[
https://issues.apache.org/jira/browse/AVRO-1777?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16717714#comment-16717714
]
ASF subversion and git services commented on AVRO-1777:
-------------------------------------------------------
Commit beef86697a0dbd09d4d99a735f8a5afe37e5d976 in avro's branch
refs/heads/master from [~shiraeeshi]
[ https://gitbox.apache.org/repos/asf?p=avro.git;h=beef866 ]
AVRO-1777: Select best matching record when writing a union in python (#95)
* make numeric schemas not valid for boolean datums
* fix formatting
* fix int and long schemas
* fix brackets
* fix brackets
* add GenericRecord type
> Select best matching record when writing a union in python
> ----------------------------------------------------------
>
> Key: AVRO-1777
> URL: https://issues.apache.org/jira/browse/AVRO-1777
> Project: Apache Avro
> Issue Type: Improvement
> Components: python
> Affects Versions: 1.7.7
> Reporter: Steven Aerts
> Priority: Major
> Fix For: 1.9.0
>
>
> Unlike javascript, python is not using wrapped types.
> So when writing a union it needs to guess find out which type it will output.
> At the moment it takes the last validating type.
> I propose to take the type with the most matching fields.
> So I propose to change in {{io.py}}:
> {code}
> # resolve union
> index_of_schema = -1
> for i, candidate_schema in enumerate(writers_schema.schemas):
> if validate(candidate_schema, datum):
> index_of_schema = i
> if index_of_schema < 0: raise AvroTypeException(writers_schema, datum)
> {code}
> into
> {code}
> # resolve union
> index_of_schema = -1
> found_fields = -1
> for i, candidate_schema in enumerate(writers_schema.schemas):
> if validate(candidate_schema, datum):
> nr_fields = candidate_schema.type in ['record', 'error', 'request'] and
> len(candidate_schema.fields) or 1
> if nr_fields > found_fields:
> index_of_schema = i
> found_fields = nr_fields
> if index_of_schema < 0: raise AvroTypeException(writers_schema, datum)
> {code}
> If you want, I can create a pull request for this. And apply it both on py3
> as py.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
