Fix platform fallback metadata and compatibility boundaries

Author: FrancescAlted

src/blosc2/ctable_nulls.py

@@ -958,7 +958,9 @@ def null_pred(self):
             return None
         if _is_nan(sentinel):
             return blosc2.isnan(col._raw_col)
-        return col._raw_col == sentinel
+        # This internal predicate must remain lazy even while expression fusion
+        # temporarily disables user-facing overloaded equality for identity checks.
+        return blosc2.LazyExpr(new_op=(col._raw_col, "==", sentinel))
 
     def valid_pred(self):
         """Lazy predicate over the raw physical array, True where the value is *not* null.

src/blosc2/expression_graph.py

@@ -1008,8 +1008,11 @@ def _root_reduction_dtype(name, receiver, arguments, backend):
     ):
         # Single-threaded WASM uses the existing NumPy reduction fallback,
         # whose default integer accumulator/index width is platform-sized.
-        # Do not replace that numerical contract with native Blosc2's int64.
-        return None
+        # Dummy-array inference can still return native int64 for positional
+        # argmin/argmax. State the NumPy fallback's width explicitly instead.
+        return np.dtype(
+            np.uintp if input_dtype.kind == "u" and name not in {"argmin", "argmax"} else np.intp
+        )
     if name in _CUMULATIVE_OPERATIONS and backend != "numpy" and requested is not None:
         # The current Blosc2 cumulative implementation does not consistently
         # apply dtype overrides. Preserve its inference rather than claiming

src/blosc2/zarr_source.py

@@ -43,6 +43,12 @@ def counting_store(zarr, store, traffic, *, metadata=None):
     class CountingStore(zarr.storage.WrapperStore):
         _blosc2_traffic = traffic
 
+        def with_read_only(self, read_only=False):
+            # Zarr 3.1's WrapperStore inherits an unimplemented base method.
+            # Retain traffic/metadata accounting when Zarr opens a writable
+            # authorized store for reading.
+            return CountingStore(self._store.with_read_only(read_only))
+
         async def get(self, key, prototype, byte_range=None):
             is_metadata = (
                 metadata is not None

tests/ctable/test_null_channel.py

@@ -269,14 +269,16 @@ class Row:
     np.testing.assert_array_equal(col.is_null(), [False, True])
 
 
-def test_null_pred_matches_is_null_for_sentinel_column():
+@pytest.mark.parametrize("identity_equal", [False, True])
+def test_null_pred_matches_is_null_for_sentinel_column(monkeypatch, identity_equal):
     @dataclass
     class Row:
         v: int = blosc2.field(blosc2.int64(null_value=-1))
 
     t = CTable(Row)
     t.extend({"v": [1, -1, 3]})
     col = t["v"]
+    monkeypatch.setattr(blosc2, "_disable_overloaded_equal", identity_equal)
     pred = np.asarray(col._nulls.null_pred().compute()[:])
     np.testing.assert_array_equal(pred[: t.nrows], col.is_null())
     valid = np.asarray(col._nulls.valid_pred().compute()[:])

tests/ctable/test_remote_ctable.py

@@ -672,6 +672,7 @@ def test_sparse_table_cache_budget(tmp_path, limit):
 
 @pytest.mark.parametrize("include_note", [False, True])
 @pytest.mark.parametrize("nrows", [0, 2, 20, 21])
+@needs_utf8
 def test_remote_example_total_timing(tmp_path, capsys, include_note, nrows):
     import runpy
     from pathlib import Path
@@ -701,6 +702,7 @@ def test_remote_example_total_timing(tmp_path, capsys, include_note, nrows):
     assert sample == str(local[start:stop]).strip()
 
 
+@needs_utf8
 def test_remote_example_cache_dir(tmp_path, capsys, monkeypatch):
     import runpy
     import sys
@@ -720,6 +722,7 @@ def test_remote_example_cache_dir(tmp_path, capsys, monkeypatch):
     assert "(0 requests," in output.split("Total network :")[1]
 
 
+@needs_utf8
 def test_remote_example_batch_columns(tmp_path, capsys):
     import runpy
     from pathlib import Path
@@ -748,6 +751,7 @@ def test_remote_example_batch_columns(tmp_path, capsys):
 
 @pytest.mark.parametrize(("format_flag", "suffix"), [("pytables", ".h5"), ("parquet", ".parquet")])
 @pytest.mark.parametrize("explicit", [False, True])
+@needs_utf8
 def test_remote_example_external_formats(tmp_path, capsys, monkeypatch, format_flag, suffix, explicit):
     import runpy
     import sys
@@ -804,6 +808,7 @@ def test_remote_example_external_formats(tmp_path, capsys, monkeypatch, format_f
 
 
 @pytest.mark.parametrize("explicit", [False, True])
+@needs_utf8
 def test_remote_example_blosc2_flag_and_inference(tmp_path, capsys, monkeypatch, explicit):
     import runpy
     import sys
@@ -822,6 +827,7 @@ def test_remote_example_blosc2_flag_and_inference(tmp_path, capsys, monkeypatch,
     assert "[Format: Blosc2 B2Z" in capsys.readouterr().out
 
 
+@needs_utf8
 def test_remote_example_unknown_extension(tmp_path, monkeypatch, capsys):
     import runpy
     import sys
@@ -836,6 +842,7 @@ def test_remote_example_unknown_extension(tmp_path, monkeypatch, capsys):
     assert "cannot infer format" in capsys.readouterr().err
 
 
+@needs_utf8
 def test_remote_example_url_extension_ignores_query_string(monkeypatch):
     import runpy
     import sys
@@ -882,6 +889,7 @@ def changed_info(self, path, **kwargs):
 @pytest.mark.parametrize("max_concurrency", [1, 8])
 @pytest.mark.parametrize("legacy", [False, True])
 @pytest.mark.usefixtures("b2z_range_reads")
+@needs_utf8
 def test_disk_cache_reuses_ctable_bootstrap(tmp_path, monkeypatch, max_concurrency, legacy):
     schema = dataclasses.make_dataclass("Sample", [("x", float), ("note", str, blosc2.field(blosc2.utf8()))])
     values = np.random.default_rng(42).random(20_000)
@@ -982,6 +990,7 @@ def unexpected_read(*args, **kwargs):
 @pytest.mark.parametrize(
     "policy", [blosc2.CachePolicy.NONE, blosc2.CachePolicy.MEMORY, blosc2.CachePolicy.DISK]
 )
+@needs_utf8
 def test_remote_ctable_refresh(tmp_path, policy):
     schema = dataclasses.make_dataclass("Sample", [("x", int), ("note", str, blosc2.field(blosc2.utf8()))])
     local = blosc2.CTable(schema, [(1, "old"), (2, "café")], create_summary_index=False)
@@ -1504,6 +1513,7 @@ def test_remote_store_returns_table_with_independent_lifetime(tmp_path):
 
 
 @pytest.mark.usefixtures("b2z_range_reads")
+@needs_utf8
 def test_remote_ctable_reference_save_roundtrip(tmp_path):
     @dataclasses.dataclass
     class TextRow:
@@ -1678,6 +1688,7 @@ class Mixed:
 @pytest.mark.parametrize("policy", list(blosc2.CachePolicy))
 @pytest.mark.parametrize("null_storage", ["mask", "sentinel"])
 @pytest.mark.parametrize("deleted", [False, True])
+@needs_utf8
 def test_remote_ctable_utf8(tmp_path, policy, null_storage, deleted):
     @dataclasses.dataclass
     class TextRow:
@@ -1806,6 +1817,7 @@ def test_remote_ctable_deleted_rows_and_disk_cache(tmp_path):
 
 
 @pytest.mark.parametrize("empty", [False, True])
+@needs_utf8
 def test_remote_utf8_nested_lifetime(tmp_path, empty):
     @dataclasses.dataclass
     class TextRow:
@@ -1843,6 +1855,7 @@ class TextRow:
 
 
 @pytest.mark.parametrize("bad_data", [None, np.arange(4, dtype="int64")])
+@needs_utf8
 def test_remote_utf8_invalid_companion(tmp_path, bad_data):
     @dataclasses.dataclass
     class TextRow:
@@ -1878,6 +1891,7 @@ class TextRow:
 @pytest.mark.parametrize("policy", list(blosc2.CachePolicy))
 @pytest.mark.parametrize("blocks", [False, True])
 @pytest.mark.usefixtures("b2z_range_reads")
+@needs_utf8
 def test_remote_utf8_bounded_transfer(tmp_path, monkeypatch, policy, blocks):
     from blosc2 import proxy_source
     from blosc2._utf8_array import UTF8Array
@@ -1970,6 +1984,7 @@ def test_remote_store_rejects_table_root(tmp_path):
 
 
 @pytest.mark.parametrize("policy", list(blosc2.CachePolicy))
+@needs_utf8
 def test_open_dispatches_local_and_remote_tables(tmp_path, policy):
     @dataclasses.dataclass
     class TextRow:
@@ -2062,6 +2077,7 @@ def counted(self, path, start=None, end=None, **kwargs):
 
 
 @pytest.mark.usefixtures("b2z_range_reads")
+@needs_utf8
 def test_parallel_metadata_benchmark(tmp_path, monkeypatch):
     import runpy
     import time
@@ -2144,6 +2160,7 @@ def fail(*args):
     "policy", [blosc2.CachePolicy.NONE, blosc2.CachePolicy.MEMORY, blosc2.CachePolicy.DISK]
 )
 @pytest.mark.parametrize("budget", [128, 1 << 20])
+@needs_utf8
 def test_parallel_rows_cache_policies(tmp_path, policy, budget):
     @dataclasses.dataclass
     class Mixed:
@@ -2304,6 +2321,7 @@ def write(self, *args):
             assert list(table[399990:]) == list(local[399990:])
 
 
+@needs_utf8
 def test_parallel_timestamp_and_projection(tmp_path):
     @dataclasses.dataclass
     class Timed:

tests/test_expression_graph_properties.py

@@ -48,12 +48,12 @@ def test_value_dependent_numpy_promotion_defers_scalar_metadata(monkeypatch):
 
 
 @pytest.mark.parametrize("name", ["sum", "prod", "argmin", "argmax", "cumulative_sum", "cumulative_prod"])
-def test_platform_accumulator_fallback_does_not_claim_native_integer_width(monkeypatch, name):
+def test_platform_accumulator_fallback_uses_numpy_integer_width(monkeypatch, name):
     module = importlib.import_module("blosc2.expression_graph")
     monkeypatch.setattr(module, "_BLOSC_PLATFORM_ACCUMULATOR_FALLBACK", True)
     source = blosc2.asarray(np.arange(1, 5, dtype="int8"))
     graph = module.parse_expression(f"x.{name}(axis=0)")
-    assert graph.infer_dtype({"x": source}) is None
+    assert graph.infer_dtype({"x": source}) == np.dtype(np.intp)
     numpy_graph = module.parse_expression("np.sum(x)")
     assert numpy_graph.infer_dtype({"x": np.arange(4, dtype="int8")}) == np.dtype(np.intp)