I have a .csv file that contains:
| created_at | actual_delivery_time |
|---|---|
| 2015-02-06 22:24:17 | 2015-02-06 23:27:16 |
| 2015-02-10 21:49:25 | 2015-02-10 22:56:29 |
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:
| created_unix |
|---|
| 1423261457 |
| 1423604965 |
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:
| actual_unix |
|---|
| 1.423265e+09 |
| 1.423609e+09 |
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.
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.