[
https://issues.apache.org/jira/browse/AVRO-30?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12709035#action_12709035
]
Doug Cutting commented on AVRO-30:
----------------------------------
Currently we define classes and protocols with two separately specified parts,
name and namespace, but we refer to them only with one, a name.
We should expand the reference to include both parts. So one might define a
field as
{code}
{"name": "foo", "type": "com.acme.Foo"}
{code}
Questions:
# Should we also permit definitions with both parts? E.g.
{"type": "record", "name":"com.acme.Foo"} instead of {"type": "record", "name":
"Foo", "namespace":"com.acme"}
# If so, should we eliminate the use of "namespace" altogether?
Regardless, unqualified names would always be in the namespace (if any) defined
by their containing definition. So usually, in a protocol, one need only
specify the namespace in the top-level definition. Note that this is
include-friendly, since an included schema file should specify it's namespace
in its top-level definition.
> name lookup should consider namespace
> -------------------------------------
>
> Key: AVRO-30
> URL: https://issues.apache.org/jira/browse/AVRO-30
> Project: Avro
> Issue Type: Bug
> Components: java
> Reporter: Doug Cutting
> Fix For: 1.0
>
>
> When a record schema is referred to by name, Avro will currently return any
> known record definition with that name, regardless of its namespace.
> Instead, unqualified references should refer to names in the current
> namespace, and namespace-qualfied names should look for a record in the given
> namespace.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
