Question

This question is somewhat related to Visually separating bar chart clusters in pandas

I am reading and plotting the tmp.csv file:

pol1 pol2 pol3
perim 0.54 0.64 0.40
mst 0.08 0.12 0.12
treeadd 0.25 0.34 0.35
health 0.14 0.17 0.17
bisort 0.48 0.56 0.56
em3d 0.14 0.17 0.17

g721d 1.41 2.58 2.58
mesa 1.16 1.8 1.8
epic 1.82 2.43 2.43
jpege 1.18 1.68 1.68

gzip 1.15 1.43 1.45
vpr 0.19 0.24 0.24
gcc 0.82 1.11 1.15
mcf 0.05 0.05 0.05
crafty 0.67 1.17 1.17

with:

#!/usr/bin/env python

from pandas import *
import matplotlib.pyplot as plt
from numpy import zeros

# Create original dataframe
df = read_csv('tmp.csv',sep='\s')

print df

df.plot(kind='bar')
plt.show()

and I get:

         pol1  pol2  pol3
perim    0.54  0.64  0.40
mst      0.08  0.12  0.12
treeadd  0.25  0.34  0.35
health   0.14  0.17  0.17
bisort   0.48  0.56  0.56
em3d     0.14  0.17  0.17
nan       NaN   NaN   NaN
g721d    1.41  2.58  2.58
mesa     1.16  1.80  1.80
epic     1.82  2.43  2.43
jpege    1.18  1.68  1.68
nan       NaN   NaN   NaN
gzip     1.15  1.43  1.45
vpr      0.19  0.24  0.24
gcc      0.82  1.11  1.15
mcf      0.05  0.05  0.05
crafty   0.67  1.17  1.17

and:

Note the separation of the clusters with the empty lines. This is the effect I want.
Is there a way to replace the ‘nan’ label with “” in the x-axis?

I tried:
df.rename(index={‘nan’: “”})

However there is an assertion failing

assert(new_axis.is_unique)

Probably because there are multiple ‘nan’ indexing the df. Ideas?

-Thanks

Answer 1

Instead of modifying the DataFrame, perhaps just change the matplotlib xtick labels:

import pandas as pd
import matplotlib.pyplot as plt

df = pd.read_csv('tmp.csv', sep = '\s')
df.plot(kind='bar')
locs, labels = plt.xticks()
plt.xticks(locs, [d if d==d else '' for d in df.index], rotation = 25)
plt.show()