[MCOL-339] bug5848.sql fails indicating casual partition failure. Created: 2016-09-29 Updated: 2023-10-25 Resolved: 2023-10-25 |
|
| Status: | Closed |
| Project: | MariaDB ColumnStore |
| Component/s: | PrimProc |
| Affects Version/s: | 1.0.3 |
| Fix Version/s: | Icebox |
| Type: | Bug | Priority: | Minor |
| Reporter: | David Hall (Inactive) | Assignee: | Leonid Fedorov |
| Resolution: | Won't Fix | Votes: | 0 |
| Labels: | None | ||
| Environment: |
CentOS 6.5 |
||
| Sprint: | 2016-19 |
| Description |
|
While running the tests in working_tpch1_compare_log_only, I noticed a failure of partitionOptimization/bug5848.sql Looking at the results, it appears that partition elimination is not occurring. This could be a significant reason for performance degradation. |
| Comments |
| Comment by David Hall (Inactive) [ 2016-10-10 ] |
|
This issue only applies to udf functions. In sql_yac.yy there's a line: For UDF functions, the following callstack is made before the value is assigned and the char_set set: This creates a convert (for char_set conversion) object. When the engine sees the convert object, it assumes a compare can't be made without it, and bypasses casual partitioning. For other types of string compares and functions that aren't udf, the value is assigned before it makes the call as above, so the char_set is already correct and no convert is generated. This issue only affects UDF functions. idbpartition and other columnstore specific functions are implemented as UDF functions. |
| Comment by David Hall (Inactive) [ 2016-10-17 ] |
|
I've renamed this test to .fixme in the test suite so we can move on until we fix this. |