flyte 0.0.1b0__py3-none-any.whl

union/extras/_container.py

@@ -0,0 +1,263 @@
+import os
+import pathlib
+from typing import Any, Dict, List, Literal, Optional, Tuple, Type, Union
+
+from flyteidl.core import tasks_pb2
+
+from union import Image, storage
+from union._datastructures import NativeInterface, SerializationContext
+from union._logging import logger
+from union._task import TaskTemplate
+
+_PRIMARY_CONTAINER_NAME_FIELD = "primary_container_name"
+
+
+def _extract_command_key(cmd: str, **kwargs) -> Any:
+    """
+    Extract the key from the command using regex.
+    """
+    import re
+
+    input_regex = r"^\{\{\s*\.inputs\.(.*?)\s*\}\}$"
+    match = re.match(input_regex, cmd)
+    if match:
+        return match.group(1)
+    return None
+
+
+def _extract_path_command_key(cmd: str, input_data_dir: Optional[str]) -> Optional[str]:
+    """
+    Extract the key from the path-like command using regex.
+    """
+    import re
+
+    input_data_dir = input_data_dir or ""
+    input_regex = rf"{re.escape(input_data_dir)}/(.+)$"
+    match = re.match(input_regex, cmd)
+    if match:
+        return match.group(1)
+    return None
+
+
+class ContainerTask(TaskTemplate):
+    """
+    This is an intermediate class that represents Flyte Tasks that run a container at execution time. This is the vast
+    majority of tasks - the typical ``@task`` decorated tasks; for instance, all run a container. An example of
+    something that doesn't run a container would be something like the Athena SQL task.
+
+    :param name: Name of the task
+    :param image: The container image to use for the task. This can be a string or an Image object.
+    :param command: The command to run in the container. This can be a list of strings or a single string.
+    :param inputs: The inputs to the task. This is a dictionary of input names to types.
+    :param arguments: The arguments to pass to the command. This is a list of strings.
+    :param outputs: The outputs of the task. This is a dictionary of output names to types.
+    :param input_data_dir: The directory where the input data is stored. This is a string or a Path object.
+    :param output_data_dir: The directory where the output data is stored. This is a string or a Path object.
+    :param metadata_format: The format of the output file. This can be "JSON", "YAML", or "PROTO".
+    :param local_logs: If True, logs will be printed to the console in the local execution.
+    """
+
+    MetadataFormat = Literal["JSON", "YAML", "PROTO"]
+
+    def __init__(
+        self,
+        name: str,
+        image: Union[str, Image],
+        command: List[str],
+        inputs: Optional[Dict[str, Type]] = None,
+        arguments: Optional[List[str]] = None,
+        outputs: Optional[Dict[str, Type]] = None,
+        input_data_dir: str | pathlib.Path = "/var/inputs",
+        output_data_dir: str | pathlib.Path = "/var/outputs",
+        metadata_format: MetadataFormat = "JSON",
+        local_logs: bool = False,
+        **kwargs,
+    ):
+        super().__init__(
+            task_type="raw-container",
+            name=name,
+            image=image,
+            interface=NativeInterface({k: (v, None) for k, v in inputs.items()}, outputs),
+            **kwargs,
+        )
+        self._image = image
+        self._cmd = command
+        self._args = arguments
+        self._input_data_dir = input_data_dir
+        self._output_data_dir = output_data_dir
+        self._metadata_format = metadata_format
+        self._inputs = inputs
+        self._outputs = outputs
+        self.local_logs = local_logs
+
+    def _render_command_and_volume_binding(self, cmd: str, **kwargs) -> Tuple[str, Dict[str, Dict[str, str]]]:
+        """
+        We support template-style references to inputs, e.g., "{{.inputs.infile}}".
+
+        For FlyteFile and FlyteDirectory commands, e.g., "/var/inputs/inputs", we extract the key from strings that
+         begin with the specified `input_data_dir`.
+        """
+        # from flytekit.types.directory import FlyteDirectory
+        # from flytekit.types.file import FlyteFile
+
+        volume_binding = {}
+        path_k = _extract_path_command_key(cmd, self._input_data_dir)
+        k = path_k if path_k else _extract_command_key(cmd)
+
+        if k:
+            input_val = kwargs.get(k)
+            # TODO: Add support file and directory transformer first
+            # if type(input_val) in [FlyteFile, FlyteDirectory]:
+            #     if not path_k:
+            #         raise AssertionError(
+            #             "FlyteFile and FlyteDirectory commands should not use the template syntax like this:
+            #             {{.inputs.infile}}\n"
+            #             "Please use a path-like syntax, such as: /var/inputs/infile.\n"
+            #             "This requirement is due to how Flyte Propeller processes template syntax inputs."
+            #         )
+            #     local_flyte_file_or_dir_path = str(input_val)
+            #     remote_flyte_file_or_dir_path = os.path.join(self._input_data_dir, k)  # type: ignore
+            #     volume_binding[local_flyte_file_or_dir_path] = {
+            #         "bind": remote_flyte_file_or_dir_path,
+            #         "mode": "rw",
+            #     }
+            #     command = remote_flyte_file_or_dir_path
+            command = str(input_val)
+        else:
+            command = cmd
+
+        return command, volume_binding
+
+    def _prepare_command_and_volumes(
+        self, cmd_and_args: List[str], **kwargs
+    ) -> Tuple[List[str], Dict[str, Dict[str, str]]]:
+        """
+        Prepares the command and volume bindings for the container based on input arguments and command templates.
+
+        Parameters:
+        - cmd_and_args (List[str]): The command and arguments to prepare.
+        - **kwargs: Keyword arguments representing task inputs.
+
+        Returns:
+        - Tuple[List[str], Dict[str, Dict[str, str]]]: A tuple containing the prepared commands and volume bindings.
+        """
+
+        commands = []
+        volume_bindings = {}
+
+        for cmd in cmd_and_args:
+            command, volume_binding = self._render_command_and_volume_binding(cmd, **kwargs)
+            commands.append(command)
+            volume_bindings.update(volume_binding)
+
+        return commands, volume_bindings
+
+    def _pull_image_if_not_exists(self, client, image: str):
+        try:
+            if not client.images.list(filters={"reference": image}):
+                logger.info(f"Pulling image: {image} for container task: {self.name}")
+                client.images.pull(image)
+        except Exception as e:
+            logger.error(f"Failed to pull image {image}: {e!s}")
+            raise
+
+    def _string_to_timedelta(self, s: str):
+        import datetime
+        import re
+
+        regex = r"(?:(\d+) days?, )?(?:(\d+):)?(\d+):(\d+)(?:\.(\d+))?"
+        parts = re.match(regex, s)
+        if not parts:
+            raise ValueError("Invalid timedelta string format")
+
+        days = int(parts.group(1)) if parts.group(1) else 0
+        hours = int(parts.group(2)) if parts.group(2) else 0
+        minutes = int(parts.group(3)) if parts.group(3) else 0
+        seconds = int(parts.group(4)) if parts.group(4) else 0
+        microseconds = int(parts.group(5)) if parts.group(5) else 0
+
+        return datetime.timedelta(
+            days=days,
+            hours=hours,
+            minutes=minutes,
+            seconds=seconds,
+            microseconds=microseconds,
+        )
+
+    def _convert_output_val_to_correct_type(self, output_val: Any, output_type: Type) -> Any:
+        import datetime
+
+        if issubclass(output_type, bool):
+            return output_val.lower() != "false"
+        elif issubclass(output_type, datetime.datetime):
+            return datetime.datetime.fromisoformat(output_val)
+        elif issubclass(output_type, datetime.timedelta):
+            return self._string_to_timedelta(output_val)
+        else:
+            return output_type(output_val)
+
+    def _get_output_dict(self, output_directory: str) -> Dict[str, Any]:
+        output_dict = {}
+        if self._outputs:
+            for k, output_type in self._outputs.items():
+                output_path = os.path.join(output_directory, k)
+                with open(output_path, "r") as f:
+                    output_val = f.read()
+                output_dict[k] = self._convert_output_val_to_correct_type(output_val, output_type)
+        return output_dict
+
+    def execute(self, **kwargs) -> Any:
+        try:
+            import docker
+        except ImportError:
+            raise ImportError("Docker is not installed. Please install Docker by running `pip install docker`.")
+
+        # Normalize the input and output directories
+        self._input_data_dir = os.path.normpath(self._input_data_dir) if self._input_data_dir else ""
+        self._output_data_dir = os.path.normpath(self._output_data_dir) if self._output_data_dir else ""
+
+        output_directory = storage.get_random_local_directory()
+        cmd_and_args = (self._cmd or []) + (self._args or [])
+        commands, volume_bindings = self._prepare_command_and_volumes(cmd_and_args, **kwargs)
+        volume_bindings[output_directory] = {"bind": self._output_data_dir, "mode": "rw"}
+
+        client = docker.from_env()
+        self._pull_image_if_not_exists(client, self._image)
+
+        container = client.containers.run(
+            self._image, command=commands, remove=True, volumes=volume_bindings, detach=True
+        )
+
+        # Wait for the container to finish the task
+        # TODO: Add a 'timeout' parameter to control the max wait time for the container to finish the task.
+
+        if self.local_logs:
+            for log in container.logs(stream=True):
+                print(f"[Local Container] {log.strip()}")
+
+        container.wait()
+
+        output_dict = self._get_output_dict(output_directory)
+        return output_dict
+
+    def data_loading_config(self, sctx: SerializationContext) -> tasks_pb2.DataLoadingConfig:
+        literal_to_protobuf = {
+            "JSON": tasks_pb2.DataLoadingConfig.JSON,
+            "YAML": tasks_pb2.DataLoadingConfig.YAML,
+            "PROTO": tasks_pb2.DataLoadingConfig.PROTO,
+        }
+
+        return tasks_pb2.DataLoadingConfig(
+            input_path=self._input_data_dir,
+            output_path=self._output_data_dir,
+            enabled=True,
+            format=literal_to_protobuf.get(self._metadata_format, "JSON"),
+        )
+
+    def container_args(self, sctx: SerializationContext) -> List[str]:
+        return self._cmd + (self._args if self._args else [])
+
+    def config(self, sctx: SerializationContext) -> Optional[Dict[str, str]]:
+        if self.pod_template is None:
+            return {}
+        return {_PRIMARY_CONTAINER_NAME_FIELD: self.primary_container_name}

union/io/__init__.py

@@ -0,0 +1,11 @@
+"""
+## IO data types
+
+This package contains additional data types beyond the primitive data types in python to abstract data flow
+of large datasets in Union.
+"""
+
+__all__ = ["Dir", "File"]
+
+from ._dir import Dir
+from ._file import File

union/io/_dir.py

@@ -0,0 +1,425 @@
+from __future__ import annotations
+
+import os
+from pathlib import Path
+from typing import AsyncIterator, Generic, Iterator, List, Optional, Type, TypeVar, Union
+
+from flyteidl.core import literals_pb2, types_pb2
+from fsspec.asyn import AsyncFileSystem
+from pydantic import BaseModel, model_validator
+
+import union.storage as storage
+from union.io._file import File
+from union.types import TypeEngine, TypeTransformer, TypeTransformerFailedError
+
+# Type variable for the directory format
+T = TypeVar("T")
+
+
+class Dir(BaseModel, Generic[T]):
+    """
+    A generic directory class representing a directory with files of a specified format.
+    Provides both async and sync interfaces for directory operations.
+    Users are responsible for handling all I/O - the type transformer for Dir does not do any automatic uploading
+    or downloading of files.
+
+    The generic type T represents the format of the files in the directory.
+
+    Example:
+        ```python
+        # Async usage
+        from pandas import DataFrame
+        data_dir = Dir[DataFrame](path="s3://my-bucket/data/")
+
+        # Walk through files
+        async for file in data_dir.walk():
+            async with file.open() as f:
+                content = await f.read()
+
+        # Sync alternative
+        for file in data_dir.walk_sync():
+            with file.open_sync() as f:
+                content = f.read()
+        ```
+    """
+
+    # Represents either a local or remote path.
+    path: str
+    name: Optional[str] = None
+    format: str = ""
+
+    class Config:
+        arbitrary_types_allowed = True
+
+    @model_validator(mode="before")
+    @classmethod
+    def pre_init(cls, data):
+        if data.get("name") is None:
+            data["name"] = Path(data["path"]).name
+        return data
+
+    async def walk(self, recursive: bool = True, max_depth: Optional[int] = None) -> AsyncIterator[File[T]]:
+        """
+        Asynchronously walk through the directory and yield File objects.
+
+        Args:
+            recursive: If True, recursively walk subdirectories
+            max_depth: Maximum depth for recursive walking
+
+        Yields:
+            File objects for each file found in the directory
+
+        Example:
+            ```python
+            async for file in directory.walk():
+                local_path = await file.download()
+                # Process the file
+            ```
+        """
+        fs = storage.get_underlying_filesystem(path=self.path)
+        if recursive is False:
+            max_depth = 2
+
+        # Note if the path is actually just a file, no walking is done.
+        if isinstance(fs, AsyncFileSystem):
+            async for parent, _, files in fs._walk(self.path, maxdepth=max_depth):
+                for file in files:
+                    full_file = fs.unstrip_protocol(parent + fs.sep + file)
+                    yield File[T](path=full_file)
+        else:
+            for parent, _, files in fs.walk(self.path, maxdepth=max_depth):
+                for file in files:
+                    if "file" in fs.protocol:
+                        full_file = os.path.join(parent, file)
+                    else:
+                        full_file = fs.unstrip_protocol(parent + fs.sep + file)
+                    yield File[T](path=full_file)
+
+    def walk_sync(
+        self, recursive: bool = True, file_pattern: str = "*", max_depth: Optional[int] = None
+    ) -> Iterator[File[T]]:
+        """
+        Synchronously walk through the directory and yield File objects.
+
+        Args:
+            recursive: If True, recursively walk subdirectories
+            file_pattern: Glob pattern to filter files
+            max_depth: Maximum depth for recursive walking
+
+        Yields:
+            File objects for each file found in the directory
+
+        Example:
+            ```python
+            for file in directory.walk_sync():
+                local_path = file.download_sync()
+                # Process the file
+            ```
+        """
+        fs = storage.get_underlying_filesystem(path=self.path)
+        for parent, _, files in fs.walk(self.path, maxdepth=max_depth):
+            for file in files:
+                if "file" in fs.protocol:
+                    full_file = os.path.join(parent, file)
+                else:
+                    full_file = fs.unstrip_protocol(parent + fs.sep + file)
+                yield File[T](path=full_file)
+
+    async def list_files(self) -> List[File[T]]:
+        """
+        Asynchronously get a list of all files in the directory (non-recursive).
+
+        Returns:
+            A list of File objects
+
+        Example:
+            ```python
+            files = await directory.list_files()
+            for file in files:
+                # Process the file
+            ```
+        """
+        # todo: this should probably also just defer to fsspec.find()
+        files = []
+        async for file in self.walk(recursive=False):
+            files.append(file)
+        return files
+
+    def list_files_sync(self) -> List[File[T]]:
+        """
+        Synchronously get a list of all files in the directory (non-recursive).
+
+        Returns:
+            A list of File objects
+
+        Example:
+            ```python
+            files = directory.list_files_sync()
+            for file in files:
+                # Process the file
+            ```
+        """
+        return list(self.walk_sync(recursive=False))
+
+    async def download(self, local_path: Optional[Union[str, Path]] = None) -> str:
+        """
+        Asynchronously download the entire directory to a local path.
+
+        Args:
+            local_path: The local path to download the directory to. If None, a temporary
+                       directory will be used.
+
+        Returns:
+            The path to the downloaded directory
+
+        Example:
+            ```python
+            local_dir = await directory.download('/tmp/my_data/')
+            ```
+        """
+        local_dest = str(local_path) if local_path else str(storage.get_random_local_path())
+        if not storage.is_remote(self.path):
+            if not local_path or local_path == self.path:
+                # Skip copying
+                return self.path
+            else:
+                # Shell out to a thread to copy
+                import asyncio
+                import shutil
+
+                async def copy_tree():
+                    loop = asyncio.get_event_loop()
+                    await loop.run_in_executor(None, lambda: shutil.copytree(self.path, local_dest, dirs_exist_ok=True))
+
+                await copy_tree()
+        return await storage.get(self.path, local_dest, recursive=True)
+
+    def download_sync(self, local_path: Optional[Union[str, Path]] = None) -> str:
+        """
+        Synchronously download the entire directory to a local path.
+
+        Args:
+            local_path: The local path to download the directory to. If None, a temporary
+                       directory will be used.
+
+        Returns:
+            The path to the downloaded directory
+
+        Example:
+            ```python
+            local_dir = directory.download_sync('/tmp/my_data/')
+            ```
+        """
+        local_dest = str(local_path) if local_path else str(storage.get_random_local_path())
+        if not storage.is_remote(self.path):
+            if not local_path or local_path == self.path:
+                # Skip copying
+                return self.path
+            else:
+                # Shell out to a thread to copy
+                import shutil
+
+                shutil.copytree(self.path, local_dest, dirs_exist_ok=True)
+
+        # Figure this out when we figure out the final synchronicity story
+        raise NotImplementedError("Sync download is not implemented for remote paths")
+
+    @classmethod
+    async def from_local(cls, local_path: Union[str, Path], remote_path: Optional[str] = None) -> Dir[T]:
+        """
+        Asynchronously create a new Dir by uploading a local directory to the configured remote store.
+
+        Args:
+            local_path: Path to the local directory
+            remote_path: Optional path to store the directory remotely. If None, a path will be generated.
+
+        Returns:
+            A new Dir instance pointing to the uploaded directory
+
+        Example:
+            ```python
+            remote_dir = await Dir[DataFrame].from_local('/tmp/data_dir/', 's3://bucket/data/')
+            ```
+        """
+        local_path_str = str(local_path)
+        dirname = os.path.basename(os.path.normpath(local_path_str))
+
+        output_path = await storage.put(from_path=local_path_str, to_path=remote_path, recursive=True)
+        return cls(path=output_path, name=dirname)
+
+    @classmethod
+    def from_local_sync(cls, local_path: Union[str, Path], remote_path: Optional[str] = None) -> Dir[T]:
+        """
+        Synchronously create a new Dir by uploading a local directory to the configured remote store.
+
+        Args:
+            local_path: Path to the local directory
+            remote_path: Optional path to store the directory remotely. If None, a path will be generated.
+
+        Returns:
+            A new Dir instance pointing to the uploaded directory
+
+        Example:
+            ```python
+            remote_dir = Dir[DataFrame].from_local_sync('/tmp/data_dir/', 's3://bucket/data/')
+            ```
+        """
+        # Implement this after we figure out the final synchronicity story
+        raise NotImplementedError("Sync upload is not implemented for remote paths")
+
+    async def exists(self) -> bool:
+        """
+        Asynchronously check if the directory exists.
+
+        Returns:
+            True if the directory exists, False otherwise
+
+        Example:
+            ```python
+            if await directory.exists():
+                # Process the directory
+            ```
+        """
+        fs = storage.get_underlying_filesystem(path=self.path)
+        if isinstance(fs, AsyncFileSystem):
+            return await fs._exists(self.path)
+        else:
+            return fs.exists(self.path)
+
+    def exists_sync(self) -> bool:
+        """
+        Synchronously check if the directory exists.
+
+        Returns:
+            True if the directory exists, False otherwise
+
+        Example:
+            ```python
+            if directory.exists_sync():
+                # Process the directory
+            ```
+        """
+        fs = storage.get_underlying_filesystem(path=self.path)
+        return fs.exists(self.path)
+
+    async def get_file(self, file_name: str) -> Optional[File[T]]:
+        """
+        Asynchronously get a specific file from the directory.
+
+        Args:
+            file_name: The name of the file to get
+
+        Returns:
+            A File instance if the file exists, None otherwise
+
+        Example:
+            ```python
+            file = await directory.get_file("data.csv")
+            if file:
+                # Process the file
+            ```
+        """
+        fs = storage.get_underlying_filesystem(path=self.path)
+        file_path = fs.sep.join([self.path, file_name])
+        file = File[T](path=file_path)
+
+        if fs.exists(file_path):
+            return file
+        return None
+
+    def get_file_sync(self, file_name: str) -> Optional[File[T]]:
+        """
+        Synchronously get a specific file from the directory.
+
+        Args:
+            file_name: The name of the file to get
+
+        Returns:
+            A File instance if the file exists, None otherwise
+
+        Example:
+            ```python
+            file = directory.get_file_sync("data.csv")
+            if file:
+                # Process the file
+            ```
+        """
+        file_path = os.path.join(self.path, file_name)
+        file = File[T](path=file_path)
+
+        if file.exists_sync():
+            return file
+        return None
+
+
+class DirTransformer(TypeTransformer[Dir]):
+    """
+    Transformer for Dir objects. This type transformer does not handle any i/o. That is now the responsibility of the
+    user.
+    """
+
+    def __init__(self):
+        super().__init__(name="Dir", t=Dir)
+
+    def get_literal_type(self, t: Type[Dir]) -> types_pb2.LiteralType:
+        """Get the Flyte literal type for a File type."""
+        return types_pb2.LiteralType(
+            blob=types_pb2.BlobType(
+                # todo: set format from generic
+                format="",  # Format is determined by the generic type T
+                dimensionality=types_pb2.BlobType.BlobDimensionality.MULTIPART,
+            )
+        )
+
+    async def to_literal(
+        self,
+        python_val: Dir,
+        python_type: Type[Dir],
+        expected: types_pb2.LiteralType,
+    ) -> literals_pb2.Literal:
+        """Convert a Dir object to a Flyte literal."""
+        if not isinstance(python_val, Dir):
+            raise TypeTransformerFailedError(f"Expected Dir object, received {type(python_val)}")
+
+        return literals_pb2.Literal(
+            scalar=literals_pb2.Scalar(
+                blob=literals_pb2.Blob(
+                    metadata=literals_pb2.BlobMetadata(
+                        type=types_pb2.BlobType(
+                            format=python_val.format, dimensionality=types_pb2.BlobType.BlobDimensionality.MULTIPART
+                        )
+                    ),
+                    uri=python_val.path,
+                )
+            )
+        )
+
+    async def to_python_value(
+        self,
+        lv: literals_pb2.Literal,
+        expected_python_type: Type[Dir],
+    ) -> Dir:
+        """Convert a Flyte literal to a File object."""
+        if not lv.scalar.HasField("blob"):
+            raise TypeTransformerFailedError(f"Expected blob literal, received {lv}")
+        if not lv.scalar.blob.metadata.type.dimensionality == types_pb2.BlobType.BlobDimensionality.MULTIPART:
+            raise TypeTransformerFailedError(
+                f"Expected multipart, received {lv.scalar.blob.metadata.type.dimensionality}"
+            )
+
+        uri = lv.scalar.blob.uri
+        filename = Path(uri).name
+        f = Dir(path=uri, name=filename, format=lv.scalar.blob.metadata.type.format)
+        return f
+
+    def guess_python_type(self, literal_type: types_pb2.LiteralType) -> Type[Dir]:
+        """Guess the Python type from a Flyte literal type."""
+        if (
+            literal_type.HasField("blob")
+            and literal_type.blob.dimensionality == types_pb2.BlobType.BlobDimensionality.MULTIPART
+        ):
+            return Dir
+        raise ValueError(f"Cannot guess python type from {literal_type}")
+
+
+TypeEngine.register(DirTransformer())