cognite-neat 0.98.0__tar.gz → 0.99.1__tar.gz

{cognite_neat-0.98.0 → cognite_neat-0.99.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: cognite-neat
-Version: 0.98.0
+Version: 0.99.1
 Summary: Knowledge graph transformation
 Home-page: https://cognite-neat.readthedocs-hosted.com/
 License: Apache-2.0
@@ -41,13 +41,13 @@ Requires-Dist: mkdocs-material-extensions ; extra == "docs"
 Requires-Dist: mkdocstrings[python] ; extra == "docs"
 Requires-Dist: networkx (>=3.4.2,<4.0.0)
 Requires-Dist: openpyxl
-Requires-Dist: oxrdflib[oxigraph] (>=0.3.3,<0.4.0) ; extra == "oxi" or extra == "all"
+Requires-Dist: oxrdflib[oxigraph] (>=0.4.0,<0.5.0) ; extra == "oxi" or extra == "all"
 Requires-Dist: packaging (>=22.0,<25.0)
 Requires-Dist: pandas
 Requires-Dist: prometheus-client (>=0,<1) ; extra == "service" or extra == "all"
 Requires-Dist: pydantic (>=2,<3)
 Requires-Dist: pymdown-extensions ; extra == "docs"
-Requires-Dist: pyoxigraph (==0.3.19) ; extra == "oxi" or extra == "all"
+Requires-Dist: pyoxigraph (==0.4.3) ; extra == "oxi" or extra == "all"
 Requires-Dist: python-multipart (==0.0.9) ; extra == "service" or extra == "all"
 Requires-Dist: pyvis (>=0.3.2,<0.4.0)
 Requires-Dist: rdflib

cognite_neat-0.99.1/cognite/neat/_client/__init__.py

@@ -0,0 +1,4 @@
+from ._api.data_modeling_loaders import DataModelingLoader, ResourceLoader
+from ._api_client import NeatClient
+
+__all__ = ["NeatClient", "ResourceLoader", "DataModelingLoader"]

cognite_neat-0.99.1/cognite/neat/_client/_api/data_modeling_loaders.py

@@ -0,0 +1,585 @@
+import warnings
+from abc import ABC, abstractmethod
+from collections.abc import Sequence
+from graphlib import TopologicalSorter
+from typing import TYPE_CHECKING, Any, Generic, Literal, TypeVar, cast
+
+from cognite.client.data_classes import filters
+from cognite.client.data_classes._base import (
+    CogniteResourceList,
+    T_CogniteResourceList,
+    T_WritableCogniteResource,
+    T_WriteClass,
+    WriteableCogniteResourceList,
+)
+from cognite.client.data_classes.data_modeling import (
+    Container,
+    ContainerApply,
+    ContainerApplyList,
+    ContainerList,
+    DataModel,
+    DataModelApply,
+    DataModelApplyList,
+    DataModelList,
+    EdgeConnection,
+    MappedProperty,
+    RequiresConstraint,
+    Space,
+    SpaceApply,
+    SpaceApplyList,
+    SpaceList,
+    View,
+    ViewApply,
+    ViewApplyList,
+    ViewList,
+)
+from cognite.client.data_classes.data_modeling.ids import (
+    ContainerId,
+    DataModelId,
+    NodeId,
+    ViewId,
+)
+from cognite.client.data_classes.data_modeling.views import (
+    EdgeConnectionApply,
+    MappedPropertyApply,
+    ReverseDirectRelation,
+    ReverseDirectRelationApply,
+)
+from cognite.client.exceptions import CogniteAPIError
+from cognite.client.utils.useful_types import SequenceNotStr
+
+from cognite.neat._issues.warnings import CDFMaxIterationsWarning
+from cognite.neat._shared import T_ID
+
+if TYPE_CHECKING:
+    from cognite.neat._client._api_client import NeatClient
+
+T_WritableCogniteResourceList = TypeVar("T_WritableCogniteResourceList", bound=WriteableCogniteResourceList)
+
+
+class ResourceLoader(
+    ABC,
+    Generic[T_ID, T_WriteClass, T_WritableCogniteResource, T_CogniteResourceList, T_WritableCogniteResourceList],
+):
+    """A resource loaders is a wrapper around the Cognite SDK that does a few things:
+
+    * Standardizes the CRUD operations for a given resource.
+    * Caches the items that have been retrieved from the CDF.
+    """
+
+    resource_name: str
+
+    def __init__(self, client: "NeatClient") -> None:
+        # This is exposed to allow for disabling the cache.
+        self.cache = True
+        self._client = client
+        # This cache is used to store the items that have been retrieved from the CDF.
+        self._items_by_id: dict[T_ID, T_WritableCogniteResource] = {}
+
+    def bust_cache(self) -> None:
+        self._items_by_id = {}
+
+    @classmethod
+    @abstractmethod
+    def get_id(cls, item: T_WriteClass | T_WritableCogniteResource | dict | T_ID) -> T_ID:
+        raise NotImplementedError
+
+    @classmethod
+    def get_ids(cls, items: Sequence[T_WriteClass | T_WritableCogniteResource]) -> list[T_ID]:
+        return [cls.get_id(item) for item in items]
+
+    def create(self, items: Sequence[T_WriteClass]) -> T_WritableCogniteResourceList:
+        created = self._create(items)
+        if self.cache:
+            self._items_by_id.update({self.get_id(item): item for item in created})
+        return created
+
+    def retrieve(self, ids: SequenceNotStr[T_ID]) -> T_WritableCogniteResourceList:
+        if not self.cache:
+            return self._retrieve(ids)
+        missing_ids = [id for id in ids if id not in self._items_by_id.keys()]
+        if missing_ids:
+            retrieved = self._retrieve(missing_ids)
+            self._items_by_id.update({self.get_id(item): item for item in retrieved})
+        # We need to check the cache again, in case we didn't retrieve all the items.
+        return self._create_list([self._items_by_id[id] for id in ids if id in self._items_by_id])
+
+    def update(self, items: Sequence[T_WriteClass]) -> T_WritableCogniteResourceList:
+        if not self.cache:
+            return self._update(items)
+        updated = self._update(items)
+        self._items_by_id.update({self.get_id(item): item for item in updated})
+        return updated
+
+    def delete(self, ids: SequenceNotStr[T_ID] | Sequence[T_WriteClass]) -> list[T_ID]:
+        id_list = [self.get_id(item) for item in ids]
+        if not self.cache:
+            return self._delete(id_list)
+        ids = [self.get_id(item) for item in ids]
+        deleted = self._delete(id_list)
+        for id in deleted:
+            self._items_by_id.pop(id, None)
+        return deleted
+
+    @abstractmethod
+    def _create(self, items: Sequence[T_WriteClass]) -> T_WritableCogniteResourceList:
+        raise NotImplementedError
+
+    @abstractmethod
+    def _retrieve(self, ids: SequenceNotStr[T_ID]) -> T_WritableCogniteResourceList:
+        raise NotImplementedError
+
+    @abstractmethod
+    def _update(self, items: Sequence[T_WriteClass]) -> T_WritableCogniteResourceList:
+        raise NotImplementedError
+
+    @abstractmethod
+    def _delete(self, ids: SequenceNotStr[T_ID]) -> list[T_ID]:
+        raise NotImplementedError
+
+    @abstractmethod
+    def _create_list(self, items: Sequence[T_WritableCogniteResource]) -> T_WritableCogniteResourceList:
+        raise NotImplementedError
+
+    def are_equal(self, local: T_WriteClass, remote: T_WritableCogniteResource) -> bool:
+        return local == remote.as_write()
+
+
+class DataModelingLoader(
+    ResourceLoader[T_ID, T_WriteClass, T_WritableCogniteResource, T_CogniteResourceList, T_WritableCogniteResourceList],
+    ABC,
+):
+    @classmethod
+    def in_space(cls, item: T_WriteClass | T_WritableCogniteResource | T_ID, space: set[str]) -> bool:
+        if hasattr(item, "space"):
+            return item.space in space
+        raise ValueError(f"Item {item} does not have a space attribute")
+
+    def sort_by_dependencies(self, items: list[T_WriteClass]) -> list[T_WriteClass]:
+        return items
+
+    def create(
+        self, items: Sequence[T_WriteClass], existing_handling: Literal["fail", "skip", "update", "force"] = "fail"
+    ) -> T_WritableCogniteResourceList:
+        if existing_handling != "force":
+            return super().create(items)
+
+        created = self._create_force(items, set())
+        if self.cache:
+            self._items_by_id.update({self.get_id(item): item for item in created})
+        return created
+
+    def _create_force(
+        self,
+        items: Sequence[T_WriteClass],
+        tried_force_deploy: set[T_ID],
+    ) -> T_WritableCogniteResourceList:
+        try:
+            return self._create(items)
+        except CogniteAPIError as e:
+            failed_ids = {self.get_id(failed) for failed in e.failed}
+            to_redeploy = [
+                item
+                for item in items
+                if self.get_id(item) in failed_ids and self.get_id(item) not in tried_force_deploy
+            ]
+            if not to_redeploy:
+                # Avoid infinite loop
+                raise e
+            tried_force_deploy.update([self.get_id(item) for item in to_redeploy])
+            self.delete(to_redeploy)
+            return self._create_force(to_redeploy, tried_force_deploy)
+
+
+class SpaceLoader(DataModelingLoader[str, SpaceApply, Space, SpaceApplyList, SpaceList]):
+    resource_name = "spaces"
+
+    @classmethod
+    def get_id(cls, item: Space | SpaceApply | str | dict) -> str:
+        if isinstance(item, Space | SpaceApply):
+            return item.space
+        if isinstance(item, dict):
+            return item["space"]
+        return item
+
+    def _create(self, items: Sequence[SpaceApply]) -> SpaceList:
+        return self._client.data_modeling.spaces.apply(items)
+
+    def _retrieve(self, ids: SequenceNotStr[str]) -> SpaceList:
+        return self._client.data_modeling.spaces.retrieve(ids)
+
+    def _update(self, items: Sequence[SpaceApply]) -> SpaceList:
+        return self._create(items)
+
+    def _delete(self, ids: SequenceNotStr[str] | Sequence[Space | SpaceApply]) -> list[str]:
+        if all(isinstance(item, Space) for item in ids) or all(isinstance(item, SpaceApply) for item in ids):
+            ids = [cast(Space | SpaceApply, item).space for item in ids]
+        return self._client.data_modeling.spaces.delete(cast(SequenceNotStr[str], ids))
+
+    def _create_list(self, items: Sequence[Space]) -> SpaceList:
+        return SpaceList(items)
+
+    def clean(self, space: str) -> None:
+        """Deletes all data in a space.
+
+        This means all nodes, edges, views, containers, and data models located in the given space.
+
+        Args:
+            client: Connected CogniteClient
+            space: The space to delete.
+
+        """
+        edges = self._client.data_modeling.instances.list(
+            "edge", limit=-1, filter=filters.Equals(["edge", "space"], space)
+        )
+        if edges:
+            instances = self._client.data_modeling.instances.delete(edges=edges.as_ids())
+            print(f"Deleted {len(instances.edges)} edges")
+        nodes = self._client.data_modeling.instances.list(
+            "node", limit=-1, filter=filters.Equals(["node", "space"], space)
+        )
+        node_types = {NodeId(node.type.space, node.type.external_id) for node in nodes if node.type}
+        node_data = set(nodes.as_ids()) - node_types
+        if node_data:
+            instances = self._client.data_modeling.instances.delete(nodes=list(node_data))
+            print(f"Deleted {len(instances.nodes)} nodes")
+        if node_types:
+            instances = self._client.data_modeling.instances.delete(nodes=list(node_types))
+            print(f"Deleted {len(instances.nodes)} node types")
+        views = self._client.data_modeling.views.list(limit=-1, space=space)
+        if views:
+            deleted_views = self._client.data_modeling.views.delete(views.as_ids())
+            print(f"Deleted {len(deleted_views)} views")
+        containers = self._client.data_modeling.containers.list(limit=-1, space=space)
+        if containers:
+            deleted_containers = self._client.data_modeling.containers.delete(containers.as_ids())
+            print(f"Deleted {len(deleted_containers)} containers")
+        if data_models := self._client.data_modeling.data_models.list(limit=-1, space=space):
+            deleted_data_models = self._client.data_modeling.data_models.delete(data_models.as_ids())
+            print(f"Deleted {len(deleted_data_models)} data models")
+        deleted_space = self._client.data_modeling.spaces.delete(space)
+        print(f"Deleted space {deleted_space}")
+
+
+class ViewLoader(DataModelingLoader[ViewId, ViewApply, View, ViewApplyList, ViewList]):
+    resource_name = "views"
+
+    @classmethod
+    def get_id(cls, item: View | ViewApply | ViewId | dict) -> ViewId:
+        if isinstance(item, View | ViewApply):
+            return item.as_id()
+        if isinstance(item, dict):
+            return ViewId.load(item)
+        return item
+
+    def _create(self, items: Sequence[ViewApply]) -> ViewList:
+        return self._client.data_modeling.views.apply(items)
+
+    def retrieve(
+        self, ids: SequenceNotStr[ViewId], include_connected: bool = False, include_ancestor: bool = False
+    ) -> ViewList:
+        if not include_connected and not include_ancestor:
+            return super().retrieve(ids)
+        # Retrieve recursively updates the cache.
+        return self._retrieve_recursive(ids, include_connected, include_ancestor)
+
+    def _retrieve(self, ids: SequenceNotStr[ViewId]) -> ViewList:
+        return self._client.data_modeling.views.retrieve(cast(Sequence, ids))
+
+    def _update(self, items: Sequence[ViewApply]) -> ViewList:
+        return self._create(items)
+
+    def _delete(self, ids: SequenceNotStr[ViewId]) -> list[ViewId]:
+        return self._client.data_modeling.views.delete(cast(Sequence, ids))
+
+    def _as_write_raw(self, view: View) -> dict[str, Any]:
+        dumped = view.as_write().dump()
+        if view.properties:
+            # All read version of views have all the properties of their parent views.
+            # We need to remove these properties to compare with the local view.
+            parents = self._retrieve_recursive(view.implements or [], include_connections=False, include_ancestors=True)
+            for parent in parents:
+                for prop_name, prop in (parent.as_write().properties or {}).items():
+                    existing = dumped["properties"].get(prop_name)
+                    if existing is None:
+                        continue
+                    if existing == prop.dump():
+                        dumped["properties"].pop(prop_name, None)
+                    # If the child overrides the parent, we keep the child's property.
+
+        if "properties" in dumped and not dumped["properties"]:
+            # All properties were removed, so we remove the properties key.
+            dumped.pop("properties", None)
+        return dumped
+
+    def are_equal(self, local: ViewApply, remote: View) -> bool:
+        local_dumped = local.dump()
+        if not remote.implements:
+            return local_dumped == remote.as_write().dump()
+
+        cdf_resource_dumped = self._as_write_raw(remote)
+
+        if "properties" in local_dumped and not local_dumped["properties"]:
+            # In case the local properties are set to an empty dict.
+            local_dumped.pop("properties", None)
+
+        return local_dumped == cdf_resource_dumped
+
+    def as_write(self, view: View) -> ViewApply:
+        return ViewApply.load(self._as_write_raw(view))
+
+    def _retrieve_recursive(
+        self, view_ids: SequenceNotStr[ViewId], include_connections: bool, include_ancestors: bool
+    ) -> ViewList:
+        """Retrieves all views with the
+
+        This will mutate the cache passed in, and return a list of views that are the ancestors
+        of the views in the parents list.
+
+        Args:
+            view_ids: The views to retrieve.
+            include_connections: Whether to include all connected views.
+            include_ancestors: Whether to include all ancestors.
+        """
+        last_batch = list(view_ids)
+        found = ViewList([])
+        found_ids: set[ViewId] = set()
+        while last_batch:
+            to_retrieve_from_cdf: set[ViewId] = set()
+            batch_ids: list[ViewId] = []
+            for view_id in last_batch:
+                if view_id in found_ids:
+                    continue
+                elif view_id in self._items_by_id:
+                    view = self._items_by_id[view_id]
+                    found.append(view)
+                    batch_ids.extend(self.get_connected_views(view, include_ancestors, include_connections, found_ids))
+                else:
+                    to_retrieve_from_cdf.add(view_id)
+
+            if to_retrieve_from_cdf:
+                retrieved_batch = self._client.data_modeling.views.retrieve(list(to_retrieve_from_cdf))
+                self._items_by_id.update({view.as_id(): view for view in retrieved_batch})
+                found.extend(retrieved_batch)
+                found_ids.update({view.as_id() for view in retrieved_batch})
+                for view in retrieved_batch:
+                    batch_ids.extend(self.get_connected_views(view, include_ancestors, include_connections, found_ids))
+
+            last_batch = batch_ids
+
+        if self.cache is False:
+            # We must update the cache to retrieve recursively.
+            # If the cache is disabled, bust the cache to avoid storing the retrieved views.
+            self.bust_cache()
+        return found
+
+    @staticmethod
+    def get_connected_views(
+        view: View | ViewApply,
+        include_parents: bool = True,
+        include_connections: bool = True,
+        skip_ids: set[ViewId] | None = None,
+    ) -> list[ViewId]:
+        connected_ids: set[ViewId] = set(view.implements or []) if include_parents else set()
+        if include_connections:
+            for prop in (view.properties or {}).values():
+                if isinstance(prop, MappedProperty | MappedPropertyApply) and prop.source:
+                    connected_ids.add(prop.source)
+                elif isinstance(
+                    prop, EdgeConnection | EdgeConnectionApply | ReverseDirectRelation | ReverseDirectRelationApply
+                ):
+                    connected_ids.add(prop.source)
+
+                if isinstance(prop, EdgeConnection | EdgeConnectionApply) and prop.edge_source:
+                    connected_ids.add(prop.edge_source)
+                elif isinstance(prop, ReverseDirectRelation | ReverseDirectRelationApply) and isinstance(
+                    prop.through.source, ViewId
+                ):
+                    connected_ids.add(prop.through.source)
+        if skip_ids:
+            return [view_id for view_id in connected_ids if view_id not in skip_ids]
+        return list(connected_ids)
+
+    def _create_list(self, items: Sequence[View]) -> ViewList:
+        return ViewList(items)
+
+
+class ContainerLoader(DataModelingLoader[ContainerId, ContainerApply, Container, ContainerApplyList, ContainerList]):
+    resource_name = "containers"
+
+    @classmethod
+    def get_id(cls, item: Container | ContainerApply | ContainerId | dict) -> ContainerId:
+        if isinstance(item, Container | ContainerApply):
+            return item.as_id()
+        if isinstance(item, dict):
+            return ContainerId.load(item)
+        return item
+
+    def sort_by_dependencies(self, items: Sequence[ContainerApply]) -> list[ContainerApply]:
+        container_by_id = {container.as_id(): container for container in items}
+        container_dependencies = {
+            container.as_id(): {
+                const.require
+                for const in container.constraints.values()
+                if isinstance(const, RequiresConstraint) and const.require in container_by_id
+            }
+            for container in items
+        }
+        return [
+            container_by_id[container_id] for container_id in TopologicalSorter(container_dependencies).static_order()
+        ]
+
+    def _create(self, items: Sequence[ContainerApply]) -> ContainerList:
+        return self._client.data_modeling.containers.apply(items)
+
+    def retrieve(self, ids: SequenceNotStr[ContainerId], include_connected: bool = False) -> ContainerList:
+        if not include_connected:
+            return super().retrieve(ids)
+        # Retrieve recursively updates the cache.
+        return self._retrieve_recursive(ids)
+
+    def _retrieve(self, ids: SequenceNotStr[ContainerId]) -> ContainerList:
+        return self._client.data_modeling.containers.retrieve(cast(Sequence, ids))
+
+    def _update(self, items: Sequence[ContainerApply]) -> ContainerList:
+        return self._create(items)
+
+    def _delete(self, ids: SequenceNotStr[ContainerId]) -> list[ContainerId]:
+        return self._client.data_modeling.containers.delete(cast(Sequence, ids))
+
+    def _create_list(self, items: Sequence[Container]) -> ContainerList:
+        return ContainerList(items)
+
+    def _retrieve_recursive(self, container_ids: SequenceNotStr[ContainerId]) -> ContainerList:
+        """Containers can reference each other through the 'requires' constraint.
+
+        This method retrieves all containers that are referenced by other containers through the 'requires' constraint,
+        including their parents.
+        """
+        max_iterations = 10  # Limiting the number of iterations to avoid infinite loops
+        found = ContainerList([])
+        found_ids: set[ContainerId] = set()
+        last_batch = list(container_ids)
+        for _ in range(max_iterations):
+            if not last_batch:
+                break
+            to_retrieve_from_cdf: set[ContainerId] = set()
+            batch_ids: list[ContainerId] = []
+            for container_id in last_batch:
+                if container_id in found_ids:
+                    continue
+                elif container_id in self._items_by_id:
+                    container = self._items_by_id[container_id]
+                    found.append(container)
+                    batch_ids.extend(self.get_connected_containers(container, found_ids))
+                else:
+                    to_retrieve_from_cdf.add(container_id)
+
+            if to_retrieve_from_cdf:
+                retrieved_batch = self._client.data_modeling.containers.retrieve(list(to_retrieve_from_cdf))
+                self._items_by_id.update({view.as_id(): view for view in retrieved_batch})
+                found.extend(retrieved_batch)
+                found_ids.update({view.as_id() for view in retrieved_batch})
+                for container in retrieved_batch:
+                    batch_ids.extend(self.get_connected_containers(container, found_ids))
+
+            last_batch = batch_ids
+        else:
+            warnings.warn(
+                CDFMaxIterationsWarning(
+                    "The maximum number of iterations was reached while resolving referenced containers."
+                    "There might be referenced containers that are not included in the list of containers.",
+                    max_iterations=max_iterations,
+                ),
+                stacklevel=2,
+            )
+
+        if self.cache is False:
+            # We must update the cache to retrieve recursively.
+            # If the cache is disabled, bust the cache to avoid storing the retrieved views.
+            self.bust_cache()
+        return found
+
+    @staticmethod
+    def get_connected_containers(
+        container: Container | ContainerApply, skip: set[ContainerId] | None = None
+    ) -> set[ContainerId]:
+        connected_containers = set()
+        for constraint in container.constraints.values():
+            if isinstance(constraint, RequiresConstraint):
+                connected_containers.add(constraint.require)
+        if skip:
+            return {container_id for container_id in connected_containers if container_id not in skip}
+        return connected_containers
+
+    def are_equal(self, local: ContainerApply, remote: Container) -> bool:
+        local_dumped = local.dump(camel_case=True)
+        if "usedFor" not in local_dumped:
+            # Setting used_for to "node" as it is the default value in the CDF.
+            local_dumped["usedFor"] = "node"
+
+        return local_dumped == remote.as_write().dump(camel_case=True)
+
+
+class DataModelLoader(DataModelingLoader[DataModelId, DataModelApply, DataModel, DataModelApplyList, DataModelList]):
+    resource_name = "data_models"
+
+    @classmethod
+    def get_id(cls, item: DataModel | DataModelApply | DataModelId | dict) -> DataModelId:
+        if isinstance(item, DataModel | DataModelApply):
+            return item.as_id()
+        if isinstance(item, dict):
+            return DataModelId.load(item)
+        return item
+
+    def _create(self, items: Sequence[DataModelApply]) -> DataModelList:
+        return self._client.data_modeling.data_models.apply(items)
+
+    def _retrieve(self, ids: SequenceNotStr[DataModelId]) -> DataModelList:
+        return self._client.data_modeling.data_models.retrieve(cast(Sequence, ids))
+
+    def _update(self, items: Sequence[DataModelApply]) -> DataModelList:
+        return self._create(items)
+
+    def _delete(self, ids: SequenceNotStr[DataModelId]) -> list[DataModelId]:
+        return self._client.data_modeling.data_models.delete(cast(Sequence, ids))
+
+    def _create_list(self, items: Sequence[DataModel]) -> DataModelList:
+        return DataModelList(items)
+
+    def are_equal(self, local: DataModelApply, remote: DataModel) -> bool:
+        local_dumped = local.dump()
+        cdf_resource_dumped = remote.as_write().dump()
+
+        # Data models that have the same views, but in different order, are considered equal.
+        # We also account for whether views are given as IDs or View objects.
+        local_dumped["views"] = sorted(
+            (v if isinstance(v, ViewId) else v.as_id()).as_tuple() for v in local.views or []
+        )
+        cdf_resource_dumped["views"] = sorted(
+            (v if isinstance(v, ViewId) else v.as_id()).as_tuple() for v in remote.views or []
+        )
+
+        return local_dumped == cdf_resource_dumped
+
+
+class DataModelLoaderAPI:
+    def __init__(self, client: "NeatClient") -> None:
+        self._client = client
+        self.spaces = SpaceLoader(client)
+        self.views = ViewLoader(client)
+        self.containers = ContainerLoader(client)
+        self.data_models = DataModelLoader(client)
+
+    def get_loader(self, items: Any) -> DataModelingLoader:
+        if isinstance(items, CogniteResourceList):
+            resource_name = type(items).__name__.casefold().removesuffix("list").removesuffix("apply")
+        elif isinstance(items, str):
+            resource_name = items
+        else:
+            raise ValueError(f"Cannot determine resource name from {items}")
+        if resource_name[-1] != "s":
+            resource_name += "s"
+        if resource_name == "datamodels":
+            resource_name = "data_models"
+        return getattr(self, resource_name)