Jan Wieck <JanWieck(at)Yahoo(dot)com> writes:
> On 3/20/2007 1:11 PM, Tom Lane wrote:
>> search_path
>> add_missing_from
>> transform_null_equals
>> sql_inheritance
> Don't we actually store the parsetree in the query cache, and doesn't
> that actually make a lot of the above rather NOT affect the resulting
> plan any more?
No, what the code now does is to store the raw grammar output --- a
replan includes a fresh pass through parse_analyze. This must happen
if we want the thing to cope with dropping and replacing temp tables,
which is one of the main use-cases ...
regards, tom lane