Merge branch 'main' into feat/numcodecs-protocol

Author: d-v-b

changes/3288.misc.rst

@@ -0,0 +1 @@
+Sort dictionary keys before returning consolidated metadata to ensure deterministic output.

docs/user-guide/consolidated_metadata.rst

@@ -45,7 +45,7 @@ that can be used.:
    >>> consolidated = zarr.open_group(store=store)
    >>> consolidated_metadata = consolidated.metadata.consolidated_metadata.metadata
    >>> from pprint import pprint
-   >>> pprint(dict(sorted(consolidated_metadata.items())))
+   >>> pprint(dict(consolidated_metadata.items()))
    {'a': ArrayV3Metadata(shape=(1,),
                           data_type=Float64(endianness='little'),
                           chunk_grid=RegularChunkGrid(chunk_shape=(1,)),
@@ -100,6 +100,14 @@ With nested groups, the consolidated metadata is available on the children, recu
    >>> consolidated['child'].metadata.consolidated_metadata
    ConsolidatedMetadata(metadata={'child': GroupMetadata(attributes={'kind': 'grandchild'}, zarr_format=3, consolidated_metadata=ConsolidatedMetadata(metadata={}, kind='inline', must_understand=False), node_type='group')}, kind='inline', must_understand=False)
 
+.. versionadded:: 3.1.1
+
+    The keys in the consolidated metadata are sorted prior to writing. Keys are
+    sorted in ascending order by path depth, where a path is defined as a sequence
+    of strings joined by ``"/"``. For keys with the same path length, lexicographic
+    order is used to break the tie.  This behaviour ensures deterministic metadata
+    output for a given group.
+
 Synchronization and Concurrency
 -------------------------------
 

src/zarr/core/group.py

@@ -4,6 +4,7 @@
 import itertools
 import json
 import logging
+import unicodedata
 import warnings
 from collections import defaultdict
 from dataclasses import asdict, dataclass, field, fields, replace
@@ -141,7 +142,16 @@ def to_dict(self) -> dict[str, JSON]:
         return {
             "kind": self.kind,
             "must_understand": self.must_understand,
-            "metadata": {k: v.to_dict() for k, v in self.flattened_metadata.items()},
+            "metadata": {
+                k: v.to_dict()
+                for k, v in sorted(
+                    self.flattened_metadata.items(),
+                    key=lambda item: (
+                        item[0].count("/"),
+                        unicodedata.normalize("NFKC", item[0]).casefold(),
+                    ),
+                )
+            },
         }
 
     @classmethod

tests/test_metadata/test_consolidated.py

@@ -467,6 +467,35 @@ def test_to_dict_empty(self):
         }
         assert result == expected
 
+    @pytest.mark.parametrize("zarr_format", [2, 3])
+    async def test_to_dict_order(
+        self, memory_store: zarr.storage.MemoryStore, zarr_format: ZarrFormat
+    ) -> None:
+        with zarr.config.set(default_zarr_format=zarr_format):
+            g = await group(store=memory_store)
+
+            # Create groups in non-lexicographix order
+            dtype = "float32"
+            await g.create_array(name="b", shape=(1,), dtype=dtype)
+            child = await g.create_group("c", attributes={"key": "child"})
+            await g.create_array(name="a", shape=(1,), dtype=dtype)
+
+            await child.create_array("e", shape=(1,), dtype=dtype)
+            await child.create_array("d", shape=(1,), dtype=dtype)
+
+            # Consolidate metadata and re-open store
+            await zarr.api.asynchronous.consolidate_metadata(memory_store)
+            g2 = await zarr.api.asynchronous.open_group(store=memory_store)
+
+            assert list(g2.metadata.consolidated_metadata.metadata) == ["a", "b", "c"]
+            assert list(g2.metadata.consolidated_metadata.flattened_metadata) == [
+                "a",
+                "b",
+                "c",
+                "c/d",
+                "c/e",
+            ]
+
     @pytest.mark.parametrize("zarr_format", [2, 3])
     async def test_open_consolidated_raises_async(self, zarr_format: ZarrFormat):
         store = zarr.storage.MemoryStore()