解决Key columns should be a ordered prefix of the schema. KeyColumns[1] (starts from zero) is xxx, but

1. 复现错误

今天在编写doris语句：

CREATE TABLE IF NOT EXISTS test3.test (
  `id` int COMMENT 'ID',
  `bb` string COMMENT '字段bb',
  `aa` date COMMENT '测试aa'
) ENGINE = olap UNIQUE KEY(`id`, `aa`) PARTITION BY RANGE(`aa`) (
  FROM
    ('2023-10-13') TO ('2023-10-14') INTERVAL 1 DAY
) DISTRIBUTED BY HASH(`id`) BUCKETS 10 PROPERTIES("replication_num" = "1");

却报出如下错误：

即Key columns should be a ordered prefix of the schema. KeyColumns[1] (starts from zero) is aa, but corresponding column is bb in the previous columns declaration.

2. 分析错误

将错误Key columns should be a ordered prefix of the schema. KeyColumns[1] (starts from zero) is aa, but corresponding column is bb in the previous columns declaration.

翻译成中文键列应该是架构的有序前缀。KeyColumns[1]（从零开始）是 aa，但在前面的列声明中对应的列是 bb。

描述的很清楚，在id后面语句应该是

`aa` date COMMENT '测试22'

而不是

`bb` string COMMENT '字段bb'

为什么是这样的呢？因为aa是分区字段，而且通过UNIQUE KEY来限制了，表示唯一键。

3. 解决错误

既然aa是分区字段，又通过UNIQUE KEY来表示唯一键。

我们便可把aa字段放在bb之前，如下代码所示：

CREATE TABLE IF NOT EXISTS test3.test (
  `id` int COMMENT 'ID',
  `aa` date COMMENT '测试aa',
  `bb` string COMMENT '字段bb'
) ENGINE = olap UNIQUE KEY(`id`, `aa`) PARTITION BY RANGE(`aa`) (
  FROM
    ('2023-10-13') TO ('2023-10-14') INTERVAL 1 DAY
) DISTRIBUTED BY HASH(`id`) BUCKETS 10 PROPERTIES("replication_num" = "1");

如此便可运行成功：