[jira] [Resolved] (CALCITE-1497) Invalid implications

["Julian Hyde (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/CALCITE-1497?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Julian Hyde resolved CALCITE-1497.
----------------------------------
       Resolution: Fixed
    Fix Version/s: 1.11.0

Fixed in http://git-wip-us.apache.org/repos/asf/calcite/commit/dccbb55a.

The issues were concerning deducing {{x IS NOT NULL}}; we could not even deduce {{x IS NOT NULL}} from {{x IS NOT NULL}}! Now we can deduce {{x IS NOT NULL}} from {{x > 5}}, among other things.

This change also makes improvements to {{RelMdPredicates}}; if you have a predicate {{x > y}} and project away {{y}}, the result of the project still has {{x IS NOT NULL}}. (Unless {{x}}'s type is declared {{NOT NULL}}, in which case the predicate would be superfluous.)

> Invalid implications
> --------------------
>
>                 Key: CALCITE-1497
>                 URL: https://issues.apache.org/jira/browse/CALCITE-1497
>             Project: Calcite
>          Issue Type: Bug
>            Reporter: Julian Hyde
>            Assignee: Julian Hyde
>             Fix For: 1.11.0
>
>
> RexImplicationChecker makes some mistakes when OR is on the left-hand size or AND is on the right-hand side. The following 3 tests should all succeed, but all currently give the wrong result. I intend to check in the tests with my fix for CALCITE-1483 (coming soon, currently in my branch https://github.com/julianhyde/calcite/tree/1483-not-in-not-null).
> {code}
>     // "s is not null or i = 5" does not imply "s is not null"
>     f.checkNotImplies(f.or(sIsNotNull, iEq5), sIsNotNull);
>     // "s is not null" implies "s is not null or i = 5"
>     f.checkImplies(sIsNotNull, f.or(sIsNotNull, iEq5));
>     // "s is not null" implies "i = 5 or s is not null"
>     f.checkImplies(sIsNotNull, f.or(iEq5, sIsNotNull));
> {code}
> [~amargoor], Can you comment on these?



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