I have a .csv file that contains:

I want to convert these columns from datetime to UNIX timestamp.

For created_at, I was able to convert:

ndf["created_unix"] = pd.to_datetime(ndf["created_at"])
ndf["created_unix"] = (ndf["created_unix"] - pd.Timestamp("1970-01-01")) // pd.Timedelta("1s")

Which outputted:

This is exactly what I'm looking for, however following the same steps for column actual_delivery_time produces a different result:

ndf["actual_unix"] = pd.to_datetime(ndf["actual_delivery_time"])
ndf["actual_unix"] = (ndf["actual_unix"] - pd.Timestamp("1970-01-01")) // pd.Timedelta("1s")

This outputs:

It is readable by the Epoch converter, but it isn't the correct timestamp. Is there something I'm doing wrong?

I'm not sure why its producing a different result.

CodePudding user response：

I'm not sure why its producing a different result.

Consult ndf.dtypes.

You are complaining that one column is of type int, while the other is a float.

Likely one or more actual_delivery_time values were blank, and pandas represented that with a floating-point NaN.