以下句子中的用户定义变量是非常有线的,
SELECT
@f:=CONVERT(
IF(@c<=>CatId AND DATEDIFF(Date, @d)=1, @f, Date), DATE
) AS Begin,
@c:=CatId, @d:=Date
FROM my_table AS init
ORDER BY CatId, Date
如此处所示:http://sqlfiddle.com/#!2/fddbd/83,my_table由:
给出CREATE TABLE my_table (
Id INT(6) UNSIGNED ZEROFILL,
CatId INT,
Date DATE,
Rate INT
);
INSERT INTO my_table
VALUES
(000001, 12, '2009-07-07', 1),
(000002, 12, '2009-07-08', 1),
(000003, 12, '2009-07-09', 1),
(000004, 12, '2009-07-10', 2),
(000005, 12, '2009-07-15', 1),
(000006, 12, '2009-07-16', 1),
(000007, 13, '2009-07-08', 1),
(000008, 13, '2009-07-09', 1),
(000009, 14, '2009-07-07', 2),
(000010, 14, '2009-07-08', 1),
(000010, 14, '2009-07-10', 1);
更确切地说,变量@f
没有预先定义,mysql将为DateDiff做什么逻辑?
答案 0 :(得分:1)
在第一行,它将使用NULL
作为所有变量的值。所以它等同于以下查询将所有显式初始化
SELECT
@f:=CONVERT(
IF(@c<=>CatId AND DATEDIFF(Date, @d)=1, @f, Date), DATE
) AS Begin,
@c:=CatId, @d:=Date
FROM my_table AS init
CROSS JOIN (SELECT @f := NULL, @c := NULL, @d := NULL) AS vars
ORDER BY CatId, Date
通常不建议将ORDER BY
子句放在与处理相同的查询级别中。 MySQL执行此操作的方式不同,具体取决于ORDER BY
子句中SELECT
中的列是否已命名。如果是,则在生成所有结果后执行排序,否则在处理之前使用它来排序表中的行。获得可预测结果的方法是使用子查询:
SELECT
@f:=CONVERT(
IF(@c<=>CatId AND DATEDIFF(Date, @d)=1, @f, Date), DATE
) AS Begin,
@c:=CatId, @d:=Date
FROM (SELECT *
FROM my_table
ORDER BY CatId, Date) AS init
CROSS JOIN (SELECT @f := NULL, @c := NULL, @d := NULL) AS vars
由于您在不同的MySQL版本上看到不同的结果,这可能还取决于版本。使用子查询应该使其一致地工作。