也许有一种更巧妙的方法来解决这个问题，比如16.10（我认为）中引入了一些新的时态逻辑，但这会起到作用：

SELECT user_id, COUNT(creation_period)
FROM
(
   SELECT NORMALIZE
      user_id,
       CASE 
         WHEN PERIOD(creation_date, NEXT(creation_date + INTERVAL '15' MINUTE)) CONTAINS LEAD(creation_date) OVER (PARTITION BY user_id ORDER BY creation_date)
            THEN PERIOD(creation_date, NEXT(LEAD(creation_date) OVER (PARTITION BY user_id ORDER BY creation_date)))
         WHEN PERIOD(creation_date - INTERVAL '15' MINUTE, NEXT(CREATION_DATE)) CONTAINS LAG(creation_date) OVER (PARTITION BY user_id ORDER BY creation_date)
            THEN PERIOD(LAG(creation_date) OVER (PARTITION BY user_id ORDER BY creation_date), NEXT(creation_date))
         ELSE
            PERIOD(creation_date, NEXT(creation_date))
         END AS creation_period
   FROM table1
) dt
GROUP BY 1;

基本上，这是做的是缝合在一起的 Period 通过查找15分钟间隔内的相邻记录来确定数据类型。那么我们 Normalize 记录每个不同的 user_id 制造周期相互交叉或重叠的地方。然后我们只计算剩余的不同时期。
输入：

CREATE MULTISET VOLATILE TABLE table1
(
   user_id VARCHAR(30),
   creation_date TIMESTAMP(0)
) ON COMMIT PRESERVE ROWS;
INSERT INTO table1 VALUES ('user', CURRENT_TIMESTAMP(0));
INSERT INTO table1 VALUES ('user', CURRENT_TIMESTAMP(0) + INTERVAL '10' MINUTE);
INSERT INTO table1 VALUES ('user', CURRENT_TIMESTAMP(0) + INTERVAL '1' DAY);
INSERT INTO table1 VALUES ('user', CURRENT_TIMESTAMP(0) + INTERVAL '1' DAY + INTERVAL '15' MINUTE);
INSERT INTO table1 VALUES ('user', CURRENT_TIMESTAMP(0) + INTERVAL '2' DAY);
INSERT INTO table1 VALUES ('user', CURRENT_TIMESTAMP(0) + INTERVAL '3' DAY);
INSERT INTO table1 VALUES ('user', CURRENT_TIMESTAMP(0) + INTERVAL '3' DAY + INTERVAL '3' MINUTE);
INSERT INTO table1 VALUES ('user' || '2', CURRENT_TIMESTAMP(0));
INSERT INTO table1 VALUES ('user' || '2', CURRENT_TIMESTAMP(0) + INTERVAL '10' MINUTE);
INSERT INTO table1 VALUES ('user' || '2', CURRENT_TIMESTAMP(0) + INTERVAL '11' MINUTE);
INSERT INTO table1 VALUES ('user' || '2', CURRENT_TIMESTAMP(0) + INTERVAL '1' DAY + INTERVAL '15' MINUTE);
INSERT INTO table1 VALUES ('user' || '2', CURRENT_TIMESTAMP(0) + INTERVAL '1' DAY + INTERVAL '20' MINUTE);
INSERT INTO table1 VALUES ('user' || '2', CURRENT_TIMESTAMP(0) + INTERVAL '3' MINUTE);
INSERT INTO table1 VALUES ('user' || '2', CURRENT_TIMESTAMP(0) + INTERVAL '3' DAY + INTERVAL '3' MINUTE);

输出：

+---------+------------------------+
| user_id | Count(creation_period) |
+---------+------------------------+
| user    |                      4 |
| user2   |                      3 |
+---------+------------------------+