automatically add filter

src/zarr/core/array.py

@@ -223,16 +223,16 @@ async def create(
     ) -> AsyncArray:
         store_path = await make_store_path(store)
 
-        dtype = parse_dtype(dtype, zarr_format)
+        dtype_parsed = parse_dtype(dtype, zarr_format)
         shape = parse_shapelike(shape)
 
         if chunks is not None and chunk_shape is not None:
             raise ValueError("Only one of chunk_shape or chunks can be provided.")
 
         if chunks:
-            _chunks = normalize_chunks(chunks, shape, dtype.itemsize)
+            _chunks = normalize_chunks(chunks, shape, dtype_parsed.itemsize)
         else:
-            _chunks = normalize_chunks(chunk_shape, shape, dtype.itemsize)
+            _chunks = normalize_chunks(chunk_shape, shape, dtype_parsed.itemsize)
 
         if zarr_format == 3:
             if dimension_separator is not None:
@@ -254,7 +254,7 @@ async def create(
             result = await cls._create_v3(
                 store_path,
                 shape=shape,
-                dtype=dtype,
+                dtype=dtype_parsed,
                 chunk_shape=_chunks,
                 fill_value=fill_value,
                 chunk_key_encoding=chunk_key_encoding,
@@ -264,6 +264,14 @@ async def create(
                 exists_ok=exists_ok,
             )
         elif zarr_format == 2:
+            if dtype is str or dtype == "str":
+                # another special case: zarr v2 added the vlen-utf8 codec
+                vlen_codec: dict[str, JSON] = {"id": "vlen-utf8"}
+                if filters and not any(x["id"] == "vlen-utf8" for x in filters):
+                    filters = list(filters) + [vlen_codec]
+                else:
+                    filters = [vlen_codec]
+
             if codecs is not None:
                 raise ValueError(
                     "codecs cannot be used for arrays with version 2. Use filters and compressor instead."
@@ -277,7 +285,7 @@ async def create(
             result = await cls._create_v2(
                 store_path,
                 shape=shape,
-                dtype=dtype,
+                dtype=dtype_parsed,
                 chunks=_chunks,
                 dimension_separator=dimension_separator,
                 fill_value=fill_value,

tests/v3/test_v2.py

@@ -2,6 +2,7 @@
 from collections.abc import Iterator
 from typing import Any
 
+import numcodecs.vlen
 import numpy as np
 import pytest
 from numcodecs import Delta
@@ -118,3 +119,4 @@ async def test_create_dtype_str(dtype: Any) -> None:
     arr = zarr.create(shape=10, dtype=dtype, zarr_format=2)
     assert arr.dtype.kind == "O"
     assert arr.metadata.to_dict()["dtype"] == "|O"
+    assert arr.metadata.filters == (numcodecs.vlen.VLenUTF8(),)