[Avro] Accept dict with only 'type': 'null' as representation of null

[Tishj]

This PR makes it so that a different serialization of the null type is accepted, this variation is produced by avro-c (apache/avro's c lang implementation)

[kevinjqliu]

this makes sense, but i cant find a corresponding mention in the avro spec

cc @Fokko to chime in here

[kevinjqliu]

should we add a test for this in

iceberg-python/tests/utils/test_schema_conversion.py

Lines 235 to 239 in 483d947

	def test_resolve_union() -> None:
	with pytest.raises(TypeError) as exc_info:
	AvroSchemaConversion()._resolve_union(["null", "string", "long"])
	assert "Non-optional types aren't part of the Iceberg specification" in str(exc_info.value)

[Tishj]

The spec seems to indicate that this is valid:
https://avro.apache.org/docs/1.11.1/specification/

Right under Schema Declaration

{"type": "typeName" ...attributes...}

where typeName is either a primitive or derived type name, as defined below

List of primitive types

Primitive Types
The set of primitive type names is:

null: no value
...

And then below the list:

Primitive type names are also defined type names. Thus, for example, the schema “string” is equivalent to:
{"type": "string"}

Which is why "null" is shorthand for {"type": "null"}

[kevinjqliu]

ty that makes sense, thanks for the pointer

[kevinjqliu]

Looks like theres a linter issue, could you run make lint locally?

[Fokko]

While the syntax looks a bit weird, I agree that this is valid:

I agree with @kevinjqliu that a test would be a very good idea, since we're moving this part to rust somewhere in the not-so-distant future :)