Torba is a Bower-less asset manager for Sprockets. It makes a local copy of a JS/CSS library and puts it under Sprockets' load path.
"Торба" [tǒːrba] in Ukrainian and "torba" in Polish, Turkic languages can mean "duffel bag", "gunny sack" or, more generally, any flexible container.
Production ready.
De facto approach, i.e. wrapping JS and CSS libraries in a gem, requires a maintainer to constantly track changes in an upstream repository. Even more so, if a gem maintainer stops using that specific library, the gem will eventually become abandoned. Additionally, many libraries still have no gem wrappers.
Other alternatives:
- rails-assets relies on Bower and it is quite complex,
- bower-rails relies on Bower, see below for why this can be an issue.
Problems with the Bower:
- it is not a part of the Ruby ecosystem,
- frontend JS libraries are usually standalone (except for a potential jQuery dependency), so there's no need for a complex Bundler-like solution with tree-dependency resolution,
- often we can't use optimistic version constraints, because the JavaScript community does not consistenly apply the principles of Semver. By specifying strict versions we use Bower as a complex facade for functionality that could be accomplished with curl.
- curl
- unzip
- gzip
- tar
- Torba doesn't do any version dependency resolution, it's up to you to specify the correct version of each asset package,
- Torba doesn't do any builds, you should use remote sources with pre-built assets.
Use torba-rails.
See this example project.
Add this line to your application's Gemfile and run bundle:
gem 'torba'
-
Create Torbafile at the project root and commit it.
-
Run
bundle exec torba pack
. -
Add "require" Sprockets directives to your "application.js" and/or "@import" Sass directives to "application.css".
If any changes made to the Torbafile, run bundle exec torba pack
again.
Torbafile is an assets specification. It is a plain text file that contains one or more sections, each of them describes one remote source of assets.
Currently only zip, tar.gz archives, Github releases and npm packages are supported.
Allows to download and unpack asset package from any source accessible by curl.
The syntax is:
zip "name", url: "..." [, import: %w(...)]
where "name" is an arbitrary name for the package, more on "import" below. For example,
zip "scroll_magic", url: "https://github.com/janpaepke/ScrollMagic/archive/v2.0.0.zip"
The syntax is same as for a zip package:
targz "name", url: "..." [, import: %w(...)]
for example,
targz "scroll_magic", url: "https://github.com/janpaepke/ScrollMagic/archive/v2.0.0.tar.gz"
This is a more readable version/shortcut for "https://github.com/.../archive/..." URLs.
The syntax is:
gh_release "name", source: "...", tag: "..." [, import: %w(...)]
where "source" is the user + repository and "tag" is the repository tag (exactly as on Github, i.e. with "v" prefix if present), more on "import" below. For example,
gh_release "scroll_magic", source: "janpaepke/ScrollMagic", tag: "v.2.0.0"
You can omit the name, it will be equal to the repository name:
gh_release source: "janpaepke/ScrollMagic", tag: "v.2.0.0" # "ScrollMagic" is assumed
Allows to download packages from npm registry.
The syntax is:
npm "name", package: "...", version: "..." [, import: %w(...)]
where "package" is the package name as published on npm registry and "version" is its version, more on "import" below. For example,
npm "coffee", package: "coffee-script", version: "1.9.2"
You can omit the name, it will be equal to the package name:
npm package: "coffee-script", version: "1.9.2"
See Torbafiles used for testing.
When you run torba pack
the following happens:
-
All remote sources are cached locally.
-
Archives are unpacked with top level directory removed. This is done for good because it usually contains the package version in the name, e.g. "react-0.13.2", and you don't want to have to reference versions inside your application code (except Torbafile).
-
Remote source's content is copied as is to the
Torba.home_path
location with package name used as a namespace.This is also done for good reason in order to avoid name collisions (since many JS projects can have assets with the same names and all packages are placed into Sprockets' shared virtual filesystem). The downside is that you have to use namespaces in each require directive, which can lead to duplication:
// application.js //= require 'underscore/underscore'
Hint: use "require_directory" if you're strongly against such duplication:
//= require_directory 'underscore'
-
Stylesheets (if any) are converted to ".css.erb" with "asset_path" helpers used in "url(...)" statements.
Copying whole remote source's content has the disadvantage of using remote source specific paths in your require/import directives. For example, if an archive contains files in the "dist/css" directory, you'll have to mention it:
/* application.css */
@import 'lightslider/dist/css/lightslider';
To mitigate this you can cherry-pick files from the source via the "import" option, for example:
gh_release "lightslider", source: "sachinchoolur/lightslider", tag: "1.1.2", import: %w[
dist/css/lightslider.css
]
Such files will be copied directly to the package root (i.e. file tree becomes flatten), thus you can omit unnecessary paths:
@import 'lightslider/lightslider';
You can use any Dir.glob pattern:
gh_release "lightslider", source: "sachinchoolur/lightslider", tag: "1.1.2", import: %w[
dist/css/lightslider.css
dist/img/*.png
]
In addition to this "path/" is treated as a shortcut for "path/**/*" glob pattern.