You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@phoenix.apache.org by "Lars Hofhansl (JIRA)" <ji...@apache.org> on 2014/10/21 05:14:33 UTC

[jira] [Commented] (PHOENIX-294) Avoid adding delete markers for rows when dropping a column

    [ https://issues.apache.org/jira/browse/PHOENIX-294?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14177873#comment-14177873 ] 

Lars Hofhansl commented on PHOENIX-294:
---------------------------------------

Brainstormed with James a bit today. Other options include:
* mark column deleted and disallow re-creating the same column again
* mark column deleted and clean up lazily during compactions (don't even have to wait for a major compaction). Would need to keep track of major compaction that followed a drop of a column to know when it is safe to reuse that column name again

> Avoid adding delete markers for rows when dropping a column
> -----------------------------------------------------------
>
>                 Key: PHOENIX-294
>                 URL: https://issues.apache.org/jira/browse/PHOENIX-294
>             Project: Phoenix
>          Issue Type: Task
>            Reporter: larsh
>
> Talked to James a bit. Currently when a column is dropped we place delete markers for this column on each row. This is very inefficient when there are many row.
> Another option is to drop the column from just the metadata.
> Since the schema is versioned the only problematic queries would be a wildcard (select * from...) queries.
> For those wildcard queries we can examine the current schema (at the appropriate timestamp) and expand the * to the actual set of columns.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)