[
https://issues.apache.org/jira/browse/AVRO-2226?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16703746#comment-16703746
]
ASF subversion and git services commented on AVRO-2226:
-------------------------------------------------------
Commit 3e5cf4a94dd1f71e42a45c812f26a53c7b89e945 in avro's branch
refs/heads/master from [~keatskelleher]
[ https://gitbox.apache.org/repos/asf?p=avro.git;h=3e5cf4a ]
[AVRO-2226] Fixes UnionSchema specificity
Trouble arises in the python library when deducing the appropriate schema from
a list of schemas given a particular datum.
When "null" values are allowed for fields in two separate schemas, there is no
way to differentiate which should schema be used given of a list of schemas
that are set as the type definition for a record.
This PR checks to ensure all fields defined on a given datum are _also_ defined
in the schema being validated to use for that datum.
With this bugfix, datums such as `{"foo": "a"}` will not "cast" to schemas such
as `{"name": "bar", "type": ["long", "null"]}`, which is currently the case.
> UnionSchema deduction is too permissive
> ---------------------------------------
>
> Key: AVRO-2226
> URL: https://issues.apache.org/jira/browse/AVRO-2226
> Project: Apache Avro
> Issue Type: Bug
> Components: python
> Affects Versions: 1.8.2
> Reporter: Andrew Kelleher
> Priority: Major
> Attachments: AVRO-2226.patch
>
> Original Estimate: 24h
> Remaining Estimate: 24h
>
> When given a schema of the form
> {code:java}
> {
> "type" : "record",
> "name" : "A",
> "namespace" : "com.example",
> "fields" : [
> {
> "name" : "foo",
> "type" : ["string", "null"]
> }
> ]
> }
> {
> "type" : "record",
> "name" : "B",
> "namespace" : "com.example",
> "fields" : [
> {
> "name" : "bar",
> "type" : ["string", "null"]
> }
> ]
> }
> {
> "type" : "record",
> "name" : "AOrB",
> "namespace" : "com.example",
> "fields" : [
> {
> "name" : "entity",
> "type" : [
> "com.example.A",
> "com.example.B"
> ]
> }
> ]
> }
> {code}
> And a datum of the form
> {code}
> {'entity': {'foo': 'this is an instance of schema A'}}{code}
> Converting to a message, and then from a message chooses the incorrect
> `entity` schema:
> {code}
> {'entity': {'bar': None}}{code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
