vmware-nsx/neutron/openstack/common/jsonutils.py

# vim: tabstop=4 shiftwidth=4 softtabstop=4

# Copyright 2010 United States Government as represented by the
# Administrator of the National Aeronautics and Space Administration.
# Copyright 2011 Justin Santa Barbara
# All Rights Reserved.
#
#    Licensed under the Apache License, Version 2.0 (the "License"); you may
#    not use this file except in compliance with the License. You may obtain
#    a copy of the License at
#
#         http://www.apache.org/licenses/LICENSE-2.0
#
#    Unless required by applicable law or agreed to in writing, software
#    distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
#    WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
#    License for the specific language governing permissions and limitations
#    under the License.

'''
JSON related utilities.

This module provides a few things:

    1) A handy function for getting an object down to something that can be
    JSON serialized.  See to_primitive().

    2) Wrappers around loads() and dumps().  The dumps() wrapper will
    automatically use to_primitive() for you if needed.

    3) This sets up anyjson to use the loads() and dumps() wrappers if anyjson
    is available.
'''


import datetime
import functools
import inspect
import itertools
import json
import types
import xmlrpclib

import six

from neutron.openstack.common import timeutils


_nasty_type_tests = [inspect.ismodule, inspect.isclass, inspect.ismethod,
                     inspect.isfunction, inspect.isgeneratorfunction,
                     inspect.isgenerator, inspect.istraceback, inspect.isframe,
                     inspect.iscode, inspect.isbuiltin, inspect.isroutine,
                     inspect.isabstract]

_simple_types = (types.NoneType, int, basestring, bool, float, long)


def to_primitive(value, convert_instances=False, convert_datetime=True,
                 level=0, max_depth=3):
    """Convert a complex object into primitives.

    Handy for JSON serialization. We can optionally handle instances,
    but since this is a recursive function, we could have cyclical
    data structures.

    To handle cyclical data structures we could track the actual objects
    visited in a set, but not all objects are hashable. Instead we just
    track the depth of the object inspections and don't go too deep.

    Therefore, convert_instances=True is lossy ... be aware.

    """
    # handle obvious types first - order of basic types determined by running
    # full tests on nova project, resulting in the following counts:
    # 572754 <type 'NoneType'>
    # 460353 <type 'int'>
    # 379632 <type 'unicode'>
    # 274610 <type 'str'>
    # 199918 <type 'dict'>
    # 114200 <type 'datetime.datetime'>
    #  51817 <type 'bool'>
    #  26164 <type 'list'>
    #   6491 <type 'float'>
    #    283 <type 'tuple'>
    #     19 <type 'long'>
    if isinstance(value, _simple_types):
        return value

    if isinstance(value, datetime.datetime):
        if convert_datetime:
            return timeutils.strtime(value)
        else:
            return value

    # value of itertools.count doesn't get caught by nasty_type_tests
    # and results in infinite loop when list(value) is called.
    if type(value) == itertools.count:
        return six.text_type(value)

    # FIXME(vish): Workaround for LP bug 852095. Without this workaround,
    #              tests that raise an exception in a mocked method that
    #              has a @wrap_exception with a notifier will fail. If
    #              we up the dependency to 0.5.4 (when it is released) we
    #              can remove this workaround.
    if getattr(value, '__module__', None) == 'mox':
        return 'mock'

    if level > max_depth:
        return '?'

    # The try block may not be necessary after the class check above,
    # but just in case ...
    try:
        recursive = functools.partial(to_primitive,
                                      convert_instances=convert_instances,
                                      convert_datetime=convert_datetime,
                                      level=level,
                                      max_depth=max_depth)
        if isinstance(value, dict):
            return dict((k, recursive(v)) for k, v in value.iteritems())
        elif isinstance(value, (list, tuple)):
            return [recursive(lv) for lv in value]

        # It's not clear why xmlrpclib created their own DateTime type, but
        # for our purposes, make it a datetime type which is explicitly
        # handled
        if isinstance(value, xmlrpclib.DateTime):
            value = datetime.datetime(*tuple(value.timetuple())[:6])

        if convert_datetime and isinstance(value, datetime.datetime):
            return timeutils.strtime(value)
        elif hasattr(value, 'iteritems'):
            return recursive(dict(value.iteritems()), level=level + 1)
        elif hasattr(value, '__iter__'):
            return recursive(list(value))
        elif convert_instances and hasattr(value, '__dict__'):
            # Likely an instance of something. Watch for cycles.
            # Ignore class member vars.
            return recursive(value.__dict__, level=level + 1)
        else:
            if any(test(value) for test in _nasty_type_tests):
                return six.text_type(value)
            return value
    except TypeError:
        # Class objects are tricky since they may define something like
        # __iter__ defined but it isn't callable as list().
        return six.text_type(value)


def dumps(value, default=to_primitive, **kwargs):
    return json.dumps(value, default=default, **kwargs)


def loads(s):
    return json.loads(s)


def load(s):
    return json.load(s)


try:
    import anyjson
except ImportError:
    pass
else:
    anyjson._modules.append((__name__, 'dumps', TypeError,
                                       'loads', ValueError, 'load'))
    anyjson.force_implementation(__name__)