cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Sylvain Lebresne (JIRA)" <>
Subject [jira] [Commented] (CASSANDRA-4929) Deal with broken ALTER DROP support in CQL3
Date Wed, 14 Nov 2012 14:26:12 GMT


Sylvain Lebresne commented on CASSANDRA-4929:

bq. If we think we could do 3919 for 1.2.1 or .2 then my preferred option would be "leave
DROP out entirely for now."

As said above, I'm fine with that. Worst case scenario, if #3919 proves trickier than we sought
and/or if we find some compelling use case for a drop that only drops the metadata, we can
add that later. For now, attaching patch to remove the support (the patch simply makes the
syntax invalid, I figured there is no reason to remove code internally that we'll need for

bq. I think there's a valid use case in wanting to drop metadata for individual column definitions

Which would be?
> Deal with broken ALTER DROP support in CQL3
> -------------------------------------------
>                 Key: CASSANDRA-4929
>                 URL:
>             Project: Cassandra
>          Issue Type: Bug
>    Affects Versions: 1.2.0 beta 1
>            Reporter: Sylvain Lebresne
>            Assignee: Sylvain Lebresne
>            Priority: Minor
>             Fix For: 1.2.0 rc1
>         Attachments: 4929.txt
> Currently, {{ALTER DROP}} only remove the metadata for the column, making it unavailable,
but don't reclaim the data. This is unintuive and CASSANDRA-3919 is opened to fix it. However,
that later issue won't make it for 1.2, and I think we should be very careful into shipping
1.2 with the current behavior because 1) it's unintuitive and 2) as unintuitive as it is,
we don't want people to start relying on that behavior. So I thing we should do one of:
> * remove support for {{ALTER DROP}} until CASSANDRA-3919 reintroduce it. After all, there
is no real performance impact in keeping a colum that you don't use and if you really really
want to get rid of the metadata, you still have the workaround of trashing the schema and
recreating it without that column (obviously not user friendly, but at least it's vaguely
> * add a specific syntax for the current behavior of {{ALTER DROP}}, one that clearly
imply that the data is not deleted, if we consider that this behavior can be sometimes useful
(that is, even after CASSANDRA-3919 is resolved). One such syntax could one of (not sure which
one I prefer):
> {noformat}
> {noformat}
> I have a slight preference for solution 2, but honestly because it will it easier to
drop a column you've just added but maybe mispelled the name until CASSANDRA-3919. Once CASSANDRA-3919
is in, I'm not sure this will be so useful anymore.

This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see:

View raw message