Automatic color cycling when plotting consecutive timeseries without arguments disappeared in 0.8.1

[lesteve]

The fact that it took so long for anyone to realise the behaviour had changed just goes to show that it is a very minor problem. I was pretty sure it used to work at one point so I chased it down. I'd say it is quite useful from time to time when you are working interactively and feeling lazy at the same time.

In case the behaviour change was not intentional I'd be willing to look into it but some pointers would be appreciated to get started.

Here is a snippet:

import matplotlib.pyplot as plt
import pandas as pd

for i in range(3): pd.util.testing.makeTimeSeries().cumsum().plot()
plt.show()

Here is the output with pandas 0.8.0:

Here is the one with pandas 0.8.1:

[wesm]

Is it still broken in v0.10.1? Thought @changhiskhan fixed it

[changhiskhan]

The bugs I fixed had to do with using the color keyword. @lesteve the bug is a result of the fact that LinePlot always ends up assigning a color from a default or custom color cycle so when you plot consecutive Series it always uses the same color. The right thing to do is just to have it omit the color specification if unspecified and matplotlib will automatically choose a different color. Look forward to a PR!

[lesteve]

With the 0.11 release scheduled in 4 days, I was wondering whether @wesm or @changhiskhan could have a quick look at the associated pull request. I completely understand this is a very minor bug which has easy work-arounds but still it would be nice if a fix could make it to the next release.

#2836 merged.