Workaround for later initialization, recovers document_ttl

[asteppke]

This is a short fix for #315.

In the future this can be updated when the changed traitlets infrastructure is fully taken into account but for now this restores the document_ttl functionality and should have no negative side effects.

[github-actions]

👈 Launch a Binder on branch asteppke/jupyter-collaboration/main

[davidbrochart]

Thanks @asteppke.

In the future this can be updated when the changed traitlets infrastructure is fully taken into account

What do you mean by "changed traitlets infrastructure"?

[davidbrochart]

I don't understand what this changes. Could you write a test that shows that this fixes something?

[asteppke]

I added a test that checks whether the store has the correct parameter from the command line arguments.

Without the changeset above the store writing fails immediately if document_ttl is set.

[asteppke]

Thanks @asteppke.

In the future this can be updated when the changed traitlets infrastructure is fully taken into account

What do you mean by "changed traitlets infrastructure"?

I did not bisect this in detail but when you implemented the document_ttl functionality in the past I think it worked as intended. Instead of going through the normal traitlets initialization code the attributes are set directly here in jupyter_collaboration (I guess because of late initialization). The traitlet guys mention in passing in their documentation that since 5.0 there is a deferral happening of the conversion from command line arguments to traits. So therefore jupyter_collaboration is now setting the 'still-unconverted-strings' as attributes instead of traits.

So this adding the missing conversion step. I can see if this can be shown also within the a test in the test suite.

[asteppke]

I added the corresponding tests for the document_ttl. Other unrelated tests fail though already without these commits.