[IPython-user] Parallel w/ scikits.timeseries dates

Robert Ferrell ferrell@diablotech....
Thu Dec 10 08:43:26 CST 2009


On Dec 10, 2009, at 7:04 AM, Robert Ferrell wrote:

>
> On Dec 9, 2009, at 4:56 PM, David Warde-Farley wrote:
>
>> On 9-Dec-09, at 5:33 PM, Robert Ferrell wrote:
>>
>>> I want to do a task parallel computation for each date in a
>>> timeseries.DateArray.  My naive attempt produces:
>>>
>>> 	PicklingError: Can't pickle <type 'timeseries.Date'>: import of
>>> module timeseries failed
>>
>>
>> You might try asking on the SciPy-user list, I know the developers
>> of scikits.timeseries answer a lot of questions there, and if this
>> is the case then it'll affect all modes of parallelism (the
>> multiprocessing module, for example) and not just IPython's (and I
>> really hope they have a solution for parallel processing -- it's not
>> even clear to me why a Date object shouldn't be picklable).
>>
>> Best,
>>
>> David
>
> I'll send a note to SciPy.  I know that pickling has been an issue.
> My work around has been to convert dates to strings.  But, when the
> objects get complicated that's not so simple.

For closure- on SciPy-User, Pierre pointed out to me that DateArray  
and TimeSeries objects are picklable.  So this isn't a big an issue as  
I had thought.  Date instances can be contained in a single-element  
DateArray, for instance.

-robert



More information about the IPython-user mailing list