#73 Table redistribution: use of other ALTER TABLE commands

[Michael Paquier]

ALTER TABLE redistribution is implemented in parallel of other ALTER TABLE commands, but other commands like column addition or deletion cannot be used. The table rewrite needs to be extended to provide support at cluster level.
However, there are 2 issues related to the way tuples of the table are scanned:
- If tuples are fetched 1 by one and redistributed 1 by 1 in the way PostgreSQL does, a batch processing needing 2 connections to remote nodes is necessary. This is... well... challenging
- Or the current tuplestore system could be used and rewrite could be extended by replacing the current tuple scan by a tuplestore scan. However there is an issue with the timing of ALTER TABLE commands sent to remote nodes. For example, if data transformed with a column addition is sent before ALTER TABLE reaches the remote nodes. they will complain. So here the timing at which ALTER TABLE commands is sent to remote nodes is critical