On Mon, 2002-05-20 at 16:08, Tom Lane wrote:
> Hannu Krosing <hannu@tm.ee> writes:
> > On Sun, 2002-05-19 at 19:37, Tom Lane wrote:
> >> I'd rather expect the toast indexes to grow given the lack-of-btree-
> >> collapse-logic issue.
>
> > Why sould the toast indexes grow significantly more than the primary key
> > of main table ?
>
> Well, the toast indexes will grow because they're using an OID key,
> and so the range of indexed values keeps increasing. AFAIR Mark didn't
> say whether he *had* a primary key, let alone what it was --- but it's
> possible that he has one that has a range that's not changing over the
> test.
his table is this:
CREATE TABLE grow (id integer,body text,CONSTRAINT grow_pk PRIMARY KEY (id))
> In particular, if the test consists simply of updating the toasted
> field, that will not change the primary keys at all ... but it will
> change the toast table's key range, because each new value will get
> a new toast OID.
But does PG not have a new index entry for each _version_ of table row ?
Or does lack-of-btree-collapse-logic affect only keys where there are
many _different_ keys and not many repeating keys?
--------------
Hannu