Add new testIsolation configuration option

What would you like?

What I would suggest to be improved is when we are using the cy.session() command or to be more specific experimentalSessionAndOrigin=true, currently the state between the tests within one cy/spec file is cleaned and we need to reset everything, but when we do not use this flag the state is kept and we can link together some tests. It would be great to be able to keep the state as we do without the flag and not set the page url to _blank_

Why is this needed?

This would save a lot of execution time for tests where we do need to stay on the same page, or depend on some setup from the previous test where instead of logging in, setting everything up again, we can just continue where we left off (like we do when experimentalSessionAndOrigin is false)

Other

No response

[emilyrohrbough]

Test isolation is writing tests such that tests to no rely on the test before it. Clearing the page and setting the url to about:blank is one mechanism Cypress used to ensure test isolation. This behavior allows each test to have a clean slate without app state leaking, whether initially or unexpectedly, into the next test.

In general, we recommend writing tests that do no rely on each other, but with this said, we recognized there are use-cases where clearing the app’s current state is undesirable.

The request to add a new configuration option to opt-out of test isolation when experimentalSessionAndOrigin=true has been taken into consideration. There are plans to add a new configuration optional which will give you the flexibility to turn this behavior off.

@vergjor Can you create a separate issue to discuss your request to remove the clear page after the session validation is ran? We need evaluate it independently since it alters the existing behaviors of the cy.session() command.

[jnsrnhld]

Totally agree on this. I have to test a long chain of forms and with the new session behavior I have to slam all assertions and actions into one huge test or the setup effort for each test gets huge.

I suggest, for example, to add a flag to the spec context that enables to stay on the current page instead of starting with a blank page for each test. So the new behavior can be kept, but the old can still be enabled via a flag.