[SciPy-User] failing to create a scikit.timeseries object

Pierre GM pgmdevlist@gmail....
Sun Jun 6 19:38:26 CDT 2010


On Jun 3, 2010, at 3:36 AM, eneide.odissea wrote:
> 
> I Thank you all and I apologize for my very bad code snippet.
> Do you know whether in scikits.timeseries there is a command / option / configuration that allows to store time using long instead of integer?

It's already done internally in the C code:  the basic element that corresponds to a Date is the date_info struct, that uses long to store the absolute date and double the absolute time. For more details, please refer to cdates.c (in the src directory of the distribution).

> Probably it might be necessary also to setup a callback somewhere  able to convert the datetime into this internally stored number ; have you any idea about it?
 All done in C...

Now, there are some limitations w the current approach, the most obvious one being that you can't define your own frequencies. There's some new code in numpy that should allow it, I need to work on that...


More information about the SciPy-User mailing list