After an update to seaborn 0.11.1 from 0.9.0 I noticed that swarmplot
and stripplot
became MUCH slower when adding a hue variable.
I am plotting ~4000 data points across ~100 categories and 3 hue levels.
In seaborn 0.9.0 this was much faster and took ~5 seconds. Now it takes 30-60 seconds.
Without adding a hue variable, the speed is comparable between the versions.
I determined this changed with the update to 0.9.1 by trying various versions of seaborn. Is there a new setting that I missed that I can use to recapitulate the old performance in newer versions of seaborn?
EDIT: Here is an example of what I am trying to do:
dfTest = sns.load_dataset('planets')
sns.stripplot(data=dfTest,
x='method',
hue='number',
y='orbital_period',
dodge=True)
plt.legend().remove()
plt.ylim([0,1000])
plt.xticks(rotation=80);
This dataset it not quite as big (1000 rows) as the one I am working with (4000 rows) but it illustrates the point.
from Speed up swarmplot/stripplot in seaborn > 0.9.0?
No comments:
Post a Comment