Define end-to-end test experimental features utility as lifecycle helper #22712
Conversation
aduth
added
[Type] Automated Testing
|
Size Change: 0 B Total Size: 1.12 MB ℹ️ View Unchanged
|
| @@ -11,7 +11,7 @@ import { | |||
| /** | |||
| * Internal dependencies | |||
| */ | |||
| import { enableExperimentalFeatures } from '../../experimental-features'; | |||
| import { useExperimentalFeatures } from '../../experimental-features'; | |||
There was a problem hiding this comment.
It reads like React hook because of the use prefix 🙃
Edit: Now I read description, it's intended 👍
There was a problem hiding this comment.
Yeah, intentional in that the purpose is largely the same in utilities to manage lifecycle. But I grant they are quite different as far as the underlying implementation, and there is some potential for confusion. It feels like the chance of actual conflict is fairly low. I'm open to different naming. In the original Slack conversation, I mentioned withFooEnabled as an example, though there's a similar issue there with how we tend to use with prefixes for higher-order functions. I also don't think that React should hold exclusive rights to this prefix, since it's quite generic.
🤷
It's an internal utility, so it can be changed in the future. I do like the idea of propagating this sort of pattern for the general requirement of test lifecycle utilities, so at some point a pattern should be established if it becomes more commonplace.
There was a problem hiding this comment.
It’s totally fine. I like this parallel. I just wanted to emphasize how this concept changed the perception when reading code outside of the React context (it wasn’t intentional to use double meaning here but now I find it funny).
There was a problem hiding this comment.
I do like the idea of propagating this sort of pattern for the general requirement of test lifecycle utilities
Related: #22804
|
Awesome. Thanks for this! |
Related Slack conversation (link requires registration): https://wordpress.slack.com/archives/C02QB2JS7/p1590677131416000
This pull request seeks to resolve intermittent failures of end-to-end tests, where the block directory end-to-end test suite does not currently tear down to reset experimental features to the initial state, causing subsequent tests run in the same container to unexpectedly run with the block directory experiment enabled. This causes builds to fail due to a related issue with ARIA attributes not being applied:
Example: https://travis-ci.com/github/WordPress/gutenberg/jobs/340966385
Implementation Notes:
The implementation here addresses this by encapsulating the entire lifecycle requirements of enabling experimental features into a single utility, thus avoiding the potential to overlook the need to provide a
afterEachtear-down.The idea here is to promote a pattern that in cases where something should be "activated" before and "deactivated" after a test is run, we should prefer to create a single utility which manages the full lifecycle of the behavior in a test. In this way, it behaves a lot like a React Hook.
Before:
After:
Testing Instructions:
Ensure end-to-end tests pass: