Specifying histtype='stepfilled' and normed=True when using plt.hist causes ymax to be set incorrectly

[joelostblom]

The combination of histtype='stepfilled' and normed=True seem to cause some problems with the max ylimit.

import matplotlib.pyplot as plt
import seaborn.apionly as sns
iris = sns.load_dataset('iris')

plt.figure()
for species, species_df in iris.groupby("species"):
    plt.hist(species_df["petal_length"].values, normed=True, histtype="stepfilled")
plt.legend(sorted(iris.species.unique()))

Which is not the case when only using histtype='stepfilled'.

plt.figure()
for species, species_df in iris.groupby("species"):
    plt.hist(species_df["petal_length"].values, histtype="stepfilled")

Or when only using normed=True.

plt.figure()
for species, species_df in iris.groupby("species"):
    plt.hist(species_df["petal_length"].values, normed=True)

It seems like in the erroneous case above, the max of the yaxis is set by the last plotted distribution, which can be tested by changing the order of when the different histograms are plotted.

plt.figure()
iris.loc[iris.species == 'setosa', 'species'] = 'wsetosa'
for species, species_df in iris.groupby("species"):
    plt.hist(species_df["petal_length"].values, normed=True, histtype="stepfilled")
plt.legend(sorted(iris.species.unique()))

[tacaswell]

I suspect that some place there is a ax.set_ylim(max(..)) call which just needs to be changed to an autoscale/limit call, but given that this is hist one can never be sure.

[QuLogic]

This was corrected by #7476 and should be fixed in 2.0.0.