最后连续相同数据的 SQL 查询

Question

我的表格中有数据

+----+------------+------+
| Id | data_date  | data |
+----+------------+------+
|  1 | 01/01/2017 |    1 |
|  1 | 01/02/2017 |    0 |
|  1 | 01/03/2017 |    0 |
|  1 | 01/04/2017 |    0 |
|  2 | 01/02/2017 |    0 |
|  2 | 01/03/2017 |    0 |
|  2 | 01/04/2017 |    1 |
|  2 | 01/05/2017 |    1 |
+----+------------+------+

数据列只有值 0 和 1

如何获取最后几天连续相同的数据类型，即每个 ID 的“0”或“1”

+----+-----------+-----------------+
| Id | data_Type | Last_Data_Count |
+----+-----------+-----------------+
|  1 |         0 |               3 |
|  2 |         1 |               2 |
+----+-----------+-----------------+

Answer 1

Oracle 查询：

SELECT id,
       MAX( last_data ) AS data_type, 
       COUNT(*) AS last_data_count
FROM   (
  SELECT d.*,
         MAX ( CASE WHEN data <> last_data THEN data_date END )
           OVER ( PARTITION BY id ) AS last_date
  FROM   (
    SELECT d.*,
           MAX( data ) KEEP ( DENSE_RANK LAST ORDER BY data_date )
             OVER ( PARTITION BY id ) AS last_data
    FROM   data d
  ) d
)
WHERE  data_date > last_date
GROUP BY id;

Answer 2

假设您需要每个id 的data 计数max(data_date)，您可以试试这个。它应该适用于几乎所有的 RDBMS。

Rexter Sample for MYSQL

select t3.id
    ,max(data) as data_type
    ,count(data) as last_data_count
    from table t3
    where exists
        (select 1 from Table t1
            where exists  
            (select 1 from table t2
             where t1.id=t2.id
             group by t2.id
             having t1.data_date=max(t2.data_date)
             ) 
     and t1.id=t3.id and t1.data=t3.data
)
group by t3.id
;

解释：内部exists 用于根据max(date) 获取每个id 的最后data_date。现在要获取计数，我使用外部 exists 来匹配 id 和 data，然后使用 group by id 来获取计数。

Answer 3

这可以很容易地在 Oracle 中通过使用Tabibitosan 然后对结果进行分组，就像这样（注意：我假设您的意思是“最后几天连续”是指连续的行，不是连续的日期）：

WITH sample_data AS (SELECT 1 ID, to_date('01/01/2017', 'dd/mm/yyyy') data_date, 1 DATA FROM dual UNION ALL
                     SELECT 1 ID, to_date('01/02/2017', 'dd/mm/yyyy') data_date, 0 DATA FROM dual UNION ALL
                     SELECT 1 ID, to_date('01/03/2017', 'dd/mm/yyyy') data_date, 0 DATA FROM dual UNION ALL
                     SELECT 1 ID, to_date('01/04/2017', 'dd/mm/yyyy') data_date, 0 DATA FROM dual UNION ALL
                     SELECT 2 ID, to_date('01/01/2017', 'dd/mm/yyyy') data_date, 0 DATA FROM dual UNION ALL
                     SELECT 2 ID, to_date('01/02/2017', 'dd/mm/yyyy') data_date, 0 DATA FROM dual UNION ALL
                     SELECT 2 ID, to_date('01/03/2017', 'dd/mm/yyyy') data_date, 1 DATA FROM dual UNION ALL
                     SELECT 2 ID, to_date('01/04/2017', 'dd/mm/yyyy') data_date, 1 DATA FROM dual UNION ALL
                     SELECT 3 ID, to_date('01/01/2017', 'dd/mm/yyyy') data_date, 0 DATA FROM dual UNION ALL
                     SELECT 3 ID, to_date('01/02/2017', 'dd/mm/yyyy') data_date, 1 DATA FROM dual UNION ALL
                     SELECT 3 ID, to_date('01/03/2017', 'dd/mm/yyyy') data_date, 0 DATA FROM dual UNION ALL
                     SELECT 3 ID, to_date('01/05/2017', 'dd/mm/yyyy') data_date, 0 DATA FROM dual)
-- end of mimicking data in a table called "sample_data"
-- see below for the main SQL query:
SELECT ID,
       DATA,
       COUNT(*) last_data_count
FROM   (SELECT ID,
               data_date,
               DATA,
               MAX(data_date) OVER (PARTITION BY ID) max_data_date,
               row_number() OVER (PARTITION BY ID ORDER BY data_date)
                 - row_number() OVER (PARTITION BY ID, DATA ORDER BY data_date) grp
        FROM   sample_data)
GROUP BY ID,
         DATA,
         grp,
         max_data_date
HAVING   max_data_date = MAX(data_date);

        ID       DATA LAST_DATA_COUNT
---------- ---------- ---------------
         1          0               3
         2          1               2
         3          0               2

tabibitosan 部分（即row_number() over (<overall set of data>) - row_number() over (<subset of data>)）为具有数据列中的相同值。

获得此信息后，很容易找到每组连续行的计数。但是，由于您在最新计数之后，我使用MAX() 分析函数来查找每个id 的最新data_date。

然后，我们可以比较每组连续行的最大 data_date，并使用它来查找与 id 的最大 data_date 匹配的连续行集（我们在 having 子句中执行此操作）。瞧！