* [http client]: re-export transport types.

Closing #605

* cargo fmt

* deps: use `FxHashMap` instead `fnv`

* fmt

* clients: introduce tls feature flag

* Update tests/tests/integration_tests.rs

* fix: don't rebuild tls connector of every connect

* fix tests + remove url dep

* fix tests again

* Bump version –> 0.5
Fix try-build tests

* Changelog

* Update CHANGELOG.md

Co-authored-by: Niklas Adolfsson <[email protected]>

* Update CHANGELOG.md

Co-authored-by: Niklas Adolfsson <[email protected]>

* Update hyper-rustls requirement from 0.22 to 0.23

Updates the requirements on [hyper-rustls](https://github.com/ctz/hyper-rustls) to permit the latest version.
- [Release notes](https://github.com/ctz/hyper-rustls/releases)
- [Commits](https://github.com/ctz/hyper-rustls/compare/v/0.22.0...v/0.23.0

)

---
updated-dependencies:
- dependency-name: hyper-rustls
  dependency-type: direct:production
...

Signed-off-by: dependabot[bot] <[email protected]>

* make it work

Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: Niklas Adolfsson <[email protected]>

Closes https://github.com/paritytech/jsonrpsee/issues/448

This PR adds an implementation for `SubscriptionClient` to the `HttpClient` struct, which makes it possible for http clients to use macro-generated RPC servers. If an http client tries to set up a subscription it will fail with a `HttpNotImplemented` error.

* request ID as RAII guard

* clippify

* fmt

* address grumbles: naming

`RequestIdGuard` -> `RequestIdManager`
`RequestId` -> `RequestIdGuard`

* ci: test each feature individually

* fix nit: --all-targets is an arg

* fix rustdoc link

* get rid of cargo hack test; too slow

* Update tokio-rustls requirement from 0.22 to 0.23

Updates the requirements on [tokio-rustls](https://github.com/tokio-rs/tls) to permit the latest version.
- [Release notes](https://github.com/tokio-rs/tls/releases)
- [Commits](https://github.com/tokio-rs/tls/commits

)

---
updated-dependencies:
- dependency-name: tokio-rustls
  dependency-type: direct:production
...

Signed-off-by: dependabot[bot] <[email protected]>

* push fixes but requires rustls-native-certs v0.6

* update native certs to 0.6.0

* fix clippy warnings

* remove webpki roots support

* Revert "remove webpki roots support"

This reverts commit 1144d567b343049ab7c967d320fc2fe162ba0f7c.

* support both native cert store and webpki

* sort deps in Cargo.toml

* Update ws-client/src/transport.rs

Co-authored-by: David <[email protected]>

Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: Niklas Adolfsson <[email protected]>
Co-authored-by: David <[email protected]>

* update CHANGELOG

* bump version number

* fix nit

* Update CHANGELOG.md

Co-authored-by: David <[email protected]>

* bullet list in CHANGELOG

* add last item to CHANGELOG

Co-authored-by: David <[email protected]>

* fix(ParamsSer): serialize to empty array

Serialize `ParamsSer::NoParams` to an empty array inorder to comply with the jsonrpc v2 spec.

* fix tests

* remove ParamsSer::NoParams

* address grumbles: adjust rpc params macro

* feat(ws client): support redirections

* reuse socket

* reuse socket

* add hacks

* fix build

* remove hacks

* fix bad merge

* address grumbles

* fix grumbles

* fix grumbles

* fix nit

* add redirection test

* Update test-utils/src/types.rs

* Resolved todo

* Check that redirected client actually works

* Rename test-utils "types" to "mocks"

* Fix windows test (?)

* fmt

* What is wrong with you windows?

* Ignore redirect test on windows

* fix bad transport errors

* debug windows tests

* update soketto

* maybe fix windows test

* add config flag for max redirections

* revert faulty change.

Relative reference must start with either `/` or `//`

* revert windows path

* use manual join paths

* remove url dep

* Update ws-client/src/tests.rs

* default max redirects 5

* remove needless clone vec

* fix bad merge

* cmon CI run

Co-authored-by: David Palm <[email protected]>

* Move RequestIdGuard to types::client

* Use the RequestIdGuard type for the http client as well

* Fix batch request ID handling (one batch uses one slot)

* Add a few tests to check that max_concurrent_requests work for http clients

* Docs

* Re-export v2::params, error, request and response types

* Use shallow type paths

* Renames

* More renames

* fmt

* s/RpcParams/Params/

* Add a type alias `SubscriptionResponse` to `Notification` which is parametrized to `SubscriptionPayload`
Move `SubscriptionResponse` to the `response` module

* Fix doc links

* Cleanup and some docs

* Review feedback

* cleanup

* Add license headers where missing and update year

* Update examples/http.rs

Co-authored-by: Niklas Adolfsson <[email protected]>

Co-authored-by: Niklas Adolfsson <[email protected]>

* rewrite me

* require proc macro API to return result

* send unknown message when error is not CallError

* show example that auto cast StdError doesn't work

* register_*_method Into<Error>

* clippy

* replace generic errors with anyhow::Error

* fix nits

* example that anyhow::Error in register_method works

* CallError: add missing From impl

* [types]: add helper methods for Error types

The rationale is to make it possible for users to either use anyhow::Error or use the helper methods.

* fmt

* Revert "register_*_method Into<Error>"

This reverts commit 33b4fa28730b72647ba150659d3c0ab1937e524a.

* add better comment

* fix nit

* Sniff the first byte to glean if the incoming request is a single or batch request

This works around the serde limitations around `untagged` enums and `RawValue`.

* fmt

* For http server, check first byte before allocating space for the body

Also, rework the way we return errors: prefer JSON-RPC errors according to spec (application/json) wherever sensible.

* Review feedback

* Don't assume there is a first byte to read

* ty clipyp

* Review concerns

* Cleanup

Co-authored-by: David Palm <[email protected]>

To make sure that we are not affected by:
	- https://github.com/hyperium/hyper/security/advisories/GHSA-f3pg-qwvg-p99c
	- https://github.com/hyperium/hyper/security/advisories/GHSA-5h46-h7hh-c6x9

However, `hyper 13` is not fixed but I think we should remove that support anyway?!

* Bump versions to 0.3.0

* Add draft release checklist

* Changelog for v0.3.0

* Mention possibility of passing params to `publish.sh`

* Added PR 412

* Include "macros" in the "server" feature

* Re-export SubscriptionSink

* Include the "types" feature in both the "client" and "server" features
Export types::* from façade when the "types" is active
Export types::* from servers

* fmt

* Export jsonrpsee_types under types

* fmt

* broken doc link

* fix benches

* fmt

* Fix test imports

* [clients]: use request timeout by-default

* add timeout for notif

* more feature flag mess

* rexport tokio types

* Update ws-client/src/client.rs

Co-authored-by: Maciej Hirsz <[email protected]>

* Impose a timeout on all requests

Variant of #367

This PR takes a more opinionated stance than #367, where timeouts are optional. In this PR I suggest we make a all requests use a timeout and only let users choose the length.

* fmt

* Address review grumbles

* fmt

* Use tokio::select! for cleaner code

Co-authored-by: Niklas Adolfsson <[email protected]>
Co-authored-by: Maciej Hirsz <[email protected]>
Co-authored-by: Maciej Hirsz <[email protected]>

* Release prep

* Mention proc macro limitations

* Mention publish script

This PR requires the internal dependencies to require the exact version
because `cargo` will otherwise fetch the latest version.

This could also be achievied by `cargo update -p package --precise x.y.z`.

Should be reverted once we release 0.2

* Add extension trait to timeout the futures execution

* Use timeout for ws-server tests

* Use timeout for ws-client tests

* Use timeout for http-server tests

* Use timeout for http-client tests

* Make TimeoutFutureExt interafaces more explicit

* get started

* add additional test

* fix nits

* cargo fmt

* [types]: write some tests.

* [http server]: send empty response on notifs

* [http server]: fix tests

* [rpc module]: send subscription response

* Update types/src/v2/error.rs

* fix nits

* cargo fmt

* Update types/src/v2/params.rs

* remove needless clone

* remove dead code

* [types]: impl PartialEq for JsonErrorObject + test

* use beef::Cow

* Update http-server/src/tests.rs

* ret err if context/params fails

* address grumbles: specific error_code context fail

* address grumbles: make env_logger dev-dependency

* address grumbles: add tests

* chore(deps): remove unused deps

* address grumbles: rename types and docs

* address grumbles: more renaming.

* fix build

* [types]: allow `data` field in jsonrpc error obj.

* fmt

* revert change to make fail

* remove redundant test

* [types]: bring back `data` and `message` in error.

Basically deducing the error message from server defined error doesn't work
and we should not strip out this information from to user which might be useful.

* remove boiler plate code

* [types]: impl From<ErrorCode> for JsonRpcError

* address grumbles

* remove more boiler plate

I went over all the atomic operations and realised I could not convince myself that all uses of `Relaxed` were ok, so better safe than sorry: let's switch to `SeqCst`.