feat: from_sample for IO descriptor (#3143)

* feat: from_sample impl

Signed-off-by: Aaron Pham <29749331+aarnphm@users.noreply.github.com>

* feat: from_sample

Signed-off-by: Aaron Pham <29749331+aarnphm@users.noreply.github.com>

* fix: sample shouldn't be a memoryview [skip ci]

depends on #3144 to be merged.

Signed-off-by: Aaron Pham <29749331+aarnphm@users.noreply.github.com>

* chore: move from_sample deps to optional via io-file

Signed-off-by: Aaron Pham <29749331+aarnphm@users.noreply.github.com>

* chore: dispatch by types for sample

Signed-off-by: Aaron Pham <29749331+aarnphm@users.noreply.github.com>

* feat: openapi and dispatcher fix

Signed-off-by: Aaron Pham <29749331+aarnphm@users.noreply.github.com>

* chore: refactor and __slots__ implementation for Base

Signed-off-by: Aaron Pham <29749331+aarnphm@users.noreply.github.com>

* fix: types

Signed-off-by: Aaron Pham <29749331+aarnphm@users.noreply.github.com>

* fix: tests [skip ci]

Signed-off-by: Aaron Pham <29749331+aarnphm@users.noreply.github.com>

* fix: not using singledispatch

Signed-off-by: Aaron Pham <29749331+aarnphm@users.noreply.github.com>

* fix: tests (incremental)

Signed-off-by: Aaron Pham <29749331+aarnphm@users.noreply.github.com>

* fix: different output

Signed-off-by: Aaron Pham <29749331+aarnphm@users.noreply.github.com>

* chore: update _from_sample implementation

Signed-off-by: Aaron Pham <29749331+aarnphm@users.noreply.github.com>

Signed-off-by: Aaron Pham <29749331+aarnphm@users.noreply.github.com>

pyproject.toml

@@ -112,17 +112,22 @@ include = [
 [project.optional-dependencies]
 all = [
   "bentoml[aws]",
-  "bentoml[io-json]",
-  "bentoml[io-image]",
-  "bentoml[io-pandas]",
+  "bentoml[io]",
   "bentoml[grpc]",
   "bentoml[grpc-reflection]",
   "bentoml[grpc-channelz]",
   "bentoml[tracing]",
 ]
 aws = ["fs-s3fs"]
+io = [
+  "bentoml[io-json]",
+  "bentoml[io-image]",
+  "bentoml[io-pandas]",
+  "bentoml[io-file]",
+] # syntatic sugar for bentoml[io-json,io-image,io-pandas,io-file]
+io-file = ["filetype"] # Currently use for from_sample
 io-json = ["pydantic<2"] # currently we don't have support for pydantic 2.0
-io-image = ["Pillow"]
+io-image = ["bentoml[io-file]", "Pillow"]
 io-pandas = ["pandas", "pyarrow"]
 grpc = [
   # Restrict maximum version due to breaking protobuf 4.21.0 changes
@@ -164,6 +169,7 @@ source = ["src"]
 
 [tool.coverage.run]
 branch = true
+parallel = true
 source = ["src/bentoml/"]
 omit = [
   "src/bentoml/__main__.py",
@@ -202,7 +208,8 @@ exclude_lines = [
   "^\\s*except ImportError",
   "if __name__ == .__main__.:",
   "^\\s*if TYPE_CHECKING:",
-  "^\\s*@overload( |$)",
+  "^\\s*@(t\\.)?overload( |$)",
+  "@(abc\\.)?abstractmethod",
 ]
 
 [tool.black]

src/bentoml/_internal/io_descriptors/base.py

@@ -19,6 +19,7 @@
     from ..types import LazyType
     from ..context import InferenceApiContext as Context
     from ..service.openapi.specification import Schema
+    from ..service.openapi.specification import MediaType
     from ..service.openapi.specification import Reference
 
     InputType = (
@@ -27,6 +28,7 @@
         | LazyType[t.Any]
         | dict[str, t.Type[t.Any] | UnionType | LazyType[t.Any]]
     )
+    OpenAPIResponse = dict[str, str | dict[str, MediaType] | dict[str, t.Any]]
 
 
 IO_DESCRIPTOR_REGISTRY: dict[str, type[IODescriptor[t.Any]]] = {}
@@ -40,19 +42,48 @@ def from_spec(spec: dict[str, str]) -> IODescriptor[t.Any]:
     return IO_DESCRIPTOR_REGISTRY[spec["id"]].from_spec(spec)
 
 
-class IODescriptor(ABC, t.Generic[IOType]):
+class _OpenAPIMeta:
+    @abstractmethod
+    def openapi_schema(self) -> Schema | Reference:
+        raise NotImplementedError
+
+    @abstractmethod
+    def openapi_components(self) -> dict[str, t.Any] | None:
+        raise NotImplementedError
+
+    @abstractmethod
+    def openapi_example(self) -> t.Any | None:
+        raise NotImplementedError
+
+    @abstractmethod
+    def openapi_request_body(self) -> dict[str, t.Any]:
+        raise NotImplementedError
+
+    @abstractmethod
+    def openapi_responses(self) -> dict[str, t.Any]:
+        raise NotImplementedError
+
+
+class IODescriptor(ABC, _OpenAPIMeta, t.Generic[IOType]):
     """
     IODescriptor describes the input/output data format of an InferenceAPI defined
     in a :code:`bentoml.Service`. This is an abstract base class for extending new HTTP
     endpoint IO descriptor types in BentoServer.
     """
 
+    __slots__ = ("_args", "_kwargs", "_proto_fields", "_mime_type", "descriptor_id")
+
     HTTP_METHODS = ["POST"]
 
+    descriptor_id: str | None
+
     _mime_type: str
     _rpc_content_type: str = "application/grpc"
     _proto_fields: tuple[ProtoField]
-    descriptor_id: str | None
+    _sample: IOType | None = None
+    _initialized: bool = False
+    _args: t.Sequence[t.Any]
+    _kwargs: dict[str, t.Any]
 
     def __init_subclass__(cls, *, descriptor_id: str | None = None):
         if descriptor_id is not None:
@@ -63,52 +94,82 @@ def __init_subclass__(cls, *, descriptor_id: str | None = None):
             IO_DESCRIPTOR_REGISTRY[descriptor_id] = cls
         cls.descriptor_id = descriptor_id
 
-    @abstractmethod
-    def to_spec(self) -> dict[str, t.Any]:
-        raise NotImplementedError
+    if TYPE_CHECKING:
 
-    @classmethod
-    @abstractmethod
-    def from_spec(cls, spec: dict[str, t.Any]) -> Self:
-        raise NotImplementedError
+        def __init__(self, **kwargs: t.Any) -> None:
+            ...
+
+    def __getattr__(self, name: str) -> t.Any:
+        if not self._initialized:
+            self._lazy_init()
+        assert self._initialized
+        return object.__getattribute__(self, name)
+
+    def __dir__(self) -> t.Iterable[str]:
+        if not self._initialized:
+            self._lazy_init()
+        assert self._initialized
+        return object.__dir__(self)
 
     def __repr__(self) -> str:
         return self.__class__.__qualname__
 
-    @abstractmethod
-    def input_type(self) -> InputType:
-        raise NotImplementedError
+    def _lazy_init(self) -> None:
+        self._initialized = True
+        self.__init__(*self._args, **self._kwargs)
+        del self._args
+        del self._kwargs
+
+    @property
+    def sample(self) -> IOType | None:
+        return self._sample
+
+    @sample.setter
+    def sample(self, value: IOType) -> None:
+        self._sample = value
+
+    @classmethod
+    def from_sample(cls, sample: IOType | t.Any, **kwargs: t.Any) -> Self:
+        klass = cls(**kwargs)
+        sample = klass._from_sample(sample)
+        klass.sample = sample
+        return klass
 
     @abstractmethod
-    def openapi_schema(self) -> Schema | Reference:
+    def _from_sample(self, sample: t.Any) -> IOType:
         raise NotImplementedError
 
+    @property
+    def mime_type(self) -> str:
+        return self._mime_type
+
     @abstractmethod
-    def openapi_components(self) -> dict[str, t.Any] | None:
+    def to_spec(self) -> dict[str, t.Any]:
         raise NotImplementedError
 
+    @classmethod
     @abstractmethod
-    def openapi_request_body(self) -> dict[str, t.Any]:
+    def from_spec(cls, spec: dict[str, t.Any]) -> Self:
         raise NotImplementedError
 
     @abstractmethod
-    def openapi_responses(self) -> dict[str, t.Any]:
+    def input_type(self) -> InputType:
         raise NotImplementedError
 
     @abstractmethod
     async def from_http_request(self, request: Request) -> IOType:
-        ...
+        raise NotImplementedError
 
     @abstractmethod
     async def to_http_response(
         self, obj: IOType, ctx: Context | None = None
     ) -> Response:
-        ...
+        raise NotImplementedError
 
     @abstractmethod
     async def from_proto(self, field: t.Any) -> IOType:
-        ...
+        raise NotImplementedError
 
     @abstractmethod
     async def to_proto(self, obj: IOType) -> t.Any:
-        ...
+        raise NotImplementedError

src/bentoml/_internal/io_descriptors/file.py

@@ -1,6 +1,7 @@
 from __future__ import annotations
 
 import io
+import os
 import typing as t
 import logging
 from typing import TYPE_CHECKING
@@ -12,10 +13,12 @@
 
 from .base import IODescriptor
 from ..types import FileLike
+from ..utils import resolve_user_filepath
 from ..utils.http import set_cookies
 from ...exceptions import BadInput
 from ...exceptions import InvalidArgument
 from ...exceptions import BentoMLException
+from ...exceptions import MissingDependencyException
 from ..service.openapi import SUCCESS_DESCRIPTION
 from ..service.openapi.specification import Schema
 from ..service.openapi.specification import MediaType
@@ -112,19 +115,33 @@ async def predict(input_pdf: io.BytesIO[Any]) -> NDArray[Any]:
 
     _proto_fields = ("file",)
 
-    def __new__(  # pylint: disable=arguments-differ # returning subclass from new
-        cls, kind: FileKind = "binaryio", mime_type: str | None = None
+    def __new__(
+        cls, kind: FileKind = "binaryio", mime_type: str | None = None, **kwargs: t.Any
     ) -> File:
         mime_type = mime_type if mime_type is not None else "application/octet-stream"
         if kind == "binaryio":
-            res = object.__new__(BytesIOFile)
+            res = super().__new__(BytesIOFile, **kwargs)
         else:
             raise ValueError(f"invalid File kind '{kind}'")
         res._mime_type = mime_type
         return res
 
-    def to_spec(self) -> dict[str, t.Any]:
-        raise NotImplementedError
+    def _from_sample(self, sample: FileType | str) -> FileType:
+        try:
+            import filetype
+        except ModuleNotFoundError:
+            raise MissingDependencyException(
+                "'filetype' is required to use 'from_sample'. Install it with 'pip install bentoml[io-file]'."
+            )
+        if isinstance(sample, t.IO):
+            sample = FileLike[bytes](sample, "<sample>")
+            self._mime_type = filetype.guess_mime(sample)
+        elif isinstance(sample, (str, os.PathLike)):
+            p = resolve_user_filepath(sample, ctx=None)
+            self._mime_type = filetype.guess_mime(p)
+            with open(p, "rb") as f:
+                sample = FileLike[bytes](f, "<sample>")
+        return sample
 
     @classmethod
     def from_spec(cls, spec: dict[str, t.Any]) -> Self:
@@ -141,6 +158,9 @@ def openapi_schema(self) -> Schema:
     def openapi_components(self) -> dict[str, t.Any] | None:
         pass
 
+    def openapi_example(self):
+        pass
+
     def openapi_request_body(self) -> dict[str, t.Any]:
         return {
             "content": {self._mime_type: MediaType(schema=self.openapi_schema())},
@@ -192,7 +212,10 @@ async def to_proto(self, obj: FileType) -> pb.File:
     async def from_proto(self, field: pb.File | bytes) -> FileLike[bytes]:
         raise NotImplementedError
 
-    async def from_http_request(self, request: Request) -> t.IO[bytes]:
+    async def from_http_request(self, request: Request) -> FileLike[bytes]:
+        raise NotImplementedError
+
+    def to_spec(self) -> dict[str, t.Any]:
         raise NotImplementedError
 
 
@@ -206,7 +229,7 @@ def to_spec(self) -> dict[str, t.Any]:
             },
         }
 
-    async def from_http_request(self, request: Request) -> t.IO[bytes]:
+    async def from_http_request(self, request: Request) -> FileLike[bytes]:
         content_type, _ = parse_options_header(request.headers["content-type"])
         if content_type.decode("utf-8") == "multipart/form-data":
             form = await request.form()
@@ -228,7 +251,7 @@ async def from_http_request(self, request: Request) -> t.IO[bytes]:
             return res  # type: ignore
         if content_type.decode("utf-8") == self._mime_type:
             body = await request.body()
-            return t.cast(t.IO[bytes], FileLike(io.BytesIO(body), "<request body>"))
+            return FileLike[bytes](io.BytesIO(body), "<request body>")
         raise BentoMLException(
             f"File should have Content-Type '{self._mime_type}' or 'multipart/form-data', got {content_type} instead"
         )