如何使用连接加速查询

Question

我有几张桌子需要加入。我有一个员工表（约 40 万行）、一个公司表（约 1000 万行）和一个存储某人工作地点的员工公司表。

基本上，我需要让所有符合某些条件的员工（他们在拥有网站的公司工作，位于某个国家/地区等）。我进行了查询以获取此信息，但花费的时间太长。我需要加快速度。

SELECT  DISTINCT "employees".* 
FROM "employees" 
INNER JOIN "employee_companies" ON "employee_companies"."employee_id" = "employees"."id" 
INNER JOIN "companies" ON "companies"."id" = "employee_companies"."company_id" 
WHERE (employee_companies.employee_id IS NOT NULL)
AND (companies.website IS NOT NULL) 
AND (employees.country = 'Uruguay') 
ORDER BY employees.connections DESC

这是该查询的计划：

Unique  (cost=877170.24..880752.72 rows=62304 width=1064) (actual time=24023.736..26001.876 rows=73318 loops=1)
  ->  Sort  (cost=877170.24..877326.00 rows=62304 width=1064) (actual time=24023.733..24305.989 rows=77579 loops=1)
        Sort Key: employees.connections DESC, employees.id, employees.name, employees.link, employees.role, employees.area, employees.profile_picture, employees.summary, employees.current_companies, employees.previous_companies, employees.skills, employees.education, employees.languages, employees.volunteer, employees.groups, employees.interests, employees.search_vector, employees.secondary_search_vector, employees.email_status, employees.languages_count, employees.role_hierarchy
        Sort Method: external merge  Disk: 85816kB
        ->  Nested Loop  (cost=2642.38..843246.15 rows=62304 width=1064) (actual time=139.870..23056.234 rows=77579 loops=1)
              ->  Hash Join  (cost=2641.95..221744.50 rows=77860 width=1068) (actual time=139.841..22617.587 rows=77579 loops=1)
                    Hash Cond: (employees.id = employee_companies.employee_id)
                    ->  Seq Scan on employees  (cost=0.00..212178.88 rows=409672 width=1064) (actual time=8.145..22369.166 rows=393725 loops=1)
                          Filter: ((country)::text = 'Uruguay'::text)
                          Rows Removed by Filter: 1075
                    ->  Hash  (cost=1666.42..1666.42 rows=78042 width=8) (actual time=44.675..44.675 rows=78042 loops=1)
                          Buckets: 131072  Batches: 1  Memory Usage: 4073kB
                          ->  Seq Scan on employee_companies  (cost=0.00..1666.42 rows=78042 width=8) (actual time=0.007..22.901 rows=78042 loops=1)
                                Filter: (employee_id IS NOT NULL)
              ->  Index Scan using companies_pkey on companies  (cost=0.43..7.97 rows=1 width=4) (actual time=0.004..0.004 rows=1 loops=77579)
                    Index Cond: (id = employee_companies.company_id)
                    Filter: (website IS NOT NULL)
Planning time: 1.957 ms
Execution time: 26025.045 ms

这些是我桌子上的相关索引：

员工：

"employees_pkey" PRIMARY KEY, btree (id)
"ix_employees_country" btree (country)

公司：

"companies_pkey" PRIMARY KEY, btree (id)
"empty_websites" btree (website) WHERE website IS NULL
"index_companies_on_website" btree (website)
"not_empty_websites" btree (website) WHERE website IS NOT NULL

员工_公司：

"employee_companies_pkey" PRIMARY KEY, btree (id)
"index_employee_companies_on_company_id" btree (company_id)
"index_employee_companies_on_employee_id" btree (employee_id)
"index_employee_companies_on_employee_id_and_company_id" btree (employee_id, company_id)
"not_empty_employee_id" btree (employee_id) WHERE employee_id IS NOT NULL

有没有其他更好的方法来做我想做的更高效/性能更好的事情？

谢谢！

Answer 1

基于一些猜测模拟，我认为您可以通过以下方式稍微改进您的查询：

1. 避免使用外部DISTINCT子句（尽管会有一个隐含的DISTINCT）。
2. 子选择一部分数据，以便减少需要的数据JOIN。

查询如下：

SELECT  
    employees.* 
FROM 
    employees 
WHERE
    employee_id IN
    (SELECT 
        -- Choose all employees from companies with website
        employee_id 
     FROM 
        employee_companies
        JOIN companies ON companies.company_id = employee_companies.company_id
     WHERE
        companies.website IS NOT NULL
    )
    -- Now filter only employees from 'Germany'
    AND employees.country = 'Germany' 
ORDER BY 
    employees.connections DESC ;

用于生成模拟的数据如下：

表和索引定义：

CREATE TABLE employees
(
    employee_id integer PRIMARY KEY,
    country text,
    connections integer,
    something_else text
) ;

CREATE INDEX idx_employee_country 
   ON employees (country) ;

CREATE TABLE companies
(
    company_id integer PRIMARY KEY,
    website text,
    something_else text
) ;

CREATE INDEX not_empty_websites 
    ON companies(company_id, website) WHERE website IS NOT NULL ;

CREATE TABLE employee_companies
(
    employee_id integer NOT NULL REFERENCES employees(employee_id),
    company_id integer NOT NULL REFERENCES companies(company_id),
    PRIMARY KEY (employee_id, company_id)
) ;

CREATE INDEX company_employee
    ON employee_companies(company_id, employee_id) ;

1.000.000 家公司（更改为 10M 没有太大区别）。我假设 90% 有一个网站。

INSERT INTO 
   companies
   (company_id, website)
SELECT
   generate_series(1, 1000000), 
   CASE WHEN random() > 0.1 THEN 'web.com' END AS website ;

80k 员工（大约 10% 是德国人）

INSERT INTO
   employees 
   (employee_id, country, connections)
SELECT
    generate_series(1, 80000),
    case (random()*10)::integer
    when 0 then 'Germany'
    when 1 then 'United Kingdon'
    when 2 then 'United States'
    else 'Angola'
    end AS country,
    (random()*10)::integer AS connections ;

200K 员工 x 公司（这意味着人们平均在大约 3 家公司工作过）：

INSERT INTO 
    employee_companies
    (employee_id, company_id)
SELECT DISTINCT
    (random()*79999)::integer + 1,
    (random()*999999)::integer + 1
FROM
    generate_series (1, 200000) ;

您可以在dbfiddle 此处查看此模拟的缩小版本。如果此模拟数据与您的场景非常相似，则更改查询可使服务器执行时间提高 3 倍。我建议你试一试。

模拟数据（按比例缩小 25 倍）与真实场景更相似的场景并没有提供如此可观的性能提升……不过，它提高了 1.5 倍。

在这个dbfiddle检查它