DataFrame.set_value() fails with numpy types

[dov]

Code Sample, a copy-pastable example if possible

import pandas as pd
import numpy as np

df = pd.DataFrame([[0,0,None],
                   [0,0,None],
                   [1,1,None]],
                  columns = ['A','B','C']).set_index(['A','B'])

# This succeeds
idx = (0,0)
df.set_value(idx,'C',5)
print df

# This fails
idx = (np.int64(0),np.int64(0))
df.set_value(idx,'C',8)
print df

Problem description

The above does a multirow assignment through the DataFrame.set_value() function by two different indices. In the first case the index is built by plain python values. This succeeds. In the second case, which is superficially equal, the index is built with numpy.int64 integers (that I actually got back from a groupby call). This case fails.

Is this behavior expected or is this a bug?

Expected Output

        C
A B      
0 0     5
  0     5
1 1  None
        C
A B      
0 0     8
  0     8
1 1  None

Output of pd.show_versions()

[paste the output of pd.show_versions() here below this line]

INSTALLED VERSIONS

commit: None
python: 2.7.13.final.0
python-bits: 64
OS: Linux
OS-release: 4.11.9-300.fc26.x86_64
machine: x86_64
processor: x86_64
byteorder: little
LC_ALL: None
LANG: None
LOCALE: en_US.UTF-8

pandas: 0.20.1
pytest: 3.1.1
pip: 9.0.1
setuptools: 36.2.7
Cython: None
numpy: 1.12.0
scipy: 0.19.1
xarray: None
IPython: 5.4.1
sphinx: 1.6.3
patsy: None
dateutil: 2.6.0
pytz: 2016.10
blosc: None
bottleneck: 1.2.1
tables: 3.3.0
numexpr: 2.6.1
feather: None
matplotlib: 2.0.0
openpyxl: None
xlrd: 0.9.4
xlwt: 1.1.2
xlsxwriter: None
lxml: None
bs4: 4.6.0
html5lib: 0.999
sqlalchemy: 1.1.11
pymysql: None
psycopg2: None
jinja2: 2.9.6
s3fs: None
pandas_gbq: None
pandas_datareader: None

[chris-b1]

This probably should work, but more idiomatic to use loc which handles this already

In [27]: df.loc[(np.int64(0),np.int64(0)), 'C'] = 5

In [28]: df
Out[28]: 
        C
A B      
0 0     5
  0     5
1 1  None

[chris-b1]

xref #15269 - proposal to deprecate set_value

[jorisvandenbossche]

Closing this, as we deprecated set_value/get_value

(although at has the same behaviour, but I would say that if you want to use numpy scalars, you need to use loc, to keep at as the more advanced but faster one)