SQL：如何使用索引加速查询

Question

我正在尝试加快查询速度，以查找所有购买了 1970 年之前制造的 MOTORCYCLE 并购买了 2010 年之后制造的另一辆 MOTORCYCLE 的客户。由于我的查询运行非常缓慢，我认为我需要帮助以找到更好的索引.我的尝试记录如下：

表格

CREATE TABLE CUSTOMER (
  id int PRIMARY KEY, 
  fname varchar(30),     
  lname varchar(30)
);

CREATE TABLE MOTORCYCLE (
  id int PRIMARY KEY, 
  name varchar(30), 
  year int -- Manufactured year
);

CREATE TABLE SALES (
  cid int,
  mid int,
  FOREIGN KEY(cid) REFERENCES CUSTOMER(id), 
  FOREIGN KEY(mid) REFERENCES MOTOCYCLE(id),
  PRIMARY KEY(pid, mid, role)
);

索引

这是我的索引（我有些猜测，但这是我的尝试）：

CREATE UNIQUE INDEX customerID on CUSTOMER(id);
CREATE INDEX customerName on CUSTOMER(fname, lname);

CREATE UNIQUE INDEX motorcycleID on MOTORCYCLE(id);
CREATE INDEX motorcycleName on MOTORCYCLE(name);
CREATE INDEX motorcycleYear on MOTORCYCLE(year);

CREATE INDEX salesCustomerMotorcycleID on SALES(cid, mid);       
CREATE INDEX salesCustomerID on SALES(cid);
CREATE INDEX castsMotorcycleID on SALES(mid);

查询

我要查找购买 1970 年之前和 2010 年之后制造的自行车的客户的查询如下：

SELECT fname, lname
FROM (SALES INNER JOIN CUSTOMER ON SALES.cid=CUSTOMER.id) INNER JOIN MOTORCYCLE ON MOTORCYCLE.id=SALES.mid
GROUP BY CUSTOMER.id
HAVING MIN(MOTORCYCLE.year) < 1970 AND MAX(MOTORCYCLE.year) > 2010;

这是另一个有效的查询，它避免了 GROUP BY 和 HAVING 子句：

SELECT DISTINCT C.id, fname, lname
FROM (CUSTOMER as C inner join (SALES as S1 INNER JOIN MOTORCYCLE as M1 ON M1.id=S1.mid) on C.id=S1.cid) inner join (SALES as S2 inner join MOTORCYCLE as M2 on S2.mid=M2.id) on C.id=S2.cid
WHERE (M1.year < 1970 AND M2.year > 2010);

关于我可以用来加快查询速度的索引种类有什么建议吗？还是我应该更改我的查询？

更新

我发现另一个查询也有效，但它也太慢了。它已在上面添加。不过，在查找索引以加快速度时，它可能会有所帮助。

Answer 1

当您使用 EXPLAIN QUERY PLAN 查看查询时，您会发现在这两种情况下，数据库都会先查找许多相关记录，然后再过滤掉不需要的记录（包括不需要的年份）。

以下查询在匹配之前查找摩托车 ID；哪个更快取决于您的数据的详细信息，并且必须由您衡量：

SELECT *
FROM Customer
WHERE EXISTS (SELECT 1
              FROM Sales
              WHERE cid = Customer.id
                AND mid IN (SELECT id
                            FROM Motorcycle
                            WHERE year < 1970))
  AND EXISTS (SELECT 1
              FROM Sales
              WHERE cid = Customer.id
                AND mid IN (SELECT id
                            FROM Motorcycle
                            WHERE year > 2010));

SELECT *
FROM Customer
WHERE EXISTS (SELECT 1
              FROM Sales AS s1
              JOIN Sales AS s2 ON s1.cid = s2.cid
              WHERE s1.cid = Customer.id
                AND s1.mid IN (SELECT id
                               FROM Motorcycle
                               WHERE year < 1970)
                AND s2.mid IN (SELECT id
                               FROM Motorcycle
                               WHERE year > 2010));

SQL Fiddle

Answer 2

当查询中没有使用聚合函数时，为什么要使用 group by？ 如果您不想看到任何重复，请改用 distinct