4:54
  "   %ˆl–ßi~”ÊeJvãa¸&TÅ£ÌƒuæŸÀÊ ©    %"""JOSE interfaces."""
import abc
import json

import six

from josepy import errors, util

try:
    from collections.abc import Sequence, Mapping  # pylint: disable=import-error
except ImportError:
    from collections import Sequence, Mapping

# pylint: disable=no-self-argument,no-method-argument,no-init,inherit-non-class
# pylint: disable=too-few-public-methods


@six.add_metaclass(abc.ABCMeta)
class JSONDeSerializable(object):
    # pylint: disable=too-few-public-methods
    """Interface for (de)serializable JSON objects.

    Please recall, that standard Python library implements
    :class:`json.JSONEncoder` and :class:`json.JSONDecoder` that perform
    translations based on respective :ref:`conversion tables
    <conversion-table>` that look pretty much like the one below (for
    complete tables see relevant Python documentation):

    .. _conversion-table:

    ======  ======
     JSON   Python
    ======  ======
    object  dict
    ...     ...
    ======  ======

    While the above **conversion table** is about translation of JSON
    documents to/from the basic Python types only,
    :class:`JSONDeSerializable` introduces the following two concepts:

      serialization
        Turning an arbitrary Python object into Python object that can
        be encoded into a JSON document. **Full serialization** produces
        a Python object composed of only basic types as required by the
        :ref:`conversion table <conversion-table>`. **Partial
        serialization** (accomplished by :meth:`to_partial_json`)
        produces a Python object that might also be built from other
        :class:`JSONDeSerializable` objects.

      deserialization
        Turning a decoded Python object (necessarily one of the basic
        types as required by the :ref:`conversion table
        <conversion-table>`) into an arbitrary Python object.

    Serialization produces **serialized object** ("partially serialized
    object" or "fully serialized object" for partial and full
    serialization respectively) and deserialization produces
    **deserialized object**, both usually denoted in the source code as
    ``jobj``.

    Wording in the official Python documentation might be confusing
    after reading the above, but in the light of those definitions, one
    can view :meth:`json.JSONDecoder.decode` as decoder and
    deserializer of basic types, :meth:`json.JSONEncoder.default` as
    serializer of basic types, :meth:`json.JSONEncoder.encode`  as
    serializer and encoder of basic types.

    One could extend :mod:`json` to support arbitrary object
    (de)serialization either by:

      - overriding :meth:`json.JSONDecoder.decode` and
        :meth:`json.JSONEncoder.default` in subclasses

      - or passing ``object_hook`` argument (or ``object_hook_pairs``)
        to :func:`json.load`/:func:`json.loads` or ``default`` argument
        for :func:`json.dump`/:func:`json.dumps`.

    Interestingly, ``default`` is required to perform only partial
    serialization, as :func:`json.dumps` applies ``default``
    recursively. This is the idea behind making :meth:`to_partial_json`
    produce only partial serialization, while providing custom
    :meth:`json_dumps` that dumps with ``default`` set to
    :meth:`json_dump_default`.

    To make further documentation a bit more concrete, please, consider
    the following imaginatory implementation example::

      class Foo(JSONDeSerializable):
          def to_partial_json(self):
              return 'foo'

          @classmethod
          def from_json(cls, jobj):
              return Foo()

      class Bar(JSONDeSerializable):
          def to_partial_json(self):
              return [Foo(), Foo()]

          @classmethod
          def from_json(cls, jobj):
              return Bar()

    """

    @abc.abstractmethod
    def to_partial_json(self):  # pragma: no cover
        """Partially serialize.

        Following the example, **partial serialization** means the following::

          assert isinstance(Bar().to_pa