[jira] [Commented] (IGNITE-13667) Add schema columns relocation table to map from user order to system order

["Andrey Mashenkov (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/IGNITE-13667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17294564#comment-17294564 ] 

Andrey Mashenkov commented on IGNITE-13667:
-------------------------------------------

Why we may need to restore column order defined in table configuration by a user?
AFAIK, "Select * From table" doesn't guarantee any column order. 

And I can't remember any cases where column order made sense, 
column values can be retrieved from BinaryObjects\Rows via their names,
TableViews doesn't imply direct access to columns neither via name nor via an index.
Internally, we always use native order (defined in internal schema descriptor) only.

> Add schema columns relocation table to map from user order to system order
> --------------------------------------------------------------------------
>
>                 Key: IGNITE-13667
>                 URL: https://issues.apache.org/jira/browse/IGNITE-13667
>             Project: Ignite
>          Issue Type: Improvement
>            Reporter: Alexey Goncharuk
>            Priority: Major
>              Labels: iep-54, ignite-3
>
> When a schema is defined, the key chunk columns and value chunk columns are sorted so that fixlen columns go first and varlen columns go second, so the sorted column order differs from the order of the user-defined columns.
> We need to add a simple relocation table which is a permutation of indices {{[0..n)}}, so that an internal column order for user index {{n}} is {{relocationTbl[n]}}.
> NB: the tuple assembler will still need to access the internal sorted order for proper tuple assembly.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)