用于自定义类的 Python 等效于 C++ begin() 和 end()

Question

假设您有一个字典，其键是整数。这些值也是字典，其键是字符串，其值是 numpy 数组。 比如：

custom = {1: {'a': np.zeros(10), 'b': np.zeros(100)}, 2:{'c': np.zeros(20), 'd': np.zeros(200)}}

我一直在代码中大量使用这种自定义数据结构，每次我需要遍历这个结构的 numpy 数组中的每一行时，我都必须这样做：

for d, delem in custom.items():
    for k, v in delem.items():
        for row in v:
            print(row)

是否可以将此行为封装在函数à la C++ 中，您可以在其中实际实现自定义begin() 和end()？此外，迭代器还应该具有有关其相应字典中的键的信息。我的设想是：

for it in custom:
    d, e, row = *it
    # then do something with these

Answer 1

有多种方法可以做到这一点。 yield 可能是最简单的，因为它为您构建了一个交互器类的繁重工作。

def custom_dict_iter(custom):
    for d, delem in custom.items():
        for k, v in delem.items():
            for row in v:
                yield d, k, row

for d, k, row in custom_dict_iter(my_custom_dict):
    print(d, k, row)

Answer 2

查找iterator protocol - 这更类似于Java 的Iterable 或C# 的IEnumerable，而不是C++ 的开始/结束。您可以通过将__iter__ 方法定义为generator 来更轻松地定义它。

唯一的问题是，您需要使用这些方法让您的 custom 拥有自己的类，而不是普通的字典，但我认为在 C++ 中也是如此。

Answer 3

import numpy as np

custom = {
    1: {'a': np.zeros(10), 'b': np.zeros(100)}, 
    2:{'c': np.zeros(20), 'd': np.zeros(200)}
}

my_gen = (
    (key, subkey, np_array) 
    for (key, a_dict) in custom.items() 
    for subkey, np_array in a_dict.items() 
)

for key, subkey, np_array in my_gen:
    print(key, subkey, np_array)

--output:--
1 b [ 0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.]
1 a [ 0.  0.  0.  0.  0.  0.  0.  0.  0.  0.]
2 d [ 0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.]
2 c [ 0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.]

或者，您可以将数据结构重构为对您的目的更有用的东西：

import numpy as np

custom = {
    1: {'a': np.zeros(10), 'b': np.zeros(100)}, 
    2:{'c': np.zeros(20), 'd': np.zeros(200)}
}

#Create a *list* of tuples:
converted_data = [
    (np_array, subkey, key)
    for (key, a_dict) in custom.items() 
    for subkey, np_array in a_dict.items() 
]

for np_array, subkey, key in converted_data:
    print(key, subkey, np_array)

创建自定义迭代器：

class Dog:
    def __init__(self, data):
        self.data = data
        self.max = len(data)
        self.index_pointer = 0

    def __next__(self):
        index = self.index_pointer

        if index < self.max:
            current_val = self.data[index]
            self.index_pointer += 1
            return current_val
        else:
            raise StopIteration


class MyIter:
    def __iter__(self):
        return Dog([1, 2, 3])


for i in MyIter():
    print(i)

--output:--
1
2
3

__iter__() 只需要返回一个实现__next__() 方法的对象，因此您可以像这样组合这两个类：

class MyIter:
    def __init__(self, data):
        self.data = data
        self.max = len(data)
        self.index_pointer = 0

    def __iter__(self):
        return self  #I have a __next__() method, so let's return me!

    def __next__(self):
        index = self.index_pointer

        if index < self.max:
            current_val = self.data[index]
            self.index_pointer += 1
            return current_val
        else:
            raise StopIteration

for i in MyIter([1, 2, 3]):
    print(i)

--output:--
1
2
3

更复杂的__next__() 方法：

import numpy as np

class CustomIter:
    def __init__(self, data):
        self.data = data
        self.count = 0


    def __iter__(self):
        return self

    def __next__(self):
        count = self.count
        self.count += 1

        if count == 0:  #On first iteration, retun a sum of the keys
            return sum(self.data.keys())

        elif count == 1: #On second iteration, return the subkeys in tuples
            subkeys =  [ 
                a_dict.keys()
                for a_dict in self.data.values()
            ]

            return subkeys

        elif count == 2: #On third iteration, return the count of np arrays
            np_arrays = [
                np_array
                for a_dict in self.data.values()
                for np_array in a_dict.values()
            ]

            return len(np_arrays)

        else:  #Quit after three iterations
            raise StopIteration


custom = {
    1: {'a': np.zeros(10), 'b': np.zeros(100)}, 
    2:{'c': np.zeros(20), 'd': np.zeros(200)}
}

for i in CustomIter(custom):
    print(i)


--output:--
3
[dict_keys(['b', 'a']), dict_keys(['d', 'c'])]
4

Answer 4

作为一种更 Python 的方式，您可以使用嵌套列表推导，它在解释器中以 C 语言速度执行：

>>> [[(i,key,t) for t in value] for i,j in custom.items() for key,value in j.items()]

如果你想获得一个迭代器，你可以使用生成器表达式而不是列表推导。

>>> ([(i,key,t) for t in value] for i,j in custom.items() for key,value in j.items())