CODEWITHSUNDEEP
pythonpandas
stackflower10
stackflower10

Reputation: 25

np.where with datetimes producing Unix timestamp

I have a dataframe with a column that has time stamp data with some nulls. I am trying to replace the nulls with the earliest date in the column using np.where.

The dataframe looks like this:

index   date
1       2019-06-30 22:40:25.799000+00:00
2       2019-06-30 22:40:25.799000+00:00
3       NaN

I'm writing my code as:

mini = df['date'].min()
df['date'] = np.where(df['date'].isnull(), mini, df['date'])

but the resulting column for date gives me a Unix Timestamp, with the NaN's filled correctly:

index       date
1       1552685510470841000
2       1555706405810536000
3       2015-05-07 13:49:51.072000+00:00

Why does this happen and how do I get it to all be timestamps?

Upvotes: 1

Views: 161

Answers (1)

Code Different
Code Different

Reputation: 93181

numpy has a bias toward treating array's elements as floats. It saw NaN and Timestamp are both representable as floats so it converts df['date'] to float.

You can use fillna instead:

df['date'].fillna(mini, inplace=True)

Upvotes: 2

Related Questions