| epriestley |
| Blessed Reviewers |
Some linters have a high startup cost, such that the time required to start the external tool are unproportionally high compared to the amount of time required to actually lint the file contents. A good example is ESLint, which has been reported internally to be nine times slower when run through arc lint compared to running eslint directly.
I think that the best solution here is to allow paths to be passed to external linters in batches.
This needs some more work, but I wanted to know what @epriestly thought before I continued working on it. With a batch-size of 20, the time required to run arc lint on 60 JavaScript files reduced from 30 seconds to 7 seconds locally.
| Lint OK |
| Unit Test Errors |
| Buildable 15623 | |
| Build 20619: Run Core Tests | |
| Build 20618: arc lint + arc unit |
| Time | Test | |
|---|---|---|
| 333 ms | ArcanistJSHintLinterTestCase::testLinter Assertion failed, expected values to be equal (at ArcanistLinterTestCase.php:131): Expect one result returned by linter.
Expected: 1
Actual: 2 | |
| 126 ms | ArcanistPhpLinterTestCase::testLinter Assertion failed, expected values to be equal (at ArcanistLinterTestCase.php:131): Expect one result returned by linter.
Expected: 1
Actual: 2 | |
| 10 ms | ArcanistChmodLinterTestCase::testLinter 5 assertions passed. | |
| 0 ms | ArcanistClosureLinterTestCase::testVersion 1 assertion passed. | |
| 0 ms | ArcanistCpplintLinterTestCase::testVersion 1 assertion passed. | |
| View Full Test Results (2 Failed · 19 Passed · 33 Skipped) |