You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@hive.apache.org by "Mark Grover (JIRA)" <ji...@apache.org> on 2013/01/11 00:48:12 UTC

[jira] [Resolved] (HIVE-3859) Expand UDF method selection logic in Function Registry

     [ https://issues.apache.org/jira/browse/HIVE-3859?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Mark Grover resolved HIVE-3859.
-------------------------------

    Resolution: Invalid

It's non-trivial to separate out the UDF method selection part of the patch (see Gunther's comment regarding it for more details). The code for this patch already exists in the parent JIRA's patch (HIVE-2693) so this JIRA no longer applies.
                
> Expand UDF method selection logic in Function Registry
> ------------------------------------------------------
>
>                 Key: HIVE-3859
>                 URL: https://issues.apache.org/jira/browse/HIVE-3859
>             Project: Hive
>          Issue Type: Sub-task
>          Components: UDF
>    Affects Versions: 0.9.0
>            Reporter: Mark Grover
>
> Presently, the Function Registry uses a constant cost of conversion to decide which method to call. This has worked well until now but wouldn't work when Decimal types get introduced (HIVE-2693). The reason being that a double and a decimal may be good candidates for conversion of an int but with the present costing strategy, they both will have the same cost of conversion. This leads to ambiguity when Hive tries to decide which method in the UDF to call if the UDF implements methods for both double and decimal.
> This needs to get resolved before the decimal code gets committed.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira