Any smart approach for determining what Holoviz package broke my notebook?

Panel
1

I have a HRRR Explorer notebook that I use in a lot of my Pangeo demos that works with this environment:

$ conda list | grep -E "bokeh|datashader|holoviews|geoviews|hvplot|panel"
bokeh                     2.4.3              pyhd8ed1ab_3    conda-forge
bokeh-root-cmd            0.1.2              pyhd8ed1ab_0    conda-forge
datashader                0.14.3             pyh1a96a4e_0    conda-forge
geoviews                  1.9.5              pyhd8ed1ab_0    conda-forge
geoviews-core             1.9.5              pyha770c72_0    conda-forge
holoviews                 1.15.2             pyhd8ed1ab_0    conda-forge
hvplot                    0.8.2              pyhd8ed1ab_0    conda-forge
jupyter-panel-proxy       0.1.0                      py_0    conda-forge
jupyter_bokeh             3.0.5              pyhd8ed1ab_0    conda-forge
panel                     0.14.1             pyhd8ed1ab_0    conda-forge

2023-08-29_16-57-07
2023-08-29_16-57-071175×652 215 KB

but not with this environment:

14:56 $ conda list | grep -E "bokeh|datashader|holoviews|geoviews|hvplot|panel"
bokeh                     2.4.3              pyhd8ed1ab_3    conda-forge
bokeh-root-cmd            0.1.2              pyhd8ed1ab_0    conda-forge
datashader                0.15.2             pyhd8ed1ab_0    conda-forge
geoviews                  1.9.6              pyhd8ed1ab_0    conda-forge
geoviews-core             1.9.6              pyha770c72_0    conda-forge
holoviews                 1.17.1             pyhd8ed1ab_0    conda-forge
hvplot                    0.8.4              pyhd8ed1ab_1    conda-forge
jupyter-panel-proxy       0.1.0                      py_0    conda-forge
jupyter_bokeh             3.0.5              pyhd8ed1ab_0    conda-forge
panel                     0.14.4             pyhd8ed1ab_0    conda-forge

2023-08-29_16-57-37
2023-08-29_16-57-371166×668 254 KB

(It also doesn’t work with bokeh=3, panel=1)

So the culprit could be datashader, geoviews, holoviews, panel or hvplot. Any good way to determine which version of what broke things?

2

That looks like a projection issue so I would guess geoviews / cartopy / hvplot.

3

It is a regression in hvplot 0.8.4. We plan to make a release next week, which will fix it. Reference issue: #1090 (comment). I checked and confirmed that it was the same commit that was the culprit.

So the culprit could be datashader, geoviews, holoviews, panel or hvplot. Any good way to determine which version of what broke things?

My suggestion would be to upgrade them one at a time and see which one breaks it.

4

Yep! Pinning hvplot==0.8.2 in the environment solved the problem:

10:07 $ conda list | grep -E "bokeh|datashader|holoviews|geoviews|hvplot|panel"
bokeh                     3.2.2              pyhd8ed1ab_0    conda-forge
bokeh-root-cmd            0.1.2              pyhd8ed1ab_0    conda-forge
datashader                0.15.2             pyhd8ed1ab_0    conda-forge
geoviews                  1.10.1             pyhd8ed1ab_0    conda-forge
geoviews-core             1.10.1             pyha770c72_0    conda-forge
holoviews                 1.17.1             pyhd8ed1ab_0    conda-forge
hvplot                    0.8.2              pyhd8ed1ab_0    conda-forge
jupyter-panel-proxy       0.1.0                      py_0    conda-forge
jupyter_bokeh             3.0.7              pyhd8ed1ab_0    conda-forge
panel                     1.2.1              pyhd8ed1ab_0    conda-forge

and it works!

2023-08-30_11-05-53
2023-08-30_11-05-531142×670 238 KB

5

@Hoxbro The current release of hvplot on conda-forge is still 0.8.4 so keep this pinned at 0.8.2 for now, right?

6

Pinning to 0.8.3 should also be fine. We hope to release hvplot 0.9.0 in the coming week.

2 Likes
7

@Hoxbro, Any updated ETA for the hvplot 0.9.0 release?

8

Next week. I promise this time!

A release candidate has been made today :slightly_smiling_face:

2 Likes
9

hvplot 0.9.0 is now released!