[ 
https://issues.apache.org/jira/browse/CASSANDRA-12705?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15564372#comment-15564372
 ] 

Stefania commented on CASSANDRA-12705:
--------------------------------------

Thank for the review.

bq. Don't know what it looked like in the original version, but in 
system_schema.columns we store the lowercase values of the enum names. Look at 
kind and clustering_order for examples. Would prefer it to be the same in 
system_schema.dropped_columns, for consistency.

I had noticed this behavior, but I preferred not to do the same because I could 
not understand why convert to lowercase before writing only to convert back to 
uppercase after reading. It seemed like a waste of CPU cycles, but thinking 
about it again, it must be for the benefit of humans reading the tables?

Regardless, for the sake of consistency, I've updated the patch to store 
lowercase values. I've restarted CI and, unless objections or failures, I will 
commit once the results are available.

> Add column definition kind to system schema dropped columns
> -----------------------------------------------------------
>
>                 Key: CASSANDRA-12705
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-12705
>             Project: Cassandra
>          Issue Type: Bug
>          Components: Distributed Metadata
>            Reporter: Stefania
>            Assignee: Stefania
>             Fix For: 4.0
>
>
> Both regular and static columns can currently be dropped by users, but this 
> information is currently not stored in {{SchemaKeyspace.DroppedColumns}}. As 
> a consequence, {{CFMetadata.getDroppedColumnDefinition}} returns a regular 
> column and this has caused problems such as CASSANDRA-12582.
> We should add the column kind to {{SchemaKeyspace.DroppedColumns}} so that 
> {{CFMetadata.getDroppedColumnDefinition}} can create the correct column 
> definition. However, altering schema tables would cause inter-node 
> communication failures during a rolling upgrade, see CASSANDRA-12236. 
> Therefore we should wait for a full schema migration when upgrading to the 
> next major version.



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

Reply via email to