为什么此子查询导致返回的行与直接查询不同？

Question

Here's the relevant data

SELECT item_group_id
      ,item_id
      ,invoice_id
FROM item
WHERE item_group_id = 92480

# Results
92480   215342  88902
92480   215343  88902
92480   215344  88902
92480   215345  88902
92480   215346  90690
92480   215347  90690
92480   215348  NULL
92480   215349  NULL

Here's the direct query

SELECT MAX(item_group_id) [item_group_id]
      ,MAX(invoice_id) [invoice_id]
FROM item
WHERE item_group_id = 92480
GROUP BY invoice_id

# Results
92480   NULL
92480   88902
92480   90690

Here's the subquery

SELECT i.item_group_id
      ,i.invoice_id
FROM (
  SELECT MAX(item_group_id) [item_group_id]
        ,MAX(invoice_id) [invoice_id]
  FROM item
  GROUP BY invoice_id
) i
WHERE i.item_group_id = 92480

# Results
92480   90690
92480   88902

I'm not sure what's going on here, I would assume that all result rows from the subquery will make up the ad-hoc table queries, but those rows with null invoice_ids are not available.

EDIT: As the selected answer below alludes to, this is a grouping problem. The item table has more item entries in it, and with different item_group_id s. Since item_group_id is not grouped by, in the subquery all the items with a null invoice_id field are grouped together. The MAX(item_group_id) selection then returns the highest item_group_id , not any of a set of identical item_group_ids as in the direct query.

Changing GROUP BY invoice_id to GROUP BY item_group_id, invoice_id yields what I'm looking for.

Answer 1

The result is different because the WHERE filter is applied at different stages of query execution in the two queries.

In the first query WHERE is applied first and then the result is grouped, whereas in the second query the GROUP ing is done first (because it is in a subquery) and the where is applied on the result of the grouping.