[jira] [Commented] (HIVE-22728) Limit the scope of uniqueness of constraint name to database

["Zoltan Haindrich (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HIVE-22728?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17033539#comment-17033539 ] 

Zoltan Haindrich commented on HIVE-22728:
-----------------------------------------

I think we should even consider moving the uniqueness to the table level; because as of now the users doesn't gain any advantage of the gloablly unique contraint name; because they may only drop a contraint on a specific table:
{code}
ALTER TABLE numericDataType DROP CONSTRAINT check1;
{code}
dropping {{check1}} on a different table results in an error; so I suppose constraints do "belong" to tables...

independently from the choosen path there is one more thing which might need to be addressed: when a contraint violation occurs; the contraint's name is mentioned in the error message - if it is allowed to use the same name in different databases; then the error message should make where the constraint belongs to....(I don't think this would happen frequently; so if it complicates things a lot - it could be left out)


https://github.com/apache/hive/blob/0d9deba3c15038df4c64ea9b8494d554eb8eea2f/ql/src/test/queries/clientpositive/check_constraint.q#L130

> Limit the scope of uniqueness of constraint name to database
> ------------------------------------------------------------
>
>                 Key: HIVE-22728
>                 URL: https://issues.apache.org/jira/browse/HIVE-22728
>             Project: Hive
>          Issue Type: Wish
>            Reporter: Jesus Camacho Rodriguez
>            Assignee: Miklos Gergely
>            Priority: Major
>
> Currently, constraint names are globally unique across all databases (assumption is that this may have done by design). Nevertheless, though behavior seems to be implementation specific, it would be interesting to limit the scope to uniqueness per database.
> Currently we do not store database information with the constraints. To change the scope to one db, we would need to store the DB_ID in the KEY_CONSTRAINTS table in metastore when we create a constraint and add the DB_ID to the PRIMARY KEY of that table. Some minor changes to the error messages would be needed too, since otherwise it would be difficult to identify the correct violation in queries that span across multiple databases. Additionally, the SQL scripts will need to be updated to populate the DB_ID when we upgrade to new version.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)