[jira] [Commented] (IGNITE-16141) IndexQuery: consistency with query after IO filter

["Maksim Timonin (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/IGNITE-16141?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17485728#comment-17485728 ] 

Maksim Timonin commented on IGNITE-16141:
-----------------------------------------

> Currently it returns GridCursor<IndexRow> but underlied IndexRow isn't initialized with CacheDataRow

 

it's not true, as CacheDataRow initialized in moment of preparing buffer for cursor in BPlusTree. See IO#getRow(). 

 

Closed as not a bug.

> IndexQuery: consistency with query after IO filter
> --------------------------------------------------
>
>                 Key: IGNITE-16141
>                 URL: https://issues.apache.org/jira/browse/IGNITE-16141
>             Project: Ignite
>          Issue Type: New Feature
>            Reporter: Maksim Timonin
>            Assignee: Maksim Timonin
>            Priority: Major
>              Labels: IEP-71
>
> As we use InlineIO for internal filtering we should guarantee to return correct data row from this stage. Currently it returns GridCursor<IndexRow> but underlied IndexRow isn't initialized with CacheDataRow, and then it is affected by concurrent updates.
>  
> Alternative: additional filter step after extracting row - we should extract cache rows in loop, so we can apply our filter here again.



--
This message was sent by Atlassian Jira
(v8.20.1#820001)