[
https://issues.apache.org/jira/browse/IGNITE-26438?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Aleksandr Chesnokov updated IGNITE-26438:
-----------------------------------------
Description:
H2 automatically generates system aliases for derived tables in the form
{{_<number>}} (e.g., {{{}_10045{}}}).
These aliases have no business meaning but cause logically identical SQL plans
to be stored as different texts, increasing storage size.
This has two consequences:
# The *PerfStat* report becomes heavier than necessary.
# The *sql.plans.history* system view grows larger, storing redundant entries.
An example is given below.
Before:
{noformat}
SELECT _10045.id, _10045.name
FROM (
SELECT id, name
FROM users
) _10045
WHERE _10045.id = 1;{noformat}
After:
{noformat}
SELECT id, name
FROM (
SELECT id, name
FROM users
)
WHERE id = 1;{noformat}
was:
H2 automatically generates system aliases for derived tables in the form
{{_<number>}} (e.g., {{{}_10045{}}}).
These aliases have no business meaning but cause logically identical SQL plans
to be stored as different texts, increasing storage size.
This has two consequences:
# The *PerfStat* report becomes heavier than necessary.
# The *sql.plans.history* system view grows larger, storing redundant entries.
Example:
Before:
{noformat}
SELECT _10045.id, _10045.name
FROM (
SELECT id, name
FROM users
) _10045
WHERE _10045.id = 1;{noformat}
After:
{noformat}
SELECT id, name
FROM (
SELECT id, name
FROM users
)
WHERE id = 1;{noformat}
> Normalize SQL plan: remove H2 system-generated subquery aliases
> ---------------------------------------------------------------
>
> Key: IGNITE-26438
> URL: https://issues.apache.org/jira/browse/IGNITE-26438
> Project: Ignite
> Issue Type: Improvement
> Reporter: Aleksandr Chesnokov
> Priority: Major
> Labels: ise
>
> H2 automatically generates system aliases for derived tables in the form
> {{_<number>}} (e.g., {{{}_10045{}}}).
> These aliases have no business meaning but cause logically identical SQL
> plans to be stored as different texts, increasing storage size.
> This has two consequences:
> # The *PerfStat* report becomes heavier than necessary.
> # The *sql.plans.history* system view grows larger, storing redundant
> entries.
> An example is given below.
> Before:
> {noformat}
> SELECT _10045.id, _10045.name
> FROM (
> SELECT id, name
> FROM users
> ) _10045
> WHERE _10045.id = 1;{noformat}
> After:
> {noformat}
> SELECT id, name
> FROM (
> SELECT id, name
> FROM users
> )
> WHERE id = 1;{noformat}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
