MySQL: Can this Join query be optimized? - mysql

Question

I have two tables. create command for those tables are as follows
CREATE TABLE `invoice` (
`Id` bigint(20) NOT NULL,
`VersionNumber` int(11) NOT NULL,
`CreationDate` datetime default NULL,
`ModificationDate` datetime default NULL,
`CreateBy` bigint(20) default NULL,
`ModifyBy` bigint(20) default NULL,
`BusinessId` varchar(255) character set utf8 default NULL,
`Status` int(11) default NULL,
`VendorId` bigint(20) default NULL,
`CustomerId` bigint(20) default NULL,
`OrderDate` date default NULL,
`ExpectedDate` date default NULL,
`DeliveryDate` date default NULL,
`InvoiceFor` int(11) default NULL,
`Reference` varchar(255) character set utf8 default NULL,
`Agent` varchar(255) character set utf8 default NULL,
`BnAgent` varchar(255) character set utf8 default NULL,
`Note` varchar(255) character set utf8 default NULL,
`HasVoucher` char(1) character set utf8 default NULL,
PRIMARY KEY (`Id`),
KEY `CustomerId` (`CustomerId`)
) ENGINE=InnoDB DEFAULT CHARSET=latin1;
CREATE TABLE `invoiceitem` (
`Id` bigint(20) NOT NULL,
`VersionNumber` int(11) NOT NULL,
`CreationDate` datetime default NULL,
`ModificationDate` datetime default NULL,
`CreateBy` bigint(20) default NULL,
`ModifyBy` bigint(20) default NULL,
`BusinessId` varchar(255) default NULL,
`Status` int(11) default NULL,
`InvoiceId` bigint(20) default NULL,
`ProductId` bigint(20) default NULL,
`PackageQty` decimal(19,5) default NULL,
`PackagePrice` decimal(19,5) default NULL,
`ItemPerPackage` decimal(19,5) default NULL,
`ItemQty` decimal(19,5) default NULL,
`ItemPrice` decimal(19,5) default NULL,
`StoreQty` decimal(19,5) default NULL,
`RevenuePercent` decimal(19,5) default NULL,
`PurchasePrice` decimal(19,5) default NULL,
`Vat` decimal(19,5) default NULL,
`TotalAmount` decimal(19,5) default NULL,
`InvoiceItemFor` int(11) default NULL,
`LifeTimeUptoDate` datetime default NULL,
PRIMARY KEY (`Id`),
KEY `invoiceid_productid` (`InvoiceId`,`ProductId`),
KEY `ProductId` (`ProductId`),
CONSTRAINT `FK_invoiceitem` FOREIGN KEY (`InvoiceId`) REFERENCES `invoice` (`Id`)
) ENGINE=InnoDB DEFAULT CHARSET=latin1;
Purpose of this query
I have to show the list of invoices to the admin. He need to see the invoice date, Invoice number, customer Id, Product id and if the customer take that product previously then that previous date.
Now the Query
SELECT iv.id, iv.CustomerId, ii.ProductId, MAX(iv2.orderdate) AS PriviousDate
FROM invoice AS iv
INNER JOIN invoice AS iv2
ON iv.CustomerId=iv2.CustomerId
AND iv.OrderDate>iv2.OrderDate
INNER JOIN invoiceItem AS ii
ON iv.id=ii.invoiceId
INNER JOIN invoiceItem AS ii2
ON iv2.id=ii2.invoiceId
AND ii.ProductId=ii2.ProductId
WHERE iv.Status=0
AND ii.Status=0
AND iv2.Status=0
AND ii2.Status=0
GROUP BY ii.ProductId, iv.CustomerId
Here iv.id and ii.id is primary key
customerId and Productid fields are also indexes
and invoiceId is the foreginkey (iv.id=ii.invoiceId )
I am running this query in my MySql server (local)
But Most of the time I got timeout. How can I optimized this query?
the Explain for this query is as follows:
now I apply
create index invoiceid_productid
on invoiceItem(invoiceId, productId)
after that the explain result is

Answer 1

Not much here (it would have been helpful to see what you're indexes are) but assuming the invoice ids increase in line with invoice date....I suspect this might be slightly faster:
SELECT iv.id, iv.CustomerId, ii.ProductId, MAX(iv2.orderdate) AS PriviousDate
FROM invoice AS iv
INNER JOIN invoice AS iv2
ON iv.CustomerId=iv2.CustomerId
AND iv.id>iv2.id
INNER JOIN invoiceItem AS ii
ON iv.id=ii.invoiceId
INNER JOIN invoiceItem AS ii2
ON iv2.id=ii2.invoiceId
AND ii.ProductId=ii2.ProductId
AND ii.invoiceId>ii2.invoiceId
WHERE iv.Status=0
AND ii.Status=0
AND iv2.Status=0
AND ii2.Status=0
GROUP BY ii.ProductId, iv.CustomerId
But you really need an index on ProductId, Status and InvoiceId in the invoiceItem table before it's going to be appreciable faster.

Answer 2

Looking at your EXPLAIN results, the problem is clearly the invoiceItem table. It has 200,000 rows, and MySQL isn't using any indexes to access it.
So you should look at creating indexes to speed access to it.
I mocked up your tables:
create table invoice (
id int auto_increment not null,
customerid int not null,
orderdate timestamp not null,
status int default 0 not null,
primary key(id)
);
create table invoiceItem (
id int auto_increment not null,
invoiceId int not null,
productId int not null,
status int default 0 not null,
primary key (id)
);
I created these indexes:
create index invoiceId on invoiceItem(invoiceId);
create index cod on invoice(customerid, orderdate);
create index stat on invoice(status);
My EXPLAIN (on empty tables) now has all tables using type ref.
You can also see this tutorial which is for MySQL ;-)