remove insertion of vlen-string codec for v2 metadata creation

changes/3100.bugfix.rst

@@ -0,0 +1,3 @@
+For Zarr format 2, allow fixed-length string arrays to be created without automatically inserting a
+``Vlen-UT8`` codec in the array of filters. Fixed-length string arrays do not need this codec. This
+change fixes a regression where fixed-length string arrays created with Zarr Python 3 could not be read with Zarr Python 2.18.

src/zarr/core/array.py

@@ -768,14 +768,6 @@ def _create_metadata_v2(
 
         dtype = parse_dtype(dtype, zarr_format=2)
 
-        # inject VLenUTF8 for str dtype if not already present
-        if np.issubdtype(dtype, np.str_):
-            filters = filters or []
-            from numcodecs.vlen import VLenUTF8
-
-            if not any(isinstance(x, VLenUTF8) or x["id"] == "vlen-utf8" for x in filters):
-                filters = list(filters) + [VLenUTF8()]
-
         return ArrayV2Metadata(
             shape=shape,
             dtype=np.dtype(dtype),

tests/test_array.py

@@ -1245,7 +1245,7 @@ async def test_invalid_v3_arguments(
             zarr.create(store=store, dtype="uint8", shape=(10,), zarr_format=3, **kwargs)
 
     @staticmethod
-    @pytest.mark.parametrize("dtype", ["uint8", "float32", "str"])
+    @pytest.mark.parametrize("dtype", ["uint8", "float32", "str", "U10", "S10", ">M8[10s]"])
     @pytest.mark.parametrize(
         "compressors",
         [