Oracle 查询查找第一行或第二行

Question

我有一组记录人员及其条款的数据。我只关心术语 201030、201040 和 201110。所有其他的都将被忽略。有的人会有几条记录，有的人只有一条。我正在写一个查询来获取每个人的第一个学期。如果第一项是 201030，它应该查看下一行。我需要 return 第二行的术语，如果它是 201040 或 201110，否则只是 return 201030。Return 第一行术语，如果它是 201040 或 201110。这是有点混乱所以我试着用不同场景的图像来。绿色框表示我应该保留的记录。

这是我目前的查询。它运行但不应用案例逻辑。它只是 return 的第一学期。我应该如何更改此查询？

select person_id, 
       min(term) as min_term,  -- including this field to verify the 1st term before changes
       case 
          when term = '201030' 
          then lead(min(term), 1, min(term)) over (order by min(term))
          else term
       end as term
from my_table 
group by person_id

Answer 1

这是一种选择。阅读代码中的注释。

SQL> with test (person_id, term) as
  2    -- sample data
  3    (select 53303, 201030 from dual union all
  4     select 53303, 201040 from dual union all
  5     select 53303, 201110 from dual union all
  6     select 53303, 201140 from dual union all
  7     --
  8     select 14627, 201030 from dual union all
  9     select 14627, 201110 from dual union all
 10     select 14627, 201510 from dual union all
 11     --
 12     select 14702, 201030 from dual union all
 13     --
 14     select 28103, 201030 from dual union all
 15     select 28103, 201230 from dual union all
 16     select 28103, 201240 from dual union all
 17     select 28103, 201310 from dual union all
 18     --
 19     select 33634, 201040 from dual union all
 20     select 33634, 201110 from dual union all
 21     select 33634, 201130 from dual union all
 22     --
 23     select 32356, 201510 from dual
 24    ),

---

 25  rns as
 26    -- select next term and row numbers
 27    (select person_id,
 28       term,
 29       lead(term) over (partition by person_id order by term) next_term,
 30       row_number() over (partition by person_id order by term) rn
 31     from test
 32    )
 33  select person_id,
 34    max(case when rn = 1 and term = 201030 then
 35               case when next_term in (201040, 201110) then next_term
 36                    else term
 37               end
 38             when rn = 1 and term in (201040, 201110) then term
 39        end) result
 40  from rns
 41  group by person_id;

 PERSON_ID     RESULT
---------- ----------
     14702     201030
     32356
     33634     201040
     14627     201110
     28103     201030
     53303     201040

6 rows selected.

SQL>

Answer 2

从 Oracle 12 开始，您可以使用 MATCH_RECOGNIZE:

SELECT *
FROM   (
  SELECT *
  FROM   my_table
  WHERE  term IN (201030, 201040, 201110)
)
MATCH_RECOGNIZE (
  PARTITION BY person_id
  ORDER     BY term_order
  ALL ROWS PER MATCH
  PATTERN ( ^ {- IS30? -} ANY_ROW )
  DEFINE
    IS30 AS term = 201030
)

在此之前，您可以使用解析函数：

SELECT person_id,
       CASE
       WHEN term = 201030
       THEN next_term
       ELSE term
       END AS term
FROM   (
  SELECT t.*,
         ROW_NUMBER() OVER (PARTITION BY person_id ORDER BY term_order) AS rn,
         LEAD(term, 1, term) OVER (PARTITION BY person_id ORDER BY term_order)
           AS next_term
  FROM   my_table t
  WHERE  term IN (201030, 201040, 201110)
)
WHERE  rn = 1;

其中，对于示例数据：

CREATE TABLE my_table (person_id, term, term_order) As
  SELECT 14627, 201030, 1 FROM DUAL UNION ALL -- Not this one
  SELECT 14627, 201110, 2 FROM DUAL UNION ALL -- This one
  SELECT 14627, 201510, 3 FROM DUAL UNION ALL -- Ignore this
  SELECT 14702, 201030, 1 FROM DUAL UNION ALL -- This one
  SELECT 28103, 201030, 1 FROM DUAL UNION ALL -- This one
  SELECT 28103, 201230, 2 FROM DUAL UNION ALL -- Ignore this
  SELECT 28103, 201240, 3 FROM DUAL UNION ALL -- Ignore this
  SELECT 28103, 201310, 4 FROM DUAL UNION ALL -- Ignore this
  SELECT 33634, 201040, 1 FROM DUAL UNION ALL -- This one
  SELECT 33634, 201110, 2 FROM DUAL UNION ALL -- Not this one
  SELECT 33634, 201130, 3 FROM DUAL UNION ALL -- Ignore this
  SELECT 32356, 201510, 1 FROM DUAL UNION ALL -- Ignore this
  SELECT 53303, 201030, 1 FROM DUAL UNION ALL -- Not this one
  SELECT 53303, 201040, 2 FROM DUAL UNION ALL -- This one.
  SELECT 53303, 201110, 3 FROM DUAL UNION ALL -- Not this one
  SELECT 53303, 201140, 4 FROM DUAL;          -- Ignore this

注意：在 SQL 中，表格是无序的。如果您希望它们按特定顺序排列，那么您需要一些东西来应用该顺序（例如 task_order 列）并使用 ORDER BY 子句。

输出：

PERSON_ID	TERM_ORDER	TERM
14627	2	201110
14702	1	201030
28103	1	201030
33634	1	201040
53303	2	201040

db<>fiddle here