Conrad Ludgate
Conrad Ludgate
> could you check if it's still effective? Construction time seems to be the same (+-5% in the majority of tests). Lookup time is still 2.5x faster
Tested some more using foldhash::fast, and it seems to still have good performance (slightly better), so I switched to that and removed the code that was no longer necessary
I'll _fold_ in those changes soon, thanks!
Looks like this will break `tokio-postgres`, cc @sfackler. https://github.com/sfackler/rust-postgres/blob/490c4852492bb0adf63a83e3c4e8b2afc9e6bf54/codegen/src/sqlstate.rs#L139-L152 Since this codegen uses the phf version in the repo lockfile, which isn't guaranteed to match the phf version eventually used...
Superceded by #12498
Yes, that is expected. I've been considering changing it to properly respect what flags are supported under interactive or not, but I haven't managed to find a decent solution yet
> I'm pretty sure ring has been deprecated, so let's not touch that. It has new maintainers (same as rustls), so it's still fine for now. Long term, graviola will...
I am working on adding some integration tests in a separate branch. Will open a PR in the coming days.
Seems I forgot about this. Will check today
> Would the change in #7252 preclude a non breaking change in this API? Can we merge this without 7252 then merge in 7252 as an optimization later? Yes, that's...