gitnexus 1.6.6-rc.86 → 1.6.6-rc.87

package/dist/core/ingestion/cobol-processor.js

@@ -101,9 +101,26 @@ export const processCobol = (graph, files, allPathSet) => {
         const entry = copybookMap.get(name.toUpperCase());
         return entry ? entry.path : null;
     };
+    // Memoize preprocessed copybook content for the duration of this
+    // processCobol call. A single copybook is COPYed by many programs (and at
+    // many COPY sites within a program); without this cache
+    // preprocessCobolSource would re-run once per COPY site —
+    // O(programs × copybooks) preprocessing passes over the same content.
+    // Keyed by the resolved copybook path. REPLACING is applied later by the
+    // expander on the returned (pre-REPLACING) content (see
+    // cobol-copy-expander.ts readFile→applyReplacing), so caching the
+    // pre-REPLACING preprocessed text here is safe and per-call-scoped.
+    const preprocessedCopyCache = new Map();
     const readCopy = (copyPath) => {
+        const cached = preprocessedCopyCache.get(copyPath);
+        if (cached !== undefined)
+            return cached;
         const content = copybookByPath.get(copyPath);
-        return content ? preprocessCobolSource(content) : null;
+        if (!content)
+            return null; // preserves original falsy→null (missing/empty)
+        const preprocessed = preprocessCobolSource(content);
+        preprocessedCopyCache.set(copyPath, preprocessed);
+        return preprocessed;
     };
     // Track module names for cross-program CALL resolution
     const moduleNodeIds = new Map(); // uppercase program name -> node id

package/dist/core/ingestion/languages/cobol/captures.js

@@ -59,7 +59,7 @@ export function emitCobolScopeCaptures(sourceText, _filePath, _cachedTree) {
             ? rangeOf(progIdLine, 7, progIdLine, lines[progIdLine - 1]?.length ?? endCol)
             : rangeOf(startLine, startCol, endLine, endCol);
         const grouped = {
-            '@scope.module': capture('@scope.module', nameRange, name),
+            '@scope.module': capture('@scope.module', rangeOf(startLine, startCol, endLine, endCol), name),
             '@declaration.program': capture('@declaration.program', rangeOf(startLine, startCol, endLine, endCol), name),
             '@declaration.name': capture('@declaration.name', nameRange, name),
         };
@@ -83,7 +83,7 @@ export function emitCobolScopeCaptures(sourceText, _filePath, _cachedTree) {
             ? rangeOf(progIdLine, 7, progIdLine, lines[progIdLine - 1]?.length ?? endCol)
             : rangeOf(startLine, startCol, endLine, endCol);
         const grouped = {
-            '@scope.module': capture('@scope.module', nameRange, prog.name),
+            '@scope.module': capture('@scope.module', rangeOf(startLine, startCol, endLine, endCol), prog.name),
             '@declaration.program': capture('@declaration.program', rangeOf(startLine, startCol, endLine, endCol), prog.name),
             '@declaration.name': capture('@declaration.name', nameRange, prog.name),
         };

package/dist/core/ingestion/registry-primary-flag.js

@@ -79,6 +79,7 @@ export const MIGRATED_LANGUAGES = new Set([
     SupportedLanguages.Java,
     SupportedLanguages.Rust,
     SupportedLanguages.Ruby,
+    SupportedLanguages.Cobol,
 ]);
 /**
  * Return the env-var name that controls a given language's registry-

package/dist/core/ingestion/scope-resolution/pipeline/phase.js

@@ -81,6 +81,20 @@ export const scopeResolutionPhase = {
         for (const pf of workerParsedFiles) {
             preExtractedByPath.set(pf.filePath, pf);
         }
+        // Drop pre-extracted entries for standalone providers — these
+        // languages are skipped by the canonical guard below (line 164)
+        // and never consume preExtractedByPath, so holding onto their
+        // entries leaks memory until the cleanup loop at 262-264 which
+        // also never runs for skipped providers.
+        for (const [path] of preExtractedByPath) {
+            const lang = getLanguageFromFilename(path);
+            if (lang === null)
+                continue;
+            const provider = SCOPE_RESOLVERS.get(lang);
+            if (provider?.languageProvider.parseStrategy === 'standalone') {
+                preExtractedByPath.delete(path);
+            }
+        }
         let totalFiles = 0;
         let totalImports = 0;
         let totalRefs = 0;
@@ -114,6 +128,14 @@ export const scopeResolutionPhase = {
         for (const [lang, provider] of SCOPE_RESOLVERS) {
             if (!isRegistryPrimary(lang))
                 continue;
+            // Standalone providers (COBOL, JCL) don't emit graph edges yet
+            // through the scope-resolution path. This is the canonical guard:
+            // runScopeResolution is never called for standalone providers, which
+            // keeps cobolPhase as the sole IMPORTS edge producer. Keep this guard
+            // in sync with any additional standalone providers added to
+            // SCOPE_RESOLVERS.
+            if (provider.languageProvider.parseStrategy === 'standalone')
+                continue;
             const langFiles = scannedFiles.filter((f) => getLanguageFromFilename(f.path) === lang);
             if (langFiles.length === 0)
                 continue;

package/dist/core/ingestion/workers/parse-worker.js

@@ -473,9 +473,29 @@ const processBatch = (files, onProgress) => {
     for (const [language, langFiles] of byLanguage) {
         const provider = getProvider(language);
         const queryString = provider.treeSitterQueries;
-        if (!queryString)
+        if (!queryString) {
+            // Standalone providers (regex-based, no tree-sitter) that implement
+            // emitScopeCaptures feed into the scope-resolution pipeline via
+            // extractParsedFile directly — no tree-sitter involved.
+            if (provider.emitScopeCaptures) {
+                for (const file of langFiles) {
+                    const parsedFile = extractParsedFile(provider, file.content, file.path, (message) => {
+                        if (parentPort) {
+                            parentPort.postMessage({ type: 'warning', message });
+                        }
+                        else {
+                            logger.warn(message);
+                        }
+                    }, undefined);
+                    if (parsedFile !== undefined) {
+                        result.parsedFiles.push(parsedFile);
+                        result.fileCount++;
+                        onFileProcessed?.();
+                    }
+                }
+            }
             continue;
-        // Track if we need to handle tsx separately
+        }
         const tsxFiles = [];
         const regularFiles = [];
         if (language === SupportedLanguages.TypeScript) {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "gitnexus",
-  "version": "1.6.6-rc.86",
+  "version": "1.6.6-rc.87",
   "description": "Graph-powered code intelligence for AI agents. Index any codebase, query via MCP or CLI.",
   "author": "Abhigyan Patwari",
   "license": "PolyForm-Noncommercial-1.0.0",