如何防止 SQL 连接出现重复？答案

【问题标题】：How do I prevent duplicates from a SQL join?如何防止 SQL 连接出现重复？
【发布时间】：2021-08-06 02:18:43
【问题描述】：

我有以下表格：

customer:

id	customer_number	company	firstname	lastname	account_manager_email	email	comments	terms	tax_id_number	lead_source	default_catalog	credit_limit
99453	C00123456	Serenity Inc.	Malcom	Reynolds	jim.smith@example.com	mal@example.com	The cap'n	1	NULL	NULL	12345	NULL
99468	C00123456	Serenity Inc.	Zoe	Washburne	jim.smith@example.com	zoe@example.com	NULL	1	NULL	NULL	NULL	NULL
99960	C00123456	Serenity Inc.	Hoban	Washburne	jim.smith@example.com	wash@example.com	NULL	1	NULL	NULL	NULL	NULL
100088	C00123456	Serenity Inc.	Inara	Serra	jim.smith@example.com	inara@example.com	NULL	1	NULL	NULL	12345	NULL

customer_address:

id	company	street	city	state_abbreviation	postcode	telephone	firstname	lastname	created_at
133996	Serenity, Inc	123 Any St.	Anytown	AX	12345	123-456-7890	Malcom	Reynolds	2017-05-08 12:45:53.000
134452	Serenity, Inc	123 Any St.	Anytown	AX	12345	123-456-7890	Accounts	Payable	2017-05-09 10:19:59.000
134961	Serenity, Inc	123 Any St.	Anytown	AX	12345	123-456-7890	REF	987654321	2017-05-09 10:19:59.000
134962	Serenity, Inc	123 Any St.	Anytown	AX	12345	123-456-7890	REF	192837465	2017-05-09 10:19:59.000
133995	Serenity, Inc	123 Any St.	Anytown	AX	12345	123-456-7890	Accounts	Payable	2017-05-09 10:19:59.000
133669	Serenity, Inc	123 Any St.	Anytown	AX	12345	123-456-7890	REF	123456789	2017-05-18 10:29:42.000
133667	Serenity, Inc	123 Any St.	Anytown	AX	12345	123-456-7890	Accounts	Payable	2017-05-18 07:56:45.000
133666	Serenity, Inc	123 Any St.	Anytown	AX	12345	123-456-7890	Accounts	Payable	2017-05-31 07:56:46.000
133626	Serenity, Inc	123 Any St.	Anytown	AX	12345	123-456-7890	Accounts	Payable	2017-06-16 12:45:08.000
133668	Serenity, Inc	123 Any St.	Anytown	AX	12345	123-456-7890	REF	PO	2017-06-16 12:45:08.000

我正在运行此查询以生成可用于将数据导入另一个系统的 CSV：

     SELECT '"' + CAST(c.customer_number AS VARCHAR) + '"' AS 'Customer Number',
            '"' + CAST(c.company AS VARCHAR) + '"' AS 'Company Name',
            '"' + CAST(a.street AS VARCHAR) + '"' AS 'Company Address Line 1',
            '"' + CAST(a.city AS VARCHAR) + '"' AS 'Company City',
            '"' + CAST(a.state_abbreviation AS VARCHAR) + '"' AS 'Company State',
            '"' + CAST(a.postcode AS VARCHAR) + '"' AS 'Company Zip Code',
            '"' + CAST(a.telephone AS VARCHAR) + '"' AS 'Company Phone',
            '"' + CAST(c.firstname AS VARCHAR) + '"' AS 'Contact First Name',
            '"' + CAST(c.lastname AS VARCHAR) + '"' AS 'Contact Last Name',
            '"' + CAST(c.account_manager_email AS VARCHAR) + '"' AS 'Account Manager Email',
            '"' + CAST(a.company AS VARCHAR) + '"' AS 'Contact Company Name',
            '"' + CAST(a.street AS VARCHAR) + '"' AS 'Contact Address Line 1',
            '"' + CAST(a.city AS VARCHAR) + '"' AS 'Contact City',
            '"' + CAST(a.state_abbreviation AS VARCHAR) + '"' AS 'Contact State',
            '"' + CAST(a.postcode AS VARCHAR) + '"' AS 'Contact Zip Code',
            '"' + CAST(a.telephone AS VARCHAR) + '"' AS 'Contact Phone',
            '"' + CAST(c.email AS VARCHAR) + '"' AS 'Contact Email',
            '"' + CAST(c.comments AS VARCHAR) + '"' AS 'Internal Notes',
            '"' + CAST(c.terms AS VARCHAR) + '"' AS 'Terms',
            '"' + CAST(c.tax_id_number AS VARCHAR) + '"' AS 'Tax ID (US)',
            '"' + CAST(c.lead_source AS VARCHAR) + '"' AS 'Lead Source',
            '"' + CAST(c.default_catalog AS VARCHAR) + '"' AS 'Catalog',
            '"' + CAST(c.credit_limit AS VARCHAR) + '"' AS 'Credit Limit'
       FROM customer c,
            customer_address a
      WHERE c.customer_number = 'C00123456'
        AND a.company = c.company
   ORDER BY c.customer_number,
            c.created_at;

但是，当我运行该查询时，我得到了 40 行，customer 中的每个条目有 10 行。我尝试了不同的连接类型，但结果是一样的。

其中很多是遗留数据，所以我似乎唯一能够可靠地加入的是公司名称（“Serenity, Inc.”）

我实际上需要这个输出的两个版本。第一个是每个公司的单行，其中包含customer 表中具有最旧created_at 值的条目。第二个是所有其他记录。

注意：这是在 SQL Server 2005 上（我知道...计划升级，但我必须先完成这个）

【问题讨论】：

您确定不能可靠地加入 fristname 和 lastname 吗？我回答说这两列看起来像是导致你的重复，但现在我很好奇为什么你不能使用这两列？如果你不能可靠地使用它们，你将不得不创造性地逐步按摩，以弄清楚如何正确地做到这一点。我还假设您不需要“应付账款”的记录？
您有 4 行连接到 10 行的匹配值，所以当然 4x10=40。您的Customers 表没有created_at，它在您的customers_address 中。如果你只想要一行，大概是最小客户Id？
您还尝试了哪些其他联接？可能值得学习 SQL 基础知识，例如在以下站点：w3schools.com/sql/sql_join.asp
customer_address 应该有一个 customer_id 列，它是将每个 customer_address 与一个 customer 相关联的外键。为什么不呢？
@DavidBrowne-Microsoft 我同意并希望我能得到这个问题的答案，但不幸的是，我没有。

标签： sql-server sql-server-2005

【解决方案1】：

要删除重复项，您需要枚举每一行并根据排序条件分配一个值。

您可以使用cte 轻松做到这一点 - 我相信它们在 SQL Server 2005 中可用，我当然无法检查。

with c as(
  select *, Row_Number() over(partition by customer_number order by id) rn
  from customer
),
ca as (
  select *, Row_Number() over(partition by company order by created_at) rn
  from customer_address
)
select <columns>
from c join ca on c.company=ca.company
where c.rn=1 and ca.rn=1 and c.customer_number='C00123456'

【讨论】：

【解决方案2】：

您还尝试了哪些其他联接？

这是为每个客户记录选择最近的 customer_address 记录的一种方法：

SELECT c.*, a.* FROM
customer c
LEFT JOIN 
  ( SELECT x.* FROM customer_address x
    INNER JOIN 
    (SELECT company, MAX(created_at) AS created_at FROM customer_address 
       GROUP BY company) u
    ON u.company=x.company AND u.created_at=x.created_at
  ) a
ON a.company=c.company
WHERE c.customer_number = 'C00123456';

因此，所有其他 customer_address 记录将由：

  ( SELECT x.* FROM customer_address x
    INNER JOIN 
    (SELECT company, MAX(created_at) AS created_at FROM customer_address 
       GROUP BY company) u
    ON u.company=x.company AND NOT u.created_at=x.created_at
  ) a

【讨论】：