Python：将函数应用于嵌套字典中的值

Question

我有一组任意深度的嵌套字典：

x = {'a': 1, 'b': {'c': 6, 'd': 7, 'g': {'h': 3, 'i': 9}}, 'e': {'f': 3}}

我想基本上将一个函数应用于字典中的所有整数，所以我猜像map，但对于嵌套字典。

所以：map_nested_dicts(x, lambda v: v + 7) 就是这样的目标。

我不知道存储密钥层然后将修改后的值放回正确位置的最佳方法。

最好的方法/方法是什么？

Answer 1

递归访问所有嵌套值：

import collections

def map_nested_dicts(ob, func):
    if isinstance(ob, collections.Mapping):
        return {k: map_nested_dicts(v, func) for k, v in ob.iteritems()}
    else:
        return func(ob)

map_nested_dicts(x, lambda v: v + 7)
# Creates a new dict object:
#    {'a': 8, 'b': {'c': 13, 'g': {'h': 10, 'i': 16}, 'd': 14}, 'e': {'f': 10}}

在某些情况下，需要修改原始 dict 对象（以避免重新创建它）：

import collections

def map_nested_dicts_modify(ob, func):
    for k, v in ob.iteritems():
        if isinstance(v, collections.Mapping):
            map_nested_dicts_modify(v, func)
        else:
            ob[k] = func(v)

map_nested_dicts_modify(x, lambda v: v + 7)
# x is now
#    {'a': 8, 'b': {'c': 13, 'g': {'h': 10, 'i': 16}, 'd': 14}, 'e': {'f': 10}}

如果您使用的是 Python 3：

• 将dict.iteritems替换为dict.items

• 将import collections 替换为import collections.abc

• 将collections.Mapping替换为collections.abc.Mapping

Answer 2

只是为了扩展 vaultah 的答案，如果您的元素之一可以是列表，并且您也想处理这些：

import collections

def map_nested_dicts_modify(ob, func):
for k, v in ob.iteritems():
    if isinstance(v, collections.Mapping):
        map_nested_dicts_modify(v, func)
    elif isinstance(v, list):
        ob[k] = map(func, v)
    else:
        ob[k] = func(v)

Answer 3

如果您需要它在任意嵌套中同时适用于列表和字典：

def apply_recursive(func, obj):
    if isinstance(obj, dict):  # if dict, apply to each key
        return {k: apply_recursive(func, v) for k, v in obj.items()}
    elif isinstance(obj, list):  # if list, apply to each element
        return [apply_recursive(func, elem) for elem in obj]
    else:
        return func(obj)

Answer 4

我有一个更通用的实现，它可以接受任意数量的任何类型的容器作为参数。

from collections.abc import Iterable
import types
def dict_value_map(fun, *dicts):
    keys = dicts[0].keys()
    for d in dicts[1:]:
        assert d.keys() == keys
    return {k:fun(*(d[k] for d in dicts)) for k in keys}
def collection_map(fun, *collections):
    assert len(collections) > 0
    if isinstance(collections[0], dict):
        return dict_value_map(fun, *collections)
    if isinstance(collections[0], (tuple, list, set)):
        return type(collections[0])(map(fun, *collections))
    else:
        return map(fun, *collections)
iscollection = lambda v:(isinstance(v,Iterable)and(not isinstance(v,str)))

def apply(fun, *collections, at=lambda collections: not iscollection(collections[0])):
    '''
    like standard map, but can apply the fun to inner elements.
    at: a int, a function or sometype. 
    at = 0 means fun(*collections)
    at = somefunction. fun will applied to the elements when somefunction(elements) is True
    at = sometype. fun will applied to the elements when elements are sometype.
    '''
    if isinstance(at, int):
        assert at >= 0
        if at == 0:
            return fun(*collections)
        else:
            return collection_map(lambda *cs:apply(fun, *cs, at=at-1), *collections)
    if isinstance(at, types.FunctionType):
        if at(collections):
            return fun(*collections)
        else:
            return collection_map(lambda *cs:apply(fun, *cs, at=at), *collections)
    else:
        return apply(fun, *collections, at=lambda eles:isinstance(eles[0], at))

例子：

> apply(lambda x:2*x, [(1,2),(3,4)])  
[(2, 4), (6, 8)]

> apply(lambda a,b: a+b, ([1,2],[3,4]), ([5,6],[7,8]))
([6, 8], [10, 12])

> apply(lambda a,b: a+b, ([1,2],[3,4]), ([5,6],[7,8]), at=1)  
([1, 2, 5, 6], [3, 4, 7, 8])

> apply(lambda a,b: a+b, ([1,2],[3,4]), ([5,6],[7,8]), at=0)  
([1, 2], [3, 4], [5, 6], [7, 8])

> apply(lambda a,b:a+b, {'m':[(1,2),[3,{4}]], 'n':5}, {'m':[(6,7),[8,{9}]],'n':10})  
{'m': [(7, 9), [11, {13}]], 'n': 15}

> apply(str.upper, [('a','b'),('c','d')], at=str)  
[('A', 'B'), ('C', 'D')]

和

> apply(lambda v:v+7, {'a': 1, 'b': {'c': 6, 'd': 7, 'g': {'h': 3, 'i': 9}}, 'e': {'f': 3}})
{'a': 8, 'b': {'c': 13, 'd': 14, 'g': {'h': 10, 'i': 16}}, 'e': {'f': 10}}