Stream Directories and Files using Flyte (#1512)

Signed-off-by: Ketan Umare <ketan.umare@gmail.com>
Signed-off-by: Niels Bantilan <niels.bantilan@gmail.com>
Signed-off-by: Yee Hing Tong <wild-endeavor@users.noreply.github.com>

Dockerfile.dev

@@ -12,27 +12,21 @@ MAINTAINER Flyte Team <users@flyte.org>
 LABEL org.opencontainers.image.source https://github.com/flyteorg/flytekit
 
 WORKDIR /root
-ENV PYTHONPATH /root
 
 ARG VERSION
-ARG DOCKER_IMAGE
 
 RUN apt-get update && apt-get install build-essential vim -y
 
-COPY . /code/flytekit
-WORKDIR /code/flytekit
+COPY . /flytekit
 
 # Pod tasks should be exposed in the default image
-RUN pip install -e .
-RUN pip install -e plugins/flytekit-k8s-pod
-RUN pip install -e plugins/flytekit-deck-standard
+RUN pip install -e /flytekit
+RUN pip install -e /flytekit/plugins/flytekit-k8s-pod
+RUN pip install -e /flytekit/plugins/flytekit-deck-standard
 RUN pip install scikit-learn
 
-ENV PYTHONPATH "/code/flytekit:/code/flytekit/plugins/flytekit-k8s-pod:/code/flytekit/plugins/flytekit-deck-standard:"
+ENV PYTHONPATH "/flytekit:/flytekit/plugins/flytekit-k8s-pod:/flytekit/plugins/flytekit-deck-standard:"
 
-WORKDIR /root
 RUN useradd -u 1000 flytekit
 RUN chown flytekit: /root
 USER flytekit
-
-ENV FLYTE_INTERNAL_IMAGE "$DOCKER_IMAGE"

flytekit/clis/sdk_in_container/run.py

@@ -81,7 +81,7 @@ def convert(
  raise ValueError(
  f"Currently only directories containing one file are supported, found [{len(files)}] files found in {p.resolve()}"
  )
- return Directory(dir_path=value, local_file=files[0].resolve())
+ return Directory(dir_path=str(p), local_file=files[0].resolve())
  raise click.BadParameter(f"parameter should be a valid directory path, {value}")
 
 

flytekit/core/data_persistence.py

@@ -107,16 +107,16 @@ def data_config(self) -> DataConfig:
  return self._data_config
 
  def get_filesystem(
- self, protocol: typing.Optional[str] = None, anonymous: bool = False
+ self, protocol: typing.Optional[str] = None, anonymous: bool = False, **kwargs
  ) -> typing.Optional[fsspec.AbstractFileSystem]:
  if not protocol:
  return self._default_remote
- kwargs = {} # type: typing.Dict[str, typing.Any]
  if protocol == "file":
- kwargs = {"auto_mkdir": True}
+ kwargs["auto_mkdir"] = True
  elif protocol == "s3":
- kwargs = s3_setup_args(self._data_config.s3, anonymous=anonymous)
- return fsspec.filesystem(protocol, **kwargs) # type: ignore
+ s3kwargs = s3_setup_args(self._data_config.s3, anonymous=anonymous)
+ s3kwargs.update(kwargs)
+ return fsspec.filesystem(protocol, **s3kwargs) # type: ignore
  elif protocol == "gs":
  if anonymous:
  kwargs["token"] = _ANON
@@ -128,9 +128,9 @@ def get_filesystem(
 
  return fsspec.filesystem(protocol, **kwargs) # type: ignore
 
- def get_filesystem_for_path(self, path: str = "", anonymous: bool = False) -> fsspec.AbstractFileSystem:
+ def get_filesystem_for_path(self, path: str = "", anonymous: bool = False, **kwargs) -> fsspec.AbstractFileSystem:
  protocol = get_protocol(path)
- return self.get_filesystem(protocol, anonymous=anonymous)
+ return self.get_filesystem(protocol, anonymous=anonymous, **kwargs)
 
  @staticmethod
  def is_remote(path: Union[str, os.PathLike]) -> bool:

flytekit/core/type_engine.py

@@ -129,7 +129,6 @@ def to_python_value(self, ctx: FlyteContext, lv: Literal, expected_python_type:
  f"Conversion to python value expected type {expected_python_type} from literal not implemented"
  )
 
- @abstractmethod
  def to_html(self, ctx: FlyteContext, python_val: T, expected_python_type: Type[T]) -> str:
  """
  Converts any python val (dataframe, int, float) to a html string, and it will be wrapped in the HTML div

flytekit/types/directory/types.py

@@ -2,20 +2,25 @@
 
 import os
 import pathlib
+import random
 import typing
 from dataclasses import dataclass, field
 from pathlib import Path
+from typing import Any, Generator, Tuple
+from uuid import UUID
 
+import fsspec
 from dataclasses_json import config, dataclass_json
+from fsspec.utils import get_protocol
 from marshmallow import fields
 
-from flytekit.core.context_manager import FlyteContext
+from flytekit.core.context_manager import FlyteContext, FlyteContextManager
 from flytekit.core.type_engine import TypeEngine, TypeTransformer
 from flytekit.models import types as _type_models
 from flytekit.models.core import types as _core_types
 from flytekit.models.literals import Blob, BlobMetadata, Literal, Scalar
 from flytekit.models.types import LiteralType
-from flytekit.types.file import FileExt
+from flytekit.types.file import FileExt, FlyteFile
 
 T = typing.TypeVar("T")
 PathType = typing.Union[str, os.PathLike]
@@ -148,6 +153,18 @@ def __fspath__(self):
  def extension(cls) -> str:
  return ""
 
+ @classmethod
+ def new_remote(cls) -> FlyteDirectory:
+ """
+ Create a new FlyteDirectory object using the currently configured default remote in the context (i.e.
+ the raw_output_prefix configured in the current FileAccessProvider object in the context).
+ This is used if you explicitly have a folder somewhere that you want to create files under.
+ If you want to write a whole folder, you can let your task return a FlyteDirectory object,
+ and let flytekit handle the uploading.
+ """
+ d = FlyteContext.current_context().file_access.get_random_remote_directory()
+ return FlyteDirectory(path=d)
+
  def __class_getitem__(cls, item: typing.Union[typing.Type, str]) -> typing.Type[FlyteDirectory]:
  if item is None:
  return cls
@@ -176,6 +193,12 @@ def downloaded(self) -> bool:
  def remote_directory(self) -> typing.Optional[str]:
  return self._remote_directory
 
+ @property
+ def sep(self) -> str:
+ if os.name == "nt" and get_protocol(self.path or self.remote_source or self.remote_directory) == "file":
+ return "\\"
+ return "/"
+
  @property
  def remote_source(self) -> str:
  """
@@ -184,9 +207,67 @@ def remote_source(self) -> str:
  """
  return typing.cast(str, self._remote_source)
 
+ def new_file(self, name: typing.Optional[str] = None) -> FlyteFile:
+ """
+ This will create a new file under the current folder.
+ If given a name, it will use the name given, otherwise it'll pick a random string.
+ Collisions are not checked.
+ """
+ # TODO we may want to use - https://github.com/fsspec/universal_pathlib
+ if not name:
+ name = UUID(int=random.getrandbits(128)).hex
+ new_path = self.sep.join([str(self.path).rstrip(self.sep), name]) # trim trailing sep if any and join
+ return FlyteFile(path=new_path)
+
+ def new_dir(self, name: typing.Optional[str] = None) -> FlyteDirectory:
+ """
+ This will create a new folder under the current folder.
+ If given a name, it will use the name given, otherwise it'll pick a random string.
+ Collisions are not checked.
+ """
+ if not name:
+ name = UUID(int=random.getrandbits(128)).hex
+
+ new_path = self.sep.join([str(self.path).rstrip(self.sep), name]) # trim trailing sep if any and join
+ return FlyteDirectory(path=new_path)
+
  def download(self) -> str:
  return self.__fspath__()
 
+ def crawl(
+ self, maxdepth: typing.Optional[int] = None, topdown: bool = True, **kwargs
+ ) -> Generator[Tuple[typing.Union[str, os.PathLike[Any]], typing.Dict[Any, Any]], None, None]:
+ """
+ Crawl returns a generator of all files prefixed by any sub-folders under the given "FlyteDirectory".
+ if details=True is passed, then it will return a dictionary as specified by fsspec.
+
+ Example:
+
+ >>> list(fd.crawl())
+ [("/base", "file1"), ("/base", "dir1/file1"), ("/base", "dir2/file1"), ("/base", "dir1/dir/file1")]
+
+ >>> list(x.crawl(detail=True))
+ [('/tmp/test', {'my-dir/ab.py': {'name': '/tmp/test/my-dir/ab.py', 'size': 0, 'type': 'file',
+ 'created': 1677720780.2318847, 'islink': False, 'mode': 33188, 'uid': 501, 'gid': 0,
+ 'mtime': 1677720780.2317934, 'ino': 1694329, 'nlink': 1}})]
+ """
+ final_path = self.path
+ if self.remote_source:
+ final_path = self.remote_source
+ elif self.remote_directory:
+ final_path = self.remote_directory
+ ctx = FlyteContextManager.current_context()
+ fs = ctx.file_access.get_filesystem_for_path(final_path)
+ base_path_len = len(fsspec.core.strip_protocol(final_path)) + 1 # Add additional `/` at the end
+ for base, _, files in fs.walk(final_path, maxdepth, topdown, **kwargs):
+ current_base = base[base_path_len:]
+ if isinstance(files, dict):
+ for f, v in files.items():
+ yield final_path, {os.path.join(current_base, f): v}
+ else:
+ for f in files:
+ yield final_path, os.path.join(current_base, f)
+
  def __repr__(self):
  return self.path
 

flytekit/types/file/file.py

@@ -3,12 +3,13 @@
 import os
 import pathlib
 import typing
+from contextlib import contextmanager
 from dataclasses import dataclass, field
 
 from dataclasses_json import config, dataclass_json
 from marshmallow import fields
 
-from flytekit.core.context_manager import FlyteContext
+from flytekit.core.context_manager import FlyteContext, FlyteContextManager
 from flytekit.core.type_engine import TypeEngine, TypeTransformer, TypeTransformerFailedError
 from flytekit.loggers import logger
 from flytekit.models.core.types import BlobType
@@ -27,7 +28,9 @@ def noop():
 @dataclass_json
 @dataclass
 class FlyteFile(os.PathLike, typing.Generic[T]):
- path: typing.Union[str, os.PathLike] = field(default=None, metadata=config(mm_field=fields.String())) # type: ignore
+ path: typing.Union[str, os.PathLike] = field(
+ default=None, metadata=config(mm_field=fields.String())
+ ) # type: ignore
  """
  Since there is no native Python implementation of files and directories for the Flyte Blob type, (like how int
  exists for Flyte's Integer type) we need to create one so that users can express that their tasks take
@@ -148,6 +151,15 @@ def t2() -> flytekit_typing.FlyteFile["csv"]:
  def extension(cls) -> str:
  return ""
 
+ @classmethod
+ def new_remote_file(cls, name: typing.Optional[str] = None) -> FlyteFile:
+ """
+ Create a new FlyteFile object with a remote path.
+ """
+ ctx = FlyteContextManager.current_context()
+ remote_path = ctx.file_access.get_random_remote_path(name)
+ return cls(path=remote_path)
+
  def __class_getitem__(cls, item: typing.Union[str, typing.Type]) -> typing.Type[FlyteFile]:
  from . import FileExt
 
@@ -226,6 +238,57 @@ def remote_source(self) -> str:
  def download(self) -> str:
  return self.__fspath__()
 
+ @contextmanager
+ def open(
+ self,
+ mode: str,
+ cache_type: typing.Optional[str] = None,
+ cache_options: typing.Optional[typing.Dict[str, typing.Any]] = None,
+ ):
+ """
+ Returns a streaming File handle
+
+ .. code-block:: python
+
+ @task
+ def copy_file(ff: FlyteFile) -> FlyteFile:
+ new_file = FlyteFile.new_remote_file(ff.name)
+ with ff.open("rb", cache_type="readahead", cache={}) as r:
+ with new_file.open("wb") as w:
+ w.write(r.read())
+ return new_file
+
+ Alternatively
+
+ .. code-block:: python
+
+ @task
+ def copy_file(ff: FlyteFile) -> FlyteFile:
+ new_file = FlyteFile.new_remote_file(ff.name)
+ with fsspec.open(f"readahead::{ff.remote_path}", "rb", readahead={}) as r:
+ with new_file.open("wb") as w:
+ w.write(r.read())
+ return new_file
+
+
+ :param mode: str Open mode like 'rb', 'rt', 'wb', ...
+ :param cache_type: optional str Specify if caching is to be used. Cache protocol can be ones supported by
+ fsspec https://filesystem-spec.readthedocs.io/en/latest/api.html#readbuffering,
+ especially useful for large file reads
+ :param cache_options: optional Dict[str, Any] Refer to fsspec caching options. This is strongly coupled to the
+ cache_protocol
+ """
+ ctx = FlyteContextManager.current_context()
+ final_path = self.path
+ if self.remote_source:
+ final_path = self.remote_source
+ elif self.remote_path:
+ final_path = self.remote_path
+ fs = ctx.file_access.get_filesystem_for_path(final_path)
+ f = fs.open(final_path, mode, cache_type=cache_type, cache_options=cache_options)
+ yield f
+ f.close()
+
  def __repr__(self):
  return self.path
 

flytekit/types/structured/basic_dfs.py

@@ -62,22 +62,6 @@ def encode(
  structured_dataset_type.format = PARQUET
  return literals.StructuredDataset(uri=uri, metadata=StructuredDatasetMetadata(structured_dataset_type))
 
- def ddencode(
- self,
- ctx: FlyteContext,
- structured_dataset: StructuredDataset,
- structured_dataset_type: StructuredDatasetType,
- ) -> literals.StructuredDataset:
-
- path = typing.cast(str, structured_dataset.uri) or ctx.file_access.get_random_remote_directory()
- df = typing.cast(pd.DataFrame, structured_dataset.dataframe)
- local_dir = ctx.file_access.get_random_local_directory()
- local_path = os.path.join(local_dir, f"{0:05}")
- df.to_parquet(local_path, coerce_timestamps="us", allow_truncated_timestamps=False)
- ctx.file_access.upload_directory(local_dir, path)
- structured_dataset_type.format = PARQUET
- return literals.StructuredDataset(uri=path, metadata=StructuredDatasetMetadata(structured_dataset_type))
-
 
 class ParquetToPandasDecodingHandler(StructuredDatasetDecoder):
  def __init__(self):
@@ -101,20 +85,6 @@ def decode(
  kwargs = get_storage_options(ctx.file_access.data_config, uri, anon=True)
  return pd.read_parquet(uri, columns=columns, storage_options=kwargs)
 
- def dcccecode(
- self,
- ctx: FlyteContext,
- flyte_value: literals.StructuredDataset,
- current_task_metadata: StructuredDatasetMetadata,
- ) -> pd.DataFrame:
- path = flyte_value.uri
- local_dir = ctx.file_access.get_random_local_directory()
- ctx.file_access.get_data(path, local_dir, is_multipart=True)
- if current_task_metadata.structured_dataset_type and current_task_metadata.structured_dataset_type.columns:
- columns = [c.name for c in current_task_metadata.structured_dataset_type.columns]
- return pd.read_parquet(local_dir, columns=columns)
- return pd.read_parquet(local_dir)
-
 
 class ArrowToParquetEncodingHandler(StructuredDatasetEncoder):
  def __init__(self):