Re: Generating code for query jumbling through gen_node_support.pl

On 11/7/2023 05:35, Michael Paquier wrote:
> On Mon, Jan 30, 2023 at 11:48:45AM +0100, Peter Eisentraut wrote:
>> On 27.01.23 03:59, Michael Paquier wrote:
>>> At the end, that would be unnoticeable for the average user, I guess,
>>> but here are the numbers I get on my laptop :)
>>
>> Personally, I think we do not want the two jumble methods in parallel.
>>
>> Maybe there are other opinions.
>
> (Thanks Jonathan for the poke.)
>
> Now that we are in mid-beta for 16, it would be a good time to
> conclude on this open item:
> "Reconsider a utility_query_id GUC to control if query jumbling of
> utilities can go through the past string-only mode and the new mode?"
>
> In Postgres ~15, utility commands used a hash of the query string to
> compute their query ID. The current query jumbling code uses a Query
> instead, like any other queries. I have registered this open item as
> a self-reminder, mostly in case there would be an argument to have a
> GUC where users could switch from one mode to another. See here as
> well for some computation times for each method (table is in ns, wiht
> millions of iterations):
> https://www.postgresql.org/message-id/Y9eeYinDb1AcpWrG@paquier.xyz
>
> I still don't think that we need both methods based on these numbers,
> but there may be more opinions about that? Are people OK if this open
> item is discarded?
I vote for only one method based on a query tree structure.
BTW, did you think about different algorithms of queryId generation?
Auto-generated queryId code can open a way for extensions to have
easy-supporting custom queryIds.

--
regards,
Andrey Lepikhov
Postgres Professional