改进子图中饼图上的标签

Question

我正在绘制一个包含三个子图的数字：

fig = plt.figure(fignum)
ax11 = plt.subplot2grid((2,2), (0,0))
ax12 = plt.subplot2grid((2,2), (0,1))
ax13 = plt.subplot2grid((2,2), (1,0), colspan=2)
ax13.axis('off')

我正在使用Pandas作为我的数据，我想绘制条形图，饼图和表格。我这样做：

d = pd.Series(data,index)

#plotting the first subplot
d.plot(kind='bar',ax=ax11)

#plotting the second one and adjusting size of labels and percentages
(wedges, texts, autotexts) = ax12.pie(d,labels=d.index,autopct='%1.1f%%')
ax12.axis('equal')
for t in texts:
        t.set_size(7)
for t in autotexts:
        t.set_size(7)

#plotting the third axes, the table
ax13.table(cellText=table,colLabels=columns,loc='bottom')

结果如下：

如何通过不重叠标签使饼图看起来更好？我尝试添加一个图例，但它覆盖了图表，我甚至不知道它是否在我指定的位置结束：

ax12.legend(index,loc='lower right',fontsize=7)

有没有办法将图例向下移动到饼图下的空白区域？一旦图例看起来不错，我将从饼图中删除标签。

Answer 1

bbox_to_anchor可让您将图例放在图的画布上。

请参阅：http://matplotlib.org/users/legend_guide.html

import matplotlib.pyplot as plt
import pandas 

fig = plt.figure()
ax11 = plt.subplot2grid((2,2), (0,0))
ax11.bar([1,2,3,4,3,2,1],[1,2,3,4,3,2,1])
ax11.set_xticklabels(["local user","whetever works","L'ours tourne","en rond","dans sa cage",2,1], rotation="vertical")

ax12 = plt.subplot2grid((2,2), (0,1))
ax12.pie( [10,20,30,5,5,6,4,10,10], labels = ["bewarfefvl","easdfgvagfvd","asasdfve.sd","rgdegaf","adfewga","qargw","qaerghrttw","errrrd","ejjjjd"],autopct='%1.1f%%')
ax12.axis('equal')
ax12.legend(loc='lower right',fontsize=7, bbox_to_anchor = (0.75, -01.0) )


ax13 = plt.subplot2grid((2,2), (1,0), colspan=2)
ax13.axis('off')

plt.show()

虽然您可以通过调整它来跳过图例并提高饼图的可读性。我可以想到三种方法，但这绝不是一个详尽的清单：

1. 将列表作为参数传递时，可以通过交换列表中值的顺序来影响图表中每个切片的相对位置，以避免小切片的簇。有可能在pandas中实现这一目标。

2. labeldistance可让您控制标签与图表的距离。

3. 将第二个数组作为explode参数传递，可以偏移每个切片的位置。

以下简单示例：

ax12 = plt.subplot2grid((2,2), (0,1))
labels = ["bewarfefvl","easdfgvagfvd","asasdfve.sd",
              "rgdegaf","adfewga","qargw","qaerghrttw","errrrd","ejjjjd"]
ax12.pie( [10,20,30,5,5,6,4,10,10], [0,0,0.0,0.1,0.3,.5,0,0,0], 
                          labels = labels, labeldistance=1.2)