Question

我有一张桌子：

CREATE TABLE spots(
    datetime TIMESTAMP,
    market VARCHAR(15),
    spot NUMERIC(10, 5),
    PRIMARY KEY (market, datetime)
);

我需要选择p_market的行，其最大值为datetime，该值小于或等于p_datetime，因此我有两个选项：

SELECT * FROM spots WHERE market = p_market AND datetime = (
    SELECT MAX(datetime) FROM spots WHERE market = p_market AND datetime <= p_datetime
);

和

SELECT * FROM spots WHERE market = p_market AND datetime <= p_datetime
ORDER BY datetime DESCENDING LIMIT 1;

所以，问题是 - 从绩效角度来看，哪种变体更好。

Answer 1

Postgres支持window function，理论上应该在处理行时逐步计算，因此它应该会对大型数据集提供一些性能提升：

SELECT * 
FROM   (SELECT *,
               RANK() OVER (ORDER BY datetime DESC) AS rk
        FROM   spots
        WHERE  market = p_market AND datetime <= p_datetime) t
FROM   spots 
WHERE  rk = 1

在主键字段选择具有最大值的行的最快方法

1 个答案: