Fix negative slicing of Zarr arrays

Closes pydata#8252
Closes pydata#3921

Author: dcherian

xarray/backends/zarr.py

@@ -86,10 +86,17 @@ def get_array(self):
     def _oindex(self, key):
         return self._array.oindex[key]
 
+    def _getitem(self, key):
+        return self._array[key]
+
     def __getitem__(self, key):
         array = self._array
         if isinstance(key, indexing.BasicIndexer):
-            return array[key.tuple]
+            # this will convert negative slices to positive slices
+            # The latter are all that Zarr supports
+            return indexing.explicit_indexing_adapter(
+                key, array.shape, indexing.IndexingSupport.VECTORIZED, self._getitem
+            )
         elif isinstance(key, indexing.VectorizedIndexer):
             return array.vindex[
                 indexing._arrayize_vectorized_indexer(key, self.shape).tuple

xarray/tests/test_backends.py

@@ -2929,14 +2929,6 @@ def test_attributes(self, obj) -> None:
             with pytest.raises(TypeError, match=r"Invalid attribute in Dataset.attrs."):
                 ds.to_zarr(store_target, **self.version_kwargs)
 
-    def test_vectorized_indexing_negative_step(self) -> None:
-        if not has_dask:
-            pytest.xfail(
-                reason="zarr without dask handles negative steps in slices incorrectly"
-            )
-
-        super().test_vectorized_indexing_negative_step()
-
 
 @requires_zarr
 class TestZarrDictStore(ZarrBase):