CODEWITHSUNDEEP
pythonpandasdataframe
Fabio Lamanna
Fabio Lamanna

Reputation: 21552

pandas - mask dataframe by column name

Starting from this simple dataframe df:

col1,col2
1,3
2,1
3,8

I would like to apply a boolean mask in function of the name of the column. I know that it is easy for values:

mask = df <= 1

df = df[mask]

which returns:

mask:

    col1   col2
0   True  False
1  False   True
2  False  False

df:

   col1  col2
0     1   NaN
1   NaN     1
2   NaN   NaN

as expected. Now I would like to obtain a boolean mask based on the column name, something like:

mask = df == df['col_1']

which should return:

mask

    col1   col2
0   True  False
1   True  False
2   True  False

EDIT:

This seems weird, but I need those kind of masks to later filtering by columns seaborn heatmaps.

Upvotes: 5

Views: 24282

Answers (2)

Anton Protopopov
Anton Protopopov

Reputation: 31672

You could transpose your dataframe than compare it with the columns and then transpose back. A bit weird but working example:

import pandas as pd
from io import StringIO

data = """
col1,col2
1,3
2,1
3,8
"""

df = pd.read_csv(StringIO(data))
mask = (df.T == df['col1']).T

In [176]: df
Out[176]:
   col1  col2
0     1     3
1     2     1
2     3     8


In [178]: mask
Out[178]:
   col1   col2
0  True  False
1  True  False
2  True  False

EDIT

I found another answer for that, you could use isin method:

In [41]: df.isin(df.col1)
Out[41]:
   col1   col2
0  True  False
1  True  False
2  True  False

EDIT2

As @DSM show in the comment that these two cases not working correctly. So you should use @KT. method. But.. Let's play more with transpose: 

df.col2 = df.col1

In [149]: df
Out[149]:
   col1  col2
0     1     1
1     2     2
2     3     3

In [147]: df.isin(df.T[df.columns == 'col1'].T)
Out[147]:
   col1   col2
0  True  False
1  True  False
2  True  False

Upvotes: 0

KT.
KT.

Reputation: 11430

As noted in the comments, situations where you would need to get a "mask" like that seem rare (and chances are, you not in one of them). Consequently, there is probably no nice "built-in" solution for them in Pandas.

None the less, you can achieve what you need, using a hack like the following, for example:

mask = (df == df) & (df.columns == 'col_1')

Update:. As noted in the comments, if your data frame contains nulls, the mask computed this way will always be False at the corresponding locations. If this is a problem, the safer option is:

mask = ((df == df) | df.isnull()) & (df.columns == 'col_1')

Upvotes: 7

Related Questions