Alvaro Herrera <alvherre@xxxxxxxxxxxxxx> writes: > On Mon, Sep 12, 2005 at 10:05:36AM -0400, Reid Thompson wrote: >> Why is that? In order to use an index, does the query have to utilize >> the 'first' element of the index? > The "leftmost part." There's no way to scan an index if you don't know > the key. On a btree index, the key is ordered, and the columns at the > left are more significant than those at the right. If you don't provide > a value for the leftmost (first) column, there's no way to start > scanning the index because there's no starting point. Actually, btree doesn't have any particular problem with that --- it just starts the scan at the beginning of the index. However the other index types do all require a constraint on the first index column; for instance hash has to be able to determine a hash value. Greg Stark suggests here: http://archives.postgresql.org/pgsql-hackers/2005-05/msg00966.php that GiST could also be fixed to work with any subset of the index columns, but it hasn't been done yet, unless Teodor and Oleg snuck something in during that last round of GiST work. regards, tom lane ---------------------------(end of broadcast)--------------------------- TIP 9: In versions below 8.0, the planner will ignore your desire to choose an index scan if your joining column's datatypes do not match
