[
https://issues.apache.org/jira/browse/KYLIN-5309?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
pengfei.zhan updated KYLIN-5309:
--------------------------------
Description:
This issue will propose the behavior of the excluded table, it makes the
feature more flexible. Previously, excluded columns were subject to their
corresponding table, which meant we either excluded all columns or none of
them. However, we can rule out some of them of a table without participating in
pre-computation from now on.
1. This issue separates the non-pre-computation of lookup tables from this
feature. Usually, non-pre-computation is used for the scenario of many-to-many
to avoid data blowing up dramatically when building the flat table.
2. If the statement query the excluded column and index with foreign key is
preferred, snapshot join index to get the newest data.
3. Reloading table keeps the unaffected columns. Two important cases are as
follows: 1) Add a column to the excluded table, and the added column will be
treated as an excluded column automatically; 2) If the last un-excluded column
is removed, this table will be treated as an excluded table.
5. By design limitation, the excluded column should not participate in any
parameters of measure.
6. By design limitation, the excluded column in the snow model should not be
used as the foreign key.
> Propose more flexible runtime join scenarios for Kylin
> ------------------------------------------------------
>
> Key: KYLIN-5309
> URL: https://issues.apache.org/jira/browse/KYLIN-5309
> Project: Kylin
> Issue Type: New Feature
> Components: Metadata, Query Engine
> Affects Versions: 5.0-alpha
> Reporter: pengfei.zhan
> Assignee: pengfei.zhan
> Priority: Major
> Fix For: 5.0-alpha
>
>
> This issue will propose the behavior of the excluded table, it makes the
> feature more flexible. Previously, excluded columns were subject to their
> corresponding table, which meant we either excluded all columns or none of
> them. However, we can rule out some of them of a table without participating
> in pre-computation from now on.
> 1. This issue separates the non-pre-computation of lookup tables from this
> feature. Usually, non-pre-computation is used for the scenario of
> many-to-many to avoid data blowing up dramatically when building the flat
> table.
> 2. If the statement query the excluded column and index with foreign key is
> preferred, snapshot join index to get the newest data.
> 3. Reloading table keeps the unaffected columns. Two important cases are as
> follows: 1) Add a column to the excluded table, and the added column will be
> treated as an excluded column automatically; 2) If the last un-excluded
> column is removed, this table will be treated as an excluded table.
> 5. By design limitation, the excluded column should not participate in any
> parameters of measure.
> 6. By design limitation, the excluded column in the snow model should not be
> used as the foreign key.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
