基于WORKS_ON复合主键表,统计各工人的同项目其他工人数
解决WORKS_ON表的工人项目匹配计数问题
当然可行!这个需求核心是判断两个工人的项目集合完全相等,下面给你两种实用的实现思路,第一种效率更高,适合大多数场景:
方法一:生成项目签名统计(推荐)
这个方法的核心是给每个工人的项目集合生成一个唯一的排序签名——把参与的项目号按固定顺序拼接成字符串,这样项目完全相同的工人会有一模一样的签名,之后只要统计每个签名的总人数,再减去工人自己就是结果。
MySQL 实现代码
WITH worker_project_signatures AS ( SELECT ID, GROUP_CONCAT(ProjectNum ORDER BY ProjectNum SEPARATOR ',') AS project_signature FROM WORKS_ON GROUP BY ID ), signature_counts AS ( SELECT project_signature, COUNT(*) AS total_workers FROM worker_project_signatures GROUP BY project_signature ) SELECT wps.ID, CASE WHEN sc.total_workers > 1 THEN sc.total_workers - 1 ELSE 0 END AS same_project_count FROM worker_project_signatures wps JOIN signature_counts sc ON wps.project_signature = sc.project_signature ORDER BY wps.ID;
PostgreSQL/SQL Server 实现代码
这两个数据库用STRING_AGG替代MySQL的GROUP_CONCAT:
WITH worker_project_signatures AS ( SELECT ID, STRING_AGG(CAST(ProjectNum AS VARCHAR), ',' ORDER BY ProjectNum) AS project_signature FROM WORKS_ON GROUP BY ID ), signature_counts AS ( SELECT project_signature, COUNT(*) AS total_workers FROM worker_project_signatures GROUP BY project_signature ) SELECT wps.ID, GREATEST(sc.total_workers - 1, 0) AS same_project_count FROM worker_project_signatures wps JOIN signature_counts sc ON wps.project_signature = sc.project_signature ORDER BY wps.ID;
代码说明
- 第一个CTE
worker_project_signatures:给每个工人生成排序后的项目拼接字符串,确保项目顺序不影响签名(比如项目2、3和3、2会生成相同的签名)。 - 第二个CTE
signature_counts:统计每个签名对应的工人总数。 - 最后关联两个CTE,每个工人的相同项目人数就是总人数减1(排除自己),如果只有自己一个人,返回0。
方法二:集合直接比较(适合小数据集)
如果你的数据量不大,也可以用集合的EXCEPT操作直接判断两个工人的项目是否完全一致,不过这种方法效率会低一些:
SELECT w1.ID, COUNT(DISTINCT w2.ID) AS same_project_count FROM WORKS_ON w1 LEFT JOIN WORKS_ON w2 ON w1.ID != w2.ID GROUP BY w1.ID HAVING -- w1的所有项目w2都包含 NOT EXISTS ( SELECT ProjectNum FROM WORKS_ON WHERE ID = w1.ID EXCEPT SELECT ProjectNum FROM WORKS_ON WHERE ID = w2.ID ) AND -- w2的所有项目w1都包含 NOT EXISTS ( SELECT ProjectNum FROM WORKS_ON WHERE ID = w2.ID EXCEPT SELECT ProjectNum FROM WORKS_ON WHERE ID = w1.ID );
用你举的例子测试:ID1和ID3的项目签名都是2,3,总人数是2,所以各自的计数是1;ID2的签名是2,总人数是1,计数是0,完全符合预期。
内容的提问来源于stack exchange,提问作者H. Wilde
相关产品推荐
相关产品推荐

