前端构建链路日常巡检要点-育师

前端构建链路日常巡检要点

大型项目的构建时间和产物体积会随着依赖、代码组织和构建配置逐步变化。与其依赖一次性优化，不如保存可比较的基线，并在合并请求中展示变化。构建变慢不必然意味着首屏变差，两类指标应分别观察。

本文给出一个巡检思路：检查依赖和产物变化，超过团队约定范围时附上报告交给人工判断。阈值需要按应用规模、压缩方式和访问路径设定。

1. 巡检痛点：重复依赖包（Duplicate Packages）默默拖垮体积

在日常巡检中，最常抓到的构建隐形杀手就是重复依赖。例如项目里同时存在lodash和lodash-es，或者不同微前端子包依赖了axios@0.21.1与axios@1.6.0。

多个版本的同一依赖可能增加安装体积，并可能进入多个输出 chunk；是否真的重复打包取决于入口、依赖图和 Rollup 的去重结果，需要结合产物分析确认。

// 巡检脚本中检测 package-lock.json / pnpm-lock.yaml 重复依赖的核心逻辑 import fs from "fs"; import path from "path"; export interface DuplicateDependencyReport { packageName: string; versions: string[]; } export function scanDuplicateDependencies(lockfilePath: string): DuplicateDependencyReport[] { if (!fs.existsSync(lockfilePath)) { throw new Error(`[Inspection Error] 未找到 Lock 文件: ${lockfilePath}`); } const content = fs.readFileSync(lockfilePath, "utf-8"); // 简化的 lockfile 版本匹配提取逻辑 const packageVersionMap = new Map<string, Set<string>>(); // 假设为 pnpm-lock 语法解析范例 const regex = /^\/([a-z0-9-@./]+)@([0-9.]+):/gm; let match; while ((match = regex.exec(content)) !== null) { const [, pkgName, version] = match; if (!packageVersionMap.has(pkgName)) { packageVersionMap.set(pkgName, new Set()); } packageVersionMap.get(pkgName)!.add(version); } const duplicates: DuplicateDependencyReport[] = []; packageVersionMap.forEach((versions, packageName) => { if (versions.size > 1) { duplicates.push({ packageName, versions: Array.from(versions) }); } }); return duplicates; }

2. 确定性巡检方案：编写 Vite 构建增量漂移诊断器

除了依赖扫描，日常巡检还需要精准比对构建产物体积与构建耗时的增量漂移（Incremental Drift）。如果单次 PR 导致打包产物增量超过 50KB，巡检脚本应立即发出预警。

我们实现了一套自动化巡检探针：

import { build, type InlineConfig } from "vite"; import { performance } from "node:perf_hooks"; export interface BuildInspectionMetrics { durationMs: number; totalBundleSizeBytes: number; chunkCount: number; } export class ViteBuildInspector { private baselineMetrics: BuildInspectionMetrics | null = null; public async runInspection(viteConfig: InlineConfig): Promise<BuildInspectionMetrics> { const startTime = performance.performance.now(); // 触发打包构建 const output = (await build({ ...viteConfig, build: { ...viteConfig.build, write: false }, // 不实际写入磁盘，加速巡检 })) as any; const durationMs = performance.performance.now() - startTime; let totalSizeBytes = 0; let chunkCount = 0; if (output && output.output) { for (const item of output.output) { if (item.type === "chunk") { chunkCount++; totalSizeBytes += Buffer.byteLength(item.code, "utf8"); } } } const currentMetrics: BuildInspectionMetrics = { durationMs: Number(durationMs.toFixed(2)), totalBundleSizeBytes: totalSizeBytes, chunkCount, }; this.auditMetrics(currentMetrics); this.baselineMetrics = currentMetrics; return currentMetrics; } private auditMetrics(current: BuildInspectionMetrics): void { console.log(`[Vite Daily Inspection] 构建耗时: ${current.durationMs}ms, 总体积: ${(current.totalBundleSizeBytes / 1024).toFixed(2)}KB, Chunk 数: ${current.chunkCount}`); if (this.baselineMetrics) { const sizeDriftRatio = (current.totalBundleSizeBytes - this.baselineMetrics.totalBundleSizeBytes) / this.baselineMetrics.totalBundleSizeBytes; if (sizeDriftRatio > 0.05) { console.warn(`[Inspection Warning] 产物体积比上次基准增长了 ${(sizeDriftRatio * 100).toFixed(2)}%! 请检查近期合并的 PR！`); } } } }