我会建议使用发电机与产量声明和产量从。

from collections.abc import Iterable

def flatten(items, ignore_types=(bytes, str)):
    """
       Flatten all of the nested lists to the one. Ignoring flatting of iterable types str and bytes by default.
    """
    for x in items:
        if isinstance(x, Iterable) and not isinstance(x, ignore_types):
            yield from flatten(x)
        else:
            yield x

values = [7, [4, 3, 5, [7, 3], (3, 4), ('A', {'B', 'C'})]]

for v in flatten(values):
    print(v)

如果你有一个Numpy Array A:

a = np.array([[1,2], [3,4]])
a.flatten('C')

生产:

[1, 2, 3, 4]

np.flatten 也接受其他参数:

C: F A K

有关参数的详细信息可在这里找到。

這裡有一個通用方法,適用於數字、線條、粘列表和混合容器,這可以讓簡單和複雜的容器混合起來(請參閱Demo)。

代码

from typing import Iterable 
#from collections import Iterable                            # < py38


def flatten(items):
    """Yield items from any nested iterable; see Reference."""
    for x in items:
        if isinstance(x, Iterable) and not isinstance(x, (str, bytes)):
            for sub_x in flatten(x):
                yield sub_x
        else:
            yield x

笔记:

在 Python 3 中,从 flatten(x) 获取可以取代 sub_x 在 flatten(x): 获取 sub_x 在 Python 3.8 中,从 collection.abc 转移到输入模块。

演示

simple = [[1, 2, 3], [4, 5, 6], [7], [8, 9]]
list(flatten(simple))
# [1, 2, 3, 4, 5, 6, 7, 8, 9]

complicated = [[1, [2]], (3, 4, {5, 6}, 7), 8, "9"]              # numbers, strs, nested & mixed
list(flatten(complicated))
# [1, 2, 3, 4, 5, 6, 7, 8, '9']

参考

此解決方案是從 Beazley, D. 和 B. Jones. Recipe 4.14, Python Cookbook 3rd Ed., O'Reilly Media Inc. Sebastopol, CA: 2013 發現以前的 SO 帖子,可能是原來的展示。

要插入深厚的数据结构,请使用 iteration_utilities.deepflatten1:

>>> from iteration_utilities import deepflatten

>>> l = [[1, 2, 3], [4, 5, 6], [7], [8, 9]]
>>> list(deepflatten(l, depth=1))
[1, 2, 3, 4, 5, 6, 7, 8, 9]

>>> l = [[1, 2, 3], [4, [5, 6]], 7, [8, 9]]
>>> list(deepflatten(l))
[1, 2, 3, 4, 5, 6, 7, 8, 9]

这是一个发电机,所以你需要将结果投到列表中,或者明确地对其进行调解。

要单一的平面,如果每一个项目本身是不可分割的,你也可以使用 iteration_utilities.flatten 它本身只是一个薄的旋转器周围 itertools.chain.from_iterable:

>>> from iteration_utilities import flatten
>>> l = [[1, 2, 3], [4, 5, 6], [7], [8, 9]]
>>> list(flatten(l))
[1, 2, 3, 4, 5, 6, 7, 8, 9]

只需添加一些时间表(基于Nico Schlömer的答案,其中不包含此答案中的功能):

此分類上一篇

结果表明,如果 iterable 只包含几个内部 iterables 那么 总数将是最快的,但是,对于长期 iterables 只有 itertools.chain.from_iterable, iteration_utilities.deepflatten 或 nested 理解具有合理的性能, itertools.chain.from_iterable 是最快的(如 Nico Schlömer 已经注意到)。

from itertools import chain
from functools import reduce
from collections import Iterable  # or from collections.abc import Iterable
import operator
from iteration_utilities import deepflatten

def nested_list_comprehension(lsts):
    return [item for sublist in lsts for item in sublist]

def itertools_chain_from_iterable(lsts):
    return list(chain.from_iterable(lsts))

def pythons_sum(lsts):
    return sum(lsts, [])

def reduce_add(lsts):
    return reduce(lambda x, y: x + y, lsts)

def pylangs_flatten(lsts):
    return list(flatten(lsts))

def flatten(items):
    """Yield items from any nested iterable; see REF."""
    for x in items:
        if isinstance(x, Iterable) and not isinstance(x, (str, bytes)):
            yield from flatten(x)
        else:
            yield x

def reduce_concat(lsts):
    return reduce(operator.concat, lsts)

def iteration_utilities_deepflatten(lsts):
    return list(deepflatten(lsts, depth=1))


from simple_benchmark import benchmark

b = benchmark(
    [nested_list_comprehension, itertools_chain_from_iterable, pythons_sum, reduce_add,
     pylangs_flatten, reduce_concat, iteration_utilities_deepflatten],
    arguments={2**i: [[0]*5]*(2**i) for i in range(1, 13)},
    argument_name='number of inner lists'
)

b.plot()

1 Disclaimer:我是该图书馆的作者

你可以简单地使用Pandas这样做:

import pandas as pd
pd.Series([[1, 2, 3], [4, 5, 6], [7], [8, 9]]).sum()

如果你愿意放弃一小量的速度,以便更清洁的外观,那么你可以使用numpy.concatenate().tolist() 或 numpy.concatenate().ravel().tolist():

import numpy

l = [[1, 2, 3], [4, 5, 6], [7], [8, 9]] * 99

%timeit numpy.concatenate(l).ravel().tolist()
1000 loops, best of 3: 313 µs per loop

%timeit numpy.concatenate(l).tolist()
1000 loops, best of 3: 312 µs per loop

%timeit [item for sublist in l for item in sublist]
1000 loops, best of 3: 31.5 µs per loop

您可以在文档中了解更多, numpy.concatenate 和 numpy.ravel。