问题
The following SQL query:
SELECT messages.id, messages.created_at, comments.created_at FROM messages
LEFT JOIN comments ON comments.message_id = messages.id
WHERE (messages.id IN (429,443))
ORDER BY GREATEST(messages.created_at, comments.created_at) DESC
returns:
id messages.created_at comments.created_at
--------------------------------------------------------
443 2 5
429 1 4
443 2 3
(I replaced dates with numbers for readability)
To get each id only once I added DISTINCT:
SELECT DISTINCT messages.id FROM messages
LEFT JOIN comments ON comments.message_id = messages.id
WHERE (messages.id IN (429,443))
ORDER BY GREATEST(messages.created_at, comments.created_at) DESC
But, in the result the id values changed order:
id
---
429
443
What could be the reason for that ?
How could I keep the order ?
回答1:
the distinct key word is doing what it's supposed to do, return one row each with a given column value. Distinct doesn't allow you to specify which such row will be returned, and it's clear from the original query that such an ordering is allowed (there is a row with id 443 that follows a row with id 429).
To take control of what rows will be returned, you need to reformulate the query. A typical solution I'll take is to use a group by, selecting the group column and the desired row from each group, something to the effect of
SELECT message.id, MAX(message.created_at) FROM message GROUP BY message.id;
If I need to do more, I'll use this sort of query as a subselect in a larger query, possibly joining on the id field to get more fields from the preferred row, Or ordering the query in a particular way.
来源：https://stackoverflow.com/questions/7026234/sql-how-to-keep-rows-order-with-distinct