[MCOL-1864] Very long IN clause causing wrong results Created: 2018-11-08 Updated: 2022-11-05 Resolved: 2022-11-05 |
|
| Status: | Closed |
| Project: | MariaDB ColumnStore |
| Component/s: | N/A |
| Affects Version/s: | 1.1.6 |
| Fix Version/s: | Icebox |
| Type: | Bug | Priority: | Major |
| Reporter: | Dev OneOone | Assignee: | Unassigned |
| Resolution: | Won't Do | Votes: | 0 |
| Labels: | Compatibility | ||
| Environment: |
Ubuntu 18.04.1 LTS |
||
| Description |
|
A very long IN clause is causing wrong results. My testcase works fine with INNODB engine though. Steps to reproduce (attachment would be too long):
Resultset counts 2015667 records.
and send them explicitly in the query.
Resultset counts 49587 records.
I'm suspecting some undocumented limit or a bug. |
| Comments |
| Comment by Dev OneOone [ 2018-11-12 ] |
|
In addition: the columnstore query takes about 5 times as long as querying the innodb table. |
| Comment by Roman [ 2018-12-08 ] |
|
Many thanks for the clue. We will look into it. |
| Comment by Patrick LeBlanc (Inactive) [ 2018-12-10 ] |
|
IIRC there is an implicit limit on the in-clause length leading up to execution. Either in the component that sends the maria execution plan to exemgr, or in the conversion to a CS execution plan. |
| Comment by patrice [ 2018-12-12 ] |
|
there is an issue for this : https://jira.mariadb.org/browse/MCOL-557 . we ran into this, and it is not recommended to have more than 50k in the in clause. |
| Comment by Dev OneOone [ 2018-12-17 ] |
|
@all thank you for taking notice Does anyone know if this is a hard limit or is it somehow configurable? Where's the exact threshold? |
| Comment by Nivesh [ 2019-01-24 ] |
|
Hi I have a workaround for this issue however is it very dependant on the amount of memory is available on the UM. Running a trim on the in list fixes this issue. |
| Comment by Todd Stoffel (Inactive) [ 2022-11-05 ] |
|
Item is out of date. Closing due to inactivity. If you feel this was done in error please open a new ticket. |