thanks! this is certainly interesting!

The thing that worries me about removing the ajv validation is that it's extremely critical to all subsequent codepaths that the input is a valid HAR. If you replaced it with hand-rolled code that did the same checks as AJV I'd be totally cool with that, but it looks like it was just outright removed. Did I read that correctly in the diff?

Also, could you please say more about the advantages you see with switching to vitest? I'm certainly open-minded on the topic, but I don't know what we substantively gain by making the switch. For example, can you please post some before-and-after runs for any performance differences so I can compare to my own runs?

So ajv has a compiled mode - it uses the exact same logic as normal ajv, but creates a single binary with no dependency on ajv. You can see what the compiled code looks like here https://github.com/jonluca/har-validator/blob/main/src/validate.js

This is exactly the same logic that ajv and the har schema would be using, just compiled at build time instead of runtime. We should see bundle size improvements and runtime speed improvements by using the compiled version.

For example, can you please post some before-and-after runs for any performance differences so I can compare to my own runs?

Sure, here's a screenshot

The nice part is that 1) it has an enormous emphasis on speed (as the entire vite toolchain does) and 2) it comes with support for code coverage reporting out of the box, in a way that's faster and nicer (IMO) to jest.

And vitest is a much smaller and more lightweight library to jest, it'll improve npm install times as well.

All committers have signed the CLA.

Closing as stale

Are you not interested in using the compiled one? It would improve the install size significantly, and speed it up as well