如何在numpy中对元组数组进行布尔切片答案

【问题标题】：How to do boolean slicing on array of tuples in numpy如何在numpy中对元组数组进行布尔切片
【发布时间】：2017-10-04 12:58:43
【问题描述】：

我有一个这样的 .csv 文件

vehicle,speed,datetime,x,y
61C22276,0.0,1.4926212E9,106.33695,11.12652
60C28912,0.0,1.4926212E9,106.84327166666667,10.90424
51D06538,0.0,1.4926212E9,106.7806,10.765768333333334
50LD08650,0.0,1.4926212E9,106.91705,10.746173333333333
50LD08519,41.0,1.4926212E9,106.95493,10.739623333333334
50LD07182,0.0,1.4926212E9,106.917225,10.746073333333333

我通过

将这些数据导入到 numpy 中

my_data = genfromtxt('data/2017-04-20.csv',names=True,delimiter=',')

输出是：

[(b'61C22276',  0.,   1.49262120e+09,  106.33695   ,  11.12652   )
 (b'60C28912',  0.,   1.49262120e+09,  106.84327167,  10.90424   )
 (b'51D06538',  0.,   1.49262120e+09,  106.7806    ,  10.76576833) ...,
 (b'61C18919',  0.,   1.49265726e+09,  106.77865833,  11.03690667)
 (b'61C18919',  0.,   1.49265729e+09,  106.77865833,  11.03690667)
 (b'61C18919',  0.,   1.49265732e+09,  106.77865833,  11.036905  )]

这是一个元组数组（因为我的数据由多种类型组成）

如何根据列的值对my_data 进行切片？（例如：列出vehicle61C2226的所有行）

【问题讨论】：

标签： arrays csv numpy tuples slice

【解决方案1】：

你有一个structured array。然后选择“行”，方法如下：

boolindex=my_data['vehicle']=='50LD08519'
selection=my_data[boolindex] 

#array([('50LD08519', 0.0, 1492621184.0, 106.91705322265625, 10.746172904968262),
#      ('50LD08519', 41.0, 1492621184.0, 106.9549331665039, 10.739623069763184)], 
#      dtype=[('vehicle', '<U'), ('speed', '<f4'), ('datetime', '<f4'), 
#      ('x', '<f4'), ('y', '<f4')])

'Pandas' 为您提供更友好的 I/O 和直观的语法：

In [521]: my_data=pd.read_csv('data.csv')

         vehicle  speed      datetime   x  y
0   61C22276      0 1,492,621,200 106 11
1   60C28912      0 1,492,621,200 107 11
2   51D06538      0 1,492,621,200 107 11
3  50LD08519      0 1,492,621,200 107 11
4  50LD08519     41 1,492,621,200 107 11
5  50LD07182      0 1,492,621,200 107 11

In [522]: my_data[my_data['vehicle']=='50LD08519']
Out[522]: 
     vehicle  speed      datetime   x  y
3  50LD08519      0 1,492,621,200 107 11
4  50LD08519     41 1,492,621,200 107 11

【讨论】：