v1.4.8

.github/workflows/coverage.yml

@@ -7,6 +7,8 @@ jobs:
     steps:
       # Your original steps
       - uses: actions/checkout@v6
+        with:
+          persist-credentials: false
       - uses: actions/setup-node@v6
       - name: Install
         run: npm install

CHANGELOG.md

@@ -1,5 +1,11 @@
 # Changelog
 
+## v1.4.8
+
+### Improvements
+
+- [x] support display-p3-linear color space. 
+
 ## v1.4.7
 
 ### Improvements

benchmark/README.md

@@ -0,0 +1,54 @@
+# CSS minification benchmark (Vitest)
+
+using [Vitest's built-in `bench()`](https://vitest.dev/guide/features.html#benchmarking)
+(powered by tinybench) for timing, plus a separate deterministic pass for
+output size, merged into a self-contained dark-themed HTML report.
+
+Same library set as the official page: `clean-css`, `cssnano`, `csso`,
+`css-tree`, `esbuild`, `lightningcss`, `@tbela99/css-parser`.
+
+## Layout
+
+```
+fixtures/               real-world CSS files used as benchmark input
+src/minifiers.js        adapter: unifies each library's API + reads its installed version
+src/fixtures.js         loads the fixture files
+src/compute-sizes.js    one deterministic pass -> results/sizes.json (byte size per file x library)
+src/generate-report.js  merges sizes.json + bench-results.json -> results/benchmark.html
+bench/minify.bench.js   the actual Vitest bench file (one describe per file, one bench per library)
+```
+
+## Run it
+
+```bash
+npm install
+npm run all       # sizes -> bench -> report
+open results/benchmark.html
+```
+
+Or step by step:
+
+```bash
+npm run sizes      # writes results/sizes.json
+npm run bench       # writes results/bench-results.json (vitest bench --outputJson)
+npm run report      # writes results/benchmark.html
+```
+
+## Notes
+
+- Each (file x library) benchmark is time-boxed to 300ms with a 5-iteration
+  floor (`bench/minify.bench.js`'s `BENCH_OPTIONS`) to keep the full 6-file
+  x 7-library suite finishing in a reasonable time. Raise `time`/`iterations`
+  for tighter confidence intervals if you have a few more minutes to spare.
+- Size and timing are computed in separate passes on purpose: byte size is
+  deterministic per input and doesn't benefit from repeated sampling, so
+  bundling it inside the timed `bench()` function would just add noise to
+  the timing without adding any information.
+- If a library throws on a given file (this happened with `lightningcss` on
+  `foundation.css` in testing -- "Invalid media query"), both passes catch
+  it and render `n/a` for that cell rather than crashing the whole run.
+- To add more fixture files: drop a `.css` file in `fixtures/` and add its
+  name to the `files` array in `src/fixtures.js`.
+- To add another library: add an entry to the `minifiers` array in
+  `src/minifiers.js` with an `id`, a `label`, and a `minify(css)` function
+  (sync or async).

benchmark/bench/minify.bench.mjs

@@ -0,0 +1,22 @@
+import { bench, describe } from "vitest";
+import { minifiers } from "../src/minifiers.js";
+import { fixtures } from "../src/fixtures.js";
+
+// Time budget per (file x library) benchmark. Keeps the whole suite
+// (6 files x 7 libraries = 42 benchmarks) finishing in a reasonable time
+// while still giving Vitest/tinybench enough samples for a stable mean.
+const BENCH_OPTIONS = { time: 300, iterations: 5 };
+
+for (const fixture of fixtures) {
+    describe(fixture.name, () => {
+        for (const minifier of minifiers) {
+            bench(
+                minifier.id,
+                async () => {
+                    await minifier.minify(fixture.css);
+                },
+                BENCH_OPTIONS,
+            );
+        }
+    });
+}