I'm writing a query to exclude a certain group of employees from a table. Let's say I have Table 1, transaction_details, that has information that I want to select. Table 2, employee_task_associations, maps each employee to a task that they were assigned to on a particular day. It has a field for the employee_id, and a field called map_id which is the id of the different tasks. This table is an associative table so that Tables 1 and 3 can have a many-to-many relationship. Table 3, employee_tasks, has a list of all tasks that an employee can have.
I have written this query, which is functional, but not optimized:
SELECT someInfo FROM transaction_details TD
WHERE TD.employee_id NOT IN
(SELECT employee_id from employee_task_associations ETA
WHERE map_id IN
(SELECT id FROM employee_tasks ET
WHERE ET.taskName = "The task I want to exclude"))
While this works, it will run multiple queries. I want to speed things up by replacing my nested NOT IN and IN statements with JOINS.
I know that I can replace the bottom four lines with the following:
SELECT employee_id FROM employee_task_assocations ETA
LEFT OUTER JOIN employee_tasks ET
ON ETA.map_id = ET.id
WHERE ET.taskName = "The task I want to exclude"
This will return a list of all ids of the employees that have had this task. I want to exclude these from my SELECT statement from transaction_details by using a JOIN instead of a subquery. I have tried using a LEFT OUTER JOIN WHERE ETA.id IS NULL, but this does not work. How can I use a JOIN to exclude certain employees in this case?
You appear to be close on your initial query, but why not have the NOT IN query joined and get distinct employees... Something like
SELECT
TD.someInfo
FROM
transaction_details TD
WHERE
TD.employee_id NOT IN
(SELECT DISTINCT
employee_id
from
employee_task_associations ETA
JOIN employee_tasks ET
ON ETA.map_id = ET.ID
AND ET.taskName = "The task I want to exclude")
You seem to think that an outer join is more performant than a subquery, but that's not true. It all depends on the SQL planner, SQL optimizer, existing indexes, table stats, and definitively on the repertoire of data operators your database engine offers.
Also, you should consider that after parsing, the query enters the transformation phase where the database engine is free to rewrite your query in a more efficient way. This means that behind the scenes your query may actually be executed using outer joins, even if you write subqueries. After that the [rewritten] query enters the query planner, and later the SQL optimizer.
The only way of optimizing it is to get the execution plan of all query options and compare them.
Related
I have some MySQL views which define a number of extra columns based on some relatively straightforward subqueries. The database is also multi-tenanted so each row has a company ID against it.
The problem I have is my views are evaluated for every row before being filtered by the company ID, giving huge performance issues. Is there any way to lazily evaluate the view so the 'where' clause in the outer query applies to the subqueries in the view. Or is there something similar to views that I can use to add the extra fields. I want to calculate them in SQL so the calculated fields can be used for filtering/searching/sorting/pagination.
I've taken a look at the MySQL docs that explain the algorithms available and am aware that the views can't be proccessed as a 'merge' since they contain subqueries.
view
create view companies_view as
select *,
(
select count(id) from company_user where company_user.company_id = companies.id
) as user_count,
(
select count(company_user.user_id)
from company_user join users on company_user.user_id = users.id
where company_user.company_id = companies.id
and users.active = 1
) as active_user_count,
(
select count(company_user.user_id)
from company_user join users on company_user.user_id = users.id
where company_user.company_id = companies.id
and users.active = 0
as inactive_user_count
from companies;
query
select * from companies_view where company_id = 123;
I want the subqueries in the view to be evaluated AFTER applying the 'where company_id = 123' from the main query scope. I can't hard code the company ID in the view since I want the view to be usable for any company ID.
You cannot change the order of evaluation, that is set by the MySQL server.
However, in this particular case you could rewrite the whole sql statement to use joins and conditional counts instead of subqueries:
select c.*,
count(u.id) as user_count,
count(if(u.active=1, 1, null)) as active_user_count,
count(if(u.active=0, 1, null)) as inactive_user_count
from companies c
left join company_user cu on c.id=cu.company_id
left join users u on cu.user_id = u.id
group by c.company_id, ...
If you have MySQL v5.7, then you may not need to add any further fields to the group by clause since the other fields in the companies table would be functionally dependent on the company_id. In earlier versions you may have to list all fields in the companies table (depends on the sql mode settings).
Another way to optimalise such query would be using denormalisation. Your users and company_user table probably have a lot more records than your companies table. You could add a user_count, an active_user_count, and an inactive_user_count field to the companies table, add after insert / update / delete triggers to the company_user table and an after update to the users table and update these 2 fields there. This way you would not need to do the joins and the conditional counts in the view.
It is possible to convince the optimizer to handle a view with scalar subqueries using the MERGE algorithm... you just have to beat the optimizer at its own game.
This will seem quite unorthodox to some, but it is a pattern I use with success in cases where this is needed.
Create a stored function to encapsulate each subquery, then reference the stored function in the view. The optimizer remains blissfully unaware that the functions will invoke the subqueries.
CREATE FUNCTION user_count (_cid INT) RETURNS INT
DETERMINISTIC
READS SQL DATA
RETURN (SELECT count(id) FROM company_user WHERE company_user.company_id = _cid);
Note that a stored function with a single statement does not need BEGIN/END or a change of DELIMITER.
Then in the view, replace the subquery with:
user_count(id) AS user_count,
And repeat the process for each subquery.
The optimizer will then process the view as a MERGE view, select the one appropriate row from the companies table based on the outer WHERE, invoke the functions, and... problem solved.
Lets say I have the following query:
SELECT occurs.*, events.*
FROM occurs
INNER JOIN events ON (events.event_id = occurs.event_id)
WHERE event.event_state = 'visible'
Another way to do the same query and get the same results would be:
SELECT occurs.*, events.*
FROM occurs
INNER JOIN events ON (events.event_id = occurs.event_id
AND event.event_state = 'visible')
My question. Is there a real difference? Is one way faster than the other? Why would I choose one way over the other?
For an INNER JOIN, there's no conceptual difference between putting a condition in ON and in WHERE. It's a common practice to use ON for conditions that connect a key in one table to a foreign key in another table, such as your event_id, so that other people maintaining your code can see how the tables relate.
If you suspect that your database engine is mis-optimizing a query plan, you can try it both ways. Make sure to time the query several times to isolate the effect of caching, and make sure to run ANALYZE TABLE occurs and ANALYZE TABLE events to provide more info to the optimizer about the distribution of keys. If you do find a difference, have the database engine EXPLAIN the query plans it generates. If there's a gross mis-optimization, you can create an Oracle account and file a feature request against MySQL to optimize a particular query better.
But for a LEFT JOIN, there's a big difference. A LEFT JOIN is often used to add details from a separate table if the details exist or return the rows without details if they do not. This query will return result rows with NULL values for b.* if no row of b matches both conditions:
SELECT a.*, b.*
FROM a
LEFT JOIN b
ON (condition_one
AND condition_two)
WHERE condition_three
Whereas this one will completely omit results that do not match condition_two:
SELECT a.*, b.*
FROM a
LEFT JOIN b ON some_condition
WHERE condition_two
AND condition_three
Code in this answer is dual licensed: CC BY-SA 3.0 or the MIT License as published by OSI.
i have two tables as below:
Table 1 "customer" with fields "Cust_id", "first_name", "last_name" (10 customers)
Table 2 "cust_order" with fields "order_id", "cust_id", (26 orders)
I need to display "Cust_id" "first_name" "last_name" "order_id"
to where i need count of order_id group by cust_id like list total number of orders placed by each customer.
I am running below query, however, it is counting all the 26 orders and applying that 26 orders to each of the customer.
SELECT COUNT(order_id), cus.cust_id, cus.first_name, cus.last_name
FROM cust_order, customer cus
GROUP BY cust_id;
Could you please suggest/advice what is wrong in the query?
You issue here is that you have told the database how these two tables are 'connected', or what they should be connected by:
Have a look at this image:
~IMAGE SOURCE
This effectively allows you to 'join' two tables together, and use a query between them.
so you might want to use something like:
SELECT COUNT(B.order_id), A.cust_id, A.first_name, A.last_name
FROM customer A
LEFT JOIN cust_order B //this is using a left join, but an inner may be appropriate also
ON (A.cust_id= B.Cust_id) //what links them together
GROUP BY A.cust_id; // the group by clause
As per your comment requesting some further info:
Left Join (right joins are almost identical, only the other way around):
The SQL LEFT JOIN returns all rows from the left table, even if there are no matches in the right table. This means that if the ON clause matches 0 (zero) records in right table, the join will still return a row in the result, but with NULL in each column from right table. ~Tutorials Point.
This means that a left join returns all the values from the left table, plus matched values from the right table or NULL in case of no matching join predicate.
LEFT joins will be used in the cases where you wish to retrieve all the data from the table in the left hand side, and only data from the right that match.
Execution Time
While the accepted answer in this case may work well in small datasets, it may however become 'heavy' in larger databases. This is because it was not actually designed for this type of operation.
This was the purpose of Joins to be introduced.
Much work in database-systems has aimed at efficient implementation of joins, because relational systems commonly call for joins, yet face difficulties in optimising their efficient execution. The problem arises because inner joins operate both commutatively and associatively. ~Wikipedia
In practice, this means that the user merely supplies the list of tables for joining and the join conditions to use, and the database system has the task of determining the most efficient way to perform the operation. A query optimizer determines how to execute a query containing joins. So, by allowing the dbms to choose the way your data is queried, you can save a lot of time.
Other Joins/Summary
AN INNER JOIN will return data from both tables where the keys in each table match
A LEFT JOIN or RIGHT JOIN will return all the rows from one table and matching data from the other table.
Use a join when you want to query multiple tables.
Joins are much faster than other ways of querying >=2 tables (speed can be seen much better on larger datasets).
You could try this one:
SELECT COUNT(cus_order.order_id), cus.cust_id, cus.first_name, cus.last_name
FROM cust_order cus_order, customer cus
WHERE cus_order.cust_id = cus.cust_id
GROUP BY cust_id;
Maybe an left join will help you
SELECT COUNT(order_id), cus.cust_id, cus.first_name, cus.last_name ]
FROM customer cus
LEFT JOIN cust_order co
ON (co.cust_id= cus.Cust_id )
GROUP BY cus.cust_id;
I have what may be a basic performance question. I've done a lot of SQL queries, but not much in terms of complex inner joins and such. So, here it is:
I have a database with 4 tables, countries, territories, employees, and transactions.
The transactions links up with the employees and countries. The employees links up with the territories. In order to produce a required report, I'm running a PHP script that processes a SQL query against a mySQL database.
SELECT trans.transactionDate, agent.code, agent.type, trans.transactionAmount, agent.territory
FROM transactionTable as trans
INNER JOIN
(
SELECT agent1.code as code, agent1.type as type, territory.territory as territory FROM agentTable as agent1
INNER JOIN territoryTable as territory
ON agent1.zip=territory.zip
) AS agent
ON agent.code=trans.agent
ORDER BY trans.agent
There are about 50,000 records in the agent table, and over 200,000 in the transaction table. The other two are relatively tiny. It's taking about 7 minutes to run this query. And I haven't even inserted the fourth table yet, which needs to relate a field in the transactionTable (country) to a field in the countryTable (country) and return a field in the countryTable (region).
So, two questions:
Where would I logically put the connection between the transactionTable and the countryTable?
Can anyone suggest a way that this can be quickened up?
Thanks.
Your query should be equivalent to this:
SELECT tx.transactionDate,
a.code,
a.type,
tx.transactionAmount,
t.territory
FROM transactionTable tx,
agentTable a,
territoryTable t
WHERE tx.agent = a.code
AND a.zip = t.zip
ORDER BY tx.agent
or to this if you like to use JOIN:
SELECT tx.transactionDate,
a.code,
a.type,
tx.transactionAmount,
t.territory
FROM transactionTable tx
JOIN agentTable a ON tx.agent = a.code
JOIN territoryTable t ON a.zip = t.zip
ORDER BY tx.agent
In order to work fast, you must have following indexes on your tables:
CREATE INDEX transactionTable_agent ON transactionTable(agent);
CREATE INDEX territoryTable_zip ON territoryTable(zip);
CREATE INDEX agentTable_code ON agentTable(code);
(basically any field that is part of WHERE or JOIN constraint should be indexed).
That said, your table structure looks suspicious in a sense that it is joined by apparently non-unique fields like zip code. You really want to join by more unique entities, like agent id, transaction id and so on - otherwise expect your queries to generate a lot of redundant data and be really slow.
One more note: INNER JOIN is equivalent to simply JOIN, there is no reason to type redundant clause.
I have the following query:
SELECT c.*
FROM companies AS c
JOIN users AS u USING(companyid)
JOIN jobs AS j USING(userid)
JOIN useraccounts AS us USING(userid)
WHERE j.jobid = 123;
I have the following questions:
Is the USING syntax synonymous with ON syntax?
Are these joins evaluated left to right? In other words, does this query say: x = companies JOIN users; y = x JOIN jobs; z = y JOIN useraccounts;
If the answer to question 2 is yes, is it safe to assume that the companies table has companyid, userid and jobid columns?
I don't understand how the WHERE clause can be used to pick rows on the companies table when it is referring to the alias "j"
Any help would be appreciated!
USING (fieldname) is a shorthand way of saying ON table1.fieldname = table2.fieldname.
SQL doesn't define the 'order' in which JOINS are done because it is not the nature of the language. Obviously an order has to be specified in the statement, but an INNER JOIN can be considered commutative: you can list them in any order and you will get the same results.
That said, when constructing a SELECT ... JOIN, particularly one that includes LEFT JOINs, I've found it makes sense to regard the third JOIN as joining the new table to the results of the first JOIN, the fourth JOIN as joining the results of the second JOIN, and so on.
More rarely, the specified order can influence the behaviour of the query optimizer, due to the way it influences the heuristics.
No. The way the query is assembled, it requires that companies and users both have a companyid, jobs has a userid and a jobid and useraccounts has a userid. However, only one of companies or user needs a userid for the JOIN to work.
The WHERE clause is filtering the whole result -- i.e. all JOINed columns -- using a column provided by the jobs table.
I can't answer the bit about the USING syntax. That's weird. I've never seen it before, having always used an ON clause instead.
But what I can tell you is that the order of JOIN operations is determined dynamically by the query optimizer when it constructs its query plan, based on a system of optimization heuristics, some of which are:
Is the JOIN performed on a primary key field? If so, this gets high priority in the query plan.
Is the JOIN performed on a foreign key field? This also gets high priority.
Does an index exist on the joined field? If so, bump the priority.
Is a JOIN operation performed on a field in WHERE clause? Can the WHERE clause expression be evaluated by examining the index (rather than by performing a table scan)? This is a major optimization opportunity, so it gets a major priority bump.
What is the cardinality of the joined column? Columns with high cardinality give the optimizer more opportunities to discriminate against false matches (those that don't satisfy the WHERE clause or the ON clause), so high-cardinality joins are usually processed before low-cardinality joins.
How many actual rows are in the joined table? Joining against a table with only 100 values is going to create less of a data explosion than joining against a table with ten million rows.
Anyhow... the point is... there are a LOT of variables that go into the query execution plan. If you want to see how MySQL optimizes its queries, use the EXPLAIN syntax.
And here's a good article to read:
http://www.informit.com/articles/article.aspx?p=377652
ON EDIT:
To answer your 4th question: You aren't querying the "companies" table. You're querying the joined cross-product of ALL four tables in your FROM and USING clauses.
The "j.jobid" alias is just the fully-qualified name of one of the columns in that joined collection of tables.
In MySQL, it's often interesting to ask the query optimizer what it plans to do, with:
EXPLAIN SELECT [...]
See "7.2.1 Optimizing Queries with EXPLAIN"
Here is a more detailed answer on JOIN precedence. In your case, the JOINs are all commutative. Let's try one where they aren't.
Build schema:
CREATE TABLE users (
name text
);
CREATE TABLE orders (
order_id text,
user_name text
);
CREATE TABLE shipments (
order_id text,
fulfiller text
);
Add data:
INSERT INTO users VALUES ('Bob'), ('Mary');
INSERT INTO orders VALUES ('order1', 'Bob');
INSERT INTO shipments VALUES ('order1', 'Fulfilling Mary');
Run query:
SELECT *
FROM users
LEFT OUTER JOIN orders
ON orders.user_name = users.name
JOIN shipments
ON shipments.order_id = orders.order_id
Result:
Only the Bob row is returned
Analysis:
In this query the LEFT OUTER JOIN was evaluated first and the JOIN was evaluated on the composite result of the LEFT OUTER JOIN.
Second query:
SELECT *
FROM users
LEFT OUTER JOIN (
orders
JOIN shipments
ON shipments.order_id = orders.order_id)
ON orders.user_name = users.name
Result:
One row for Bob (with the fulfillment data) and one row for Mary with NULLs for fulfillment data.
Analysis:
The parenthesis changed the evaluation order.
Further MySQL documentation is at https://dev.mysql.com/doc/refman/5.5/en/nested-join-optimization.html
SEE http://dev.mysql.com/doc/refman/5.0/en/join.html
AND start reading here:
Join Processing Changes in MySQL 5.0.12
Beginning with MySQL 5.0.12, natural joins and joins with USING, including outer join variants, are processed according to the SQL:2003 standard. The goal was to align the syntax and semantics of MySQL with respect to NATURAL JOIN and JOIN ... USING according to SQL:2003. However, these changes in join processing can result in different output columns for some joins. Also, some queries that appeared to work correctly in older versions must be rewritten to comply with the standard.
These changes have five main aspects:
The way that MySQL determines the result columns of NATURAL or USING join operations (and thus the result of the entire FROM clause).
Expansion of SELECT * and SELECT tbl_name.* into a list of selected columns.
Resolution of column names in NATURAL or USING joins.
Transformation of NATURAL or USING joins into JOIN ... ON.
Resolution of column names in the ON condition of a JOIN ... ON.
Im not sure about the ON vs USING part (though this website says they are the same)
As for the ordering question, its entirely implementation (and probably query) specific. MYSQL most likely picks an order when compiling the request. If you do want to enforce a particular order you would have to 'nest' your queries:
SELECT c.*
FROM companies AS c
JOIN (SELECT * FROM users AS u
JOIN (SELECT * FROM jobs AS j USING(userid)
JOIN useraccounts AS us USING(userid)
WHERE j.jobid = 123)
)
as for part 4: the where clause limits what rows from the jobs table are eligible to be JOINed on. So if there are rows which would join due to the matching userids but don't have the correct jobid then they will be omitted.
1) Using is not exactly the same as on, but it is short hand where both tables have a column with the same name you are joining on... see: http://www.java2s.com/Tutorial/MySQL/0100__Table-Join/ThekeywordUSINGcanbeusedasareplacementfortheONkeywordduringthetableJoins.htm
It is more difficult to read in my opinion, so I'd go spelling out the joins.
3) It is not clear from this query, but I would guess it does not.
2) Assuming you are joining through the other tables (not all directly on companyies) the order in this query does matter... see comparisons below:
Origional:
SELECT c.*
FROM companies AS c
JOIN users AS u USING(companyid)
JOIN jobs AS j USING(userid)
JOIN useraccounts AS us USING(userid)
WHERE j.jobid = 123
What I think it is likely suggesting:
SELECT c.*
FROM companies AS c
JOIN users AS u on u.companyid = c.companyid
JOIN jobs AS j on j.userid = u.userid
JOIN useraccounts AS us on us.userid = u.userid
WHERE j.jobid = 123
You could switch you lines joining jobs & usersaccounts here.
What it would look like if everything joined on company:
SELECT c.*
FROM companies AS c
JOIN users AS u on u.companyid = c.companyid
JOIN jobs AS j on j.userid = c.userid
JOIN useraccounts AS us on us.userid = c.userid
WHERE j.jobid = 123
This doesn't really make logical sense... unless each user has their own company.
4.) The magic of sql is that you can only show certain columns but all of them are their for sorting and filtering...
if you returned
SELECT c.*, j.jobid....
you could clearly see what it was filtering on, but the database server doesn't care if you output a row or not for filtering.