Re: Merge Append Patch merged up to 85devel

On Sat, Jul 25, 2009 at 8:12 PM, Jaime
Casanova<jcasanov(at)systemguards(dot)com(dot)ec> wrote:
> i was trying to test this one but i can't find a query that produces a
> diferent plan than in 8.4.0, attached my current test just in case...
> what kind of query is this intended to help?

You may have to disable enable_seqscan to get simple examples like this to work:

select * from partitioned_table order by indexed_column;

more complex examples would trigger it naturally such as:

select * from partitioned_table where active order by indexed_column
(with an index on indexed_column where active)

or

select * from partitioned_table where indexed_column between x and y
order by indexed_column

> something, maybe style dependant, that i don't like is the definition
> of  LAPPEND_PATH_FLATTEN_APPENDPATHS macro at some point in the middle
> of the file i prefer they be defined at the top (just my
> preference)... or there is a reason for doing it there?

well it's only used in that one function, it's just some code which is
repeated three times and would obscure what's going on if it were
inlined.

> another thing a don't like is those #ifdef FIXME surrounding already
> existing if, why are those? and if they need to be fixed why there
> isn't a comment explaining what the fix is or what it should behave?

Yeah, if I knew how to fix them then this patch wouldn't be stuck
waiting for feedback... :(

--
greg
http://mit.edu/~gsstark/resume.pdf