Inconsistant datatype with type()?

[grgsolymosi]

[in]

for cl in train.columns:
     print cl, type(train.loc[0, cl])
     print cl, type(train[cl][0])
     break

[out]

record_ID <type 'numpy.float64'> 
record_ID <type 'numpy.int64'>

Though I can solve it in many cases with train[cl].dtype, which gives the correct output:

[out]

record_ID int64

But if we work with string data types the output will be object instead of str as data type.

Something shortstop by me or what happens?
Thank you.

[jreback]

pls show train.info(), train.head() and pd.show_versions()

[jorisvandenbossche]

A smaller reproducible example:

In [5]: df = pd.DataFrame({'a':[0,1,2], 'b':[0.1,0.2,0.3]})

In [6]: df
Out[6]:
   a    b
0  0  0.1
1  1  0.2
2  2  0.3

In [7]: type(df.loc[0, 'a'])
Out[7]: numpy.float64

In [8]: type(df['a'][0])
Out[8]: numpy.int64

[jreback]

dupe of #11617

[grgsolymosi]

@jreback0
Why the "dupe" label, please?
@jorisvandenbossche 's example shows the same issue, doesn't it?

[jreback]

dupe means duplicate. See the referenced issue which is the same. We get a fair number of duplicates here (no fault of the poster of course). So try to have a single issue.

If you look at the referenced issue and it IS not the same, pls let me know.

[grgsolymosi]

@jreback thank you for the info and your efforts.

[grgsolymosi]

@jreback
Just one more info as I saw, that the other issue was on other system config:
My system:
Python 2.7.11
Ubuntu 14.04 LTS

[jreback]

@temmplar ok, these ops are not system dependent. but thanks for the info.

[grgsolymosi]

@jreback Furthermore I have some other additional info, which maybe would be a help to backcode the bug, if you need it, please contact me.