Add ada-url dependency, initial impl of jsg::Url

[jasnell]

This will serve as the new underlying implementation of the url-standard.{h|c++} class while also supporting handling of module import specifiers as URLs.

Why using ada-url at all? It's fast and spec compliant (verified against the web platform tests for the URL spec). And it allows us to delete a bunch of code relating to our current standard URL parser. That'll be done in a separate URL as we need to verify that switching won't break any edge cases.

/cc @anonrig

[irvinebroque]

🎉 exciting! two questions

1. Re: edge cases — maybe a way to run ada in an observe only mode for some period of time, for some subset of requests, log any differences in output between our current implementation and ada-url? or do we know enough with confidence to hunt these down upfront?
2. Ada is fast, wonder how much faster than what we have today?

[jasnell]

@irvinebroque:

Re: edge cases — maybe a way to run ada in an observe only mode for some period of time, for some subset of requests, log any differences in output between our current implementation and ada-url? or do we know enough with confidence to hunt these down upfront?

I do not plan on replacing the underlying implementation of url-standard.c++ right away. What I want to do is set up a parallel of tests first to ensure that all of our existing URL tests are handled just fine,.

Ada is fast, wonder how much faster than what we have today?

The current url-standard implementation was based entirely off of the original Node.js implementation. I would expect the performance improvement here to be able the same as we see from Node.js' switch to ada.

[jasnell]

This is currently having challenges on windows due to the simd instructions. Will need to get that figured out.

[fhanau]

I believe the SIMD problem here is the same problem we had with other code after the Github runner image upgraded to LLVM 16 – this caused bazel to include the wrong SIMD header. Orion developed a workaround and contributed a fix (bazelbuild/bazel#19391) that will be available in bazel 6.4.0. To solve the issues for this PR, you can adjust .bazelrc to modify the includes like Orion did in the workaround. 6.4.0 should be released next week though, so you can also see if that fixes things here too by testing with the release candidate as in #1277.

[jasnell]

woo green ci! next step is to update the internal repo workspace to make sure this will compile.

[jasnell]

Internal CI looks good.

[jasnell]

Added basic parse tests based on the corpus of WPT url tests. These don't test all of the getters/setters on jsg::Url but they do test that URLs that should be parseable are parseable and that the url.getHref() value is correct. These prove that we're able to successfully parse even more than the current standard workerd::api::URL supports!

[jasnell]

Going to wait to land this until after we also get the url search params iterator support landed up stream, that way we can convert URLSearchParams over to this also. ada-url/ada#532

[jasnell]

Ok, jsg::UrlSearchParams added

[jasnell]

Merging this. There's still a lot to do before we can really start using it.

Just a taste:

• Need to update the URLPattern implementation to decouple it from the old url-standard impl
• Need to add more fuzz testing for the ada url implementation (just for our own testing)
• Need to update the url-standard impl to use this instead