有什么方法可以不在1次查询中多次扫描全表

Question

我想询问有什么想法可以避免扫描整个表超过 1 次。 我的问题是我需要统计公司每个部门每年的招聘和离职人数。该数据集为我提供了有关招聘日期和离职日期的数据。我当前的查询通过使用 2 个 CTE 来解决这个问题，1 个用于招聘计数，2 个用于离职计数（每个部门每年都在两个 CTE 中），然后将它们连接起来以显示最终结果。但是，我注意到这种方式会强制查询扫描整个表两次（第一次用于计算总招聘量，第二次用于计算总离职量）。

我想问是否有其他方法可以在一次扫描中计算总招聘人数（通过计算一个部门一年内的招聘人数）和总离职人数。我认为这将有助于大幅提高性能。我使用的 DBMS 是 Postgres

这是我的询问：

WITH hiring_count AS (
    SELECT EXTRACT(YEAR FROM effective_department_from_date) AS year,
           department,  
           COUNT(*) AS total_hiring
    FROM employee_department
    JOIN departments
        ON employee_department.department_id = departments.department_id
    GROUP BY year, department
), departure_count AS (
    SELECT EXTRACT(YEAR FROM effective_department_end_date) AS year, 
           department,
           COUNT(*) AS total_departure
    FROM employee_department
    JOIN departments
        ON employee_department.department_id = departments.department_id
    GROUP BY year, department
)
SELECT hiring_count.year, 
       hiring_count.department,
       total_hiring, 
       total_departure
FROM hiring_count
JOIN departure_count
    ON hiring_count.year = departure_count.year
    AND hiring_count.department = departure_count.department

Answer 1

您可以使用

grouping sets

：演示

WITH hiring_count AS (
    SELECT EXTRACT(YEAR FROM effective_department_from_date) AS year_hired,
           EXTRACT(YEAR FROM effective_department_end_date) AS year_departed,
           department,
           d.department_id,
           COUNT(*) AS total
    FROM employee_department ed
    JOIN departments d
        using (department_id)
    GROUP BY grouping sets ( (year_hired, d.department_id)
                            ,(year_departed, d.department_id)) )
SELECT coalesce( a.year_hired
                ,b.year_departed) as year
      ,a.department
      ,a.total as total_hiring
      ,b.total as total_departure
FROM hiring_count a join hiring_count b
  using (year_departed,department_id)
order by year,a.department_id
limit 10;

年	部门	总招聘	总出发时间
2013	部门_0	4	4
2013	部门_1	9	9
2013	部门_2	17	17
2013	部门_3	19	19
2013	部门_4	11	11
2013	部门_5	17	17
2013	部门_6	18	18
2013	部门_7	11	11
2013	部门_8	17	17
2013	部门_9	15	15