Pandas: [Improvement] Deterministic value_counts

[tomspur]

Code Sample, a copy-pastable example if possible

import pandas as pd

df = pd.DataFrame(["a", "b", "c"], columns=["test"])

print(df["test"].value_counts())

Problem description

Using value_counts in a testsuite can be a problem, when the resulting values have the same count as they permutade on each call, e.g.:

$ python pandas_value_counts.py
a    1
b    1
c    1
Name: test, dtype: int64

$ python pandas_value_counts.py
c    1
a    1
b    1
Name: test, dtype: int64

Expected Output

Some stable/deterministic output or optionally additionally sorting of the keys, if they have the same counts

Output of pd.show_versions()

INSTALLED VERSIONS

commit: None
python: 3.6.0.final.0
python-bits: 32
OS: Windows
OS-release: 10
machine: AMD64
processor: Intel64 Family 6 Model 94 Stepping 3, GenuineIntel
byteorder: little
LC_ALL: None
LANG: de_DE.UTF-8
LOCALE: None.None

pandas: 0.19.2
nose: 1.3.7
pip: 9.0.1
setuptools: 27.2.0
Cython: 0.25.2
numpy: 1.11.3
scipy: 0.18.1
statsmodels: 0.6.1
xarray: None
IPython: 5.1.0
sphinx: 1.5.1
patsy: 0.4.1
dateutil: 2.6.0
pytz: 2016.10
blosc: None
bottleneck: 1.2.0
tables: 3.2.2
numexpr: 2.6.1
matplotlib: 2.0.0
openpyxl: 2.4.1
xlrd: 1.0.0
xlwt: 1.2.0
xlsxwriter: 0.9.6
lxml: 3.7.2
bs4: 4.5.3
html5lib: None
httplib2: None
apiclient: None
sqlalchemy: 1.1.5
pymysql: None
psycopg2: None
jinja2: 2.9.4
boto: 2.45.0
pandas_datareader: None

[jreback]

see discussion in these related issues.

xref #12679
xref #11227
xref #14860

This is not guaranteed in any way, nor is performant to do so. Further why should this be anything but an arbitrary ordering? This is a mapping of value -> count.

[jreback]

If you need this for testing, the easiest / best is simply to .sort_index() and compare.

[jreback]

actually this is a duplicate of #12679

the guarantee on sort=False (not the default) is not there. This could be done as a post-processing step

The .unique() has a guarantee that shows the orderings as seen.

In [23]: s.value_counts(sort=False).reindex(s.unique())
Out[23]: 
a    1
b    1
c    1
dtype: int64

In [24]: s = Series(list('bac'))

In [25]: s.value_counts(sort=False).reindex(s.unique())
Out[25]: 
b    1
a    1
c    1
dtype: int64

If you'd like to do a PR for #12679 would be great.

[tomspur]

Yes, I would need this for testing and would like to have the highest count of the data. With just getting the first item of the value_counts, this is some random value, that has the maximum count and it would be great to have always the same value.

[jreback]

love to have a PR as above!