Re: interesting trigger behaviour in 8.3

"Ivan Zolotukhin" <ivan(dot)zolotukhin(at)gmail(dot)com> writes:
> In pseudo code it looks like the following. There are 2 tables, empty
> abstract_table with 3 columns (id, col1, col2) and many tables (e.g.
> inherited_table1_with_data) that inherit abstract_table.
> Constraint_exclusion is set up on id column and works perfectly. So
> we've got update like this

> UPDATE abstract_table SET col1 = 1, col2 = 2 WHERE id = 12345;

I bet it does not *really* look like that, but has a parameterized
WHERE clause. As per the fine manual:

Constraint exclusion only works when the query's WHERE clause
contains constants. A parameterized query will not be optimized,
since the planner cannot know which partitions the parameter value
might select at run time. For the same reason, "stable" functions
such as CURRENT_DATE must be avoided.

regards, tom lane