Pythonfold/reduce多个词典的组合

Question

我想实现以下目标。它本质上是任意数量的字典的组合或合并，参考 'seed' 或根字典，在最终结果中累积所有未更改和更新的值。

seed = {
    'update': False,
    'data': {
        'subdata': {
            'field1': 5,
            'field2': '2018-01-30 00:00:00'
        },
        'field3': 2,
        'field4': None
    },
    'data_updates': {},
    'subdata_updates': {},
    'diffs': {}
}

update_1 = {
    'update': True,
    'data': {
        'subdata': {
            'field1': 6,
            'field2': '2018-01-30 00:00:00'
        },
        'field3': 2,
        'field4': None
    },
    'data_updates': {},
    'subdata_updates': {'field1': 6},
    'diffs': {
        'field1': {
            'field': 'field1',
            'before': 5,
            'after': 6
        }
    }
}

update_2 = {
    'update': True,
    'data': {
        'subdata': {
            'field1': 5,
            'field2': '2018-01-30 00:00:00',
        },
        'field3': 2,
        'field4': 1
    },
    'data_updates': {'field4': 1},
    'subdata_updates': {},
    'diffs': {
        'field4': {
            'field': 'field4',
            'before': None,
            'after': 1
        }
    }
}

# I want to be able to pass in an arbitrary number of updates.
assert reduce_maps(seed, *[update_1, update_2]) == {
    'update': True,
    'data': {
        'subdata': {
            'field1': 6,
            'field2': '2018-01-30 00:00:00',
        },
        'field3': 2,
        'field4': 1
    },
    'data_updates': {'field4': 1},
    'subdata_updates': {'field1': 6},
    'diffs': {
        'field1': {
            'field': 'field1',
            'before': 5,
            'after': 6
        },
        'field4': {
            'field': 'field4',
            'before': None,
            'after': 1
        }
    }
}

您可以假设数据将始终采用这种形状，您还可以假设每个有效载荷只更新一个字段，并且没有两次更新会更新同一个字段。

我可以隐约感觉到 fold 的类似物潜伏在此处的背景中，在 seed 周围构建数据。

Answer 1

给你：

from pprint import pprint


def merge_working(pre, post):
    if not (isinstance(pre, dict) and isinstance(post, dict)):
        return post

    new = pre.copy()  # values for unique keys of pre will be preserved
    for key, post_value in post.items():
        new[key] = merge_working(new.get(key), post_value)

    return new


def merge_simplest(pre, post):
    if not isinstance(pre, dict):
        return post
    return {key: merge_simplest(pre[key], post[key])
            for key in pre}


merge = merge_working


def reduce_maps(*objects):
    new = objects[0]
    for post in objects[1:]:
        new = merge(new, post)
    return new


seed = {
    'update': False,
    'data': {
        'subdata': {
            'field1': 5,
            'field2': '2018-01-30 00:00:00'
        },
        'field3': 2,
        'field4': None
    },
    'data_updates': {},
    'subdata_updates': {},
    'diffs': {}
}

update_1 = {
    'update': True,
    'data': {
        'subdata': {
            'field1': 6,
            'field2': '2018-01-30 00:00:00'
        },
        'field3': 2,
        'field4': None
    },
    'data_updates': {},
    'subdata_updates': {'field1': 6},
    'diffs': {
        'field1': {
            'field': 'field 1',
            'before': 5,
            'after': 6
        }
    }
}

update_2 = {
    'update': True,
    'data': {
        'subdata': {
            'field1': 5,
            'field2': '2018-01-30 00:00:00',
        },
        'field3': 2,
        'field4': 1
    },
    'data_updates': {'field4': 1},
    'subdata_updates': {},  # was subdata_update
    'diffs': {
        'field4': {
            'field': 'field 4',
            'before': None,
            'after': 1
        }
    }
}

result = reduce_maps(*[seed, update_1, update_2])

golden = {
    'update': True,
    'data': {
        'subdata': {
            'field1': 5,  # was 6
            'field2': '2018-01-30 00:00:00',
        },
        'field3': 2,
        'field4': 1
    },
    'data_updates': {'field4': 1},
    'subdata_updates': {'field1': 6},  # was subdata_update
    'diffs': {
        'field1': {
            'field': 'field 1',
            'before': 5,
            'after': 6
        },
        'field4': {
            'field': 'field 4',
            'before': None,
            'after': 1
        }
    }
}

pprint(result)
pprint(golden)

assert result == golden

我已经修复了您数据中我认为的错别字（请参阅代码中的注释）。

请注意，merge 可能需要根据确切的合并规则和可能的数据进行调整。要明白我的意思，请使用 merge = merge_simplest 并理解它失败的原因。如果 "data-agnostic" 形状（理解为不考虑叶子值的字典树）真的相同，则不会。

Answer 2

import copy
from functools import partial, reduce

def traverse(seed, update, sentinel):
    for key, value in update.items():
        if isinstance(value, dict):
            try:
                traverse(seed[key], update[key], sentinel)
            except KeyError:
                seed[key] = value
        else:
            if key not in seed or value != seed[key] \
                    and key not in sentinel:
                seed[key] = value
                sentinel.add(key)
    return seed


def reduce_maps(seed, *updates):
    seed = copy.deepcopy(seed)
    return reduce(
        partial(traverse, sentinel=set()), [seed, *updates]
    )

Pythonfold/reduce多个词典的组合

Python fold/reduce composition of multiple dictionaries

python

reduce

dictionary

fold