[Charset iso-8859-1 unsupported, filtering to ASCII...]
> > > And, while we are on heap subject - using index (RECNO) for heap
> > > means that all our secondary-index scans will performe TWO
> > > index scans - first, to find recno in secondary-index, and
> > > second, to find heap tuple using recno (now indices give us
> > > TID, which is physical address).
> >
> > Yes, that was one of my questions. Why use recno at all? We already
> > have heap access which is very fast. Why switch to SDB which gives us
> > a recno ordering of heap that doesn't do us any real good, except to
> > allow tuple update without changing indexes.
>
> But if we'll use our heap AM, then we'll have to implement redo/undo
> for it... no sence to switch to SDB for btree/hash WAL support -:)
Yes, SDB would give us redo/undo in heap, and that would make things
easier. However, if there is the overhead of a double-index lookup when
using indexes, it seems like a very high cost.
-- Bruce Momjian | http://www.op.net/~candle pgman@candle.pha.pa.us | (610)
853-3000+ If your life is a hard drive, | 830 Blythe Avenue + Christ can be your backup. | Drexel Hill,
Pennsylvania19026