django rest framework 反序列化过程剖析

前天写了序列化过程，今天就水一篇反序列化过程吧。

反序列化一般都在create,update方法中被使用，如：

class CreateModelMixin:
    """
    Create a model instance.
    """
    def create(self, request, *args, **kwargs):
        serializer = self.get_serializer(data=request.data)
        serializer.is_valid(raise_exception=True)
        self.perform_create(serializer)
        headers = self.get_success_headers(serializer.data)
        return Response(serializer.data, status=status.HTTP_201_CREATED, headers=headers)

    def perform_create(self, serializer):
        serializer.save()

    def get_success_headers(self, data):
        try:
            return {'Location': str(data[api_settings.URL_FIELD_NAME])}
        except (TypeError, KeyError):
            return {}

根据前天所述的继承关系这里serializer.is_valid()调用的就是rest_framework.serilizers.BaseSerializer的is_valid方法，用该方法去反序列化去检查输入的数据是不是符合serilizer类中定义的数据格式

class BaseSerializer(Field):
    def is_valid(self, raise_exception=False):
        assert not hasattr(self, 'restore_object'), (
            'Serializer `%s.%s` has old-style version 2 `.restore_object()` '
            'that is no longer compatible with REST framework 3. '
            'Use the new-style `.create()` and `.update()` methods instead.' %
            (self.__class__.__module__, self.__class__.__name__)
        )

        assert hasattr(self, 'initial_data'), (
            'Cannot call `.is_valid()` as no `data=` keyword argument was '
            'passed when instantiating the serializer instance.'
        )

        if not hasattr(self, '_validated_data'):
            try:
                self._validated_data = self.run_validation(self.initial_data)
            except ValidationError as exc:
                self._validated_data = {}
                self._errors = exc.detail
            else:
                self._errors = {}

        if self._errors and raise_exception:
            raise ValidationError(self.errors)

        return not bool(self._errors)

run_validation方法在rest_framework.serilizers.Serializer类中被定义

class Serializer(BaseSerializer, metaclass=SerializerMetaclass):
    def run_validation(self, data=empty):
        """
        We override the default `run_validation`, because the validation
        performed by validators and the `.validate()` method should
        be coerced into an error dictionary with a 'non_fields_error' key.
        """
        (is_empty_value, data) = self.validate_empty_values(data)
        if is_empty_value:
            return data

        value = self.to_internal_value(data)
        try:
            self.run_validators(value)
            value = self.validate(value)
            assert value is not None, '.validate() should return the validated data'
        except (ValidationError, DjangoValidationError) as exc:
            raise ValidationError(detail=as_serializer_error(exc))

        return value
    def run_validators(self, value):
        """
        Add read_only fields with defaults to value before running validators.
        """
        if isinstance(value, dict):
            to_validate = self._read_only_defaults()
            to_validate.update(value)
        else:
            to_validate = value
        super().run_validators(to_validate)

    def to_internal_value(self, data):
        """
        Dict of native values <- Dict of primitive datatypes.
        """
        if not isinstance(data, Mapping):
            message = self.error_messages['invalid'].format(
                datatype=type(data).__name__
            )
            raise ValidationError({
                api_settings.NON_FIELD_ERRORS_KEY: [message]
            }, code='invalid')

        ret = OrderedDict()
        errors = OrderedDict()
        fields = self._writable_fields

        for field in fields:
            validate_method = getattr(self, 'validate_' + field.field_name, None)
            primitive_value = field.get_value(data)
            try:
                validated_value = field.run_validation(primitive_value)
                if validate_method is not None:
                    validated_value = validate_method(validated_value)
            except ValidationError as exc:
                errors[field.field_name] = exc.detail
            except DjangoValidationError as exc:
                errors[field.field_name] = get_error_detail(exc)
            except SkipField:
                pass
            else:
                set_value(ret, field.source_attrs, validated_value)

        if errors:
            raise ValidationError(errors)

        return ret

在to_internal_value中对字段做反序列化校验，然后在run_validators中对serilizers类中定义的validators规则做校验

校验通过后使用serializer.save保存数据，serializer.save()调用的就是rest_framework.serilizers.BaseSerializer的save方法

class BaseSerializer(Field):
    def save(self, **kwargs):
        assert not hasattr(self, 'save_object'), (
            'Serializer `%s.%s` has old-style version 2 `.save_object()` '
            'that is no longer compatible with REST framework 3. '
            'Use the new-style `.create()` and `.update()` methods instead.' %
            (self.__class__.__module__, self.__class__.__name__)
        )

        assert hasattr(self, '_errors'), (
            'You must call `.is_valid()` before calling `.save()`.'
        )

        assert not self.errors, (
            'You cannot call `.save()` on a serializer with invalid data.'
        )

        # Guard against incorrect use of `serializer.save(commit=False)`
        assert 'commit' not in kwargs, (
            "'commit' is not a valid keyword argument to the 'save()' method. "
            "If you need to access data before committing to the database then "
            "inspect 'serializer.validated_data' instead. "
            "You can also pass additional keyword arguments to 'save()' if you "
            "need to set extra attributes on the saved model instance. "
            "For example: 'serializer.save(owner=request.user)'.'"
        )

        assert not hasattr(self, '_data'), (
            "You cannot call `.save()` after accessing `serializer.data`."
            "If you need to access data before committing to the database then "
            "inspect 'serializer.validated_data' instead. "
        )

        validated_data = dict(
            list(self.validated_data.items()) +
            list(kwargs.items())
        )

        if self.instance is not None:
            self.instance = self.update(self.instance, validated_data)
            assert self.instance is not None, (
                '`update()` did not return an object instance.'
            )
        else:
            self.instance = self.create(validated_data)
            assert self.instance is not None, (
                '`create()` did not return an object instance.'
            )

        return self.instance

在这里用self.instance判断是不是创建还是更新（即是create还是update）所以也会调用相应的方法，就拿create来说

在这里调用的就是rest_framework.serilizers.ModelSerializer类的create方法（update时也是同理）

class ModelSerializer(Serializer):
    def create(self, validated_data):
        """
        We have a bit of extra checking around this in order to provide
        descriptive messages when something goes wrong, but this method is
        essentially just:

            return ExampleModel.objects.create(**validated_data)

        If there are many to many fields present on the instance then they
        cannot be set until the model is instantiated, in which case the
        implementation is like so:

            example_relationship = validated_data.pop('example_relationship')
            instance = ExampleModel.objects.create(**validated_data)
            instance.example_relationship = example_relationship
            return instance

        The default implementation also does not handle nested relationships.
        If you want to support writable nested relationships you'll need
        to write an explicit `.create()` method.
        """
        raise_errors_on_nested_writes('create', self, validated_data)

        ModelClass = self.Meta.model

        # Remove many-to-many relationships from validated_data.
        # They are not valid arguments to the default `.create()` method,
        # as they require that the instance has already been saved.
        info = model_meta.get_field_info(ModelClass)
        many_to_many = {}
        for field_name, relation_info in info.relations.items():
            if relation_info.to_many and (field_name in validated_data):
                many_to_many[field_name] = validated_data.pop(field_name)

        try:
            instance = ModelClass._default_manager.create(**validated_data)
        except TypeError:
            tb = traceback.format_exc()
            msg = (
                'Got a `TypeError` when calling `%s.%s.create()`. '
                'This may be because you have a writable field on the '
                'serializer class that is not a valid argument to '
                '`%s.%s.create()`. You may need to make the field '
                'read-only, or override the %s.create() method to handle '
                'this correctly.
Original exception was:
 %s' %
                (
                    ModelClass.__name__,
                    ModelClass._default_manager.name,
                    ModelClass.__name__,
                    ModelClass._default_manager.name,
                    self.__class__.__name__,
                    tb
                )
            )
            raise TypeError(msg)

        # Save many-to-many relationships after the instance is created.
        if many_to_many:
            for field_name, value in many_to_many.items():
                field = getattr(instance, field_name)
                field.set(value)

        return instance

    def update(self, instance, validated_data):
        raise_errors_on_nested_writes('update', self, validated_data)
        info = model_meta.get_field_info(instance)

        # Simply set each attribute on the instance, and then save it.
        # Note that unlike `.create()` we don't need to treat many-to-many
        # relationships as being a special case. During updates we already
        # have an instance pk for the relationships to be associated with.
        m2m_fields = []
        for attr, value in validated_data.items():
            if attr in info.relations and info.relations[attr].to_many:
                m2m_fields.append((attr, value))
            else:
                setattr(instance, attr, value)

        instance.save()

        # Note that many-to-many fields are set after updating instance.
        # Setting m2m fields triggers signals which could potentially change
        # updated instance and we do not want it to collide with .update()
        for attr, value in m2m_fields:
            field = getattr(instance, attr)
            field.set(value)

        return instance

注意了这里self.Meta.model是在serilizer类中定义了的数据库模型类

这里调用_default_manager.create去将数据保存到数据库