Broadcasting TimeSeries

[Terry Reedy]

On 9/4/2013 4:06 PM, Davide Dalmasso wrote:

> I opened my Python Shell and I wrote the following:

This is with Idle. I believe you are running on Windows.

>>>> import numpy as np
>>>> import pandas as pd
>
> then I made a function
>
>>>> def afunc(aframe):
>      return aframe - aframe.mean(axis=1)
>
> and I defined a DataFrame with time as index
>
>>>> A = pd.DataFrame(np.random.randn(5,3),columns=['a','b','c'],index=pd.date_range(start='1984-12-20',periods=5))
>>>> A
>                     a         b         c
> 1984-12-20 -0.257916 -0.137923 -0.669796
> 1984-12-21 -1.632874 -1.850365  1.571715
> 1984-12-22  1.185828 -0.149839 -1.565930
> 1984-12-23 -0.757311  0.034627  0.794608
> 1984-12-24  0.548785 -1.126786 -0.438457
>
> now, if I call the function the following error is generated:
>
>>>> afunc(A)
> Traceback (most recent call last):
>    File "<pyshell#6>", line 1, in <module>
>      afunc(A)
>    File "<pyshell#3>", line 2, in afunc
>      return aframe - aframe.mean(axis=1)
>    File "C:\Python33\lib\site-packages\pandas\core\frame.py", line 217, in f
>      return self._combine_series(other, na_op, fill_value, axis, level)
>    File "C:\Python33\lib\site-packages\pandas\core\frame.py", line 3601, in _combine_series
>      return self._combine_series_infer(other, func, fill_value)
>    File "C:\Python33\lib\site-packages\pandas\core\frame.py", line 3619, in _combine_series_infer
>      FutureWarning)
>    File "C:\Python33\lib\idlelib\PyShell.py", line 60, in idle_showwarning
>      file.write(warnings.formatwarning(message, category, filename,
> AttributeError: 'NoneType' object has no attribute 'write'

This was a bug in Idle when run directly on Windows, from Start or an 
icon. In this circumstance, there is no stderr stream to write to. I 
fixed it last June by catching the AttributeError and ignoring the 
warning. The fix will be in the next 2.7 and 3.3 releases and is in the 
3.4 releases (currently 0a1, soon 0a2).

The problem (of no stderr) is avoided now, while still using Idle, by 
starting Idle from a Command Prompt window (Start/Accessories):
   python -m idlelib
or by starting the standard interpreter (from Start or icon) and entering
   >>> import idlelib.idle

> then I recall the same function (without any change) and the function works!

pandas must have a mechanism to only issue the warning once.

>>>> afunc(A)
>                     a         b         c
> 1984-12-20  0.097295  0.217289 -0.314584
> 1984-12-21 -0.995699 -1.213190  2.208890
> 1984-12-22  1.362475  0.026808 -1.389283
> 1984-12-23 -0.781285  0.010652  0.770633
> 1984-12-24  0.887604 -0.787967 -0.099637
>
> Why does this happen? Why I must call the function two times before it finally works?
> How can I overcome this problem?

Answered, I believe.

-- 
Terry Jan Reedy