Most people are looking for limit and offset, so they can implement
result paging. Without support for limit and offset most people use
setMaxRows in the statement for limit and something like relative(int)
in the result set for the offset. The problem I see, is the planner
does not know how much the user intends to skip until after the query
is processed, but maybe this can't be optimized anyway.
-dain
On Nov 28, 2004, at 8:03 AM, Jan Hlavat� wrote:
Daniel John Debrunner wrote:
Derby supports limiting the number of rows returned by a query through
the standard JDBC java.sql.Statement.setMaxRows() method.
And does it affect the execution plan? Knowing the maximum number of
records
would be extremely useful for query optimizer, which could select a
much faster
plan using indexes which otherwise has higher "per record" cost over
mass methods
like full table scan...
This was (is?) one of the major shortcomings in SapDB - it always
realized query for first few records starting with specific value in
an (existing) index order
by full table scan and sort, thinking the index scan would cost much
more.
Jan
