finch
finch copied to clipboard
Elixir HTTP client, focused on performance
This is to avoid eating all the memory when fetching from untrusted links.
I'd like to be able to instrument connects from my finch instances. The various connect telemetry events don't expose the pool ID or Finch instance. I'd love to see this...
Correct me if I'm wrong, but it seems there is no way of querying finch for a pool's current state, for example, number of available workers. `:poolboy` exposes this information...
In https://github.com/wojtekmach/req/issues/81, we noticed Finch exits on an invalid URL: ```elixir Mix.install([ {:finch, "~> 0.11.0"} ]) {:ok, _} = Finch.start_link(name: MyFinch) Finch.build(:get, "http://") |> Finch.request(MyFinch) |> IO.inspect() ``` ``` **...
Currently our Telemetry tests only run against HTTP1 pools. We could refactor the test suite in a way that we can also run them against HTTP2 pools. Unfortunately, there is...
The OTP team is working on exciting new capability and that's using the system provided certs store: https://github.com/erlang/otp/pull/5853. When that ships, I believe Finch would no longer have to depend...
fixes #125 Error before: ```elixir 19:09:43.864 [error] GenServer #PID terminating ** (stop) exited in: NimblePool.checkout(#PID) ** (EXIT) time out (finch 0.6.3) lib/finch/http1/pool.ex:58: Finch.HTTP1.Pool.request/5 (finch 0.6.3) lib/finch.ex:247: Finch.request/3 (tesla 1.4.0) lib/tesla/adapter/finch.ex:63:...
Hi, I'd like to report an issue of using Finch with HTTP proxy. ## Problem Requests to plain HTTP (i.e.: non-HTTPS) URLs via proxy are handled in `Mint.UnsafeProxy`, but currently...
When under very high load with many concurrent requests, some requests may fail with ```elixir {:error, %Mint.HTTPError{module: Mint.HTTP2, reason: :too_many_concurrent_requests}} ``` From the [`Mint.HTTP2.request/5` docs](https://hexdocs.pm/mint/Mint.HTTP2.html#request/5): > In HTTP/2, opening a...
While trying to add OpenTelemetry tracing to some of our applications, we couldn't find an easy solution to inject tracing headers into all requests. We'd love to hear about it...