BigQuery 的最长连续天数

Question

现在我只是汇总了用户工作了多少天。我正在尝试将此查询更改为最连续工作天数。

u12345 是 4，u1 是 2。

这可能与 BigQuery 语句有关吗？

编辑我是 Kind of close，有以下查询，但我的 u1 得到 3 而不是 2。

SELECT MIN(e.timestamp) as date_created, e.uid, COUNT(e.uid) + 1 AS streak
FROM OnSite e
LEFT JOIN OnSite ee 
  ON e.uid = ee.uid 
AND DATE(e.timestamp) = DATE(DATE_ADD(ee.timestamp, INTERVAL -1 DAY))
WHERE ee.uid IS NOT NULL
GROUP BY e.uid;

---

架构 (MySQL v5.7)

CREATE TABLE OnSite
    (`uid` varchar(55), `worksite_id`  varchar(55), `timestamp` datetime)
;

INSERT INTO OnSite
    (`uid`, `worksite_id`, `timestamp`)
VALUES
  ("u12345", "worksite_1", '2019-01-01'),
  ("u12345", "worksite_1", '2019-01-02'),
  ("u12345", "worksite_1", '2019-01-03'),
  ("u12345", "worksite_1", '2019-01-04'),
  ("u12345", "worksite_1", '2019-01-06'),
  ("u1", "worksite_1", '2019-01-01'),
  ("u1", "worksite_1", '2019-01-02'),
  ("u1", "worksite_1", '2019-01-05'),
  ("u1", "worksite_1", '2019-01-06')

;

---

查询 #1

SELECT    uid, COUNT(DISTINCT timestamp) Total
FROM      OnSite
GROUP BY  uid;

| uid    | Total |
| ------ | ----- |
| u1     | 4     |
| u12345 | 5     |

---

View on DB Fiddle

Answer 1

以下是 BigQuery 标准 SQL

如果您对同一网站上用户的最大连续天数感兴趣：

#standardSQL
SELECT uid, MAX(consecuitive_days) max_consecuitive_days
FROM (
  SELECT uid, grp, COUNT(1) consecuitive_days
  FROM (
    SELECT uid, 
      COUNTIF(step > 1) OVER(PARTITION BY uid, worksite_id ORDER BY ts) grp
    FROM (
      SELECT uid, worksite_id, ts, 
        DATE_DIFF(ts, LAG(ts) OVER(PARTITION BY uid, worksite_id ORDER BY ts), DAY) step 
      FROM `project.dataset.table`
    )
  ) GROUP BY uid, grp
) GROUP BY uid  

如果工作地点无关紧要，而您只是在寻找最多连续天数：

#standardSQL
SELECT uid, MAX(consecuitive_days) max_consecuitive_days
FROM (
  SELECT uid, grp, COUNT(1) consecuitive_days
  FROM (
    SELECT uid, 
      COUNTIF(step > 1) OVER(PARTITION BY uid ORDER BY ts) grp
    FROM (
      SELECT uid, ts, 
        DATE_DIFF(ts, LAG(ts) OVER(PARTITION BY uid ORDER BY ts), DAY) step 
      FROM `project.dataset.table`
    )
  ) GROUP BY uid, grp
) GROUP BY uid  

您可以使用您的问题中的样本数据进行测试、播放上述任何操作，如下例所示

#standardSQL
WITH `project.dataset.table` AS (
  SELECT 'u12345' uid, 'worksite_1' worksite_id, DATE '2019-01-01' ts UNION ALL
  SELECT 'u12345', 'worksite_1', '2019-01-02' UNION ALL
  SELECT 'u12345', 'worksite_1', '2019-01-03' UNION ALL
  SELECT 'u12345', 'worksite_1', '2019-01-04' UNION ALL
  SELECT 'u12345', 'worksite_1', '2019-01-06' UNION ALL
  SELECT 'u1', 'worksite_1', '2019-01-01' UNION ALL
  SELECT 'u1', 'worksite_1', '2019-01-02' UNION ALL
  SELECT 'u1', 'worksite_1', '2019-01-05' UNION ALL
  SELECT 'u1', 'worksite_1', '2019-01-06' 
)
SELECT uid, MAX(consecuitive_days) max_consecuitive_days
FROM (
  SELECT uid, grp, COUNT(1) consecuitive_days
  FROM (
    SELECT uid, 
      COUNTIF(step > 1) OVER(PARTITION BY uid ORDER BY ts) grp
    FROM (
      SELECT uid, ts, 
        DATE_DIFF(ts, LAG(ts) OVER(PARTITION BY uid ORDER BY ts), DAY) step 
      FROM `project.dataset.table`
    )
  ) GROUP BY uid, grp
) GROUP BY uid   

结果：

Row uid     max_consecuitive_days    
1   u12345  4    
2   u1      2    

Answer 2

这适合你吗？

set @gr=1;
select uid, max(cnt) max_cnt from (
  select uid, grp, count(*) cnt from (
    select uid, 
      case when ifnull(DATE_ADD(oldDate, INTERVAL 1 DAY), timestamp)= timestamp then 
        @gr 
      else  
        @gr := @gr +1 
      end grp
    from
    (
        SELECT    
          uid, 
          timestamp, 
          lag(timestamp) over (partition by uid order by timestamp asc) as oldDate
        FROM      OnSite
    ) t
  )t2
  group by uid, grp
  )t3
group by uid

结果

| uid    | max_cnt |
| ------ | ------- |
| u1     | 2       |
| u12345 | 4       |

DB Fiddle