css-trimmer
v0.1.0
Published
`css-trimmer` identifies the unused properties in your CSS.
Downloads
2
Readme
css-trimmer
css-trimmer
identifies the unused properties in your CSS.
Wait, doesn't Chrome DevTools do that already?
Yes, kinda. Here are two reasons for this tool:
The Coverage panel in Chrome DevTools will consider any rule (.some-rule { ... }
) as "used" if
any node on the page uses the selectors. It doesn't catch if a property within a rule is always
superceded by some other rule. In other words, a rule is either all-covered or all-uncovered.
For example, the property in div {}
should be considered uncovered, but Chrome DevTools doesn't
know.
css-trimmer
does:
=========
file:///Users/cjamcl/src/me/css-trimmer/test/fixtures/1/index.html (3 unused declarations)
=========
:5 div { color: red; /* unused! */ }
:13 .never-used { color: blue; /* unused! */ }
:14 .never-used { display: block; /* unused! */ }
-----
total unused declarations: 3
Additionally, this tool checks style coverage by considering many variants of a page, providing general strategies (try this viewport, try this color scheme ...), and also provides a way to configure your own variants.
Usage
Package is available on npm as css-trimmer
.
css-trimmer [urlsOrFiles] <options>
css-trimmer path/to/page.html path/to/page2.html --viewports 500,500
css-trimmer https://www.example.com
Options:
--help Show help [boolean]
--version Show version number [boolean]
--color-scheme Add a collection while emulating
`prefers-color-scheme` to the value given [string]
--config-path Loads these options from disk. Supports .js and
.json [string]
--debug Generate debug data, such as which collections were
redundant [boolean]
--disable-default-config Disable loading of the default configuration
[boolean]
--only-collections Collections to process - all others are skipped
[array]
--output Format for the generated the report
[choices: "json", "text"] [default: "text"]
--quiet Suppress logging output to stderr
[boolean] [default: false]
--skip-collections Collections to skip - all others are processed
[array]
--viewports viewports "width,height". Adds a collection while
emulate each viewport. [array]
*.js
configs can define afterNavigation(context)
, which provides access to the puppeteer
page handler. It runs before any collection for a URL.
Besides the CLI, there is the Node API.
const cssTrimmer('css-trimmer');
const context = await cssTrimmer.start(); // see src/css-trimmer.js for context type definition
await context.navigate('https://www.example.com');
await context.collect('https://www.example.com basic');
const report = context.finish(context);
// Or, use the same runner that the CLI uses:
const report = await cssTrimmer.run(urls, options);
How it works
Uses Puppeteer. Loads each URL and collects the active style declarations for every relevant node in the DOM. The range within the style sheet for each applied property is recorded. Every mass-reading of the DOM is called a collection. After all the URLs and each variant of collections runs, the recorded property sets are combined. Every property not in this set is flagged.
Every collection takes a few seconds, and with all the URLs and variants the program can take painfully long.
Running with --debug
will determine which of the collections are redundant, and --only-collection
will allow
you to skip the rest.
Future work
- Flesh out
afterNavigation
, maybe provide more lifecycle methods / easy way to add more variants. Using againstlighthouse
report should help flesh this out - Variables
- Upstream property-level coverage to Chrome DevTools
- Output text format doesn't work with minified CSS