY_PRIMARY_PYTHON=0 (legacy): 191/191.
- REGISTRY_PRIMARY_PYTHON=1 (registry): 191/191.
- Default (post MIGRATED_LANGUAGES flip): 191/191.
- tsc --noEmit clean.
This commit is contained in:
Gergo Magyar 2026-04-20 13:29:28 +01:00
parent fb824a6877
commit 8c8a3610d6
10 changed files with 453 additions and 366 deletions

View file

@ -42,6 +42,7 @@ import type {
} from 'gitnexus-shared';
import type { KnowledgeGraph } from '../../graph/types.js';
import type { GraphNodeLookup } from './graph-node-lookup.js';
import { LanguageProvider } from '../language-provider.js';
/** A LinearizeStrategy receives the full ancestor map so C3-style
* algorithms (which need to merge each parent's MRO) can implement
@ -60,6 +61,16 @@ export interface EmitProvider {
/** Identity for telemetry + per-language flag check. */
readonly language: SupportedLanguages;
/** Parsing-side hook bag consumed by `extractParsedFile`. The
* same `LanguageProvider` reference flows through both interfaces
* to keep parsing and emit semantics in sync. */
readonly languageProvider: LanguageProvider;
/** Reason text on emitted IMPORTS edges. Mirrors the legacy DAG's
* per-language convention so consumers asserting on reason keep
* working. */
readonly importEdgeReason: string;
// ─── Pipeline hooks ────────────────────────────────────────────────────────
/**
@ -69,8 +80,17 @@ export interface EmitProvider {
*
* Called once per `ParsedImport` during `finalizeScopeModel`. The
* Python implementation wraps `resolvePythonImportTarget`.
*
* `allFilePaths` is the workspace's file set needed by per-language
* resolvers that must distinguish "this module exists in the repo"
* from "this module is external" (Python's fallback resolver, for
* example).
*/
resolveImportTarget(targetRaw: string, fromFile: string): string | null;
resolveImportTarget(
targetRaw: string,
fromFile: string,
allFilePaths: ReadonlySet<string>,
): string | null;
/**
* Per-scope binding-merge precedence. The shared finalize pass

View file

@ -50,3 +50,8 @@ export { collectNamespaceTargets } from './namespace-targets.js';
export { buildPopulatedMethodDispatch } from './method-dispatch-bridge.js';
export type { ArityVerdict, EmitProvider, LinearizeStrategy } from './emit-provider.js';
export { buildMro, defaultLinearize } from './build-mro.js';
export {
runScopeResolution,
type RunScopeResolutionInput,
type RunScopeResolutionStats,
} from './orchestrator.js';

View file

@ -0,0 +1,161 @@
/**
* `runScopeResolution` generic registry-primary resolution
* orchestrator.
*
* ParsedFile[] (one per file via `extractParsedFile`)
* finalizeScopeModel( + provider hooks adapted to FinalizeHooks)
*
* ScopeResolutionIndexes
* resolveReferenceSites
*
* ReferenceIndex
* emitReceiverBoundCalls (FIRST see Contract Invariant I1)
* emitFreeCallFallback (THEN)
* emitReferencesViaLookup (LAST uses handledSites)
* emitImportEdges
*
* KnowledgeGraph
*
* Per-language entry points (e.g. `runPythonScopeResolution` in
* `languages/python/emit/index.ts`) construct an `EmitProvider` and
* delegate here.
*
* Plan: `docs/plans/2026-04-20-001-refactor-emit-pipeline-generalization-plan.md`.
*/
import type { ParsedFile, RegistryProviders } from 'gitnexus-shared';
import type { KnowledgeGraph } from '../../graph/types.js';
import { extractParsedFile } from '../scope-extractor-bridge.js';
import { finalizeScopeModel } from '../finalize-orchestrator.js';
import { resolveReferenceSites, type ResolveStats } from '../resolve-references.js';
import { buildGraphNodeLookup } from './graph-node-lookup.js';
import { buildPopulatedMethodDispatch } from './method-dispatch-bridge.js';
import { propagateImportedReturnTypes } from './propagate-return-types.js';
import { emitReceiverBoundCalls } from './emit-receiver-bound.js';
import { emitFreeCallFallback } from './emit-free-call.js';
import { emitReferencesViaLookup } from './emit-references.js';
import { emitImportEdges } from './emit-imports.js';
import type { EmitProvider } from './emit-provider.js';
export interface RunScopeResolutionInput {
readonly graph: KnowledgeGraph;
readonly files: readonly { readonly path: string; readonly content: string }[];
readonly onWarn?: (message: string) => void;
}
export interface RunScopeResolutionStats {
readonly filesProcessed: number;
readonly filesSkipped: number;
readonly importsEmitted: number;
readonly resolve: ResolveStats;
readonly referenceEdgesEmitted: number;
readonly referenceSkipped: number;
}
export function runScopeResolution(
input: RunScopeResolutionInput,
provider: EmitProvider,
): RunScopeResolutionStats {
const { graph, files } = input;
const onWarn = input.onWarn ?? (() => {});
// ── Phase 1: extract each file → ParsedFile ────────────────────────────
const parsedFiles: ParsedFile[] = [];
let filesSkipped = 0;
for (const file of files) {
const parsed = extractParsedFile(provider.languageProvider, file.content, file.path, onWarn);
if (parsed === undefined) {
filesSkipped++;
continue;
}
provider.populateOwners(parsed);
parsedFiles.push(parsed);
}
if (parsedFiles.length === 0) {
return {
filesProcessed: 0,
filesSkipped,
importsEmitted: 0,
resolve: { sitesProcessed: 0, referencesEmitted: 0, unresolved: 0 },
referenceEdgesEmitted: 0,
referenceSkipped: 0,
};
}
// ── Phase 2: finalize → ScopeResolutionIndexes ─────────────────────────
const allFilePaths = new Set(parsedFiles.map((f) => f.filePath));
const nodeLookup = buildGraphNodeLookup(graph);
const mroByClassDefId = provider.buildMro(graph, parsedFiles, nodeLookup);
const indexes = finalizeScopeModel(parsedFiles, {
hooks: {
resolveImportTarget: (targetRaw, fromFile) =>
provider.resolveImportTarget(targetRaw, fromFile, allFilePaths),
mergeBindings: (existing, incoming, scopeId) =>
provider.mergeBindings(existing, incoming, scopeId),
},
});
// Stitch the MRO into the finalized indexes (same pattern as before
// generalization — finalizeScopeModel builds an empty
// MethodDispatchIndex by design).
(indexes as { methodDispatch: typeof indexes.methodDispatch }).methodDispatch =
buildPopulatedMethodDispatch(mroByClassDefId);
// Cross-file return-type propagation (Contract Invariant I3 timing:
// after finalize, before resolve).
if (provider.propagatesReturnTypesAcrossImports !== false) {
propagateImportedReturnTypes(parsedFiles, indexes);
}
// ── Phase 3: resolve references via Registry.lookup ────────────────────
const registryProviders: RegistryProviders = {
arityCompatibility: provider.arityCompatibility,
};
const { referenceIndex, stats: resolveStats } = resolveReferenceSites({
scopes: indexes,
providers: registryProviders,
});
// ── Phase 4: emit graph edges (LOAD-BEARING ORDER — see I1) ────────────
const handledSites = new Set<string>();
const receiverExtras = emitReceiverBoundCalls(
graph,
indexes,
parsedFiles,
nodeLookup,
handledSites,
provider,
);
const freeCallExtras = emitFreeCallFallback(
graph,
indexes,
parsedFiles,
nodeLookup,
referenceIndex,
handledSites,
);
const { emitted, skipped } = emitReferencesViaLookup(
graph,
indexes,
referenceIndex,
nodeLookup,
handledSites,
);
const importsEmitted = emitImportEdges(
graph,
indexes.imports,
indexes.scopeTree,
provider.importEdgeReason,
);
return {
filesProcessed: parsedFiles.length,
filesSkipped,
importsEmitted,
resolve: resolveStats,
referenceEdgesEmitted: emitted + receiverExtras + freeCallExtras,
referenceSkipped: skipped,
};
}

View file

@ -0,0 +1,27 @@
/**
* Per-language `EmitProvider` registry the lookup the generic
* `scopeResolutionPhase` uses to pick the right provider for each
* migrated language.
*
* Adding a language is two lines: implement an `EmitProvider` in
* `languages/<lang>/emit/index.ts` and register it here. The phase
* picks it up automatically no workflow changes, no per-language
* pipeline phase file.
*/
import { SupportedLanguages } from 'gitnexus-shared';
import type { EmitProvider } from './emit-core/emit-provider.js';
import { pythonEmitProvider } from './languages/python/emit/index.js';
/** Map of `SupportedLanguages` `EmitProvider`. The phase iterates
* this map intersected with `MIGRATED_LANGUAGES` (the per-language
* flag set) so adding a provider here without flipping the flag is
* safe the provider sits idle until the language is migrated. */
export const EMIT_PROVIDERS: ReadonlyMap<SupportedLanguages, EmitProvider> = new Map<
SupportedLanguages,
EmitProvider
>([[SupportedLanguages.Python, pythonEmitProvider]]);
export function getEmitProvider(lang: SupportedLanguages): EmitProvider | undefined {
return EMIT_PROVIDERS.get(lang);
}

View file

@ -0,0 +1,89 @@
/**
* Python `EmitProvider` and the `runPythonScopeResolution` entry point.
*
* The provider is a thin wiring object Python's specific bits
* (super recognizer, LEGB merge precedence, Python's relative-import
* resolver, the simplified MRO walk) plug into the generic
* `runScopeResolution` orchestrator from `emit-core/`.
*
* Migration reference: when bringing up the next language
* (TypeScript / Java / Kotlin / Ruby), copy this file's structure
* implement the 6 required `EmitProvider` fields, optionally toggle
* the 2 booleans, and call `runScopeResolution(input, provider)`.
*/
import type { ParsedFile, Scope, WorkspaceIndex } from 'gitnexus-shared';
import { SupportedLanguages } from 'gitnexus-shared';
import {
buildMro,
defaultLinearize,
populateClassOwnedMembers,
runScopeResolution,
type EmitProvider,
type RunScopeResolutionInput,
type RunScopeResolutionStats,
} from '../../../emit-core/index.js';
import { pythonProvider } from '../../python.js';
import {
pythonArityCompatibility,
pythonMergeBindings,
resolvePythonImportTarget,
type PythonResolveContext,
} from '../index.js';
const pythonEmitProvider: EmitProvider = {
language: SupportedLanguages.Python,
languageProvider: pythonProvider,
importEdgeReason: 'python-scope: import',
resolveImportTarget: (targetRaw, fromFile, allFilePaths) => {
// PythonResolveContext expects a mutable Set; orchestrator hands us
// a ReadonlySet — safe to widen since the resolver only reads.
const ws: PythonResolveContext = {
fromFile,
allFilePaths: allFilePaths as Set<string>,
};
return resolvePythonImportTarget(
{ kind: 'named', localName: '_', importedName: '_', targetRaw },
ws as unknown as WorkspaceIndex,
);
},
// Python LEGB precedence: local > import/namespace/reexport > wildcard.
mergeBindings: (existing, incoming, scopeId) => {
// pythonMergeBindings(scope, bindings) only consults BindingRef.origin
// for tier ordering, not scope.kind. A shape-stub satisfies the type
// contract without falsifying behavior. Widen the readonly result
// to a mutable BindingRef[] for the orchestrator's hook signature.
const fakeScope = { id: scopeId } as unknown as Scope;
return [...pythonMergeBindings(fakeScope, [...existing, ...incoming])];
},
// Adapter: pythonArityCompatibility predates RegistryProviders and
// uses (def, callsite). Contract is (callsite, def).
arityCompatibility: (callsite, def) => pythonArityCompatibility(def, callsite),
buildMro: (graph, parsedFiles, nodeLookup) =>
buildMro(graph, parsedFiles, nodeLookup, defaultLinearize),
populateOwners: (parsed: ParsedFile) => populateClassOwnedMembers(parsed),
isSuperReceiver: (text) => /^super\s*\(/.test(text),
// Python is dynamically typed — field-fallback heuristic on, return-
// type propagation across imports on. Both default to true; listed
// explicitly here for documentation.
fieldFallbackOnMethodLookup: true,
propagatesReturnTypesAcrossImports: true,
};
export { pythonEmitProvider };
export interface RunPythonScopeResolutionInput extends RunScopeResolutionInput {}
export interface RunPythonScopeResolutionStats extends RunScopeResolutionStats {}
export function runPythonScopeResolution(
input: RunPythonScopeResolutionInput,
): RunPythonScopeResolutionStats {
return runScopeResolution(input, pythonEmitProvider);
}

View file

@ -16,7 +16,7 @@ export { routesPhase, type RoutesOutput, type RouteEntry } from './routes.js';
export { toolsPhase, type ToolsOutput, type ToolDef } from './tools.js';
export { ormPhase, type ORMOutput } from './orm.js';
export { crossFilePhase, type CrossFileOutput } from './cross-file.js';
export { pythonScopePhase, type PythonScopeOutput } from './python-scope.js';
export { scopeResolutionPhase, type ScopeResolutionOutput } from './scope-resolution.js';
export { mroPhase, type MROOutput } from './mro.js';
export { communitiesPhase, type CommunitiesOutput } from './communities.js';
export { processesPhase, type ProcessesOutput } from './processes.js';

View file

@ -1,113 +0,0 @@
/**
* Phase: pythonScope
*
* Registry-primary resolution for Python files (RFC #909 Ring 4).
*
* Gated by `REGISTRY_PRIMARY_PYTHON=1` (via `isRegistryPrimary`). When
* the flag is OFF (default), this phase is a no-op: the legacy paths in
* `import-processor.ts` and `call-processor.ts` continue to handle
* Python imports + calls and this phase contributes nothing to the
* graph.
*
* When the flag is ON, this phase:
* 1. Reads every `.py` file in the workspace.
* 2. Drives the scope-based pipeline end-to-end (extract finalize
* resolve emit) via `runPythonScopeResolution`.
* 3. Emits IMPORTS / CALLS / ACCESSES / INHERITS / USES edges directly.
*
* Pairs with the matching gates in `import-processor.ts` and
* `call-processor.ts` that skip Python files when this phase is active
* so we don't double-emit edges from both code paths.
*
* @deps parse (needs Symbol nodes already in the graph so emit-references
* can attach edges to existing Function/Method/Class nodes)
* @reads scannedFiles
* @writes graph (IMPORTS, CALLS, ACCESSES, INHERITS, USES)
*/
import type { PipelinePhase, PipelineContext, PhaseResult } from './types.js';
import { getPhaseOutput } from './types.js';
import type { StructureOutput } from './structure.js';
import { isRegistryPrimary } from '../registry-primary-flag.js';
import { SupportedLanguages, getLanguageFromFilename } from 'gitnexus-shared';
import { readFileContents } from '../filesystem-walker.js';
import { runPythonScopeResolution } from '../python-scope-emit.js';
import { isDev } from '../utils/env.js';
export interface PythonScopeOutput {
/** True when the flag was on and the phase actually ran. */
readonly ran: boolean;
/** Python files seen by the phase. `0` when `ran === false`. */
readonly filesProcessed: number;
/** IMPORTS edges emitted by this phase. */
readonly importsEmitted: number;
/** Reference (CALLS/ACCESSES/INHERITS/USES) edges emitted. */
readonly referenceEdgesEmitted: number;
}
const NOOP_OUTPUT: PythonScopeOutput = Object.freeze({
ran: false,
filesProcessed: 0,
importsEmitted: 0,
referenceEdgesEmitted: 0,
});
export const pythonScopePhase: PipelinePhase<PythonScopeOutput> = {
name: 'pythonScope',
// Depends on `parse` because emit-references attaches edges to
// already-existing Symbol nodes (Function/Method/Class). The legacy
// `parse` phase still creates those nodes; we only replace the
// import + call resolution layer.
deps: ['parse', 'structure'],
async execute(
ctx: PipelineContext,
deps: ReadonlyMap<string, PhaseResult<unknown>>,
): Promise<PythonScopeOutput> {
if (!isRegistryPrimary(SupportedLanguages.Python)) {
return NOOP_OUTPUT;
}
const { scannedFiles } = getPhaseOutput<StructureOutput>(deps, 'structure');
// Only `.py` files; the per-language flag scopes by extension via
// `getLanguageFromFilename`.
const pythonScanned = scannedFiles.filter(
(f) => getLanguageFromFilename(f.path) === SupportedLanguages.Python,
);
if (pythonScanned.length === 0) return NOOP_OUTPUT;
// Read source for every Python file. `runPythonScopeResolution`
// re-parses each file via `pythonProvider.emitScopeCaptures`; we
// accept the duplicate parse cost in this first cut and can revisit
// by sharing `astCache` across phases if needed.
const filePaths = pythonScanned.map((f) => f.path);
const contents = await readFileContents(ctx.repoPath, filePaths);
const files: { path: string; content: string }[] = [];
for (const fp of filePaths) {
const content = contents.get(fp);
if (content !== undefined) files.push({ path: fp, content });
}
const stats = runPythonScopeResolution({
graph: ctx.graph,
files,
onWarn: (msg) => {
if (isDev) console.warn(`[python-scope] ${msg}`);
},
});
if (isDev) {
console.log(
`🐍 python-scope: ${stats.filesProcessed} files → ${stats.importsEmitted} IMPORTS + ${stats.referenceEdgesEmitted} reference edges (${stats.resolve.unresolved} unresolved sites, ${stats.referenceSkipped} skipped)`,
);
}
return {
ran: true,
filesProcessed: stats.filesProcessed,
importsEmitted: stats.importsEmitted,
referenceEdgesEmitted: stats.referenceEdgesEmitted,
};
},
};

View file

@ -0,0 +1,147 @@
/**
* Phase: scopeResolution
*
* Generic registry-primary resolution phase (RFC #909 Ring 4).
*
* For every language in `MIGRATED_LANGUAGES` (per-language flag set)
* whose provider is registered in `EMIT_PROVIDERS`:
* 1. Filter scanned files by language extension.
* 2. Read file contents.
* 3. Drive the scope-based pipeline end-to-end via the generic
* `runScopeResolution(input, provider)` orchestrator.
* 4. Emit IMPORTS / CALLS / ACCESSES / INHERITS / USES edges.
*
* Pairs with the per-language gates in `import-processor.ts` and
* `call-processor.ts` that skip files when their language is registry-
* primary, so we don't double-emit edges from both code paths.
*
* Adding a language is two changes:
* - Implement `EmitProvider` in `languages/<lang>/emit/index.ts`
* and register it in `emit-providers-registry.ts`.
* - Add the language to `MIGRATED_LANGUAGES` in
* `registry-primary-flag.ts`.
*
* @deps parse (needs Symbol nodes already in the graph so emit-references
* can attach edges to existing Function/Method/Class nodes)
* @reads scannedFiles
* @writes graph (IMPORTS, CALLS, ACCESSES, INHERITS, USES)
*/
import type { PipelinePhase, PipelineContext, PhaseResult } from './types.js';
import { getPhaseOutput } from './types.js';
import type { StructureOutput } from './structure.js';
import { isRegistryPrimary } from '../registry-primary-flag.js';
import { SupportedLanguages, getLanguageFromFilename } from 'gitnexus-shared';
import { readFileContents } from '../filesystem-walker.js';
import { runScopeResolution } from '../emit-core/index.js';
import { EMIT_PROVIDERS } from '../emit-providers-registry.js';
import { isDev } from '../utils/env.js';
export interface ScopeResolutionOutput {
/** True when at least one language ran. */
readonly ran: boolean;
/** Files seen across all languages. `0` when `ran === false`. */
readonly filesProcessed: number;
/** IMPORTS edges emitted across all languages. */
readonly importsEmitted: number;
/** Reference (CALLS / ACCESSES / INHERITS / USES) edges emitted. */
readonly referenceEdgesEmitted: number;
/** Per-language breakdown for telemetry / shadow-parity. */
readonly perLanguage: ReadonlyMap<
SupportedLanguages,
{
readonly filesProcessed: number;
readonly importsEmitted: number;
readonly referenceEdgesEmitted: number;
}
>;
}
const NOOP_OUTPUT: ScopeResolutionOutput = Object.freeze({
ran: false,
filesProcessed: 0,
importsEmitted: 0,
referenceEdgesEmitted: 0,
perLanguage: new Map(),
});
export const scopeResolutionPhase: PipelinePhase<ScopeResolutionOutput> = {
name: 'scopeResolution',
// Depends on `parse` because emit-references attaches edges to
// already-existing Symbol nodes (Function/Method/Class). The legacy
// `parse` phase still creates those nodes; we only replace the
// import + call resolution layer.
deps: ['parse', 'structure'],
async execute(
ctx: PipelineContext,
deps: ReadonlyMap<string, PhaseResult<unknown>>,
): Promise<ScopeResolutionOutput> {
const { scannedFiles } = getPhaseOutput<StructureOutput>(deps, 'structure');
let totalFiles = 0;
let totalImports = 0;
let totalRefs = 0;
let anyRan = false;
const perLanguage = new Map<
SupportedLanguages,
{
readonly filesProcessed: number;
readonly importsEmitted: number;
readonly referenceEdgesEmitted: number;
}
>();
for (const [lang, provider] of EMIT_PROVIDERS) {
if (!isRegistryPrimary(lang)) continue;
const langFiles = scannedFiles.filter((f) => getLanguageFromFilename(f.path) === lang);
if (langFiles.length === 0) continue;
const filePaths = langFiles.map((f) => f.path);
const contents = await readFileContents(ctx.repoPath, filePaths);
const files: { path: string; content: string }[] = [];
for (const fp of filePaths) {
const content = contents.get(fp);
if (content !== undefined) files.push({ path: fp, content });
}
const stats = runScopeResolution(
{
graph: ctx.graph,
files,
onWarn: (msg) => {
if (isDev) console.warn(`[scope-resolution:${lang}] ${msg}`);
},
},
provider,
);
anyRan = true;
totalFiles += stats.filesProcessed;
totalImports += stats.importsEmitted;
totalRefs += stats.referenceEdgesEmitted;
perLanguage.set(lang, {
filesProcessed: stats.filesProcessed,
importsEmitted: stats.importsEmitted,
referenceEdgesEmitted: stats.referenceEdgesEmitted,
});
if (isDev) {
console.log(
`🐍 scope-resolution:${lang}: ${stats.filesProcessed} files → ${stats.importsEmitted} IMPORTS + ${stats.referenceEdgesEmitted} reference edges (${stats.resolve.unresolved} unresolved sites, ${stats.referenceSkipped} skipped)`,
);
}
}
if (!anyRan) return NOOP_OUTPUT;
return {
ran: true,
filesProcessed: totalFiles,
importsEmitted: totalImports,
referenceEdgesEmitted: totalRefs,
perLanguage,
};
},
};

View file

@ -30,7 +30,7 @@ import {
toolsPhase,
ormPhase,
crossFilePhase,
pythonScopePhase,
scopeResolutionPhase,
mroPhase,
communitiesPhase,
processesPhase,
@ -81,7 +81,7 @@ function buildPhaseList(options?: PipelineOptions): PipelinePhase[] {
toolsPhase,
ormPhase,
crossFilePhase,
pythonScopePhase,
scopeResolutionPhase,
];
if (!options?.skipGraphPhases) {

View file

@ -1,249 +0,0 @@
/**
* `runPythonScopeResolution` drive the registry-primary resolution
* pipeline end-to-end for the Python files in a workspace and emit
* graph edges (RFC #909 Ring 3 Python migration).
*
* ParsedFile[] (one per .py via `extractParsedFile`)
* finalizeScopeModel( + Python hooks adapted to FinalizeHooks)
*
* ScopeResolutionIndexes
* resolveReferenceSites
*
* ReferenceIndex
* emitReferencesViaLookup (shared emit-core)
* + emitReceiverBoundCalls (Python-specific; moves to
* languages/python/emit/ in Unit 11)
* + emitImportEdges (shared emit-core)
*
* KnowledgeGraph
*
* The orchestrator is the public seam between the gitnexus pipeline and
* the language-agnostic scope-resolution machinery in `gitnexus-shared`.
* It wires the Python provider's hooks into `FinalizeOrchestratorOptions`
* and threads the workspace index through the import-target resolver.
*
* Gating lives in the pipeline phase (`pipeline-phases/python-scope.ts`),
* not here this function is "what to do" once we've decided to do it.
*/
import type { ParsedFile, RegistryProviders, Scope, WorkspaceIndex } from 'gitnexus-shared';
import type { KnowledgeGraph } from '../graph/types.js';
import { extractParsedFile } from './scope-extractor-bridge.js';
import { finalizeScopeModel } from './finalize-orchestrator.js';
import { resolveReferenceSites, type ResolveStats } from './resolve-references.js';
import { pythonProvider } from './languages/python.js';
import {
pythonArityCompatibility,
pythonMergeBindings,
resolvePythonImportTarget,
type PythonResolveContext,
} from './languages/python/index.js';
import {
buildGraphNodeLookup,
buildMro,
buildPopulatedMethodDispatch,
defaultLinearize,
emitFreeCallFallback,
emitImportEdges,
emitReceiverBoundCalls,
emitReferencesViaLookup,
populateClassOwnedMembers,
propagateImportedReturnTypes,
type EmitProvider,
} from './emit-core/index.js';
import { SupportedLanguages } from 'gitnexus-shared';
// ─── Public API ─────────────────────────────────────────────────────────────
export interface RunPythonScopeResolutionInput {
readonly graph: KnowledgeGraph;
readonly files: readonly { readonly path: string; readonly content: string }[];
/** Optional warning sink (e.g. for telemetry). Failures per-file are non-fatal. */
readonly onWarn?: (message: string) => void;
}
export interface RunPythonScopeResolutionStats {
readonly filesProcessed: number;
readonly filesSkipped: number;
readonly importsEmitted: number;
readonly resolve: ResolveStats;
readonly referenceEdgesEmitted: number;
readonly referenceSkipped: number;
}
/**
* Run the full registry-primary resolution path for `files` and emit the
* resulting CALLS / ACCESSES / INHERITS / USES / IMPORTS edges into
* `graph`. Caller is responsible for ensuring `files` are Python only.
*
* Returns telemetry; never throws on per-file failures (warnings flow
* through `onWarn`).
*/
export function runPythonScopeResolution(
input: RunPythonScopeResolutionInput,
): RunPythonScopeResolutionStats {
const { graph, files } = input;
const onWarn = input.onWarn ?? (() => {});
// ── Phase 1: extract each file → ParsedFile ─────────────────────────────
const parsedFiles: ParsedFile[] = [];
let filesSkipped = 0;
for (const file of files) {
const parsed = extractParsedFile(pythonProvider, file.content, file.path, onWarn);
if (parsed === undefined) {
filesSkipped++;
continue;
}
populateClassOwnedMembers(parsed);
parsedFiles.push(parsed);
}
if (parsedFiles.length === 0) {
return {
filesProcessed: 0,
filesSkipped,
importsEmitted: 0,
resolve: { sitesProcessed: 0, referencesEmitted: 0, unresolved: 0 },
referenceEdgesEmitted: 0,
referenceSkipped: 0,
};
}
// ── Phase 2: finalize → ScopeResolutionIndexes ─────────────────────────
const allFilePaths = new Set(parsedFiles.map((f) => f.filePath));
// Pre-build a graph-node lookup (used both for MRO bridging and for
// edge emission below). EXTENDS edges already in the graph (from the
// legacy heritage processor in `parse`) drive the MRO chain — we
// mirror them into a `MethodDispatchIndex` so receiver-typed
// resolution can walk inherited methods.
const nodeLookup = buildGraphNodeLookup(graph);
const mroByClassDefId = buildMro(graph, parsedFiles, nodeLookup, defaultLinearize);
const indexes = finalizeScopeModel(parsedFiles, {
hooks: {
// Adapter: shared `finalize()` calls `resolveImportTarget(targetRaw,
// fromFile, ws)` with `targetRaw` already extracted; the Python
// provider's signature takes a synthetic `ParsedImport`. Wrap it so
// the hook contract is satisfied without leaking provider internals.
resolveImportTarget: (targetRaw, fromFile) => {
const ws: PythonResolveContext = { fromFile, allFilePaths };
return resolvePythonImportTarget(
{ kind: 'named', localName: '_', importedName: '_', targetRaw },
ws as unknown as WorkspaceIndex,
);
},
// Python LEGB precedence: local > import/namespace/reexport > wildcard.
mergeBindings: (existing, incoming, scopeId) => {
// `pythonMergeBindings(scope, bindings)` only consults
// `BindingRef.origin` for tier ordering, not `scope.kind`. A
// shape-stub satisfies the type contract without falsifying
// behavior.
const fakeScope = { id: scopeId } as unknown as Scope;
return pythonMergeBindings(fakeScope, [...existing, ...incoming]);
},
},
});
// Stitch the MRO into the finalized indexes. `finalizeScopeModel`
// builds an empty MethodDispatchIndex (the comment in
// `finalize-orchestrator.ts:124-129` notes this is a known gap); we
// overwrite with a populated index that wraps the same shape.
(indexes as { methodDispatch: typeof indexes.methodDispatch }).methodDispatch =
buildPopulatedMethodDispatch(mroByClassDefId);
// Propagate return-type typeBindings across imports. The shared
// finalize pass copies callable bindings (`from x import f` puts
// `f` in the importer's bindings), but typeBindings stay file-local.
// Without this step, `u = get_user(); u.save()` works only when
// get_user is in the same file as the call. Done as a post-finalize
// mutation since `Scope.typeBindings` is a plain Map (per
// `draftToScope` line 302).
propagateImportedReturnTypes(parsedFiles, indexes);
// ── Phase 3: resolve references via Registry.lookup ─────────────────────
const providers: RegistryProviders = {
// The Python provider's `arityCompatibility` predates the
// RegistryProviders contract and uses `(def, callsite)` argument
// order. The contract is `(callsite, def)`. Adapt at the boundary
// so the provider source stays untouched.
arityCompatibility: (callsite, def) => pythonArityCompatibility(def, callsite),
};
const { referenceIndex, stats: resolveStats } = resolveReferenceSites({
scopes: indexes,
providers,
});
// ── Phase 4: emit graph edges ───────────────────────────────────────────
// Order matters: run the Python-specific receiver-bound and free-call
// passes FIRST so they record (filePath, line, col) keys for sites
// they emit edges for. The shared resolver then skips those sites in
// `emitReferencesViaLookup` so its potentially-wrong fallback (e.g.
// resolving `app_metrics.get_metrics()` to a same-named local function
// instead of the namespace target) doesn't fight the precise emission.
const handledSites = new Set<string>();
const receiverExtras = emitReceiverBoundCalls(
graph,
indexes,
parsedFiles,
nodeLookup,
handledSites,
pythonEmitProviderInline,
);
const freeCallExtras = emitFreeCallFallback(
graph,
indexes,
parsedFiles,
nodeLookup,
referenceIndex,
handledSites,
);
// The shared `emit-references.ts` emits edges between
// `SymbolDefinition.nodeId` values, which use the scope-extractor's
// `def:<file>#<line>:<col>:<type>:<name>` format. The CLI's existing
// graph nodes (created by `parsing-processor.ts`) use the legacy
// `<Type>:<file>:<qualifiedName>` ID format. Bridging is required so
// edges actually link to existing graph nodes.
const { emitted, skipped } = emitReferencesViaLookup(
graph,
indexes,
referenceIndex,
nodeLookup,
handledSites,
);
// IMPORTS edges: the scope-resolution path now owns Python file→file
// IMPORTS edge emission when `REGISTRY_PRIMARY_PYTHON=1`. The legacy
// `processImports` path still runs (heritage needs its `importMap`
// population for `ctx.resolve`), but import-processor's graph edge
// emission is gated per-language in `createImportEdgeHelpers` so
// Python no longer double-emits.
const importsEmitted = emitImportEdges(
graph,
indexes.imports,
indexes.scopeTree,
'python-scope: import',
);
return {
filesProcessed: parsedFiles.length,
filesSkipped,
importsEmitted,
resolve: resolveStats,
referenceEdgesEmitted: emitted + receiverExtras + freeCallExtras,
referenceSkipped: skipped,
};
}
/** Minimal `EmitProvider` carrying only the hooks the receiver-bound
* pass currently consults. The full provider (with mergeBindings,
* resolveImportTarget, arityCompatibility, buildMro, populateOwners)
* lands in G-Unit 6 when it moves to `languages/python/emit/`. */
const pythonEmitProviderInline: Pick<
EmitProvider,
'language' | 'isSuperReceiver' | 'fieldFallbackOnMethodLookup'
> = {
language: SupportedLanguages.Python,
isSuperReceiver: (text) => /^super\s*\(/.test(text),
fieldFallbackOnMethodLookup: true,
};