Question

我需要将叙述字段（自由文本）拆分为多行。格式目前遵循：

Case_Reference | Narrative
```````````````|`````````````````````````````````````
XXXX/XX-123456 | [Endless_Text up to ~50k characters]

在作为文本的叙述字段中，单个条目（当各种代理人对案例做了一些事情时）从输入日期开始，后跟两个空格（即'dd/mm/yyyy '），日期的值随每个变化在同一领域内进入。

换句话说，在搜索更好的分隔符后，我可以使用的唯一一种是字符串格式，所以我需要在叙述文本中识别多个位置，其中格式（掩盖是一个更好的单词？）匹配'dd/mm/yyyy '。

我可以识别出多次出现的一致字符串没有问题，但是它确定了我在寻找的地方：

'%[0-9][0-9]/[0-9][0-9]/[0-9][0-9][0-9][0-9] %'

PATINDEX当然会返回第一次出现/位置，但据我所知，没有办法“修改”这个（即创建的函数）以便接收其余的我们可以使用CHARINDEX（因为PATINDEX没有起始位置参数）的出现/位置。

为了清楚起见，我不是在寻找代码来直接划分它，因为我需要进一步操作每个条目，所以它纯粹是我正在寻找的叙事文本中多次出现的字符串的位置。

非常感谢任何帮助。

为清楚起见，没有选择在导入前执行此操作，因此需要对此着陆数据执行此操作。

所需的输出将是

 Case_Reference1 | 1st_Position_of_Delimiter_String  
 Case_Reference1 | 2nd_Position_of_Delimiter_String  
 Case_Reference2 | 1st_Position_of_Delimiter_String  
 Case_Reference2 | 2nd_Position_of_Delimiter_String  
 Case_Reference2 | 3rd_Position_of_Delimiter_String

Answer 1

您可以使用递归CTE解决此问题

DECLARE @tbl TABLE (Case_Reference NVARCHAR(MAX),Narrative NVARCHAR(MAX));
INSERT INTO @tbl VALUES
 (N'C1',N'01/02/2000  Some text with     blanks 02/03/2000  More text 03/04/2000  An even more')
,(N'C2',N'01/02/2000  Test for C2 02/03/2000  One more for C2 03/04/2000  An even more 04/05/2000  Blah')
,(N'C3',N'01/02/2000  Test for C3 02/03/2000  One more for C3 03/04/2000  An even more')
 ;

WITH recCTE AS
(
    SELECT 1 AS Step,Case_Reference,Narrative,CAST(1 AS BIGINT) AS StartsAt,NewPos.EndsAt+10 AS EndsAt,LEN(Narrative) AS MaxLen
          ,SUBSTRING(Narrative,NewPos.EndsAt+10+1,999999) AS RestString
    FROM @tbl AS tbl
    CROSS APPLY(SELECT PATINDEX('%[0-3][0-9]/[0-1][0-9]/[1-2][0-9][0-9][0-9]  %',SUBSTRING(Narrative,12,9999999))) AS NewPos(EndsAt)

    UNION ALL

    SELECT r.Step+1,r.Case_Reference,r.Narrative,r.EndsAt+1,CASE WHEN NewPos.EndsAt>0 THEN r.EndsAt+NewPos.EndsAt+10 ELSE r.MaxLen END,r.MaxLen
          ,SUBSTRING(r.RestString,NewPos.EndsAt+10+1,999999) 
    FROM recCTE AS r
    CROSS APPLY(SELECT PATINDEX('%[0-3][0-9]/[0-1][0-9]/[1-2][0-9][0-9][0-9]  %',SUBSTRING(r.RestString,12,99999999))) AS NewPos(EndsAt)
    WHERE r.EndsAt<r.MaxLen
)
SELECT Step,Case_Reference,StartsAt,EndsAt
      ,SUBSTRING(Narrative,StartsAt,EndsAt-StartsAt+1) AS OutputString 
FROM recCTE

ORDER BY Case_Reference,Step

结果

+------+----------------+----------+--------+---------------------------------------+
| Step | Case_Reference | StartsAt | EndsAt | OutputString                          |
+------+----------------+----------+--------+---------------------------------------+
| 1    | C1             | 1        | 38     | 01/02/2000  Some text with     blanks |
+------+----------------+----------+--------+---------------------------------------+
| 2    | C1             | 39       | 60     | 02/03/2000  More text                 |
+------+----------------+----------+--------+---------------------------------------+
| 3    | C1             | 61       | 84     | 03/04/2000  An even more              |
+------+----------------+----------+--------+---------------------------------------+
| 1    | C2             | 1        | 24     | 01/02/2000  Test for C2               |
+------+----------------+----------+--------+---------------------------------------+
| 2    | C2             | 25       | 52     | 02/03/2000  One more for C2           |
+------+----------------+----------+--------+---------------------------------------+
| 3    | C2             | 53       | 77     | 03/04/2000  An even more              |
+------+----------------+----------+--------+---------------------------------------+
| 4    | C2             | 78       | 93     | 04/05/2000  Blah                      |
+------+----------------+----------+--------+---------------------------------------+
| 1    | C3             | 1        | 24     | 01/02/2000  Test for C3               |
+------+----------------+----------+--------+---------------------------------------+
| 2    | C3             | 25       | 52     | 02/03/2000  One more for C3           |
+------+----------------+----------+--------+---------------------------------------+
| 3    | C3             | 53       | 76     | 03/04/2000  An even more              |
+------+----------------+----------+--------+---------------------------------------+

Answer 2

试试这个递归cte

declare @t table 
(
    caseref varchar(20),
    narrative varchar(max)
)

insert into @t values('Case_Reference1', 'blah 10/11/2016  something 13/11/2016  something else');
insert into @t values('Case_Reference2', '11/11/2016  something 12/11/2016  something else 14/11/2016  something yet still');
insert into @t values('Case_Reference3', 'should find nothing');

with cte (caseref, pos, remainingstr) as 
(
    select caseref, 
        patindex('%[0-9][0-9]/[0-9][0-9]/[0-9][0-9][0-9][0-9]  %', narrative),
        substring(narrative, patindex('%[0-9][0-9]/[0-9][0-9]/[0-9][0-9][0-9][0-9]  %', narrative) + 12, len(narrative) - 12 - patindex('%[0-9][0-9]/[0-9][0-9]/[0-9][0-9][0-9][0-9]  %', narrative))
    from @t
    where patindex('%[0-9][0-9]/[0-9][0-9]/[0-9][0-9][0-9][0-9]  %', narrative) > 0

    union all

    select caseref,
        pos + 12 + patindex('%[0-9][0-9]/[0-9][0-9]/[0-9][0-9][0-9][0-9]  %', remainingstr),
        substring(remainingstr, patindex('%[0-9][0-9]/[0-9][0-9]/[0-9][0-9][0-9][0-9]  %', remainingstr) + 12, len(remainingstr) - 12 - patindex('%[0-9][0-9]/[0-9][0-9]/[0-9][0-9][0-9][0-9]  %', remainingstr))
    from cte
    where patindex('%[0-9][0-9]/[0-9][0-9]/[0-9][0-9][0-9][0-9]  %', remainingstr) > 0

)
select caseref, pos
from cte
order by caseref, pos

（SQL）标识字段中多次出现的字符串格式的位置

2 个答案: