Must the MySQL query column be consistent with the group by field?

Catalog

• scene ： Check the highest paid employees in each department .
• MySQL group by How to decide which piece of data to leave ？
• that target list and group by column Can't it be executed if it doesn't match ？

MySQL edition ：8.0.27

scene ： Check the highest paid employees in each department .

CREATE TABLE `employee`  (
  `id` int NOT NULL AUTO_INCREMENT COMMENT ' Primary key ID',
  `dept` int NOT NULL COMMENT ' department ',
  `user` varchar(32) CHARACTER SET utf8 COLLATE utf8_general_ci NOT NULL COMMENT ' staff ',
  `salary` int NULL DEFAULT NULL COMMENT ' salary ',
  `is_deleted` tinyint(1) NOT NULL DEFAULT 0 COMMENT ' Whether or not to delete ',
  `remark` varchar(512) CHARACTER SET utf8 COLLATE utf8_general_ci NULL DEFAULT NULL COMMENT ' remarks ',
  `modify_time` datetime(3) NOT NULL DEFAULT CURRENT_TIMESTAMP(3) ON UPDATE CURRENT_TIMESTAMP(3) COMMENT ' Modification time ',
  PRIMARY KEY (`id`) USING BTREE
) ENGINE = InnoDB CHARACTER SET = utf8 COLLATE = utf8_general_ci COMMENT = ' staff ' ROW_FORMAT = Dynamic;

INSERT INTO `employee` VALUES (1, 1, ' Zhang San ', 1000, 0, NULL, '2021-12-23 09:20:19.606');
INSERT INTO `employee` VALUES (2, 1, ' Li Si ', 1500, 0, NULL, '2021-12-23 09:20:21.679');
INSERT INTO `employee` VALUES (3, 1, ' Wang Wu ', 2000, 0, NULL, '2021-12-23 09:20:23.371');
INSERT INTO `employee` VALUES (4, 2, ' Zhao Liu ', 1000, 0, NULL, '2021-12-23 09:21:59.373');
INSERT INTO `employee` VALUES (5, 2, ' Sun Qi ', 1500, 0, NULL, '2021-12-23 09:22:15.000');

SELECT * FROM employee ;

Method 1 ：

SELECT
	t1.* 
FROM
	employee t1
	LEFT JOIN employee t2 ON t2.dept = t1.dept AND t1.salary < t2.salary 
WHERE
	t2.salary IS NULL;

Method 2 ：

SELECT
	* 
FROM
	( SELECT * FROM `employee` ORDER BY dept, salary DESC LIMIT 1000 ) t 
GROUP BY
	dept;

（ No addition limit It may fail ）

It seems that the result is the same , But there will be problems with the second one .

MySQL group by How to decide which piece of data to leave ？

MySQL adopt sql_mode To provide SQL Check the validity of the statement ,

By default ,MySQL Allow query Columns target list In addition to group by column、 Expressions other than aggregate functions .

however , Those who don't participate group by Which data value will be returned in the field of MySQL Is in the state of undefined rules ,

MySQL Do not promise which data will be returned .

Data before grouping ：

SELECT * FROM employee ORDER BY dept, salary DESC LIMIT 1000;

It seems that method 2 returns the data of the first item in each group ,

But it will actually work with the storage engine 、 physical position 、 Index, etc ,

If it is InnoDB Words , It depends on B+Tree The first index hit on ,

There is no explanation here , After all, it's not safe to use ,

Sometimes the result may not be what we want .

About B+Tree, Take a look at this article ：

adopt B+Tree Balanced multitree understanding InnoDB Clustered and nonclustered indexes of the engine

So for target list Ambiguous columns in ,MySQL It's not sure which piece of data is left .

For databases with strict syntax restrictions , Don't support target list Semantically ambiguous columns appear in ,

MySQL A modified is provided in sql_mode,ONLY_FULL_GROUP_BY.

SET SESSION sql_mode = 'ONLY_FULL_GROUP_BY';

Then execute the... Of method 2 SQL I was rejected ：

SELECT
	* 
FROM
	( SELECT * FROM `employee` ORDER BY dept, salary DESC LIMIT 1000 ) t 
GROUP BY
	dept
> 1055 - Expression #1 of SELECT list is not in GROUP BY clause and contains nonaggregated column 't.id' which is not functionally dependent on columns in GROUP BY clause; this is incompatible with sql_mode=only_full_group_by
>  Time : 0s

'only_full_group_by' In mode MySQL Would be right target list and group by column The base column in 、 expression 、 Don't make a strict match .

that target list and group by column Can't it be executed if it doesn't match ？

Let's look at another one SQL：

#  Order 
CREATE TABLE `order`  (
  `order_id` int NOT NULL AUTO_INCREMENT COMMENT ' Order ID',
  `order_amount` int NULL DEFAULT NULL COMMENT ' Order amount ',
  PRIMARY KEY (`order_id`) USING BTREE
) ENGINE = InnoDB CHARACTER SET = utf8 COLLATE = utf8_general_ci COMMENT = ' Order ' ROW_FORMAT = DYNAMIC;

INSERT INTO `order` VALUES (1, 100);
INSERT INTO `order` VALUES (2, 103);
INSERT INTO `order` VALUES (3, 100);

#  The order details 
CREATE TABLE `order_detail`  (
  `order_detail_id` int NOT NULL AUTO_INCREMENT COMMENT ' Primary key ID',
  `order_id` int NOT NULL COMMENT ' Order ID',
  `goods` varchar(32) CHARACTER SET utf8 COLLATE utf8_general_ci NOT NULL COMMENT ' Name of commodity ',
  `goods_amount` int NOT NULL COMMENT ' The amount of goods ',
  PRIMARY KEY (`order_detail_id`) USING BTREE
) ENGINE = InnoDB CHARACTER SET = utf8 COLLATE = utf8_general_ci COMMENT = ' The order details ' ROW_FORMAT = DYNAMIC;

INSERT INTO `order_detail` VALUES (1, 1, ' Apple ', 10);
INSERT INTO `order_detail` VALUES (2, 1, ' The oranges ', 20);
INSERT INTO `order_detail` VALUES (3, 1, ' Banana ', 70);
INSERT INTO `order_detail` VALUES (4, 2, ' a mandarin orange ', 50);
INSERT INTO `order_detail` VALUES (5, 2, ' Pineapple ', 53);

Query all items in the order

SELECT
	t1.order_id,
	t1.order_amount,
	GROUP_CONCAT( t2.goods, t2.goods_amount ) 
FROM
	`order` t1
	LEFT JOIN order_detail t2 ON t2.order_id = t1.order_id 
GROUP BY
	t1.order_id;

This article SQL Of target list and group by column It's not a strict match , But it can also be done ,

Be careful

t1.order_id Is the primary key of the order table .

So in 'only_full_group_by' In mode , If MySQL Can be determined target list Return values of all columns in the ,

that , Even if target list and group by column The base column in 、 expression 、 Aliases, columns, etc. do not match strictly ,

MySQL It will also be considered that its semantics is clear , Therefore, the statement can pass .

original text
http://www.cnblogs.com/captaincat/