如何从hv.Dataset创建分组条形图？

Question

I would like to create a grouped bar plot from a pandas.DataFrame using Holoviews Datasets.

In particular I would like to have bars grouped by their original column. Right now bars are being plotted on top of each other:

import pandas as pd
import holoviews as hv
hv.extension('bokeh')

df = pd.DataFrame({'A': list(range(10,15)),
                   'B' : list(reversed(range(20,25)))})
ds = hv.Dataset(df, kdims='index')
ds.to.bars(vdims='A')*ds.to.bars(vdims='B').opts(alpha=.5)

This results in: current snippet

However I would like to have them side-by-side, similar to the plot featured in the official docs (bottom of the page).

Answer 1

For plotting directly with pandas DataFrames I would recommend the hvPlot library which is built on top of HoloViews. HoloViews deals well with so called tidy data , while hvPlot is built to deal well with both tidy and wide data. Using hvPlot generating the plot you want is as simple as:

import hvplot.pandas
df.hvplot.bar()

Using HoloViews directly you would have to use pd.melt to convert the data into a so called tidy format. That would look like this:

hv.Bars(pd.melt(df.reset_index(), ['index']), ['index', 'variable'], 'value')

For a neat explanation for what pd.melt is doing see this visualization