availability of nanosecond periods in DatetimeIndex

[wuan]

Hi there,

when trying to create a DatetimeIndex for data sampled in the nanosecond range

pd.DatetimeIndex(start=pd.Timestamp('2012-12-01'), periods=10, freq=pd.offsets.Nano(2800))

there appears an error due to an uninitialized timestamp.

Inside the pandas code python datetime objects seem to be used for time representation of "slower" than nanosecond times, whereas the nanosecond period seems to appear as a simple int objects which cannot be handled by the internal arithmetics at the moment.

Is there a plan to switch to nanosecond aware time objects internally? As I would like to use this great library for fast data as well I can also offer to help working on that.

Best regards
Andreas

[changhiskhan]

Hi Andreas, we'd love to get your contribution on getting the nano second offsets to play nicer. There are a number of places where the internal arithmetic would have to change to handle the nano-second delta, some of which needs to be in the cython code.
Please take a look and let us know if it's something you'd have time to work on. A PR would be very welcome.

[wuan]

Hi,

in the meantime there is a working version of nanosecond period in my branch. Unfortunately getting a sub-second frequency string like freq='2800N' to work required some kind of rework of src/period.c which is still work in progress.

Yes, I would be glad to contribute to the project.

Regards
Andreas