Reset file pointer to 0 when reading file stream (#7304)

* Reset file pointer to 0 when reading file stream

Instead of raising a ValueError about the file pointer not being at the start of the file, reset the file pointer automatically to zero, and warn that the pointer has been reset.

* Expect warning rather than ValueError for test_open_twice

Fixes the `Failed: DID NOT RAISE <class 'ValueError'>`

* Remove checks for errors raised in test_open_fileobj

The ValueError and RuntimeWarning isn't raised anymore.

* Fix typo form -> from

* Add changelog entry for bugfix

* Remove warning about resetting file pointer to zero

File pointer is reset to zero after reading the magic byte number anyway, so should be ok not to warn about this.

Author: weiji14

doc/whats-new.rst

@@ -63,6 +63,8 @@ Bug fixes
   By `Michael Niklas <https://github.com/headtr1ck>`_.
 - Fix static typing of :py:meth:`xr.polyval` (:issue:`7312`, :pull:`7315`).
   By `Michael Niklas <https://github.com/headtr1ck>`_.
+- Fix multiple reads on fsspec S3 files by resetting file pointer to 0 when reading file streams (:issue:`6813`, :pull:`7304`).
+  By `David Hoese <https://github.com/djhoese>`_ and `Wei Ji Leong <https://github.com/weiji14>`_.
 
 Documentation
 ~~~~~~~~~~~~~

xarray/core/utils.py

@@ -656,15 +656,11 @@ def read_magic_number_from_file(filename_or_obj, count=8) -> bytes:
         magic_number = filename_or_obj[:count]
     elif isinstance(filename_or_obj, io.IOBase):
         if filename_or_obj.tell() != 0:
-            raise ValueError(
-                "cannot guess the engine, "
-                "file-like object read/write pointer not at the start of the file, "
-                "please close and reopen, or use a context manager"
-            )
+            filename_or_obj.seek(0)
         magic_number = filename_or_obj.read(count)
         filename_or_obj.seek(0)
     else:
-        raise TypeError(f"cannot read the magic number form {type(filename_or_obj)}")
+        raise TypeError(f"cannot read the magic number from {type(filename_or_obj)}")
     return magic_number
 
 

xarray/tests/test_backends.py

@@ -3159,13 +3159,12 @@ def test_open_badbytes(self) -> None:
     def test_open_twice(self) -> None:
         expected = create_test_data()
         expected.attrs["foo"] = "bar"
-        with pytest.raises(ValueError, match=r"read/write pointer not at the start"):
-            with create_tmp_file() as tmp_file:
-                expected.to_netcdf(tmp_file, engine="h5netcdf")
-                with open(tmp_file, "rb") as f:
+        with create_tmp_file() as tmp_file:
+            expected.to_netcdf(tmp_file, engine="h5netcdf")
+            with open(tmp_file, "rb") as f:
+                with open_dataset(f, engine="h5netcdf"):
                     with open_dataset(f, engine="h5netcdf"):
-                        with open_dataset(f, engine="h5netcdf"):
-                            pass
+                        pass
 
     @requires_scipy
     def test_open_fileobj(self) -> None:
@@ -3197,15 +3196,7 @@ def test_open_fileobj(self) -> None:
             # `raises_regex`?). Ref https://github.com/pydata/xarray/pull/5191
             with open(tmp_file, "rb") as f:
                 f.seek(8)
-                with pytest.raises(
-                    ValueError,
-                    match="match in any of xarray's currently installed IO",
-                ):
-                    with pytest.warns(
-                        RuntimeWarning,
-                        match=re.escape("'h5netcdf' fails while guessing"),
-                    ):
-                        open_dataset(f)
+                open_dataset(f)
 
 
 @requires_h5netcdf