This required the breaking change of removing the Client trait. The
intent of the Client trait was to facilitate the decorator pattern by
allowing users to create their own Clients that added behavior on top of
the base client. Unfortunately, this trait had become a maintenance
burden, consistently causing issues with lifetimes and the lack of
generic associated types. Specifically, it meant that Client impls could
not use async fns, which is no longer tenable today.
Before this commit, each request future had its own timeout and would
communicate to the client Channel when a request was no longer being
listened to. Now, instead, the Channel tracks deadlines of in-flight
requests and completes requests with deadline-exceeded errors when they
expire.
This should be functionally equivalent to the previous way. It just cuts
down on the amount of two-way processing required. Unfortunately,
dropping a response future early still requires the client to send a
cancellation message to the Channel.
Before this commit, deadlines were handled by a timeout future that
wrapped each request handler. However, request handlers can be dropped
before sending a response back to the channel, so they can't be relied
on for channel state cleanup. Additionally, clients can't be relied on
to send cancellation messages. It was therefore theoretically possible
for pathological behaviors to cause an unbounded growth in orphan
request data in the Channel.
With this change, as long as requests sent have reasonable deadlines,
then the channel will be able to clean itself up. It is still possible
for requests to be sent with very large deadlines, which would prevent
the channel from cleaning itself up.
1. Renames
Some of the items in this module were renamed to be less generic:
- Handler => Incoming
- ClientHandler => Requests
- ResponseHandler => InFlightRequest
- Channel::{respond_with => requests}
In the case of Handler: handler of *what*? Now it's a bit clearer that
this is a stream of Channels (aka *incoming* connections).
Similarly, ClientHandler was a stream of requests over a single
connection. Hopefully Requests better reflects that.
ResponseHandler was renamed InFlightRequest because it no longer
contains the serving function. Instead, it is just the request, plus
the response channel and an abort hook. As a result of this,
Channel::respond_with underwent a big change: it used to take the
serving function and return a ClientHandler; now it has been renamed
Channel::requests and does not take any args.
2. Execute methods
All methods thats actually result in responses being generated
have been consolidated into methods named `execute`:
- InFlightRequest::execute returns a future that completes when a
response has been generated and sent to the server Channel.
- Requests::execute automatically spawns response handlers for all
requests over a single channel.
- Channel::execute is a convenience for `channel.requests().execute()`.
- Incoming::execute automatically spawns response handlers for all
requests over all channels.
3. Removal of Server.
server::Server was removed, as it provided no value over the Incoming/Channel
abstractions. Additionally, server::new was removed, since it just
returned a Server.
warning: explicit lifetimes given in parameter types where they could be elided (or replaced with `'_` if needed by type declaration)
--> tarpc/src/rpc/server/filter.rs:127:5
|
127 | fn channel<'a>(self: Pin<&'a mut Self>) -> Pin<&'a mut C> {
| ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
|
= note: `#[warn(clippy::needless_lifetimes)]` on by default
= help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#needless_lifetimes
warning: 1 warning emitted
I don't know what the intention was behind using u32::MAX + 1 but since the
argument's type is usize this is the only giant value that makes sense to me.
- it's more portable (some architectures like MIPS don't support AtomicU64)
- for most 64bit architectures usize should be 64bit as well
- for most users even 32bit would probably be enough because:
- it's tied to the connection(for streaming sockets)
- the ID wraps and by the time that happens, all previous requests would have
timed out unless you send a lot of requests and have a ton of RAM
Instead, serde_transport::tcp::connect returns a future named Connect
that has methods to directly access the framing config. This is
consistent with how serde_transport::tcp::listen returns a future with
methods to access the framing config. In addition to this consistency,
it reduces the API surface and provides a simpler user transition from
"zero config" to "some config".
