LEFT JOIN latest row using a date column - mysql

Question

I am trying to select the latest row from a LEFT JOIN not on the main query. This is my SQL:
SELECT *
FROM messages
LEFT JOIN users
ON messages.message_to = users.user_id
OR messages.message_user = users.user_id
LEFT JOIN message_replies
ON messages.message_id = message_replies.message_reply_main
LEFT JOIN user_personal_information
ON users.user_id =
user_personal_information.user_personal_information_user
I want to select the latest row from:
LEFT JOIN message_replies
ON messages.message_id = message_replies.message_reply_main
My column is called: message_reply_date - how can I use that to LEFT JOIN the latest row?
message_replies:
CREATE TABLE IF NOT EXISTS `message_replies` (
`message_reply_id` int(11) NOT NULL,
`message_reply_user` int(11) NOT NULL,
`message_reply_main` int(11) NOT NULL,
`message_reply_message` text NOT NULL,
`message_reply_date` datetime NOT NULL DEFAULT CURRENT_TIMESTAMP,
`message_reply_read` int(11) NOT NULL DEFAULT '0'
) ENGINE=InnoDB DEFAULT CHARSET=latin1;
I am using this for the WHERE clause:
WHERE m.message_user = ?
OR m.message_to = ?
AND m.message_deleted=0
AND m.message_permdeleted=0
GROUP BY mr.message_reply_main
ORDER BY mr.message_reply_date DESC

Answer 1

If i understood the question, i'll do it this simple way:
SELECT *
FROM messages
LEFT JOIN users
ON messages.message_to = users.user_id OR messages.message_user = users.user_id
LEFT JOIN message_replies
ON messages.message_id = message_replies.message_reply_main
LEFT JOIN user_personal_information
ON users.user_id = user_personal_information.user_personal_information_user
WHERE message_replies.message_reply_date = (SELECT MAX(message_reply_date) FROM message_replies WHERE message_reply_main = messages.message_id)
/* AND more criterias */
No grouping in the main query but just using a MAX evaluation of message_reply_date in a subquery regarding the WHERE criteria on message_reply_date itself.

Answer 2

There is a "left join" method for getting the most recent message (as well as several others). But in keeping with the preferred join method of the question:
SELECT *
FROM messages m LEFT JOIN
users u
ON m.message_to = u.user_id OR m.message_user = u.user_id LEFT JOIN
message_replies mr
ON m.message_id = mr.message_reply_main LEFT JOIN
user_personal_information upi
ON u.user_id = upi.user_personal_information_user LEFT JOIN
message_replies mr2
ON mr2.message_reply_main = m.message_id AND
mr2.message_reply_date > mr.message_reply_date
WHERE mr2.message_reply_main IS NULL;
I also added table aliases because they make a query easier to write and to read.
The idea is to match to the table again, but only for later messages. Then, the WHERE clause checks that none exist -- so it has the latest one.

Answer 3

Give this a go.
SELECT *
FROM messages
LEFT JOIN users
ON messages.message_to = users.user_id
OR messages.message_user = users.user_id
LEFT JOIN ( SELECT *
, MAX(message_reply_date) OVER (PARTITION BY message_id) AS most_recent_message_reply_date
FROM message_replies ) message_replies
ON messages.message_id = message_replies.message_reply_main
AND message_replies.message_reply_date = message_replies.most_recent_message_reply_date
LEFT JOIN user_personal_information
ON users.user_id =
user_personal_information.user_personal_information_user
I've substituted your direct call to the message_replies table with a Select that displays the table and the maximum reply date grouped by the message_id. I've then included this in the join to filter all answers other than the one your looking for.
Have a go, any problem drop me a message, always happy to help.

Answer 4

What you want to do is emulate the rank & partition functionality seen in MSSQL & PL/SQL, essentially ranking on date & partitioning by message_reply_main.
This can be done in MySQL using an outer join & a count on your partition, while using a < or > in the join for the ranking criteria. This allows you to do your standard join, but also where rank = 1, returning the largest date.
SELECT *
FROM messages
LEFT JOIN users
ON messages.message_to = users.user_id
OR messages.message_user = users.user_id
LEFT JOIN (select mri.message_reply_main
, mri.message_reply_date
, mri.message_reply_message
, mri.message_reply_read
, count(mri2.message_reply_main) + 1 as rank
FROM message_replies mri
left join message_replies mri2 on mri.message_reply_date < mri2.message_reply_date
and mri.message_reply_main = mri2.message_reply_main
group by mri.message_reply_main, mri.message_reply_message
, mri.message_reply_read, mri.message_reply_date) mr ON messages.message_id = mr.message_reply_man
AND mr.rank = 1
LEFT JOIN user_personal_information
ON users.user_id =
user_personal_information.user_personal_information_user