[Python] Pandas cummax datetime when NaT values exist

I have a column of "Purchase Dates". The column either contains NaT or an actual date.

Date Last_Purchase Cummax_Purchase
2010-05-28 NaT NaT
2010-06-01 2010-06-01 2010-06-01
2010-06-02 2010-06-02 2010-06-02
2010-06-03 NaT NaT
2010-06-04 NaT NaT


I want to do a cummax() on the column such that it returns the most recent purchase date.

data['Purchase_Date'] = numpy.where(data['Buy Signal'] == True, data.index.astype(str), pandas.NaT)
data['Cummax_Purchase'] = pandas.to_datetime(data['Purchase_Date']).cummax()


The above cummax returns an NaT whenever their is an NaT in a corresponding row, not the cummax. But whenever I change the pandas.NaT to 0, then it works. But I want to return NaT values when there is an NaT.

Here's a small sample code:

data = pandas.DataFrame(
{"Purchase Dates" : pandas.to_datetime(['01-01-2020', '02-01-2020',None,'04-01-2020'])}, index=pandas.to_datetime(['01-01-2020','02-01-2020','03-01-2020','04-01-2020']))
data['Cummax_date'] = df['Purchase Dates'].cummax()

Continue reading...