[jira] [Updated] (CALCITE-824) Fix type inference when converting large IN clauses to semijoins

["Josh Wills (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/CALCITE-824?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Josh Wills updated CALCITE-824:
-------------------------------
    Attachment: CALCITE-824.patch

Here's the patch I wrote for this and the test I added to CsvTest that will fail w/o the fix from the patch. I kinda doubt that my implementation is the best way to fix this bug, so improvements are welcome.

> Fix type inference when converting large IN clauses to semijoins
> ----------------------------------------------------------------
>
>                 Key: CALCITE-824
>                 URL: https://issues.apache.org/jira/browse/CALCITE-824
>             Project: Calcite
>          Issue Type: Bug
>          Components: core
>    Affects Versions: 1.1.0-incubating, 1.2.0-incubating, 1.3.0-incubating, 1.4.0-incubating
>            Reporter: Josh Wills
>            Assignee: Julian Hyde
>         Attachments: CALCITE-824.patch
>
>
> I hit an issue where using an IN clause of int literals would work properly when compared against a BIGINT column when the IN clause had less than 20 values, but would fail when processing more than 20 values. I traced the issue to the way target row types are handled inside of the code that converts the IN clause into a semijoin.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)