将不完整的嵌套列表放在矩形ndarray中 [英] Putting incomplete nested lists in a rectangular ndarray

问题描述

在Python(也使用numpy)中，我有一个列表列表列表，每个列表的长度都不同.

In Python (also using numpy) I have a list of lists of lists, with each list being different lengths.

[
    [
         ["header1","header2"],
         ["---"],
         [],
         ["item1","value1"]
    ],

    [
         ["header1","header2","header3"],
         ["item2","value2"],
         ["item3","value3","value4","value5"]
    ]
]

我想使此数据结构为矩形:即，确保len(list[x])对于所有x是恒定的，len(list[x][y])对于所有x，y等都是恒定的.

I want to make this data structure rectangular: i.e. guarantee that len(list[x]) is constant for all x, len(list[x][y]) is constant for all x,y, etc.

(这是因为我要将数据结构导入numpy)

(This is because I want to import the data structure into numpy)

我可以想到做这种事情的各种非Python方法(迭代结构，记录每个级别的最大长度，具有None的第二遍和填充值，但是必须有更好的方法.

I can think of various unpythonic ways of doing such a thing (iterate over structure, record maximum length at each level, have second pass and pad values with None, but there must be a better way.

(我也希望解决方案不依赖于结构的维数；即，它也应适用于此类结构的列表...)

(I also would like the solution to not be dependant on the dimensionality of the structure; i.e. it should work on lists of such structures, too...)

有没有一种简单的方法可以使我丢失?

Is there a simple way of doing this that I'm missing?

推荐答案

您可以创建具有所需尺寸的ndarray并轻松阅读列表.由于列表不完整，因此必须捕获IndexError，可以在try / exception块中完成.

You can create a ndarray with the desired dimensions and readily read your list. Since your list is incomplete you must catch the IndexError, which can be done in a try / exception block.

使用numpy.ndenumerate使解决方案可以轻松扩展到更大的维度(在下面的for循环中添加更多索引i,j,k,l,m,n,...):

Using numpy.ndenumerate allows the solution to be easily extensible to more dimensions (adding more indexes i,j,k,l,m,n,... in the for loop below):

import numpy as np
test = [ [ ["header1","header2"],
           ["---"],
           [],
           ["item1","value1"] ],
         [ ["header1","header2","header3"],
           ["item2","value2"],
           ["item3","value3","value4","value5"] ] ]


collector = np.empty((2,4,4),dtype='|S20')

for (i,j,k), v in np.ndenumerate( collector ):
    try:
        collector[i,j,k] = test[i][j][k]
    except IndexError:
        collector[i,j,k] = ''


print collector
#array([[['header1', 'header2', '', ''],
#        ['---', '', '', ''],
#        ['', '', '', ''],
#        ['item1', 'value1', '', '']],
#       [['header1', 'header2', 'header3', ''],
#        ['item2', 'value2', '', ''],
#        ['item3', 'value3', 'value4', 'value5'],
#        ['', '', '', '']]],  dtype='|S10')

这篇关于将不完整的嵌套列表放在矩形ndarray中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！