ES6 polyfill vs. feature performance tests.
Report is located at http://fhinkel.github.io/six-speed/
clone, npm install,
npm start. Open localhost:9999/#destructuring to see timings for all tests
related to destructuring.
We currently only care about es5 vs es6.
Generated files are in build. If you makes changes to checked in tests, don't
forget to run ./node_modules/gulp/bin/gulp.js build.
On 403 or 404, run ./node_modules/gulp/bin/gulp.js build and maybe npm run test.
npm run test:sauce
Test against all registered Sauce Labs browsers.
npm run test:node
Tests against the current node version.
npm run profile:node -- --testName=$name --type=$type --count=$iterationCount
Profiles a given test within the current Node environment. Type may be one of:
- babel
- babel-runtime
- traceur
- typescript
- es5
- es6
npm start
Starts a server instance for manual browser testing. Tests may be accessed via http://machineName:9999/ and the # component may be used to filter the tests to be executed, i.e. http://machineName:9999/#promises
Profiling of specific tests may be done through http://machineName:9999/profile.html?testName=$testName&type=$type&count=$number, i.e. http://localhost:9999/profile.html?testName=generator&type=babel&count=1000000.
Firefox browsers need to use /moz/index.html and /moz/profile.html respectively to enable all supported ES6 features.
npm run report
Generates the data report.
For each of the ES6 features in question, an ES5 implementation of that functionality was written along with an ES6 version. It should be noted that the functionality is frequently the same, but in some cases the "common" vs. "correct" version was written, i.e. using x[key] = value vs. defineProperty which is faster but can be hit but a particular nasty edge case for those who deem it fun to extend Object.prototype.
Babel, in both loose+runtime and runtime mode, and Traceur were then used to compile the ES6 version to an ES5 compliant version, utilizing the runtime over polyfill to maintain test isolation and avoid native implementations where possible.
All of these test instances were then benchmarked in the given JavaScript engine using Benchmark.js and then the operations per second compared to the ES5 implementation. Cross browser and cross execution comparisons are avoided as much as possible to isolate environmental issues when executing on VMs in the cloud.
./bin/test-all.shnpm run report- Checkin changes to site sub-repository.
The Windows 10 VM used must be manually setup to ensure the proper state prior to testing. This can be done with this command:
mkdir browsers
./node_modules/.bin/browser-downloader vm ./browsers
After this the image should be restarted a few times until all setup and update processes have completed and then a snapshot named six-speed taken from the idle desktop screen. The test-all.sh script will check that the VM image is up to date and will halt execution if the image is not setup properly, as a sanity check.
Thanks to BrowserStack and Sauce Labs for providing open source accounts which the majority of this testing was performed on.