CanonicalSchema and SchemaCRC64Avro fail to inline sub-schemas referenced by name

[joshng]

If the type of a record field, array, or map-value refers to a previously-declared type by name (eg, "type": "com.example.MyEmbeddedType"), then the structure used to compute the schemaCanonical string (and thus the schema fingerprint) should embed the fully-specified sub-schema, rather than just the type name.

The current implementation fails to dereference the named schema, resulting in incorrect canonicalization and fingerprinting.

I can follow up later with a concrete example, and I'm working on a fix which I will hopefully submit for review soon.

[przemekd]

On top of that, logical types are not properly processed. I've added my test and got:

--- FAIL: TestCanonicalSchema (0.00s)
    canonical_test.go:267: Test failed for schema: {"type":"record","name":"test","namespace":"com.linkedin.avro","fields":[{"name":"astring","type":"string"},{"name":"acomplextype","type":{"type":"long","logicalType":"timestamp-millis"}},{"name":"arecord",  "type":{"type":"record","name":"innerRecord","fields":[{"name":"value","type":"long"}]}}]}
                got canonical:          {"name":"com.linkedin.avro.test","type":"record","fields":[{"name":"astring","type":"string"},{"name":"acomplextype","type":{"type":"long"}},{"name":"arecord","type":{"name":"com.linkedin.avro.innerRecord","type":"record","fields":[{"name":"value","type":"long"}]}}]}
                expected canonical:     {"name":"com.linkedin.avro.test","type":"record","fields":[{"name":"astring","type":"string"},{"name":"acomplextype","type":"long"},{"name":"arecord","type":{"name":"com.linkedin.avro.innerRecord","type":"record","fields":[{"name":"value","type":"long"}]}}]}
FAIL
exit status 1
FAIL    github.com/linkedin/goavro/v2   0.049s

[przemekd]

I have created separate issue for logical types bug -> #174