Ufdjrw

I have a simple dataframe as follows:

 Condition State Value
0 A AM 0.775651
1 B XP 0.700265
2 A HML 0.688315
3 A RMSML 0.666956
4 B XAD 0.636014
5 C VAP 0.542897
6 C RMSML 0.486664
7 B XMA 0.482742
8 D VCD 0.469553

Now I would like to have a barplot with each value, and same color for each state if Condition is the same. I tried the following python code:

Data_Control = pd.ExcelFile('Bar_plot_example.xlsx') 
df_Control= Data_Control.parse('Sheet2')# my dataframe
s = pd.Series(df_Control.iloc[:,2].values, index=df_Control.iloc[:,1])
colors = 'A': 'r', 'B': 'b', 'C': 'g', 'D':'k'
s.plot(kind='barh', color=[colors[i] for i in df_Control['Condition']])
plt.legend()

But I am not able to get legend correctly for each condition. I am getting the following plot.

So how should I get correct legend for each condition? Any help is highly appreciated, Thanks.

You can create the handles and labels for the legend directly from the data:

labels = df['Condition'].unique()
handles = [plt.Rectangle((0,0),1,1, color=colors[l]) for l in labels]
plt.legend(handles, labels, title="Conditions")

Complete example:

u = """ Condition State Value
0 A AM 0.775651
1 B XP 0.700265
2 A HML 0.688315
3 A RMSML 0.666956
4 B XAD 0.636014
5 C VAP 0.542897
6 C RMSML 0.486664
7 B XMA 0.482742
8 D VCD 0.469553"""

import io
import pandas as pd
import matplotlib.pyplot as plt


df = pd.read_csv(io.StringIO(u),sep="s+" )
s = pd.Series(df.iloc[:,2].values, index=df.iloc[:,1])
colors = 'A': 'r', 'B': 'b', 'C': 'g', 'D':'k'
s.plot(kind='barh', color=[colors[i] for i in df['Condition']])

labels = df['Condition'].unique()
handles = [plt.Rectangle((0,0),1,1, color=colors[l]) for l in labels]
plt.legend(handles, labels, title="Conditions")

plt.show()

So I haven't worked much with plotting directly from pandas, but you'll have to access the handles and use that to construct lists of handles and labels that you can pass to plt.legend.

s.plot(kind='barh', color=[colors[i] for i in df['Condition']])
# Get the original handles.
original_handles = plt.gca().get_legend_handles_labels()[0][0]

# Hold the handles and labels that will be passed to legend in lists.
handles = []
labels = []
conditions = df['Condition'].values
# Seen conditions helps us make sure that each label is added only once.
seen_conditions = set()
# Iterate over the condition and handle together.
for condition, handle in zip(conditions, original_handles):
 # If the condition was already added to the labels, then ignore it.
 if condition in seen_conditions:
 continue
 # Add the handle and label.
 handles.append(handle)
 labels.append(condition)
 seen_conditions.add(condition)

# Call legend with the stored handles and labels.
plt.legend(handles, labels)