If the write buffer was filled with large bufs from the user, such that
it couldn't be fully written to the transport, the write buffer could
start to grow significantly as it moved its cursor without shifting over
the unwritten bytes.
This will now try to shift over the unwritten bytes if the next buf
wouldn't fit in the already allocated space.
Decouple preserving header case from FFI:
The feature is now supported in both the server and the client
and can be combined with the title case feature, for headers
which don't have entries in the header case map.
Closes#2313
Make it possible to refer to Connected, Connection, HttpConnector, etc.
without enabling either of the http1/http2 features. This makes feature
selection work better for downstream libraries like hyper-openssl, which
don't want to commit to any particular protocol.
Fix#2376.
Tokio's `AsyncWrite` trait once again has support for vectored writes in
Tokio 0.3.4 (see tokio-rs/tokio#3149).
This branch re-enables vectored writes in Hyper for HTTP/1. Using
vectored writes in HTTP/2 will require an upstream change in the `h2`
crate as well.
I've removed the adaptive write buffer implementation
that attempts to detect whether vectored IO is or is not available,
since the Tokio 0.3.4 `AsyncWrite` trait exposes this directly via the
`is_write_vectored` method. Now, we just ask the IO whether or not it
supports vectored writes, and configure the buffer accordingly. This
makes the implementation somewhat simpler.
This also removes `http1_writev()` methods from the builders. These are
no longer necessary, as Hyper can now determine whether or not
to use vectored writes based on `is_write_vectored`, rather than trying
to auto-detect it.
Closes#2320
BREAKING CHANGE: Removed `http1_writev` methods from `client::Builder`,
`client::conn::Builder`, `server::Builder`, and `server::conn::Builder`.
Vectored writes are now enabled based on whether the `AsyncWrite`
implementation in use supports them, rather than though adaptive
detection. To explicitly disable vectored writes, users may wrap the IO
in a newtype that implements `AsyncRead` and `AsyncWrite` and returns
`false` from its `AsyncWrite::is_write_vectored` method.
This branch updates `bytes` and `http-body` to the latest versions. The
`http-body` version that uses `bytes` 0.6 hasn't been released yet, so
we depend on it via a git dep for now. Presumably Hyper and `http-body`
will synchronize their releases.
Other than that, this is a pretty mechanical update. Should fix the
build and unblock the `h2` update to use vectored writes.
cc #2223
BREAKING CHANGE: The HTTP server code is now an optional feature. To
enable the server, add `features = ["server"]` to the dependency in
your `Cargo.toml`.
cc #2223
BREAKING CHANGE: The HTTP client of hyper is now an optional feature. To
enable the client, add `features = ["client"]` to the dependency in
your `Cargo.toml`.
cc #2251
BREAKING CHANGE: This puts all HTTP/1 methods and support behind an
`http1` cargo feature, which will not be enabled by default. To use
HTTP/1, add `features = ["http1"]` to the hyper dependency in your
`Cargo.toml`.
Previously, calling `http1_writev(true)` would just keep the default behavior, which was to auto detect if writev was optimal. Now, the auto-detection is still default, but explicitly calling `http1_writev(true)` will skip the auto-detection, and always use writev queue strategy.
Closes#2282
Adds utility functions to `hyper::body` to help asynchronously
collecting all the buffers of some `HttpBody` into one.
- `aggregate` will collect all into an `impl Buf` without copying the
contents. This is ideal if you don't need a contiguous buffer.
- `to_bytes` will copy all the data into a single contiguous `Bytes`
buffer.
The default read strategy for HTTP/1 connections is now adaptive. It
increases or decreases the size of the read buffer depending on the
number of bytes that are received in a `read` call. If a transport
continuously fills the read buffer, it will continue to grow (up to the
`max_buf_size`), allowing for reading faster. If the transport
consistently only fills a portion of the read buffer, it will be shrunk.
This doesn't provide much benefit to small requests/responses, but
benchmarks show it to be a noticeable improvement to throughput when
streaming larger bodies.
Closes#1708
- Adds `Body::on_upgrade()` that returns an `OnUpgrade` future.
- Adds `hyper::upgrade` module containing types for dealing with
upgrades.
- Adds `server::conn::Connection::with_upgrades()` method to enable
these upgrades when using lower-level API (because of a missing
`Send` bound on the transport generic).
- Client connections are automatically enabled.
- Optimizes request parsing, to make up for extra work to look for
upgrade requests.
- Returns a smaller `DecodedLength` type instead of the fatter
`Decoder`, which should also allow a couple fewer branches.
- Removes the `Decode::Ignore` wrapper enum, and instead ignoring
1xx responses is handled directly in the response parsing code.
Ref #1563Closes#1395
- When the `Body` is created from a buffer of bytes (such as
`Body::from("hello")`), we can skip some bookkeeping that is
normally required for streaming bodies.
- Orthogonally, optimize encoding body chunks when the strategy
is to flatten into the headers buf, by skipping the EncodedBuf
enum.
The `hyper::Server` is now a proper higher-level API for running HTTP
servers. There is a related `hyper::server::Builder` type, to construct
a `Server`. All other types (`Http`, `Serve`, etc) were moved into the
"lower-level" `hyper::server::conn` module.
The `Server` is a `Future` representing a listening HTTP server. Options
needed to build one are set on the `Builder`.
As `Server` is just a `Future`, it no longer owns a thread-blocking
executor, and can thus be run next to other servers, clients, or
what-have-you.
Closes#1322Closes#1263
BREAKING CHANGE: The `Server` is no longer created from `Http::bind`,
nor is it `run`. It is a `Future` that must be polled by an
`Executor`.
The `hyper::server::Http` type has move to
`hyper::server::conn::Http`.
**The `Error` is now an opaque struct**, which allows for more variants to
be added freely, and the internal representation to change without being
breaking changes.
For inspecting an `Error`, there are several `is_*` methods to check for
certain classes of errors, such as `Error::is_parse()`. The `cause` can
also be inspected, like before. This likely seems like a downgrade, but
more inspection can be added as needed!
The `Error` now knows about more states, which gives much more context
around when a certain error occurs. This is also expressed in the
description and `fmt` messages.
**Most places where a user would provide an error to hyper can now pass
any error type** (`E: Into<Box<std::error::Error>>`). This error is passed
back in relevant places, and can be useful for logging. This should make
it much clearer about what error a user should provide to hyper: any it
feels is relevant!
Closes#1128Closes#1130Closes#1431Closes#1338
BREAKING CHANGE: `Error` is no longer an enum to pattern match over, or
to construct. Code will need to be updated accordingly.
For body streams or `Service`s, inference might be unable to determine
what error type you mean to return. Starting in Rust 1.26, you could
just label that as `!` if you never return an error.
