SQL - 如何从表中的 XML 字段中提取多个属性

Question

我正在尝试对一个表运行 SQL 查询，该表具有一个包含 XML 数据的字段，但该 XML 包含多个需要转换为一个字段的值。请注意，该字段是 XML 内容，但实际字段类型设置为 nvarchar(max)，而不是 xml。

编辑：版本是SQL Server 2014 Express Edition

我有一张这样的桌子： [有市场的客户名单]

我想提取同一行中的“marketCode”值（逗号分隔）：

|CompanyCode|CompanyName|MarketCode,MarketCode,MarketCode,etc.|Phone|

示例的预期输出（见截图）：

|ABC123|JOHN DEERE|AA,BB,CC,DD|555-123-000|
|DEF456|NEW HOLLLAND|AA,FF,GG,HH,KK|555-456-0000|

Answer 1

样本数据

下次请提供文字而不是图片；-)

列Markets 定义为nvarchar(max)。数据以 unicode 形式插入（带有 N 前缀）。

create table Company2
(
    Code nvarchar(6),
    Name nvarchar(11),
    Markets nvarchar(max),
    Phone nvarchar(12)
);

insert into Company2 (Code, Name, Markets, Phone) values
(N'ABC123',
 N'JOHN DEERE',
 N'<?xml version="1.0" encoding="UTF-8" standalone="yes"?>
  <license>
    <company companyCode="ABC123">
      <markets>
        <market marketCode="AA"/>
        <market marketCode="BB"/>
        <market marketCode="CC"/>
        <market marketCode="DD"/>
      </markets>
    </company>
  </license>',
 N'555-123-0000'),
(N'DEF456',
 N'NEW HOLLAND',
 N'<?xml version="1.0" encoding="UTF-8" standalone="yes"?>
  <license>
    <company companyCode="DEF456">
      <markets>
        <market marketCode="AA"/>
        <market marketCode="FF"/>
        <market marketCode="GG"/>
        <market marketCode="HH"/>
        <market marketCode="KK"/>
      </markets>
    </company>
  </license>',
 N'555-456-0000');

解决方案

无法将Markets 直接转换为XML，因为nvarchar(max) 编码与数据中的“utf-8”冲突。我将转换移动到一个单独的公用表表达式（CTE，cte_convert），从nvarchar(max) 到varchar(max) 到XML。

下一个 CTE (cte_parse) 现在可以使用 c.MarketsXML.nodes() 将 XML 中的 <market> 节点提取到新列 m.Market 中。从该列中提取@marketCode 属性作为所需值。

然后使用带有for xml path('') 的子查询来连接这些值。

with cte_convert as
(
    select c.Code, c.Name, convert(XML, convert(varchar(max), c.Markets)) as MarketsXML, c.Phone
    from Company2 c
),
cte_parse as
(
    select c.Code, c.Name, m.Market.value('@marketCode', 'nvarchar(10)') as MarketCode, c.Phone
    from cte_convert c
    outer apply c.MarketsXML.nodes('/license/company/markets/market') as m(Market)
)
select  cp.Code,
        cp.Name,
        stuff(( select ',' + cp2.MarketCode as MC
                from cte_parse cp2
                where cp2.Code = cp.Code
                for xml path(''), type).value('.', 'nvarchar(max)'),1,1,'') as MarketCodes,
        cp.Phone
from cte_parse cp
group by cp.Code, cp.Name, cp.Phone;

结果

Code   Name        MarketCodes    Phone
------ ----------- -------------- ------------
ABC123 JOHN DEERE  AA,BB,CC,DD    555-123-0000
DEF456 NEW HOLLAND AA,FF,GG,HH,KK 555-456-0000

---

此原始解决方案使用从 SQL Server 2017 开始可用的 string_agg() 函数。

样本数据

备注：Markets 列定义为XML 以反映其内容。

create table Company
(
    Code nvarchar(6),
    Name nvarchar(11),
    Markets XML,
    Phone nvarchar(12)
);

insert into Company (Code, Name, Markets, Phone) values
('ABC123',
 'JOHN DEERE',
 '<?xml version="1.0" encoding="UTF-8" standalone="yes"?>
  <license>
    <company companyCode="ABC123">
      <markets>
        <market marketCode="AA"/>
        <market marketCode="BB"/>
        <market marketCode="CC"/>
        <market marketCode="DD"/>
      </markets>
    </company>
  </license>',
  '555-123-0000'),
('DEF456',
 'NEW HOLLAND',
 '<?xml version="1.0" encoding="UTF-8" standalone="yes"?>
  <license>
    <company companyCode="DEF456">
      <markets>
        <market marketCode="AA"/>
        <market marketCode="FF"/>
        <market marketCode="GG"/>
        <market marketCode="HH"/>
        <market marketCode="KK"/>
      </markets>
    </company>
  </license>',
  '555-456-0000');

解决方案

with cte_parse as
(
    select c.Code, c.Name, m.Market.value('@marketCode', 'nvarchar(10)') as MarketCode, c.Phone
    from Company c
    outer apply c.Markets.nodes('/license/company/markets/market') as m(Market)
)
select cp.Code, cp.Name, string_agg(cp.MarketCode, ',') as MarketCodes, cp.Phone
from cte_parse cp
group by cp.Code, cp.Name, cp.Phone;

Fiddle

Answer 2

declare @t table(CompanyMarkets nvarchar(max));
insert into @t(CompanyMarkets)
values(N'<?xml version="1.0" encoding="UTF-8" standalone="yes"?>
<license>
<company companyCode="ABC123">
<markets>
<market marketCode="AA"/>
<market marketCode="BB"/>
<market marketCode="CC"/>
<market marketCode="DD"/>
</markets>
</company>
</license>');

select
    --remove encoding ...
    try_cast(replace(CompanyMarkets, 'encoding="UTF-8"', '') as xml),
    --... or transform any prolog to a weird&harmless processing instruction 
    try_cast(concat(case when CompanyMarkets like N'<?xml%' then '<?x ' end, CompanyMarkets) as xml) 
from @t;


select *,
    try_cast(concat(case when CompanyMarkets like N'<?xml%' then '<?x ' end, CompanyMarkets) as xml).query('
    for $i in (data(/license/company/markets/market/@marketCode)[1], (for $k in data(/license/company/markets/market/@marketCode)[position()>1] return concat(",", $k)))
    return text {$i}
    ').value('.', 'nvarchar(max)'),
    
    --?? no spaces in marketCodes
    replace(
    try_cast(concat(case when CompanyMarkets like N'<?xml%' then '<?x ' end, CompanyMarkets) as xml).query('data(/license/company/markets/market/@marketCode)').value('.', 'nvarchar(max)'),
    ' ', ','),
    
    stuff(
    try_cast(concat(case when CompanyMarkets like N'<?xml%' then '<?x ' end, CompanyMarkets) as xml).query('
    for $i in data(/license/company/markets/market/@marketCode) return text {concat(",", $i)}
    ').value('.', 'nvarchar(max)'), 1, 1, '')

from @t;