CODEWITHSUNDEEP
pythonpandas
cfort
cfort

Reputation: 2776

Map dataframe index using dictionary

Why doesn't df.index.map(dict) work like df['column_name'].map(dict)?

Here's a little example of trying to use index.map:

import pandas as pd

df = pd.DataFrame({'one': {'A': 10, 'B': 20, 'C': 30, 'D': 40, 'E': 50}})
map_dict = {'A': 'every', 'B': 'good', 'C': 'boy', 'D': 'does', 'E': 'fine'}
df
'''
    one
A   10
B   20
C   30
D   40
E   50
'''

df['two'] = df.index.map(mapper=map_dict)

This raises TypeError: 'dict' object is not callable

Feeding it a lambda works:

df['two'] = df.index.map(mapper=(lambda x: map_dict[x])); df
'''
   one    two
A   10  every
B   20   good
C   30    boy
D   40   does
E   50   fine
'''

However, resetting the index and mapping on a column works as expected without complaint:

df.reset_index(inplace=True)
df.rename(columns={'index': 'old_ndx'}, inplace=True) #so there's no index name confusion
df['two'] = df.old_ndx.map(map_dict); df

'''
  old_ndx  one    two
0       A   10  every
1       B   20   good
2       C   30    boy
3       D   40   does
4       E   50   fine
'''

Upvotes: 40

Views: 50076

Answers (6)

KenHBS
KenHBS

Reputation: 7174

As of pandas version 0.23.x (released at May 15th, 2018) this problem is fixed:

import pandas as pd
pd.__version__        # 0.23.4

df = pd.DataFrame({'one': {'A': 10, 'B': 20, 'C': 30, 'D': 40, 'E': 50}})
map_dict = {'A': 'every', 'B': 'good', 'C': 'boy', 'D': 'does', 'E': 'fine'}
df
#    one
# A   10
# B   20
# C   30
# D   40
# E   50
df.index.map(map_dict)
#        one
# every   10
# good    20
# boy     30
# does    40
# fine    50

From the What's New page for pandas 0.23.0 it says:

Index.map() can now accept Series and dictionary input objects (GH12756, GH18482, GH18509).

For more information, check the help page of Index.map

Upvotes: 16

BENY
BENY

Reputation: 323396

Adding get at the end 

df['Two']=df.index.map(map_dict.get)
df
Out[155]: 
   one    Two
A   10  every
B   20   good
C   30    boy
D   40   does
E   50   fine

Upvotes: 24

tozCSS
tozCSS

Reputation: 6134

A shorter alternative --with no explicit call to to_series or pd.Series:

df['two'] = df.rename(map_dict).index

Upvotes: 4

T. Ray
T. Ray

Reputation: 641

An alternative workaround to calling map:

df['two'] = pd.Series(map_dict)

df

   one    two
A   10  every
B   20   good
C   30    boy
D   40   does
E   50   fine

In any case, until the mapping issue gets resolved (per juanpa.arrivillaga's comment) you have to convert either the index or the dict-to-map to a pandas Series.

Upvotes: 6

piRSquared
piRSquared

Reputation: 294576

I'm not answering your question... Just giving you a better work around.
Use to_series() them map

df = pd.DataFrame({'one': {'A': 10, 'B': 20, 'C': 30, 'D': 40, 'E': 50}})
map_dict = {'A': 'every', 'B': 'good', 'C': 'boy', 'D': 'does', 'E': 'fine'}

df['two'] = df.index.to_series().map(map_dict)

df

   one    two
A   10  every
B   20   good
C   30    boy
D   40   does
E   50   fine

Upvotes: 47

JacobIRR
JacobIRR

Reputation: 8966

map (a python keyword) is apparently being used as a method of df.index

Because this has its own internal demands, passing it an argument which has no __call__ method is not allowed.

lambda and functions are callable, a simple test:

def foo():
    pass
if foo.__call__:
    print True
# Prints True

bar = lambda x: x+1
if bar.__call__:
    print True
# Prints True

print {'1':'one'}.__call__
# AttributeError: 'dict' object has no attribute '__call__'

Upvotes: 1

Related Questions