-
Notifications
You must be signed in to change notification settings - Fork 183
Extension package format
Status: Implemented
Note: this will likely undergo revision during the extension API research
Brackets extensions follow a format established by CommonJS and npm.
Package files are zip files, with a package.json file in the top-level directory.
npm uses gzipped tar files, but the standard for Brackets extensions is zip files because:
- they are more convenient for users to create on Windows and Mac
- they can be created with a single click from GitHub
If more compatibility with npm proves to be desirable, we can always repack the extensions into gzipped tar files.
Brackets uses the same package.json format as npm. There is a sample package.json in the section below. Many of the fields that make sense for Node are ignored in Brackets (such as scripts and man pages).
-
name
andversion
are required as in Node - Brackets adds an optional
title
that is the pretty display form ofname
-
description
is required, but will optionally be pulled from a README.txt or README.md file at the top of the package (note that pulling from README files is not currently implemented). -
homepage
can be a URL that will provide the user with more information (and a "More Info" link is provided in the Extension Manager for this purpose) - the
keywords
field will help users find the extensions they need (see below) -
files
works the same as it does for npm, including the optional use of an.npmignore
file - Brackets does not support any of npm's dependencies-related features on installation. When we implement extension dependencies, those will likely be expressed as peerDependencies
-
engines
is used to specify the compatible Brackets versions. For most extensions, the version should be expressed as>= (some Brackets version)
(in other words, there's no max version). - We should honor the
private
flag to not publish private extensions
note by reusing engines
and peerDependencies
, we may run into friction with npm, assuming that we allow users to install and use npm modules. If need be, we can change to using two other Brackets-specific fields in package.json for these needs.
name
and version
are required. title
, description
and license
are also recommended.
{
"name": "unique-lowercase-identifier",
"title": "Pretty Title for the UI",
"description": "More useful details to let someone know what your extension is all about.",
"homepage": "https://github.com/yourname/yourproject",
"version": "1.0.0",
"author": "Your Name <your@email> (http://your.url)",
"license": "MIT",
"engines": {
"brackets": ">=0.24.0"
}
}
Our package.json files include a "categories" field which is not present in npm's format. Categories, unlike keywords, are primarily a "browsing" feature rather than a searching one.
The categories
field can be either a string (the common case, since most extensions have only one), or an array of strings.
Each of the strings should be one of the following boldface values:
- editing – Editing and Navigation
- snippets – Snippets and Shorthand
- formatting – Formatting
- codegen – Code Generation
- language – Language Support
- general – General Functionality
- livedev – Live Development
- visual – Visual Editing
- external – External Tools and Online Content
- docs – Documentation
- linting – Linting and Warnings
- testing – Testing and Code Metrics
Keywords can be anything that will help users find the right extension. Some conventions will help Brackets steer users to the right extensions quicker:
- Relevant file extensions including a leading ".". For example, if your extension is specifically for JavaScript files, include a ".js" keyword.
Additional keyword conventions will be created as we review the categories of extensions.
Subject to change from the Extension API research, Brackets will automatically execute main.js
at the root of the package when the extension is loaded.
Modules that are imported via require
are searched for from the root of the extension. For example, require("foo/bar")
will look for foo/bar.js
in the extension.
Brackets extensions will be able to provide modules that run in a Node process. The exact location of these modules will depend on the answers to the open questions at the end of this section.
Brackets extensions will also be able to use modules downloaded via npm. The node_modules
directory will be incorporated directly into the extension package. Including dependencies directly in a package for deployment is considered a best practice.
Node modules (or any modules for that matter) are not shared between extensions.
There are open questions surrounding the lifecycle of the Node code and the module format differences between Brackets and Node and these will be resolved as we head toward the Extension API research story.
If there is a CHANGELOG.md file at the package root, the h2 headers (delimited by ##) are assumed to contain version numbers and the sections of the file will be pulled apart to display update information to users.
A unit test module can be located at unittests.js
at the root of the package.
Open question: npm uses CommonJS modules, Brackets uses AMD, how do we reconcile?
(This will be investigated during the extension API research. We have discussed a move to pure CommonJS modules.)
The pattern for localizing extensions is to put an nls
directory at the root, containing a strings.js
module. Next to that module, there is a root
directory with the baseline set of strings and a directory for each locale's specific strings. See the Localization Example extension.