Most of our preferred style when writing typescript is configured in our .eslintrc.yml
files.
- Use camelCase for methods
- Use PascalCase for class names
- ESLint in your editor
We currently use JSDoc even though we don't currently generate any documentation or verify the format. We're using JSDoc over other formats because the typescript compiler has built-in support for parsing jsdoc and presenting it in IDEs.
While there doesn't appear to be any well-used typescript documentation export utilities out there at the moment we hope that it's only a matter of time. JSDoc uses a lot of metadata that is already self-documented in the typescript type system such as visibility, inheritance, membership.
For now all you need to know is that you can document classes, methods, properties and fields by using the following formatted comment on the line above whatever you're trying to document.
/** This is a documentation string */
The double start /**
opener is the key. It has to be exactly two stars for it to be a valid JSDoc open token.
If you need multiple lines to describe the subject try to sum up the thing you're describing in a short title and leave a blank line before you go into detail, similar to a git commit message.
/**
* This is a title, keep it short and sweet
*
* Go nuts with documentation here and in more paragraphs if you need to.
*/
The Dispatcher
is the entry point for most interactions with the application which update state,
and for most usages this work is then delegated to the AppStore
.
Due to this coupling, we need to discourage callers directly manipulating
specific methods in the AppStore
unless there's a compelling reason.
We do this by making the methods look unappealing:
- underscore prefix on method name
- comment indicating that you should be looking elsewhere
/** This shouldn't be called directly. See `Dispatcher`. */
public async _repositoryWithRefreshedGitHubRepository(repository: Repository): Promise<Repository> {
// ...
}
We should be using asynchronous core APIs throughout the application, unless
there's a compelling reason and no asynchronous alternative. In those cases the
method should be suffixed with Sync
to make it clear to the caller what's
happening. We also fall back to Sync
methods for readability in tests.
We use an ESLint rule to enforce this standard.
For scripts we should favor synchronous APIs as the asynchronous benefits are not so important there, and it makes the code easier to read.