Re: AW: [HACKERS] triggers, views and rules (not instead)

Zeugswetter Andreas SARZ wrote:
> Ok, to sum it up:
>         1. We need and want the select part of the rewrite rules.

Agreed.

>         2. for the insert/update/delete rules the old instance rules system
>             was much more appropriate. TODO: dig up the old code
>             and merge it with the current trigger Implementation
>             it must be pretty much the wanted functionality (it
>             supported sql)

??? Old instance rules system was removed by Jolly & Andrew and so
it never supported SQL. I hope that Jan will give us PL/pgSQL soon
and it will be used for triggers, without changing current trigger
implementation...

>                 3. the CURRENT keyword in the i/u/d rewrite rules is stupid
>                    and should be disabled, destroyed and burned in hell

Agreed, if standard hasn't it. I know that OLD & NEW are in standard,
for triggers atleast.

>                 4. To stick to the mainstream we should enhance the trigger
>                     syntax, and forget the rule stuff for i/u/d

Yes. Statement level triggers give the same functionality as rewrite
i/u/d rules. We could let them to return something special to skip
user' i/u/d itself, isn't it the same as INSTEAD ?

Vadim