@colbymchenry/codegraph-win32-arm64 1.4.1 → 1.6.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (314) hide show
  1. package/bin/codegraph.cmd +1 -1
  2. package/lib/dist/bin/codegraph.js +206 -44
  3. package/lib/dist/bin/codegraph.js.map +1 -1
  4. package/lib/dist/context/formatter.d.ts +7 -1
  5. package/lib/dist/context/formatter.d.ts.map +1 -1
  6. package/lib/dist/context/formatter.js +12 -6
  7. package/lib/dist/context/formatter.js.map +1 -1
  8. package/lib/dist/context/index.d.ts +12 -0
  9. package/lib/dist/context/index.d.ts.map +1 -1
  10. package/lib/dist/context/index.js +101 -25
  11. package/lib/dist/context/index.js.map +1 -1
  12. package/lib/dist/db/index.d.ts +179 -0
  13. package/lib/dist/db/index.d.ts.map +1 -1
  14. package/lib/dist/db/index.js +368 -7
  15. package/lib/dist/db/index.js.map +1 -1
  16. package/lib/dist/db/migrations.d.ts +1 -1
  17. package/lib/dist/db/migrations.d.ts.map +1 -1
  18. package/lib/dist/db/migrations.js +25 -1
  19. package/lib/dist/db/migrations.js.map +1 -1
  20. package/lib/dist/db/queries.d.ts +234 -5
  21. package/lib/dist/db/queries.d.ts.map +1 -1
  22. package/lib/dist/db/queries.js +687 -53
  23. package/lib/dist/db/queries.js.map +1 -1
  24. package/lib/dist/db/schema.sql +16 -3
  25. package/lib/dist/db/sqlite-adapter.d.ts +3 -1
  26. package/lib/dist/db/sqlite-adapter.d.ts.map +1 -1
  27. package/lib/dist/db/sqlite-adapter.js +22 -4
  28. package/lib/dist/db/sqlite-adapter.js.map +1 -1
  29. package/lib/dist/db/wal-valve.d.ts +15 -2
  30. package/lib/dist/db/wal-valve.d.ts.map +1 -1
  31. package/lib/dist/db/wal-valve.js +88 -6
  32. package/lib/dist/db/wal-valve.js.map +1 -1
  33. package/lib/dist/directory.d.ts +29 -0
  34. package/lib/dist/directory.d.ts.map +1 -1
  35. package/lib/dist/directory.js +52 -0
  36. package/lib/dist/directory.js.map +1 -1
  37. package/lib/dist/extraction/extraction-version.d.ts +1 -1
  38. package/lib/dist/extraction/extraction-version.js +1 -1
  39. package/lib/dist/extraction/function-ref.d.ts.map +1 -1
  40. package/lib/dist/extraction/function-ref.js +6 -0
  41. package/lib/dist/extraction/function-ref.js.map +1 -1
  42. package/lib/dist/extraction/generated-detection.d.ts +48 -12
  43. package/lib/dist/extraction/generated-detection.d.ts.map +1 -1
  44. package/lib/dist/extraction/generated-detection.js +176 -11
  45. package/lib/dist/extraction/generated-detection.js.map +1 -1
  46. package/lib/dist/extraction/grammars.d.ts.map +1 -1
  47. package/lib/dist/extraction/grammars.js +101 -2
  48. package/lib/dist/extraction/grammars.js.map +1 -1
  49. package/lib/dist/extraction/index.d.ts +138 -3
  50. package/lib/dist/extraction/index.d.ts.map +1 -1
  51. package/lib/dist/extraction/index.js +518 -94
  52. package/lib/dist/extraction/index.js.map +1 -1
  53. package/lib/dist/extraction/kernel/decode.d.ts +9 -0
  54. package/lib/dist/extraction/kernel/decode.d.ts.map +1 -0
  55. package/lib/dist/extraction/kernel/decode.js +162 -0
  56. package/lib/dist/extraction/kernel/decode.js.map +1 -0
  57. package/lib/dist/extraction/kernel/index.d.ts +64 -0
  58. package/lib/dist/extraction/kernel/index.d.ts.map +1 -0
  59. package/lib/dist/extraction/kernel/index.js +279 -0
  60. package/lib/dist/extraction/kernel/index.js.map +1 -0
  61. package/lib/dist/extraction/kernel/layout.d.ts +101 -0
  62. package/lib/dist/extraction/kernel/layout.d.ts.map +1 -0
  63. package/lib/dist/extraction/kernel/layout.js +104 -0
  64. package/lib/dist/extraction/kernel/layout.js.map +1 -0
  65. package/lib/dist/extraction/kernel/loader.d.ts +103 -0
  66. package/lib/dist/extraction/kernel/loader.d.ts.map +1 -0
  67. package/lib/dist/extraction/kernel/loader.js +146 -0
  68. package/lib/dist/extraction/kernel/loader.js.map +1 -0
  69. package/lib/dist/extraction/languages/c-cpp.d.ts +14 -0
  70. package/lib/dist/extraction/languages/c-cpp.d.ts.map +1 -1
  71. package/lib/dist/extraction/languages/c-cpp.js +868 -16
  72. package/lib/dist/extraction/languages/c-cpp.js.map +1 -1
  73. package/lib/dist/extraction/languages/erlang.d.ts.map +1 -1
  74. package/lib/dist/extraction/languages/erlang.js +58 -22
  75. package/lib/dist/extraction/languages/erlang.js.map +1 -1
  76. package/lib/dist/extraction/languages/objc.d.ts.map +1 -1
  77. package/lib/dist/extraction/languages/objc.js +4 -0
  78. package/lib/dist/extraction/languages/objc.js.map +1 -1
  79. package/lib/dist/extraction/languages/rust.d.ts +20 -0
  80. package/lib/dist/extraction/languages/rust.d.ts.map +1 -1
  81. package/lib/dist/extraction/languages/rust.js +47 -20
  82. package/lib/dist/extraction/languages/rust.js.map +1 -1
  83. package/lib/dist/extraction/parse-pool.d.ts +15 -0
  84. package/lib/dist/extraction/parse-pool.d.ts.map +1 -1
  85. package/lib/dist/extraction/parse-pool.js +35 -1
  86. package/lib/dist/extraction/parse-pool.js.map +1 -1
  87. package/lib/dist/extraction/parse-worker.js +35 -1
  88. package/lib/dist/extraction/parse-worker.js.map +1 -1
  89. package/lib/dist/extraction/store-worker.d.ts +24 -0
  90. package/lib/dist/extraction/store-worker.d.ts.map +1 -0
  91. package/lib/dist/extraction/store-worker.js +139 -0
  92. package/lib/dist/extraction/store-worker.js.map +1 -0
  93. package/lib/dist/extraction/store-writer.d.ts +63 -0
  94. package/lib/dist/extraction/store-writer.d.ts.map +1 -0
  95. package/lib/dist/extraction/store-writer.js +174 -0
  96. package/lib/dist/extraction/store-writer.js.map +1 -0
  97. package/lib/dist/extraction/tree-sitter-types.d.ts +2 -0
  98. package/lib/dist/extraction/tree-sitter-types.d.ts.map +1 -1
  99. package/lib/dist/extraction/tree-sitter.d.ts +27 -5
  100. package/lib/dist/extraction/tree-sitter.d.ts.map +1 -1
  101. package/lib/dist/extraction/tree-sitter.js +257 -56
  102. package/lib/dist/extraction/tree-sitter.js.map +1 -1
  103. package/lib/dist/extraction/wasm/tree-sitter-c.wasm +0 -0
  104. package/lib/dist/extraction/wasm/tree-sitter-cpp.wasm +0 -0
  105. package/lib/dist/extraction/wasm/tree-sitter-dart.wasm +0 -0
  106. package/lib/dist/extraction/wasm/tree-sitter-go.wasm +0 -0
  107. package/lib/dist/extraction/wasm/tree-sitter-java.wasm +0 -0
  108. package/lib/dist/extraction/wasm/tree-sitter-javascript.wasm +0 -0
  109. package/lib/dist/extraction/wasm/tree-sitter-kotlin.wasm +0 -0
  110. package/lib/dist/extraction/wasm/tree-sitter-php.wasm +0 -0
  111. package/lib/dist/extraction/wasm/tree-sitter-python.wasm +0 -0
  112. package/lib/dist/extraction/wasm/tree-sitter-ruby.wasm +0 -0
  113. package/lib/dist/extraction/wasm/tree-sitter-rust.wasm +0 -0
  114. package/lib/dist/extraction/wasm/tree-sitter-swift.wasm +0 -0
  115. package/lib/dist/extraction/wasm/tree-sitter-tsx.wasm +0 -0
  116. package/lib/dist/extraction/wasm/tree-sitter-typescript.wasm +0 -0
  117. package/lib/dist/extraction/wasm-runtime-flags.d.ts +15 -0
  118. package/lib/dist/extraction/wasm-runtime-flags.d.ts.map +1 -1
  119. package/lib/dist/extraction/wasm-runtime-flags.js +27 -3
  120. package/lib/dist/extraction/wasm-runtime-flags.js.map +1 -1
  121. package/lib/dist/graph/queries.d.ts.map +1 -1
  122. package/lib/dist/graph/queries.js +2 -0
  123. package/lib/dist/graph/queries.js.map +1 -1
  124. package/lib/dist/graph/traversal.js +1 -1
  125. package/lib/dist/graph/traversal.js.map +1 -1
  126. package/lib/dist/index.d.ts +36 -2
  127. package/lib/dist/index.d.ts.map +1 -1
  128. package/lib/dist/index.js +305 -11
  129. package/lib/dist/index.js.map +1 -1
  130. package/lib/dist/installer/beta-signup.d.ts +54 -0
  131. package/lib/dist/installer/beta-signup.d.ts.map +1 -0
  132. package/lib/dist/installer/beta-signup.js +178 -0
  133. package/lib/dist/installer/beta-signup.js.map +1 -0
  134. package/lib/dist/installer/index.d.ts +4 -3
  135. package/lib/dist/installer/index.d.ts.map +1 -1
  136. package/lib/dist/installer/index.js +19 -6
  137. package/lib/dist/installer/index.js.map +1 -1
  138. package/lib/dist/installer/targets/claude.d.ts.map +1 -1
  139. package/lib/dist/installer/targets/claude.js +37 -5
  140. package/lib/dist/installer/targets/claude.js.map +1 -1
  141. package/lib/dist/installer/targets/codex.d.ts +22 -8
  142. package/lib/dist/installer/targets/codex.d.ts.map +1 -1
  143. package/lib/dist/installer/targets/codex.js +65 -48
  144. package/lib/dist/installer/targets/codex.js.map +1 -1
  145. package/lib/dist/installer/targets/copilot-cli.d.ts +32 -0
  146. package/lib/dist/installer/targets/copilot-cli.d.ts.map +1 -0
  147. package/lib/dist/installer/targets/copilot-cli.js +209 -0
  148. package/lib/dist/installer/targets/copilot-cli.js.map +1 -0
  149. package/lib/dist/installer/targets/copilot-jetbrains.d.ts +42 -0
  150. package/lib/dist/installer/targets/copilot-jetbrains.d.ts.map +1 -0
  151. package/lib/dist/installer/targets/copilot-jetbrains.js +238 -0
  152. package/lib/dist/installer/targets/copilot-jetbrains.js.map +1 -0
  153. package/lib/dist/installer/targets/copilot-vscode.d.ts +47 -0
  154. package/lib/dist/installer/targets/copilot-vscode.d.ts.map +1 -0
  155. package/lib/dist/installer/targets/copilot-vscode.js +219 -0
  156. package/lib/dist/installer/targets/copilot-vscode.js.map +1 -0
  157. package/lib/dist/installer/targets/registry.d.ts.map +1 -1
  158. package/lib/dist/installer/targets/registry.js +6 -0
  159. package/lib/dist/installer/targets/registry.js.map +1 -1
  160. package/lib/dist/installer/targets/toml.d.ts +6 -5
  161. package/lib/dist/installer/targets/toml.d.ts.map +1 -1
  162. package/lib/dist/installer/targets/toml.js +99 -20
  163. package/lib/dist/installer/targets/toml.js.map +1 -1
  164. package/lib/dist/installer/targets/types.d.ts +5 -5
  165. package/lib/dist/installer/targets/types.d.ts.map +1 -1
  166. package/lib/dist/mcp/daemon-manager.d.ts +1 -1
  167. package/lib/dist/mcp/daemon-manager.d.ts.map +1 -1
  168. package/lib/dist/mcp/daemon-manager.js +1 -1
  169. package/lib/dist/mcp/daemon-manager.js.map +1 -1
  170. package/lib/dist/mcp/daemon-paths.d.ts +6 -0
  171. package/lib/dist/mcp/daemon-paths.d.ts.map +1 -1
  172. package/lib/dist/mcp/daemon-paths.js +54 -0
  173. package/lib/dist/mcp/daemon-paths.js.map +1 -1
  174. package/lib/dist/mcp/daemon-registry.d.ts +10 -0
  175. package/lib/dist/mcp/daemon-registry.d.ts.map +1 -1
  176. package/lib/dist/mcp/daemon-registry.js +51 -3
  177. package/lib/dist/mcp/daemon-registry.js.map +1 -1
  178. package/lib/dist/mcp/daemon.d.ts +9 -6
  179. package/lib/dist/mcp/daemon.d.ts.map +1 -1
  180. package/lib/dist/mcp/daemon.js +10 -8
  181. package/lib/dist/mcp/daemon.js.map +1 -1
  182. package/lib/dist/mcp/engine.d.ts +3 -0
  183. package/lib/dist/mcp/engine.d.ts.map +1 -1
  184. package/lib/dist/mcp/engine.js +45 -3
  185. package/lib/dist/mcp/engine.js.map +1 -1
  186. package/lib/dist/mcp/explore-dedup.d.ts +137 -0
  187. package/lib/dist/mcp/explore-dedup.d.ts.map +1 -0
  188. package/lib/dist/mcp/explore-dedup.js +236 -0
  189. package/lib/dist/mcp/explore-dedup.js.map +1 -0
  190. package/lib/dist/mcp/explore-diagnostics.d.ts +289 -0
  191. package/lib/dist/mcp/explore-diagnostics.d.ts.map +1 -0
  192. package/lib/dist/mcp/explore-diagnostics.js +558 -0
  193. package/lib/dist/mcp/explore-diagnostics.js.map +1 -0
  194. package/lib/dist/mcp/explore-session-state.d.ts +217 -0
  195. package/lib/dist/mcp/explore-session-state.d.ts.map +1 -0
  196. package/lib/dist/mcp/explore-session-state.js +322 -0
  197. package/lib/dist/mcp/explore-session-state.js.map +1 -0
  198. package/lib/dist/mcp/index.d.ts +24 -0
  199. package/lib/dist/mcp/index.d.ts.map +1 -1
  200. package/lib/dist/mcp/index.js +67 -6
  201. package/lib/dist/mcp/index.js.map +1 -1
  202. package/lib/dist/mcp/liveness-watchdog.d.ts.map +1 -1
  203. package/lib/dist/mcp/liveness-watchdog.js +3 -1
  204. package/lib/dist/mcp/liveness-watchdog.js.map +1 -1
  205. package/lib/dist/mcp/proxy.d.ts.map +1 -1
  206. package/lib/dist/mcp/proxy.js +6 -1
  207. package/lib/dist/mcp/proxy.js.map +1 -1
  208. package/lib/dist/mcp/server-instructions.d.ts +2 -2
  209. package/lib/dist/mcp/server-instructions.d.ts.map +1 -1
  210. package/lib/dist/mcp/server-instructions.js +8 -3
  211. package/lib/dist/mcp/server-instructions.js.map +1 -1
  212. package/lib/dist/mcp/session.d.ts +16 -0
  213. package/lib/dist/mcp/session.d.ts.map +1 -1
  214. package/lib/dist/mcp/session.js +36 -13
  215. package/lib/dist/mcp/session.js.map +1 -1
  216. package/lib/dist/mcp/tools.d.ts +276 -10
  217. package/lib/dist/mcp/tools.d.ts.map +1 -1
  218. package/lib/dist/mcp/tools.js +2425 -237
  219. package/lib/dist/mcp/tools.js.map +1 -1
  220. package/lib/dist/project-config.d.ts +23 -0
  221. package/lib/dist/project-config.d.ts.map +1 -1
  222. package/lib/dist/project-config.js +39 -2
  223. package/lib/dist/project-config.js.map +1 -1
  224. package/lib/dist/resolution/c-fnptr-synthesizer.d.ts +1 -1
  225. package/lib/dist/resolution/c-fnptr-synthesizer.d.ts.map +1 -1
  226. package/lib/dist/resolution/c-fnptr-synthesizer.js +582 -138
  227. package/lib/dist/resolution/c-fnptr-synthesizer.js.map +1 -1
  228. package/lib/dist/resolution/callback-synthesizer.d.ts +54 -10
  229. package/lib/dist/resolution/callback-synthesizer.d.ts.map +1 -1
  230. package/lib/dist/resolution/callback-synthesizer.js +287 -167
  231. package/lib/dist/resolution/callback-synthesizer.js.map +1 -1
  232. package/lib/dist/resolution/cooperative-yield.d.ts +10 -2
  233. package/lib/dist/resolution/cooperative-yield.d.ts.map +1 -1
  234. package/lib/dist/resolution/cooperative-yield.js +6 -4
  235. package/lib/dist/resolution/cooperative-yield.js.map +1 -1
  236. package/lib/dist/resolution/frameworks/go.d.ts.map +1 -1
  237. package/lib/dist/resolution/frameworks/go.js +29 -6
  238. package/lib/dist/resolution/frameworks/go.js.map +1 -1
  239. package/lib/dist/resolution/frameworks/swift.d.ts.map +1 -1
  240. package/lib/dist/resolution/frameworks/swift.js +11 -1
  241. package/lib/dist/resolution/frameworks/swift.js.map +1 -1
  242. package/lib/dist/resolution/import-resolver.d.ts +2 -3
  243. package/lib/dist/resolution/import-resolver.d.ts.map +1 -1
  244. package/lib/dist/resolution/import-resolver.js +184 -9
  245. package/lib/dist/resolution/import-resolver.js.map +1 -1
  246. package/lib/dist/resolution/index.d.ts +87 -2
  247. package/lib/dist/resolution/index.d.ts.map +1 -1
  248. package/lib/dist/resolution/index.js +731 -134
  249. package/lib/dist/resolution/index.js.map +1 -1
  250. package/lib/dist/resolution/memory-budget.d.ts +48 -0
  251. package/lib/dist/resolution/memory-budget.d.ts.map +1 -0
  252. package/lib/dist/resolution/memory-budget.js +162 -0
  253. package/lib/dist/resolution/memory-budget.js.map +1 -0
  254. package/lib/dist/resolution/name-matcher.d.ts +42 -1
  255. package/lib/dist/resolution/name-matcher.d.ts.map +1 -1
  256. package/lib/dist/resolution/name-matcher.js +681 -99
  257. package/lib/dist/resolution/name-matcher.js.map +1 -1
  258. package/lib/dist/resolution/path-aliases.d.ts +6 -4
  259. package/lib/dist/resolution/path-aliases.d.ts.map +1 -1
  260. package/lib/dist/resolution/path-aliases.js +125 -17
  261. package/lib/dist/resolution/path-aliases.js.map +1 -1
  262. package/lib/dist/resolution/resolver-pool.d.ts +106 -0
  263. package/lib/dist/resolution/resolver-pool.d.ts.map +1 -0
  264. package/lib/dist/resolution/resolver-pool.js +344 -0
  265. package/lib/dist/resolution/resolver-pool.js.map +1 -0
  266. package/lib/dist/resolution/resolver-worker.d.ts +17 -0
  267. package/lib/dist/resolution/resolver-worker.d.ts.map +1 -0
  268. package/lib/dist/resolution/resolver-worker.js +141 -0
  269. package/lib/dist/resolution/resolver-worker.js.map +1 -0
  270. package/lib/dist/resolution/types.d.ts +3 -0
  271. package/lib/dist/resolution/types.d.ts.map +1 -1
  272. package/lib/dist/search/identifier-segments.d.ts +10 -0
  273. package/lib/dist/search/identifier-segments.d.ts.map +1 -1
  274. package/lib/dist/search/identifier-segments.js +29 -0
  275. package/lib/dist/search/identifier-segments.js.map +1 -1
  276. package/lib/dist/search/query-paths.d.ts +57 -0
  277. package/lib/dist/search/query-paths.d.ts.map +1 -0
  278. package/lib/dist/search/query-paths.js +298 -0
  279. package/lib/dist/search/query-paths.js.map +1 -0
  280. package/lib/dist/search/query-utils.d.ts +1 -1
  281. package/lib/dist/search/query-utils.d.ts.map +1 -1
  282. package/lib/dist/search/query-utils.js +13 -3
  283. package/lib/dist/search/query-utils.js.map +1 -1
  284. package/lib/dist/sync/watcher.d.ts +47 -1
  285. package/lib/dist/sync/watcher.d.ts.map +1 -1
  286. package/lib/dist/sync/watcher.js +127 -8
  287. package/lib/dist/sync/watcher.js.map +1 -1
  288. package/lib/dist/types.d.ts +61 -3
  289. package/lib/dist/types.d.ts.map +1 -1
  290. package/lib/dist/types.js +27 -1
  291. package/lib/dist/types.js.map +1 -1
  292. package/lib/dist/ui/color.d.ts +21 -0
  293. package/lib/dist/ui/color.d.ts.map +1 -0
  294. package/lib/dist/ui/color.js +42 -0
  295. package/lib/dist/ui/color.js.map +1 -0
  296. package/lib/dist/ui/glyphs.d.ts +35 -10
  297. package/lib/dist/ui/glyphs.d.ts.map +1 -1
  298. package/lib/dist/ui/glyphs.js +59 -12
  299. package/lib/dist/ui/glyphs.js.map +1 -1
  300. package/lib/dist/ui/shimmer-progress.d.ts.map +1 -1
  301. package/lib/dist/ui/shimmer-progress.js +78 -4
  302. package/lib/dist/ui/shimmer-progress.js.map +1 -1
  303. package/lib/dist/ui/shimmer-worker.js +25 -17
  304. package/lib/dist/ui/shimmer-worker.js.map +1 -1
  305. package/lib/dist/ui/types.d.ts +0 -2
  306. package/lib/dist/ui/types.d.ts.map +1 -1
  307. package/lib/dist/upgrade/index.d.ts +7 -0
  308. package/lib/dist/upgrade/index.d.ts.map +1 -1
  309. package/lib/dist/upgrade/index.js +18 -5
  310. package/lib/dist/upgrade/index.js.map +1 -1
  311. package/lib/kernel/codegraph-kernel.node +0 -0
  312. package/lib/node_modules/.package-lock.json +1 -1
  313. package/lib/package.json +6 -1
  314. package/package.json +6 -2
@@ -47,6 +47,7 @@ exports.buildDefaultIgnore = buildDefaultIgnore;
47
47
  exports.buildScopeIgnore = buildScopeIgnore;
48
48
  exports.discoverEmbeddedRepoRoots = discoverEmbeddedRepoRoots;
49
49
  exports.findUnindexedIgnoredRepos = findUnindexedIgnoredRepos;
50
+ exports.getGitChangedFiles = getGitChangedFiles;
50
51
  exports.scanDirectory = scanDirectory;
51
52
  exports.scanDirectoryAsync = scanDirectoryAsync;
52
53
  const fs = __importStar(require("fs"));
@@ -57,6 +58,9 @@ const crypto = __importStar(require("crypto"));
57
58
  const child_process_1 = require("child_process");
58
59
  const tree_sitter_1 = require("./tree-sitter");
59
60
  const parse_pool_1 = require("./parse-pool");
61
+ const store_writer_1 = require("./store-writer");
62
+ const kernel_1 = require("./kernel");
63
+ const generated_detection_1 = require("./generated-detection");
60
64
  const grammars_1 = require("./grammars");
61
65
  const project_config_1 = require("../project-config");
62
66
  const directory_1 = require("../directory");
@@ -807,6 +811,11 @@ function findUnindexedIgnoredRepos(rootDir) {
807
811
  if (includeIgnored?.ignores((0, utils_1.normalizePath)(dir)))
808
812
  continue; // already opted in — nothing to nag about
809
813
  for (const repo of findNestedGitRepos(path.join(rootDir, dir), dir)) {
814
+ // Per-repo opt-in check, mirroring findIgnoredEmbeddedRepos: a child
815
+ // pattern (`repos/a/`) doesn't match the parent dir above but DOES
816
+ // cover this repo — it's indexed, so don't nag about it (#1295).
817
+ if (includeIgnored?.ignores((0, utils_1.normalizePath)(repo)))
818
+ continue;
810
819
  repos.push(repo);
811
820
  if (repos.length >= UNINDEXED_IGNORED_REPO_HINT_CAP)
812
821
  return repos;
@@ -837,9 +846,22 @@ function findIgnoredEmbeddedRepos(repoDir, includeIgnored, prefix) {
837
846
  for (const dir of listIgnoredDirs(repoDir)) {
838
847
  if (defaults.ignores(dir))
839
848
  continue;
840
- if (!includeIgnored.ignores((0, utils_1.normalizePath)(prefix + dir)))
841
- continue;
842
- repos.push(...findNestedGitRepos(path.join(repoDir, dir), dir));
849
+ const nested = findNestedGitRepos(path.join(repoDir, dir), dir);
850
+ if (includeIgnored.ignores((0, utils_1.normalizePath)(prefix + dir))) {
851
+ // The whole ignored dir is opted in — every nested repo under it counts.
852
+ repos.push(...nested);
853
+ }
854
+ else {
855
+ // A single gitignore rule often covers the PARENT of the opted-in
856
+ // repos: `.gitignore: /repos/` lists `repos/` as ONE ignored entry,
857
+ // while `includeIgnored: ["repos/a/"]` (the CLI hint's own suggested
858
+ // spelling) names the child — which never matches the parent path, so
859
+ // the opt-in silently did nothing (#1295). Match each nested repo
860
+ // root individually so both spellings work. The walk is bounded
861
+ // (depth/entry caps in findNestedGitRepos) and only runs when
862
+ // includeIgnored is configured at all.
863
+ repos.push(...nested.filter((r) => includeIgnored.ignores((0, utils_1.normalizePath)(prefix + r))));
864
+ }
843
865
  }
844
866
  return repos;
845
867
  }
@@ -1036,7 +1058,14 @@ function getGitChangedFiles(rootDir) {
1036
1058
  }
1037
1059
  }
1038
1060
  function collectGitStatus(repoDir, prefix, out, overrides, includeIgnored = null, exclude = null) {
1039
- const output = (0, child_process_1.execFileSync)('git', ['status', '--porcelain', '--no-renames'], { cwd: repoDir, encoding: 'utf-8', timeout: 10000, maxBuffer: 50 * 1024 * 1024, stdio: ['pipe', 'pipe', 'pipe'], windowsHide: true });
1061
+ const output = (0, child_process_1.execFileSync)('git',
1062
+ // `-uall` lists individual untracked files instead of collapsing an
1063
+ // entirely-untracked directory into one `?? dir/` entry, which would
1064
+ // otherwise be dropped here (only embedded git repos are recursed into
1065
+ // below). Nested untracked git repos still collapse to `?? repo/` even
1066
+ // with `-uall` — git never crosses a repo boundary — so the recursion
1067
+ // still handles them. (#1213)
1068
+ ['status', '--porcelain', '--no-renames', '-uall'], { cwd: repoDir, encoding: 'utf-8', timeout: 10000, maxBuffer: 50 * 1024 * 1024, stdio: ['pipe', 'pipe', 'pipe'], windowsHide: true });
1040
1069
  // This repo's own ignore rules — built-in defaults (#407) plus its .gitignore.
1041
1070
  // Change detection must exclude the SAME files the full index does, but git
1042
1071
  // status hides neither: it ignores nothing for *tracked* paths, and the
@@ -1323,10 +1352,47 @@ class ExtractionOrchestrator {
1323
1352
  * hasn't run yet so single-file re-index paths can detect on the spot.
1324
1353
  */
1325
1354
  detectedFrameworkNames = null;
1355
+ /**
1356
+ * Scope matcher for SCOPED syncs, memoized on the mtimes of the two root
1357
+ * files it is derived from (`codegraph.json`, `.gitignore`). See
1358
+ * {@link scopedSyncMatcher}.
1359
+ */
1360
+ scopedMatcher = null;
1326
1361
  constructor(rootDir, queries) {
1327
1362
  this.rootDir = rootDir;
1328
1363
  this.queries = queries;
1329
1364
  }
1365
+ /**
1366
+ * The scope matcher a scoped sync applies to the paths it was handed — the
1367
+ * same `buildScopeIgnore` the full scan uses, so an explicitly-passed path
1368
+ * that is OUT of scope (a user `exclude` in `codegraph.json`, a `.gitignore`
1369
+ * rule, a built-in default) is treated exactly as the full walk would treat
1370
+ * it: absent, hence removed if tracked, never parsed (#1590).
1371
+ *
1372
+ * Memoized on the root config + root `.gitignore` mtimes: building the
1373
+ * matcher runs embedded-repo discovery (`git ls-files`), which would defeat
1374
+ * the scoped path's whole point (skipping O(repo) work) if paid per sync.
1375
+ * Two `stat`s per sync while nothing changed. An embedded repo created
1376
+ * between config edits joins the scoped matcher on the next full sync, the
1377
+ * same lifecycle the watcher's own matcher already has.
1378
+ */
1379
+ scopedSyncMatcher() {
1380
+ const key = [project_config_1.PROJECT_CONFIG_FILENAME, '.gitignore']
1381
+ .map((name) => {
1382
+ try {
1383
+ return String(fs.statSync(path.join(this.rootDir, name)).mtimeMs);
1384
+ }
1385
+ catch {
1386
+ return '-';
1387
+ }
1388
+ })
1389
+ .join('|');
1390
+ if (this.scopedMatcher && this.scopedMatcher.key === key)
1391
+ return this.scopedMatcher.matcher;
1392
+ const matcher = buildScopeIgnore(this.rootDir);
1393
+ this.scopedMatcher = { key, matcher };
1394
+ return matcher;
1395
+ }
1330
1396
  /**
1331
1397
  * Build a filesystem-backed ResolutionContext sufficient for framework
1332
1398
  * detection. Graph-query methods (getNodesByName etc.) return empty because
@@ -1407,8 +1473,17 @@ class ExtractionOrchestrator {
1407
1473
  // null in the normal case, or a promise to await (at this safe,
1408
1474
  // between-transactions boundary) when the WAL has outrun the off-thread
1409
1475
  // checkpointer past its hard cap. See db/wal-valve.ts.
1410
- walBackpressure) {
1476
+ walBackpressure,
1477
+ // Fresh-DB store offload (perf): when set, per-file store bundles are
1478
+ // applied by a dedicated writer thread instead of the main thread. Only
1479
+ // passed for a COMPLETELY fresh database, where the main thread performs
1480
+ // no reads/writes during the parse loop, so one writer applying bundles
1481
+ // in file order preserves the #1015 determinism exactly.
1482
+ storeWriterOpts) {
1483
+ const tGrammar = Date.now();
1411
1484
  await (0, grammars_1.initGrammars)();
1485
+ if (process.env.CODEGRAPH_SYNTH_TIMINGS)
1486
+ console.error(`[phase-timing] grammar-init: ${Date.now() - tGrammar}ms`);
1412
1487
  const startTime = Date.now();
1413
1488
  const errors = [];
1414
1489
  let filesIndexed = 0;
@@ -1443,6 +1518,10 @@ class ExtractionOrchestrator {
1443
1518
  });
1444
1519
  if (process.env.CODEGRAPH_SYNTH_TIMINGS)
1445
1520
  console.error(`[phase-timing] scan: ${Date.now() - tScan}ms (${files.length} files)`);
1521
+ // A re-index over an existing DB skips unchanged-hash files at the store,
1522
+ // which would preserve wiped zero-node rows (#1541) — drop them first so
1523
+ // this run stores their files fresh. No-op on a fresh DB.
1524
+ this.healZeroNodeRows();
1446
1525
  // Detect frameworks once per indexAll run using the scanned file list.
1447
1526
  // Names are passed to each parse call so framework-specific extractors
1448
1527
  // (route nodes, middleware, etc.) run after the tree-sitter pass.
@@ -1491,8 +1570,15 @@ class ExtractionOrchestrator {
1491
1570
  let pool = null;
1492
1571
  if (useWorker) {
1493
1572
  // CODEGRAPH_PARSE_WORKERS: explicit worker count; 1 = the old single-worker
1494
- // behaviour (the conservative rollback). Unset → clamp(cores-1, 1, 8).
1495
- const poolSize = (0, parse_pool_1.resolveParsePoolSize)(process.env.CODEGRAPH_PARSE_WORKERS, os.cpus().length);
1573
+ // behaviour (the conservative rollback). Unset → clamp(cores-1, 1, 8),
1574
+ // with cores from availableParallelism — cpuset/affinity-honest, where
1575
+ // os.cpus() enumerates the host's CPUs and spawned 8 wasm workers (and
1576
+ // their grammar heaps) inside a 2-CPU container for zero extra
1577
+ // throughput (§7a.1). Floored so a 2-core box still gets 2 workers:
1578
+ // parse is worker-side CPU, and 1 worker measured 34% slower than the
1579
+ // old oversubscribed pool on the kernel-scale 2-cpuset envelope
1580
+ // (493s vs 369s) — main + store-worker don't fill the second core.
1581
+ const poolSize = (0, parse_pool_1.resolveParsePoolSize)(process.env.CODEGRAPH_PARSE_WORKERS, Math.max(3, os.availableParallelism()));
1496
1582
  // Read each needed grammar's WASM ONCE here and hand the bytes to every
1497
1583
  // worker, so spawns/respawns load grammars from memory instead of
1498
1584
  // re-reading them from disk (#1231: on an HDD, respawn re-reads amplify
@@ -1508,11 +1594,32 @@ class ExtractionOrchestrator {
1508
1594
  grammarBuffers,
1509
1595
  });
1510
1596
  log(`Parse worker pool: ${poolSize} worker(s)`);
1597
+ // Bulk index: every core will be needed — spawn the whole pool now so
1598
+ // worker boot overlaps the first read batches instead of trickling in
1599
+ // behind queue-pressure growth.
1600
+ pool.prewarm();
1511
1601
  }
1512
1602
  else {
1513
1603
  // In-process fallback: load grammars locally and parse on the main thread.
1514
1604
  await (0, grammars_1.loadGrammarsForLanguages)(neededLanguages);
1515
1605
  }
1606
+ // Dedicated store writer thread (fresh DB only — see the parameter doc).
1607
+ // Same availability rule as the parse pool: needs the compiled worker
1608
+ // (absent when running from source in tests → main-thread fallback).
1609
+ const storeWorkerPath = path.join(__dirname, 'store-worker.js');
1610
+ let storeWriter = null;
1611
+ if (storeWriterOpts &&
1612
+ process.env.CODEGRAPH_NO_STORE_WORKER !== '1' &&
1613
+ fs.existsSync(storeWorkerPath)) {
1614
+ // Deliberately NOT awaiting ready(): worker_threads delivers messages in
1615
+ // order, so bundles posted while the worker is still booting queue
1616
+ // behind 'open'. A boot failure surfaces at the first drain() — same
1617
+ // propagation point as a store error.
1618
+ storeWriter = new store_writer_1.StoreWriter(storeWorkerPath, storeWriterOpts.dbPath, storeWriterOpts.fastInit);
1619
+ log('Store writer thread active');
1620
+ }
1621
+ /** Queue-depth bound for un-acked bundles (bundles hold whole node/edge arrays). */
1622
+ const STORE_WRITER_WINDOW = 64;
1516
1623
  /**
1517
1624
  * Parse one file: on the pool when available (the promise REJECTS on a worker
1518
1625
  * crash/timeout — the caller records it and the retry pass re-attempts), or
@@ -1554,10 +1661,34 @@ class ExtractionOrchestrator {
1554
1661
  const bp = walBackpressure?.();
1555
1662
  if (bp)
1556
1663
  await bp;
1557
- // Store in database on main thread (SQLite is not thread-safe)
1558
- if (result.nodes.length > 0 || result.errors.length === 0) {
1559
- const language = (0, grammars_1.detectLanguage)(filePath, content, overrides);
1560
- await this.storeExtractionResult(filePath, content, language, stats, result, commitYield);
1664
+ // Kernel deferred-decode results carry table sizes in kernelCounts
1665
+ // (their object arrays are empty — decode happens at the store).
1666
+ const nodeCount = result.kernelCounts?.nodes ?? result.nodes.length;
1667
+ const edgeCount = result.kernelCounts?.edges ?? result.edges.length;
1668
+ // Store: on the writer thread when active (fresh DB — bundles applied
1669
+ // in the same file order this chain dispatches them), else on the main
1670
+ // thread (SQLite connections are per-thread).
1671
+ const language = (0, grammars_1.detectLanguage)(filePath, content, overrides);
1672
+ if (storeWriter) {
1673
+ if (result.kernelBuffers) {
1674
+ // Buffers go to the writer as-is; the worker decodes + finalizes.
1675
+ // The main thread's only per-file work stays O(1) + the content hash.
1676
+ storeWriter.send({
1677
+ kernel: true,
1678
+ filePath,
1679
+ language,
1680
+ buffers: result.kernelBuffers,
1681
+ file: this.buildFileRecord(filePath, content, language, stats, nodeCount, result.errors),
1682
+ });
1683
+ }
1684
+ else {
1685
+ storeWriter.send(this.buildFreshStoreBundle(filePath, content, language, stats, result));
1686
+ }
1687
+ await storeWriter.waitBelow(STORE_WRITER_WINDOW);
1688
+ }
1689
+ else {
1690
+ const materialized = (0, kernel_1.materializeKernelResult)(result, filePath, language);
1691
+ await this.storeExtractionResult(filePath, content, language, stats, materialized, commitYield);
1561
1692
  }
1562
1693
  if (result.errors.length > 0) {
1563
1694
  for (const err of result.errors) {
@@ -1566,10 +1697,10 @@ class ExtractionOrchestrator {
1566
1697
  }
1567
1698
  errors.push(...result.errors);
1568
1699
  }
1569
- if (result.nodes.length > 0) {
1700
+ if (nodeCount > 0) {
1570
1701
  filesIndexed++;
1571
- totalNodes += result.nodes.length;
1572
- totalEdges += result.edges.length;
1702
+ totalNodes += nodeCount;
1703
+ totalEdges += edgeCount;
1573
1704
  }
1574
1705
  else if (result.errors.some((e) => e.severity === 'error')) {
1575
1706
  filesErrored++;
@@ -1588,16 +1719,19 @@ class ExtractionOrchestrator {
1588
1719
  }
1589
1720
  onProgress?.({ phase: 'parsing', current: processed, total, currentFile: filePath });
1590
1721
  };
1591
- const recordParseFailure = (filePath, err) => {
1592
- processed++;
1593
- filesErrored++;
1594
- errors.push({
1595
- message: err instanceof Error ? err.message : String(err),
1596
- filePath,
1597
- severity: 'error',
1598
- code: 'parse_error',
1722
+ const recordParseFailure = async (filePath, content, stats, err) => {
1723
+ await storeResult(filePath, content, stats, {
1724
+ nodes: [],
1725
+ edges: [],
1726
+ unresolvedReferences: [],
1727
+ errors: [{
1728
+ message: err instanceof Error ? err.message : String(err),
1729
+ filePath,
1730
+ severity: 'error',
1731
+ code: 'parse_error',
1732
+ }],
1733
+ durationMs: 0,
1599
1734
  });
1600
- onProgress?.({ phase: 'parsing', current: processed, total });
1601
1735
  };
1602
1736
  // Commit buffered parses to the DB in file order, advancing the cursor over
1603
1737
  // contiguous completed results. Runs after each parse settles (and once more
@@ -1622,7 +1756,7 @@ class ExtractionOrchestrator {
1622
1756
  if (item.ok)
1623
1757
  await storeResult(item.filePath, item.content, item.stats, item.result);
1624
1758
  else
1625
- recordParseFailure(item.filePath, item.err);
1759
+ await recordParseFailure(item.filePath, item.content, item.stats, item.err);
1626
1760
  }
1627
1761
  }
1628
1762
  catch (err) {
@@ -1642,7 +1776,7 @@ class ExtractionOrchestrator {
1642
1776
  completed.set(seq, { ok: true, filePath, content, stats, result });
1643
1777
  }
1644
1778
  catch (parseErr) {
1645
- completed.set(seq, { ok: false, filePath, err: parseErr });
1779
+ completed.set(seq, { ok: false, filePath, content, stats, err: parseErr });
1646
1780
  }
1647
1781
  flushOrdered();
1648
1782
  })();
@@ -1663,6 +1797,7 @@ class ExtractionOrchestrator {
1663
1797
  await flushOrdered();
1664
1798
  }
1665
1799
  };
1800
+ const tParseLoop = Date.now();
1666
1801
  for (let i = 0; i < files.length; i += FILE_IO_BATCH_SIZE) {
1667
1802
  if (signal?.aborted) {
1668
1803
  aborted = true;
@@ -1713,15 +1848,18 @@ class ExtractionOrchestrator {
1713
1848
  // useful symbols. The single-file extractFile path already enforces
1714
1849
  // this; the bulk path used to silently skip the check.
1715
1850
  if (stats.size > MAX_FILE_SIZE) {
1716
- processed++;
1717
- filesSkipped++;
1718
- errors.push({
1719
- message: `File exceeds max size (${stats.size} > ${MAX_FILE_SIZE})`,
1720
- filePath,
1721
- severity: 'warning',
1722
- code: 'size_exceeded',
1851
+ await storeResult(filePath, content, stats, {
1852
+ nodes: [],
1853
+ edges: [],
1854
+ unresolvedReferences: [],
1855
+ errors: [{
1856
+ message: `File exceeds max size (${stats.size} > ${MAX_FILE_SIZE})`,
1857
+ filePath,
1858
+ severity: 'warning',
1859
+ code: 'size_exceeded',
1860
+ }],
1861
+ durationMs: 0,
1723
1862
  });
1724
- onProgress?.({ phase: 'parsing', current: processed, total });
1725
1863
  continue;
1726
1864
  }
1727
1865
  // Parse on the pool (main thread stays unblocked). Errors/timeouts are
@@ -1736,10 +1874,29 @@ class ExtractionOrchestrator {
1736
1874
  if (!aborted) {
1737
1875
  await Promise.all(inFlight);
1738
1876
  await flushOrdered();
1739
- if (flushError)
1877
+ if (flushError) {
1878
+ if (storeWriter)
1879
+ await storeWriter.close();
1740
1880
  throw flushError;
1881
+ }
1882
+ // All bundles are posted; wait for the writer to apply them, then close
1883
+ // its connection BEFORE any main-thread DB work below (retry pass,
1884
+ // resolution) so exactly one connection writes at a time.
1885
+ if (storeWriter) {
1886
+ try {
1887
+ await storeWriter.drain();
1888
+ }
1889
+ finally {
1890
+ await storeWriter.close();
1891
+ storeWriter = null;
1892
+ }
1893
+ }
1741
1894
  }
1895
+ if (process.env.CODEGRAPH_SYNTH_TIMINGS)
1896
+ console.error(`[phase-timing] parse-loop: ${Date.now() - tParseLoop}ms`);
1742
1897
  if (signal?.aborted || aborted) {
1898
+ if (storeWriter)
1899
+ await storeWriter.close();
1743
1900
  if (pool)
1744
1901
  await pool.destroy();
1745
1902
  return {
@@ -1803,8 +1960,15 @@ class ExtractionOrchestrator {
1803
1960
  stillFailing.push(errEntry);
1804
1961
  continue;
1805
1962
  }
1963
+ // The pool hands kernel results back as an undecoded buffer transport
1964
+ // (`nodes`/`edges` EMPTY, tables in kernelBuffers). The main loop
1965
+ // decodes or forwards to the store worker; this path stores directly,
1966
+ // so decode here — otherwise a kernel-language retry passes the gate
1967
+ // below via `errors.length === 0`, stores nothing, and the file is
1968
+ // permanently recorded as "(0 symbols)" with the error erased (#1541).
1969
+ const language = (0, grammars_1.detectLanguage)(filePath, content, overrides);
1970
+ result = (0, kernel_1.materializeKernelResult)(result, filePath, language);
1806
1971
  if (result.nodes.length > 0 || result.errors.length === 0) {
1807
- const language = (0, grammars_1.detectLanguage)(filePath, content, overrides);
1808
1972
  const stats = await fsp.stat(path.join(this.rootDir, filePath));
1809
1973
  await this.storeExtractionResult(filePath, content, language, stats, result, commitYield);
1810
1974
  const idx = errors.indexOf(errEntry);
@@ -1851,13 +2015,20 @@ class ExtractionOrchestrator {
1851
2015
  catch {
1852
2016
  continue;
1853
2017
  }
2018
+ // Same undecoded-transport hazard as the first retry pass (#1541).
2019
+ const language = (0, grammars_1.detectLanguage)(filePath, fullContent, overrides);
2020
+ result = (0, kernel_1.materializeKernelResult)(result, filePath, language);
1854
2021
  if (result.nodes.length > 0 || result.errors.length === 0) {
1855
- const language = (0, grammars_1.detectLanguage)(filePath, fullContent, overrides);
1856
2022
  const stats = await fsp.stat(path.join(this.rootDir, filePath));
1857
2023
  await this.storeExtractionResult(filePath, fullContent, language, stats, result, commitYield);
1858
- const idx = errors.indexOf(errEntry);
1859
- if (idx >= 0)
1860
- errors.splice(idx, 1);
2024
+ // Salvaged from comment-stripped source: keep a visible trace in
2025
+ // the summary instead of erasing the failure outright — the
2026
+ // stored result may be missing whatever the failing parse choked
2027
+ // on, and a silently "clean" file here is how an index quietly
2028
+ // disagrees with a later per-file sync of the same bytes (#1565).
2029
+ errEntry.severity = 'warning';
2030
+ errEntry.code = 'salvaged_stripped';
2031
+ errEntry.message = `Indexed from comment-stripped source after repeated parse failures (symbols may be incomplete until the file is re-indexed): ${errEntry.message}`;
1861
2032
  filesErrored--;
1862
2033
  filesIndexed++;
1863
2034
  totalNodes += result.nodes.length;
@@ -1984,9 +2155,10 @@ class ExtractionOrchestrator {
1984
2155
  durationMs: 0,
1985
2156
  };
1986
2157
  }
2158
+ const language = (0, grammars_1.detectLanguage)(relativePath, content, (0, project_config_1.loadExtensionOverrides)(this.rootDir));
1987
2159
  // Check file size
1988
2160
  if (stats.size > MAX_FILE_SIZE) {
1989
- return {
2161
+ const result = {
1990
2162
  nodes: [],
1991
2163
  edges: [],
1992
2164
  unresolvedReferences: [],
@@ -2000,9 +2172,10 @@ class ExtractionOrchestrator {
2000
2172
  ],
2001
2173
  durationMs: 0,
2002
2174
  };
2175
+ await this.storeExtractionResult(relativePath, content, language, stats, result, (0, cooperative_yield_1.createYielder)());
2176
+ return result;
2003
2177
  }
2004
2178
  // Detect language (honoring the project's codegraph.json extension overrides)
2005
- const language = (0, grammars_1.detectLanguage)(relativePath, content, (0, project_config_1.loadExtensionOverrides)(this.rootDir));
2006
2179
  if (!(0, grammars_1.isLanguageSupported)(language)) {
2007
2180
  return {
2008
2181
  nodes: [],
@@ -2018,15 +2191,43 @@ class ExtractionOrchestrator {
2018
2191
  const frameworkNames = this.ensureDetectedFrameworks();
2019
2192
  const result = (0, tree_sitter_1.extractFromSource)(relativePath, content, language, frameworkNames);
2020
2193
  // Store in database
2021
- if (result.nodes.length > 0 || result.errors.length === 0) {
2022
- await this.storeExtractionResult(relativePath, content, language, stats, result, (0, cooperative_yield_1.createYielder)());
2023
- }
2194
+ await this.storeExtractionResult(relativePath, content, language, stats, result, (0, cooperative_yield_1.createYielder)());
2024
2195
  return result;
2025
2196
  }
2026
2197
  /**
2027
2198
  * Store extraction result in database
2028
2199
  */
2200
+ /**
2201
+ * Delete file rows recorded with ZERO nodes so their files re-index.
2202
+ *
2203
+ * No extraction path stores an empty, error-free result for a
2204
+ * symbol-bearing language — even an empty file keeps its file node — so a
2205
+ * zero-node row is a wiped one (#1541: an interrupted parse's retry stored
2206
+ * an undecoded kernel transport). The wiped row's content hash matches the
2207
+ * on-disk bytes, so every hash-based reconcile skips the file forever;
2208
+ * deleting the row lets the normal add path repair it. File-level-only
2209
+ * languages (yaml, twig, properties) are left alone. Deleting a zero-node
2210
+ * row cascades nothing: it has no nodes, so no edges or refs either.
2211
+ */
2212
+ healZeroNodeRows() {
2213
+ for (const f of this.queries.getAllFiles()) {
2214
+ // A zero-node row WITH recorded errors is a deliberate skip marker
2215
+ // (#1557: oversized / repeatedly-unparseable files are persisted with
2216
+ // their reason so syncs stop retrying them) — leave those alone. The
2217
+ // #1541 wipe rows are the error-FREE zero-node rows.
2218
+ if (f.nodeCount === 0 &&
2219
+ !(0, grammars_1.isFileLevelOnlyLanguage)(f.language) &&
2220
+ (f.errors === undefined || f.errors.length === 0)) {
2221
+ this.queries.deleteFile(f.path);
2222
+ }
2223
+ }
2224
+ }
2029
2225
  async storeExtractionResult(filePath, content, language, stats, result, onYield) {
2226
+ // A kernel result can arrive as an undecoded buffer transport (empty
2227
+ // node/edge arrays, tables riding in kernelBuffers). Decode it before
2228
+ // storing — persisting the transport as-is records the file as having no
2229
+ // symbols at all (#1541). No-op for already-decoded results.
2230
+ result = (0, kernel_1.materializeKernelResult)(result, filePath, language);
2030
2231
  // Bulk inserts run in bounded sub-transactions with a yield between, so a
2031
2232
  // giant generated file (tens of thousands of symbols) can't block the
2032
2233
  // event loop — and the #850 watchdog heartbeat — for the whole store.
@@ -2035,11 +2236,24 @@ class ExtractionOrchestrator {
2035
2236
  // is unchanged: a partially-stored file has no record and re-indexes.
2036
2237
  const STORE_CHUNK = 2000;
2037
2238
  const contentHash = hashContent(content);
2038
- // Check if file already exists and hasn't changed
2239
+ // Check if file already exists and hasn't changed. A skip/failure MARKER
2240
+ // row (zero nodes + recorded errors, #1557) never blocks a store carrying
2241
+ // real content: markers are written BEFORE the retry pass under the same
2242
+ // content hash, so treating them as "no changes" would silently discard a
2243
+ // successful retry's symbols — a permanent empty file presented as
2244
+ // recovered (the #1541 wipe, reintroduced through the marker path).
2039
2245
  const existingFile = this.queries.getFileByPath(filePath);
2040
2246
  if (existingFile && existingFile.contentHash === contentHash) {
2041
- return; // No changes
2247
+ const existingIsMarker = existingFile.nodeCount === 0 && (existingFile.errors?.length ?? 0) > 0;
2248
+ const incomingHasContent = result.nodes.length > 0;
2249
+ if (!existingIsMarker || !incomingHasContent) {
2250
+ return; // No changes
2251
+ }
2042
2252
  }
2253
+ // Re-decided on every re-index of a changed file, so a banner added (or
2254
+ // removed) by an edit is reflected on the next sync (#1500). Computed after
2255
+ // the unchanged-file early return so untouched files pay nothing.
2256
+ const generated = (0, generated_detection_1.detectGeneratedFile)(filePath, content);
2043
2257
  // Snapshot incoming cross-file edges BEFORE deleting this file's nodes.
2044
2258
  // `deleteFile` cascades to delete every edge whose source OR target is a
2045
2259
  // node in this file (edges.FK ... ON DELETE CASCADE). Edges whose SOURCE is
@@ -2067,15 +2281,53 @@ class ExtractionOrchestrator {
2067
2281
  // This prevents FK violations when edges reference nodes that would
2068
2282
  // be silently skipped by insertNode() (see issue #42).
2069
2283
  const validNodes = result.nodes.filter((n) => n.id && n.kind && n.name && n.filePath && n.language);
2284
+ const insertedIds = new Set(validNodes.map((n) => n.id));
2285
+ const validEdges = result.edges.filter((e) => insertedIds.has(e.source) && insertedIds.has(e.target));
2286
+ const validRefs = result.unresolvedReferences
2287
+ .filter((ref) => insertedIds.has(ref.fromNodeId))
2288
+ .map((ref) => ({
2289
+ ...ref,
2290
+ filePath: ref.filePath ?? filePath,
2291
+ language: ref.language ?? language,
2292
+ }));
2293
+ // Fast path for the common case (everything fits one chunk): the whole
2294
+ // file — nodes, edges, refs, file record — lands in ONE transaction with
2295
+ // no event-loop yields in between. Giant generated files keep the chunked
2296
+ // + yielding path below so the #850 watchdog heartbeat stays serviced.
2297
+ const fitsOneChunk = validNodes.length <= STORE_CHUNK &&
2298
+ validEdges.length <= STORE_CHUNK &&
2299
+ validRefs.length <= STORE_CHUNK;
2300
+ if (fitsOneChunk) {
2301
+ // Snapshot/re-resolution of cross-file incoming edges (below) still runs
2302
+ // for the sync path; on a fresh bulk index crossFileIncomingEdges is [].
2303
+ this.queries.storeFileBundle({
2304
+ nodes: validNodes,
2305
+ edges: validEdges,
2306
+ refs: validRefs,
2307
+ file: {
2308
+ path: filePath,
2309
+ contentHash,
2310
+ language,
2311
+ size: stats.size,
2312
+ modifiedAt: stats.mtimeMs,
2313
+ indexedAt: Date.now(),
2314
+ nodeCount: result.nodes.length,
2315
+ errors: result.errors.length > 0 ? result.errors : undefined,
2316
+ generated,
2317
+ },
2318
+ });
2319
+ if (crossFileIncomingEdges.length > 0) {
2320
+ this.reattachCrossFileEdges(crossFileIncomingEdges, validNodes);
2321
+ }
2322
+ return;
2323
+ }
2070
2324
  // Insert nodes (chunked — see STORE_CHUNK above)
2071
2325
  for (let i = 0; i < validNodes.length; i += STORE_CHUNK) {
2072
2326
  this.queries.insertNodes(validNodes.slice(i, i + STORE_CHUNK));
2073
2327
  await onYield?.();
2074
2328
  }
2075
2329
  // Filter edges to only reference nodes that were actually inserted
2076
- if (result.edges.length > 0) {
2077
- const insertedIds = new Set(validNodes.map((n) => n.id));
2078
- const validEdges = result.edges.filter((e) => insertedIds.has(e.source) && insertedIds.has(e.target));
2330
+ if (validEdges.length > 0) {
2079
2331
  for (let i = 0; i < validEdges.length; i += STORE_CHUNK) {
2080
2332
  this.queries.insertEdges(validEdges.slice(i, i + STORE_CHUNK));
2081
2333
  await onYield?.();
@@ -2100,44 +2352,12 @@ class ExtractionOrchestrator {
2100
2352
  // a ref from the target's plain name would strip receiver/qualifier
2101
2353
  // context and risk a rebind a full re-index would never make.
2102
2354
  if (crossFileIncomingEdges.length > 0) {
2103
- const newNodesByKindName = new Map();
2104
- for (const n of validNodes) {
2105
- newNodesByKindName.set(`${n.kind}\0${n.name}`, n.id);
2106
- }
2107
- const reinserted = [];
2108
- const resurrected = [];
2109
- for (const e of crossFileIncomingEdges) {
2110
- const newTargetId = newNodesByKindName.get(`${e.targetKind}\0${e.targetName}`);
2111
- if (newTargetId) {
2112
- reinserted.push({ source: e.source, target: newTargetId, kind: e.kind, metadata: e.metadata, line: e.line, column: e.column, provenance: e.provenance });
2113
- }
2114
- else {
2115
- const ref = resurrectRefFromDroppedEdge(e);
2116
- if (ref)
2117
- resurrected.push(ref);
2118
- }
2119
- }
2120
- if (reinserted.length > 0) {
2121
- this.queries.insertEdges(reinserted);
2122
- }
2123
- if (resurrected.length > 0) {
2124
- this.queries.insertUnresolvedRefsBatch(resurrected);
2125
- }
2355
+ this.reattachCrossFileEdges(crossFileIncomingEdges, validNodes);
2126
2356
  }
2127
2357
  // Insert unresolved references in batch with denormalized filePath/language
2128
- if (result.unresolvedReferences.length > 0) {
2129
- const insertedIds = new Set(validNodes.map((n) => n.id));
2130
- const refsWithContext = result.unresolvedReferences
2131
- .filter((ref) => insertedIds.has(ref.fromNodeId))
2132
- .map((ref) => ({
2133
- ...ref,
2134
- filePath: ref.filePath ?? filePath,
2135
- language: ref.language ?? language,
2136
- }));
2137
- for (let i = 0; i < refsWithContext.length; i += STORE_CHUNK) {
2138
- this.queries.insertUnresolvedRefsBatch(refsWithContext.slice(i, i + STORE_CHUNK));
2139
- await onYield?.();
2140
- }
2358
+ for (let i = 0; i < validRefs.length; i += STORE_CHUNK) {
2359
+ this.queries.insertUnresolvedRefsBatch(validRefs.slice(i, i + STORE_CHUNK));
2360
+ await onYield?.();
2141
2361
  }
2142
2362
  // Insert file record
2143
2363
  const fileRecord = {
@@ -2149,9 +2369,126 @@ class ExtractionOrchestrator {
2149
2369
  indexedAt: Date.now(),
2150
2370
  nodeCount: result.nodes.length,
2151
2371
  errors: result.errors.length > 0 ? result.errors : undefined,
2372
+ generated,
2152
2373
  };
2153
2374
  this.queries.upsertFile(fileRecord);
2154
2375
  }
2376
+ /**
2377
+ * Build one file's store bundle for the FRESH-DB path: no existing-file
2378
+ * check, no cross-file edge snapshot (both are re-index concerns — a fresh
2379
+ * database has neither). Filters mirror storeExtractionResult exactly.
2380
+ */
2381
+ /** The FileRecord for a fresh-index store (nodeCount is the PRE-filter count). */
2382
+ buildFileRecord(filePath, content, language, stats, nodeCount, resultErrors) {
2383
+ return {
2384
+ path: filePath,
2385
+ contentHash: hashContent(content),
2386
+ language,
2387
+ size: stats.size,
2388
+ modifiedAt: stats.mtimeMs,
2389
+ indexedAt: Date.now(),
2390
+ nodeCount,
2391
+ errors: resultErrors.length > 0 ? resultErrors : undefined,
2392
+ // Decided here, once, while the content is already in memory — never at
2393
+ // query time (#1500). The header scan short-circuits on a single
2394
+ // substring test for ~every hand-written file.
2395
+ generated: (0, generated_detection_1.detectGeneratedFile)(filePath, content),
2396
+ };
2397
+ }
2398
+ buildFreshStoreBundle(filePath, content, language, stats, result) {
2399
+ return (0, store_writer_1.finalizeStoreBundle)(result, filePath, language, this.buildFileRecord(filePath, content, language, stats, result.nodes.length, result.errors));
2400
+ }
2401
+ /**
2402
+ * Re-attach cross-file incoming edges snapshotted before a re-index delete
2403
+ * (#899): re-resolve each edge's target to the re-indexed node's new id by
2404
+ * (kind, name); targets that vanished are resurrected as their original
2405
+ * unresolved ref (#1240's removal-side counterpart) when the edge carries
2406
+ * its refName stamp.
2407
+ */
2408
+ reattachCrossFileEdges(crossFileIncomingEdges, validNodes) {
2409
+ const newNodesByKindName = new Map();
2410
+ for (const n of validNodes) {
2411
+ newNodesByKindName.set(`${n.kind}\0${n.name}`, n.id);
2412
+ }
2413
+ const reinserted = [];
2414
+ const resurrected = [];
2415
+ for (const e of crossFileIncomingEdges) {
2416
+ const newTargetId = newNodesByKindName.get(`${e.targetKind}\0${e.targetName}`);
2417
+ if (newTargetId) {
2418
+ reinserted.push({ source: e.source, target: newTargetId, kind: e.kind, metadata: e.metadata, line: e.line, column: e.column, provenance: e.provenance });
2419
+ }
2420
+ else {
2421
+ const ref = resurrectRefFromDroppedEdge(e);
2422
+ if (ref)
2423
+ resurrected.push(ref);
2424
+ }
2425
+ }
2426
+ if (reinserted.length > 0) {
2427
+ this.queries.insertEdges(reinserted);
2428
+ }
2429
+ if (resurrected.length > 0) {
2430
+ this.queries.insertUnresolvedRefsBatch(resurrected);
2431
+ }
2432
+ }
2433
+ /**
2434
+ * Re-open, for re-resolution, every resolution edge whose answer this sync
2435
+ * may have changed — the fix for index drift (CG-33).
2436
+ *
2437
+ * Incremental sync re-resolves only the references IN the changed files, but
2438
+ * resolution's answer is a function of the WHOLE graph: a reference binds to
2439
+ * one of the same-named definitions project-wide, so adding or removing a
2440
+ * definition of `pct` can change which `pct` every other file's `pct(...)`
2441
+ * should bind to. Those other files are never revisited, and their references
2442
+ * resolved successfully once and were deleted from `unresolved_refs`, so
2443
+ * nothing existed to revisit them with — the index kept an answer that was
2444
+ * correct against an older graph. Measured on codegraph's own long-lived
2445
+ * index: 4.3% of distinct edges differed from a clean rebuild, in BOTH
2446
+ * directions, overwhelmingly `calls`. See docs/benchmarks/index-drift-cg33.md.
2447
+ *
2448
+ * This deletes each affected edge and re-inserts it as the reference that
2449
+ * created it (the refName/refKind stamp), status='pending', for the sync's
2450
+ * resolution sweep to bind against the post-sync graph — the same input a
2451
+ * full rebuild resolves from, which is what makes the two converge.
2452
+ *
2453
+ * Deliberately conservative in three ways, because a wrong deletion is a
2454
+ * permanent edge loss while a missed rebind is only residual drift:
2455
+ * - an edge with no refName stamp (synthesized, or built by an engine older
2456
+ * than the stamp) is left ALONE rather than reconstructed from the target's
2457
+ * plain name, same rule as `resurrectRefFromDroppedEdge`;
2458
+ * - edges whose source is in a file this sync already re-extracted are
2459
+ * skipped — their references were re-resolved from scratch moments ago;
2460
+ * - very common names are skipped by the per-name ceiling in
2461
+ * `getResolutionEdgesByTargetName`.
2462
+ *
2463
+ * Returns the number of references resurrected.
2464
+ */
2465
+ resurrectStaleResolutionEdges(definitionDelta, changedFilePaths) {
2466
+ if (definitionDelta.length === 0)
2467
+ return 0;
2468
+ const alreadyFresh = new Set(changedFilePaths);
2469
+ const candidates = this.queries.getResolutionEdgesByTargetName(definitionDelta);
2470
+ const edgeIds = [];
2471
+ const refs = [];
2472
+ for (const e of candidates) {
2473
+ if (alreadyFresh.has(e.sourceFilePath))
2474
+ continue;
2475
+ const ref = resurrectRefFromDroppedEdge(e);
2476
+ if (!ref)
2477
+ continue; // no stamp — never delete what we cannot restore
2478
+ edgeIds.push(e.edgeId);
2479
+ refs.push(ref);
2480
+ }
2481
+ if (refs.length === 0)
2482
+ return 0;
2483
+ // Delete first. The sweep re-inserts whichever edge resolution now picks,
2484
+ // and `insertEdges` is INSERT OR IGNORE against idx_edges_identity — so a
2485
+ // rebind to the same target is a clean no-op, but leaving the old row in
2486
+ // place for a rebind ELSEWHERE would keep both, turning drift into
2487
+ // duplication.
2488
+ this.queries.deleteEdgesByIds(edgeIds);
2489
+ this.queries.insertUnresolvedRefsBatch(refs);
2490
+ return refs.length;
2491
+ }
2155
2492
  /**
2156
2493
  * Sync the index with the current file state.
2157
2494
  *
@@ -2160,7 +2497,17 @@ class ExtractionOrchestrator {
2160
2497
  * changes. This works in non-git projects and catches committed changes from
2161
2498
  * `git pull`/`checkout`/`merge`/`rebase` that `git status` cannot see.
2162
2499
  */
2163
- async sync(onProgress) {
2500
+ async sync(onProgress,
2501
+ /**
2502
+ * Watcher fast path: the exact project-relative paths the OS reported as
2503
+ * changed. When provided, reconciliation runs over ONLY these paths —
2504
+ * per-path logic identical to the full walk (stat pre-filter, hash
2505
+ * confirm, the #1240 removal/resurrection flow) — skipping the O(repo)
2506
+ * scan and tracked-load. Callers must pass undefined whenever the change
2507
+ * set is not exactly known (directory removals, event overflow): the full
2508
+ * scan-diff remains the ground truth those cases need (#1285).
2509
+ */
2510
+ scopedPaths) {
2164
2511
  await (0, grammars_1.initGrammars)(); // Initialize WASM runtime (grammars loaded lazily below)
2165
2512
  const startTime = Date.now();
2166
2513
  let filesChecked = 0;
@@ -2169,6 +2516,10 @@ class ExtractionOrchestrator {
2169
2516
  let filesRemoved = 0;
2170
2517
  let nodesUpdated = 0;
2171
2518
  const changedFilePaths = [];
2519
+ // `file\0name` definition pairs for the files this sync touches, sampled
2520
+ // BEFORE their nodes are replaced/deleted. Compared against the post-store
2521
+ // pairs below to derive `definitionDelta` (CG-33).
2522
+ const pairsBefore = new Set();
2172
2523
  onProgress?.({
2173
2524
  phase: 'scanning',
2174
2525
  current: 0,
@@ -2185,15 +2536,53 @@ class ExtractionOrchestrator {
2185
2536
  // changes from `git pull`/`checkout`/`merge`/`rebase` — which `git status`
2186
2537
  // cannot see, because the working tree is clean afterward.
2187
2538
  const tSyncScan = Date.now();
2188
- const currentFiles = await scanDirectoryAsync(this.rootDir);
2189
- if (process.env.CODEGRAPH_SYNTH_TIMINGS)
2190
- console.error(`[phase-timing] sync-scan: ${Date.now() - tSyncScan}ms (${currentFiles.length} files)`);
2191
- filesChecked = currentFiles.length;
2539
+ let currentFiles;
2540
+ let trackedFiles;
2541
+ if (scopedPaths && scopedPaths.length > 0) {
2542
+ // Scoped reconcile: stat only the reported paths. filesChecked counts
2543
+ // the PATHS examined (not the files found) — it must stay non-zero even
2544
+ // when every scoped path was a deletion, because CodeGraph.watch()
2545
+ // reads `filesChecked === 0 && durationMs === 0` as the
2546
+ // lock-unavailable signature (#449).
2547
+ const unique = [...new Set(scopedPaths)];
2548
+ // A scoped path is "present" only if it exists AND is in scope — the
2549
+ // same two gates the full walk applies (source extension, scope
2550
+ // matcher). Without the scope gate a caller's stale view of scope
2551
+ // leaked straight into the index: the watcher re-parsed a file the
2552
+ // user had just excluded in `codegraph.json` while `codegraph sync`
2553
+ // removed it (#1590). Out-of-scope paths fall out of `currentFiles`,
2554
+ // so a tracked one takes the removal branch below, exactly as a full
2555
+ // sync would treat it. (`include`-forced paths pass: ScopeIgnore
2556
+ // applies the include precedence itself.)
2557
+ const scope = this.scopedSyncMatcher();
2558
+ const overrides = (0, project_config_1.loadExtensionOverrides)(this.rootDir);
2559
+ currentFiles = unique.filter((p) => (0, grammars_1.isSourceFile)(p, overrides) &&
2560
+ !scope.ignores(p) &&
2561
+ fs.existsSync(path.join(this.rootDir, p)));
2562
+ trackedFiles = [];
2563
+ for (const p of unique) {
2564
+ const rec = this.queries.getFileByPath(p);
2565
+ if (rec)
2566
+ trackedFiles.push(rec);
2567
+ }
2568
+ filesChecked = unique.length;
2569
+ if (process.env.CODEGRAPH_SYNTH_TIMINGS)
2570
+ console.error(`[phase-timing] sync-scoped: ${Date.now() - tSyncScan}ms (${unique.length} paths, ${trackedFiles.length} tracked)`);
2571
+ }
2572
+ else {
2573
+ currentFiles = await scanDirectoryAsync(this.rootDir);
2574
+ if (process.env.CODEGRAPH_SYNTH_TIMINGS)
2575
+ console.error(`[phase-timing] sync-scan: ${Date.now() - tSyncScan}ms (${currentFiles.length} files)`);
2576
+ filesChecked = currentFiles.length;
2577
+ // Full reconcile only (scoped syncs must not touch rows outside their
2578
+ // scope): drop zero-node rows so the wiped files re-index as adds below.
2579
+ this.healZeroNodeRows();
2580
+ const tTracked = Date.now();
2581
+ trackedFiles = this.queries.getAllFiles();
2582
+ if (process.env.CODEGRAPH_SYNTH_TIMINGS)
2583
+ console.error(`[phase-timing] sync-tracked-load: ${Date.now() - tTracked}ms (${trackedFiles.length} tracked)`);
2584
+ }
2192
2585
  const currentSet = new Set(currentFiles);
2193
- const tTracked = Date.now();
2194
- const trackedFiles = this.queries.getAllFiles();
2195
- if (process.env.CODEGRAPH_SYNTH_TIMINGS)
2196
- console.error(`[phase-timing] sync-tracked-load: ${Date.now() - tTracked}ms (${trackedFiles.length} tracked)`);
2197
2586
  const trackedMap = new Map();
2198
2587
  for (const f of trackedFiles) {
2199
2588
  trackedMap.set(f.path, f);
@@ -2213,6 +2602,10 @@ class ExtractionOrchestrator {
2213
2602
  // failed until the symbol reappears somewhere. (A deleted file whose
2214
2603
  // CALLERS are also being deleted is fine: their nodes cascade later
2215
2604
  // in this loop and take the resurrected rows with them.)
2605
+ // Every name this file defined is about to stop existing here, which
2606
+ // narrows the candidate set for that name repo-wide (CG-33).
2607
+ for (const pair of this.queries.getNodeNamePairsByFiles([tracked.path]))
2608
+ pairsBefore.add(pair);
2216
2609
  const incoming = this.queries.getCrossFileIncomingEdgesWithTarget(tracked.path);
2217
2610
  if (incoming.length > 0) {
2218
2611
  const resurrected = incoming
@@ -2277,6 +2670,14 @@ class ExtractionOrchestrator {
2277
2670
  filesModified++;
2278
2671
  }
2279
2672
  }
2673
+ // Sampled here — after the add/modify classification, before any file is
2674
+ // re-extracted — because `storeExtractionResult` deletes a file's nodes
2675
+ // before inserting the new ones, so this is the last point the pre-edit
2676
+ // definition set is readable (CG-33).
2677
+ if (filesToIndex.length > 0) {
2678
+ for (const pair of this.queries.getNodeNamePairsByFiles(filesToIndex))
2679
+ pairsBefore.add(pair);
2680
+ }
2280
2681
  // Load only grammars needed for changed files
2281
2682
  if (filesToIndex.length > 0) {
2282
2683
  const overrides = (0, project_config_1.loadExtensionOverrides)(this.rootDir);
@@ -2300,6 +2701,28 @@ class ExtractionOrchestrator {
2300
2701
  const result = await this.indexFile(filePath);
2301
2702
  nodesUpdated += result.nodes.length;
2302
2703
  }
2704
+ // Names whose definition set this sync changed: a `file\0name` pair present
2705
+ // before but not after (removed/renamed away) or after but not before
2706
+ // (added). A pair on both sides is untouched as far as resolution's
2707
+ // candidate set is concerned — only its node id moved, which
2708
+ // reattachCrossFileEdges already follows — so an edit that only changes
2709
+ // bodies yields an empty delta and no downstream rebind work (CG-33).
2710
+ //
2711
+ // Compared per FILE, not as one name set over the whole batch: a commit
2712
+ // that adds `collect` to a new file while an unrelated changed file already
2713
+ // defined `collect` must still flag the name, and a bare name set cancels
2714
+ // exactly that case out. That miss left the largest residual class in the
2715
+ // first measurement of this fix.
2716
+ const pairsAfter = this.queries.getNodeNamePairsByFiles(filesToIndex);
2717
+ const deltaNames = new Set();
2718
+ const nameOf = (pair) => pair.slice(pair.indexOf('\0') + 1);
2719
+ for (const pair of pairsBefore)
2720
+ if (!pairsAfter.has(pair))
2721
+ deltaNames.add(nameOf(pair));
2722
+ for (const pair of pairsAfter)
2723
+ if (!pairsBefore.has(pair))
2724
+ deltaNames.add(nameOf(pair));
2725
+ const definitionDelta = [...deltaNames];
2303
2726
  return {
2304
2727
  filesChecked,
2305
2728
  filesAdded,
@@ -2308,6 +2731,7 @@ class ExtractionOrchestrator {
2308
2731
  nodesUpdated,
2309
2732
  durationMs: Date.now() - startTime,
2310
2733
  changedFilePaths: changedFilePaths.length > 0 ? changedFilePaths : undefined,
2734
+ definitionDelta: definitionDelta.length > 0 ? definitionDelta : undefined,
2311
2735
  };
2312
2736
  }
2313
2737
  /**