从 SQL 中的行中搜索和提取单词

Question

我有一张下面的表格

表 A

Text,id,Cid,CName,Aid,AName
Acc.sa is very Acc.pa and Acc.ba is awesome, 1,2,AB,1,CC
Acc.aa is awesome and Acc.sas is great,2,3,CC,1,CC
Acc.ee is not only great but Acc.sew is best,4,3,FF,1,CC

它应该获取与 Acc 关联的所有单词，因此结果应该是

 Did,id,Cid,CName,Aid,AName
 Acc.sa,1,2,AB,1,CC
 Acc.pa,1,2,AB,1,CC
 Acc.ba,1,2,AB,1,CC
 Acc.aa,2,3,CC,1,CC
 Acc.sas,2,3,CC,1,CC
 Acc.ee,4,3,FF,1,CC
 Acc.sew,4,3,FF,1,CC

即每个搜索都应该有一个新行

我尝试了 CHARINDEX 和 substring，但我不确定如何在 SELECT 语句中继续使用 CHARINDEX 和 substring 任何帮助都非常感谢。

Answer 1

with base10 as (
    select n
    from (values (0), (1), (2), (3), (4), (5), (6), (7), (8), (9)) v(n)
), k as (
    select d2.n * 100 + d1.n * 10 + d0.n + 1 as n
    from base10 d0 cross join base10 d1 cross join base10 d2
)
select
    substring(a.Text, k.n, charindex(' ', a.Text, k.n) - k.n) as Did,
    Id, Cid, CName, Aid, AName
from TableA a inner join k
    on substring(a.Text, k.n, 4) = 'Acc.'
        and k.n < len(a.Text) /* not necessary but optimizer might use it??? */

http://sqlfiddle.com/#!6/a0f87/4（输出一些额外的列）

这只会处理大约一千个字符长的字符串。我怀疑这可能已经足够了，如果这很慢，您甚至可能想要缩小搜索长度。

我假设您的“Acc”后面会立即出现一个空格。值，这意味着它不会出现在行尾。必要时可以处理。

由于您看到错误，您的输入行似乎与您指定的格式不同。除了我提到的空格字符分隔符之外，我没有看到任何其他假设。

对于调试，您可以用此输出替换整个 substring() 行，以便更好地了解正在发生的事情。还要添加where 子句以将行限制为会导致错误的行：

select
    'Bad offset' as Msg,
    a.Text, k.n as StartOfAccBlick, charindex(' ', a.Text, k.n) as EndOfAccBlock
from ...    
where
    k.n - charindex(' ', a.Text, k.n) <= 0

Answer 2

xml 节点方法可以方便地将单元格值解析为这样的行。例如：

select n.value('@s[1]', 'varchar(max)'), id, Cid, CName, Aid, AName
from Table_A ta
cross apply (select convert(xml, '<x s="' + replace(replace(replace(replace(replace(replace(ta.[Text],'&','&amp;'),'>','&gt;'),'<','&lt;'),'''','&apos;'),'"','&quot;'),' ','"/><x s="') + '"/>') xval) r
cross apply r.xval.nodes('*') x(n)
where n.value('@s[1]', 'varchar(max)') like 'Acc.%'

SqlFiddle

编辑 - 转义 5 个无效的 xml 字符

Answer 3

您将需要使用递归 CTE 将每个出现的子串取出。我今天早上工作很忙，所以我无法创建与您的特定模式匹配的东西，但这是我过去做过的一个示例，用于从字符串中提取所有 PDF 文件名。所有注释掉的行都是为了显示我如何获得需要提取的 PDF 文件名的构建块。您将需要针对您的场景相应地修改模式搜索。这是SQL Fiddle。

DECLARE @f TABLE (fieldName VARCHAR(255), IDField int)
INSERT INTO @f VALUES('>>>>>>1.pdf test> >b>c>xyz.pdf bob >hello world.pdf foo >womp womp.pdf>', 1)
INSERT INTO @f VALUES('>2.pdf other unnecssary stuff > bar.pdf', 2)

; WITH cte2 AS (
   SELECT 
       IDField,
       --PATINDEX('%.pdf%', fieldName) + 3 AS PDFLocation,
       --SUBSTRING(fieldName, 1, (PATINDEX('%.pdf%', fieldName) + 3)) AS PDFSubstring,
       --REVERSE(SUBSTRING(fieldName, 1, (PATINDEX('%.pdf%', fieldName) + 3))) AS PDFSubstringReverse,
       --PATINDEX('%>%', REVERSE(SUBSTRING(fieldName, 1, (PATINDEX('%.pdf%', fieldName) + 3)))) AS ReverseSymbolLocationBeforePDF,
       --LEN(SUBSTRING(fieldName, 1, (PATINDEX('%.pdf%', fieldName) + 3))) - PATINDEX('%>%', REVERSE(SUBSTRING(fieldName, 1, (PATINDEX('%.pdf%', fieldName) + 3)))) + 2 AS SymbolLocationBeforePDF,

       CONVERT(VARCHAR(255), SUBSTRING(fieldName, 
           LEN(SUBSTRING(fieldName, 1, (PATINDEX('%.pdf%', fieldName) + 3))) - PATINDEX('%>%', REVERSE(SUBSTRING(fieldName, 1, (PATINDEX('%.pdf%', fieldName) + 3)))) + 2, 
           PATINDEX('%.pdf%', fieldName) + 3 - (LEN(SUBSTRING(fieldName, 1, (PATINDEX('%.pdf%', fieldName) + 3))) - PATINDEX('%>%', REVERSE(SUBSTRING(fieldName, 1, (PATINDEX('%.pdf%', fieldName) + 3)))) + 2) + 1
       )) AS PDFName,

       CONVERT(VARCHAR(255), STUFF(fieldName, 1, PATINDEX('%.pdf%', fieldName) + 3, '')) AS strWhatsLeft
   FROM @f

   UNION ALL

   SELECT 
       IDField,
       --PATINDEX('%.pdf%', strWhatsLeft) + 3 AS PDFLocation,
       --SUBSTRING(strWhatsLeft, 1, (PATINDEX('%.pdf%', strWhatsLeft) + 3)) AS PDFSubstring,
       --REVERSE(SUBSTRING(strWhatsLeft, 1, (PATINDEX('%.pdf%', strWhatsLeft) + 3))) AS PDFSubstringReverse,
       --PATINDEX('%>%', REVERSE(SUBSTRING(strWhatsLeft, 1, (PATINDEX('%.pdf%', strWhatsLeft) + 3)))) AS ReverseSymbolLocationBeforePDF,
       --LEN(SUBSTRING(strWhatsLeft, 1, (PATINDEX('%.pdf%', strWhatsLeft) + 3))) - PATINDEX('%>%', REVERSE(SUBSTRING(strWhatsLeft, 1, (PATINDEX('%.pdf%', strWhatsLeft) + 3)))) + 2 AS SymbolLocationBeforePDF,

       CONVERT(VARCHAR(255), SUBSTRING(strWhatsLeft, 
           LEN(SUBSTRING(strWhatsLeft, 1, (PATINDEX('%.pdf%', strWhatsLeft) + 3))) - PATINDEX('%>%', REVERSE(SUBSTRING(strWhatsLeft, 1, (PATINDEX('%.pdf%', strWhatsLeft) + 3)))) + 2, 
           PATINDEX('%.pdf%', strWhatsLeft) + 3 - (LEN(SUBSTRING(strWhatsLeft, 1, (PATINDEX('%.pdf%', strWhatsLeft) + 3))) - PATINDEX('%>%', REVERSE(SUBSTRING(strWhatsLeft, 1, (PATINDEX('%.pdf%', strWhatsLeft) + 3)))) + 2) + 1
       )) AS PDFName,

       CONVERT(VARCHAR(255), STUFF(strWhatsLeft, 1, PATINDEX('%.pdf%', strWhatsLeft) + 3, '')) AS strWhatsLeft
   FROM cte2
   WHERE strWhatsLeft LIKE '%.pdf%'
)

SELECT * FROM cte2 ORDER BY IDField