[
https://issues.apache.org/jira/browse/AVRO-3118?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17344624#comment-17344624
]
Scott commented on AVRO-3118:
-----------------------------
I'm trying to implement this in fastavro and even if the `namespace == ""` is
fixed, I don't think this example should work based on the specification.
The specification states: "References to previously defined names are as in the
latter two cases above: if they contain a dot they are a fullname, if they do
not contain a dot, the namespace is the namespace of the enclosing definition."
To me this means that for "field_b" where the type starts as "row" should
eventually get parsed as "my_ns.row". If the "field_a" record has an explicit
null or empty string for the namespace, then that record would be named "row"
and the field_b record would be named "my_ns.row" and so they wouldn't match.
Does this type of schema currently work in Java because it seems to me like it
should result in a parse error. Or am I misreading the specificaiton?
> Namespace with empty string is not treated as null in Python API
> ----------------------------------------------------------------
>
> Key: AVRO-3118
> URL: https://issues.apache.org/jira/browse/AVRO-3118
> Project: Apache Avro
> Issue Type: Bug
> Components: python
> Affects Versions: 1.10.2
> Reporter: Frank Mieves
> Assignee: Michael A. Smith
> Priority: Blocker
>
> Hi,
> based on [the AVRO
> documentation|https://avro.apache.org/docs/current/spec.html#names] an empty
> string may also be used as a namespace to indicate the null namespace.
> The Python package doesn't do that. Following example will fail when parsing:
> {code:python}
> import avro.schema
> schema_str="""
> {"type": "record", "name": "my_schema", "namespace": "my_ns",
> "fields": [{"name": "field_a", "type": {"type": "record", "name": "row",
> "namespace": "",
> "fields": [{"name": "subfield", "type": "string"}]}},
> {"name": "field_b", "type": "row"}
> ]}
> """
> schema = avro.schema.parse(schema_str)
> {code}
> Raised exception is "SchemaParseException: Type property "row" not a valid
> Avro schema: Could not make an Avro Schema object from row."
> If I set the namespace in the in the subfield to null it's working.
> Problem for me is, that I can't change the schema definition. The schema is
> in the Kafka schema repository. The Kafka AVRO consumer receives this from
> the schema registry server with an empty string.
> I could fix this by adding a check in the parser source schema.py:
>
> {code:python}
> ...
> return writer.type == self.type and (self.type == 'request' or
> self.check_props(writer, ['fullname']))
> def __init__(self, name, namespace, fields, names=None, schema_type='record',
> doc=None, other_props=None):
> # Fixing empty namespace
> if namespace == '':
> namespace = None
>
> # Ensure valid ctor args
> if fields is None:
> ...
> {code}
>
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
