[
https://issues.apache.org/jira/browse/IMPALA-5323?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16999373#comment-16999373
]
Cliff Resnick commented on IMPALA-5323:
---------------------------------------
We store HLL intermediates in Kudu, and UTF-8 encoding (imposed in the Java
API) essentially doubles the response time of our UDAF because of the added
bloat and decoding. Patching Kudu's Java API to pass ISO-8859-1 strings is
simple, but deployment issues can be a pain many organizations and probably
requires shading. However, if the Kudu team follows through with UTF-8 in c++
as [~adar] mentions, then the party is over for most folks.
Is it really a lot of work to add a BINARY type to Impala? I imagine the
implementation would nearly equivalent to current STRING. It seems to me this
is more of a semantic problem, whether a String is a c-string or charset
encoded bytes. It's clear that Impala embraces the former, and StringVal is
used everywhere accordingly. However, Kudu clearly embraces the latter and this
could become a bad situation if not resolved. Here hoping you find a way.
> Support Kudu BINARY
> -------------------
>
> Key: IMPALA-5323
> URL: https://issues.apache.org/jira/browse/IMPALA-5323
> Project: IMPALA
> Issue Type: New Feature
> Components: Backend
> Reporter: Pavel Martynov
> Priority: Major
> Labels: kudu
>
> I trying to 'CREATE EXTERNAL TABLE STORED AS KUDU' on the table with BINARY
> Kudu column data type and got an error: Kudu type 'binary' is not supported
> in Impala.
> This limitation is not documented, checked:
> https://impala.incubator.apache.org/docs/build/html/topics/impala_kudu.html
> https://kudu.apache.org/docs/kudu_impala_integration.html#_known_issues_and_limitations
> There are some thoughts that Kudu BINARY data type may be supported by
> Impala's STRING data type:
> https://community.cloudera.com/t5/Interactive-Short-cycle-SQL/Does-impala-support-binary-data-type/td-p/24366
> https://groups.google.com/a/cloudera.org/forum/#!msg/impala-user/muguKJU3c3I/_oArmoxSlDMJ
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
