CODEWITHSUNDEEP
pythonpandasnumpydatetime
bwk
bwk

Reputation: 816

How do I get the minimum of a pandas datetime series and the (constant) max of another datetime series?

Suppose I have two datetime series:

foo = pd.to_datetime(pd.Series([
    '2020-01-01 12:00:00',
    '2020-02-02 23:12:00'
]))

bar = pd.to_datetime(pd.Series([
    '2020-01-20 01:02:03',
    '2020-01-30 03:02:01'
]))

Both are of type datetime64[ns]:

>>> foo
0   2020-01-01 12:00:00
1   2020-02-02 23:12:00
dtype: datetime64[ns]

>>> bar
0   2020-01-20 01:02:03
1   2020-01-30 03:02:01
dtype: datetime64[ns]

For each element in foo, I want to get the minimum of:

  1. That value from foo
  2. The (constant) maximum value of bar

But this produces a TypeError:

>>> np.minimum(foo, bar.max())
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
...
TypeError: '<=' not supported between instances of 'int' and 'Timestamp'

It works if I just do the Series themselves:

>>> np.minimum(foo, bar)
0   2020-01-01 12:00:00
1   2020-01-30 03:02:01
dtype: datetime64[ns]

bar.max() returns a Timestamp for some reason, instead of a datetime64, but even using an explicit python datetime object doesn't work. Why is numpy considering foo to be an int? Is there a way around this?

Upvotes: 1

Views: 1610

Answers (2)

Malcolm
Malcolm

Reputation: 839

How about

>>> barmax = bar.max()
>>> barmax
Timestamp('2020-01-30 03:02:01')
>>> foo.map(lambda x: np.minimum(x, barmax))
0   2020-01-01 12:00:00
1   2020-01-30 03:02:01
dtype: datetime64[ns]
>>>

Upvotes: 0

noah
noah

Reputation: 2776

Using pandas.Series.where:

foo.where(foo < bar.max(), bar.max())

This replaces values of foo with bar.max() if the condition (foo < bar.max()) is False.

Upvotes: 3

Related Questions