Implement multi-yaxis support

[philippjfr]

This adds multi-yaxis support to the Bokeh backend.

ToDo

• Clean up utilities
• Add autorange support for multi-axes
• Fix linking of shared axes
• Implement axis labels correctly
• Handle correct logx/logy mappings
• Correctly determine categorical axis creation per axis
• Clean up _get_factors
• Handle dynamic axis creation when DynamicMap OverlayPlot adds new elements

[jbednar]

Very cool! How does a user configure this behavior? I think we should introduce it as an option at first since it's a change in behavior that some people may previously have been relying on (e.g. by defining dimensions only on the first plot in an overlay, rather than (correctly) defining them on all elements of the overlay).

[philippjfr]

@jlstevens and I discussed this and I am still strongly of the opinion that we should never enable this by default. Right now you enable this with the multi_y option and I think we should keep that. What we discussed as an alternative is that if you have multiple clashing vdims we will simply put all of the vdim labels on the axis (probably separated by a pipe).

My concern with enabling it by default continues to be that it is likely to be much more annoying than useful. Particularly when creating plots directly from DataFrames or xarray you will often overlay things with different column/variable names and a slight mismatch shouldn't result in multiple axes.

[jbednar]

Thanks; I didn't spot multi_y in the PNG. Sounds good.

[jlstevens]

I think I am happy with multi_y as an additional parameter: it gates the functionality, preserving backwards compatibility for now. Then, to make things more convenient again (but backwards incompatible) for HoloViews 2.0 we can simply set the default to True.

[philippjfr]

As I said above and suggested to you, I vote strongly never to enable it by default.

[jlstevens]

As I said above and suggested to you, I vote strongly never to enable it by default.

As I have no strong opinion this is also fine by me! That said, I also think other people may disagree and that this is something to discuss again when the time comes. Specifying an additional flag is more cumbersome and inconvenient though in this case, I could easily be convinced being explicit is the overriding factor.

[jbednar]

We'll see! I'll try setting it on for my workflows, and see what problems it causes in practice. Meanwhile, no need to decide about future possible changes to defaults.

[orozcojd]

Is this expected to work with Datashaded plots?

[philippjfr]

Yes, absolutely.

[GeoVizNow]

Hi, I am really looking forward to having support for multiple vdim axes. Was wondering if there are plans to include some option to configure an holoviews.Area element with multiple axes such that colors are swapping when y1 and y2 referenced to different axes are crossing. i.e. to make the below example in matplotlib a oneline with bokeh (or other backends) in holoviews.

from matplotlib import pyplot as plt
import numpy as np

xs = np.linspace(0, np.pi*4, 100)
ys = np.sin(xs)
ys2 = (np.sin(xs) * .01) + .2

plt.plot(xs, ys, color='black', zorder=20)
plt.plot(xs, ys2, color='blue', zorder=20)
plt.grid(visible=True)
plt.yticks(ticks=[-1, 0, 1])
plt.fill_between(x=xs, y1=ys, y2=ys2, where=ys >= ys2, zorder=10)
plt.fill_between(x=xs, y1=ys, y2=ys2, where=ys <= ys2, color='green', zorder=10)

[jbednar]

Was wondering if there are plans to include some option to configure an holoviews.Area element with multiple axes such that colors are swapping when y1 and y2 referenced to different axes are crossing.

@GeoVizNow, that sounds unrelated to having multiple y axes, and seems like it should be doable already by overlaying multiple hv.Area plots cropped in comparison to a reference curve. Seems like a discourse.holoviz.org topic to propose and see if someone can fill in an example.

[codecov-commenter]

Codecov Report

Merging #5621 (e319a5a) into main (d48950a) will increase coverage by 0.01%.
The diff coverage is 92.19%.

@@            Coverage Diff             @@
##             main    #5621      +/-   ##
==========================================
+ Coverage   88.19%   88.20%   +0.01%     
==========================================
  Files         307      308       +1     
  Lines       63305    63628     +323     
==========================================
+ Hits        55829    56125     +296     
- Misses       7476     7503      +27     

Flag	Coverage Δ
ui-tests	22.30% <16.15%> (-0.05%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
holoviews/plotting/bokeh/util.py	66.89% <69.76%> (+0.08%)	⬆️
holoviews/plotting/plot.py	91.71% <88.46%> (-0.20%)	⬇️
holoviews/plotting/bokeh/element.py	88.17% <90.69%> (+0.21%)	⬆️
holoviews/tests/plotting/bokeh/test_multiaxis.py	98.28% <98.28%> (ø)
holoviews/plotting/bokeh/annotation.py	85.14% <100.00%> (ø)
holoviews/plotting/bokeh/chart.py	88.18% <100.00%> (-0.17%)	⬇️
holoviews/plotting/bokeh/graphs.py	91.32% <100.00%> (-0.18%)	⬇️
holoviews/plotting/bokeh/heatmap.py	91.76% <100.00%> (ø)
holoviews/plotting/bokeh/hex_tiles.py	94.73% <100.00%> (ø)
holoviews/plotting/bokeh/sankey.py	80.47% <100.00%> (ø)
... and 7 more

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[jlstevens]

Bokeh 3 tests are now green! :-)

However, something since 2025b5d has broken Bokeh 2 badly (which now hangs. My suspicion is that it could be due to the Field import but I could be wrong...

[jlstevens]

Adding a note as a reminder that we will need a proper fix for the new dask release and to unpin in setup.py!

[hoxbro]

@philippjfr and I have reviewed your code and left some comments and suggestions for you.

Something to note is we haven't reviewed the Javascript code in callback = CustomJS(code=f"""

One last thing is, please doublecheck that all get_extents has the same signature.

[hoxbro]

As discussed with @philippjfr, this PR is in a good state and can be merged, and then I will tag a new dev version.

Thank you for all your work here, @jlstevens! I know how you have loved every single second of bringing this fantastic functionality to life!

[jlstevens]

I know how you have loved every single second of bringing this fantastic functionality to life!

Of course! :-)

[GeoVizNow]

Thank you for working up this great new feature!