What is the best index for this two query? - mysql

Question

I want to avoid redundant index, so what is the best composite index for these two query? Based on my understanding, these two query cannot have the same composite index since one need country, the other one need product_id, but if I make the index as below, will it be redundant index, and effect the DB performance?
combine merchant_id, created_at and product_id
combine merchant_id, created_at and country
Query 1
SELECT * from shop_order
WHERE shop_order.merchant_id = ?
AND shop_order.created_at >= TIMESTAMP(?)
AND shop_order.created_at <= TIMESTAMP(?)
AND shop_order.product_id = ?) AS mytable
WHERE product_id IS NOT NULL GROUP BY product_id, title;
Query 2
SELECT COALESCE(SUM(total_price_usd),0) AS revenue,
COUNT(*) as total_order, COALESCE(province, 'Unknown') AS name
FROM shop_order
WHERE DATE(created_at) >= '2021-02-08 13:37:42'
AND DATE(created_at) <= '2021-02-14 22:44:13'
AND merchant_id IN (18,19,20,1)
AND country = 'Malaysia' GROUP BY province;
Table structure
CREATE TABLE `shop_order` (
`id` bigint(20) NOT NULL AUTO_INCREMENT,
`merchant_id` bigint(20) DEFAULT NULL,
`order_id` bigint(20) NOT NULL,
`customer_id` bigint(20) DEFAULT NULL,
`customer_orders_count` varchar(45) DEFAULT NULL,
`customer_total_spent` varchar(45) DEFAULT NULL,
`customer_email` varchar(100) DEFAULT NULL,
`customer_last_order_name` varchar(45) DEFAULT NULL,
`currency` varchar(10) NOT NULL,
`total_price` decimal(20,8) NOT NULL,
`subtotal_price` decimal(20,8) NOT NULL,
`transaction_fee` decimal(20,8) DEFAULT NULL,
`total_discount` decimal(20,8) DEFAULT '0.00000000',
`shipping_fee` decimal(20,8) DEFAULT '0.00000000',
`total_price_usd` decimal(20,8) DEFAULT NULL,
`transaction_fee_usd` decimal(20,8) DEFAULT NULL,
`country` varchar(50) DEFAULT NULL,
`province` varchar(45) DEFAULT NULL,
`processed_at` datetime DEFAULT NULL,
`refunds` json DEFAULT NULL,
`ffm_status` varchar(50) DEFAULT NULL,
`gateway` varchar(45) DEFAULT NULL,
`confirmed` tinyint(1) DEFAULT NULL,
`cancelled_at` datetime DEFAULT NULL,
`cancel_reason` varchar(100) DEFAULT NULL,
`created` datetime NOT NULL DEFAULT CURRENT_TIMESTAMP,
`updated` datetime DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
`order_number` bigint(1) DEFAULT NULL,
`created_at` datetime DEFAULT NULL,
`financial_status` varchar(45) DEFAULT NULL,
PRIMARY KEY (`id`),
UNIQUE KEY `shop_order_unique` (`merchant_id`,`order_id`),
KEY `merchant_id` (`merchant_id`),
KEY `combine_idx1` (`country`,`merchant_id`,`created_at`)
) ENGINE=InnoDB AUTO_INCREMENT=2237 DEFAULT CHARSET=utf8mb4;
Please help me

Answer 1

Query 1:
INDEX(merchant_id, product_id, -- Put columns for "=" tests first (in any order)
created_at) -- then range
Query 2. First, avoid hiding created_at in a function call (DATE()); it prevents using it in an index.
INDEX(country, -- "="
merchant_id, -- IN
created_at) -- range (after removing DATE)
You are correct in saying that you need separate indexes for those queries. And possibly some of your existing indexes are needed for other queries.
Also, you already have a redundant index. Drop KEY merchant_id (merchant_id), -- You have ate least one other index starting with merchant_id.
Having extra indexes is only a minor performance drag. And the hit is during INSERT or if you UPDATE any column in an index. Generally, the benefit of the 'right' index to a SELECT outweighs the hit to the writes.
Having multiple unique indexes is somewhat a burden. Do you really need id since you have a "natural" PK made up of those two columns? Check to see if other tables need to join on id.
Consider shrinking many of the datasizes. BIGINT takes 8 bytes and has a range that is rarely needed. decimal(20,8) takes 10 bytes and allows up to a trillion dollars; this seems excessive, too. Is customer_orders_count a number?