7.6.1 生成自增值

需求：通過SQL生成一個1到1000條記錄.

解決方法:通過CTE的遞迴來實現該需求.

SQLServer:

DECLARE @startINT, @endINT
SELECT @start=1, @end=1000
;WITH NumberSequence( Number) AS
(
    SELECT @start ASNumber
       UNION ALL
    SELECT Number + 1
       FROM NumberSequence
       WHERE Number <@end
)
SELECT * FROM NumberSequence OPTION (MaxRecursion 1000)
 

執行結果:

Oracle:

WITH t(num) AS (
  SELECT 1 FROM DUAL
  UNION ALL
  SELECT t.num+1
  FROM t WHERE t.num<100
  )
SELECT * FROM t;

執行結果:

Mysql(8.0及以上版本):

WITH RECURSIVE cte (num) AS
(
  SELECT 1
  UNION ALL
  SELECT num+1 FROM cte WHERE num <100
)
SELECT * FROM cte;

執行結果:

7.6.2 遍歷字串裡的每個值

需求：打印出ename為’King’的名字裡每一個字母,每個字母佔一行.

解決方法:通過自增表和emp表先cross join(笛卡爾積),然後再通過ename的len(ename的長度)進行過濾,最終得到顯示每個字母的結果.

SQLServer:

SELECT SUBSTRING (e.ENAME,seq.pos,1) AS ename_Split
FROM (SELECT ENAME FROM emp WHERE ename= 'KING' ) e,
(SELECT number AS pos FROM master.[dbo].[spt_values] WHERE type = 'P' AND number>0)seq
WHERE seq.pos <= LEN(e.ename)
 

執行結果:

注:

1: 這裡master.[dbo].[spt_values]是一張特殊的系統檢視,裡面存了從0到2047總2048條自增序列.

2: 如果不明白,可以分段來看.

Step1:

SELECT e.*,seq.* FROM
(SELECT ENAME FROM emp WHERE ename= 'KING' ) e,
(SELECT number AS pos FROM master.[dbo].[spt_values] WHERE type = 'P' AND number>0)seq

執行結果:

Sept2:

SELECT SUBSTRING(e.ENAME,seq.pos,1)AS ename_Split
FROM .. e, ..seq
WHERE seq.pos<=LEN(e.ename)

這裡通過SUBSTRNG函式,每次的開始位置不斷調整,每次僅取一個字元,再通過LEN函式過濾.所以得到最終結果.如果不熟悉SUBSTRING的語法,這裡簡單介紹下.

SUBSTRING ( expression, start, length )

1) 引數expression是要擷取的原始字串,比如這裡的” KING”

2) 引數start是要擷取的位置,比如從第2個位置開始,那應該從” I”往後數.

3) 引數length是要擷取的長度,沿用上一行的例子,如果長度定義為2,則最終擷取字串是”IN”

Oracle:

WITH t(num) AS (
  SELECT 1 FROM DUAL
  UNION ALL
  SELECT t.num+1
  FROM t WHEREt.num<100
  )
SELECT SUBSTR(e.ENAME,seq.num,1) AS ename_Split FROM
(SELECT ENAME FROM emp WHERE ename ='KING' ) e,
(SELECT num FROM t)seq
WHERE seq.num <= LENGTH(e.ename)

 Mysql 8.0:

WITH RECURSIVE cte (num)AS
(
  SELECT 1
  UNION ALL
  SELECT num+1 FROM cte WHERE num <100
)
SELECT SUBSTRING(e.ENAME,seq.num,1) AS ename_Split FROM
(SELECT ENAME FROM emp WHERE ename ='KING' ) e,
(SELECT num FROM cte)seq
WHERE seq.num <= LENGTH(e.ename)

7.6.3 處理含引號的字串

需求：往dept表裡插入dname為Test’s,loc為Beijing,deptno為100的資料.

解決方法:這裡有位引號是特殊符號,所以需要特殊處理,比如如果雙引號包裹起來.

Mysql:

INSERT INTO dept VALUES(100,'Test\'s','Beijing');

Sql Server:

BEGIN TRAN
   SET IDENTITY_INSERTdeptON;
   GO
   INSERT INTO dept(deptno,dname,loc) VALUES (100,'Test''s','Beijing');
   SELECT * FROM dept WHERE deptno=100;
   SET IDENTITY_INSERTdeptOFF;
   GO
ROLLBACK TRAN

執行結果:

7.6.4 計算某個字元出現的次數

需求：查詢emp表emptno是7499的使用者的job裡S出現的次數.

解決方法:這裡length(len)結合replace函式算出字串出現的次數.

Sql Server:

SELECT empno,job,(LEN(JOB) - LEN(REPLACE(JOB,'S','')))/LEN('S') AS StrFreq
FROM emp WHERE empno=7499;

Mysql:

SELECT empno,job,ROUND((LENGTH(JOB) - LENGTH(REPLACE(JOB,'S','')))/LENGTH('S')) AS StrFreq
FROM emp WHERE empno=7499;

注:這裡除以LENGTH('S')是為了考慮傳入的字串是2位以及以上的情況,比如’SS’.

7.6.5 字串裡過濾不需要的字元

需求：過濾tmp_v視圖裡含數字的部分. 其中tmp_v檢視的data欄位的定義是emp表的ename欄位拼接空格和deptno欄位。

解決方法:這裡通過translate函式對含數字的部分進行替換.

SQL Server:

create view tmp_v
AS SELECT ename+' '+cast(deptno as varchar)as data
from emp

SELECT data,replace(dbo.translate(data,'0123456789','@@@@@@@@@@'),'@','') as ename FROM tmp_v
order by replace(dbo.translate(data,'0123456789','@@@@@@@@@@'),'@','') desc

注：這裡需要參考之前章節裡translate函式的實現.

Mysql：

create view tmp_v
AS SELECT CONCAT(ename,' ',deptno) as data
from emp
SELECT data,replace(translate(data,'0123456789','@@@@@@@@@@'),'@','') as ename
FROM tmp_v
order by replace(translate(data,'0123456789','@@@@@@@@@@'),'@','') desc

Oracle：

7.6.6 拆分字串裡的字元和數字

需求：過濾tmp_v視圖裡data欄位拆分會原來的ename和deptno兩個欄位.

解決方法:這裡通過translate、replace、repeate（replicate、rpad）函式對含數字的部分進行替換.

SQL Server：

SELECT data,replace(dbo.translate(data,'0123456789',REPLICATE('@',10)),'@','') as ename,
replace(dbo.translate(data,'ABCDEFGHIJKLMNOPQRSTUVWXYZ',REPLICATE('@',26)),'@','') as deptno
FROM tmp_v
order by replace(dbo.translate(data,'0123456789',REPLICATE('@',10)),'@','') desc

Mysql:

SELECT data,replace(translate(data,'0123456789',REPEAT('@',10)),'@','') as ename,
replace(translate(data,'ABCDEFGHIJKLMNOPQRSTUVWXYZ',REPEAT('@',26)),'@','') as deptno
FROM tmp_v
order by replace(translate(data,'0123456789',REPEAT('@',10)),'@','') desc

結果同上

Oracle：

7.6.7 判斷字串是字串數字型

需求：檢索temp_strdata表的欄位data是字串數字型別的記錄這裡如果都是字串或者數字的也符合條件.

解決方法:這裡通過translate、replace、repeate（replicate、rpad）函式對含數字的部分進行替換.

SqlServer：

CREATE TABLE temp_str(data VARCHAR(1000));
INSERT INTO temp_str VALUES('SMITH20');
INSERT INTO temp_str VALUES('JONES30');
INSERT INTO temp_str VALUES('Jim#40');
INSERT INTO temp_str VALUES('50$Tom');
INSERT INTO temp_str VALUES('60:Mike');
INSERT INTO temp_str VALUES('70Cruz');
INSERT INTO temp_str VALUES('Jack');
INSERT INTO temp_str VALUES('J8oh0n');

SELECT data --,dbo.translate(UPPER(data),'0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ',REPLICATE('a',36)) as trans
FROM temp_str
WHERE dbo.translate(UPPER(data),'0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ',REPLICATE('a',36))=REPLICATE('a',LEN(data))

或者用如下方法,思路通過判斷擷取的每個字元的ASCII值來判斷區間是否落在[48,57],[97,122],落在的為0,反之為1,然後再通過對該Flag進行sum來判斷,如果等於0則說明數字字元型的.

SELECT data -- ,SUM(Flag)
FROM
(
SELECT LOWER(data) as data,SUBSTRING(e.data,seq.pos,1)AS ename_Split,ascii(SUBSTRING(LOWER(e.data),seq.pos,1)) AS asci,
CASE WHEN ascii(SUBSTRING(LOWER(e.data),seq.pos,1))>= 48 AND ascii(SUBSTRING(LOWER(e.data),seq.pos,1))<=57 THEN 0
WHEN ascii(SUBSTRING(LOWER(e.data),seq.pos,1))>= 97 AND ascii(SUBSTRING(LOWER(e.data),seq.pos,1))<=122 THEN 0 ELSE 1 END AS Flag
FROM temp_str e,
(SELECT number AS pos FROM master.[dbo].[spt_values]WHERE type = 'P' AND number>0) seq
WHERE seq.pos<= LEN(e.data)
-- ORDER BY data
)A
GROUP BY data
HAVING SUM(Flag) = 0

Mysql:

SELECT data--,translate(UPPER(data),'0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ',REPEAT('a',36)) AS trans
FROM temp_str
WHERE translate(UPPER(data),'0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ',REPEAT('a',36))=REPEAT('a',LENGTH(data))

-- 利用正則表示式來匹配
SELECT data
FROM temp_str
WHERE data regexp'[^0-9A-Za-z]'=0
 
-- 待驗證Mysql8

WITH RECURSIVE cte (num) AS
(
  SELECT 1
  UNION ALL
  SELECT num+1 FROM cte WHERE num <100
)
SELECT data-- ,SUM(Flag)
FROM
(
SELECT LOWER(data)as data,SUBSTRING(e.data,seq.pos,1) AS ename_Split,ascii(SUBSTRING(LOWER(e.data),seq.pos,1)) AS asci,

CASE WHEN ascii(SUBSTRING(LOWER(e.data),seq.pos,1))>=48 AND ascii(SUBSTRING(LOWER(e.data),seq.pos,1))<=57 THEN 0

WHEN ascii(SUBSTRING(LOWER(e.data),seq.pos,1))>=97 AND ascii(SUBSTRING(LOWER(e.data),seq.pos,1))<=122 THEN 0 ELSE 1 END AS Flag
FROM temp_str e,
(SELECT num FROM cte) seq
WHERE seq.pos<=LEN(e.data)
-- ORDER BY data
)A
GROUP BY data
HAVING SUM(Flag)=0

Oracle:

7.6.8 判斷字串含有漢字

需求：檢索含有漢字的字串.

解決方法:這裡通過函式CHAR_LENGTH對比LENGTH進行對比來判斷.

Mysql:

SELECT data,LENGTH(data) AS Len_data,CHAR_LENGTH(data) AS CharLen_Data,HEX(data) AS HexData FROM
(
SELECT 'Hello,World,SQL'AS data
    UNION ALL
SELECT 'Data,Arithmetic' AS data
    UNION ALL
SELECT 'Science中國' AS data
    UNION ALL
SELECT '上S海H' AS data   
)A
WHERE LENGTH(data) <> CHAR_LENGTH(data)

執行結果:

注:

LENGTH() returnsthe length of the string measured in bytes. 
CHAR_LENGTH() returns the length of the string measured incharacters.

LENGTH:是計算位元組的長度.一個漢字是算三個字元,一個數字或字母算一個字元

CHAR_LENGTH:漢字、數字、字母都算是一個字元

或者通過字串的十六進位制並結合REGEXP來判斷.

SELECT data,HEX(data) AS HexData FROM
(
SELECT 'Hello,World,SQL' AS data
    UNION ALL
SELECT 'Data,Arithmetic'AS data
    UNION ALL
SELECT 'Science中國' AS data
    UNION ALL
SELECT '上S海H' AS data   
)A
WHERE HEX(data) REGEXP'^(..)*(E[4-9])'

執行結果:

SQL Server:

SELECT data FROM
(
SELECT  'Hello,World,SQL' AS data
    UNION ALL
SELECT  'Data,Arithmetic' AS data
    UNION ALL
SELECT  'Science中國' AS data
    UNION ALL
SELECT  '上S海H' AS data   
)A
WHERE data LIKE  '%[吖-座]%'

-- 或者利用PATINDEX函式進行判斷
SELECT data  FROM
( SELECT 'Hello,World,SQL' AS data
    UNION ALL
SELECT 'Data,Arithmetic' AS data
    UNION ALL
SELECT 'Science中國' AS data
    UNION ALL
SELECT '上S海H' AS data  
)A
WHERE PATINDEX ('%[吖-座]%',data) > 0

執行結果:

Oracle:

7.6.9 合併多行到一行

需求：將emp表裡deptno相同的ename以逗號拼接在一起.

解決方法:這裡通過字串合併函式完成該效果.如group_concat

SQL Server:

SELECT DISTINCT
     deptno
    , STUFF((
       SELECT N', '+ CAST([ename]ASVARCHAR(255))
       FROM emp e1
       WHERE e1.deptno= e2.deptno
       FOR XML PATH ('')), 1, 2,'')AS StrCombine
FROM emp e2

執行結果:

或者藉助CTE:

WITH x(deptno,cnt,list,empno,len)
AS(
SELECT deptno,COUNT(*) OVER(PARTITIONBY deptno),CAST(ename AS VARCHAR(100)),empno,1
FROM emp
UNION ALL
SELECT x.deptno,x.cnt,CAST(x.list+','+e.ename AS VARCHAR(100)),e.empno,x.len+1
FROM emp e,x
WHERE e.deptno= x.deptno
AND e.empno> x.empno
)
SELECT deptno,list AS StrCombine from X
WHERE len=cnt
ORDER BY 1

步驟分析:

Step1: 首先借助COUNT(*) OVER(PARTITION BY deptno)確定每個部門裡有多少員工

Step2: 藉助len,初始值為1,完成自增(遞迴裡 x.len+1)

Step3: 藉助x.list+','+e.ename完成ename的拼接

Step4: 藉助e.deptno = x.deptno AND e.empno >x.empno確定要遞迴的結果集

Step5: 查詢滿足條件的記錄,即按deptno分組的empno數和自增序號相同的記錄.

分步檢視結果:

1)  檢視構建的遞迴資料.這裡以depto=10的為例:

SELECT * FROM x WHERE deptno= 10

2)  不難發現,這裡高亮處的資料是我們想要的,所以通過如下方式獲取最終結果:

SELECT deptno,list AS StrCombine FROM X
WHERE len=cnt
ORDER BY 1

延展閱讀:如果僅想獲得某一個分組下的字串合併,也可以按照如下方法:

DECLARE @combinedString VARCHAR(MAX)
SELECT @combinedString = COALESCE(@combinedString+', ','')+ ename
FROM emp
WHERE deptno=10
SELECT @combinedString as StringValue

Mysql:

SELECT deptno,group_concat(ename) AS StrCombine FROM emp
GROUP BY deptno
ORDER BY emp.deptno,ename

7.6.10 對字串重新按字母排序重新組合

需求：將emp表裡ename按照字母順序重新組合生成新的字元.

解決方法:這裡通過字串合併函式或者結合substring和row_number完成該效果.

SqlServer:

WITH x(ename,ename_Split) AS
(
SELECT TOP 100000 ename,SUBSTRING(e.ENAME,seq.pos,1) AS ename_Split
           FROM (SELECT ENAME FROM emp) e,
           (SELECT number AS pos FROM master.[dbo].[spt_values] WHERE type= 'P' AND number>0) seq
           WHERE seq.pos <= LEN(e.ename)
           ORDER BY ename, ename_Split

)
 SELECT DISTINCT
     ename
    , STUFF((
       SELECT N''+ CAST(ename_SplitASVARCHAR(255))
       FROM x e1
       WHERE e1.ename= e2.ename
       FOR XML PATH ('')),1,0,'')AS StrByAlph
FROM x e2

注: 如果想在CTE裡使用[]排序,需要在查詢裡指定TOP.

訊息 1033，級別 15，狀態 1，第 67 行

除非另外還指定了 TOP、OFFSET 或 FORXML，否則，ORDER BY 子句在檢視、行內函數、派生表、子查詢和公用表表達式中無效。

執行結果:

Mysql:

SELECT ename,group_concat(SUBSTRING(e.ENAME,seq.num,1) ORDER BY SUBSTRING(e.ENAME,seq.num,1) separator '') AS StrByAlph
FROM (SELECT ENAME FROM emp) e,
    (SELECT i AS num FROM tb_incr)seq
WHERE seq.num <= LENGTH(e.ename)
GROUP BY ename
ORDER BY ename,SUBSTRING(e.ENAME,seq.num,1)

注: 這裡藉助group_concat函式裡的ORDER BY關鍵字,對已經排序的字母進行合併.

這裡tb_incr是個自增表,存放從1開始步長為1的自增序列

Oracle:

7.6.11 判斷一個字元是否是數字

需求：將臨時表裡判斷data欄位裡哪些是數字.

解決方法:這裡通過函式isnumberic或者regexp完成該效果.

Mysql:

DELIMITER $$
DROP FUNCTION IF EXISTS `IsNum` $$
CREATE FUNCTION `IsNum`(str VARCHAR(25)) RETURNS INT
BEGIN
    DECLARE iResult INT DEFAULT 0;
    IF ISNULL(str) THEN return 0;END IF;-- NULL 字串
    IF str='' THEN return 0;END IF;-- 空字串
    SELECT str REGEXP '^[0-9]*$' INTO iResult;
    IF iResult =1 THEN
       RETURN 1;
    ELSE
       RETURN 0;
    END IF;
    END $$
DELIMITER ;

或者使用正則表示式:

SELECT data FROM
(
SELECT '63' AS data
UNION ALL
SELECT '36(' AS data
UNION ALL
SELECT '3(6' AS data
UNION ALL
SELECT '(36' AS data
UNION ALL
SELECT '36$' AS data
UNION ALL
SELECT '' AS data
UNION ALL
SELECT NULL AS data
)A
-- WHERE IsNum(data) = 1
WHERE data REGEXP '^[0-9]*$'=1 AND data IS NOT NULL AND data<> '';

或者直接通過函式IsNum(data) = 1來判斷,見註釋部分.

SqlServer:

SELECT data FROM
(
SELECT '63' AS data
UNION ALL
SELECT '36(' AS data
UNION ALL
SELECT '3(6' AS data
UNION ALL
SELECT '(36' AS data
UNION ALL
SELECT '36$'AS data
UNION ALL
SELECT  '' AS data
UNION ALL
SELECT NULL AS data
)A
WHERE ISNUMERIC(data)= 1

執行結果:

7.6.12 按照指定的位置擷取字元

需求：按照逗號拆分字串,取拆分出來的第二個子串.

解決方法:這裡需要自定義函式結合substring擷取字串,以達到該效果.

SQL Server:

CREATE FUNCTION strSplitIndex 
( @str VARCHAR(1024), --要分割的字串
 @split VARCHAR(10), --分隔符
 @index INT --要取元素的位置
) 
RETURNS VARCHAR(1024) 
AS 
BEGIN 
  DECLARE @location INT 
  DECLARE @start INT 
  DECLARE @next INT 
  DECLARE @seed INT 
  SET @str=LTRIM(RTRIM(@str)) 
  SET @start=1 
  SET @next=1 
  SET @seed=LEN(@split) 
  SET @location=CHARINDEX(@split,@str) 
  WHILE @location<>0and @index>@next 
  BEGIN 
    SET @[email protected][email protected] 
    SET @location=CHARINDEX(@split,@str,@start) 
    SET @[email protected]+1 
  END 
  IF @location =0 SELECT @location =LEN(@str)+1 
  RETURN SUBSTRING(@str,@start,@[email protected]) 
END
GO
SELECT dbo.strSplitIndex(data,',',1) AS StrSplit FROM
(
   SELECT 'Hello,World,SQL' AS data
   UNION ALL
   SELECT 'Data,Arithmetic' AS data
   UNION ALL
   SELECT  'Science' AS data
 )A

執行結果:

或者藉助parsename函式:

SELECT PARSENAME(REPLACE(data,',','.'),2) AS StrSplit FROM
(

SELECT  'Hello,World,SQL' AS data
   UNION ALL
   SELECT  'Data,Arithmetic' AS data
   UNION ALL
   SELECT  'Science' AS data
)A
WHERE PARSENAME(REPLACE(data,',','.'),2)IS NOT NULL

執行結果:

Mysql:

SELECT data,SUBSTRING_INDEX(SUBSTRING_INDEX(data,',',seq.num),',',-1) AS sub,seq.num AS subStrPos
FROM
(SELECT 'Hello,World,SQL' AS data
    UNION ALL
SELECT 'Data,Arithmetic' AS data
    UNION ALL
SELECT 'Science' AS data) e,
(SELECT ias num FROM tb_incr)seq
WHERE seq.num <= LENGTH(e.data)- LENGTH(REPLACE(e.data,',',''))+1
AND seq.num=2
ORDER BY data,seq.num

執行結果:

步驟解析:

Step1: 首先借助自增表將data欄位裡的資料按照逗號的數目切分,如果有2個逗號,則會切分成3部分

Step2: 藉助SUBSTRING_INDEX函式擷取逗號所在位置的子串,這裡鑑於SUBSTRING_INDEX的第三個引數的意義是子串累加,所以又套了個SUBSTRING_INDEX,第三個引數傳-1,即從右邊擷取.

Step3:藉助自增表的num,取指定分割位置的資料,這裡是2.

注: SUBSTRING_INDEX函式執行示例見下:

SELECT SUBSTRING_INDEX('Hello,World,SQL',',',1) AS SUBSTRING,1 AS pos
UNION ALL
SELECT SUBSTRING_INDEX('Hello,World,SQL',',',2) AS SUBSTRING,2 AS pos
UNION ALL
SELECT SUBSTRING_INDEX('Hello,World,SQL',',',3) AS SUBSTRING,3 AS pos

執行結果:

Oracle:

7.6.13 按照指定的分隔符擷取字元返回表形式

需求：按照逗號拆分字串,並指定返回的格式是表.

解決方法:這裡需要自定義函式結合substring擷取字串,以達到該效果.

SQLServer:

CREATE FUNCTION strSplitTable(@strNVARCHAR(2000),@splitNVARCHAR(2))
RETURNS @t TABLE(SubStr VARCHAR(1000))
AS
BEGIN
DECLARE @tmpSubStr VARCHAR(1000),@getIndexINT
SET  @getIndex=CHARINDEX(',',@str)
WHILE(@getIndex<>0)  
BEGIN
    SET @tmpSubStr=CONVERT(VARCHAR(1000),SUBSTRING(@str,1,@getIndex-1))
    INSERT INTO @t(SubStr)VALUES(@tmpSubStr)
    SET @str=STUFF(@str,1,@getIndex,'')
    SET  @getIndex=CHARINDEX(',',@str)  
END
INSERT INTO @t(SubStr)VALUES(@str)
RETURN
END
GO
SELECT * FROM strSplitTable('Hello,World,SQL',',')

執行結果:

注:  自Sql Server 2016已新增系統函式STRING_SPLIT,測試示例見下:

SELECT A.value value_A,B.value value_B 
FROM 
( SELECT value FROM STRING_SPLIT('A$B$C','$') 
)A
LEFT JOIN 
(SELECT value FROM STRING_SPLIT('A,B',',') 
)B
ON A.value = B.value

執行結果:

這裡僅支援單個字元的分隔符.

訊息 214，級別 16，狀態 11，第 12 行
Procedure expects parameter 'separator' of type 'nchar(1)/nvarchar(1)'.

Mysql:

未見該方法.