如何选择日期间隔

Question

This is my table:

+------+--------+-------------------+
| user | item   | date_time         |
|   10 |     01 | 10-10-10 20:10:05 |
|   10 |     02 | 10-10-10 20:10:10 |
|   10 |     03 | 10-10-10 20:10:10 |
|   20 |     02 | 10-10-10 20:15:10 |
|   20 |     02 | 10-10-10 20:20:10 |
|   30 |     10 | 10-10-10 20:01:10 |
|   30 |     20 | 10-10-10 20:01:20 |
|   30 |     30 | 10-10-10 20:05:20 |
+------+--------+-------------------+

I want to do a query that return a user that took multiple items in a 1min interval, like this result:

+------+--------+-------------------+
| user | item   | date_time         |
|   10 |     01 | 10-10-10 20:10:05 |
|   10 |     02 | 10-10-10 20:10:10 |
|   10 |     03 | 10-10-10 20:10:10 |
|   30 |     10 | 10-10-10 20:01:10 |
|   30 |     20 | 10-10-10 20:01:20 |
+------+--------+-------------------+

How can I do this?

---

Edit

And if I want to only display users that appear 2 or more times on this output?

Example:

+------+--------+-------------------+
| user | item   | date_time         |
|   10 |     01 | 10-10-10 20:10:05 |
|   10 |     02 | 10-10-10 20:10:10 |
|   10 |     03 | 10-10-10 20:10:10 |
+------+--------+-------------------+

Answer 1

You'll have to join the table against itself (let's call the table aliases T1 and T2). Then write a WHERE clause to filter only rows where T1.user is equal T2.user and the absolute value of the difference between T1.date_time and T2.date_time is less than a minute.

But the problem is that every row will get selected, because you have no primary key on your table, thus there's no way to detect when a row is getting joined with itself. Create a primary key (an autonumber sequence will work just fine), and add a condition to the WHERE clause that says T1.id <> T2.id .

So in (untested) code form:

SELECT *
FROM stuff T1, stuff T2
WHERE T1.user = T2.user
  AND ABS(UNIX_TIMESTAMP(T1.date_time) - UNIX_TIMESTAMP(T2.date_time)) < 60
  AND T1.id <> T2.id;

Answer 2

Yes, you need to add a primary key (id) as suggested by @Will.

To get each item once (and only once) regardless of how many matches there were within the 1-minute window, try a subquery instead of a full join:

Select user,item,date_time from my_table t1
  where id in (select t2.id from my_table t2,my_table t3
      where t2.id <> t3.id and t2.user = t3.user 
           and abs(t2.date_time - t3.date_time) < 60)

--edit-- For your edited question, that depends on exactly what you mean. Do you mean "users who bought 3 or more items within 60 seconds" or "users who appear more than 2 times in the output". The latter is easy to do: assume the results of the above query are saved in a temporary table (or view) "temp1":

select * from my_table where user in 
    (select user from temp1 group by user having count(*) > 2);

Answer 3

let's try that one:

SELECT * FROM myTable  
JOIN (SELECT MAX(date_time) AS maxi, MIN(date_time) AS mini, user AS uid FROM myTable GROUP BY uid) AS otherTable  
ON date_time<=maxi AND date_time>=mini AND user = uid AND UNIX_TIMESTAMP(maxi) - UNIX_TIMESTAMP(mini) < 60  
GROUP BY uid, date_time