m w <[EMAIL PROTECTED]> writes:
> Inconsistent behavior indicates that a different
> matching algorithm is used if one uses an index
> instead of a table scan. That scares me.
A seq scan and an index scan are inherently different algorithms,
so I don't see exactly how you think we can avoid this risk.
In particular, if you are dealing with a btree index and a "WHERE
column = constant" query, then a seq scan is only going to be concerned
with the behavior of the '=' operator --- does it return TRUE or not
for any particular row? But an index search is inherently going to make
ordered comparisons (<, =, >). So there is always a potential for
inconsistent behavior if the ordering operators produce results that are
inconsistent with simple '='. We cannot design that away --- all we can
do is fix such bugs when one is discovered in a particular datatype.
regards, tom lane
