两个Y轴条形图：自定义Xticks

Question

I am trying to add custom xticks to a relatively complicated bar graph plot and I am stuck.

I am plotting from two data frames, merged_90 and merged_15 :

merged_15
                 Volume     y_err_x      Area_2D     y_err_y
TripDate                                                    
2015-09-22  1663.016032  199.507503  1581.591701  163.473202

merged_90

                 Volume     y_err_x      Area_2D     y_err_y
TripDate                                                    
1990-06-10  1096.530711  197.377497  1531.651913  205.197493

I want to create a bar graph with two axes (ie Area_2D and Volume ) where the Area_2D and Volume bars are grouped based on their respective data frame. An example script would look like:

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


fig = plt.figure()
ax1 = fig.add_subplot(111)
merged_90.Volume.plot(ax=ax1, color='orange', kind='bar',position=2.5, yerr=merged_90['y_err_x'] ,use_index=False , width=0.1)
merged_15.Volume.plot(ax=ax1, color='red', kind='bar',position=0.9, yerr=merged_15['y_err_x'] ,use_index=False, width=0.1)


ax2 = ax1.twinx()
merged_90.Area_2D.plot(ax=ax2,color='green',  kind='bar',position=3.5, yerr=merged_90['y_err_y'],use_index=False, width=0.1)
merged_15.Area_2D.plot(ax=ax2,color='blue',  kind='bar',position=0, yerr=merged_15['y_err_y'],use_index=False, width=0.1)

ax1.set_xlim(-0.5,0.2)

x = scipy.arange(1)
ax2.set_xticks(x)
ax2.set_xticklabels(['2015'])
plt.tight_layout()
plt.show()

The resulting plot is:

One would think I could change:

x = scipy.arange(1)
ax2.set_xticks(x)
ax2.set_xticklabels(['2015'])

to

x = scipy.arange(2)
ax2.set_xticks(x)
ax2.set_xticklabels(['1990','2015'])

but that results in:

I would like to see the ticks ordered in chronological order (ie 1990,2015)

Thanks!

Answer 1

Have you considered dropping the second axis and plotting them as follows:

ind = np.array([0,0.3])
width = 0.1
fig, ax = plt.subplots()
Rects1 = ax.bar(ind, [merged_90.Volume.values, merged_15.Volume.values], color=['orange', 'red'] ,width=width)
Rects2 = ax.bar(ind + width, [merged_90.Area_2D.values, merged_15.Area_2D.values], color=['green', 'blue'] ,width=width)
ax.set_xticks([.1,.4]) 
ax.set_xticklabels(('1990','2015'))

This produces:

I omitted the error and colors but you can easily add them. That would produce a readable graph given your test data. As you mentioned in comments you would still rather have two axes, presumably for different data with proper scales. To do this you could do: fig = plt.figure() ax1 = fig.add_subplot(111) merged_90.Volume.plot(ax=ax, color='orange', kind='bar',position=2.5, use_index=False , width=0.1) merged_15.Volume.plot(ax=ax, color='red', kind='bar',position=1.0, use_index=False, width=0.1) ax2 = ax1.twinx() merged_90.Area_2D.plot(ax=ax,color='green', kind='bar',position=3.5,use_index=False, width=0.1) merged_15.Area_2D.plot(ax=ax,color='blue', kind='bar',position=0,use_index=False, width=0.1) ax1.set_xlim([-.45, .2]) ax2.set_xlim(-.45, .2]) ax1.set_xticks([-.35, 0]) ax1.set_xticklabels([1990, 2015])

This produces:

Your problem was with resetting just one axis limit and not the other, they are created as twins but do not necessarily follow the changes made to one another.