Entwicklerhandbuch

Wie man Plume auf Ihrem Computer installiert und Änderungen am Quellcode vornehmen kann. Diese Anleitung gibt Ihnen auch Tipps, um Debugging und Tests einfacher zu machen.

Plume is mostly made in Rust. The back-end uses Rocket and Diesel. The front-end is in Rust too, thanks to WebAssembly. The stylesheets are written in SCSS.

If you want to write some code but you don’t really know where to start, you can try to find an issue that interests you (GitHub).

Then, fork Plume (if you didn’t already do it), git clone your fork, and start a new branch with git checkout -b NAME-OF-THE-BRANCH. You can now start to work on the issue.

Once you have something working, do git add FILES THAT YOU CHANGED (or git add . to add them all), and then git commit. Write a message explaining your changes, and do git push origin NAME-OF-THE-BRANCH to upload your work to Gitea or GitHub. Open the URL that appears in the output of this last command to open a pull request, that we will then review, and eventually merge.

Installing the development environment

Please refer to the installation guide. Choose to compile Plume from source when asked. Instead of using cargo install, use cargo run which starts a freshly compiled debugging version of Plume.

It is recommended to enable the debug-mailer feature, especially if you need emails during development. You can do it by passing the --feature debug-mailer flags to cargo. When enabled, mails will be logged to the standard output instead of being sent for real.

Migrations

Migrations are files than can be used to update the database schema (for instance to add a new field to a model). To create new migrations you will need a tool called diesel, that can be installed with:

cargo install diesel_cli --no-default-features --features DATABASE --version '=1.3.0'

After that to create a migration, both for PostgreSQL and SQlite, you need to run these two commands:

MIGRATION_DIRECTORY=migrations/postgres diesel migration generate NAME
MIGRATION_DIRECTORY=migrations/sqlite diesel migration generate NAME

Where NAME is the name you want to give to your migration, in one “word”, for instance add_role_to_users. New files will be generated in migrations/postgres and migrations/sqlite, called up.sql and down.sql. The former should run the actual migration, and the later undo it.

You can also run some Rust code in migrations, by writing it in comments starting with #!, and wrapped in a closure taking a database connection and a path to the current directory. You can access the plume-models modules with the super module. Here is an example:

--#!|conn: &Connection, path: &Path| {
--#!    println!("Running a migration from {}", path);
--#!    println!("The admin of this instance is @{}", Instance::get_local(conn).unwrap().main_admin(conn).unwrap().name());
--#!    Ok(())
--#!}

If your function is too long, you can also put it in plume-models, and simply give it’s full identifier in the comment:

--#! crate::migrations::functions::my_migration_function

To run migrations, you can use plm migration run. To cancel and re-run them, use plm migration redo.

Testing the federation

To test the federation, you’ll need to setup another database, also owned by the “plume” user, but with a different name. Then, you’ll need to setup this instance too.

The easiest way to do it is probably to install plm and plume globally (as explained here), but with the --debug flag to avoid long compilation times. Then create a copy of your .env file in another directory, and change the DATABASE_URL and ROCKET_PORT variables. Then copy the migration files in this new directory and run them.

plm migration run

Setup the new instance with plm as explained here.

Now, all you need for your two instances to be able to communicate is a fake domain name with HTTPS for each of them. The first step to have that on your local machine is to edit your /etc/hosts file, to create two new aliases by adding the following lines.

127.0.0.1       plume01.localhost
127.0.0.1       plume02.localhost

Now, we need to create SSL certificates for each of these domains. We will use mkcert for this purpose. Here are the instructions to install it. Once you installed it, run.

mkcert -install
mkcert plume01.localhost
mkcert plume02.localhost

Finally, we need a reverse proxy to load these certificates and redirect to the correct Plume instance for each domain. We will use Caddy here as it is really simple to configure, but if you are more at ease with something else you can also use alternatives.

To install Caddy, please refer to their website. Then create a file called Caddyfile in the same directory you ran mkcert and write this inside.

plume01.localhost {
  reverse_proxy localhost:7878
  tls plume01.localhost.pem plume01.localhost-key.pem
}

plume02.localhost {
  reverse_proxy 127.0.0.1:8787
  tls plume02.localhost.pem plume02.localhost-key.pem
}

Eventually replace the ports in the proxy blocks by the one of your two instances, and then run caddy. You can now open your browser and load https://plume01.localhost and https://plume02.localhost.

Running tests

To run tests, use DATABASE_URL for tests and run migration at first.

To run tests of plume-models use RUST_TEST_THREADS=1, otherwise tests are run concurrently, which causes error because they all use the same database.

Internationalization

To mark a string as translatable wrap it in the i18n! macro. The first argument should be the catalog to load translations from (usually ctx.1 in templates), the second the string to translate. You can specify format arguments after a ;.

If your string vary depending on the number of elements, provide the plural version as the third arguments, and the number of element as the first format argument.

You can find example uses of this macro here.

There is no need to provide individual translations of i18n!-wrapped strings in pull requests. The strings will be uploaded to a third-party web service and translated automatically as a separate step.

Adding a language

Managers of the Crowdin project may add new languages to the list of target languages. But for these new languages to be visible in the interface as well, some changes need to be made to the code.

First of all, make sure that the corresponding .po file is present in po/plume, and in po/plume-front. If it is not, ask someone who has access to the Crowdin API key (like a Crowdin manager or owner) to pull it with the Crowdin CLI tool.

To find a manager, look at the list in the sidebar of this page. These people are managers because they were involved enough in the project, and especially in the translation process. If you want to become a manager on Crowdin too, ask us on Matrix.

Then, in src/main.rs and plume-front/src/main.rs, add the code corresponding to the language in the init_i18n! macro. If you don’t know the code for the language: it is the name of the .po file, without the extension. Here is an exemple of a Pull Request adding Persian.

cargo check to make sure everything compiles, and you are good to go!

Working with the front-end

When working with the front-end, we try limit our use of JavaScript as much as possible. Actually, we are not even using JavaScript since our front-end also uses Rust thanks to WebAssembly. But we want Plume to work with as little JavaScript as possible, since reading a post (Plume’s first goal) shouldn’t require a lot of interactions with the page.

When editing SCSS files, it is good to know that they are compiled by cargo too. But cargo can be a bit slow, since it recompiles all of Plume every time, not only the SCSS files. A workaround is to run cargo run in the background, and use cargo build to compile your SCSS, then kill it before the end of the build. To know when your SCSS have been compiled, wait for cargo to tell you it is compiling plume(bin) and not plume(build) (next to the progress bar).

Also, templates are using the Ructe syntax, which is a mix of Rust and HTML. They are compiled to Rust and embedded in Plume, which means you have to re-run cargo everytime you make a change to the templates.

Code Style

For Rust, use the standard style. rustfmt can help you keeping your code clean.

For SCSS, the only rules are to use One True Brace Style and two spaces to indent code.

For JavaScript, we use the JavaScript Standard Style.

For HTML/Ructe templates, we use HTML5 syntax.