Our regression suite has left about nine hundred posts behind in the sandbox and the list endpoint is getting slow. Right now the deletes are in an after-hook in our own step file, which works for the happy path and does nothing when the scenario fails halfway — which is exactly when there is a half-created record to clean.
Is there a first-class way to do this, or is a hook the answer and I should write a better one?
Accepted answer
There is a step for it, and the point of it is that you register the cleanup at the moment the thing exists rather than at the end:
When I send a POST request to "/posts" with body:
"""
{ "title": "{{title}}", "userId": 1 }
"""
When I save the response JSON path "id" as "postId"
Given I register cleanup DELETE "/posts/{{postId}}"Registered cleanups run last-in-first-out at the end of the scenario, pass or fail, so a scenario that dies three steps later still deletes what it made. LIFO matters when you create a parent and a child — the child's cleanup was registered second and runs first.
A cleanup that fails is attached to the result and never fails the scenario. That is deliberate: a green test turned red by its own teardown teaches people to ignore red.
The other half of not leaving nine hundred posts behind is not creating them through the UI in the first place. If the scenario is about the list page, seed the post over the API and drive the browser only for the thing you are actually asserting. It is faster, it fails for one reason instead of two, and the cleanup is one line next to the create.
We do both of the above and still run a nightly sweep, because cleanups do not run when a worker is killed. Everything our factories generate carries a recognisable prefix, so the sweep is a filter and a delete rather than a judgement call.
Have the same question?
Ask it with the command you ran and the output you got, and it will be answered here.
Ask a question