Re: [Pytables-users] datetime objects on tables

[Fernando P. <fs...@gm...>]

Hi Anthony,

Thanks for your suggestions. I'll try number (1) and see how it goes.
If, however, I still need number (2) and have the time to do a decent
work I let you know. Indeed it would be nice to have the
numpy.datetime64 with its arithmetic capabilities supported in
pytables.

Thanks,
-Fernando


On Thu, Nov 3, 2011 at 4:20 PM, Anthony Scopatz <sc...@gm...> wrote:
> Hello Fernando,
> I personally have always found the 64-bit time stamps to be much more useful
> (and much less ambiguous) than Python datetime objects.
> However, if against my better judgement, you decide to store datetimes, you
> effectively have three options (in preferred order):
>
> Create a numpy dtype or PyTables description that matches the structure of
> datetime to your desired precision and save them in a Table. If you want to
> save timezone information as well, I might add an extra length-3 string
> column and save the str representation of the tzinfo field ('UTC', 'KST',
> 'EDT').
> Look into the new numpy native datetime type.  I am not sure how they are
> storing this under the covers; it might be as time stamps plus some
> meta-data (which would be easy for use to store).  Then it would be nice for
> PyTables to have an appropriate Col subclass to go along with this.
> Store an array of pickled datatime objects (and possibly index it).  This is
> ill-advised as it will likely be slow.
>
> I would go with (1).  Number (2) should probably be done eventually (and it
> would be great if you did some leg work and had a pull request ~_~).
> Be Well
> Anthony
> On Thu, Nov 3, 2011 at 12:26 AM, Fernando Paolo <fs...@gm...> wrote:
>>
>> Hi Ben,
>>
>> Thanks for the reply. Actually that's the way my measurements come (a
>> single number representing seconds since a reference epoch). What I
>> need is to fractionate this number into the respective time elements
>> (year, month, day, hour, secs, micro secs) and operate with them. So I
>> would appreciate any suggestion on how to properly store (or
>> represent) these on a very large table in order to be able to perform
>> queries (without inconvenient transformations on the fly).
>>
>> Thank you - Fernando
>>
>>
>> On Wed, Nov 2, 2011 at 9:55 PM, Ben Elliston <bj...@ai...> wrote:
>> > Hi Fernando
>> >
>> > I don't know about datetime, per se, but you can also convert your
>> > times/dates into time since Unix epoch and store that as a single
>> > 64-bit integer.
>> >
>> > Cheers, Ben
>> >
>> >
>> > ------------------------------------------------------------------------------
>> > RSA(R) Conference 2012
>> > Save $700 by Nov 18
>> > Register now
>> > http://p.sf.net/sfu/rsa-sfdev2dev1
>> > _______________________________________________
>> > Pytables-users mailing list
>> > Pyt...@li...
>> > https://lists.sourceforge.net/lists/listinfo/pytables-users
>> >
>>
>>
>>
>> --
>> Fernando Paolo
>> Institute of Geophysics & Planetary Physics
>> Scripps Institution of Oceanography
>> University of California, San Diego
>> 9500 Gilman Drive
>> La Jolla, CA 92093-0225
>>
>> web: fspaolo.net
>>
>>
>> ------------------------------------------------------------------------------
>> RSA(R) Conference 2012
>> Save $700 by Nov 18
>> Register now
>> http://p.sf.net/sfu/rsa-sfdev2dev1
>> _______________________________________________
>> Pytables-users mailing list
>> Pyt...@li...
>> https://lists.sourceforge.net/lists/listinfo/pytables-users
>
>
> ------------------------------------------------------------------------------
> RSA(R) Conference 2012
> Save $700 by Nov 18
> Register now
> http://p.sf.net/sfu/rsa-sfdev2dev1
> _______________________________________________
> Pytables-users mailing list
> Pyt...@li...
> https://lists.sourceforge.net/lists/listinfo/pytables-users
>
>



-- 
Fernando Paolo
Institute of Geophysics & Planetary Physics
Scripps Institution of Oceanography
University of California, San Diego
9500 Gilman Drive
La Jolla, CA 92093-0225

web: fspaolo.net