[Numpy-discussion] RFC: A proposal for implementing some date/time types in NumPy

[Francesc Alted]

A Friday 11 July 2008, Pierre GM escrigué:
> On Friday 11 July 2008 12:28:30 Anne Archibald wrote:
> > A date/time class that tries to do everything is quickly going to
> > become unfeasibly complicated.
>
> I quite agree, and I think that's why Francesc and Ivan are
> considering different classes for different problems: one targetting
> series at a frequency at most daily, and some others when sub-second
> resolution is needed.

Exactly, that's the idea: two of the proposed types has a fixed 
sub-second resolution. ``datetime64`` has microsecond precision, while 
``timestamp64`` has a precision of nanoseconds (at the expenses of a 
much more limited time span).

For people needing geological or astronomical time scales the 
``timefloat64`` comes handy (at the expense of loosing resolution for 
large time spans).

Cheers,

-- 
Francesc Alted