Postgres嵌套查询并保持顺序

Question

I have two queries and want to combine into one.

Query 1(Pseudocode):

recipe_titles = select recipe_title
             from recipe
             where nutrient_id in (4,5)
             group by recipe_title
             order by sum(nutrient_amount) desc

Then, I want to use each recipe_title to select the whole recipe row:

Query 2(Pseudocode):

select * from recipe where recipe_title = recipe_titles[n].    (n is 1, then 2, then 3...)

If the first query returns 100 recipe_title , I will have to run the second query for 100 times, which is not efficient.

Is there a way to do these two selection in a single query?

IN will not work for my case since the order is lost:

select * from recipe where recipe_title in recipe_titles. (Not work)

Answer 1

You can use window functions. I would recommend a window conditional window sum() , that you can then use for filtering and ordering:

select *
from (
    select 
        r.*,
        sum(nutrient_amount) 
            filter(where nutrient_id in (4,5)) 
            over(partition by recipe_title) as sum_nutrient_amount
    from recipe r
) t
where sum_nutrient_amount > 0
order by sum_nutrient_amount desc, recipe_title, nutrient_id 

On the other hand, if you don't need to filter on recipe_title s that have either nutrient_id 4 or 5 , then the subquery is not necessary:

select r.* 
from recipe r
order by 
    sum(nutrient_amount) 
        filter(where nutrient_id in (4,5)) 
        over(partition by recipe_title) desc,
    recipe_title, 
    nutrient_id