Add AioS3FileSystem for native asyncio S3 operations

Implement AioS3FileSystem as an fsspec AsyncFileSystem that composes
a sync S3FileSystem internally and dispatches operations via
asyncio.to_thread and asyncio.gather for parallelism.

AioS3File is a minimal S3File subclass — all async behavior is
provided by the injected S3AioExecutor, eliminating the need
for method overrides.

Wire AioS3FSCursor to use AioS3FileSystem and add pluggable
filesystem_class parameter to AthenaS3FSResultSet.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: laughingman7743

pyathena/aio/s3fs/cursor.py

@@ -8,6 +8,7 @@
 from pyathena.aio.common import WithAsyncFetch
 from pyathena.common import CursorIterator
 from pyathena.error import OperationalError, ProgrammingError
+from pyathena.filesystem.s3_async import AioS3FileSystem
 from pyathena.model import AthenaQueryExecution
 from pyathena.s3fs.converter import DefaultS3FSTypeConverter
 from pyathena.s3fs.result_set import AthenaS3FSResultSet, CSVReaderType
@@ -16,11 +17,12 @@
 
 
 class AioS3FSCursor(WithAsyncFetch):
-    """Native asyncio cursor that reads CSV results via S3FileSystem.
+    """Native asyncio cursor that reads CSV results via AioS3FileSystem.
 
-    Uses ``asyncio.to_thread()`` for result set creation and fetch operations
-    because ``AthenaS3FSResultSet`` lazily streams rows from S3 via a CSV
-    reader, making fetch calls blocking I/O.
+    Uses ``AioS3FileSystem`` for S3 operations, which replaces
+    ``ThreadPoolExecutor`` parallelism with ``asyncio.gather`` +
+    ``asyncio.to_thread``. Fetch operations are wrapped in
+    ``asyncio.to_thread()`` because CSV reading is blocking I/O.
 
     Example:
         >>> async with await pyathena.aio_connect(...) as conn:
@@ -127,6 +129,7 @@ async def execute(  # type: ignore[override]
                 arraysize=self.arraysize,
                 retry_config=self._retry_config,
                 csv_reader=self._csv_reader,
+                filesystem_class=AioS3FileSystem,
                 **kwargs,
             )
         else:

pyathena/filesystem/s3_async.py

@@ -0,0 +1,357 @@
+# -*- coding: utf-8 -*-
+from __future__ import annotations
+
+import asyncio
+import logging
+from multiprocessing import cpu_count
+from typing import TYPE_CHECKING, Any, Dict, List, Optional, Tuple, Union, cast
+
+from fsspec.asyn import AsyncFileSystem
+from fsspec.callbacks import _DEFAULT_CALLBACK
+
+from pyathena.filesystem.s3 import S3File, S3FileSystem
+from pyathena.filesystem.s3_executor import S3AioExecutor
+from pyathena.filesystem.s3_object import S3Object
+
+if TYPE_CHECKING:
+    from datetime import datetime
+
+    from pyathena.connection import Connection
+
+_logger = logging.getLogger(__name__)
+
+
+class AioS3FileSystem(AsyncFileSystem):
+    """An async filesystem interface for Amazon S3 using fsspec's AsyncFileSystem.
+
+    This class wraps ``S3FileSystem`` to provide native asyncio support. Instead of
+    using ``ThreadPoolExecutor`` for parallel operations, it uses ``asyncio.gather``
+    with ``asyncio.to_thread`` for natural integration with the asyncio event loop.
+
+    The implementation uses composition: an internal ``S3FileSystem`` instance handles
+    all boto3 calls, while this class delegates to it via ``asyncio.to_thread()``.
+    This avoids diamond inheritance issues and keeps all boto3 logic in one place.
+
+    File handles created by ``_open`` use ``S3AioExecutor`` so that parallel
+    operations (range reads, multipart uploads) are dispatched via the event loop
+    instead of spawning additional threads.
+
+    Attributes:
+        _sync_fs: The internal synchronous S3FileSystem instance.
+
+    Example:
+        >>> from pyathena.filesystem.s3_async import AioS3FileSystem
+        >>> fs = AioS3FileSystem(asynchronous=True)
+        >>>
+        >>> # Use in async context
+        >>> files = await fs._ls('s3://my-bucket/data/')
+        >>>
+        >>> # Sync wrappers also available (auto-generated by fsspec)
+        >>> files = fs.ls('s3://my-bucket/data/')
+    """
+
+    # https://docs.aws.amazon.com/AmazonS3/latest/API/API_DeleteObjects.html
+    DELETE_OBJECTS_MAX_KEYS: int = 1000
+
+    protocol = ("s3", "s3a")
+    mirror_sync_methods = True
+    async_impl = True
+    _extra_tokenize_attributes = ("default_block_size",)
+
+    def __init__(
+        self,
+        connection: Optional["Connection[Any]"] = None,
+        default_block_size: Optional[int] = None,
+        default_cache_type: Optional[str] = None,
+        max_workers: int = (cpu_count() or 1) * 5,
+        s3_additional_kwargs: Optional[Dict[str, Any]] = None,
+        asynchronous: bool = False,
+        loop: Optional[Any] = None,
+        batch_size: Optional[int] = None,
+        **kwargs,
+    ) -> None:
+        super().__init__(
+            asynchronous=asynchronous,
+            loop=loop,
+            batch_size=batch_size,
+            **kwargs,
+        )
+        self._sync_fs = S3FileSystem(
+            connection=connection,
+            default_block_size=default_block_size,
+            default_cache_type=default_cache_type,
+            max_workers=max_workers,
+            s3_additional_kwargs=s3_additional_kwargs,
+            **kwargs,
+        )
+        # Share dircache for cache coherence between async and sync instances
+        self.dircache = self._sync_fs.dircache
+
+    @staticmethod
+    def parse_path(path: str) -> Tuple[str, Optional[str], Optional[str]]:
+        return S3FileSystem.parse_path(path)
+
+    async def _info(self, path: str, **kwargs) -> S3Object:
+        return await asyncio.to_thread(self._sync_fs.info, path, **kwargs)
+
+    async def _ls(
+        self, path: str, detail: bool = False, **kwargs
+    ) -> Union[List[S3Object], List[str]]:
+        return await asyncio.to_thread(self._sync_fs.ls, path, detail=detail, **kwargs)
+
+    async def _cat_file(
+        self, path: str, start: Optional[int] = None, end: Optional[int] = None, **kwargs
+    ) -> bytes:
+        return await asyncio.to_thread(self._sync_fs.cat_file, path, start=start, end=end, **kwargs)
+
+    async def _exists(self, path: str, **kwargs) -> bool:
+        return await asyncio.to_thread(self._sync_fs.exists, path, **kwargs)
+
+    async def _rm_file(self, path: str, **kwargs) -> None:
+        await asyncio.to_thread(self._sync_fs.rm_file, path, **kwargs)
+
+    async def _pipe_file(self, path: str, value: bytes, **kwargs) -> None:
+        await asyncio.to_thread(self._sync_fs.pipe_file, path, value, **kwargs)
+
+    async def _put_file(self, lpath: str, rpath: str, callback=_DEFAULT_CALLBACK, **kwargs) -> None:
+        await asyncio.to_thread(self._sync_fs.put_file, lpath, rpath, callback=callback, **kwargs)
+
+    async def _get_file(self, rpath: str, lpath: str, callback=_DEFAULT_CALLBACK, **kwargs) -> None:
+        await asyncio.to_thread(self._sync_fs.get_file, rpath, lpath, callback=callback, **kwargs)
+
+    async def _mkdir(self, path: str, create_parents: bool = True, **kwargs) -> None:
+        await asyncio.to_thread(self._sync_fs.mkdir, path, create_parents=create_parents, **kwargs)
+
+    async def _makedirs(self, path: str, exist_ok: bool = False) -> None:
+        await asyncio.to_thread(self._sync_fs.makedirs, path, exist_ok=exist_ok)
+
+    async def _rm(self, path: Union[str, List[str]], recursive: bool = False, **kwargs) -> None:
+        """Remove files or directories using async parallel batch deletion.
+
+        For multiple paths, chunks into batches of 1000 (S3 API limit) and uses
+        ``asyncio.gather`` with ``asyncio.to_thread`` instead of ThreadPoolExecutor.
+        """
+        if isinstance(path, str):
+            path = [path]
+
+        bucket, _, _ = self.parse_path(path[0])
+
+        expand_paths: List[str] = []
+        for p in path:
+            expanded = await asyncio.to_thread(self._sync_fs.expand_path, p, recursive=recursive)
+            expand_paths.extend(expanded)
+
+        if not expand_paths:
+            return
+
+        quiet = kwargs.pop("Quiet", True)
+        delete_objects: List[Dict[str, Any]] = []
+        for p in expand_paths:
+            _, key, version_id = self.parse_path(p)
+            if key:
+                object_: Dict[str, Any] = {"Key": key}
+                if version_id:
+                    object_["VersionId"] = version_id
+                delete_objects.append(object_)
+
+        if not delete_objects:
+            return
+
+        chunks = [
+            delete_objects[i : i + self.DELETE_OBJECTS_MAX_KEYS]
+            for i in range(0, len(delete_objects), self.DELETE_OBJECTS_MAX_KEYS)
+        ]
+
+        async def _delete_chunk(chunk: List[Dict[str, Any]]) -> None:
+            request = {
+                "Bucket": bucket,
+                "Delete": {
+                    "Objects": chunk,
+                    "Quiet": quiet,
+                },
+            }
+            await asyncio.to_thread(
+                self._sync_fs._call, self._sync_fs._client.delete_objects, **request
+            )
+
+        await asyncio.gather(*[_delete_chunk(chunk) for chunk in chunks])
+
+        for p in expand_paths:
+            self._sync_fs.invalidate_cache(p)
+
+    async def _cp_file(self, path1: str, path2: str, **kwargs) -> None:
+        """Copy an S3 object, using async parallel multipart upload for large files."""
+        kwargs.pop("onerror", None)
+        bucket1, key1, version_id1 = self.parse_path(path1)
+        bucket2, key2, version_id2 = self.parse_path(path2)
+        if version_id2:
+            raise ValueError("Cannot copy to a versioned file.")
+        if not key1 or not key2:
+            raise ValueError("Cannot copy buckets.")
+
+        info1 = await self._info(path1)
+        size1 = info1.get("size", 0)
+        if size1 <= S3FileSystem.MULTIPART_UPLOAD_MAX_PART_SIZE:
+            await asyncio.to_thread(
+                self._sync_fs._copy_object,
+                bucket1=bucket1,
+                key1=key1,
+                version_id1=version_id1,
+                bucket2=bucket2,
+                key2=key2,
+                **kwargs,
+            )
+        else:
+            await self._copy_object_with_multipart_upload(
+                bucket1=bucket1,
+                key1=key1,
+                version_id1=version_id1,
+                size1=size1,
+                bucket2=bucket2,
+                key2=key2,
+                **kwargs,
+            )
+        self._sync_fs.invalidate_cache(path2)
+
+    async def _copy_object_with_multipart_upload(
+        self,
+        bucket1: str,
+        key1: str,
+        size1: int,
+        bucket2: str,
+        key2: str,
+        block_size: Optional[int] = None,
+        version_id1: Optional[str] = None,
+        **kwargs,
+    ) -> None:
+        block_size = block_size if block_size else S3FileSystem.MULTIPART_UPLOAD_MAX_PART_SIZE
+        if (
+            block_size < S3FileSystem.MULTIPART_UPLOAD_MIN_PART_SIZE
+            or block_size > S3FileSystem.MULTIPART_UPLOAD_MAX_PART_SIZE
+        ):
+            raise ValueError("Block size must be greater than 5MiB and less than 5GiB.")
+
+        copy_source: Dict[str, Any] = {
+            "Bucket": bucket1,
+            "Key": key1,
+        }
+        if version_id1:
+            copy_source["VersionId"] = version_id1
+
+        ranges = S3File._get_ranges(
+            0,
+            size1,
+            self._sync_fs.max_workers,
+            block_size,
+        )
+        multipart_upload = await asyncio.to_thread(
+            self._sync_fs._create_multipart_upload,
+            bucket=bucket2,
+            key=key2,
+            **kwargs,
+        )
+
+        async def _upload_part(i: int, range_: Tuple[int, int]) -> Dict[str, Any]:
+            result = await asyncio.to_thread(
+                self._sync_fs._upload_part_copy,
+                bucket=bucket2,
+                key=key2,
+                copy_source=copy_source,
+                upload_id=cast(str, multipart_upload.upload_id),
+                part_number=i + 1,
+                copy_source_ranges=range_,
+            )
+            return {
+                "ETag": result.etag,
+                "PartNumber": result.part_number,
+            }
+
+        parts = await asyncio.gather(*[_upload_part(i, r) for i, r in enumerate(ranges)])
+        parts_list = sorted(parts, key=lambda x: x["PartNumber"])
+
+        await asyncio.to_thread(
+            self._sync_fs._complete_multipart_upload,
+            bucket=bucket2,
+            key=key2,
+            upload_id=cast(str, multipart_upload.upload_id),
+            parts=parts_list,
+        )
+
+    async def _find(
+        self,
+        path: str,
+        maxdepth: Optional[int] = None,
+        withdirs: bool = False,
+        **kwargs,
+    ) -> Union[Dict[str, S3Object], List[str]]:
+        detail = kwargs.pop("detail", False)
+        files = await asyncio.to_thread(
+            self._sync_fs._find, path, maxdepth=maxdepth, withdirs=withdirs, **kwargs
+        )
+        if detail:
+            return {f.name: f for f in files}
+        return [f.name for f in files]
+
+    def _open(
+        self,
+        path: str,
+        mode: str = "rb",
+        block_size: Optional[int] = None,
+        cache_type: Optional[str] = None,
+        autocommit: bool = True,
+        cache_options: Optional[Dict[Any, Any]] = None,
+        **kwargs,
+    ) -> "AioS3File":
+        if block_size is None:
+            block_size = self._sync_fs.default_block_size
+        if cache_type is None:
+            cache_type = self._sync_fs.default_cache_type
+        max_workers = kwargs.pop("max_worker", self._sync_fs.max_workers)
+        s3_additional_kwargs = kwargs.pop("s3_additional_kwargs", {})
+        s3_additional_kwargs.update(self._sync_fs.s3_additional_kwargs)
+
+        return AioS3File(
+            self._sync_fs,
+            path,
+            mode,
+            version_id=None,
+            max_workers=max_workers,
+            executor=S3AioExecutor(loop=self._loop),
+            block_size=block_size,
+            cache_type=cache_type,
+            autocommit=autocommit,
+            cache_options=cache_options,
+            s3_additional_kwargs=s3_additional_kwargs,
+            **kwargs,
+        )
+
+    def sign(self, path: str, expiration: int = 3600, **kwargs) -> str:
+        return cast(str, self._sync_fs.sign(path, expiration=expiration, **kwargs))
+
+    def checksum(self, path: str, **kwargs) -> int:
+        return cast(int, self._sync_fs.checksum(path, **kwargs))
+
+    def created(self, path: str) -> "datetime":
+        return self._sync_fs.created(path)
+
+    def modified(self, path: str) -> "datetime":
+        return self._sync_fs.modified(path)
+
+    def invalidate_cache(self, path: Optional[str] = None) -> None:
+        self._sync_fs.invalidate_cache(path)
+
+    async def _touch(self, path: str, truncate: bool = True, **kwargs) -> None:
+        await asyncio.to_thread(self._sync_fs.touch, path, truncate=truncate, **kwargs)
+
+
+class AioS3File(S3File):
+    """Async-aware S3 file handle using ``S3AioExecutor``.
+
+    Functionally identical to ``S3File``; exists as a distinct type for
+    ``isinstance`` checks and to document the async execution model.
+    All parallel operations (range reads, multipart uploads) are dispatched
+    through the ``S3Executor`` interface — the ``S3AioExecutor``
+    provided by ``AioS3FileSystem`` uses the event loop instead of threads.
+    """
+
+    pass