Re: IDLE in transaction introspection

On Fri, Nov 4, 2011 at 9:48 AM, Magnus Hagander <magnus(at)hagander(dot)net> wrote:
> On Fri, Nov 4, 2011 at 14:42, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>> Marti Raudsepp <marti(at)juffo(dot)org> writes:
>>> While we're already breaking everything, we could remove the "waiting"
>>> column and use a state with value 'waiting' instead.
>>
>> -1 ... I think it's useful to see the underlying state as well as the
>> waiting flag.  Also, this would represent breakage of part of the API
>> that doesn't need to be broken.
>
> I guess with the changes that showed different thing like fastpath,
> that makes sense. But if we just mapped the states that are there
> today straight off, is there any case where waiting can be true, when
> we're either idle or idle in transaction? I think not..

Maybe if we get a sinval reset while we're just sitting there? Not sure.

In any case, I agree with Tom. I think that most people will be happy
to have a "query" column and a "state" column rather than a
"current_query" column that amalgamates the two, but I see no reason
to tinker with the waiting column if the changes we want to make don't
otherwise require it.

>>> Also, returning these as text seems a little lame. Should there be an
>>> enum type for that?
>>
>> Perhaps, but we don't really use enum types in any other system views,
>> so inventing one here would be out of character.
>
> Yeha, that seems inconsistent. Using a single character might work -
> but it's not particularly user-friendly to do that in the view itself.

+1 for keeping it as text, but loosing the angle brackets.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company