[zebra] Zebra Column Group Naming Support

                 Key: PIG-986
                 URL: https://issues.apache.org/jira/browse/PIG-986
             Project: Pig
          Issue Type: New Feature
          Components: impl
            Reporter: Chao Wang
            Assignee: Chao Wang
             Fix For: 0.5.0

We introduce column group name to Zebra and make it a first-class citizen in 
Zebra. This can ease management of column groups.
We plan to introduce an "as" clause for column group name in Zebra's syntax.

Functional Specifications:
1) Column group names are optional. For column groups which do not have a 
user-provided name, Zebra will assign some default column group names 
internally that is unique for that table - CG0, CG1, CG2 ... Note: If CGx is 
used by user, then it can not be used for internal names.

2) We introduce an "AS" clause in Zebra's syntax for column group names. If it 
occurs, it has to immediately follow [ ]. For example, "[a1, a2] as PI secure 
by user:joe group:secure perm:640; [a3, a4] as General compress by lzo". Note 
that keyword "AS" is case insensitive.

3) Column group names are unique within one table and are case sensitive, i.e., 
c1 and C1 are different.

4) Column group names will be used as the physical column group directory path 

5) Zebra V2 will support dropColumnGroup by column group names (will integrate 
with Raghu's A29 drop column work).

6) Zebra V2 can support backward compatibility (If there are Zebra V1 created 
tables in production when V2 is released). More specifically, this means that 
Zebra V2 can load from V1-created tables and do dropColumnGroup on it.

7) Does NOT support renaming.

This message is automatically generated by JIRA.
You can reply to this email to add a comment to the issue online.

Reply via email to