直接从数据框熊猫对象中带有颜色条的水平条形图

Question

I have the below code and I am wondering if I can produce a horizontal bar chart plot where the bars change colour horizontally (overtime in my case) according to a given colour map directly from a data frame object.

from matplotlib import pyplot as plt
from matplotlib import cm
import pandas as pd
import numpy as np

# generate dataframe
df = pd.DataFrame(np.random.randn(1000, 4), index=pd.date_range('1/1/2000', periods=ndays), columns=list('ABCD'))
df = df.cumsum()

# plot dataframe
df.plot(figsize=(10,5))

# TODO
# make a plot with four horizontal bars each for one column (A, B, C, D)
# the bars represent the time variation and thus should change colour according to the values in the ABCD columns mapped to cmap colour map

I also include below a temporary solution using scatter plot for the purpose of visualisation what I aim for.

# this is what I would like plot to look like
# I use plt.scatter to show
fig = plt.figure(figsize=(15,2))
ax = fig.add_subplot(111)
cmap = cm.gnuplot
ax.scatter(x=df.index, y=[0]*df.A.shape[0], c=df.A, cmap=cmap)
ax.scatter(x=df.index, y=[1]*df.A.shape[0], c=df.B, cmap=cmap)
ax.scatter(x=df.index, y=[2]*df.A.shape[0], c=df.C, cmap=cmap)
cax = ax.scatter(x=df.index, y=[3]*df.A.shape[0], c=df.D, cmap=cmap)

# vertical color bar
clim = cax.get_clim()
cbar = fig.colorbar(cax, ticks=[min(clim),np.mean(clim),max(clim)], orientation='vertical')
cbar.ax.set_xticklabels(['Low', 'Medium', 'High'])

Note: The question is not about scatter plot but if and how to make the result obtain with my current solution directly from the data frame object.

Answer 1

You can kind of do this through pandas, but it is really ugly. You need to make new columns just for scatter plotting, and then pass the color argument a list of colors based on the actual data.

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.colors as colors

df = pd.DataFrame(np.random.randn(1000, 4), index=pd.date_range('1/1/2000', periods=1000), 
                  columns=list('ABCD')).cumsum()

# create color normalizer
cnormer = colors.Normalize(df.min().min(), df.max().max())

# add new columns for the scatter plot
df['i'] = np.arange(1000)
for x in range(4):
    df['z{}'.format(x)] = x

# create an axis to plot on
fig, ax = plt.subplots(1,1,figsize=(10,3))

# make scatter plots
for x in range(4):
    df.plot(x='i', y='z{}'.format(x), kind='scatter', marker='o', edgecolor='none', 
            c=plt.cm.viridis(cnormer(df.iloc[:,x])), ax=ax)

You would still need to reformat the labels and ticks.