Let's say I have a DataFrame with a MultiIndex of columns like this:
In [29]: df = pd.DataFrame([[0] * 8], columns = pd.MultiIndex.from_product(
[['a', 'b'], [1, 2], [2000, 2001]])
)
In [30]: df
Out[30]:
a b
1 2 1 2
2000 2001 2000 2001 2000 2001 2000 2001
0 0 0 0 0 0 0 0 0
In [46]: df.columns.levels
Out[46]: FrozenList([[u'a', u'b'], [1, 2], [2000, 2001]])
I need to know, for all values of level 0 and some specific value of level 1, what are all the existing unique values of level 2 (say the DataFrame goes through some process in which for some values of level 1 and level 0, level 2 is dropped). The best I've been able to come up with so far is this:
In [54]: level_1_val = 2
In [55]: cols_series = df.columns.to_series()
In [56]: cols_series[
....: cols_series.index.get_level_values(1) == level_1_val
....: ].index.get_level_values(2).unique()
array([2000, 2001])
What's a better way to do this?
IIUC
df.xs(2, axis=1, level=1).groupby(axis=1, level=1).first().columns.values
array([2000, 2001])
Or
df.xs(2, axis=1, level=1).columns.get_level_values(level=1).unique()
Collected from the Internet
Please contact [email protected] to delete if infringement.
Comments