PonyORM: how to get rid of "subquery uses ungrouped column"?

Question

Please assume this data model. This is a simplified model of what I really have here but all important fields are there. Database: Postgres. I want to calculate some stats on those models and I stumbled upon this problem:

c = Customer.select().random(1)[0]  # example
query = left_join(
        (
            p,
            count(i.order.customer == c),
            count((b.is_used == True) for b in i.bonuses),
            count(i)
        ) for p in Product
        for i in p.order_items)

Now trying to show results (in reality I have some more filtering on the aggregations to do)

query.show()

yields this:

ProgrammingError: subquery uses ungrouped column "i.id" from outer query
LINE 4:     WHERE "i"."id" = "b"."order_item"

What can I do to correct this?

Resulting SQL looks like this:

SELECT "p"."id", COUNT(case when "order"."customer" = ? then 1 else null end), (
    SELECT COUNT(DISTINCT "b"."is_used" = 1)
    FROM "Bonus" "b"
    WHERE "i"."id" = "b"."order_item"
    ), COUNT(DISTINCT "i"."id")
FROM "Product" "p"
  LEFT JOIN "OrderItem" "i"
    ON "p"."id" = "i"."product"
  LEFT JOIN "Order" "order"
    ON "i"."order" = "order"."id"
GROUP BY "p"."id"

EDIT: My real models are made to fit a db used by Django, so I have _table_ in each class and column=something_id in all foreign keys. But aside that everything looks OK to the point I can make any simpler query with ease.

EDIT2: Here's the gist with my test data.

Answer 1

It appears that I needed to rephrase the code to this form:

query = left_join(
    (
        p,
        count(i.order.customer == c),
        count(b.is_used == True),
        count(i)
    ) for p in Product
    for i in p.order_items
    for b in i.bonuses)