@pratik7368patil/anchor-core 0.1.28 → 0.1.30

package/dist/index.d.ts

@@ -781,6 +781,18 @@ type CodeIndexProgress = {
     repo: string;
     chunks: number;
     patterns: number;
+} | {
+    stage: "deleting_code_fts";
+    repo: string;
+    current: number;
+    total: number;
+    chunks: number;
+} | {
+    stage: "deleting_architecture_fts";
+    repo: string;
+    current: number;
+    total: number;
+    patterns: number;
 } | {
     stage: "writing_code_files";
     repo: string;

package/dist/index.js

@@ -707,6 +707,21 @@ CREATE INDEX IF NOT EXISTS idx_org_consumers_provider ON org_api_consumers(org,
 CREATE INDEX IF NOT EXISTS idx_org_consumers_consumer ON org_api_consumers(org, consumer_repo);
 CREATE INDEX IF NOT EXISTS idx_org_anomalies_org ON org_anomaly_events(org, severity);
 CREATE INDEX IF NOT EXISTS idx_org_graph_state_status ON org_graph_state(org, last_status);
+
+-- Foreign-key indexes backing per-repo / per-PR bulk deletes and re-index scans.
+-- Without these, replaceCodeIndex/deleteExistingPrData full-scan each table per repo/PR.
+CREATE INDEX IF NOT EXISTS idx_code_chunks_repo ON code_chunks(repo_id);
+CREATE INDEX IF NOT EXISTS idx_code_files_repo ON code_files(repo_id);
+CREATE INDEX IF NOT EXISTS idx_code_imports_repo ON code_imports(repo_id);
+CREATE INDEX IF NOT EXISTS idx_test_files_repo ON test_files(repo_id);
+CREATE INDEX IF NOT EXISTS idx_test_links_repo ON test_links(repo_id);
+CREATE INDEX IF NOT EXISTS idx_architecture_components_repo ON architecture_components(repo_id);
+CREATE INDEX IF NOT EXISTS idx_architecture_patterns_repo ON architecture_patterns(repo_id);
+CREATE INDEX IF NOT EXISTS idx_architecture_map_edges_repo ON architecture_map_edges(repo_id);
+CREATE INDEX IF NOT EXISTS idx_wisdom_units_repo ON wisdom_units(repo_id);
+CREATE INDEX IF NOT EXISTS idx_regression_events_repo ON regression_events(repo_id);
+CREATE INDEX IF NOT EXISTS idx_pr_files_pr ON pr_files(pr_id);
+CREATE INDEX IF NOT EXISTS idx_pr_comments_pr ON pr_comments(pr_id);
 `;
 
 // src/rules/team-rules.ts
@@ -1695,9 +1710,28 @@ function calculateCoverage(input) {
 
 // src/db/database.ts
 var CODE_WRITE_PROGRESS_INTERVAL = 150;
+var FTS_DELETE_BATCH_SIZE = 500;
 function shouldEmitCodeWriteProgress(current, total) {
   return current === 0 || current === 1 || current === total || current % CODE_WRITE_PROGRESS_INTERVAL === 0;
 }
+function shouldEmitFtsDeleteProgress(current, total) {
+  return current === 0 || current === 1 || current === total || current % FTS_DELETE_BATCH_SIZE === 0;
+}
+function deleteFtsRowsByRowId(db, ftsTable, rowIds, onProgress) {
+  if (rowIds.length === 0) {
+    onProgress?.(0, 0);
+    return;
+  }
+  const deleteRow = db.prepare(`DELETE FROM ${ftsTable} WHERE rowid = ?`);
+  onProgress?.(0, rowIds.length);
+  for (const [index, rowId] of rowIds.entries()) {
+    deleteRow.run(rowId);
+    const current = index + 1;
+    if (shouldEmitFtsDeleteProgress(current, rowIds.length)) {
+      onProgress?.(current, rowIds.length);
+    }
+  }
+}
 function defaultDatabasePath(cwd) {
   return path4.join(cwd, ".anchor", "index.sqlite");
 }
@@ -1707,14 +1741,22 @@ function openAnchorDatabase(cwd, databasePath = defaultDatabasePath(cwd)) {
   db.pragma("busy_timeout = 5000");
   db.pragma("journal_mode = WAL");
   db.pragma("foreign_keys = ON");
+  applyPerformancePragmas(db);
   return db;
 }
 function openAnchorDatabaseReadOnly(databasePath) {
   const db = new Database(databasePath, { readonly: true, fileMustExist: true });
   db.pragma("busy_timeout = 5000");
   db.pragma("foreign_keys = ON");
+  applyPerformancePragmas(db);
   return db;
 }
+function applyPerformancePragmas(db) {
+  db.pragma("synchronous = NORMAL");
+  db.pragma("cache_size = -65536");
+  db.pragma("mmap_size = 268435456");
+  db.pragma("temp_store = MEMORY");
+}
 function initializeSchema(db) {
   db.exec(SCHEMA_SQL);
   ensureColumn(db, "sync_state", "history_coverage", "TEXT");
@@ -1882,9 +1924,12 @@ function clearGraphQLFetchCheckpoint(db, repo, scope) {
   ).run((/* @__PURE__ */ new Date()).toISOString(), repo);
 }
 function deleteExistingPrData(db, prId) {
-  db.prepare(
-    "DELETE FROM wisdom_units_fts WHERE unitId IN (SELECT id FROM wisdom_units WHERE pr_id = ?)"
-  ).run(prId);
+  const wisdomRowIds = db.prepare("SELECT rowid FROM wisdom_units WHERE pr_id = ?").all(prId);
+  deleteFtsRowsByRowId(
+    db,
+    "wisdom_units_fts",
+    wisdomRowIds.map((row) => row.rowid)
+  );
   db.prepare("DELETE FROM regression_events WHERE pr_id = ?").run(prId);
   db.prepare("DELETE FROM wisdom_units WHERE pr_id = ?").run(prId);
   db.prepare("DELETE FROM pr_comments WHERE pr_id = ?").run(prId);
@@ -1996,11 +2041,11 @@ function upsertPullRequest(db, pr, wisdomUnits, regressionEvents = []) {
     );
     const insertFts = db.prepare(
       `INSERT INTO wisdom_units_fts
-       (unitId, sanitizedText, filePaths, symbols, prTitle, prBody, category)
-       VALUES (?, ?, ?, ?, ?, ?, ?)`
+       (rowid, unitId, sanitizedText, filePaths, symbols, prTitle, prBody, category)
+       VALUES (?, ?, ?, ?, ?, ?, ?, ?)`
     );
     for (const unit of wisdomUnits) {
-      insertWisdom.run(
+      const wisdomInsert = insertWisdom.run(
         unit.id,
         repoId,
         prRow.id,
@@ -2019,6 +2064,7 @@ function upsertPullRequest(db, pr, wisdomUnits, regressionEvents = []) {
         unit.confidence
       );
       insertFts.run(
+        Number(wisdomInsert.lastInsertRowid),
         unit.id,
         unit.sanitizedText,
         unit.filePaths.join(" "),
@@ -2075,22 +2121,31 @@ function replaceCodeIndex(db, repo, codeFiles, codeChunks, skippedFiles, cwd, ar
   });
   options.onProgress?.({ stage: "writing_code_index", repo, phase: "Writing code index" });
   const transaction = db.transaction(() => {
-    const existingChunkCount = db.prepare("SELECT COUNT(*) AS count FROM code_chunks WHERE repo_id = ?").get(repoId).count;
-    const existingPatternCount = db.prepare("SELECT COUNT(*) AS count FROM architecture_patterns WHERE repo_id = ?").get(repoId).count;
+    const existingChunkRowIds = db.prepare("SELECT rowid FROM code_chunks WHERE repo_id = ?").all(repoId);
+    const existingPatternRowIds = db.prepare("SELECT rowid FROM architecture_patterns WHERE repo_id = ?").all(repoId);
     options.onProgress?.({
       stage: "deleting_existing_code_index",
       repo,
-      chunks: existingChunkCount,
-      patterns: existingPatternCount
+      chunks: existingChunkRowIds.length,
+      patterns: existingPatternRowIds.length
     });
-    db.prepare(
-      "DELETE FROM code_chunks_fts WHERE chunkId IN (SELECT id FROM code_chunks WHERE repo_id = ?)"
-    ).run(repoId);
+    deleteFtsRowsByRowId(
+      db,
+      "code_chunks_fts",
+      existingChunkRowIds.map((row) => row.rowid),
+      (current, total) => options.onProgress?.({
+        stage: "deleting_code_fts",
+        repo,
+        current,
+        total,
+        chunks: existingChunkRowIds.length
+      })
+    );
     db.prepare("DELETE FROM code_chunks WHERE repo_id = ?").run(repoId);
     db.prepare("DELETE FROM code_files WHERE repo_id = ?").run(repoId);
     db.prepare("DELETE FROM test_links WHERE repo_id = ? AND reason != 'PR co-change'").run(repoId);
     db.prepare("DELETE FROM test_files WHERE repo_id = ?").run(repoId);
-    deleteExistingArchitectureData(db, repoId);
+    deleteExistingArchitectureData(db, repoId, repo, existingPatternRowIds, options);
     const insertFile = db.prepare(
       `INSERT INTO code_files
        (repo_id, path, language, size_bytes, content_hash, updated_at)
@@ -2132,8 +2187,8 @@ function replaceCodeIndex(db, repo, codeFiles, codeChunks, skippedFiles, cwd, ar
     );
     const insertFts = db.prepare(
       `INSERT INTO code_chunks_fts
-       (chunkId, sanitizedText, filePath, symbols, language)
-       VALUES (?, ?, ?, ?, ?)`
+       (rowid, chunkId, sanitizedText, filePath, symbols, language)
+       VALUES (?, ?, ?, ?, ?, ?)`
     );
     options.onProgress?.({
       stage: "writing_code_chunks",
@@ -2146,7 +2201,7 @@ function replaceCodeIndex(db, repo, codeFiles, codeChunks, skippedFiles, cwd, ar
     for (const [index, chunk] of codeChunks.entries()) {
       const fileId = fileIds.get(chunk.filePath);
       if (!fileId) continue;
-      insertChunk.run(
+      const chunkInsert = insertChunk.run(
         chunk.id,
         repoId,
         fileId,
@@ -2161,6 +2216,7 @@ function replaceCodeIndex(db, repo, codeFiles, codeChunks, skippedFiles, cwd, ar
         chunk.updatedAt
       );
       insertFts.run(
+        Number(chunkInsert.lastInsertRowid),
         chunk.id,
         chunk.sanitizedText,
         chunk.filePath,
@@ -2222,10 +2278,19 @@ function replaceCodeIndex(db, repo, codeFiles, codeChunks, skippedFiles, cwd, ar
     databasePath: defaultDatabasePath(cwd)
   };
 }
-function deleteExistingArchitectureData(db, repoId) {
-  db.prepare(
-    "DELETE FROM architecture_patterns_fts WHERE patternId IN (SELECT id FROM architecture_patterns WHERE repo_id = ?)"
-  ).run(repoId);
+function deleteExistingArchitectureData(db, repoId, repo, patternRowIds, options = {}) {
+  deleteFtsRowsByRowId(
+    db,
+    "architecture_patterns_fts",
+    patternRowIds.map((row) => row.rowid),
+    (current, total) => options.onProgress?.({
+      stage: "deleting_architecture_fts",
+      repo,
+      current,
+      total,
+      patterns: patternRowIds.length
+    })
+  );
   db.prepare("DELETE FROM architecture_patterns WHERE repo_id = ?").run(repoId);
   db.prepare("DELETE FROM architecture_components WHERE repo_id = ?").run(repoId);
   db.prepare("DELETE FROM code_imports WHERE repo_id = ?").run(repoId);
@@ -2308,8 +2373,8 @@ function insertArchitectureData(db, repoId, repo, architecture, options = {}) {
      VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)`
   );
   const insertFts = db.prepare(
-    `INSERT INTO architecture_patterns_fts (patternId, summary, area, sourceFiles, symbols)
-     VALUES (?, ?, ?, ?, ?)`
+    `INSERT INTO architecture_patterns_fts (rowid, patternId, summary, area, sourceFiles, symbols)
+     VALUES (?, ?, ?, ?, ?, ?)`
   );
   options.onProgress?.({
     stage: "writing_architecture_data",
@@ -2319,7 +2384,7 @@ function insertArchitectureData(db, repoId, repo, architecture, options = {}) {
     kind: "patterns"
   });
   for (const [index, pattern] of architecture.patterns.entries()) {
-    insertPattern.run(
+    const patternInsert = insertPattern.run(
       pattern.id,
       repoId,
       pattern.repo,
@@ -2333,6 +2398,7 @@ function insertArchitectureData(db, repoId, repo, architecture, options = {}) {
       pattern.createdAt
     );
     insertFts.run(
+      Number(patternInsert.lastInsertRowid),
       pattern.id,
       pattern.sanitizedSummary,
       pattern.area,
@@ -2921,14 +2987,24 @@ function buildRelatedTestIndex(allPaths) {
   const testPaths = allPaths.filter((candidate) => isTestFilePath(candidate));
   const byBase = /* @__PURE__ */ new Map();
   const byDirectory = /* @__PURE__ */ new Map();
+  const byDotPrefix = /* @__PURE__ */ new Map();
   for (const testPath of testPaths) {
     addToStringMap(byBase, testBaseFor(testPath), testPath);
-    const segments = path6.posix.dirname(testPath).split("/").filter(Boolean);
-    for (let index = 1; index <= segments.length; index += 1) {
-      addToStringMap(byDirectory, segments.slice(0, index).join("/"), testPath);
+    const dirSegments = path6.posix.dirname(testPath).split("/").filter(Boolean);
+    for (let index = 1; index <= dirSegments.length; index += 1) {
+      addToStringMap(byDirectory, dirSegments.slice(0, index).join("/"), testPath);
+    }
+    const pathSegments2 = testPath.split("/");
+    const dotPrefixes = /* @__PURE__ */ new Set();
+    for (let i = 1; i < pathSegments2.length; i += 1) {
+      const segment = pathSegments2[i] ?? "";
+      for (let dot = segment.indexOf("."); dot >= 0; dot = segment.indexOf(".", dot + 1)) {
+        dotPrefixes.add(segment.slice(0, dot));
+      }
     }
+    for (const prefix of dotPrefixes) addToStringMap(byDotPrefix, prefix, testPath);
   }
-  return { testPaths, byBase, byDirectory };
+  return { testPaths, byBase, byDirectory, byDotPrefix };
 }
 function relatedTestsFor(filePath, index) {
   if (isTestFilePath(filePath)) return [];
@@ -2945,8 +3021,8 @@ function relatedTestsFor(filePath, index) {
   if (parsed.dir) {
     for (const testPath of index.byDirectory.get(parsed.dir) ?? []) add(testPath);
   }
-  for (const testPath of index.testPaths) {
-    if (testPath.includes(`/${basename}.`)) add(testPath);
+  for (const testPath of index.byDotPrefix.get(basename) ?? []) {
+    add(testPath);
     if (related.length >= 8) break;
   }
   return related.slice(0, 8);
@@ -4094,8 +4170,7 @@ function symbolMatch2(unit, querySymbols) {
     const lower = symbol.toLowerCase();
     if (unitSymbols.includes(lower)) best = Math.max(best, 1);
     else if (text.includes(`\`${lower}\``)) best = Math.max(best, 1);
-    else if (new RegExp(`\\b${escapeRegExp(lower)}\\b`, "i").test(text))
-      best = Math.max(best, 0.66);
+    else if (symbolBoundaryRegex(lower).test(text)) best = Math.max(best, 0.66);
     else if (unitSymbols.some((candidate) => candidate.includes(lower) || lower.includes(candidate))) {
       best = Math.max(best, 0.35);
     }
@@ -4178,6 +4253,15 @@ function scoreUnit(unit, input, duplicateCount, repeatedEvidenceCount, freshness
 function escapeRegExp(value) {
   return value.replace(/[.*+?^${}()|[\]\\]/g, "\\$&");
 }
+var symbolBoundaryRegexCache = /* @__PURE__ */ new Map();
+function symbolBoundaryRegex(lower) {
+  let regex = symbolBoundaryRegexCache.get(lower);
+  if (!regex) {
+    regex = new RegExp(`\\b${escapeRegExp(lower)}\\b`, "i");
+    symbolBoundaryRegexCache.set(lower, regex);
+  }
+  return regex;
+}
 function loadCandidates(db, input) {
   const ftsQuery = buildFtsQuery(input);
   const categories = "categories" in input ? input.categories ?? [] : [];
@@ -4213,9 +4297,11 @@ function loadClaimRepetitionCounts(db) {
   }
   return new Map([...grouped.entries()].map(([key, prs]) => [key, prs.size]));
 }
-function loadFeedbackAdjustments(db) {
-  const rows = db.prepare("SELECT result_id, rating FROM feedback_events").all();
+function loadFeedbackAdjustments(db, resultIds) {
   const adjustments = /* @__PURE__ */ new Map();
+  if (resultIds.length === 0) return adjustments;
+  const placeholders = resultIds.map(() => "?").join(", ");
+  const rows = db.prepare(`SELECT result_id, rating FROM feedback_events WHERE result_id IN (${placeholders})`).all(...resultIds);
   for (const row of rows) {
     const delta = row.rating === "useful" ? 0.03 : -0.03;
     adjustments.set(row.result_id, (adjustments.get(row.result_id) ?? 0) + delta);
@@ -4235,7 +4321,10 @@ function rankWisdomUnits(db, input) {
   const candidates = loadCandidates(db, input);
   const codeSnapshot = loadCurrentCodeSnapshot(db);
   const repetitionCounts = loadClaimRepetitionCounts(db);
-  const feedbackAdjustments = loadFeedbackAdjustments(db);
+  const feedbackAdjustments = loadFeedbackAdjustments(
+    db,
+    candidates.map((unit) => unit.id)
+  );
   const duplicates = /* @__PURE__ */ new Map();
   for (const unit of candidates) {
     const key = claimKeyFor(unit.category, unit.sanitizedText);
@@ -9897,8 +9986,9 @@ function queryTerms(input) {
 function matchesRepo(repo, repos) {
   return !repos || repos.length === 0 || repos.includes(repo);
 }
-function rowScore(input, text, files, symbols) {
+function rowScore(input, text, files, symbols, lowerTerms) {
   let score = 0;
+  const lowerText = text.toLowerCase();
   for (const file of input.files ?? []) {
     if (files.includes(file)) score += 5;
     else if (files.some((candidate) => candidate.endsWith(`/${file.split("/").pop() ?? file}`)))
@@ -9906,10 +9996,10 @@ function rowScore(input, text, files, symbols) {
   }
   for (const symbol of input.symbols ?? []) {
     if (symbols.includes(symbol)) score += 4;
-    else if (text.toLowerCase().includes(symbol.toLowerCase())) score += 1;
+    else if (lowerText.includes(symbol.toLowerCase())) score += 1;
   }
-  for (const term of queryTerms(input)) {
-    if (text.toLowerCase().includes(term.toLowerCase())) score += 0.5;
+  for (const term of lowerTerms) {
+    if (lowerText.includes(term)) score += 0.5;
   }
   return score;
 }
@@ -9920,9 +10010,10 @@ function getWisdom(db, input, limit) {
        ORDER BY confidence DESC, created_at DESC
        LIMIT 500`
   ).all();
+  const lowerTerms = queryTerms(input).map((term) => term.toLowerCase());
   return rows.filter((row) => matchesRepo(row.repo, input.repos)).map((row) => ({
     row,
-    score: rowScore(input, row.sanitized_text, parseStringArray(row.file_paths_json), [])
+    score: rowScore(input, row.sanitized_text, parseStringArray(row.file_paths_json), [], lowerTerms)
   })).filter((item) => item.score > 0 || (input.files ?? []).length === 0).sort((a, b) => b.score - a.score || b.row.confidence - a.row.confidence).slice(0, limit).map((item) => item.row);
 }
 function getCodeEvidence(db, input, limit) {
@@ -9932,13 +10023,15 @@ function getCodeEvidence(db, input, limit) {
        ORDER BY updated_at DESC
        LIMIT 800`
   ).all();
+  const lowerTerms = queryTerms(input).map((term) => term.toLowerCase());
   return rows.filter((row) => matchesRepo(row.repo, input.repos)).map((row) => ({
     row,
     score: rowScore(
       input,
       row.sanitized_text,
       [row.file_path],
-      parseStringArray(row.symbols_json)
+      parseStringArray(row.symbols_json),
+      lowerTerms
     )
   })).filter((item) => item.score > 0).sort((a, b) => b.score - a.score).slice(0, limit).map((item) => item.row);
 }
@@ -9949,23 +10042,30 @@ function getArchitecture(db, input, limit) {
        ORDER BY confidence DESC, created_at DESC
        LIMIT 300`
   ).all();
+  const lowerTerms = queryTerms(input).map((term) => term.toLowerCase());
   return rows.filter((row) => matchesRepo(row.repo, input.repos)).map((row) => ({
     row,
-    score: rowScore(input, row.summary_sanitized, parseStringArray(row.source_files_json), [])
+    score: rowScore(
+      input,
+      row.summary_sanitized,
+      parseStringArray(row.source_files_json),
+      [],
+      lowerTerms
+    )
   })).filter((item) => item.score > 0 || (input.files ?? []).length === 0).sort((a, b) => b.score - a.score || b.row.confidence - a.row.confidence).slice(0, limit).map((item) => item.row);
 }
 function findOrgApiConsumers(db, config, input) {
   initializeSchema(db);
+  const repoClause = input.repo ? " AND (provider_repo = ? OR consumer_repo = ?)" : "";
+  const repoParams = input.repo ? [input.repo, input.repo] : [];
   const rows = db.prepare(
     `SELECT provider_repo, provider_path, consumer_repo, consumer_path, contract, evidence_json, confidence
        FROM org_api_consumers
-       WHERE org = ?
+       WHERE org = ?${repoClause}
        ORDER BY confidence DESC`
-  ).all(config.org);
+  ).all(config.org, ...repoParams);
   const limit = Math.max(1, Math.min(input.maxResults ?? 8, 25));
-  return rows.filter(
-    (row) => !input.repo || row.provider_repo === input.repo || row.consumer_repo === input.repo
-  ).filter((row) => {
+  return rows.filter((row) => {
     const files = input.files ?? [];
     if (files.length === 0 && !input.query) return true;
     return files.some((file) => row.provider_path === file || row.consumer_path === file) || Boolean(input.query && row.contract.toLowerCase().includes(input.query.toLowerCase()));