Groupwise maximum query for getting records for latest date

问题

I'm trying to build a query that only gets the latest record on a group of records based on date.

The layout of the table is following:

| date | category | action | label | label2 | count_today | count_total | period |

The primary key is based on the columns date, category, action, label, label2, period. date has format yyyy-mm-dd and period can have the values Day, Week, month.

For each unique combination of category | action | label | label2 I need to have the record with the latest date.

My first attempt at this was this:

SELECT * FROM `statistic` 
WHERE 
 (action='total' OR action='' OR category='user')
 AND
 (period='day' 
   OR (period='week' AND DATEDIFF(now(), `date`) > 30)
   OR (period = 'Month' AND DATEDIFF(now(), `date`) > 7*26)
 )
GROUP BY category, action, label, label2
ORDER BY date DESC

The problem with this query is that it does the GROUP BY before the ORDER BY, causing incorrect records to be returned.

After searching, I found that what I want is called a group-wise maximum query.

My next attempt was this:

SELECT s1.* FROM `statistic` AS s1
LEFT JOIN statistic AS s2 
ON 
 s1.category = s2.category
 AND s1.action = s2.action
 AND s1.label = s2.label
 AND s1.label2 = s2.label2
 AND s1.date > s2.date
WHERE 
   (s1.action='total' OR s1.action='' OR s1.category='user')
   AND
   (s1.period='day' 
   OR (s1.period='week' AND DATEDIFF(now(), s1.`date`) > 30)
   OR (s1.period = 'Month' AND DATEDIFF(now(), s1.`date`) > 7*26)
 )
GROUP BY category, action, label, label2

But this query doesn't give me the correct results either (it looks similar to the first query).

Any clue how I can get the data that I need?

回答1:

You are right that you want the group-wise maximum, but you accomplish this by joining your table with a subquery that finds the latest date for each group:

SELECT * FROM statistic NATURAL JOIN (
  SELECT   category, action, label, label2, MAX(date) date
  FROM     statistic
  GROUP BY category, action, label, label2
) t

And then, if the following filters are still required:

WHERE 
 (action='total' OR action='' OR category='user')
 AND
 (period='day' 
   OR (period='week' AND DATEDIFF(now(), `date`) > 30)
   OR (period = 'Month' AND DATEDIFF(now(), `date`) > 7*26)
 )

来源：https://stackoverflow.com/questions/13451472/groupwise-maximum-query-for-getting-records-for-latest-date

标签

mysql

aggregate-functions