[jira] [Commented] (CASSANDRA-2777) Pig storage handler should implement LoadMetadata

["Jeremy Hanna (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/CASSANDRA-2777?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13050159#comment-13050159 ] 

Jeremy Hanna commented on CASSANDRA-2777:
-----------------------------------------

while we're add it can we remove the redundant addMutation call on line 505 and on line 513 add the e param on:
{quote}
throw new IOException(e + " Output must be (key, {(column,value)...}) for ColumnFamily or (key, {supercolumn:{(column,value)...}...}) for SuperColumnFamily", e);
{quote}

> Pig storage handler should implement LoadMetadata
> -------------------------------------------------
>
>                 Key: CASSANDRA-2777
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-2777
>             Project: Cassandra
>          Issue Type: Improvement
>          Components: Contrib
>            Reporter: Brandon Williams
>            Assignee: Brandon Williams
>            Priority: Minor
>             Fix For: 0.7.7
>
>         Attachments: 2777.txt
>
>
> The reason for this is many builtin functions like SUM won't work on longs (you can workaround using LongSum, but that's lame) because the query planner doesn't know about the types beforehand, even though we are casting to native longs.
> There is some impact to this, though.  With LoadMetadata implemented, existing scripts that specify schema will need to remove it (since LM is doing it for them) and they will need to conform to LM's terminology (key, columns, name, value) within the script.  This is trivial to change, however, and the increased functionality is worth the switch.

--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira