reraise the KeyError for unknown codecs with a more specific error class (#3395)

* create a error class for unknown codecs

* reraise as a unknown codec error

* check that `parse_codecs` raises (v3 only)

* monkeypatch to always work with a empty registry

* immediatly format the error

* changelog

Author: keewis

changes/3395.bugfix.rst

@@ -0,0 +1 @@
+Raise a Zarr-specific error class when a codec can't be found by name when deserializing the given codecs. This avoids hiding this error behind a "not part of a zarr hierarchy" warning.

src/zarr/core/metadata/v3.py

@@ -34,7 +34,7 @@
 )
 from zarr.core.config import config
 from zarr.core.metadata.common import parse_attributes
-from zarr.errors import MetadataValidationError, NodeTypeValidationError
+from zarr.errors import MetadataValidationError, NodeTypeValidationError, UnknownCodecError
 from zarr.registry import get_codec_class
 
 
@@ -63,7 +63,11 @@ def parse_codecs(data: object) -> tuple[Codec, ...]:
             out += (c,)
         else:
             name_parsed, _ = parse_named_configuration(c, require_configuration=False)
-            out += (get_codec_class(name_parsed).from_dict(c),)
+
+            try:
+                out += (get_codec_class(name_parsed).from_dict(c),)
+            except KeyError as e:
+                raise UnknownCodecError(f"Unknown codec: {e.args[0]!r}") from e
 
     return out
 

src/zarr/errors.py

@@ -101,6 +101,14 @@ class MetadataValidationError(BaseZarrError):
     _msg = "Invalid value for '{}'. Expected '{}'. Got '{}'."
 
 
+class UnknownCodecError(BaseZarrError):
+    """
+    Raised when a unknown codec was used.
+    """
+
+    _msg = "{}"
+
+
 class NodeTypeValidationError(MetadataValidationError):
     """
     Specialized exception when the node_type of the metadata document is incorrect.

tests/test_metadata/test_v3.py

@@ -17,10 +17,11 @@
 from zarr.core.group import GroupMetadata, parse_node_type
 from zarr.core.metadata.v3 import (
     ArrayV3Metadata,
+    parse_codecs,
     parse_dimension_names,
     parse_zarr_format,
 )
-from zarr.errors import MetadataValidationError, NodeTypeValidationError
+from zarr.errors import MetadataValidationError, NodeTypeValidationError, UnknownCodecError
 
 if TYPE_CHECKING:
     from collections.abc import Sequence
@@ -323,3 +324,17 @@ async def test_special_float_fill_values(fill_value: str) -> None:
     elif fill_value == "-Infinity":
         assert np.isneginf(m.fill_value)
         assert d["fill_value"] == "-Infinity"
+
+
+def test_parse_codecs_unknown_codec_raises(monkeypatch: pytest.MonkeyPatch) -> None:
+    from collections import defaultdict
+
+    import zarr.registry
+    from zarr.registry import Registry
+
+    # to make sure the codec is always unknown (not sure if that's necessary)
+    monkeypatch.setattr(zarr.registry, "__codec_registries", defaultdict(Registry))
+
+    codecs = [{"name": "unknown"}]
+    with pytest.raises(UnknownCodecError):
+        parse_codecs(codecs)