Question

I am building a blog and i have a problem. There should be some sticky posts. All i want is to get first the sticky posts and then the rest.

One query that works is

select * from 

(
(select *,true as st from blog where "stickyUntil" > current_timestamp) 
UNION  all
(SELECT *,false as st from blog where "stickyUntil" < current_timestamp  or "stickyUntil" is null  ) 
) q 


order by st desc, "stickyUntil" DESC ,publish DESC OFFEST x LIMIT z

Another simpler query that workz is

select * from blog order by case when "stickyUntil" > current_timestamp then "stickyUntil" end desc nulls last, publish desc;

But this would force 200.000 rows to be sorted in memory not very fast..

Is there a way to optimize it ??

Would it be better to use two separate queries ?
Thanks

Answer 1

I would go with a CASE statement to avoid two passes of the table:

select *, 
   CASE 
   WHEN "stickyUntil" > current_timestamp THEN true
   ELSE false
   END as st
FROM blog
ORDER BY st DESC ,publish DESC OFFEST x LIMIT z

Postgres supports creating an INDEX on a computed field which would help here but there is a restriction:

All functions and operators used in an index definition must be “immutable”,
that is, their results must depend only on their arguments and never on any
outside influence (such as the contents of another table or the current time).

So you cannot index the st calculation. If the calculation of st does not need to be too accurate another option is to add the st field to the table:

ALTER TABLE blog
ADD COLUMN st boolean default true

INDEX st as a regular column:

CREATE INDEX sti ON blog(st)

And periodically run:

UPDATE blog
SET st = false
WHERE st = true
AND "stickyUntil" < current_timestamp

but having a polling process doing an update is much less attractive than a straight up query. This would only make sense if your query is really slow or there are a lot of reads of the blog table.