[ 
https://issues.apache.org/jira/browse/IMPALA-15247?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18114112#comment-18114112
 ] 

ASF subversion and git services commented on IMPALA-15247:
----------------------------------------------------------

Commit 904ad4e3c1c0f41f322957673f97d903a9a36d50 in impala's branch 
refs/heads/master from Csaba Ringhofer
[ https://gitbox.apache.org/repos/asf?p=impala.git;h=904ad4e3c ]

IMPALA-15247: Unify not-comparable type checks in Type.throwIfNotComparable

Several analysis checks independently rejected complex/VARIANT types with
their own ad-hoc error messages, e.g.  ORDER BY, GROUP BY ...

Consolidate them into Type.supportsComparison() and
Type.throwIfNotComparable(exprDesc), which emit a single uniform message:
  "<exprDesc> with type '<type>' is not supported."

This slightly changes some error messages. Existing complex/VARIANT test
messages are updated to the unified wording.

This is a geometry-free refactor extracted from the adding GEOMETRY type:
https://gerrit.cloudera.org/#/c/24534/

Change-Id: I2dc2c4579508cf1f8954ac251087c52f2eddf36a
Reviewed-on: http://gerrit.cloudera.org:8080/24661
Reviewed-by: Impala Public Jenkins <[email protected]>
Tested-by: Impala Public Jenkins <[email protected]>


> Unify missing type capability handling in analyzer
> --------------------------------------------------
>
>                 Key: IMPALA-15247
>                 URL: https://issues.apache.org/jira/browse/IMPALA-15247
>             Project: IMPALA
>          Issue Type: Improvement
>          Components: Frontend
>            Reporter: Csaba Ringhofer
>            Priority: Major
>
> Currently "roles" in queries are often rejected based on an ad-hoc list of 
> types.
> Example:
> https://github.com/apache/impala/blob/664f455ad1b573d53cac3f6e91329d25bbfca9a3/fe/src/main/java/org/apache/impala/analysis/AnalyticExpr.java#L434
> it says that a complex type or variant can't be partitioning expression in 
> analytic functions, and not the root cause, which is not equality and hashing 
> defined for these types. Relying on type list is cumbersome for new types, 
> for example GEOMETRY/GEOGRAPHY are in the same shoes.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to