autosnippet 3.2.7 → 3.2.8

package/lib/core/analysis/CallGraphAnalyzer.js

@@ -0,0 +1,367 @@
+/**
+ * @module CallGraphAnalyzer
+ * @description Phase 5: 顶层编排器 — 协调 Call Graph 分析的全流程
+ *
+ * 流水线:
+ *   1. CallSiteExtractor  — 从 AST 提取调用点 (已在 AstAnalyzer 二次遍历中完成)
+ *   2. SymbolTableBuilder  — 构建全局符号表
+ *   3. ImportPathResolver  — 导入路径解析器
+ *   4. CallEdgeResolver    — 调用点 → 调用边
+ *   5. DataFlowInferrer    — 调用边 → 数据流边
+ *
+ * 输出:
+ *   { callEdges, dataFlowEdges, stats }
+ */
+
+import { SymbolTableBuilder } from './SymbolTableBuilder.js';
+import { ImportPathResolver } from './ImportPathResolver.js';
+import { CallEdgeResolver } from './CallEdgeResolver.js';
+import { DataFlowInferrer } from './DataFlowInferrer.js';
+
+/**
+ * @typedef {object} CallGraphResult
+ * @property {import('./CallEdgeResolver.js').ResolvedEdge[]} callEdges
+ * @property {import('./DataFlowInferrer.js').DataFlowEdge[]} dataFlowEdges
+ * @property {CallGraphStats} stats
+ */
+
+/**
+ * @typedef {object} CallGraphStats
+ * @property {number} totalCallSites — 总调用点数
+ * @property {number} resolvedCallSites — 成功解析的调用点数
+ * @property {number} resolvedRate — 解析成功率 (0-1)
+ * @property {number} totalEdges — 总边数 (call + data_flow)
+ * @property {number} filesProcessed — 处理的文件数
+ * @property {number} symbolCount — 符号表大小
+ * @property {number} durationMs — 总耗时
+ */
+
+/**
+ * @typedef {object} AnalyzeOptions
+ * @property {number} [timeout=15000] — 超时(ms)
+ * @property {number} [maxCallSitesPerFile=500] — 每文件最多调用点
+ * @property {number} [minConfidence=0.5] — 最低置信度
+ */
+
+export class CallGraphAnalyzer {
+  /**
+   * @param {string} projectRoot
+   */
+  constructor(projectRoot) {
+    this.projectRoot = projectRoot;
+  }
+
+  /**
+   * 执行完整的调用图分析
+   *
+   * @param {object} astProjectSummary — analyzeProject() 的输出 (需包含 callSites)
+   * @param {AnalyzeOptions} [options]
+   * @returns {Promise<CallGraphResult>}
+   */
+  async analyze(astProjectSummary, options = {}) {
+    const t0 = Date.now();
+    const timeout = options.timeout || 15_000;
+    const maxCallSitesPerFile = options.maxCallSitesPerFile || 500;
+
+    if (!astProjectSummary?.fileSummaries?.length) {
+      return this._emptyResult(Date.now() - t0);
+    }
+
+    // ── 渐进式超时: 逐文件检查超时，返回 partial result ──
+    const deadline = t0 + timeout;
+    const result = await this._doAnalyze(astProjectSummary, maxCallSitesPerFile, deadline);
+    result.stats.durationMs = Date.now() - t0;
+    return result;
+  }
+
+  /**
+   * 增量分析 — 仅重新分析变更文件及其依赖方
+   *
+   * @param {object} astProjectSummary — analyzeProject() 的全量输出
+   * @param {string[]} changedFiles — 变更文件的相对路径列表
+   * @param {AnalyzeOptions} [options]
+   * @returns {Promise<CallGraphResult>}
+   */
+  async analyzeIncremental(astProjectSummary, changedFiles, options = {}) {
+    const t0 = Date.now();
+    const timeout = options.timeout || 15_000;
+    const maxCallSitesPerFile = options.maxCallSitesPerFile || 500;
+
+    if (!astProjectSummary?.fileSummaries?.length || !changedFiles?.length) {
+      return this._emptyResult(Date.now() - t0);
+    }
+
+    // ── 超过 10 个文件变更 → 回退全量分析 ──
+    if (changedFiles.length > 10) {
+      return this.analyze(astProjectSummary, options);
+    }
+
+    const deadline = t0 + timeout;
+    const changedSet = new Set(changedFiles);
+
+    // ── Step 1: 构建全局符号表 (始终全量，确保跨文件符号可解析) ──
+    const symbolTable = SymbolTableBuilder.build(astProjectSummary);
+
+    // ── Step 2: 构建 ImportPathResolver ──
+    const allFiles = astProjectSummary.fileSummaries.map((f) => f.file);
+    const importResolver = new ImportPathResolver(this.projectRoot, allFiles);
+
+    // ── Step 3: 找到依赖变更文件的所有文件 (reverse dependency) ──
+    const affectedFiles = new Set(changedFiles);
+    for (const fileSummary of astProjectSummary.fileSummaries) {
+      const imports = symbolTable.fileImports.get(fileSummary.file) || [];
+      for (const imp of imports) {
+        const resolved = importResolver.resolve(String(imp), fileSummary.file);
+        if (resolved && changedSet.has(resolved)) {
+          affectedFiles.add(fileSummary.file);
+          break;
+        }
+      }
+    }
+
+    // ── Step 4: 仅对受影响文件解析调用边 ──
+    const fileCount = astProjectSummary.fileSummaries.length;
+    const tier = _computeTier(fileCount);
+    const useCHA = tier === 'full-cha';
+    const inheritanceGraph = useCHA ? (astProjectSummary.inheritanceGraph || []) : [];
+    const callEdgeResolver = new CallEdgeResolver(symbolTable, importResolver, inheritanceGraph);
+    const allCallEdges = [];
+    let totalCallSites = 0;
+    let processedFiles = 0;
+
+    for (const fileSummary of astProjectSummary.fileSummaries) {
+      if (!affectedFiles.has(fileSummary.file)) continue;
+
+      const callSites = fileSummary.callSites || [];
+      if (callSites.length === 0) continue;
+
+      // 超时检查
+      if (Date.now() > deadline) {
+        return {
+          callEdges: allCallEdges,
+          dataFlowEdges: DataFlowInferrer.infer(allCallEdges),
+          stats: {
+            totalCallSites,
+            resolvedCallSites: allCallEdges.length,
+            resolvedRate: totalCallSites > 0 ? allCallEdges.length / totalCallSites : 0,
+            totalEdges: allCallEdges.length,
+            filesProcessed: processedFiles,
+            symbolCount: symbolTable.declarations.size,
+            durationMs: Date.now() - t0,
+            tier,
+            partial: true,
+            incremental: true,
+            processedFiles,
+            totalFiles: affectedFiles.size,
+            changedFiles: changedFiles.length,
+            affectedFiles: affectedFiles.size,
+          },
+        };
+      }
+
+      const limitedCallSites =
+        callSites.length > maxCallSitesPerFile
+          ? callSites.slice(0, maxCallSitesPerFile)
+          : callSites;
+
+      totalCallSites += limitedCallSites.length;
+
+      const edges = callEdgeResolver.resolveFile(limitedCallSites, fileSummary.file);
+      allCallEdges.push(...edges);
+      processedFiles++;
+    }
+
+    // ── Step 5: 推断数据流 ──
+    const dataFlowEdges = DataFlowInferrer.infer(allCallEdges);
+
+    return {
+      callEdges: allCallEdges,
+      dataFlowEdges,
+      stats: {
+        totalCallSites,
+        resolvedCallSites: allCallEdges.length,
+        resolvedRate: totalCallSites > 0 ? allCallEdges.length / totalCallSites : 0,
+        totalEdges: allCallEdges.length + dataFlowEdges.length,
+        filesProcessed: processedFiles,
+        symbolCount: symbolTable.declarations.size,
+        durationMs: Date.now() - t0,
+        tier,
+        incremental: true,
+        changedFiles: changedFiles.length,
+        affectedFiles: affectedFiles.size,
+      },
+    };
+  }
+
+  /**
+   * @private 实际分析逻辑
+   *
+   * 分级降级策略 (§5.2):
+   *   - <100 文件  → 完整分析 (含 CHA)
+   *   - 100-500   → 完整分析，禁用 CHA
+   *   - 500-2000  → 抽样分析 (核心目录优先)
+   *   - >2000     → 仅模块级 import graph (跳过调用边解析)
+   *
+   * 渐进式超时 (§13 Issue #15):
+   *   每处理完一个文件检查 deadline，超时时返回已有的 partial result
+   *
+   * @param {object} astProjectSummary
+   * @param {number} maxCallSitesPerFile
+   * @param {number} deadline — Date.now() + timeout
+   */
+  async _doAnalyze(astProjectSummary, maxCallSitesPerFile, deadline) {
+    const fileCount = astProjectSummary.fileSummaries.length;
+
+    // ── 分级降级 ──
+    const tier = _computeTier(fileCount);
+    let fileSummaries = astProjectSummary.fileSummaries;
+
+    if (tier === 'import-only') {
+      // >2000 文件: 仅返回 import graph，不解析调用边
+      return {
+        callEdges: [],
+        dataFlowEdges: [],
+        stats: {
+          totalCallSites: 0,
+          resolvedCallSites: 0,
+          resolvedRate: 0,
+          totalEdges: 0,
+          filesProcessed: fileCount,
+          symbolCount: 0,
+          durationMs: 0,
+          tier: 'import-only',
+        },
+      };
+    }
+
+    if (tier === 'sampled') {
+      // 500-2000 文件: 抽样核心目录 (仅限制 call site 解析范围)
+      fileSummaries = _sampleCoreFiles(fileSummaries, 500);
+    }
+
+    // ── Step 2: 构建符号表 (始终使用全量文件，确保跨文件符号可解析) ──
+    const symbolTable = SymbolTableBuilder.build(astProjectSummary);
+
+    // ── Step 3: 构建 ImportPathResolver (全量文件索引) ──
+    const allFiles = astProjectSummary.fileSummaries.map((f) => f.file);
+    const importResolver = new ImportPathResolver(this.projectRoot, allFiles);
+
+    // ── Step 4: 解析调用边 (逐文件 + 超时检查) ──
+    const useCHA = tier === 'full-cha';
+    const inheritanceGraph = useCHA ? (astProjectSummary.inheritanceGraph || []) : [];
+    const callEdgeResolver = new CallEdgeResolver(symbolTable, importResolver, inheritanceGraph);
+    const allCallEdges = [];
+    let totalCallSites = 0;
+    let processedFiles = 0;
+    const totalFiles = fileSummaries.filter((f) => f.callSites?.length > 0).length;
+
+    for (const fileSummary of fileSummaries) {
+      const callSites = fileSummary.callSites || [];
+      if (callSites.length === 0) continue;
+
+      // ── 渐进式超时: 每文件检查 deadline ──
+      if (Date.now() > deadline) {
+        const dataFlowEdges = DataFlowInferrer.infer(allCallEdges);
+        return {
+          callEdges: allCallEdges,
+          dataFlowEdges,
+          stats: {
+            totalCallSites,
+            resolvedCallSites: allCallEdges.length,
+            resolvedRate: totalCallSites > 0 ? allCallEdges.length / totalCallSites : 0,
+            totalEdges: allCallEdges.length + dataFlowEdges.length,
+            filesProcessed: processedFiles,
+            symbolCount: symbolTable.declarations.size,
+            durationMs: 0,
+            tier,
+            partial: true,
+            processedFiles,
+            totalFiles,
+          },
+        };
+      }
+
+      // 防护: 限制每文件调用点数 (防止超大文件)
+      const limitedCallSites =
+        callSites.length > maxCallSitesPerFile
+          ? callSites.slice(0, maxCallSitesPerFile)
+          : callSites;
+
+      totalCallSites += limitedCallSites.length;
+
+      const edges = callEdgeResolver.resolveFile(limitedCallSites, fileSummary.file);
+      allCallEdges.push(...edges);
+      processedFiles++;
+    }
+
+    // ── Step 5: 推断数据流 ──
+    const dataFlowEdges = DataFlowInferrer.infer(allCallEdges);
+
+    // ── Stats ──
+    const stats = {
+      totalCallSites,
+      resolvedCallSites: allCallEdges.length,
+      resolvedRate: totalCallSites > 0 ? allCallEdges.length / totalCallSites : 0,
+      totalEdges: allCallEdges.length + dataFlowEdges.length,
+      filesProcessed: processedFiles,
+      symbolCount: symbolTable.declarations.size,
+      durationMs: 0, // 由外层填充
+      tier,
+    };
+
+    return { callEdges: allCallEdges, dataFlowEdges, stats };
+  }
+
+  /**
+   * @private 空结果
+   */
+  _emptyResult(durationMs) {
+    return {
+      callEdges: [],
+      dataFlowEdges: [],
+      stats: {
+        totalCallSites: 0,
+        resolvedCallSites: 0,
+        resolvedRate: 0,
+        totalEdges: 0,
+        filesProcessed: 0,
+        symbolCount: 0,
+        durationMs,
+      },
+    };
+  }
+}
+
+// ── 分级降级辅助 ──────────────────────────────────────────
+
+/**
+ * 根据文件数量确定分析层级
+ * @param {number} fileCount
+ * @returns {'full-cha'|'full'|'sampled'|'import-only'}
+ */
+function _computeTier(fileCount) {
+  if (fileCount < 100) return 'full-cha';
+  if (fileCount <= 500) return 'full';
+  if (fileCount <= 2000) return 'sampled';
+  return 'import-only';
+}
+
+/**
+ * 抽样核心文件 — 优先选取 src/、lib/、app/、core/ 等核心目录
+ * @param {object[]} fileSummaries
+ * @param {number} limit
+ * @returns {object[]}
+ */
+function _sampleCoreFiles(fileSummaries, limit) {
+  const CORE_DIRS = /\/(src|lib|app|core|pkg|internal|domain|service|controller|handler|api)\//i;
+  const scored = fileSummaries.map((f) => ({
+    f,
+    score: CORE_DIRS.test(f.file) ? 2 : 1,
+    callSiteCount: f.callSites?.length || 0,
+  }));
+  // 排序: 核心目录优先，有调用点的优先
+  scored.sort((a, b) => b.score - a.score || b.callSiteCount - a.callSiteCount);
+  return scored.slice(0, limit).map((s) => s.f);
+}
+
+export default CallGraphAnalyzer;