Is there any way to get interactive matplotlib in Jupyterlab without conda-forge?

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

Is there any way to get interactive matplotlib in Jupyterlab without conda-forge?

Brendan Barnwell
        I was recently told (
file:///C:/Users/BrenBarn/AppData/Roaming/jupyter/runtime/nbserver-6156-open.html
) that Jupyterlab is supposed to be ready for prime time.  For me that
would include interactive use of Matplotlib.  Apparently the way to do
this in Jupyterlab is with ipympl, aka "jupyter-matplotlib" at
https://github.com/matplotlib/jupyter-matplotlib .

        Looking at the instructions, I see it tells me to install from
conda-forge.  However, I have found several warnings (e.g., here
https://github.com/conda/conda/issues/8920#issuecomment-519105949 ) that
mixing conda defaults and conda-forge channels can cause problems.  I
have experienced this myself with conda going crazy upgrading and
downgrading packages due to slight version differences between the
packages in different channels.

        I am going to be teaching a class using Python.  I do not want to use
conda-forge because I want to keep installation of everything as simple
as possible for the students and avoid the possibility of such problems.

        So my question is: is it possible to get fully functional interactive
Matplotlib in Jupyterlab WITHOUT ever using conda-forge?  If not, I
would say that Jupyterlab is not really ready for primetime!  (I realize
that, if the answer is no, that may be more an Anaconda issue than a
Matplotlib or Jupyter issue, but I'm just trying to figure out whether
it's actually possible to replicate my working Jupyter notebook
environment in Jupyterlab using only the conda defaults channel.)

Thanks,
--
Brendan Barnwell
"Do not follow where the path may lead.  Go, instead, where there is no
path, and leave a trail."
    --author unknown
_______________________________________________
Matplotlib-users mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/matplotlib-users
Reply | Threaded
Open this post in threaded view
|

Re: Is there any way to get interactive matplotlib in Jupyterlab without conda-forge?

Jason Grout
We can see from https://anaconda.org/search?q=ipympl that the Anaconda channel is not publishing ipympl.

You can install everything in a new environment from conda-forge, which would avoid the problem of mixing channels. That is what I typically do, adding a couple of lines to the config file or a couple of command line switches to the install command.

As you say, publishing this package is an Anaconda issue, totally outside of our control in developing and publishing JupyterLab. Maybe you could request Anaconda look into shipping ipympl?

I suppose from your perspective it would be more correct to use the exact same logic to say that Anaconda's support for JupyterLab is not quite ready for prime time, given that shipping this package is an Anaconda issue and not a JupyterLab issue :). (I would not make that statement personally, but I understand your perspective.)

Thanks,

Jason


On Thu, Aug 8, 2019 at 11:51 PM Brendan Barnwell <[hidden email]> wrote:
        I was recently told (
file:///C:/Users/BrenBarn/AppData/Roaming/jupyter/runtime/nbserver-6156-open.html
) that Jupyterlab is supposed to be ready for prime time.  For me that
would include interactive use of Matplotlib.  Apparently the way to do
this in Jupyterlab is with ipympl, aka "jupyter-matplotlib" at
https://github.com/matplotlib/jupyter-matplotlib .

        Looking at the instructions, I see it tells me to install from
conda-forge.  However, I have found several warnings (e.g., here
https://github.com/conda/conda/issues/8920#issuecomment-519105949 ) that
mixing conda defaults and conda-forge channels can cause problems.  I
have experienced this myself with conda going crazy upgrading and
downgrading packages due to slight version differences between the
packages in different channels.

        I am going to be teaching a class using Python.  I do not want to use
conda-forge because I want to keep installation of everything as simple
as possible for the students and avoid the possibility of such problems.

        So my question is: is it possible to get fully functional interactive
Matplotlib in Jupyterlab WITHOUT ever using conda-forge?  If not, I
would say that Jupyterlab is not really ready for primetime!  (I realize
that, if the answer is no, that may be more an Anaconda issue than a
Matplotlib or Jupyter issue, but I'm just trying to figure out whether
it's actually possible to replicate my working Jupyter notebook
environment in Jupyterlab using only the conda defaults channel.)

Thanks,
--
Brendan Barnwell
"Do not follow where the path may lead.  Go, instead, where there is no
path, and leave a trail."
    --author unknown
_______________________________________________
Matplotlib-users mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/matplotlib-users

_______________________________________________
Matplotlib-users mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/matplotlib-users