更好的MySQL查询性能

Question

I want to get the SUM of QTYs of an item# grouped be months, the query takes too long (15 - 20) seconds to fetch.

-Total rows: 1495873 -Total Fetched rows: 9 - 12

The relation between two tables ( invoice_header and invoice_detail ) is (one to many) that the invoice_header is the header of an invoice, with only totals. Which is linked to invoice_detail using location ID ( loc_id ) and Invoice number ( invo_no ), as each location has its own serial number. The invoice detail contains the details of each invoice.

Is there's a better way to enhance the performance of that query, here it's:

SELECT SUM(invoice_detail.qty) AS qty, Month(invoice_header.date) AS month
FROM invoice_detail
JOIN invoice_header ON invoice_detail.invo_no = invoice_header.invo_no
AND  invoice_detail.loc_id = invoice_header.loc_id
WHERE invoice_detail.item_id = {$itemId}
GROUP BY Month(invoice_header.date)
ORDER BY Month(invoice_header.date)

EXPLAIN:

invoice_header table structure:

CREATE TABLE `invoice_header` (
 `invo_type` varchar(1) NOT NULL,
 `invo_no` int(20) NOT NULL AUTO_INCREMENT,
 `invo_code` varchar(50) NOT NULL,
 `date` date NOT NULL,
 `time` time NOT NULL,
 `cust_id` int(11) NOT NULL,
 `loc_id` int(3) NOT NULL,
 `cash_man_id` int(11) NOT NULL,
 `sales_man_id` int(11) NOT NULL,
 `ref_invo_no` int(20) NOT NULL,
 `total_amount` decimal(19,2) NOT NULL,
 `tax` decimal(19,2) NOT NULL,
 `discount_amount` decimal(19,2) NOT NULL,
 `net_value` decimal(19,2) NOT NULL,
 `split` decimal(19,2) NOT NULL,
 `qty` int(11) NOT NULL,
 `payment_type_id` varchar(20) NOT NULL,
 `comments` varchar(255) NOT NULL,
 PRIMARY KEY (`invo_no`,`loc_id`)
) ENGINE=InnoDB AUTO_INCREMENT=20286 DEFAULT CHARSET=utf8

invoice_detail table structure:

CREATE TABLE `invoice_detail` (
 `invo_no` int(11) NOT NULL,
 `loc_id` int(3) NOT NULL,
 `serial` int(11) NOT NULL,
 `item_id` varchar(11) NOT NULL,
 `size_id` int(5) NOT NULL,
 `qty` int(11) NOT NULL,
 `rtp` decimal(19,2) NOT NULL,
 `type` tinyint(1) NOT NULL,
 PRIMARY KEY (`invo_no`,`loc_id`,`serial`),
 KEY `item_id` (`item_id`),
 KEY `size_id` (`size_id`),
 KEY `invo_no` (`invo_no`),
 KEY `serial` (`serial`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8

Answer 1

According to see the result of explain invoice_header can not use any multiple index.

You could create index on field invo_no on invoice_header Table.

Answer 2

How long does it take by following SQL?

SELECT count(*)
FROM invoice_detail
WHERE invoice_detail.item_id = {$itemId}

If this SQL takes 15-20 seconds, you should add an index on field item_id on inovice_detail table.

The invoice_header already had the primary key on the join columns of invo_no and loc_id, so you don't need add other indexes on invoice_header table. But if you add an index on fields invo_no, loc_id and date, that may enhance a little performance by only index scan.

Answer 3

Try creating a few indexes on some columns: invo_no, loc_id, item_id, date.

based on your explain, your query is scanning all rows in invoice_header. Index the columns that make the join... invo_no, loc_id