to_datetime passed string return datetime rather than Timestamp

[hayd]

It seems it returns different objects when passed array vs string:

In [1]: pd.to_datetime(['2012'])[0]
Out[1]: Timestamp('2012-01-01 00:00:00', tz=None)

In [2]: pd.to_datetime('2012')
Out[2]: datetime.datetime(2012, 1, 1, 0, 0)

Also worth mentioning (though different):

In [3]: pd.Timestamp('2012')
Out[3]: Timestamp('2012-06-13 00:00:00', tz=None)

Seems a little inconsistent... (cc #3859)

[jreback]

to_datetime('')
NaT

now....before return the empty_string...

better?

[hayd]

So it'll always do Timestamps? It's also weird it justs lets junk/integers through...

In [1]: pd.to_datetime('ssds')
Out[1]: 'ssds'

In [2]: pd.Timestamp(0)
Out[2]: Timestamp('1970-01-01 00:00:00', tz=None)

In [3]: pd.to_datetime(0)
Out[3]: 0

[jreback]

with 0 now returns a Timestamp, so that I think is ok

the strings OTOH, you need to supply error='raise', default
is ignore, change the default?

??

In [2]: pd.to_datetime('ssds')
Out[2]: 'ssds'

In [4]: pd.to_datetime('ssds',errors='raise')
ValueError: unknown string format

In [5]: pd.to_datetime(0,errors='raise')
Out[5]: Timestamp('1970-01-01 00:00:00', tz=None)

In [6]: pd.to_datetime(0)
Out[6]: Timestamp('1970-01-01 00:00:00', tz=None)

[hayd]

Lol, I should have read the docstring there. Yeah, can't think why leaving it untouched would ever be what's wanted. ?

[hayd]

Thoughts on this one:

In [56]: pd.Timestamp('2012')
Out[56]: Timestamp('2012-06-13 00:00:00', tz=None)

(picks todays date, but on the year of the string...)

Surely to_datetime(x) should be the same as map(Tinestamp, x) ?