Nullable integer documentation examples are incorrect

[toniprada]

Code Sample

import pandas as pd
import numpy as np

print(pd.array([23842938553820651, np.nan], dtype=pd.Int64Dtype())) # from the docs
print()
print(pd.array([23842938553820651, pd.NaT], dtype=pd.Int64Dtype()))

Problem description

Examples in the documentation of Nullable integers don't work as expected https://pandas.pydata.org/pandas-docs/stable/user_guide/integer_na.html

The integers are converted to float before casting to the new Nullable integer, so they lose precision anyway (which is the the most usual reason for using them, as stated in the docs). Using the native pd.NaT solves the problem.

Expected Output

<IntegerArray>
[23842938553820651, NaN]
Length: 2, dtype: Int64

<IntegerArray>
[23842938553820651, NaN]
Length: 2, dtype: Int64

Output of code sample

<IntegerArray>
[23842938553820652, NaN]
Length: 2, dtype: Int64

<IntegerArray>
[23842938553820651, NaN]
Length: 2, dtype: Int64

Output of pd.show_versions()

commit           : None

pandas           : 0.25.3
numpy            : 1.17.4
pytz             : 2018.9
dateutil         : 2.8.0
pip              : 19.0.3
setuptools       : 41.0.1
Cython           : None
pytest           : None
hypothesis       : None
sphinx           : None
blosc            : None
feather          : None
xlsxwriter       : None
lxml.etree       : None
html5lib         : None
pymysql          : None
psycopg2         : None
jinja2           : 2.10.1
IPython          : 7.5.0
pandas_datareader: None
bs4              : None
bottleneck       : None
fastparquet      : None
gcsfs            : None
lxml.etree       : None
matplotlib       : 3.1.1
numexpr          : None
odfpy            : None
openpyxl         : None
pandas_gbq       : None
pyarrow          : 0.14.1
pytables         : None
s3fs             : None
scipy            : None
sqlalchemy       : 1.3.2
tables           : None
xarray           : None
xlrd             : None
xlwt             : None
xlsxwriter       : None

[TomAugspurger]

The np.nan is causing in intermediate floating-point ndarray to be created. And that value is to large to be represented exactly as a float.

Perhaps we should update the docs to use None instead of np.nan there. That works correctly

In [3]: print(pd.array([23842938553820651, None], dtype=pd.Int64Dtype()))
<IntegerArray>
[23842938553820651, NaN]
Length: 2, dtype: Int64

OTOH, that's going to create an intermediate object-dtype ndarray, which will be slow for large arrays.

[lucassa3]

@TomAugspurger would this simple edit be enough to make the docs coherent with expected functionality?