fix: be more more caution when claiming a backend can open a URL

Author: ianhi

xarray/backends/netCDF4_.py

@@ -702,7 +702,12 @@ class NetCDF4BackendEntrypoint(BackendEntrypoint):
 
     def guess_can_open(self, filename_or_obj: T_PathFileOrDataStore) -> bool:
         if isinstance(filename_or_obj, str) and is_remote_uri(filename_or_obj):
-            return True
+            # For remote URIs, check file extension to avoid claiming non-netCDF URLs
+            # (e.g., remote Zarr stores)
+            _, ext = os.path.splitext(filename_or_obj.rstrip("/"))
+            # Accept remote URIs with netCDF extensions or no extension
+            # (OPeNDAP endpoints often have no extension)
+            return ext in {".nc", ".nc4", ".cdf", ""}
 
         magic_number = (
             bytes(filename_or_obj[:8])

xarray/backends/pydap_.py

@@ -1,5 +1,6 @@
 from __future__ import annotations
 
+import os
 from collections.abc import Iterable
 from typing import TYPE_CHECKING, Any
 
@@ -209,7 +210,14 @@ class PydapBackendEntrypoint(BackendEntrypoint):
     url = "https://docs.xarray.dev/en/stable/generated/xarray.backends.PydapBackendEntrypoint.html"
 
     def guess_can_open(self, filename_or_obj: T_PathFileOrDataStore) -> bool:
-        return isinstance(filename_or_obj, str) and is_remote_uri(filename_or_obj)
+        if not (isinstance(filename_or_obj, str) and is_remote_uri(filename_or_obj)):
+            return False
+
+        # Check file extension to avoid claiming non-OPeNDAP URLs (e.g., remote Zarr stores)
+        _, ext = os.path.splitext(filename_or_obj.rstrip("/"))
+        # Pydap handles OPeNDAP endpoints, which typically have no extension or .nc/.nc4
+        # Reject URLs with non-OPeNDAP extensions like .zarr
+        return ext not in {".zarr", ".zip", ".tar", ".gz"}
 
     def open_dataset(
         self,

xarray/tests/test_backends.py

@@ -7252,6 +7252,77 @@ def test_zarr_entrypoint(tmp_path: Path) -> None:
     assert not entrypoint.guess_can_open("something.zarr.txt")
 
 
+@requires_netCDF4
+@requires_pydap
+@requires_zarr
+def test_remote_url_backend_auto_detection() -> None:
+    """
+    Test that remote URLs are correctly claimed by appropriate backends.
+
+    This tests the fix for issue where netCDF4 and pydap backends were
+    claiming ALL remote URLs, preventing remote Zarr stores from being
+    auto-detected.
+
+    See: https://github.com/pydata/xarray/issues/XXXXX
+    """
+    from xarray.backends.netCDF4_ import NetCDF4BackendEntrypoint
+    from xarray.backends.pydap_ import PydapBackendEntrypoint
+    from xarray.backends.zarr import ZarrBackendEntrypoint
+
+    netcdf4_entrypoint = NetCDF4BackendEntrypoint()
+    pydap_entrypoint = PydapBackendEntrypoint()
+    zarr_entrypoint = ZarrBackendEntrypoint()
+
+    # Remote Zarr URLs should be claimed by Zarr backend, not netCDF4/pydap
+    remote_zarr_urls = [
+        "https://example.com/store.zarr",
+        "http://example.com/data.zarr/",
+        "s3://bucket/path/to/data.zarr",
+    ]
+
+    for url in remote_zarr_urls:
+        assert zarr_entrypoint.guess_can_open(url), f"Zarr should claim {url}"
+        assert not netcdf4_entrypoint.guess_can_open(url), (
+            f"NetCDF4 should not claim {url}"
+        )
+        assert not pydap_entrypoint.guess_can_open(url), f"Pydap should not claim {url}"
+
+    # Remote netCDF URLs with extensions should be claimed by netCDF4, not Zarr
+    remote_netcdf_urls_with_ext = [
+        "https://example.com/file.nc",
+        "http://example.com/data.nc4",
+        "https://example.com/test.cdf",
+    ]
+
+    for url in remote_netcdf_urls_with_ext:
+        assert not zarr_entrypoint.guess_can_open(url), f"Zarr should not claim {url}"
+        assert netcdf4_entrypoint.guess_can_open(url), f"NetCDF4 should claim {url}"
+
+    # OPeNDAP endpoints (no extension) should be claimed by both netCDF4 and pydap
+    opendap_urls = [
+        "http://opendap.example.com/data",
+        "https://test.opendap.org/dataset",
+    ]
+
+    for url in opendap_urls:
+        assert not zarr_entrypoint.guess_can_open(url), f"Zarr should not claim {url}"
+        assert netcdf4_entrypoint.guess_can_open(url), f"NetCDF4 should claim {url}"
+        assert pydap_entrypoint.guess_can_open(url), f"Pydap should claim {url}"
+
+    # Other file types should not be claimed
+    other_urls = [
+        "https://example.com/data.zip",
+        "https://example.com/data.tar.gz",
+    ]
+
+    for url in other_urls:
+        assert not zarr_entrypoint.guess_can_open(url), f"Zarr should not claim {url}"
+        assert not netcdf4_entrypoint.guess_can_open(url), (
+            f"NetCDF4 should not claim {url}"
+        )
+        assert not pydap_entrypoint.guess_can_open(url), f"Pydap should not claim {url}"
+
+
 @requires_netCDF4
 @pytest.mark.parametrize("str_type", (str, np.str_))
 def test_write_file_from_np_str(str_type: type[str | np.str_], tmpdir: str) -> None: