convert integer to datetime bug, Python int too large to convert to C long

[poc7667]

pandas version pandas (0.16.1)

for field in ["release_date", "date"]:
    if field in df:
        df[field] = pd.to_datetime(df[field], unit="ms")

[jorisvandenbossche]

Which version are you using?

In [30]: pd.to_datetime(1430784000000, unit='ms')
Out[30]: Timestamp('2015-05-05 00:00:00')

In [31]: pd.__version__
Out[31]: '0.16.2'

[jorisvandenbossche]

Sorry, you did specify the version. Can you give a reprodicible example? (make up some dummy data that gives you this error)
For a small example (see above), it works for me on 0.16.2

[jreback]

ok on 0.16.1 as well.

In [1]: pd.to_datetime(1430784000000, unit='ms')
Out[1]: Timestamp('2015-05-05 00:00:00')

In [2]: pd.__version__
Out[2]: '0.16.1'

You get this error if the unit you pass is incompatible

In [4]: pd.to_datetime(1430784000000, unit='s')
OverflowError: Python int too large to convert to C long

[wmacmillan]

Just to clarify on @jreback's point here, if the time since epoch is recorded in milliseconds, and you select seconds as your unit, you get this very unhelpful error.

[jreback]

yeah, we should prob just catch this and re-raise as a more helpful error message. Want to do a PR?

[dreamgonfly]

Here is the error message i got when i tried to regenerate this issue. It looks like it already has been solved. Should we close this issue?

In [1]: pd.to_datetime(1430784000000, unit='s')
Out[1]: OutOfBoundsDatetime: cannot convert input with unit 's'

In [2]: pd.__version__
Out[2]: '0.18.1'