Viewer is no longer working in Colab

[fedorov]

This IDC tutorial used to work in Colab https://github.com/ImagingDataCommons/IDC-Tutorials/blob/master/notebooks/getting_started/part2_searching_basics.ipynb, as demonstrated in the captured image, but as of the latest release, invoking viewer has no effect - no viewer, no errors, no warnings.

Possibly related to #763 (comment) - the regression was noted in the release following the fixes to the zarr incompatibility.

[fedorov]

I now tested in 1.0a54 and 1.0a53, and those versions do not work either, so it is likely that something changed in Colab.

Please let me know if there is a commitment and resources to maintain compatibility with Google Colab. If not, we probably should look for alternative visualization recipes for the IDC users of our tutorial notebooks.

[thewtex]

@fedorov thanks for the report.

I tried a few things for Colab, which does break occasionally when Google changes it. Colab, a proprietary platform, is more of a black box than others. Support for Colab is a high priority, but it will take a while implement a more robust approach that is less coupled to Colab's infrastructure.

Colab links are removed in #774 for now.

[fedorov]

Colab, a proprietary platform, is more of a black box than others.

I think they share publicly the docker images they use here us-docker.pkg.dev/colab-images/public/runtime:latest (https://console.cloud.google.com/artifacts/docker/colab-images/us/public/runtime), so not quite black box I would say. Maybe even can be automatically tested for regressions.

[oeway]

Aha, it's nice that they provide the docker image now, however I'm afraid that this does not immediately help with tracking api changes, for the interaction between the viewer and Python kernel, colab uses non-standard communication mechanism (at least previously) and we had to work around it with some tricks. We need to check if there is any document describing the new api for the Jupyter comm alternative if the old one breaks.

[fedorov]

Did you consider opening an issue under https://github.com/googlecolab/colabtools and describe the problem and maybe discuss what can be done to avoid such regressions in the future?

My understanding is that Colab is maintained by a small team, but they do seem to at least acknowledge issues and opening this one would allow for other users to upvote.