Re: possible TODO: read-only tables, select from indexes only.

Hannu Krosing wrote:
> On R, 2005-04-22 at 11:40 -0400, Bruce Momjian wrote:
> > See this TODO:
> > 	
> > 	* Allow data to be pulled directly from indexes
> > 	
> > 	  Currently indexes do not have enough tuple visibility information
> > 	  to allow data to be pulled from the index without also accessing
> > 	  the heap.  One way to allow this is to set a bit to index tuples
> > 	  to indicate if a tuple is currently visible to all transactions
> > 	  when the first valid heap lookup happens. 
> 
> I don't think this is implementable in any reasonably cheap way (i.e.
> this will be a general performance loss). 
> 
> This has all the downsides of storing full visibility in index tuples,
> except the size.

Yea, I suppose.

> >           This bit would have to be cleared when a heap tuple is expired.
> 
> Does "expired" here mean marked for deletion ?

Yes.

> This can be prohibitively pricey for big tables with lots of indexes, as
> marking the tuple means alsn finding and possibly writing to all index
> enytries pointing to this tuple.

Yep, it could be very ugly, but it would help with our COUNT(*) problem
too.  Isn't there a solution?  If there isn't, I can remove the TODO
item.

> > I think this is the direction we should be heading because it has more
> > general usefulness.
> 
> OTOH this will probably never be implemented for the same reason that
> full visibility in index tuples will not, whereas my proposition can be
> implemented quite easily (it's just a SMOP).

I would like to find something that has more general usefulness.

--  Bruce Momjian                        |  http://candle.pha.pa.us pgman@candle.pha.pa.us               |  (610) 359-1001 +  If your life is a hard drive,     |  13 Roberts Road +  Christ can be your backup.        |  Newtown Square, Pennsylvania 19073