Improve fsspec DataPipe to accept extra keyword arguments

test/requirements.txt

@@ -1,6 +1,7 @@
 pytest
 expecttest
 fsspec
+s3fs
 iopath == 0.1.9
 numpy
 rarfile

test/test_remote_io.py

@@ -19,13 +19,33 @@
 from torchdata.datapipes.iter import (
     EndOnDiskCacheHolder,
     FileOpener,
+    FSSpecFileLister,
+    FSSpecFileOpener,
     HttpReader,
     IterableWrapper,
     OnDiskCacheHolder,
     S3FileLister,
     S3FileLoader,
 )
 
+try:
+    import fsspec
+    import s3fs
+
+    HAS_FSSPEC_S3 = True
+except ImportError:
+    HAS_FSSPEC_S3 = False
+skipIfNoFSSpecS3 = unittest.skipIf(not HAS_FSSPEC_S3, "no FSSpec with S3fs")
+
+try:
+    from torchdata._torchdata import S3Handler
+
+    HAS_AWS = True
+except ImportError:
+    HAS_AWS = False
+skipIfAWS = unittest.skipIf(HAS_AWS, "AWSSDK Enabled")
+skipIfNoAWS = unittest.skipIf(not HAS_AWS, "No AWSSDK Enabled")
+
 
 class TestDataPipeRemoteIO(expecttest.TestCase):
     def setUp(self):
@@ -183,20 +203,43 @@ def _read_and_decode(x):
         if not IS_WINDOWS:
             dl = DataLoader(file_cache_dp, num_workers=3, multiprocessing_context="fork", batch_size=1)
             expected = [[os.path.join(self.temp_dir.name, root_dir, f"{i}.csv")] for i in range(3)] * 3
-            self.assertEqual(sorted(expected), sorted(list(dl)))
+            res = list(dl)
+            self.assertEqual(sorted(expected), sorted(res))
 
-    def test_s3_io_iterdatapipe(self):
-        # sanity test
+    @skipIfNoFSSpecS3
+    def test_fsspec_io_iterdatapipe(self):
+        input_list = [
+            (["s3://ai2-public-datasets"], 39),  # bucket without '/'
+            (["s3://ai2-public-datasets/charades/"], 18),  # bucket with '/'
+            (
+                [
+                    "s3://ai2-public-datasets/charades/Charades_v1.zip",
+                    "s3://ai2-public-datasets/charades/Charades_v1_flow.tar",
+                    "s3://ai2-public-datasets/charades/Charades_v1_rgb.tar",
+                    "s3://ai2-public-datasets/charades/Charades_v1_480.zip",
+                ],
+                4,
+            ),  # multiple files
+        ]
+        for urls, num in input_list:
+            fsspec_lister_dp = FSSpecFileLister(IterableWrapper(urls), anon=True)
+            self.assertEqual(sum(1 for _ in fsspec_lister_dp), num, f"{urls} failed")
+
+        url = "s3://ai2-public-datasets/charades/"
+        fsspec_loader_dp = FSSpecFileOpener(FSSpecFileLister(IterableWrapper([url]), anon=True), anon=True)
+        res = list(fsspec_loader_dp)
+        self.assertEqual(len(res), 18, f"{input} failed")
+
+    @skipIfAWS
+    def test_disabled_s3_io_iterdatapipe(self):
         file_urls = ["s3://ai2-public-datasets"]
-        try:
-            s3_lister_dp = S3FileLister(IterableWrapper(file_urls))
-            s3_loader_dp = S3FileLoader(IterableWrapper(file_urls))
-        except ModuleNotFoundError:
-            warnings.warn(
-                "S3 IO datapipes or C++ extension '_torchdata' isn't built in the current 'torchdata' package"
-            )
-            return
+        with self.assertRaisesRegex(ModuleNotFoundError, "TorchData must be built with"):
+            _ = S3FileLister(IterableWrapper(file_urls))
+        with self.assertRaisesRegex(ModuleNotFoundError, "TorchData must be built with"):
+            _ = S3FileLoader(IterableWrapper(file_urls))
 
+    @skipIfNoAWS
+    def test_s3_io_iterdatapipe(self):
         # S3FileLister: different inputs
         input_list = [
             [["s3://ai2-public-datasets"], 77],  # bucket without '/'

torchdata/datapipes/iter/load/fsspec.py

@@ -42,6 +42,8 @@ class FSSpecFileListerIterDataPipe(IterDataPipe[str]):
     Args:
         root: The root `fsspec` path directory or list of path directories to list files from
         masks: Unix style filter string or string list for filtering file name(s)
+        kwargs: Extra options that make sense to a particular storage connection,
+            e.g. host, port, username, password, etc.
 
     Example:
         >>> from torchdata.datapipes.iter import FSSpecFileLister
@@ -52,6 +54,7 @@ def __init__(
         self,
         root: Union[str, Sequence[str], IterDataPipe],
         masks: Union[str, List[str]] = "",
+        **kwargs,
     ) -> None:
         _assert_fsspec()
 
@@ -64,10 +67,11 @@ def __init__(
         else:
             self.datapipe = root
         self.masks = masks
+        self.kwargs = kwargs
 
     def __iter__(self) -> Iterator[str]:
         for root in self.datapipe:
-            fs, path = fsspec.core.url_to_fs(root)
+            fs, path = fsspec.core.url_to_fs(root, **self.kwargs)
 
             if isinstance(fs.protocol, str):
                 protocol_list = [fs.protocol]
@@ -88,8 +92,10 @@ def __iter__(self) -> Iterator[str]:
                     else:
                         if is_local:
                             abs_path = os.path.join(path, file_name)
-                        else:
+                        elif not file_name.startswith(path):
                             abs_path = posixpath.join(path, file_name)
+                        else:
+                            abs_path = file_name
 
                         starts_with = False
                         for protocol in protocol_list:
@@ -111,22 +117,25 @@ class FSSpecFileOpenerIterDataPipe(IterDataPipe[Tuple[str, StreamWrapper]]):
     Args:
         source_datapipe: Iterable DataPipe that provides the pathnames or URLs
         mode: An optional string that specifies the mode in which the file is opened (``"r"`` by default)
+        kwargs: Extra options that make sense to a particular storage connection,
+            e.g. host, port, username, password, etc.
 
     Example:
         >>> from torchdata.datapipes.iter import FSSpecFileLister
         >>> datapipe = FSSpecFileLister(root=dir_path)
         >>> file_dp = datapipe.open_files_by_fsspec()
     """
 
-    def __init__(self, source_datapipe: IterDataPipe[str], mode: str = "r") -> None:
+    def __init__(self, source_datapipe: IterDataPipe[str], mode: str = "r", **kwargs) -> None:
         _assert_fsspec()
 
         self.source_datapipe: IterDataPipe[str] = source_datapipe
         self.mode: str = mode
+        self.kwargs = kwargs
 
     def __iter__(self) -> Iterator[Tuple[str, StreamWrapper]]:
         for file_uri in self.source_datapipe:
-            fs, path = fsspec.core.url_to_fs(file_uri)
+            fs, path = fsspec.core.url_to_fs(file_uri, **self.kwargs)
             file = fs.open(path, self.mode)
             yield file_uri, StreamWrapper(file)
 
@@ -149,6 +158,8 @@ class FSSpecSaverIterDataPipe(IterDataPipe[str]):
         source_datapipe: Iterable DataPipe with tuples of metadata and data
         mode: Mode in which the file will be opened for write the data (``"w"`` by default)
         filepath_fn: Function that takes in metadata and returns the target path of the new file
+        kwargs: Extra options that make sense to a particular storage connection,
+            e.g. host, port, username, password, etc.
 
     Example:
         >>> from torchdata.datapipes.iter import IterableWrapper
@@ -165,17 +176,19 @@ def __init__(
         source_datapipe: IterDataPipe[Tuple[Any, U]],
         mode: str = "w",
         filepath_fn: Optional[Callable] = None,
+        **kwargs,
     ):
         _assert_fsspec()
 
         self.source_datapipe: IterDataPipe[Tuple[Any, U]] = source_datapipe
         self.mode: str = mode
         self.filepath_fn: Optional[Callable] = filepath_fn
+        self.kwargs = kwargs
 
     def __iter__(self) -> Iterator[str]:
         for meta, data in self.source_datapipe:
             filepath = meta if self.filepath_fn is None else self.filepath_fn(meta)
-            fs, path = fsspec.core.url_to_fs(filepath)
+            fs, path = fsspec.core.url_to_fs(filepath, **self.kwargs)
             with fs.open(path, self.mode) as f:
                 f.write(data)
             yield filepath

torchdata/datapipes/iter/load/s3io.py

@@ -51,7 +51,7 @@ class S3FileListerIterDataPipe(IterDataPipe[str]):
 
     def __init__(self, source_datapipe: IterDataPipe[str], length: int = -1, request_timeout_ms=-1, region="") -> None:
         if not hasattr(torchdata, "_torchdata") or not hasattr(torchdata._torchdata, "S3Handler"):
-            raise ModuleNotFoundError("Torchdata must be built with BUILD_S3=1 to use this datapipe.")
+            raise ModuleNotFoundError("TorchData must be built with BUILD_S3=1 to use this datapipe.")
 
         self.source_datapipe: IterDataPipe[str] = source_datapipe
         self.length: int = length
@@ -113,7 +113,7 @@ def __init__(
         multi_part_download=None,
     ) -> None:
         if not hasattr(torchdata, "_torchdata") or not hasattr(torchdata._torchdata, "S3Handler"):
-            raise ModuleNotFoundError("Torchdata must be built with BUILD_S3=1 to use this datapipe.")
+            raise ModuleNotFoundError("TorchData must be built with BUILD_S3=1 to use this datapipe.")
 
         self.source_datapipe: IterDataPipe[str] = source_datapipe
         self.handler = torchdata._torchdata.S3Handler(request_timeout_ms, region)