使用 PyYAML 选择性地转储对象属性

Question

我可以使用 YAML 转储 python 对象的层次结构，所以：

import yaml

class C():
    def __init__(self, x, y):
        self.x = x
        self.y = y

class D():
    def __init__(self, c, d):
        self.c = c
        self.d = d

d = D(c=C(x=1, y='hi'),
      d='bye')

print yaml.dump(d)

产生输出：

!!python/object:__main__.D
c: !!python/object:__main__.C {x: 1, y: hi}
d: bye

但我想选择性地隐藏一些属性。所以假设我有一个函数attribs_to_dump(obj)，它对于任何对象都返回我要转储的属性名称列表，例如：

def attribs_to_dump(obj):
    if obj.__class__ == C: return ['x']
    if obj.__class__ == D: return ['c']

我的问题是，如何将attribs_to_dump 挂钩到yaml.dump 以便获得以下输出？

!!python/object:__main__.D
c: !!python/object:__main__.C {x: 1}

有一个复杂的因素：我想通过挂钩到 Yaml 来实现效果，因为它在对象层次结构上爬行，而不是通过自己预处理对象层次结构。原因是由于我正在使用的某些库中存在setattr/getattr/__dict__ 魔法，并非层次结构中的所有对象都易于自省:-(...

非常感谢所有帮助！

Answer 1

这是一个有趣的问题，我很乐意解决它，谢谢:)

from copy import deepcopy


class C(object):
    def __init__(self, x, y):
        self.x = x
        self.y = y

class D(object):
    def __init__(self, c, d):
        self.c = c
        self.d = d

d = D(
    c=C(x=1, y='hi'),
    d='bye'
)


FILTER_PARAMS = (
    #(class_reference, process_recursively, ['attributes', 'on', 'whitelist'])
    (D, True, ['c']),
    (C, False, ['x']),
)

def attr_filter(obj, filter_params):
    for attr in dir(obj):
        if attr.startswith('__'):
            # ignore builtins
            continue

        attr_val = obj.__getattribute__(attr)
        # loop through filter params
        for (cls, do_recursive, whitelist) in filter_params:
            if isinstance(obj, cls) and attr in whitelist:
                # filter class matches the current obj's class and
                # current attribute is on the whitelist
                if do_recursive:
                    # must process this attribute the same way as the parent
                    setattr(obj, attr, attr_filter(attr_val, filter_params))                
                # break will avoid the execution of the else clause meaning
                # the attribute was on the white list so don't delete it
                break
        else:
            # delete the current attribute of the instance as it was
            # not on the whitelist
            delattr(obj, attr)

    return obj


# do a deepcopy of the object you pass in, so the original will be preserved
filtered_instance = attr_filter(deepcopy(d), FILTER_PARAMS)
print dir(filtered_instance)
print dir(filtered_instance.c)

# now pass filtered_instance to yaml's dump function