label keyword is not interpreted proporly in errorbar() for pandas.DataFrame-like objects

[monodera]

I encountered an issue of the legend in matplotlib.pyplot.errorbar when I plot a pandas' DataFrame object. A minimum example based on the release note of v1.5 is the following.

import pandas
import numpy as np
import matplotlib.pyplot as plt

df = pandas.DataFrame({"var1":[1,2,3,4,5,6], "var2":[1,2,3,4,5,6]})

Here is the same as one in the release note, but with a legend.

plt.plot(df["var1"], df["var2"], label='test')
plt.legend()

The output is the following.

The following code makes the issue.

plt.errorbar(df["var1"], df["var2"], label='test')
plt.legend()

As you can see, the 'var2' key is not overridden by the label='test' and there are two lines instead of one in the legend.

A quick workaround is to convert the y values to numpy array with np.array(df["var2"]) when plotting.

I'm using Python 2.7.8 installed via homebrew on OS X 10.10.5, matplotlib 1.5.0, and pandas 0.17.0.

[tacaswell]

I think this is a duplicate of #5393