[
https://issues.apache.org/jira/browse/IGNITE-19964?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Pavel Pereslegin updated IGNITE-19964:
--------------------------------------
Description:
Currently boolean index is used by an expression like {{WHERE ... field IS
<TRUE | FALSE, NULL>}}, but is not used by expression like {{WHERE ... field IS
NOT <TRUE | FALSE | NULL>}}. This is based on the assumption that without
knowing the data distribution, on average, performing multiple index scans
will be slower than a single table scan (for example {{field = TRUE or field =
NULL}}).
We need to investigate the possibility of improving this behavior when the
planner will considers distribution of data.
You may also need to rewrite the corresponding expression (see
{{RexUtils#expandBooleanFieldComparison}}).
was:
Currently boolean index is used by an expression like {{WHERE ... field IS
<TRUE | FALSE, NULL>}}, but is not used by expression like {{WHERE ... field IS
NOT <TRUE | FALSE | NULL>}}.
This is based on the assumption that without knowing the data distribution, on
average, performing multiple index scans will be slower than a single table
scan (for example {{field = TRUE or field = NULL}}).
We need to investigate the possibility of improving this behavior when the
planner will considers distribution of data.
You may also need to rewrite the corresponding expression (see
{{RexUtils#expandBooleanFieldComparison}}).
> Sql. Improve index scan on boolean field.
> -----------------------------------------
>
> Key: IGNITE-19964
> URL: https://issues.apache.org/jira/browse/IGNITE-19964
> Project: Ignite
> Issue Type: Improvement
> Components: sql
> Reporter: Pavel Pereslegin
> Priority: Minor
> Labels: ignite-3
>
> Currently boolean index is used by an expression like {{WHERE ... field IS
> <TRUE | FALSE, NULL>}}, but is not used by expression like {{WHERE ... field
> IS NOT <TRUE | FALSE | NULL>}}. This is based on the assumption that without
> knowing the data distribution, on average, performing multiple index scans
> will be slower than a single table scan (for example {{field = TRUE or field
> = NULL}}).
> We need to investigate the possibility of improving this behavior when the
> planner will considers distribution of data.
> You may also need to rewrite the corresponding expression (see
> {{RexUtils#expandBooleanFieldComparison}}).
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
