Re: select single entry and its neighbours using direct-acess to index?

=?iso-8859-15?Q?Pierre-Fr=E9d=E9ric_Caillaud?= <lists(at)boutiquenumerique(dot)com> writes:
> SELECT * FROM test WHERE (name='b' and name2>'a') OR (name>'b') ORDER BY
> name,name2 ASC LIMIT 1;
>> Write that WHERE clause instead as:
>> WHERE name>='b' AND (name>'b' OR (name='b' AND name2>'a'))
>> This is logically equivalent, but it gives the planner a better handle on
>> how to use an index scan to satisfy the query.

> I thought the planner had an automatic rewriter for these situations.

No. There was a prior discussion of this, saying that we really ought
to support the SQL-spec row comparison syntax:
... WHERE (name, name2) > ('b', 'a');
which would map directly onto the semantics of a 2-column index. We
don't have that functionality yet, though (we take the syntax but the
semantics are not SQL-compliant) let alone any ability to pass it
through to a 2-column index.

regards, tom lane