Re: [PATCHES] Partition: use triggers instead of rules

Tom Lane wrote:
> "Joshua D. Drake" <jd(at)commandprompt(dot)com> writes:
>> Rules are extremely slow in comparisons and not anywhere near as
>> flexible. As I said up post yesterday... they work well in the basic
>> partitioning configuration but anything else they are extremely deficient.
>
> I think that the above claim is exceedingly narrow-minded.

We are talking about partitioning. It is supposed to be narrow-minded.

> A trigger
> will probably beat a rule for inserts/updates involving a small number
> of rows.

Which is exactly what partitioning is doing.

For large numbers of rows, like an INSERT/SELECT from another
> large table, the rule is likely to win, because its overhead is paid
> once per query not once per row. Also, if you implement the trigger
> with an EXECUTE (forcing a planning cycle) intead of hard-coded
> commands, the speed advantage becomes even more dubious.

Not for partitioning. Although I agree with your sentiments for normal
operation.

Sincerely,

Joshua D. Drake