Show HN: Malleon – 将真实用户会话转化为自动化测试
2 分•作者: godelshalt•6 个月前
嗨，HN，
我一直在构建一个名为 Malleon 的工具，因为我厌倦了无法反映用户实际操作的端到端测试。而且我一直痴迷于使用真实用户会话进行负载测试。在使用 Tsung 和 Gatling 等工具进行负载测试后，我经常希望可以直接重放昨天的流量 x2 或 x5，而不是使用合成会话。Malleon 最初是试图弥合这两个世界之间的差距。
基本思路是：与其从头开始编写测试，不如记录真实的用户会话，并将它们转化为可重放的测试。
SDK 记录的内容包括：
* DOM 交互
* 网络请求
* 控制台输出
* 操作之间的时间
重放不是视频。它会重建并重放针对您的应用程序的实际浏览器交互。
因此，真实的用户会话就变成了一个可复现的测试用例。
典型的流程如下：
* 在您的应用程序中嵌入一个小的 JS SDK
* 用户正常与网站交互
* 会话被记录
* 您浏览重放并找到一些有趣的东西（错误、异常、奇怪的行为）
* 将该会话转化为测试并在 CI 中运行它
测试运行器是自托管的。您可以拉取 Docker 镜像并在任何您想运行的地方运行它。它驱动一个浏览器（无头或有头）并重放交互序列。
在构建过程中，有些事情出乎意料地棘手：
* 当 DOM 自记录以来发生变化时，重放交互
* 处理视口/布局差异
* 使网络重放对应用程序透明
* 在保持测试速度的同时保持时间真实
该系统还收集日志、错误和请求时间，以便可以搜索会话，并且您可以跟踪网络性能，查看 p90/p95/p99 统计数据，以及所有这些好东西。
链接：
* Malleon: [https://malleon.io](https://malleon.io)
* Replay SDK: [https://www.npmjs.com/package/@malleon/replay](https://www.npmjs.com/package/@malleon/replay)
* Replay CLI: [https://www.npmjs.com/package/@malleon/replay-cli](https://www.npmjs.com/package/@malleon/replay-cli)
* 文档: [https://github.com/malleonio/malleon-documentation](https://github.com/malleonio/malleon-documentation)
提供免费套餐。
好奇是否还有其他人遇到过“我们的测试无法反映用户的实际操作（或他们的扩展方式）”这个问题。
查看原文
Hey HN,<p>I've been building a tool called Malleon because I got tired of e2e tests that don't reflect what users actually do. And I have long been obsessed with using real user sessions for load testing. Having worked with tools like Tsung and Gatling for load testing, I often wished I could just replay yesterday's traffic x2 or x5 instead of using synthetic sessions. Malleon started as an attempt to bridge those two worlds.<p>The basic idea is: instead of writing tests from scratch, record real user sessions and turn them into replayable tests.<p>The SDK records things like:
- DOM interactions
- network requests
- console output
- timing between actions<p>The replay isn't a video. It reconstructs and replays the actual browser interactions against your app.<p>So a real user session becomes a reproducible test case.<p>Typical flow looks like:
- drop a small JS SDK into your app
- users interact with the site normally
- sessions get recorded
- you browse replays and find something interesting (bug, error, weird behavior)
- turn that session into a test and run it in CI<p>The test runner is self-hosted. You pull the Docker image and run it wherever you want. It drives a browser (headless or headful) and replays the interaction sequence.<p>Some things that turned out to be surprisingly tricky while building this:
- replaying interactions when the DOM has changed since recording
- handling viewport/layout differences
- making network replay transparent to the app
- keeping timing realistic without making tests slow<p>The system also collects logs, errors, and request timings so sessions are searchable and you can track network performance, see p90/p95/p99 stats, all that good stuff.<p>Links:
- Malleon: <a href="https://malleon.io" rel="nofollow">https://malleon.io</a>
- Replay SDK: <a href="https://www.npmjs.com/package/@malleon/replay" rel="nofollow">https://www.npmjs.com/package/@malleon/replay</a>
- Replay CLI: <a href="https://www.npmjs.com/package/@malleon/replay-cli" rel="nofollow">https://www.npmjs.com/package/@malleon/replay-cli</a>
- Docs: <a href="https://github.com/malleonio/malleon-documentation" rel="nofollow">https://github.com/malleonio/malleon-documentation</a><p>Free tier available.<p>Curious if anyone else has run into the "our tests don't reflect what users actually do (or how they scale)" problem.