Conrad Ludgate
Conrad Ludgate
Implements bidirectional map, fixes #278. I'm not interested in driving this forward, as I don't have a need for it. But since the issue said PRs welcome, I thought I'd...
Because of the requirement that `S: Unpin`, there's no need for awkward pin projection code. Alternatively, tokio already depends on [pin-project-lite](https://docs.rs/pin-project-lite/latest/pin_project_lite/) [unconditionally](https://github.com/tokio-rs/tokio/blob/0dbdd196b6a9af9a106c654a8613088b81a64655/tokio/Cargo.toml#L91). So if you don't want to rely on...
While [updating our test suite](https://github.com/neondatabase/neon/pull/9613) to use the latest tokio-postgres-rustls, I noticed that our channel binding tests started failing. I suspected this might be an incompatibility with rcgen but when...
`WakerList` is currently doing too much - which makes it a difficult abstraction to manage. Splitting it into `ThinArcList` and `WakerList`. `ThinArcList` is like [`triomphe::ThinArc`](https://docs.rs/triomphe/latest/triomphe/struct.ThinArc.html), except that each entry stores...
See #11942 Idea: * if connections are short lived, they can get enqueued and then also remove themselves later if they never made it to redis. This reduces the load...
With high TCP/WS connect load, we cannot always process cancellation events fast enough. This causes the queue to fill up and the backpressure causes some slow downs. Ideas: * Accept...
SQL-over-http flow makes heavy use of password hashing. Can we optimise the implementation of pbkdf2? We currently use the sha2 crate but there's other sha2 implementations in our dep tree...
We parse unstructured, user-provided json during sql-over-http. On large inputs, this might be slow/memory intensive. Maybe consider https://github.com/conradludgate/sonny-jim?
#11962 A few changes: * refine hashmap usage, faster hashing. * replace some hashmaps with vecs based on field index * use linear search for extract - less effort than...
In our continous profiles, tracing is noticeable in the profile output. We should try and optimise the json formatting layer since we interact with it everywhere in the proxy codebase.