jupyter_bokeh version in conda channel
qiyubio opened this issue · comments
Hi, the jupyter_bokeh version in conda bokeh channel is still 2.0.4, and the one in conda-forge channel is 3.0.0. Would you want to update the README to point that out, so that it will not confuse python3 users? Thanks.
@qiyubio The README already indicates to install from the bokeh channel on anaconda.org, which will always work:
Updates will reach conda-forge pretty quickly as well.
The 2.0.4 still work with Jupyterlab>3, except when you run "jupyter labextension list", it doesn't show up correctly or it will keep request to rebuild. Would you add jupyter_bokeh 3.0.0 to bokeh channel as well?
only has 2.0.4
https://anaconda.org/bokeh/jupyter_bokeh/labels
https://anaconda.org/bokeh/jupyter_bokeh/files
has 2.0.4 and 3.0.0
https://anaconda.org/conda-forge/jupyter_bokeh/labels
https://anaconda.org/conda-forge/jupyter_bokeh/files
Hrm I guess it did not get published to anaconda.org after all. @philippjfr @mattpap thoughts? I would be fine only supporting pip and conda-forge and unpublishing the extension from the anaconda.org channel entirely (one less thing to deal with)
@philippjfr @mattpap actually I am a bit confused, what even are the pre-3.0 python packages? I thought prior to 3.0 the only way to install was via labextension with the NPM packages?
In any case, I reiterate the question: do we need to publish this package to the bokeh channel at all, or is pip + conda-forge sufficient?
The jupyter_bokeh package prior to the recent release included the BokehWidget which is an ipywidget wrapper around any Bokeh component. This allows bi-directional communication just like on a Bokeh server. Since the recent release it now serves a dual purpose since we also bundle the JLab extension.
I think we should also get the package on defaults and be done with it. I don't think publishing it to the bokeh channel adds much at that point.
That said, I think we may have to reconsider the npm package since that is the only mechanism by which we can support rendering and interactivity in VSCode since that only works via the ipywidget support.
I don't think publishing it to the bokeh channel adds much at that point.
Can the entire package there just be deleted, then?
I think we should also get the package on defaults and be done with it.
I guess I am not confident that the update cycle for defaults is fast enough for users, I think we have to keep conda-forge as well and perhaps advise that as the default?
That said, I think we may have to reconsider the npm package since that is the only mechanism by which we can support the VSCode via the ipywidget support.
That's annoying. Can you make a separate issue?
I've updated the README to use -c conda-forge for now at least.
@philippjfr is any action needed here still?
I am not sure what if any action is still needed, so I will close this now.
Version 3.0.4 is definitely uploaded to the bokeh channel so I think we're good.