将SQL Server中的文本值从UTF8转换为ISO 8859-1

Question

我在SQL Server中有一个使用utf8 SQL_Latin1_General_CP1_CI_AS编码的列.如何转换和保存ISO 8859-1编码的文本？我想在SQL Server的查询中做一些事情.有小费吗？

Olá.Gostei做jogo.Quando"baixei"até©achei quenão iria curtir muito

Answer 1

我编写了一个函数来修复存储在varchar字段中的UTF-8文本.

要检查固定值,您可以像这样使用它:

CREATE TABLE #Table1 (Column1 varchar(max))

INSERT #Table1
VALUES ('Olá. Gostei do jogo. Quando "baixei" até achei que não iria curtir muito')

SELECT *, NewColumn1 = dbo.DecodeUTF8String(Column1)
FROM Table1
WHERE Column1 <> dbo.DecodeUTF8String(Column1)

输出:

Column1
-------------------------------
Olá. Gostei do jogo. Quando "baixei" até achei que não iria curtir muito

NewColumn1
-------------------------------
Olá. Gostei do jogo. Quando "baixei" até achei que não iria curtir muito

代码:

CREATE FUNCTION dbo.DecodeUTF8String (@value varchar(max))
RETURNS nvarchar(max)
AS
BEGIN
    -- Transforms a UTF-8 encoded varchar string into Unicode
    -- By Anthony Faull 2014-07-31
    DECLARE @result nvarchar(max);

    -- If ASCII or null there's no work to do
    IF (@value IS NULL
        OR @value NOT LIKE '%[^ -~]%' COLLATE Latin1_General_BIN
    )
        RETURN @value;

    -- Generate all integers from 1 to the length of string
    WITH e0(n) AS (SELECT TOP(POWER(2,POWER(2,0))) NULL FROM (VALUES (NULL),(NULL)) e(n))
        , e1(n) AS (SELECT TOP(POWER(2,POWER(2,1))) NULL FROM e0 CROSS JOIN e0 e)
        , e2(n) AS (SELECT TOP(POWER(2,POWER(2,2))) NULL FROM e1 CROSS JOIN e1 e)
        , e3(n) AS (SELECT TOP(POWER(2,POWER(2,3))) NULL FROM e2 CROSS JOIN e2 e)
        , e4(n) AS (SELECT TOP(POWER(2,POWER(2,4))) NULL FROM e3 CROSS JOIN e3 e)
        , e5(n) AS (SELECT TOP(POWER(2.,POWER(2,5)-1)-1) NULL FROM e4 CROSS JOIN e4 e)
    , numbers(position) AS
    (
        SELECT TOP(DATALENGTH(@value)) ROW_NUMBER() OVER (ORDER BY (SELECT NULL))
        FROM e5
    )
    -- UTF-8 Algorithm (http://en.wikipedia.org/wiki/UTF-8)
    -- For each octet, count the high-order one bits, and extract the data bits.
    , octets AS
    (
        SELECT position, highorderones, partialcodepoint
        FROM numbers a
        -- Split UTF8 string into rows of one octet each.
        CROSS APPLY (SELECT octet = ASCII(SUBSTRING(@value, position, 1))) b
        -- Count the number of leading one bits
        CROSS APPLY (SELECT highorderones = 8 - FLOOR(LOG( ~CONVERT(tinyint, octet) * 2 + 1)/LOG(2))) c
        CROSS APPLY (SELECT databits = 7 - highorderones) d
        CROSS APPLY (SELECT partialcodepoint = octet % POWER(2, databits)) e
    )
    -- Compute the Unicode codepoint for each sequence of 1 to 4 bytes
    , codepoints AS
    (
        SELECT position, codepoint
        FROM
        (
            -- Get the starting octect for each sequence (i.e. exclude the continuation bytes)
            SELECT position, highorderones, partialcodepoint
            FROM octets
            WHERE highorderones <> 1
        ) lead
        CROSS APPLY (SELECT sequencelength = CASE WHEN highorderones in (1,2,3,4) THEN highorderones ELSE 1 END) b
        CROSS APPLY (SELECT endposition = position + sequencelength - 1) c
        CROSS APPLY
        (
            -- Compute the codepoint of a single UTF-8 sequence
            SELECT codepoint = SUM(POWER(2, shiftleft) * partialcodepoint)
            FROM octets
            CROSS APPLY (SELECT shiftleft = 6 * (endposition - position)) b
            WHERE position BETWEEN lead.position AND endposition
        ) d
    )
    -- Concatenate the codepoints into a Unicode string
    SELECT @result = CONVERT(xml,
        (
            SELECT NCHAR(codepoint)
            FROM codepoints
            ORDER BY position
            FOR XML PATH('')
        )).value('.', 'nvarchar(max)');

    RETURN @result;
END
GO

Answer 2

Jason Penny还编写了一个 SQL 函数来将 UTF-8 转换为 Unicode（MIT 许可证），它为我提供了一个简单的例子：

CREATE FUNCTION dbo.UTF8_TO_NVARCHAR(@in VarChar(MAX))
   RETURNS NVarChar(MAX)
AS
BEGIN
   DECLARE @out NVarChar(MAX), @i int, @c int, @c2 int, @c3 int, @nc int

   SELECT @i = 1, @out = ''

   WHILE (@i <= Len(@in))
   BEGIN
      SET @c = Ascii(SubString(@in, @i, 1))

      IF (@c < 128)
      BEGIN
         SET @nc = @c
         SET @i = @i + 1
      END
      ELSE IF (@c > 191 AND @c < 224)
      BEGIN
         SET @c2 = Ascii(SubString(@in, @i + 1, 1))

         SET @nc = (((@c & 31) * 64 /* << 6 */) | (@c2 & 63))
         SET @i = @i + 2
      END
      ELSE
      BEGIN
         SET @c2 = Ascii(SubString(@in, @i + 1, 1))
         SET @c3 = Ascii(SubString(@in, @i + 2, 1))

         SET @nc = (((@c & 15) * 4096 /* << 12 */) | ((@c2 & 63) * 64 /* << 6 */) | (@c3 & 63))
         SET @i = @i + 3
      END

      SET @out = @out + NChar(@nc)
   END
   RETURN @out
END
GO

Anthony 打勾的答案对我来说“看起来”更好，但如果进行转换并调查任何差异，也许可以同时运行？！

我们还使用下面非常丑陋的代码来检测 BMP 页​​面 unicode 字符，这些字符被编码为 UTF-8，然后从 varchar 转换为 nvarchar 字段，可以转换为 UCS-16。

LIKE (N'%[' + CONVERT(NVARCHAR,(CHAR(192))) + CONVERT(NVARCHAR,(CHAR(193))) + CONVERT(NVARCHAR,(CHAR(194))) + CONVERT(NVARCHAR,(CHAR(195))) + CONVERT(NVARCHAR,(CHAR(196))) + CONVERT(NVARCHAR,(CHAR(197))) + CONVERT(NVARCHAR,(CHAR(198))) + CONVERT(NVARCHAR,(CHAR(199))) + CONVERT(NVARCHAR,(CHAR(200))) + CONVERT(NVARCHAR,(CHAR(201))) + CONVERT(NVARCHAR,(CHAR(202))) + CONVERT(NVARCHAR,(CHAR(203))) + CONVERT(NVARCHAR,(CHAR(204))) + CONVERT(NVARCHAR,(CHAR(205))) + CONVERT(NVARCHAR,(CHAR(206))) + CONVERT(NVARCHAR,(CHAR(207))) + CONVERT(NVARCHAR,(CHAR(208))) + CONVERT(NVARCHAR,(CHAR(209))) + CONVERT(NVARCHAR,(CHAR(210))) + CONVERT(NVARCHAR,(CHAR(211))) + CONVERT(NVARCHAR,(CHAR(212))) + CONVERT(NVARCHAR,(CHAR(213))) + CONVERT(NVARCHAR,(CHAR(214))) + CONVERT(NVARCHAR,(CHAR(215))) + CONVERT(NVARCHAR,(CHAR(216))) + CONVERT(NVARCHAR,(CHAR(217))) + CONVERT(NVARCHAR,(CHAR(218))) + CONVERT(NVARCHAR,(CHAR(219))) + CONVERT(NVARCHAR,(CHAR(220))) + CONVERT(NVARCHAR,(CHAR(221))) + CONVERT(NVARCHAR,(CHAR(222))) + CONVERT(NVARCHAR,(CHAR(223))) + CONVERT(NVARCHAR,(CHAR(224))) + CONVERT(NVARCHAR,(CHAR(225))) + CONVERT(NVARCHAR,(CHAR(226))) + CONVERT(NVARCHAR,(CHAR(227))) + CONVERT(NVARCHAR,(CHAR(228))) + CONVERT(NVARCHAR,(CHAR(229))) + CONVERT(NVARCHAR,(CHAR(230))) + CONVERT(NVARCHAR,(CHAR(231))) + CONVERT(NVARCHAR,(CHAR(232))) + CONVERT(NVARCHAR,(CHAR(233))) + CONVERT(NVARCHAR,(CHAR(234))) + CONVERT(NVARCHAR,(CHAR(235))) + CONVERT(NVARCHAR,(CHAR(236))) + CONVERT(NVARCHAR,(CHAR(237))) + CONVERT(NVARCHAR,(CHAR(238))) + CONVERT(NVARCHAR,(CHAR(239)))
    + N'][' + CONVERT(NVARCHAR,(CHAR(128))) + CONVERT(NVARCHAR,(CHAR(129))) + CONVERT(NVARCHAR,(CHAR(130))) + CONVERT(NVARCHAR,(CHAR(131))) + CONVERT(NVARCHAR,(CHAR(132))) + CONVERT(NVARCHAR,(CHAR(133))) + CONVERT(NVARCHAR,(CHAR(134))) + CONVERT(NVARCHAR,(CHAR(135))) + CONVERT(NVARCHAR,(CHAR(136))) + CONVERT(NVARCHAR,(CHAR(137))) + CONVERT(NVARCHAR,(CHAR(138))) + CONVERT(NVARCHAR,(CHAR(139))) + CONVERT(NVARCHAR,(CHAR(140))) + CONVERT(NVARCHAR,(CHAR(141))) + CONVERT(NVARCHAR,(CHAR(142))) + CONVERT(NVARCHAR,(CHAR(143))) + CONVERT(NVARCHAR,(CHAR(144))) + CONVERT(NVARCHAR,(CHAR(145))) + CONVERT(NVARCHAR,(CHAR(146))) + CONVERT(NVARCHAR,(CHAR(147))) + CONVERT(NVARCHAR,(CHAR(148))) + CONVERT(NVARCHAR,(CHAR(149))) + CONVERT(NVARCHAR,(CHAR(150))) + CONVERT(NVARCHAR,(CHAR(151))) + CONVERT(NVARCHAR,(CHAR(152))) + CONVERT(NVARCHAR,(CHAR(153))) + CONVERT(NVARCHAR,(CHAR(154))) + CONVERT(NVARCHAR,(CHAR(155))) + CONVERT(NVARCHAR,(CHAR(156))) + CONVERT(NVARCHAR,(CHAR(157))) + CONVERT(NVARCHAR,(CHAR(158))) + CONVERT(NVARCHAR,(CHAR(159))) + CONVERT(NVARCHAR,(CHAR(160))) + CONVERT(NVARCHAR,(CHAR(161))) + CONVERT(NVARCHAR,(CHAR(162))) + CONVERT(NVARCHAR,(CHAR(163))) + CONVERT(NVARCHAR,(CHAR(164))) + CONVERT(NVARCHAR,(CHAR(165))) + CONVERT(NVARCHAR,(CHAR(166))) + CONVERT(NVARCHAR,(CHAR(167))) + CONVERT(NVARCHAR,(CHAR(168))) + CONVERT(NVARCHAR,(CHAR(169))) + CONVERT(NVARCHAR,(CHAR(170))) + CONVERT(NVARCHAR,(CHAR(171))) + CONVERT(NVARCHAR,(CHAR(172))) + CONVERT(NVARCHAR,(CHAR(173))) + CONVERT(NVARCHAR,(CHAR(174))) + CONVERT(NVARCHAR,(CHAR(175))) + CONVERT(NVARCHAR,(CHAR(176))) + CONVERT(NVARCHAR,(CHAR(177))) + CONVERT(NVARCHAR,(CHAR(178))) + CONVERT(NVARCHAR,(CHAR(179))) + CONVERT(NVARCHAR,(CHAR(180))) + CONVERT(NVARCHAR,(CHAR(181))) + CONVERT(NVARCHAR,(CHAR(182))) + CONVERT(NVARCHAR,(CHAR(183))) + CONVERT(NVARCHAR,(CHAR(184))) + CONVERT(NVARCHAR,(CHAR(185))) + CONVERT(NVARCHAR,(CHAR(186))) + CONVERT(NVARCHAR,(CHAR(187))) + CONVERT(NVARCHAR,(CHAR(188))) + CONVERT(NVARCHAR,(CHAR(189))) + CONVERT(NVARCHAR,(CHAR(190))) + CONVERT(NVARCHAR,(CHAR(191)))
    + N']%') COLLATE Latin1_General_BIN

以上：

• 检测编码 U+0080 到 U+FFFF 的多字节序列（U+0080 到 U+07FF 被编码为 110xxxxx 10xxxxxx，U+0800 到 U+FFFF 被编码为 1110xxxx 10xxxxxx 10xxxxxx）
• 即它检测到十六进制字节 0xC0 到 0xEF 后跟十六进制字节 0x80 到 0xBF
• 忽略 ASCII 控制字符 U+0000 到 U+001F
• 忽略已经正确编码为 un​​icode >= U+0100（即不是 UTF-8）的字符
• 忽略 unicode 字符 U+0080 到 U+00FF，如果它们看起来不是 UTF-8 序列的一部分，例如“coöperatief”。
• 由于潜在的整理问题，对于 X=0x80 到 Y=0xBF 不使用 LIKE "%[XY]"
• 使用 CONVERT(VARCHAR,CHAR(X)) 而不是 NCHAR 因为我们在将 NCHAR 转换为错误值（对于某些值）时遇到了问题。
• 忽略大于 U+FFFF 的 UTF 字符（4 到 6 字节序列，第一个字节是十六进制 0xF0 到 0xFD）