ormar/ormar/models/newbasemodel.py

import sys
import uuid
from typing import (
    AbstractSet,
    Any,
    Callable,
    Dict,
    List,
    Mapping,
    MutableSequence,
    Optional,
    Sequence,
    Set,
    TYPE_CHECKING,
    Type,
    TypeVar,
    Union,
)

try:
    import orjson as json
except ImportError:  # pragma: no cover
    import json  # type: ignore


import databases
import pydantic
import sqlalchemy
from pydantic import BaseModel

import ormar  # noqa I100
from ormar.exceptions import ModelError
from ormar.fields import BaseField
from ormar.fields.foreign_key import ForeignKeyField
from ormar.models.helpers import register_relation_in_alias_manager
from ormar.models.helpers.relations import expand_reverse_relationship
from ormar.models.helpers.sqlalchemy import (
    populate_meta_sqlalchemy_table_if_required,
    update_column_definition,
)
from ormar.models.metaclass import ModelMeta, ModelMetaclass
from ormar.models.modelproxy import ModelTableProxy
from ormar.queryset.utils import translate_list_to_dict
from ormar.relations.alias_manager import AliasManager
from ormar.relations.relation_manager import RelationsManager

if TYPE_CHECKING:  # pragma no cover
    from ormar import Model
    from ormar.signals import SignalEmitter

    T = TypeVar("T", bound=Model)

    IntStr = Union[int, str]
    DictStrAny = Dict[str, Any]
    AbstractSetIntStr = AbstractSet[IntStr]
    MappingIntStrAny = Mapping[IntStr, Any]


class NewBaseModel(pydantic.BaseModel, ModelTableProxy, metaclass=ModelMetaclass):
    """
    Main base class of ormar Model.
    Inherits from pydantic BaseModel and has all mixins combined in ModelTableProxy.
    Constructed with ModelMetaclass which in turn also inherits pydantic metaclass.

    Abstracts away all internals and helper functions, so final Model class has only
    the logic concerned with database connection and data persistance.
    """

    __slots__ = ("_orm_id", "_orm_saved", "_orm", "_pk_column")

    if TYPE_CHECKING:  # pragma no cover
        __model_fields__: Dict[str, Type[BaseField]]
        __table__: sqlalchemy.Table
        __fields__: Dict[str, pydantic.fields.ModelField]
        __pydantic_model__: Type[BaseModel]
        __pkname__: str
        __tablename__: str
        __metadata__: sqlalchemy.MetaData
        __database__: databases.Database
        _orm_relationship_manager: AliasManager
        _orm: RelationsManager
        _orm_saved: bool
        _related_names: Optional[Set]
        _related_names_hash: str
        _pydantic_fields: Set
        _quick_access_fields: Set
        Meta: ModelMeta

    # noinspection PyMissingConstructor
    def __init__(self, *args: Any, **kwargs: Any) -> None:  # type: ignore
        """
        Initializer that creates a new ormar Model that is also pydantic Model at the
        same time.

        Passed keyword arguments can be only field names and their corresponding values
        as those will be passed to pydantic validation that will complain if extra
        params are passed.

        If relations are defined each relation is expanded and children models are also
        initialized and validated. Relation from both sides is registered so you can
        access related models from both sides.

        Json fields are automatically loaded/dumped if needed.

        Models marked as abstract=True in internal Meta class cannot be initialized.

        Accepts also special __pk_only__ flag that indicates that Model is constructed
        only with primary key value (so no other fields, it's a child model on other
        Model), that causes skipping the validation, that's the only case when the
        validation can be skipped.

        Accepts also special __excluded__ parameter that contains a set of fields that
        should be explicitly set to None, as otherwise pydantic will try to populate
        them with their default values if default is set.

        :raises ModelError: if abstract model is initialized, model has ForwardRefs
         that has not been updated or unknown field is passed
        :param args: ignored args
        :type args: Any
        :param kwargs: keyword arguments - all fields values and some special params
        :type kwargs: Any
        """
        self._verify_model_can_be_initialized()
        object.__setattr__(self, "_orm_id", uuid.uuid4().hex)
        object.__setattr__(self, "_orm_saved", False)
        object.__setattr__(self, "_pk_column", None)
        object.__setattr__(
            self,
            "_orm",
            RelationsManager(related_fields=self.extract_related_fields(), owner=self,),
        )

        pk_only = kwargs.pop("__pk_only__", False)
        excluded: Set[str] = kwargs.pop("__excluded__", set())

        if "pk" in kwargs:
            kwargs[self.Meta.pkname] = kwargs.pop("pk")

        # build the models to set them and validate but don't register
        # also remove property fields values from validation
        try:
            new_kwargs: Dict[str, Any] = {
                k: self._convert_json(
                    k,
                    self.Meta.model_fields[k].expand_relationship(
                        v, self, to_register=False, relation_name=k
                    ),
                    "dumps",
                )
                for k, v in kwargs.items()
                if k not in object.__getattribute__(self, "Meta").property_fields
            }
        except KeyError as e:
            raise ModelError(
                f"Unknown field '{e.args[0]}' for model {self.get_name(lower=False)}"
            )

        # explicitly set None to excluded fields
        # as pydantic populates them with default if set
        for field_to_nullify in excluded:
            new_kwargs[field_to_nullify] = None

        values, fields_set, validation_error = pydantic.validate_model(
            self, new_kwargs  # type: ignore
        )
        if validation_error and not pk_only:
            raise validation_error

        object.__setattr__(self, "__dict__", values)
        object.__setattr__(self, "__fields_set__", fields_set)

        # register the columns models after initialization
        for related in self.extract_related_names():
            self.Meta.model_fields[related].expand_relationship(
                new_kwargs.get(related), self, to_register=True, relation_name=related
            )

    def __setattr__(self, name: str, value: Any) -> None:  # noqa CCR001
        """
        Overwrites setattr in object to allow for special behaviour of certain params.

        Parameter "pk" is translated into actual primary key field name.

        Relations are expanded (child model constructed if needed) and registered on
        both ends of the relation. The related models are handled by RelationshipManager
        exposed at _orm param.

        Json fields converted if needed.

        Setting pk, foreign key value or any other field value sets Model save status
        to False. Setting a reverse relation or many to many relation does not as it
        does not modify the state of the model (but related model or through model).

        To short circuit all checks and expansions the set of attribute names present
        on each model is gathered into _quick_access_fields that is looked first and
        if field is in this set the object setattr is called directly.

        :param name: name of the attribute to set
        :type name: str
        :param value: value of the attribute to set
        :type value: Any
        :return: None
        :rtype: None
        """
        if name in object.__getattribute__(self, "_quick_access_fields"):
            object.__setattr__(self, name, value)
        elif name == "pk":
            object.__setattr__(self, self.Meta.pkname, value)
            self.set_save_status(False)
        elif name in self._orm:
            model = self.Meta.model_fields[name].expand_relationship(
                value=value, child=self, relation_name=name
            )
            if isinstance(self.__dict__.get(name), list):
                # virtual foreign key or many to many
                self.__dict__[name].append(model)
            else:
                # foreign key relation
                self.__dict__[name] = model
                self.set_save_status(False)
        else:
            value = (
                self._convert_json(name, value, "dumps")
                if name in self.__fields__
                else value
            )
            super().__setattr__(name, value)
            self.set_save_status(False)

    def __getattribute__(self, item: str) -> Any:
        """
        Because we need to overwrite getting the attribute by ormar instead of pydantic
        as well as returning related models and not the value stored on the model the
        __getattribute__ needs to be used not __getattr__.

        It's used to access all attributes so it can be a big overhead that's why a
        number of short circuits is used.

        To short circuit all checks and expansions the set of attribute names present
        on each model is gathered into _quick_access_fields that is looked first and
        if field is in this set the object setattr is called directly.

        To avoid recursion object's getattribute is used to actually get the attribute
        value from the model after the checks.

        Even the function calls are constructed with objects functions.

        Parameter "pk" is translated into actual primary key field name.

        Relations are returned so the actual related model is returned and not current
        model's field. The related models are handled by RelationshipManager exposed
        at _orm param.

        Json fields are converted if needed.

        :param item: name of the attribute to retrieve
        :type item: str
        :return: value of the attribute
        :rtype: Any
        """
        if item in object.__getattribute__(self, "_quick_access_fields"):
            return object.__getattribute__(self, item)
        if item == "pk":
            return object.__getattribute__(self, "__dict__").get(self.Meta.pkname, None)
        if item in object.__getattribute__(self, "extract_related_names")():
            return object.__getattribute__(
                self, "_extract_related_model_instead_of_field"
            )(item)
        if item in object.__getattribute__(self, "Meta").property_fields:
            value = object.__getattribute__(self, item)
            return value() if callable(value) else value
        if item in object.__getattribute__(self, "_pydantic_fields"):
            value = object.__getattribute__(self, "__dict__").get(item, None)
            value = object.__getattribute__(self, "_convert_json")(item, value, "loads")
            return value
        return object.__getattribute__(self, item)  # pragma: no cover

    def _verify_model_can_be_initialized(self) -> None:
        """
        Raises exception if model is abstract or has ForwardRefs in relation fields.

        :return: None
        :rtype: None
        """
        if self.Meta.abstract:
            raise ModelError(f"You cannot initialize abstract model {self.get_name()}")
        if self.Meta.requires_ref_update:
            raise ModelError(
                f"Model {self.get_name()} has not updated "
                f"ForwardRefs. \nBefore using the model you "
                f"need to call update_forward_refs()."
            )

    def _extract_related_model_instead_of_field(
        self, item: str
    ) -> Optional[Union["T", Sequence["T"]]]:
        """
        Retrieves the related model/models from RelationshipManager.

        :param item: name of the relation
        :type item: str
        :return: related model, list of related models or None
        :rtype: Optional[Union[Model, List[Model]]]
        """
        if item in self._orm:
            return self._orm.get(item)
        return None  # pragma no cover

    def __eq__(self, other: object) -> bool:
        """
        Compares other model to this model. when == is called.
        :param other: other model to compare
        :type other: object
        :return: result of comparison
        :rtype: bool
        """
        if isinstance(other, NewBaseModel):
            return self.__same__(other)
        return super().__eq__(other)  # pragma no cover

    def __same__(self, other: "NewBaseModel") -> bool:
        """
        Used by __eq__, compares other model to this model.
        Compares:
        * _orm_ids,
        * primary key values if it's set
        * dictionary of own fields (excluding relations)
        :param other: model to compare to
        :type other: NewBaseModel
        :return: result of comparison
        :rtype: bool
        """
        return (
            self._orm_id == other._orm_id
            or (self.pk == other.pk and self.pk is not None)
            or self.dict(exclude=self.extract_related_names())
            == other.dict(exclude=other.extract_related_names())
        )

    @classmethod
    def get_name(cls, lower: bool = True) -> str:
        """
        Returns name of the Model class, by default lowercase.

        :param lower: flag if name should be set to lowercase
        :type lower: bool
        :return: name of the model
        :rtype: str
        """
        name = cls.__name__
        if lower:
            name = name.lower()
        return name

    @property
    def pk_column(self) -> sqlalchemy.Column:
        """
        Retrieves primary key sqlalchemy column from models Meta.table.
        Each model has to have primary key.
        Only one primary key column is allowed.

        :return: primary key sqlalchemy column
        :rtype: sqlalchemy.Column
        """
        if object.__getattribute__(self, "_pk_column") is not None:
            return object.__getattribute__(self, "_pk_column")
        pk_columns = self.Meta.table.primary_key.columns.values()
        pk_col = pk_columns[0]
        object.__setattr__(self, "_pk_column", pk_col)
        return pk_col

    @property
    def saved(self) -> bool:
        """Saved status of the model. Changed by setattr and loading from db"""
        return self._orm_saved

    @property
    def signals(self) -> "SignalEmitter":
        """Exposes signals from model Meta"""
        return self.Meta.signals

    @classmethod
    def pk_type(cls) -> Any:
        """Shortcut to models primary key field type"""
        return cls.Meta.model_fields[cls.Meta.pkname].__type__

    @classmethod
    def db_backend_name(cls) -> str:
        """Shortcut to database dialect,
        cause some dialect require different treatment"""
        return cls.Meta.database._backend._dialect.name

    def remove(self, parent: "T", name: str) -> None:
        """Removes child from relation with given name in RelationshipManager"""
        self._orm.remove_parent(self, parent, name)

    def set_save_status(self, status: bool) -> None:
        """Sets value of the save status"""
        object.__setattr__(self, "_orm_saved", status)

    @classmethod
    def get_properties(
        cls, include: Union[Set, Dict, None], exclude: Union[Set, Dict, None]
    ) -> Set[str]:
        """
        Returns a set of names of functions/fields decorated with
        @property_field decorator.

        They are added to dictionary when called directly and therefore also are
        present in fastapi responses.

        :param include: fields to include
        :type include: Union[Set, Dict, None]
        :param exclude: fields to exclude
        :type exclude: Union[Set, Dict, None]
        :return: set of property fields names
        :rtype: Set[str]
        """

        props = cls.Meta.property_fields
        if include:
            props = {prop for prop in props if prop in include}
        if exclude:
            props = {prop for prop in props if prop not in exclude}
        return props

    @classmethod
    def update_forward_refs(cls, **localns: Any) -> None:
        """
        Processes fields that are ForwardRef and need to be evaluated into actual
        models.

        Expands relationships, register relation in alias manager and substitutes
        sqlalchemy columns with new ones with proper column type (null before).

        Populates Meta table of the Model which is left empty before.

        Calls the pydantic method to evaluate pydantic fields.

        :param localns: local namespace
        :type localns: Any
        :return: None
        :rtype: None
        """
        globalns = sys.modules[cls.__module__].__dict__.copy()
        globalns.setdefault(cls.__name__, cls)
        fields_to_check = cls.Meta.model_fields.copy()
        for field_name, field in fields_to_check.items():
            if issubclass(field, ForeignKeyField):
                field.evaluate_forward_ref(globalns=globalns, localns=localns)
                expand_reverse_relationship(
                    model=cls,  # type: ignore
                    model_field=field,
                )
                register_relation_in_alias_manager(
                    cls,  # type: ignore
                    field,
                    field_name,
                )
                update_column_definition(model=cls, field=field)
        populate_meta_sqlalchemy_table_if_required(meta=cls.Meta)
        super().update_forward_refs(**localns)
        cls.Meta.requires_ref_update = False

    def _get_related_not_excluded_fields(
        self, include: Optional[Dict], exclude: Optional[Dict],
    ) -> List:
        """
        Returns related field names applying on them include and exclude set.

        :param include: fields to include
        :type include: Union[Set, Dict, None]
        :param exclude: fields to exclude
        :type exclude: Union[Set, Dict, None]
        :return:
        :rtype: List of fields with relations that is not excluded
        """
        fields = [field for field in self.extract_related_names()]
        if include:
            fields = [field for field in fields if field in include]
        if exclude:
            fields = [
                field
                for field in fields
                if field not in exclude or exclude.get(field) is not Ellipsis
            ]
        return fields

    @staticmethod
    def _extract_nested_models_from_list(
        models: MutableSequence,
        include: Union[Set, Dict, None],
        exclude: Union[Set, Dict, None],
    ) -> List:
        """
        Converts list of models into list of dictionaries.

        :param models: List of models
        :type models: List
        :param include: fields to include
        :type include: Union[Set, Dict, None]
        :param exclude: fields to exclude
        :type exclude: Union[Set, Dict, None]
        :return: list of models converted to dictionaries
        :rtype: List[Dict]
        """
        result = []
        for model in models:
            try:
                result.append(
                    model.dict(nested=True, include=include, exclude=exclude,)
                )
            except ReferenceError:  # pragma no cover
                continue
        return result

    def _skip_ellipsis(
        self, items: Union[Set, Dict, None], key: str
    ) -> Union[Set, Dict, None]:
        """
        Helper to traverse the include/exclude dictionaries.
        In dict() Ellipsis should be skipped as it indicates all fields required
        and not the actual set/dict with fields names.

        :param items: current include/exclude value
        :type items: Union[Set, Dict, None]
        :param key: key for nested relations to check
        :type key: str
        :return: nested value of the items
        :rtype: Union[Set, Dict, None]
        """
        result = self.get_child(items, key)
        return result if result is not Ellipsis else None

    def _extract_nested_models(  # noqa: CCR001
        self,
        nested: bool,
        dict_instance: Dict,
        include: Optional[Dict],
        exclude: Optional[Dict],
    ) -> Dict:
        """
        Traverse nested models and converts them into dictionaries.
        Calls itself recursively if needed.

        :param nested: flag if current instance is nested
        :type nested: bool
        :param dict_instance: current instance dict
        :type dict_instance: Dict
        :param include: fields to include
        :type include: Optional[Dict]
        :param exclude: fields to exclude
        :type exclude: Optional[Dict]
        :return: current model dict with child models converted to dictionaries
        :rtype: Dict
        """

        fields = self._get_related_not_excluded_fields(include=include, exclude=exclude)

        for field in fields:
            if self.Meta.model_fields[field].virtual and nested:
                continue
            nested_model = getattr(self, field)
            if isinstance(nested_model, MutableSequence):
                dict_instance[field] = self._extract_nested_models_from_list(
                    models=nested_model,
                    include=self._skip_ellipsis(include, field),
                    exclude=self._skip_ellipsis(exclude, field),
                )
            elif nested_model is not None:
                dict_instance[field] = nested_model.dict(
                    nested=True,
                    include=self._skip_ellipsis(include, field),
                    exclude=self._skip_ellipsis(exclude, field),
                )
            else:
                dict_instance[field] = None
        return dict_instance

    def dict(  # type: ignore # noqa A003
        self,
        *,
        include: Union[Set, Dict] = None,
        exclude: Union[Set, Dict] = None,
        by_alias: bool = False,
        skip_defaults: bool = None,
        exclude_unset: bool = False,
        exclude_defaults: bool = False,
        exclude_none: bool = False,
        nested: bool = False,
    ) -> "DictStrAny":  # noqa: A003'
        """

        Generate a dictionary representation of the model,
        optionally specifying which fields to include or exclude.

        Nested models are also parsed to dictionaries.

        Additionally fields decorated with @property_field are also added.

        :param include: fields to include
        :type include: Union[Set, Dict, None]
        :param exclude: fields to exclude
        :type exclude: Union[Set, Dict, None]
        :param by_alias: flag to get values by alias - passed to pydantic
        :type by_alias: bool
        :param skip_defaults: flag to not set values - passed to pydantic
        :type skip_defaults: bool
        :param exclude_unset: flag to exclude not set values - passed to pydantic
        :type exclude_unset: bool
        :param exclude_defaults: flag to exclude default values - passed to pydantic
        :type exclude_defaults: bool
        :param exclude_none: flag to exclude None values - passed to pydantic
        :type exclude_none: bool
        :param nested: flag if the current model is nested
        :type nested: bool
        :return:
        :rtype:
        """
        dict_instance = super().dict(
            include=include,
            exclude=self._update_excluded_with_related_not_required(exclude, nested),
            by_alias=by_alias,
            skip_defaults=skip_defaults,
            exclude_unset=exclude_unset,
            exclude_defaults=exclude_defaults,
            exclude_none=exclude_none,
        )

        if include and isinstance(include, Set):
            include = translate_list_to_dict(include)
        if exclude and isinstance(exclude, Set):
            exclude = translate_list_to_dict(exclude)

        dict_instance = self._extract_nested_models(
            nested=nested,
            dict_instance=dict_instance,
            include=include,  # type: ignore
            exclude=exclude,  # type: ignore
        )

        # include model properties as fields in dict
        if object.__getattribute__(self, "Meta").property_fields:
            props = self.get_properties(include=include, exclude=exclude)
            if props:
                dict_instance.update({prop: getattr(self, prop) for prop in props})

        return dict_instance

    def update_from_dict(self, value_dict: Dict) -> "NewBaseModel":
        """
        Updates self with values of fields passed in the dictionary.

        :param value_dict: dictionary of fields names and values
        :type value_dict: Dict
        :return: self
        :rtype: NewBaseModel
        """
        for key, value in value_dict.items():
            setattr(self, key, value)
        return self

    def _convert_json(self, column_name: str, value: Any, op: str) -> Union[str, Dict]:
        """
        Converts value to/from json if needed (for Json columns).

        :param column_name: name of the field
        :type column_name: str
        :param value: value fo the field
        :type value: Any
        :param op: operator on json
        :type op: str
        :return: converted value if needed, else original value
        :rtype: Any
        """
        if not self._is_conversion_to_json_needed(column_name):
            return value

        condition = (
            isinstance(value, str) if op == "loads" else not isinstance(value, str)
        )
        operand: Callable[[Any], Any] = (
            json.loads if op == "loads" else json.dumps  # type: ignore
        )

        if condition:
            try:
                value = operand(value)
            except TypeError:  # pragma no cover
                pass
        return value.decode("utf-8") if isinstance(value, bytes) else value

    def _is_conversion_to_json_needed(self, column_name: str) -> bool:
        """
        Checks if given column name is related to JSON field.

        :param column_name: name of the field
        :type column_name: str
        :return: result of the check
        :rtype: bool
        """
        return (
            column_name in self.Meta.model_fields
            and self.Meta.model_fields[column_name].__type__ == pydantic.Json
        )

    def _extract_own_model_fields(self) -> Dict:
        """
        Returns a dictionary with field names and values for fields that are not
        relations fields (ForeignKey, ManyToMany etc.)

        :return: dictionary of fields names and values.
        :rtype: Dict
        """
        related_names = self.extract_related_names()
        self_fields = self.dict(exclude=related_names)
        return self_fields

    def _extract_model_db_fields(self) -> Dict:
        """
        Returns a dictionary with field names and values for fields that are stored in
        current model's table.

        That includes own non-relational fields ang foreign key fields.

        :return: dictionary of fields names and values.
        :rtype: Dict
        """
        self_fields = self._extract_own_model_fields()
        self_fields = {
            k: v
            for k, v in self_fields.items()
            if self.get_column_alias(k) in self.Meta.table.columns
        }
        for field in self._extract_db_related_names():
            target_pk_name = self.Meta.model_fields[field].to.Meta.pkname
            target_field = getattr(self, field)
            self_fields[field] = getattr(target_field, target_pk_name, None)
        return self_fields

    def get_relation_model_id(self, target_field: Type["BaseField"]) -> Optional[int]:
        """
        Returns an id of the relation side model to use in prefetch query.

        :param target_field: field with relation definition
        :type target_field: Type["BaseField"]
        :return: value of pk if set
        :rtype: Optional[int]
        """
        if target_field.virtual or issubclass(
            target_field, ormar.fields.ManyToManyField
        ):
            return self.pk
        related_name = target_field.name
        related_model = getattr(self, related_name)
        return None if not related_model else related_model.pk