提高SQL查询性能，需要澄清

Question

我有一个股票信息数据库，并正在尝试从中挖掘数据。

首先设置索引：

CREATE INDEX IF NOT EXISTS s1 on Income (symbol, period);
CREATE INDEX IF NOT EXISTS s2 on BalanceSheet (symbol, period);
CREATE INDEX IF NOT EXISTS s3 on CashFlow (symbol, period);

然后我建立了一个临时表：

DROP TABLE IF EXISTS _;
CREATE TABLE _(symbol, period);
INSERT INTO  _(symbol, period) VALUES ('AAPL', 'Annual');

然后我做我的选择：

    SELECT 
        a.yearmonth [Date], a.symbol, a.periodtype [Period],

        -- Income and revenues
        COALESCE(MAX(CASE WHEN a.statementitem = 'Revenues' THEN a.value END), 0) Revenue,
        COALESCE(MAX(CASE WHEN a.statementitem = 'Other Operating Expense/(Income)' THEN a.value END),0) [Other Operating Expense/(Income)],
        -- ...
        -- ...
        -- ...
        COALESCE(MAX(CASE WHEN c.statementitem = 'Depreciation & Amortization, Total' THEN c.value END),0) [Depreciation & Amortization, Total],  -- This one comes from CashFlow
        -- ...
        -- ...
        -- ...
        COALESCE(MAX(CASE WHEN b.statementitem = 'Cash And Equivalents' THEN b.value END),0) [Cash And Equivalents],
        COALESCE(MAX(CASE WHEN b.statementitem = 'Total Cash & ST Investments' THEN b.value END),0) [Total Cash & ST Investments],
        -- ...
        -- ...
        -- ...
FROM _ 
INNER JOIN Income a ON a.symbol = _.symbol AND a.periodtype = _.period 
INNER JOIN BalanceSheet b ON b.symbol = _.symbol AND b.periodtype = _.period 
INNER JOIN CashFlow c ON c.symbol=_.symbol AND c.periodtype = _.period 
GROUP BY a.yearmonth, a.symbol, a.periodtype

以上两件事：

1. 此查询有效
2. 此查询很慢

请问如何做才能使其更快？

Answer 1

所有表上的索引应在(symbol, period)上。但是，我认为这不会产生很大的影响。

据我所知，您无能为力。数据需要聚合大量数据，这大概是花费时间。

Answer 2

您所有的联接都是INNER联接，但实际上您是在对表进行CROSS联接：Income，BalanceSheet和CashFlow过滤掉symbol = 'AAPL' and periodtype = 'Annual'所在的行。我看不到临时表的需要。3个表的cross联接总是很昂贵，索引可能会有所帮助。您可以做的是使crossjoinσ的结果更浅，是在连接之前删除临时表并进行过滤：

...............................
FROM (SELECT * FROM Income WHERE symbol = 'AAPL' AND periodtype = 'Annual') a 
CROSS JOIN (SELECT * FROM BalanceSheet WHERE symbol = 'AAPL' AND periodtype = 'Annual') b
CROSS JOIN (SELECT * FROM CashFlow WHERE symbol = 'AAPL' AND periodtype = 'Annual') c
GROUP BY a.yearmonth, a.symbol, a.periodtype