Support for sub schemas and unknown logical types

I am debating adding support for “schema” types and “unknown” logical types. I’m thinking of using fastavro for some avro data I have. It’s change data from a database, so there are two scenarios here that fastavro doesn’t like. The example here:

{
    "name":"special_nullable_string",
    "type": ["null", {
        "type":"string",
        "logicalType":"varchar",
        "metadata":30 } ]
}```

The first is having a logicaltype of "varchar" for a string base type.  The second is having a subschema as a type, which is perfectly valid avro (this is how you get an avro schema that is not flat).

Would this be a welcome addition, and if so, I believe the reader is easy (I've already written two tests I'll add in a comment here) but the writer is escaping my immediate thoughts.  Any pointers on how to best approach that would be appreciated.

[rhaarm]

@todpunk Looks like the fix will take a bit of time, validate function needs to be updated or how the write_union function, since it’s using validate. I’ve already had an idea around this for a month now because I wanted to make changes to how validation works anyways, but this may take a few days.

[rhaarm]

@todpunk your most recent schema is broken, I just pasted that code in Schema Registry and it got rejected, looks like you are missing a closing bracket ‘}’ between “name” and “address” field.

@scottbelden I’ll see if I can get the writer and reader updated sooner with the extra logic around logical_type. If anything the logicalType should default to the standard type if the logicalType doesn’t exist in the “Known” logicalTypes. This will also help with #173 which should just be a function “registers” to the LOGICAL_READERS and LOGICAL_WRITERS, if the writer or reader functions are left out, then it should refer to the base type.