如何通过任意长度的两列列表对熊猫数据框进行子集

Question

I have tried different combinations of Boolean arrays and .isin constructions, but my pandas fu is not strong enough.

If I have the following example dataframe:

In[1]:  import pandas as pd
        exampledf = pd.DataFrame({ 'factor1' : ['a', 'b', 'c', 'd', 'a', 'b', 'c', 'd'],
                                   'factor2' : ['e', 'e', 'e', 'e', 'f', 'f', 'f', 'f'],
                                   'numeric' : [1., 2., 3., 4., 5., 6., 7., 8.] })

I need to pass a list of factor1, factor2 pairs of any length to return the subset of the dataframe that has that combination of factors.

For example:

In[2]:  def factorfilter(df, factorlist):
           # code goes here
           # returns a dataframe

        factorfilter(exampledf, [['a', 'e'], ['c', 'f']])

Out[2]:   factor1 factor2  numeric
        0       a       e        1
        6       f       f        7

(If there's a better way to set this up than with lists, I'm all ears, it's just what occurred to me and is easy to produce and pass to a function).

Answer 1

You can utilize a multi-index (index off more than one column). Two ways of building an index from the example schema come to mind.

import pandas as pd
index = pd.MultiIndex.from_product([list('abcd'),list('ef')],
                                   names=['factor1','factor2'])

or

factor1 = list('abcdabcd')
factor2 = list('eeeeffff')
index = pd.MultIndex.from_tuples(list(zip(factor1, factor2)),
                                 names=['factor1', 'factor2'])

from this, you can create a multi-index DataFrame by

numerics = list(range(1,9))
df = pd.DataFrame({'numeric': numerics}, index=index)

df outputs

                 numeric
factor1 factor2
a       e              1
        f              2
b       e              3
        f              4
c       e              5
        f              6
d       e              7
        f              8

[8 rows x 1 columns]

Then, you can retrieve a subset of indices, by passing a list of tuples to the ix property.

subdf = df.ix[[('a','e'), ('c','f')]]

subdf outputs

                 numeric
factor1 factor2
a       e              1
c       f              6

[2 rows x 1 columns]