如何对matplot lib中条形图中的字符串列名进行排序

Question

这是我的一些数据的一个例子：

from pandas import *
df = DataFrame({"Experience":['8 to 9 years', '12 to 13 years', '13 to 14 years', '17 to 18 years',
                 '5 to 6 years', '19 to 20 years', '20 or more years', '14 to 15 years', '3 to 4 years',
                 '10 to 11 years', 'Less than a year', '4 to 5 years', '6 to 7 years',
                 '2 to 3 years', '15 to 16 years', '11 to 12 years', '16 to 17 years', '18 to 19 years',
                 '1 to 2 years', '9 to 10 years', '7 to 8 years', '8 to 9 years',
                 '12 to 13 years', '13 to 14 years', '14 to 15 years', '3 to 4 years',
                 '17 to 18 years', '5 to 6 years', '19 to 20 years', '20 or more years',
                 '10 to 11 years', 'Less than a year', '4 to 5 years', '6 to 7 years',
                 '2 to 3 years', '15 to 16 years', '11 to 12 years', '16 to 17 years',
                 '18 to 19 years', '1 to 2 years', '9 to 10 years', '7 to 8 years'],
               "Salary":[50000, 20000, 80000, 60000, 70000, 50000, 45000, 47000, 36000, 74000, 50000, 20000, 80000, 
                         60000, 70000, 50000, 45000, 47000, 36000, 74000, 90000, 50000, 20000, 80000, 60000, 70000, 
                         50000, 45000, 47000, 36000, 74000, 50000, 20000, 80000, 60000, 70000, 50000, 45000, 60000, 
                         70000, 50000, 45000]})
df

df['Salary'] = df['Salary'].astype('int64')

这是我制作的条形图，用来比较每个经验水平的平均工资：

from numpy import median
%matplotlib inline
group = df.groupby('Experience')
group.aggregate(median).plot(kind='barh')

这给了我这个图表：

我希望条形图的列名按顺序排列（如“不到一年”、“1到2年”等），但我正在努力。对于初学熊猫的人来说，最干净的方法是什么？你知道吗

Answer 1

两种方法，第一种是简单和自动的，第二种是用一种系统分类和用另一种系统贴标签。你知道吗

最初的问题是，在“经验”中给定文本字符串的条形图是按字母顺序排序的。我们要数字订单。一种快速的方法是从字符串中提取数字（使用函数to_min_number）并分组/绘图，而不是编辑轴标签，这样图形仍然是自解释的。你知道吗

from pandas import *
from matplotlib.pyplot import show
df = DataFrame({"Experience":['8 to 9 years', '12 to 13 years', '13 to 14 years', '17 to 18 years',
                 '5 to 6 years', '19 to 20 years', '20 or more years', '14 to 15 years', '3 to 4 years',
                 '10 to 11 years', 'Less than a year', '4 to 5 years', '6 to 7 years',
                 '2 to 3 years', '15 to 16 years', '11 to 12 years', '16 to 17 years', '18 to 19 years',
                 '1 to 2 years', '9 to 10 years', '7 to 8 years', '8 to 9 years',
                 '12 to 13 years', '13 to 14 years', '14 to 15 years', '3 to 4 years',
                 '17 to 18 years', '5 to 6 years', '19 to 20 years', '20 or more years',
                 '10 to 11 years', 'Less than a year', '4 to 5 years', '6 to 7 years',
                 '2 to 3 years', '15 to 16 years', '11 to 12 years', '16 to 17 years',
                 '18 to 19 years', '1 to 2 years', '9 to 10 years', '7 to 8 years'],
               "Salary":[50000, 20000, 80000, 60000, 70000, 50000, 45000, 47000, 36000, 74000, 50000, 20000, 80000, 
                         60000, 70000, 50000, 45000, 47000, 36000, 74000, 90000, 50000, 20000, 80000, 60000, 70000, 
                         50000, 45000, 47000, 36000, 74000, 50000, 20000, 80000, 60000, 70000, 50000, 45000, 60000, 
                         70000, 50000, 45000]})
df

df['Salary'] = df['Salary'].astype('int64')

# Making a new column of Experience values that will plot gracefully
def to_min_number(experience):
    t = experience.split(' ')[0]
    if t == 'Less': return 0
    return int(t)

df['Minimum experience'] = map(to_min_number, df['Experience'])

from numpy import median
group = df.groupby('Minimum experience')
barplot = group.aggregate(median).plot(kind='barh', legend=None)
barplot.set_ylabel('Minimum years experience, non-overlapping')
barplot.set_xlabel('Salary, USD')
show()

如果必须使用原始文本字符串，可以根据Minimum experience列中的值更改y-tick-labels的文本。自动pandas打印只为数字标签腾出空间，因此我们在要打印的轴的左边距中强制留出更多空间：

# We are overriding the barplot defaults, so enforcing a new axis layout
fig, ax = subplots()
subplots_adjust(left=0.3)  # Argument is proportion of figure width; found by trial-and-error

barplot = group.aggregate(median).plot(ax=ax, kind='barh', legend=None) # pass it the ax
barplot.set_ylabel('Experience')
barplot.set_xlabel('Salary, USD')

# Need a list of new tick labels in lower-to-upper order. Use the group object, since we have it:
labellist = []
for i, v in group:
    labellist.append({'I':int(i), 'T':v.Experience.values[0]})
labeldf = DataFrame(labellist)

barplot.set_yticklabels(labeldf.sort_values(by='I')['T'])


show()

请注意，如果原始文本字符串不是由具有有限选择的程序生成的，则应该对变体进行更多检查：如果有人编写了“最多1年”呢“20多年”？你知道吗