I like it. I don't want to break backward compatibility.
The new column could be named like query_qn, and will have here the
unambitious text of the query
where all the objects names are extended to the qualified names.
Sergei Agalakov
On 11/27/2018 2:17 PM, Alvaro Herrera wrote:
> On 2018-Nov-27, Sergei Agalakov wrote:
>
>> We do see that the queries are different but we can't see why they are so
>> much different in the execution time.
>> If the pg_stat_statements module would extend the object name to the
>> qualified names like s1.t1 and s2.t2 then we would see the report as
>> 3004391594 select count(*) from s1.t1
>> 1336375111 select count(*) from s2.t1
>> with an immediate understanding of what's going on.
> I think this is hard to do in a really useful manner -- if we qualify
> all names in the query, it becomes mangled to the point that it's no
> longer easily recognizable (users want to "grep" their applications to
> see where the queries are). Maybe we can have a second column that
> shows the query with all object names schema-qualified, keeping the
> original one that uses the query as entered by the user.
>