Merge pull request #2 from nrbnlulu/support-deffered-types

Support deffered types

.github/workflows/test.yaml

@@ -1,5 +1,5 @@
 name: Test
-on: [push]
+on: [push, pull_request]
 jobs:
   test:
     runs-on: ubuntu-latest

.gitignore

@@ -18,3 +18,7 @@ coverage.xml
 # virtualenv
 .venv
 venv*/
+
+
+# python cached files
+*.py[cod]

aioinject/_store.py

@@ -34,23 +34,27 @@ def __init__(self) -> None:
         self._sync_exit_stack = contextlib.ExitStack()
 
     def get(self, provider: Provider[T]) -> T | Literal[NotInCache.sentinel]:
-        return self._cache.get(provider.type_, NotInCache.sentinel)
+        return self._cache.get(provider.resolve_type(), NotInCache.sentinel)
 
     def add(self, provider: Provider[T], obj: T) -> None:
         if provider.lifetime is not DependencyLifetime.transient:
-            self._cache[provider.type_] = obj
+            self._cache[provider.resolve_type()] = obj
 
     def lock(
         self,
         provider: Provider[Any],
     ) -> AbstractAsyncContextManager[bool]:
-        return contextlib.nullcontext(provider.type_ not in self._cache)
+        return contextlib.nullcontext(
+            provider.resolve_type() not in self._cache,
+        )
 
     def sync_lock(
         self,
         provider: Provider[Any],
     ) -> AbstractContextManager[bool]:
-        return contextlib.nullcontext(provider.type_ not in self._cache)
+        return contextlib.nullcontext(
+            provider.resolve_type() not in self._cache,
+        )
 
     @typing.overload
     async def enter_context(
@@ -124,9 +128,9 @@ def __init__(self) -> None:
 
     @contextlib.asynccontextmanager
     async def lock(self, provider: Provider[Any]) -> AsyncIterator[bool]:
-        if provider.type_ not in self._cache:
-            async with self._locks[provider.type_]:
-                yield provider.type_ not in self._cache
+        if provider.resolve_type() not in self._cache:
+            async with self._locks[provider.resolve_type()]:
+                yield provider.resolve_type() not in self._cache
                 return
         yield False
 
@@ -135,8 +139,8 @@ def sync_lock(
         self,
         provider: Provider[Any],
     ) -> Iterator[bool]:
-        if provider.type_ not in self._cache:
-            with self._sync_locks[provider.type_]:
-                yield provider.type_ not in self._cache
+        if provider.resolve_type() not in self._cache:
+            with self._sync_locks[provider.resolve_type()]:
+                yield provider.resolve_type() not in self._cache
                 return
         yield False

aioinject/containers.py

@@ -18,16 +18,33 @@ class Container:
     def __init__(self) -> None:
         self.providers: _Providers[Any] = {}
         self._singletons = SingletonStore()
+        self._unresolved_providers: list[Provider[Any]] = []
+        self.type_context: dict[str, type[Any]] = {}
+
+    def _resolve_unresolved_provider(self) -> None:
+        for provider in self._unresolved_providers:
+            with contextlib.suppress(NameError):
+                self._register_impl(provider)
+                self._unresolved_providers.remove(provider)
+
+    def _register_impl(self, provider: Provider[Any]) -> None:
+        provider_type = provider.resolve_type(self.type_context)
+        if provider_type in self.providers:
+            msg = f"Provider for type {provider_type} is already registered"
+            raise ValueError(msg)
+        self.providers[provider_type] = provider
+        if klass_name := getattr(provider_type, "__name__", None):
+            self.type_context[klass_name] = provider_type
 
     def register(
         self,
         provider: Provider[Any],
     ) -> None:
-        if provider.type_ in self.providers:
-            msg = f"Provider for type {provider.type_} is already registered"
-            raise ValueError(msg)
-
-        self.providers[provider.type_] = provider
+        try:
+            self._register_impl(provider)
+        except NameError:
+            self._unresolved_providers.append(provider)
+        self._resolve_unresolved_provider()
 
     def get_provider(self, type_: type[_T]) -> Provider[_T]:
         try:
@@ -53,14 +70,14 @@ def override(
         self,
         provider: Provider[Any],
     ) -> Iterator[None]:
-        previous = self.providers.get(provider.type_)
-        self.providers[provider.type_] = provider
+        previous = self.providers.get(provider.resolve_type())
+        self.providers[provider.resolve_type()] = provider
 
         yield
 
-        del self.providers[provider.type_]
+        del self.providers[provider.resolve_type()]
         if previous is not None:
-            self.providers[provider.type_] = previous
+            self.providers[provider.resolve_type()] = previous
 
     async def __aenter__(self) -> Self:
         return self

aioinject/context.py

@@ -60,7 +60,9 @@ async def resolve(
             return cached
 
         dependencies = {}
-        for dependency in provider.dependencies:
+        for dependency in provider.resolve_dependencies(
+            self._container.type_context,
+        ):
             dependencies[dependency.name] = await self.resolve(
                 type_=dependency.type_,
             )
@@ -151,7 +153,7 @@ def resolve(
             return cached
 
         dependencies = {}
-        for dependency in provider.dependencies:
+        for dependency in provider.resolve_dependencies():
             dependencies[dependency.name] = self.resolve(
                 type_=dependency.type_,
             )

aioinject/providers.py

@@ -23,6 +23,7 @@
 
 from aioinject.markers import Inject
 from aioinject.utils import (
+    _get_type_hints,
     is_context_manager_function,
     remove_annotation,
 )
@@ -62,13 +63,13 @@ def _find_inject_marker_in_annotation_args(
 
 def collect_dependencies(
     dependant: typing.Callable[..., object] | dict[str, Any],
+    ctx: dict[str, type[Any]] | None = None,
 ) -> typing.Iterable[Dependency[object]]:
     if not isinstance(dependant, dict):
         with remove_annotation(dependant.__annotations__, "return"):
-            type_hints = typing.get_type_hints(dependant, include_extras=True)
+            type_hints = _get_type_hints(dependant, context=ctx)
     else:
         type_hints = dependant
-
     for name, hint in type_hints.items():
         dep_type, args = _get_annotation_args(hint)
         inject_marker = _find_inject_marker_in_annotation_args(args)
@@ -81,15 +82,18 @@ def collect_dependencies(
         )
 
 
-def _get_provider_type_hints(provider: Provider[Any]) -> dict[str, Any]:
+def _get_provider_type_hints(
+    provider: Provider[Any],
+    context: dict[str, Any] | None = None,
+) -> dict[str, Any]:
     source = provider.impl
     if inspect.isclass(source):
         source = source.__init__
 
     if isinstance(source, functools.partial):
         return {}
 
-    type_hints = typing.get_type_hints(source, include_extras=True)
+    type_hints = _get_type_hints(source, context=context)
     for key, value in type_hints.items():
         _, args = _get_annotation_args(value)
         for arg in args:
@@ -120,11 +124,14 @@ def _get_provider_type_hints(provider: Provider[Any]) -> dict[str, Any]:
 )
 
 
-def _guess_return_type(factory: _FactoryType[_T]) -> type[_T]:
+def _guess_return_type(
+    factory: _FactoryType[_T],
+    context: dict[str, type[Any]] | None = None,
+) -> type[_T]:
     if isclass(factory):
         return typing.cast(type[_T], factory)
 
-    type_hints = typing.get_type_hints(factory)
+    type_hints = _get_type_hints(factory, context=context)
     try:
         return_type = type_hints["return"]
     except KeyError as e:
@@ -159,34 +166,59 @@ class DependencyLifetime(enum.Enum):
 
 @runtime_checkable
 class Provider(Protocol[_T]):
-    type_: type[_T]
     impl: Any
     lifetime: DependencyLifetime
+    _cached_dependencies: tuple[Dependency[object], ...]
+    _cached_type: type[_T]  # TODO: I think it is redundant.
 
     async def provide(self, kwargs: Mapping[str, Any]) -> _T:
         ...
 
     def provide_sync(self, kwargs: Mapping[str, Any]) -> _T:
         ...
 
-    @property
-    def type_hints(self) -> dict[str, Any]:
+    def _resolve_type_impl(
+        self,
+        context: dict[str, Any] | None = None,
+    ) -> type[_T]:
+        ...
+
+    def resolve_type(self, context: dict[str, Any] | None = None) -> type[_T]:
+        try:
+            return self._cached_type
+        except AttributeError:
+            self._cached_type = self._resolve_type_impl(context)
+            return self._cached_type
+
+    def type_hints(self, context: dict[str, Any] | None) -> dict[str, Any]:
         ...
 
     @property
     def is_async(self) -> bool:
         ...
 
-    @functools.cached_property
-    def dependencies(self) -> tuple[Dependency[object], ...]:
-        return tuple(collect_dependencies(self.type_hints))
+    def resolve_dependencies(
+        self,
+        context: dict[str, Any] | None = None,
+    ) -> tuple[Dependency[object], ...]:
+        try:
+            return self._cached_dependencies
+        except AttributeError:
+            self._cached_dependencies = tuple(
+                collect_dependencies(self.type_hints(context), ctx=context),
+            )
+            return self._cached_dependencies
 
     @functools.cached_property
     def is_generator(self) -> bool:
         return is_context_manager_function(self.impl)
 
-    def __repr__(self) -> str:
-        return f"{self.__class__.__qualname__}(type={self.type_}, implementation={self.impl})"
+    def __repr__(self) -> str:  # pragma: no cover
+        try:
+            type_ = repr(self.resolve_type())
+        except NameError:
+            type_ = "UNKNOWN"
+        return f"{self.__class__.__qualname__}(type={type_}, implementation={self.impl})"
 
 
 class Scoped(Provider[_T]):
@@ -197,9 +229,15 @@ def __init__(
         factory: _FactoryType[_T],
         type_: type[_T] | None = None,
     ) -> None:
-        self.type_ = type_ or _guess_return_type(factory)
+        self.type_ = type_
         self.impl = factory
 
+    def _resolve_type_impl(
+        self,
+        context: dict[str, Any] | None = None,
+    ) -> type[_T]:
+        return self.type_ or _guess_return_type(self.impl, context=context)
+
     def provide_sync(self, kwargs: Mapping[str, Any]) -> _T:
         return self.impl(**kwargs)  # type: ignore[return-value]
 
@@ -209,9 +247,11 @@ async def provide(self, kwargs: Mapping[str, Any]) -> _T:
 
         return self.provide_sync(kwargs)
 
-    @functools.cached_property
-    def type_hints(self) -> dict[str, Any]:
-        type_hints = _get_provider_type_hints(self)
+    def type_hints(
+        self,
+        context: dict[str, Any] | None = None,
+    ) -> dict[str, Any]:
+        type_hints = _get_provider_type_hints(self, context=context)
         if "return" in type_hints:
             del type_hints["return"]
         return type_hints
@@ -244,7 +284,7 @@ class Transient(Scoped[_T]):
 
 
 class Object(Provider[_T]):
-    type_hints: ClassVar[dict[str, Any]] = {}
+    _type_hints: ClassVar[dict[str, Any]] = {}
     is_async = False
     impl: _T
     lifetime = DependencyLifetime.scoped  # It's ok to cache it
@@ -257,8 +297,14 @@ def __init__(
         self.type_ = type_ or type(object_)
         self.impl = object_
 
+    def _resolve_type_impl(self, _: dict[str, Any] | None = None) -> type[_T]:
+        return self.type_
+
     def provide_sync(self, kwargs: Mapping[str, Any]) -> _T:  # noqa: ARG002
         return self.impl
 
     async def provide(self, kwargs: Mapping[str, Any]) -> _T:  # noqa: ARG002
         return self.impl
+
+    def type_hints(self, _: dict[str, Any] | None = None) -> dict[str, Any]:
+        return self._type_hints

aioinject/utils.py

@@ -94,3 +94,12 @@ def remove_annotation(
     yield
     if annotation is not sentinel:
         annotations[name] = annotation
+
+
+def _get_type_hints(
+    obj: Any,
+    context: dict[str, type[Any]] | None = None,
+) -> dict[str, Any]:
+    if not context:
+        context = {}
+    return typing.get_type_hints(obj, include_extras=True, localns=context)

aioinject/validation/_builtin.py

@@ -15,11 +15,14 @@ def all_dependencies_are_present(
 ) -> Sequence[ContainerValidationError]:
     errors = []
     for provider in container.providers.values():
-        for dependency in provider.dependencies:
-            if dependency.type_ not in container.providers:
+        for dependency in provider.resolve_dependencies(
+            container.type_context,
+        ):
+            dep_type = dependency.type_
+            if dep_type not in container.providers:
                 error = DependencyNotFoundError(
-                    message=f"Provider for type {dependency.type_} not found",
-                    dependency=dependency.type_,
+                    message=f"Provider for type {dep_type} not found",
+                    dependency=dep_type,
                 )
                 errors.append(error)
 
@@ -44,9 +47,12 @@ def __call__(
             if not self.dependant(provider):
                 continue
 
-            for dependency in provider.dependencies:
+            for dependency in provider.resolve_dependencies(
+                container.type_context,
+            ):
+                dep_type = dependency.type_
                 dependency_provider = container.get_provider(
-                    type_=dependency.type_,
+                    type_=dep_type,
                 )
                 if self.dependency(dependency_provider):
                     msg = f"Provider {provider!r} cannot depend on {dependency_provider!r}"