Re: I: About "Our CLUSTER implementation is pessimal" patch

Sorry for the very delayed review.

On Wed, Jul 21, 2010 at 11:15 PM, Leonardo Francalanci <m_lists@yahoo.it> wrote:
> 2) what other areas can I comment more?

I think the patch is almost ready to commit, but still
have some comments for the usability and documentations.
I hope native English speakers would help improving docs.

* Documentation could be a bit more simplified like as
  "CLUSTER requires twice disk spaces of your original table".
  The added description seems too difficult for standard users.

* How to determine which method was used?
  We can get some information from trace_sort logs,
  but CLUSTER VERBOSE would be better to log
  which CLUSTER method was used.

* How to control which method will be used?
  It might be good to explain we can control the method
  with enable_seqscan/indexscan.

-- 
Itagaki Takahiro