[
https://issues.apache.org/jira/browse/CALCITE-7818?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18118650#comment-18118650
]
Mihai Budiu commented on CALCITE-7818:
--------------------------------------
Why is this a Druid-specific problem?
> Druid interval extraction drops rows for mixed OR predicates
> ------------------------------------------------------------
>
> Key: CALCITE-7818
> URL: https://issues.apache.org/jira/browse/CALCITE-7818
> Project: Calcite
> Issue Type: Bug
> Reporter: Bruno Volpato
> Priority: Major
>
> A Druid filter can combine a timestamp range with a predicate that cannot be
> represented as an interval. For example:
> `"timestamp" < TIMESTAMP '2020-01-01 00:00:00' OR EXTRACT(DAY FROM
> "timestamp") = 15`
> A row dated `2021-01-15` satisfies the second branch. Interval extraction
> currently keeps only the first branch's range. The filter rule treats those
> partial intervals as the full condition and removes the filter, so the row is
> lost.
> Expected: retain the full OR predicate and return the matching row.
> Actual: restrict the scan to dates before 2020 and drop the filter.
> Interval extraction should stop when any OR branch cannot be represented as
> an interval.
> Fix and regression tests: https://github.com/apache/calcite/pull/5243
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
