Bad x-axis dates when plotting two time-series with different indexes (start, end, and freq)

[ocefpaf]

The two DataFrames are from 2014, but from the figure you can see that pandas throws one of them to the year 2050!

I am using pandas 0.15.2 and the notebook below describes the problem and show a few workarounds that might help debug this:

http://nbviewer.ipython.org/gist/ocefpaf/e7e085c84fd3ea7dbc24

Here is a quick SSCCE to generate the figure above:

import numpy as np                                                                               
import pandas as pd                                                                              

index = pd.date_range(start='2014-06-01 00:00:00', end='2014-09-01 00:00:00', freq='H')          
data = np.random.rand(len(index))                                                                
df0 = pd.DataFrame(data=data, index=index, columns=['DF0'])                                      
index = pd.date_range(start='2014-04-07 10:57:00', end='2014-10-29 09:32:00', freq='D')          
data = np.random.rand(len(index))                                                                
df1 = pd.DataFrame(data=data, index=index, columns=['DF1'])                                      

ax = df0.plot(legend=True, figsize=(12, 4))                                                      
ax = df1.plot(ax=ax)                                                                             

Thanks!

[TomAugspurger]

Thanks for the report. It seems to be tripping up our Autoformatter.

The easiest workaround is to pass the x_compat=True parameter to your first plot call.

ax = df0.plot(legend=True, figsize=(12, 4), x_compat=True)
ax = df1.plot(ax=ax)                                                                             

[ocefpaf]

Thanks @TomAugspurger! That worked nicely.

I updated the notebook to show x_compat=True as a preferable workaround to this bug.

PS: x_compat is not documented, correct?

[shoyer]

Thanks @ocefpaf for the report and @TomAugspurger for the fix! I actually encountered the same issue today when comparing daily and hourly data:

[jreback]

@TomAugspurger close this? as this is a usage question, or need docs? or is it a bug?

[TomAugspurger]

Closing for now. This would need a big refactor of our datetime plotting to support this properly, which I won't be able to get to anytime soon.