Add API to create ridge / joy plots?

[maximlt]

The subcoordinate_y feature added to HoloViews opens the door to the implementation of ridge / joy plots in hvPlot. I'm opening this issue to gauge the interest in this feature and discuss it.

HoloViews code (stolen from Demetris'!)

import numpy as np
import holoviews as hv
from scipy.stats import gaussian_kde

hv.extension('bokeh')

categories = ['A', 'B', 'C', 'D', 'E']
data = {cat: np.random.normal(loc=i-2, scale=1.0, size=100) for i, cat in enumerate(categories)}
x = np.linspace(-5, 5, 100)

curves = []
areas = []
for i, (cat, values) in enumerate(data.items()):
    pdf = gaussian_kde(values)(x)    
    area = hv.Area((x, pdf), label=cat).opts(
        subcoordinate_y=True, 
        subcoordinate_scale=1.5,
    )
    areas.append(area)

ridge_plot_areas = hv.Overlay(areas).opts(
    width=900,
    height=400,
)

ridge_plot_areas.opts(show_legend=False)

Python users can already create such plots using the usual plotting libraries, albeit usually by having to resort to low-level APIs:

• https://python-graph-gallery.com/ridgeline/
• Bokeh: https://docs.bokeh.org/en/latest/docs/examples/plotting/ridgeplot_subcoordinates.html
• Seaborn: https://seaborn.pydata.org/examples/kde_ridgeplot.html
• Plotly: https://plotly.com/python/violin/, Ridgeline Plot plotly/plotly_express#25
• Matplotlib: https://matplotlib.org/matplotblog/posts/create-ridgeplots-in-matplotlib/

I also found two packages implementing high-level APIs:

• ridgeplot (7K downloads/month): https://github.com/tpvasconcelos/ridgeplot
• joypy (15K downloads/month): https://github.com/leotac/joypy

And a few other packages from other ecosystems:

• ggridges: https://github.com/wilkelab/ggridges
• stata-joyplot: https://github.com/asjadnaqvi/stata-joyplot

(#1160 should be implemented first)

[droumis]

Is this theoretical joyplot API separate or the same API as would support a multi-chan timeseries plot? I think they are probably different.

[maximlt]

Note to self. For a little while I got confused with what exactly a ridgeplot is, trying to adapt it to fit and replace subcoordinate_y. However, a ridgeplot most often is an overlay of kde plots, this is the output this API should offer.

import pandas as pd
import numpy as np
import hvplot.pandas

df = pd.DataFrame(np.random.random((1000, 4)), columns=list("ABCD"))
df.hvplot.ridge()  # somewhat better version of `df.hvplot.kde(subcoordinate_y={'subcoordinate_scale': 2})`