BUG: dataframe.apply() loops on first row when applied method attempts to modify the row

[nicolasrozain]

With pandas 1.1.0 on Python 3.6.8:

>>> import pandas as pd
>>> df = pd.DataFrame({'a': list(range(0,100)), 'b': list(range(100,200))})
>>> def func(row):
...     row.loc['a'] += 1
...     return row
... 
>>> df
     a    b
0    0  100
1    1  101
2    2  102
3    3  103
4    4  104
..  ..  ...
95  95  195
96  96  196
97  97  197
98  98  198
99  99  199
[100 rows x 2 columns]
>>> df.apply(func, axis=1)
      a    b
0   100  100
1   100  100
2   100  100
3   100  100
4   100  100
..  ...  ...
95  100  100
96  100  100
97  100  100
98  100  100
99  100  100
[100 rows x 2 columns]
>>> df
      a    b
0   100  100
1     1  101
2     2  102
3     3  103
4     4  104
..  ...  ...
95   95  195
96   96  196
97   97  197
98   98  198
99   99  199
[100 rows x 2 columns]
>>> pd.show_versions()
INSTALLED VERSIONS
------------------
commit           : d9fff2792bf16178d4e450fe7384244e50635733
python           : 3.6.8.final.0
python-bits      : 64
OS               : Windows
OS-release       : 10
Version          : 10.0.17763
machine          : AMD64
processor        : Intel64 Family 6 Model 142 Stepping 10, GenuineIntel
byteorder        : little
LC_ALL           : None
LANG             : None
LOCALE           : None.None
pandas           : 1.1.0
numpy            : 1.18.2
pytz             : 2019.3
dateutil         : 2.8.1
pip              : 20.0.2
setuptools       : 42.0.2
Cython           : None
pytest           : 5.4.1
hypothesis       : None
sphinx           : None
blosc            : None
feather          : None
xlsxwriter       : None
lxml.etree       : 4.5.0
html5lib         : 1.0.1
pymysql          : None
psycopg2         : None
jinja2           : None
IPython          : None
pandas_datareader: None
bs4              : 4.9.1
bottleneck       : 1.3.2
fsspec           : None
fastparquet      : None
gcsfs            : None
matplotlib       : 3.2.1
numexpr          : 2.7.1
odfpy            : None
openpyxl         : 2.5.14
pandas_gbq       : None
pyarrow          : None
pytables         : None
pyxlsb           : None
s3fs             : None
scipy            : 1.3.3
sqlalchemy       : None
tables           : None
tabulate         : None
xarray           : None
xlrd             : 1.2.0
xlwt             : 1.3.0
numba            : None

With pandas 1.0.5:

>>> import pandas as pd
>>> df = pd.DataFrame({'a': list(range(0,100)), 'b': list(range(100,200))})
>>> def func(row):
...     row.loc['a'] += 1
...     return row
... 
>>> df
     a    b
0    0  100
1    1  101
2    2  102
3    3  103
4    4  104
..  ..  ...
95  95  195
96  96  196
97  97  197
98  98  198
99  99  199
[100 rows x 2 columns]
>>> df.apply(func, axis=1)
      a    b
0     1  100
1     2  101
2     3  102
3     4  103
4     5  104
..  ...  ...
95   96  195
96   97  196
97   98  197
98   99  198
99  100  199
[100 rows x 2 columns]
>>> df
      a    b
0     1  100
1     2  101
2     3  102
3     4  103
4     5  104
..  ...  ...
95   96  195
96   97  196
97   98  197
98   99  198
99  100  199
[100 rows x 2 columns]
>>> pd.show_versions()
INSTALLED VERSIONS
------------------
commit           : None
python           : 3.6.8.final.0
python-bits      : 64
OS               : Windows
OS-release       : 10
machine          : AMD64
processor        : Intel64 Family 6 Model 142 Stepping 10, GenuineIntel
byteorder        : little
LC_ALL           : None
LANG             : None
LOCALE           : None.None
pandas           : 1.0.5
numpy            : 1.18.2
pytz             : 2019.3
dateutil         : 2.8.1
pip              : 20.0.2
setuptools       : 42.0.2
Cython           : None
pytest           : 5.4.1
hypothesis       : None
sphinx           : None
blosc            : None
feather          : None
xlsxwriter       : None
lxml.etree       : 4.5.0
html5lib         : 1.0.1
pymysql          : None
psycopg2         : None
jinja2           : None
IPython          : None
pandas_datareader: None
bs4              : 4.9.1
bottleneck       : 1.3.2
fastparquet      : None
gcsfs            : None
lxml.etree       : 4.5.0
matplotlib       : 3.2.1
numexpr          : 2.7.1
odfpy            : None
openpyxl         : 2.5.14
pandas_gbq       : None
pyarrow          : None
pytables         : None
pytest           : 5.4.1
pyxlsb           : None
s3fs             : None
scipy            : 1.3.3
sqlalchemy       : None
tables           : None
tabulate         : None
xarray           : None
xlrd             : 1.2.0
xlwt             : 1.3.0
xlsxwriter       : None
numba            : None

I expected the behavior of 1.0.5 in 1.1.0, did I misunderstood the apply method?
Thank you for your help.

[simonjayhawkins]

Thanks @nicolasrozain for the report. is adding row=row.copy() to the function before the mutation a suitable workaround in the short term?

[simonjayhawkins]

@nicolasrozain in 1.0.5 the original DataFrame is being mutated by this function so removing the regression tag pending further discussion/investigation of correct behaviour or whether we want to restore the 1.0.5 behaviour.

>>> pd.__version__
'1.0.5'
>>>
>>> df = pd.DataFrame({"a": list(range(0, 3)), "b": list(range(100, 103))})
>>> orig = df.copy()
>>>
>>>
>>> def func(row):
...     row.loc["a"] += 1
...     return row
...
>>>
>>> df
   a    b
0  0  100
1  1  101
2  2  102
>>>
>>> res = df.apply(func, axis=1)
>>> print(res)
   a    b
0  1  100
1  2  101
2  3  102
>>>
>>> df
   a    b
0  1  100
1  2  101
2  3  102
>>>
>>> tm.assert_frame_equal(df, orig)
Traceback (most recent call last):
...
AssertionError: DataFrame.iloc[:, 0] (column name="a") are different

[simonjayhawkins]

OK so the commit that caused the change is not the one I expected. it is #34909. This PR is labelled PERF and so should not have changed behaviour. re-instating regression tag. cc @jbrockmendel

91802a9 is the first bad commit
commit 91802a9
Author: jbrockmendel jbrockmendel@gmail.com
Date: Thu Jun 25 16:06:10 2020 -0700

PERF: avoid creating many Series in apply_standard (#34909)

[GusBite]

I have experienced the same behaviour. Modifying a copy of the series instead of the original solved the issue in the short term. I guess i'll wait a fix before enjoying the long waited 'apply does not alterate first row twice' functionnality. Keep up the good work guys.

[seal9]

Same problem for me, I feel like the 1.0.5 behavior was much better. I often use the apply method to modify rows like this, as I feel the ability to call an arbitary Python function on any row is very powerful. I'll downgrade to 1.0.5 until this is fixed.