geom_violin fails when the same values is often in the data

[brechtmann]

plotnine gives unclear error and shows no violin plot, when a values appears to often in the to be plotted variable.

See the following example code:

import plotnine as p9
import pandas as pd
import numpy as np

x = pd.DataFrame(data = np.array([np.ones(10), np.repeat(['A', 'B'], 5)]).T, columns=['ones', 'cat'])

p9.ggplot(x, p9.aes('factor(cat)', 'ones')) + p9.geom_violin()

Executing this example gives the following error:
plotnine.exceptions.PlotnineError: 'Width not defined. Set with position_dodge(width = ?)'

Running the same example in R ggplot2 produces a plot:

library(ggplot2)
library(data.table)

 x = data.table(ones = rep(1, 10), cat = rep(c('A', 'B')))
ggplot(x, aes(cat, ones)) + geom_violin()

The same happens, when there is only a small fraction of other values in the data.

Thanks!

[TyberiusPrime]

The error message with master-2 (4bba2e4) seems to be slightly different:

/anysnake/code_venv/lib/python3.7/site-packages/statsmodels/nonparametric/kde.py:487: RuntimeWarning: invalid value encountered in true_divide
  binned = fast_linbin(X, a, b, gridsize) / (delta * nobs)
/anysnake/code_venv/lib/python3.7/site-packages/statsmodels/nonparametric/kdetools.py:34: RuntimeWarning: invalid value encountered in double_scalars
  FAC1 = 2*(np.pi*bw/RANGE)**2
/anysnake/code_venv/lib/python3.7/site-packages/statsmodels/sandbox/nonparametric/kernels.py:204: RuntimeWarning: divide by zero encountered in double_scalars
  w = 1. / (h * n) * np.sum(self((xs-x)/h), axis=0)
/anysnake/code_venv/lib/python3.7/site-packages/statsmodels/sandbox/nonparametric/kernels.py:204: RuntimeWarning: invalid value encountered in true_divide
  w = 1. / (h * n) * np.sum(self((xs-x)/h), axis=0)

[brechtmann]

It seems to be a probablem in the density estimation. A similar error happens for:

p9.ggplot(x, p9.aes('factor(cat)', 'ones')) + p9.geom_density()