Deterministic chunk padding

changes/2755.bugfix.rst

@@ -0,0 +1 @@
+Upon creation, an empty ``zarr.core.buffer.cpu.NDBuffer`` is filled with zeros to ensure deterministic behavior.

src/zarr/core/buffer/cpu.py

@@ -154,10 +154,9 @@ def create(
         order: Literal["C", "F"] = "C",
         fill_value: Any | None = None,
     ) -> Self:
-        ret = cls(np.empty(shape=tuple(shape), dtype=dtype, order=order))
-        if fill_value is not None:
-            ret.fill(fill_value)
-        return ret
+        return cls(
+            np.full(shape=tuple(shape), fill_value=fill_value or 0, dtype=dtype, order=order)
+        )
 
     @classmethod
     def from_numpy_array(cls, array_like: npt.ArrayLike) -> Self:

tests/test_store/test_memory.py

@@ -1,12 +1,18 @@
 from __future__ import annotations
 
+from typing import TYPE_CHECKING
+
 import pytest
 
+import zarr
 from zarr.core.buffer import Buffer, cpu, gpu
 from zarr.storage import GpuMemoryStore, MemoryStore
 from zarr.testing.store import StoreTests
 from zarr.testing.utils import gpu_test
 
+if TYPE_CHECKING:
+    from zarr.core.common import ZarrFormat
+
 
 class TestMemoryStore(StoreTests[MemoryStore, cpu.Buffer]):
     store_cls = MemoryStore
@@ -46,6 +52,28 @@ def test_store_supports_partial_writes(self, store: MemoryStore) -> None:
     def test_list_prefix(self, store: MemoryStore) -> None:
         assert True
 
+    @pytest.mark.parametrize("dtype", ["uint8", "float32", "int64"])
+    @pytest.mark.parametrize("zarr_format", [2, 3])
+    async def test_deterministic_size(
+        self, store: MemoryStore, dtype, zarr_format: ZarrFormat
+    ) -> None:
+        def padding_size() -> int:
+            a = zarr.empty(
+                store=store,
+                shape=(3,),
+                chunks=(1000,),
+                dtype=dtype,
+                zarr_format=zarr_format,
+                overwrite=True,
+            )
+            a[...] = 1
+            key = "0" if zarr_format == 2 else "c/0"
+            return len(store._store_dict[key])
+
+        l1 = padding_size()
+        l2 = padding_size()
+        assert l1 == l2
+
 
 @gpu_test
 class TestGpuMemoryStore(StoreTests[GpuMemoryStore, gpu.Buffer]):