Re: Triggers with DO functionality

On 17 February 2012 21:07, Andrew Dunstan <andrew(at)dunslane(dot)net> wrote:
>
>
> On 02/17/2012 03:58 PM, Thom Brown wrote:
>>
>> On 17 February 2012 20:40, Dimitri Fontaine<dimitri(at)2ndquadrant(dot)fr>
>>  wrote:
>>>
>>> Thom Brown<thom(at)linux(dot)com>  writes:
>>>>
>>>> And thinking about it, DO is a bit nonsense here, so maybe we'd just
>>>> have something like:
>>>>
>>>> CREATE TRIGGER...
>>>> AS $$
>>>> BEGIN
>>>> END;
>>>> $$;
>>>>
>>>> i.e. the same as a function.
>>>
>>> I like that.  How do you tell which language the trigger is written in?
>>
>> Exactly the same as a function I'd imagine.  Just tack LANGUAGE
>> <language>; at the end.
>>
>>> I'm not so sure about other function properties (SET, COST, ROWS,
>>> SECURITY DEFINER etc) because applying default and punting users to go
>>> use the full CREATE FUNCTION syntax would be a practical answer here.
>>
>> *shrug* There's also the question about the stability of the trigger's
>> own in-line function too (i.e. IMMUTABLE, STABLE, VOLATILE).
>>
>
> This is going to be pretty much a piece of syntactic sugar. Would it matter
> that much if the trigger functions made thus are all volatile? If someone
> wants the full function feature set they can always use CREATE FUNCTION
> first. I think I'm with Dimitri - let's keep it simple.

Yes, always best to start with essential functionality.

--
Thom