DataFrame.iat will create new column if .iat is used to set None on int Series

[jimmywan]

Code Sample, a copy-pastable example if possible

>>> df = pd.DataFrame({'a':[0,1],'b':[4,5]})
>>> df
   a  b
0  0  4
1  1  5
>>> df.iat[0, 0] = None
>>> df
   a  b   0
0  0  4 NaN
1  1  5 NaN

Problem description

This is problematic for multiple reasons.

• inconsistency between iloc and iat.
• creation of a brand new column is almost surely not the intended/expected behavior
• At the very least, I would expect it to simply bail on the operation with a warning about incompatible types.

This is likely related to the non-intuitive behavior of Series which has already been documented here:
#20643 (comment)

Expected Output

I would expect it to do what it does when using .iloc:

>>> df = pd.DataFrame({'a':[0,1],'b':[4,5]})
>>> df
   a  b
0  0  4
1  1  5
>>> df.iloc[0, 0] = None
>>> df
     a  b
0  NaN  4
1  1.0  5

Output of pd.show_versions()

INSTALLED VERSIONS ------------------ commit: None python: 3.6.6.final.0 python-bits: 64 OS: Linux OS-release: 3.13.0-24-generic machine: x86_64 processor: x86_64 byteorder: little LC_ALL: None LANG: en_US.utf8 LOCALE: en_US.UTF-8

pandas: 0.22.0
pytest: None
pip: 9.0.3
setuptools: 39.0.1
Cython: 0.27.3
numpy: 1.14.0
scipy: 1.0.0
pyarrow: None
xarray: None
IPython: None
sphinx: None
patsy: 0.5.0
dateutil: 2.6.1
pytz: 2017.3
blosc: None
bottleneck: None
tables: None
numexpr: None
feather: None
matplotlib: None
openpyxl: None
xlrd: None
xlwt: None
xlsxwriter: None
lxml: None
bs4: None
html5lib: None
sqlalchemy: None
pymysql: None
psycopg2: None
jinja2: None
s3fs: 0.1.2
fastparquet: None
pandas_gbq: None
pandas_datareader: None

[jimmywan]

I tried it with 0.23.4 and had the same output.

INSTALLED VERSIONS ------------------ commit: None python: 3.6.6.final.0 python-bits: 64 OS: Linux OS-release: 3.13.0-24-generic machine: x86_64 processor: x86_64 byteorder: little LC_ALL: None LANG: en_US.utf8 LOCALE: en_US.UTF-8

pandas: 0.23.4
pytest: None
pip: 9.0.3
setuptools: 39.0.1
Cython: 0.27.3
numpy: 1.14.0
scipy: 1.0.0
pyarrow: None
xarray: None
IPython: None
sphinx: None
patsy: 0.5.0
dateutil: 2.6.1
pytz: 2017.3
blosc: None
bottleneck: None
tables: None
numexpr: None
feather: None
matplotlib: None
openpyxl: None
xlrd: None
xlwt: None
xlsxwriter: None
lxml: None
bs4: None
html5lib: None
sqlalchemy: None
pymysql: None
psycopg2: None
jinja2: None
s3fs: 0.1.2
fastparquet: None
pandas_gbq: None
pandas_datareader: None

[mroeschke]

Thanks for the report. Investigations and PRs welcome!

[aditya0811]

I think we can conclude that as df.iat is capable of creating and setting values,it will create index for 0 which is not present as column index(a,b).
We cannot access those values using iat if column index are char type.

[mroeschke]

iat indexes by integer position and not label, so it shouldn't matter if 0 is not in the columns; it should modify the value in row 0 column A

IIRC iat and at doesn't perform as many data validation checks, so this may be a "fallback" assignment and broadcasting.

[jimmywan]

I do not agree with @aditya0811 :

I think we can conclude that as df.iat is capable of creating and setting values,it will create index for 0 which is not present as column index(a,b).

As explained by @mroeschke here:

iat indexes by integer position and not label, so it shouldn't matter if 0 is not in the columns; it should modify the value in row 0 column A