cmf.timeseries should get a from_pandas factory function about cmf

[philippkraft]

424ff31 creates the from_pandas function with this code:

cmf/cmf/timetools.py

Lines 229 to 253 in 71926a8

	def pandas_series_to_timeseries(cls, series, regularize='warning'):
	import numpy as np
	if (series.index[1:] - series.index[:-1] == series.index[1] - series.index[0]).all():
	s = series
	else:
	if regularize:
	s = series.resample(series.index[1] - series.index[0]).interpolate('time')
	if regularize in ['w', 'warn', 'warning']:
	from logging import warning
	warning(('While converting pandas series "{n}" to cmf.timeseries, '
	'data has been regularized and consists now of {o} original values '
	'and {i} interpolated values').format(
	n=series.name, o=series.count(), i=s.count() - series.count())
	)
	else:
	raise ValueError('The pandas timeseries is not regular. '
	'Try cmf.timeseries.from_series(series, regularize=True)'
	'to allow for interpolation')
	tstep = sec * (s.index[1] - s.index[0]).total_seconds()
	tstart = datetime_to_cmf(s.index[0].to_pydatetime())
	array = np.array(s)
	return cls.from_array(tstart, tstep, array)

and attaches it to cmf.timeseries here:

cmf/cmf/timetools.py

Line 269 in 71926a8

timeseries.from_pandas = classmethod(pandas_series_to_timeseries)

[philippkraft]

Usage:

import pandas as pd
import cmf
import numpy as np
ts = cmf.timeseries.from_array(cmf.Time(),cmf.h, np.random.uniform(0,1,1000))    
pts = ts.to_pandas()                                                             
ts2 = cmf.timeseries.from_pandas(pts)