@mnemonik/shared 7.142.0 → 7.144.0

package/src/ast/astChunker.ts

@@ -57,7 +57,12 @@
 
 import type { Node, Query, Tree } from 'web-tree-sitter';
 import { Parser } from 'web-tree-sitter';
-import { loadGrammar, type AstLanguageId, type GrammarBinding } from './grammars.js';
+import {
+  loadGrammar,
+  objcppSyntaxIssue,
+  type AstLanguageId,
+  type GrammarBinding,
+} from './grammars.js';
 
 export interface AstChunk {
   /** Verbatim contiguous source: exactly the lines `startLine..endLine`. */
@@ -298,6 +303,10 @@ export async function chunkWithAst(
   const parseMs = performance.now() - startedAt;
 
   try {
+    if (languageId === 'objcpp') {
+      const issue = objcppSyntaxIssue(tree.rootNode);
+      if (issue) return { unsupported: 'parse_failed', detail: issue };
+    }
     const lines = content.split('\n');
     // A trailing newline TERMINATES the last line, it does not open a new one,
     // but `split` yields a phantom '' for it. Dropping it here - before

package/src/ast/grammars.ts

@@ -33,7 +33,7 @@
 import { existsSync, readFileSync, statSync } from 'node:fs';
 import { dirname, extname, join } from 'node:path';
 import { fileURLToPath } from 'node:url';
-import { Language, Parser, Query } from 'web-tree-sitter';
+import { Language, Parser, Query, type Node } from 'web-tree-sitter';
 import { warn } from '../logger.js';
 
 const AST_LANGUAGE_ID_LIST = [
@@ -45,6 +45,7 @@ const AST_LANGUAGE_ID_LIST = [
   'java',
   'c',
   'cpp',
+  'objcpp',
   'csharp',
   'ruby',
   'php',
@@ -55,6 +56,12 @@ const AST_LANGUAGE_ID_LIST = [
   'bash',
   'groovy',
   'powershell',
+  'swift',
+  'sql',
+  'lua',
+  'terraform',
+  'nix',
+  'cmake',
 ] as const;
 
 /** A language this build can AST-parse. Exactly the vendored set, no aspirations. */
@@ -113,6 +120,7 @@ export const QUERY_CHAIN: Readonly<Record<AstLanguageId, readonly AstLanguageId[
   java: ['java'],
   c: ['c'],
   cpp: ['c', 'cpp'],
+  objcpp: ['objcpp'],
   csharp: ['csharp'],
   ruby: ['ruby'],
   php: ['php'],
@@ -123,6 +131,12 @@ export const QUERY_CHAIN: Readonly<Record<AstLanguageId, readonly AstLanguageId[
   bash: ['bash'],
   groovy: ['groovy'],
   powershell: ['powershell'],
+  swift: ['swift'],
+  sql: ['sql'],
+  lua: ['lua'],
+  terraform: ['terraform'],
+  nix: ['nix'],
+  cmake: ['cmake'],
 };
 
 /**
@@ -135,8 +149,30 @@ export const QUERY_CHAIN: Readonly<Record<AstLanguageId, readonly AstLanguageId[
  */
 const EXTENSION_GRAMMARS: Readonly<Record<string, AstLanguageId>> = {
   '.tsx': 'tsx',
+  '.m': 'objcpp',
+  '.mm': 'objcpp',
 };
 
+/** Upstream permits a bodyless else for highlighting across preprocessor boundaries. */
+export function objcppSyntaxIssue(root: Node): string | null {
+  if (root.hasError) return 'Objective-C++ contains ERROR or MISSING nodes';
+  const cursor = root.walk();
+  try {
+    for (;;) {
+      const node = cursor.currentNode;
+      if (node.type === 'else_clause' && node.namedChildCount === 0)
+        return 'Objective-C++ else clause has no body; conditional-source recovery is not admitted';
+      if (cursor.gotoFirstChild()) continue;
+      for (;;) {
+        if (cursor.gotoNextSibling()) break;
+        if (!cursor.gotoParent()) return null;
+      }
+    }
+  } finally {
+    cursor.delete();
+  }
+}
+
 /**
  * Scanner language string -> grammar. Keyed on what `languageForExtension`
  * actually returns, so this table and the extension table cannot disagree about
@@ -144,7 +180,7 @@ const EXTENSION_GRAMMARS: Readonly<Record<string, AstLanguageId>> = {
  *
  * Absences are deliberate answers, not gaps. `markdown` is missing because
  * headers already are a document's semantic unit and 10,483 production chunks
- * depend on `chunkMarkdown` keeping them; `kotlin`, `swift`, `dart` and friends
+ * depend on `chunkMarkdown` keeping them; `kotlin`, `dart` and friends
  * are missing because no loadable grammar is vendored yet. Both resolve to
  * `null` and the heuristic chunker takes over.
  */
@@ -166,6 +202,12 @@ const LANGUAGE_GRAMMARS: Readonly<Record<string, AstLanguageId>> = {
   shell: 'bash',
   groovy: 'groovy',
   powershell: 'powershell',
+  swift: 'swift',
+  sql: 'sql',
+  lua: 'lua',
+  terraform: 'terraform',
+  nix: 'nix',
+  cmake: 'cmake',
 };
 
 /** A loaded grammar plus its compiled query. Compiled once, reused per file. */

package/src/codeScanner.ts

@@ -532,6 +532,7 @@ const EXTENSION_LANGUAGES: Readonly<Record<string, string>> = {
  * preference) need the same answer without constructing a scanner.
  */
 export function languageForExtension(filePathOrExt: string): string {
+  if (isCmakeListsPath(filePathOrExt)) return 'cmake';
   // `extname` FIRST. A dotfile that carries a real extension ('.eslintrc.js',
   // '.mocharc.cjs', '.prettierrc.ts') starts with '.' and contains no '/', so
   // a bare-extension-first reading swallowed the whole name and answered
@@ -563,6 +564,13 @@ export function languageForExtension(filePathOrExt: string): string {
 const isMigrationSqlAuthorityPath = (posixRelPath: string): boolean =>
   /^migrations\/.*\.sql$/.test(posixRelPath);
 
+/** Content-only build authority, shared by scanner collection and server admission. */
+export const XCODE_AUTHORITY_PATH = /(?:^|\/)[^/]+\.xcodeproj\/project\.pbxproj$|\.xcconfig$/;
+export const isXcodeAuthorityPath = (path: string): boolean => XCODE_AUTHORITY_PATH.test(path);
+/** Exact CMake build entrypoint; ordinary .txt files remain outside collection. */
+export const CMAKE_LISTS_PATH = /(?:^|\/)CMakeLists\.txt$/;
+export const isCmakeListsPath = (path: string): boolean => CMAKE_LISTS_PATH.test(path);
+
 /**
  * Predicates for paths whose verbatim content is ALREADY shipped by
  * `collectAuthorityFiles` and that carry no additional value as embedded code
@@ -582,6 +590,8 @@ const isMigrationSqlAuthorityPath = (posixRelPath: string): boolean =>
  */
 const AUTHORITY_ONLY_PATH_PREDICATES: readonly ((posixRelPath: string) => boolean)[] = [
   isMigrationSqlAuthorityPath,
+  isXcodeAuthorityPath,
+  isCmakeListsPath,
 ];
 
 /**
@@ -627,6 +637,8 @@ export const AUTHORITY_FILE_MATCHERS: Array<(relPath: string) => boolean> = [
   // `AUTHORITY_ONLY_PATH_PREDICATES` by reference, not by a copied regex, so
   // collection and the chunking exclusion cannot disagree.
   isMigrationSqlAuthorityPath,
+  isXcodeAuthorityPath,
+  isCmakeListsPath,
 ];
 
 /**

package/src/index.ts

@@ -26,6 +26,10 @@ export {
   isFixturePath,
   isSecretFile,
   isAuthorityOnlyPath,
+  XCODE_AUTHORITY_PATH,
+  isXcodeAuthorityPath,
+  isCmakeListsPath,
+  CMAKE_LISTS_PATH,
   isCursorRulePath,
   logAstCapabilityOnce,
   type CodeChunk,
@@ -75,6 +79,7 @@ export {
   loadGrammar,
   loadedGrammarIds,
   resolveAstLanguage,
+  objcppSyntaxIssue,
   type AstLanguageId,
   type GrammarBinding,
   type GrammarLoad,
@@ -175,3 +180,5 @@ export {
 } from './systemdControl.js';
 
 export { pidIsScanner } from './scannerPid.js';
+
+export { parseVueSource, walkVueTemplate, VUE_FIDELITY_MARKER } from './vueSource.js';