Code Sample

import pandas as pd
import matplotlib.pyplot as plt

data = {"A":0, "B":3, "C":-4}
df = pd.DataFrame.from_dict(data, orient = "index", columns = ["Value"])
ax = df.plot.barh()

df = df.sort_values("Value") * - 2
df.plot.barh(ax = ax, color = "red")
plt.show()

Problem description

The second plot overwrites the first index even if it is not in the same order, producing incoherent figure where the ylabel does not match with the data displayed.
The current output (pandas 0.24.1) creates

instead of

I don't know if this is related to the way pandas handles index or matplotlib.

cc @TomAugspurger

I dug a bit to see if it was pandas or matplotlib related, but it seems that it is correctly handled by matplotlib.
I reused the code and add the "pure matplotlib" version, and obtained the following figures.

data = {"A":0, "B":3, "C":-4}
df = pd.DataFrame.from_dict(data, orient = "index", columns = ["Value"])
ax = df.plot.barh()

df2 = df.sort_values("Value") * - 2
df2.plot.barh(ax = ax, color = "red")
plt.show()

fig = plt.figure()
plt.barh(df.index.values, df.Value.values)
ax = plt.gca()
ax.barh(df2.index.values, df2.Value.values)
plt.show()

Pandas plotting

Matplotlib

It seems that pandas.Dataframe.plot doesn't reuse the axes info the same way that matplotlib does.

The PR that fixed this issue has been reverted.