CODEWITHSUNDEEP
pythonserializationyamlpyyaml
SimonD
SimonD

Reputation: 638

Selectively dump object attributes with PyYAML

I can use YAML to dump a hierarchy of python objects, so:

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)

produces the output:

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

But I want to selectively hide some attributes. So suppose I have a function attribs_to_dump(obj) which for any object returns the list of attribute names I want to dump, for example:

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

My question is, how do I hook attribs_to_dump into yaml.dump so that I get the following output?

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

There's a complicating factor: I want to achieve the effect by hooking into Yaml as it crawls over the object hierarchy rather than by pre-processing the object hierarchy myself. The reason is that not all objects in the hierarchy are easily amenable to introspection due to setattr/getattr/__dict__ magic that is present in some libraries I am using :-(...

All help much appreciated!

Upvotes: 2

Views: 1475

Answers (1)

andrean
andrean

Reputation: 6796

This is an interesting question, I enjoyed solving it, thanks :)

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

Upvotes: 1

Related Questions