Welcome! Thanks for looking into contributing to our project!
Here is a list of helpful resources you can consult:
- Ruma Developer Matrix room: #ruma-dev:matrix.org
- Matrix Developer room: #matrix-dev:matrix.org
If you find any bugs, inconsistencies or other problems, feel free to submit a GitHub issue.
If you have a quick question, it may be easier to leave a message in #ruma:matrix.org.
Also, if you have trouble getting on board, let us know so we can help future contributors to the project overcome that hurdle too.
Ready to write some code? Great! Here are some guidelines to follow to help you on your way:
In general, try to replicate the coding style that is already present. Specifically:
When writing endpoint definitions, use the following mapping from request / response field types listed in the specification to Rust types:
Specification type | Rust type |
---|---|
boolean |
bool |
integer |
js_int::UInt (unless denoted as signed, then js_int::Int ) |
string |
If for an identifier (e.g. user ID, room ID), use one of the types from ruma-identifiers . Otherwise, use String . |
object |
serde_json::Value |
[…] |
Vec<…> |
{string: …} |
BTreeMap<String, …> (or BTreeMap<SomeId, …> ) |
Generally, all struct
s that are mirroring types defined in the Matrix specification should have
all their fields pub
lic. Where there are restrictions to the fields value beyond their type, these
should generally be implemented by creating or using a more constrained type than the spec uses for
that field – for example, we have a number of identifier types but the Matrix spec uses string
for
fields that hold user IDs / room IDs and so on.
Almost all types in ruma-events
and the API crates use the #[non_exhaustive]
attribute, to allow
us to adapt to new minor releases of the Matrix specification without having a major release of our
crates. You can generally just apply #[non_exhaustive]
to everything – it's a backwards compatible
change to remove it in the rare case it is not warranted.
Due to this combination of public fields and non-exhaustiveness, all struct
s generally need a
constructor function or From
/ TryFrom
implementation to be able to create them in a
straight-forward way (always going through Deserialize
would be quite ugly).
Organize your imports into three groups separated by blank lines:
std
imports- External imports (from other crates)
- Local imports (
self::
,super::
,crate::
and things likeLocalEnum::*
)
For example,
use std::collections::BTreeMap;
use ruma_api::ruma_api;
use super::MyType;
Also, group imports by module. For example, do this:
use std::{
collections::BTreeMap,
convert::TryFrom,
fmt::{self, Debug, Display, Formatter},
};
as opposed to:
use std::collections::BTreeMap;
use std::convert::TryFrom;
use std::fmt::{self, Debug, Display, Formatter};
When importing methods and types from serde_json
, methods should be such as
serde_json::{from,to}_{slice,string,value,vec}
should be imported as
{from,to}_json_{slice,string,value,vec}
.
For example:
use serde_json::{
from_value as from_json_value,
to_str as to_json_str,
};
Also, serde_json::Value
should be imported as JsonValue
.
Use rustfmt
to format your code and clippy
to lint your code¹. Before
committing your changes, go ahead and run cargo fmt
and cargo clippy
² on the
repository to make sure that the formatting and linting checks pass in CI. Note
that clippy
warnings are reported as errors in CI builds, so make sure to
handle those before comitting as well.
¹ To install the tools, run rustup component add rustfmt clippy
.
² If you modified feature-gated code (#[cfg(feature = "something")]
), you will
have to pass --all-features
or --features something
to clippy for it to
check that code.
Write commit messages using the imperative mood, as if completing the sentence: "If applied, this commit will ___." For example, use "Fix some bug" instead of "Fixed some bug" or "Add a feature" instead of "Added a feature".
(Take a look at this blog post for more information on writing good commit messages.)
Use the latest r0.x.x documentation when adding or modifying code. We target the latest minor version of the Matrix specification. (Note: We might reconsider this when the Client-Server API hits r1.0.0.)
Add a comment to the top of each endpoint file that includes the path and a link to the documentation of the spec. You can use the latest version at the time of the commit. For example:
//! [GET /.well-known/matrix/client](https://matrix.org/docs/spec/client_server/r0.4.0#get-well-known-matrix-client)
When adding new endpoints, select the module that fits the purpose of the endpoint. When naming the endpoint itself, you can use the following guidelines:
- The name should be a verb describing what the client is requesting, e.g.
get_some_resource
. - Endpoints which are basic CRUD operations should use the prefixes
create
,get
,update
, anddelete
. - The prefix
set
is preferred to create if the resource is a singleton. In other words, when there's no distinction betweencreate
andupdate
. - Try to use names that are as descriptive as possible and distinct from
other endpoints in all other modules. (For example, instead of
r0::room::get_event
, user0::room::get_room_event
). - If you're not sure what to name it, pick any name and we can help you with it.
In order to reduce the number of .clone()
s necessary to send requests
to a server, we support borrowed types for request types. (We hope to support
borrowing in response types sometime, but that is dependent on GATs.)
Field type | Borrowed type | Notes |
---|---|---|
strings | &'a str |
|
identifiers | &'a IdentifierType |
|
Vec<_> |
&'a [_] |
The inner type should not be borrowed. |
Inner types of Vec
s should not be borrowed, nor should BTreeMap
s and such.
Structs also should not be borrowed, with the exception that if a struct:
- has fields that should be borrowed according to the table
above (strings, identifiers,
Vec
s), and - is only used inside request blocks (i.e. not in response blocks or in events),
then the struct should be lifetime-parameterized and apply the same rules to their fields. So instead of
ruma_api! {
request: {
my_field: MyStruct,
}
// ...
}
pub struct MyStruct {
inner_string_field: String,
}
use
ruma_api! {
request: {
my_field: MyStruct<'a>
}
// ...
}
pub struct MyStruct<'a> {
inner_string_field: &'a str,
}
If your changes affect the API of a user-facing crate (all except the -macros
crates and
ruma-identifiers-validation
), add an entry about them to the change log (CHANGELOG.md
)
of that crate. Where applicable, try to find and denote the version of the spec that
included the change you are making.
Once you're ready to submit your code, create a pull request, and one of our maintainers will review it. Once your PR has passed review, a maintainer will merge the request and you're done! 🎉
If this is your first contribution to the project, we recommend taking a look at one of the open issues we've marked for new contributors.
Before committing, run cargo check
to make sure that your changes can build, as well as running the formatting and linting tools mentioned above.
Thanks again for being a contributor! If you have any questions, join us at #ruma:matrix.org.