The HttpConnector's connect future was lazy, but if any custom connector
did not use a lazy future, then a connect would always be started, even
if an idle connection was available.
If a request sees an error on a pooled connection before ever writing
any bytes, it will now retry with a new connection.
This can be configured with `Config::retry_canceled_requests(bool)`.
Currently, if the remote closes the connection at the same time that the
pool selects it to use for a new request, the connection may actually
hang. This fix will now more allow the keep-alive read to check the
socket even when the `Conn` think it's busy.
If the connection was closed before the request write happened, returns
back an `Error::Cancel`, letting the user know they could safely retry
it.
Closes#1439
Setting this to false will force HTTP/1 connections to always flatten
all buffers (headers and body) before writing to the transport. The
default is true.
This allows using a future `Executor` other than a `Handle` to execute
the background (connection) tasks needed for sending requests and
responses.
This also deprecates `Client::handle()`, since the executor may not be
a `Handle`.
- Deprecates the `no_proto` configuration on `Server`. It is always
enabled.
- Deprecates all pieces related to tokio-proto.
- Makes the tokio-proto crate optional, and the `server-proto` feature
can be used to completely remove the dependency. It is enabled by
default.
Additionally fixes if there were idle connections when a `Client` is
dropped.
Only fixes with the no-proto dispatcher, as changing internals for the
tokio-proto dispatcher would be much harder, and it will replace it very
soon.
Closes#1397
For now, this adds `client::Config::no_proto`, `server::Http::no_proto`,
and `server::Server::no_proto` to skip tokio-proto implementations, and
use an internal dispatch system instead.
`Http::no_proto` is similar to `Http::bind_connection`, but returns a
`Connection` that is a `Future` to drive HTTP with the provided service.
Any errors prior to parsing a request, and after delivering a response
(but before flush the response body) will be returned from this future.
See #1342 for more.
If a `Request`'s version is `Http09`, `H2`, or `H2c`, `client.request`
will return a `hyper::Error::Version`, and a message is logged at
`error!` level.
Closes#1283
Remove requirement when calling client::Config::connector() that the connector implements Connect.
Removing this requirement allows users to set the connector back to UseDefaultConnector. Previously,
this was not possible.
Request and Response are now visible from:
- hyper::{Request, Response}
- hyper::server::{Request, Response}
- hyper::client::{Request, Response}
They truly exist in the http module, but are re-exported to reduce the number of breaking changes.
request::new and response::new were renamed to ::from_wire to reduce confusion with Request::new
and Response::new. See issue #1126
Request now has an optional Body, because not all requests have bodies.
Use body_ref() to determine if a body exists.
Use body() to take the body, or construct one if no body exists.
Closes#1155
BREAKING CHANGE: Response::body() now consumes the response
BREAKING CHANGE: The `Url` type is no longer used. Any instance in the
`Client` API has had it replaced with `hyper::Uri`.
This also means `Error::Uri` has changed types to
`hyper::error::UriError`.
The type `hyper::header::parsing::HTTP_VALUE` has been made private,
as an implementation detail. The function `http_percent_encoding`
should be used instead.
This commit updates to the most recent versions (released today) of the various
Tokio libraries in use. Namely the `tokio_core::io` module has now been
deprecated in favor of an external `tokio-io` crate. This commit pulls in that
crate and uses the `AsyncRead + AsyncWrite` abstraction instead of `Io` from
tokio-core.
BREAKING CHANGE: Any external types that were using that had implemented `Io` will need to
implement `AsyncRead + AsyncWrite` from tokio_io.
This allows us to improve the performance. For now, a Cow is used
internally, so clients can set the host to a static value and no longer
need copies.
Later, we can change it to also possibly have a MemSlice.
BREAKING CHANGE: The fields of the `Host` header are no longer
available. Use the getter methods instead.
There are many changes involved with this, but let's just talk about
user-facing changes.
- Creating a `Client` and `Server` now needs a Tokio `Core` event loop
to attach to.
- `Request` and `Response` both no longer implement the
`std::io::{Read,Write}` traits, but instead represent their bodies as a
`futures::Stream` of items, where each item is a `Chunk`.
- The `Client.request` method now takes a `Request`, instead of being
used as a builder, and returns a `Future` that resolves to `Response`.
- The `Handler` trait for servers is no more, and instead the Tokio
`Service` trait is used. This allows interoperability with generic
middleware.
BREAKING CHANGE: A big sweeping set of breaking changes.
The previous keep-alive strategy was to cycle connections in a
round-robin style. However, that will always keep more connections
around than are needed. This new strategy will allow extra connections
to expire when only a few are needed. This is accomplished by prefering
to reuse a connection that was just released to the pool over one that
has been there for a long time.
We've been seeing a strange number of timeouts in our benchmarking.
Handling spurious timeouts as in this patch seems to fix it!
Note that managing the `timeout_start` needs to be done carefully. If
the current time is provided in the wrong place, it's possible requests
would never timeout.
I've had a couple of instances during stress testing now where
Conn::ready would overflow its stack due to recursing on itself. This
moves subsequent calls to ready() into a loop outside the function.
When loading up a client suddenly with thousands of connections, the
default DNS worker count of four cannot keep up and many requests
timeout as a result. Most people don't need a large pool, so making this
configurable is a natural choice.
In the scenario where a request is started on the `Client`, the client
has a full slab, and sockets for a *different* domain are idling in
keep-alive, the new request would previously cause the client to panic!.
This patch adds a `spawn_error` handler which attempts to evict an idle
connection to make space for the new request. If space cannot be made,
the error handler is run (passed `Error::Full`) and the `Handler` is
dropped.
This is a breaking change because of the new variant of `Error`.
Some inefficient use of `Vec` in the client was replaced with `VecDeque`
to support push/pop from either end.
Closes#896Closes#897
BREAKING CHANGE: `RequestUri::AbsolutePath` variant is changed to a struct variant. Consider using `req.path()` or `req.query()` to get the relevant slice.
