Pandas 0.15.1 pandas.t0_datetime() sometimes returns pandas.core.series.Series and sometimes pandas.tseries.index.DatetimeIndex

[sorenwacker]

pandas.to_datetime() should not return different types.
Otherwise, all output should have methods like .head() and .tail() and .month()

---

import pandas as pd
import numpy as np
pd.options.display.mpl_style = 'default'
import matplotlib.pyplot as plt
x0['Month']=x0['Date'].month
data=np.random.rand(5)
data[4]=NaN
dfdata=pd.DataFrame({'Data':data})
times=['19990103','19990106','19990109','19990112','NaT']
dftimes = pd.to_datetime(times,format='%Y%m%d')

dftimes

Out[10]:
class 'pandas.tseries.index.DatetimeIndex'

dftimes.head

AttributeError Traceback (most recent call last)
in ()
----> 1 dftimes.head
AttributeError: 'DatetimeIndex' object has no attribute 'head'

---

'Download files'
import urllib
urllib.urlretrieve ("http://www.epa.gov/ttn/airs/airsaqs/detaildata/501files/Rd_501_88101_1999.Zip", "./data/Rd_501_88101_1999.zip")
'Unzip the files...'
import zipfile
import os.path
zfile = zipfile.ZipFile("./data/Rd_501_88101_1999.zip")
for name in zfile.namelist():
zfile.extract(name,'./data/')

import pandas as pd
import numpy as np
pd.options.display.mpl_style = 'default'
import matplotlib.pyplot as plt
%pylab inline

pm0=pd.io.parsers.read_table("./data/RD_501_88101_1999-0.txt","|",index_col=False,low_memory=False).drop(0)
x0=pm0[['Sample Value']].astype(float)
t0=pd.to_datetime(pm0['Date'],format='%Y%m%d')
print type(t0)

class 'pandas.core.series.Series'

t0.head()

1 1999-01-03
2 1999-01-06
3 1999-01-09
4 1999-01-12
5 1999-01-15
Name: Date, dtype: datetime64[ns]

[jreback]

No sometimes about it; Its a function of the type of the input.

With a list-like, you will get a DatetimeIndex
with a scalar, you will get a Timestamp
with a Series you will get a Series

So you get an output of like-kind to the input.

In [13]: pd.to_datetime(['20140101'])
Out[13]: 
<class 'pandas.tseries.index.DatetimeIndex'>
[2014-01-01]
Length: 1, Freq: None, Timezone: None

In [14]: pd.to_datetime(Series(['20140101']))
Out[14]: 
0   2014-01-01
dtype: datetime64[ns]

[jorisvandenbossche]

Maybe this should be added to the docstring?

[jorisvandenbossche]

@jreback @teramind see #8921