How to get multiple column-slices of a dataframe in pandas

Question

for example,

from pandas import DataFrame
df = DataFrame(np.arange(8).reshape(1, 8), columns = list('abcdefgh'))

I want to select the columns 'b':'d' and 'f':'h', say both inclusive.

I know I could select 'b':'d' by executing:

df2 = df.loc[:, 'b':'d'], but something like df2 = df.loc[:, ['b':'d', 'f':'h']] which is like the syntax in MATLAB wouldn't work here, so how can I select several columns in dataframe?

Answer 1

You can use numpy.r_ for concatenation of indices, but it works only with positions, so need get_loc or searchsorted + iloc:

df = pd.DataFrame(np.arange(8).reshape(1, 8), columns = list('abcdefgh'))
print (df)
   a  b  c  d  e  f  g  h
0  0  1  2  3  4  5  6  7

b = df.columns.get_loc('b')
d = df.columns.get_loc('d')
f = df.columns.get_loc('f')
h = df.columns.get_loc('h')
print (b,d,f,h)
1 3 5 7

b = df.columns.searchsorted('b')
d = df.columns.searchsorted('d')
f = df.columns.searchsorted('f')
h = df.columns.searchsorted('h')
print (b,d,f,h)
1 3 5 7

df = df.iloc[:, np.r_[b:c+1, f:h+1]]
print (df)
   b  c  d  f  g  h
0  1  2  3  5  6  7

It is same as:

df =  df.iloc[:, np.r_[1:4, 5:8]]
print (df)
   b  c  d  f  g  h
0  1  2  3  5  6  7

---

df =  df.iloc[:, np.r_['b':'d', 'f':'h']]
print (df)
#TypeError: unsupported operand type(s) for -: 'str' and 'str'

Another solution with loc + join:

df =  df.loc[:,'b':'d'].join(df.loc[:,'f':'h'])
print (df)
   b  c  d  f  g  h
0  1  2  3  5  6  7

Answer 2

Another way to do this using pd.concat:

pd.concat([df.loc[:,'b':'d'],df.loc[:,'f':'h']],1)
Out[608]: 
   b  c  d  f  g  h
0  1  2  3  5  6  7

Answer 3

One option for selecting multiple slices is with select_columns from pyjanitor:

# pip install pyjanitor
import pandas as pd
import janitor

df.select_columns(slice('b', 'd'), slice('f', 'h'))

   b  c  d  f  g  h
0  1  2  3  5  6  7

The caveat here is that you have to explicitly use python's builtin slice.