How can I do an interpolating reindex in pandas us

问题:

I have a series with a datetime index, and what I'd like is to interpolate this data using some other, arbitrary datetime index. Essentially what I want is how to make the following code snippet more or less work:

from pandas import Series
import datetime

datetime_index = [datetime.datetime(2010, 1, 5), datetime.datetime(2010, 1, 10)]
data_series = Series([5, 15], [datetime.datetime(2010, 1, 5), datetime.datetime(2010, 1, 15)])

def interpolating_reindex(data_series, datetime_index):
    """?????"""

goal_series = interpolating_reindex(data_series, datetime_index) 

assert(goal_series == Series([5, 10], datetime_index))

reindex doesn't do what I want because it can't interpolate, and also my series might not have the same indices anyway. resample isn't what I want because I want to use an arbitrary, already defined index which isn't necessarily periodic. I've also tried combining indices using Index.join in the hopes that I could then do reindex and then interpolate, but that didn't work as I expected. Any pointers?

回答1:

Try this:

from pandas import Series
import datetime

datetime_index = [datetime.datetime(2010, 1, 5), datetime.datetime(2010, 1, 10)]
s1 = Series([5, 15], [datetime.datetime(2010, 1, 5), datetime.datetime(2010, 1, 15)])
s2 = Series(None, datetime_index)
s3 = s1.combine_first(s2)
s3.interpolate()

Based on the comments, the result interpolated to the target index would be:

goal_series  = s3.interpolate().reindex(datetime_index)

assert((goal_series == Series([5, 10], datetime_index)).all())