[
https://issues.apache.org/jira/browse/IMPALA-12742?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17811607#comment-17811607
]
ASF subversion and git services commented on IMPALA-12742:
----------------------------------------------------------
Commit 595212b4ea4f0cabd2504090626f5753f8feb865 in impala's branch
refs/heads/master from Zoltan Borok-Nagy
[ https://gitbox.apache.org/repos/asf?p=impala.git;h=595212b4e ]
IMPALA-12742: DELETE/UPDATE Iceberg table partitioned by DATE fails with error
Iceberg tables can be identity partitioned by any type, e.g. int, date
and even float. If a table is partitioned, the file path contains the
partition value in human readable form, and this form is expected to
be passed to CatalogD. When an UPDATE or DELETE command is executed,
we don't transform the integer date value to human readable format,
which causes errors in CatalogD.
With this patch, we transform identity-partitioned date values to
human-readable format.
Note on floating point numbers:
When users partition their data via floating point values (users should
not do that), then the file paths created for delete files might not
correspond to the data files (e.g. '1.1' vs '1.100000023841858'). Though
the values are the same in the Iceberg metadata layer, so it doesn't
cause correctness issues.
Testing:
* added e2e tests for DELETEs
* added e2e tests for UPDATEs
Change-Id: I506f95527e741efe18c71706e2cdea51b45958b8
Reviewed-on: http://gerrit.cloudera.org:8080/20954
Reviewed-by: Impala Public Jenkins <impala-public-jenk...@cloudera.com>
Tested-by: Impala Public Jenkins <impala-public-jenk...@cloudera.com>
> DELETE/UPDATE Iceberg table partitioned by DATE fails with error
> ----------------------------------------------------------------
>
> Key: IMPALA-12742
> URL: https://issues.apache.org/jira/browse/IMPALA-12742
> Project: IMPALA
> Issue Type: Bug
> Components: Backend, Catalog
> Reporter: Noemi Pap-Takacs
> Assignee: Zoltán Borók-Nagy
> Priority: Major
> Labels: impala-iceberg
>
> Iceberg tables can be identity partitioned by any type, e.g. int, date and
> even float.
> If a table is partitioned, the file path contains the partition value in
> human readable form. When an UPDATE or DELETE command is executed, the delete
> file contains the file path to the referenced data file. It seems that DATE
> type is converted to this form incorrectly, and cannot be parsed by the
> Catalog and throws an error.
>
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-all-unsubscr...@impala.apache.org
For additional commands, e-mail: issues-all-h...@impala.apache.org
