Updating a DataFrame by iteratively indexing into a columns

[Jenders74]

I am initializing a DataFrame with 0 and then update it by iteratively indexing into indvidual columns. The behavior of my code has changed with pandas 0.13.0 such that resulting DataFrame out[['A']] remains 0 but series out['A'] has the correct values:

>>> print out[['A']]
            A
2014-05-07  0
2014-05-08  0
2014-05-09  0

 >>> print out['A']
2014-05-07    600
2014-05-08    600
2014-05-09    600

Is this a bug?

import pandas
#initialize a DataFrame with 0 values
out = pandas.DataFrame({'A': [0, 0, 0]})
out.index = pandas.date_range('5/7/2014', '5/9/2014')

#DataFrame to update out with
df = pandas.DataFrame({'C': ['A', 'A', 'A'], 'D': [100, 200, 300]})

#loop through df to update out
for ix, row in df.iterrows():
    six = pandas.Timestamp('5/7/2014')
    eix = pandas.Timestamp('5/9/2014')
    out[row['C']][six:eix] = out[row['C']][six:eix] + row['D']

print out
print out[['A']]
print out['A']

INSTALLED VERSIONS
------------------
Python: 2.7.6.final.0
OS: Windows
Release: 7
Processor: Intel64 Family 6 Model 58 Stepping 9, GenuineIntel
byteorder: little
LC_ALL: None
LANG: en_US

pandas: 0.13.0
Cython: 0.20
Numpy: 1.7.1
Scipy: 0.12.0
statsmodels: 0.5.0
    patsy: 0.2.1
scikits.timeseries: Not installed
dateutil: 1.5
pytz: 2013.9
bottleneck: 0.7.0
PyTables: 3.1.0rc2
    numexpr: 2.2.2
matplotlib: 1.3.1
openpyxl: 1.8.2
xlrd: 0.9.2
xlwt: 0.7.5
xlsxwriter: Not installed
sqlalchemy: Not installed
lxml: Not installed
bs4: 4.3.2
html5lib: Not installed
bigquery: Not installed
apiclient: Not installed

[cpcloud]

This kind of looks like a regression. I say "kind of", because you should avoid this kind of indexing like the plague but also because we like to keep backcompat and breaking this is kind against our de facto rules. I'll call this a bug.

[cpcloud]

For future reference, instead of

out[row['C']][six:eix] = out[row['C']][six:eix] + row['D']

do

out.loc[six:eix, row['C']] += row['D']

[jreback]

this is chained indexing and hence is not guaranteed to work
it happens to work sometimes
don't do this

not a bug

[jorisvandenbossche]

@jreback just curious, but how is it possible that out['A'] and out[['A']] give different results?

[jtratner]

Isn't this a caching bug?

[Jenders74]

@cpcloud nice one! thanks a ton.

[jreback]

@Jenders74 have a read here, you should never use chain indexing

http://pandas-docs.github.io/pandas-docs-travis/indexing.html#indexing-view-versus-copy

that said this was a bug, thanks for reporting

closed by #7087