[
https://issues.apache.org/jira/browse/AVRO-3833?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Oscar Westra van Holthe - Kind updated AVRO-3833:
-------------------------------------------------
Description:
The specification currently says nothing about the uniqueness of names. At most
it says something about equality of names, assuming readers understand that a
name must uniquely identify a type.
I suggest making the rules for uniqueness explicit.
This would mean:
# Full names and aliases of types (including namespace) must be unique
# For each enum type, symbol names must be unique
# For each record/error type, field names and aliases must be unique
Also, I propose to add that aliases should be, but to not need be, valid names.
Using an invalid name as alias allows schema evolution to correct illegal names
used in the past.
This allows us to explicitly describe how to fix invalid, but previously
accepted, schemata.
was:
The specification currently says nothing about the uniqueness of names. At most
it says something about equality of names, assuming readers understand that a
name must uniquely identify a type.
I suggest making the rules for uniqueness explicit.
This would mean:
# Full names and aliases of types (including namespace) must be unique
# For each enum type, symbol names must be unique
# For each record/error type, field names and aliases must be unique
Also, I propose to add that aliases should be, but to not need be, valid names.
Using an invalid name as alias allows schema evolution to correct illegal names
used in the past.
> Spec: clarify using names and aliases
> -------------------------------------
>
> Key: AVRO-3833
> URL: https://issues.apache.org/jira/browse/AVRO-3833
> Project: Apache Avro
> Issue Type: Improvement
> Components: spec
> Affects Versions: 1.11.2
> Reporter: Oscar Westra van Holthe - Kind
> Priority: Minor
> Labels: pull-request-available
> Time Spent: 10m
> Remaining Estimate: 0h
>
> The specification currently says nothing about the uniqueness of names. At
> most it says something about equality of names, assuming readers understand
> that a name must uniquely identify a type.
> I suggest making the rules for uniqueness explicit.
> This would mean:
> # Full names and aliases of types (including namespace) must be unique
> # For each enum type, symbol names must be unique
> # For each record/error type, field names and aliases must be unique
> Also, I propose to add that aliases should be, but to not need be, valid
> names. Using an invalid name as alias allows schema evolution to correct
> illegal names used in the past.
> This allows us to explicitly describe how to fix invalid, but previously
> accepted, schemata.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
