使MySQL查询更快

Question

I need to make this query faster and need some help.

The code is :

sqlFrom = "FROM tx_apartment_buy a where result_id" + 
"not in (select result_id from tx_user_view2" + 
"where product_type='apartmentBuy'" + 
 "and  date_published>='"+yesterdayDate+"' and user_list like '%;"+uid +";%')";

if(StringUtility.isSet(userSql)) {
  sqlFrom+=" AND "+userSql;
}

sqlFrom+=" and a.source=? order by batch_no desc," + 
"a.date_created asc, a.order_no limit 0,15";
sql = "SELECT * " + sqlFrom;

insertSql = "INSERT INTO tx_user_view2 " + 
"(result_id, date_created, product_type, user_list) " +
"SELECT result_id, SYSDATE(), '"+product+"', ';"+uid+";\r\n'" + sqlFrom;

insertSql += " ON DUPLICATE KEY UPDATE user_list = CONCAT(user_list, VALUES(user_list))";

Answer 1

I cannot say or tune this query for you. But I recommend you construct a final query by looking at the explain plan. The explain plan will tell you whether indexes are being used, how many row scans you are making and probably you can start from there. For explaining a query do

explain select * from dummy_table where x = 'abcd' and y like '%dfs% 
       and z not in ('ab','cd','ef')

Plus, it is better you don't have queries with %like% since it will never hit indexes.

Query performance will further improve if you use a stored procedure. The application layer is separate from a database layer. Stored procedures are native to databases. For example when you want to iterate on a select query which is having 10000 rows, you can do it both via application or a stored procedure. The disadvantage of doing this in application layer is you will have to bring records in chunks (JDBC seamlessly does this for you using resultset object), some sort of data transfer has to happen. This involves latency with respect to memory size, transfer speed b/w app and db etc.

Whereas when you do this in a stored procedure however, the memory is local to database and all the processing can happen there. I am not saying that you should totally isolate application layer from database, but still SPs are far better especially when you are dealing with huge records and complex queries.

Answer 2

Avoid building the String from scratch. Use prepared statements, stored procedures and also add an index to your database.

Answer 3

If you can avoid a nested select query and use a left join instead. Especially a left join can be useful for adjacent tree model or a parent-child relationship.