Feat: User defined output frequency

[pgierz]

Closes #9

@chrisdane: Can you see if this does what you want, and is adequately documented?

[pgierz]

Documentation Preview: https://pymorize.readthedocs.io/en/feat-user_frequency/pymorize_saving_output.html

[chrisdane]

Whats the difference between YE and YS? Documentation is ok but cannot say much to the code.

[pgierz]

As far as I am aware, it controls the timestamp sampling (beginning vs. end) that is puts into the records:

>>> import xarray as xr
>>> ds = xr.tutorial.load_dataset("air_temperature")
>>> ds.resample(time="MS").groups
{numpy.datetime64('2013-01-01T00:00:00.000000000'): slice(0, 124, None),
 numpy.datetime64('2013-02-01T00:00:00.000000000'): slice(124, 236, None),
 numpy.datetime64('2013-03-01T00:00:00.000000000'): slice(236, 360, None),
 numpy.datetime64('2013-04-01T00:00:00.000000000'): slice(360, 480, None),
 numpy.datetime64('2013-05-01T00:00:00.000000000'): slice(480, 604, None),
 numpy.datetime64('2013-06-01T00:00:00.000000000'): slice(604, 724, None),
 numpy.datetime64('2013-07-01T00:00:00.000000000'): slice(724, 848, None),
 numpy.datetime64('2013-08-01T00:00:00.000000000'): slice(848, 972, None),
 numpy.datetime64('2013-09-01T00:00:00.000000000'): slice(972, 1092, None),
 numpy.datetime64('2013-10-01T00:00:00.000000000'): slice(1092, 1216, None),
 numpy.datetime64('2013-11-01T00:00:00.000000000'): slice(1216, 1336, None),
 numpy.datetime64('2013-12-01T00:00:00.000000000'): slice(1336, 1460, None),
 numpy.datetime64('2014-01-01T00:00:00.000000000'): slice(1460, 1584, None),
 numpy.datetime64('2014-02-01T00:00:00.000000000'): slice(1584, 1696, None),
 numpy.datetime64('2014-03-01T00:00:00.000000000'): slice(1696, 1820, None),
 numpy.datetime64('2014-04-01T00:00:00.000000000'): slice(1820, 1940, None),
 numpy.datetime64('2014-05-01T00:00:00.000000000'): slice(1940, 2064, None),
 numpy.datetime64('2014-06-01T00:00:00.000000000'): slice(2064, 2184, None),
 numpy.datetime64('2014-07-01T00:00:00.000000000'): slice(2184, 2308, None),
 numpy.datetime64('2014-08-01T00:00:00.000000000'): slice(2308, 2432, None),
 numpy.datetime64('2014-09-01T00:00:00.000000000'): slice(2432, 2552, None),
 numpy.datetime64('2014-10-01T00:00:00.000000000'): slice(2552, 2676, None),
 numpy.datetime64('2014-11-01T00:00:00.000000000'): slice(2676, 2796, None),
 numpy.datetime64('2014-12-01T00:00:00.000000000'): slice(2796, None, None)}
>>> ds.resample(time="ME").groups
{numpy.datetime64('2013-01-31T00:00:00.000000000'): slice(0, 124, None),
 numpy.datetime64('2013-02-28T00:00:00.000000000'): slice(124, 236, None),
 numpy.datetime64('2013-03-31T00:00:00.000000000'): slice(236, 360, None),
 numpy.datetime64('2013-04-30T00:00:00.000000000'): slice(360, 480, None),
 numpy.datetime64('2013-05-31T00:00:00.000000000'): slice(480, 604, None),
 numpy.datetime64('2013-06-30T00:00:00.000000000'): slice(604, 724, None),
 numpy.datetime64('2013-07-31T00:00:00.000000000'): slice(724, 848, None),
 numpy.datetime64('2013-08-31T00:00:00.000000000'): slice(848, 972, None),
 numpy.datetime64('2013-09-30T00:00:00.000000000'): slice(972, 1092, None),
 numpy.datetime64('2013-10-31T00:00:00.000000000'): slice(1092, 1216, None),
 numpy.datetime64('2013-11-30T00:00:00.000000000'): slice(1216, 1336, None),
 numpy.datetime64('2013-12-31T00:00:00.000000000'): slice(1336, 1460, None),
 numpy.datetime64('2014-01-31T00:00:00.000000000'): slice(1460, 1584, None),
 numpy.datetime64('2014-02-28T00:00:00.000000000'): slice(1584, 1696, None),
 numpy.datetime64('2014-03-31T00:00:00.000000000'): slice(1696, 1820, None),
 numpy.datetime64('2014-04-30T00:00:00.000000000'): slice(1820, 1940, None),
 numpy.datetime64('2014-05-31T00:00:00.000000000'): slice(1940, 2064, None),
 numpy.datetime64('2014-06-30T00:00:00.000000000'): slice(2064, 2184, None),
 numpy.datetime64('2014-07-31T00:00:00.000000000'): slice(2184, 2308, None),
 numpy.datetime64('2014-08-31T00:00:00.000000000'): slice(2308, 2432, None),
 numpy.datetime64('2014-09-30T00:00:00.000000000'): slice(2432, 2552, None),
 numpy.datetime64('2014-10-31T00:00:00.000000000'): slice(2552, 2676, None),
 numpy.datetime64('2014-11-30T00:00:00.000000000'): slice(2676, 2796, None),
 numpy.datetime64('2014-12-31T00:00:00.000000000'): slice(2796, None, None)}

[chrisdane]

Ok. Is the middle of a time period also defined? I cannot find this in the documentation.

The middle is usually used, so e.g. yyyy-mm-15 or -16 in case of monthly output. See e.g. the documentation from cdo (e.g. cdo -h monmean):

DESCRIPTION
The time of outfile is determined by the time in the middle of all contributing timesteps of infile.
This can be change with the CDO option --timestat_date <first|middle|last>.

[pgierz]

Is the middle of a time period also defined?

Not natively, we would need to implement that as a separate correction step. I'll see what I can do.

[pgierz]

Alright, I have something that sort-of works:

https://pymorize.readthedocs.io/en/feat-user_frequency/api/pymorize.html#pymorize.timeaverage.custom_resample

Maybe @siligam can have a look and hook it up to the rest of his averaging functionality.