[
https://issues.apache.org/jira/browse/HIVE-19433?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16469017#comment-16469017
]
Jesus Camacho Rodriguez commented on HIVE-19433:
------------------------------------------------
[~vgarg], could you verify whether changes in CALCITE-1960 fix the issue?
http://git-wip-us.apache.org/repos/asf/calcite/commit/f1a002ea
It seems to me that this solution will work by bailing out when you have a
single element (no permutation needed), but that solution is more general in
avoiding unnecessary permutations and also cover the single element case. If
that is the case, probably we should pick up that piece of code, as it will
also help to minimize diffs between Calcite and Hive codebase.
> HiveJoinPushTransitivePredicatesRule hangs
> ------------------------------------------
>
> Key: HIVE-19433
> URL: https://issues.apache.org/jira/browse/HIVE-19433
> Project: Hive
> Issue Type: Bug
> Components: Logical Optimizer
> Reporter: Vineet Garg
> Assignee: Vineet Garg
> Priority: Major
> Attachments: HIVE-19433.1.patch
>
>
> *Reproducer*
> {code:sql}
> CREATE TABLE `table1`(
> `idp_warehouse_id` bigint,
> `idp_audit_id` bigint,
> `idp_effective_date` date,
> `idp_end_date` date,
> `idp_delete_date` date,
> `pruid` varchar(32),
> `prid` bigint,
> `prtimesheetid` bigint,
> `prassignmentid` bigint,
> `prchargecodeid` bigint,
> `prtypecodeid` bigint,
> `prsequence` bigint,
> `prmodby` varchar(96),
> `prmodtime` timestamp,
> `prrmexported` bigint,
> `prrmckdel` bigint,
> `slice_status` int,
> `role_id` bigint,
> `user_lov1` varchar(30),
> `user_lov2` varchar(30),
> `incident_id` bigint,
> `incident_investment_id` bigint,
> `odf_ss_actuals` bigint,
> `practsum` decimal(38,20));
> CREATE TABLE `table2`(
> `idp_warehouse_id` bigint,
> `idp_audit_id` bigint,
> `idp_effective_date` date,
> `idp_end_date` date,
> `idp_delete_date` date,
> `pruid` varchar(32),
> `prid` bigint,
> `prtimesheetid` bigint,
> `prassignmentid` bigint,
> `prchargecodeid` bigint,
> `prtypecodeid` bigint,
> `prsequence` bigint,
> `prmodby` varchar(96),
> `prmodtime` timestamp,
> `prrmexported` bigint,
> `prrmckdel` bigint,
> `slice_status` int,
> `role_id` bigint,
> `user_lov1` varchar(30),
> `user_lov2` varchar(30),
> `incident_id` bigint,
> `incident_investment_id` bigint,
> `odf_ss_actuals` bigint,
> `practsum` decimal(38,20));
> explain SELECT s.idp_warehouse_id AS source_warehouse_id
> FROM table1 s
> JOIN
> table2 d
> ON (
> s.prid = d.prid )
> JOIN
> table2 e
> ON
> s.prid = e.prid
> WHERE
> concat(
> CASE
> WHEN s.prid IS NULL THEN 1
> ELSE s.prid
> END,',',
> CASE
> WHEN s.prtimesheetid IS NULL THEN 1
> ELSE s.prtimesheetid
> END,',',
> CASE
> WHEN s.prassignmentid IS NULL THEN 1
> ELSE s.prassignmentid
> END,',',
> CASE
> WHEN s.prchargecodeid IS NULL THEN 1
> ELSE s.prchargecodeid
> END,',',
> CASE
> WHEN (s.prtypecodeid) IS NULL THEN ''
> ELSE s.prtypecodeid
> END,',',
> CASE
> WHEN s.practsum IS NULL THEN 1
> ELSE s.practsum
> END,',',
> CASE
> WHEN s.prsequence IS NULL THEN 1
> ELSE s.prsequence
> END,',',
> CASE
> WHEN length(s.prmodby) IS NULL THEN ''
> ELSE s.prmodby
> END,',',
> CASE
> WHEN s.prmodtime IS NULL THEN
> cast(from_unixtime(unix_timestamp('2017-12-08','yyyy-MM-dd') ) AS timestamp)
> ELSE s.prmodtime
> END,',',
> CASE
> WHEN s.prrmexported IS NULL THEN 1
> ELSE s.prrmexported
> END,',',
> CASE
> WHEN s.prrmckdel IS NULL THEN 1
> ELSE s.prrmckdel
> END,',',
> CASE
> WHEN s.slice_status IS NULL THEN 1
> ELSE s.slice_status
> END,',',
> CASE
> WHEN s.role_id IS NULL THEN 1
> ELSE s.role_id
> END,',',
> CASE
> WHEN length(s.user_lov1) IS NULL THEN ''
> ELSE s.user_lov1
> END,',',
> CASE
> WHEN length(s.user_lov2) IS NULL THEN ''
> ELSE s.user_lov2
> END,',',
> CASE
> WHEN s.incident_id IS NULL THEN 1
> ELSE s.incident_id
> END,',',
> CASE
> WHEN s.incident_investment_id IS NULL
> THEN 1
> ELSE s.incident_investment_id
> END,',',
> CASE
> WHEN s.odf_ss_actuals IS NULL THEN 1
> ELSE s.odf_ss_actuals
> END ) != concat(
> CASE
> WHEN length(d.pruid) IS NULL THEN ''
> ELSE d.pruid
> END,',',
> CASE
> WHEN d.prid IS NULL THEN 1
> ELSE d.prid
> END,',',
> CASE
> WHEN d.prtimesheetid IS NULL THEN 1
> ELSE d.prtimesheetid
> END,',',
> CASE
> WHEN d.prassignmentid IS NULL THEN 1
> ELSE d.prassignmentid
> END,',',
> CASE
> WHEN d.prchargecodeid IS NULL THEN 1
> ELSE d.prchargecodeid
> END,',',
> CASE
> WHEN (d.prtypecodeid) IS NULL THEN ''
> ELSE d.prtypecodeid
> END,',',
> CASE
> WHEN d.practsum IS NULL THEN 1
> ELSE d.practsum
> END,',',
> CASE
> WHEN d.prsequence IS NULL THEN 1
> ELSE d.prsequence
> END,',',
> CASE
> WHEN length(d.prmodby) IS NULL THEN ''
> ELSE d.prmodby
> END,',',
> CASE
> WHEN d.prmodtime IS NULL THEN
> cast(from_unixtime(unix_timestamp('2017-12-08','yyyy-MM-dd') ) AS timestamp)
> ELSE d.prmodtime
> END,',',
> CASE
> WHEN d.prrmexported IS NULL THEN 1
> ELSE d.prrmexported
> END,',',
> CASE
> WHEN d.prrmckdel IS NULL THEN 1
> ELSE d.prrmckdel
> END,',',
> CASE
> WHEN d.slice_status IS NULL THEN 1
> ELSE d.slice_status
> END,',',
> CASE
> WHEN d.role_id IS NULL THEN 1
> ELSE d.role_id
> END,',',
> CASE
> WHEN length(d.user_lov1) IS NULL THEN ''
> ELSE d.user_lov1
> END,',',
> CASE
> WHEN length(d.user_lov2) IS NULL THEN ''
> ELSE d.user_lov2
> END,',',
> CASE
> WHEN d.incident_id IS NULL THEN 1
> ELSE d.incident_id
> END,',',
> CASE
> WHEN d.incident_investment_id IS NULL
> THEN 1
> ELSE d.incident_investment_id
> END,',',
> CASE
> WHEN d.odf_ss_actuals IS NULL THEN 1
> ELSE d.odf_ss_actuals
> END );
> {code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
