Conda package incompatibility between JupyterLab 3.6 and nbgrader 8.1

[nthiery]

Operating system: Ubuntu + micromamba

nbgrader --version

0.8.1

Expected behavior

Being able to install simultaneously JupyterLab 3.6 and nbgrader 8.1

Actual behavior

Failure to install.

Steps to reproduce the behavior

1. Install micromamba (or presumably any other conda installation)
2. Try to create a fresh environment:

> mamba create -n toto "jupyterlab>=3.6" nbgrader=0.8.1
Encountered problems while solving:
  - package jupyterlab-3.6.0-pyhd8ed1ab_0 requires jupyter_server_ydoc >=0.6.0,<0.7.0, but none of the providers can be installed
The environment can't be solved, aborting the operation

[brichet]

Thanks @nthiery for reporting this.
The error message is quite confusing, as nbgrader does not directly depend on jupyter_server_ydoc.
If we first install "jupyterlab>=3.6" in a fresh environment, and then try to install nbgrader=0.8.1, the error message is the following:

Could not solve for environment specs
Encountered problems while solving:
  - package jupyter_events-0.6.3-pyhd8ed1ab_0 requires traitlets >=5.3, but none of the providers can be installed

Actually the pinned version of traitlets in conda is not the same as the one in pip (traitlets <5.2.0 vs "traitlets<5.7.0".
It might be the error here.

[brichet]

I opened conda-forge/nbgrader-feedstock#58 which pins jupyterlab<3.6.
We should probably release a 0.8.2 working with jupyterlab 3.6.x.

EDIT:
Sorry, I forgot the first expectation, the PR on conda forge is not necessary. I'll release 0.8.2 soon.

[brichet]

I forgot to mention in this issue that a 0.8.2 version as been released, compatible with Jupyterlab=3.6.x.
It's also available on conda-forge.