In any thread involving SQL NULL, I see a lot of not-quite-right explanations of what SQL NULL is, conceptually. I challenge anyone who feels like they understand NULL conceptually to explain the following query:

    -- find orders with a total less than $10000
    select order_id, sum(price)
      from orders o left join order_lineitems l using (order_id)
      group by order_id having sum(price) < 10000;

This query is actually incorrect. Orders with no line items at all are clearly less than $10000, but they will be excluded because: first, the left outer join produces a NULL for the price; second, the group aggregation with SUM over that NULL will result in NULL; and third, the HAVING clause treats that NULL as false-like and excludes the order from the result.

Of course, we can explain procedurally what's happening here, and each individual step makes some sense. But the end result has no conceptual integrity.

Extra challenge: explain why using COUNT instead of SUM in the query does correctly return orders with fewer than 4 items:

    -- find orders with fewer than 4 line items
    select order_id, count(price)
      from orders o left join order_lineitems l using (order_id)
      group by order_id having count(price) < 4;

PS: thank you to the author for a developer-friendly feature that adds flexibility here!