BUG: Bug in .to_datetime() when passing integers or floats, no unit and errors=coerce

[jreback]

closes #13180

[jreback]

@jorisvandenbossche

note that prior to 0.18.1 we would handle

pd.to_datetime([1]) with unit='ns' as the default. This PR addresses this, but after looking again, should we make the user specify it here (when it IS ns), e.g. as in the accompanied issue.

[jreback]

any comments @jorisvandenbossche

[melzoghbi]

I have pandas version 0.19.2 and i still get this error when using to_datetime() method. My column type is int64 and here is the error message when i use the following:

1. pd.to_datetime(ds['createdTime'], unit='s') --> OutOfBoundsDatetime: cannot convert input with unit 's'

2. pd.to_datetime(ds['createdTime']) --> no conversion

Any idea how to fix this? Thanks a lot.

[jorisvandenbossche]

@melzoghbi if you still have an issue, please open a new issue (as your example is not related to the above issue). However, please first make sure to test on the latest pandas (0.21.0) and make sure to have an actual reproducible examples (so copy-pasteable code, not as an image).

Further, for me this actually converts to a datetime:

In [121]: s = pd.Series([1508258340299])

In [122]: pd.to_datetime(s)
Out[122]: 
0   1970-01-01 00:25:08.258340299
dtype: datetime64[ns]

[melzoghbi]

Thanks for your help and this actually shows the datetime but not the actual date that i am looking for.

For example:
import pandas as pd
s = pd.Series([1510675225])
pd.to_datetime(s)

Should show a date as 11/14/27 but it actually shows epoch time as of 1970 datetime.

Any way i will open a new issue for this.

[hashamMunir]

issue persists in pd 0.23.3
converting the unixtime to date time.
1970-01-01 00:00:01.529009633
it returns 1970

[pushyamig]

this is still an issue in the Pandas 0.23.4, is there any workaround?

[jorisvandenbossche]

This is a closed issue that is fixed. If you think there is still a bug, please open a new issue with a reproducible example.

If it is about the year 1970, that is not a bug but because to_datetime interprets integers as nanoseconds by default:

In [11]: s = pd.Series([1510675225])

In [12]: pd.to_datetime(s)
Out[12]: 
0   1970-01-01 00:00:01.510675225
dtype: datetime64[ns]

In [13]: pd.to_datetime(s, unit='s')
Out[13]: 
0   2017-11-14 16:00:25
dtype: datetime64[ns]