设置名称时使用numpy.ones创建的意外大数组 [英] Unexpectedly large array created with numpy.ones when setting names

问题描述

我正在尝试使用下面的numpy代码获取10 x 8 array

Im trying to get a 10 x 8 array using the code below with numpy

import numpy as np
columns = ["Port Wt", "Bench Wt", "Port Retn", 
               "Bench Retn", "Attrib", "Select", "Inter", "Total"]
a = np.ones([10,len(columns)], 
               dtype={"names":columns, "formats":["f8"]*len(columns)})

我是numpy的新手，但是我遇到了意外的行为-我得到的是10 x 8 x 8 grid.

I'm new to numpy and I get unexpected behaviour - I'm getting a 10 x 8 x 8 grid instead.

我尝试过

a.dtype.names = columns

并获得一个ValueError: there are no fields defined

我在做什么错了，我怎么会得到一个带有名称的10 x 8网格?

What am I doing wrong and how would I get a 10 x 8 grid as desired with the names?

谢谢

推荐答案

您的代码确实会生成10 x 8数组，即a.shape == (10, 8).但是，数组中的每个元素都有8个字段，总共增加了10 x 8 x 8个字段.

Your code does produce a 10 x 8 array, i.e. a.shape == (10, 8). However, each element in the array has 8 fields, adding to a total of 10 x 8 x 8 fields.

所以您可能想要的是一个形状为(10,)且每个元素有8个字段的数组:

So what you probably want is an array with shape (10,) and 8 fields per element:

a = np.ones((10,), dtype={"names":columns, "formats":["f8"]*len(columns)})

这篇关于设置名称时使用numpy.ones创建的意外大数组的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！