Transcript CSS122_WINTER09_LECTURE08

Using Relational Databases and SQL
Lecture 8:
Correlated Subqueries
and the
Data Manipulation Language
Steven Emory
Department of Computer Science
California State University, Los Angeles
SELECT Clause Subqueries
A SELECT clause subquery must return a single
value (not a list or table)
Examples:
SELECT (SELECT 1) + (SELECT 2); -- 3
SELECT (SELECT COUNT(*) FROM Movies); -- 6
SELECT (SELECT * FROM Movies); -- ERROR!!!
SELECT Clause Subqueries
SELECT clause subqueries are good for singlevalue calculations, such as percentages
Example:
-- What percent of accounts are male?
SELECT Clause Subqueries
Example
-- OUTER QUERY
SELECT 100*(X)/(Y);
-- INNER QUERY X = number of male accounts
SELECT COUNT(*) FROM Accounts
WHERE Gender = 'M';
-- INNER QUERY Y = number of total accounts
SELECT COUNT(*) FROM Accounts;
-- SOLUTION
SELECT 100*(SELECT COUNT(*) FROM
Accounts WHERE Gender = 'M')/(SELECT
COUNT(*) FROM Accounts);
SELECT Clause Subqueries
A SELECT clause subquery is even more useful
when the outer query and inner query are
correlated (the inner query is dependent on data
from the outer query)
Correlated Subqueries
Previous subqueries have been non-correlated.
non-correlated means ‘no dependencies’
which means you can run the inner query separately
Correlated subqueries are inner queries that are
‘dependent’ on data from outer queries.
correlated means ‘with dependencies’
which means you can’t run the inner query separately
the result of the inner query ‘depends on’ data given
to it from the outer query
Correlated Subqueries
Some FROM clause subquery problems can be
rewritten using correlated subqueries in the
SELECT clause.
Let’s try an example:
List each movie title along with the number of ratings
and the number of genres for that movie.
+--------------------------------+---------+--------+
| Title
| Ratings | Genres |
+--------------------------------+---------+--------+
| Star Trek: Generations
|
10 |
4 |
| X-Men
|
12 |
4 |
| X-Men: The Last Stand
|
12 |
4 |
| Things We Lost in the Fire
|
9 |
1 |
| The X Files
|
12 |
5 |
| The X Files: I Want to Believe |
11 |
3 |
+--------------------------------+---------+--------+
Correlated Subqueries
Non-correlated solution:
SELECT Title, X.Ratings, Y.Genres
FROM Movies M LEFT JOIN (SELECT MovieID,
COUNT(Rating) AS Ratings FROM Ratings
GROUP BY MovieID) X ON M.MovieID =
X.MovieID LEFT JOIN (SELECT MovieID,
COUNT(Genre) AS Genres FROM
XRefGenresMovies GROUP BY MovieID) Y ON
M.MovieID = Y.MovieID;
Take out the inner queries and try running them. Both
run because they are independent of the outer query!
Correlated Subqueries
Correlated setup:
-- OUTER QUERY
SELECT Title, (X) AS Ratings, (Y) AS Genres
FROM Movie M;
-- INNER QUERY X
SELECT COUNT(Rating)
FROM Ratings
WHERE MovieID = M.MovieID;
-- INNER QUERY Y
SELECT COUNT(Genre)
FROM XRefGenresMovies
WHERE MovieID = M.MovieID;
Correlated Subqueries
Correlated solution (dependencies are underlined):
SELECT Title, (SELECT COUNT(Rating)
FROM Ratings WHERE MovieID = M.MovieID) AS
Ratings, (SELECT COUNT(Genre) FROM
XRefGenresMovies WHERE MovieID = M.MovieID)
AS Genres FROM Movies M;
Take out the inner queries and try running them. They
won’t run because they are dependent on the MovieID
attribute from the outer query.
Correlated Subqueries
So which one do you choose? Subqueries in the
FROM clause or correlated subqueries in the
SELECT clause?
Whichever one runs faster!
Our database is too small to do any real testing.
Notice the correlated version is shorter and looks nicer.
Data Manipulation Language
DML for short
Contains commands for modifying table data
Insertion commands (INSERT INTO)
Deletion commands (DELETE)
Update commands (UPDATE)
Not a query
Queries extract data from the database
Commands do not extract data from the database
Before We Start
When modifying the database data, you are going
to mess up because nobody is perfect
If you mess up there are two ways to restore the
original database:
Remove and restore the tables
Use transactions (use BEGIN and ROLLBACK)
I prefer using BEGIN and ROLLBACK
Use BEGIN before entering DML commands
Use ROLLBACK to undo all changes
USE COMMIT to accept all changes
Transactions
ACID
Atomicity
Consistency
Isolation
Durability
SQL Keywords
BEGIN/START TRANSACTION
COMMIT
ROLLBACK
Inserting Records
Two syntaxes:
INSERT INTO
Insert one record at a time
INSERT SELECT
Insert one or more records at a time
Inserting Records
INSERT INTO Syntax
-- Form #1: Insert whole record.
INSERT INTO tablename
VALUES(value1, value2, ..., valuen);
-- Form #2: Insert partial record. Non-specified
fieldnames are assigned default values.
INSERT INTO tablename(field1, field2, ..., fieldn)
VALUES(value1, value2, ..., valuen);
IGNORE
You can use the IGNORE keyword between INSERT
and INTO to suppress duplicate error messages.
Inserting Records
INSERT SELECT Syntax
-- Form #1: Insert whole record.
INSERT INTO destination_table
SELECT field1, field2, ..., fieldn
FROM source_tables
WHERE conditions;
-- Form #2: Insert partial record. Non-specified
fieldnames are assigned default values.
INSERT INTO destination_table(df1, df2, ..., dfn)
SELECT sf1, sf2, ..., sfn
FROM source_tables
WHERE conditions;
INSERT INTO Example
Example:
Add Kung Fu Panda into the database.
INSERT INTO Movies
VALUES(7, 'Kung Fu Panda', '2008-06-06', 'G', 92,
'USA', 'English', (SELECT CompanyID FROM
Companies WHERE Name = 'Dreamworks
Pictures'));
INSERT INTO and Subqueries
As in the previous example, subqueries in the
INSERT command work, but only if the update
table and the subquery table are different.
This rule only applies to MySQL, other database
management systems may behave differently.
INSERT SELECT Example
Example:
Associate all movies that have ‘The X Files’ anywhere
in the title with the romance genre.
INSERT IGNORE INTO XRefGenresMovies
SELECT MovieID, 'Romance'
FROM Movies
WHERE Title LIKE '%The X Files%';
Deleting Records
Deletes one or more rows from a table
Deletes all rows without WHERE condition
Two syntaxes
Single-Table DELETE Syntax
Multi-Table DELETE Syntax
Single-Table DELETE Syntax
Deletes one or more rows from a table
Deletes all rows without WHERE condition
Syntax:
DELETE
FROM tablename
WHERE conditions;
Single-Table DELETE Syntax
Examples:
Delete all ratings.
Delete all ratings by semory.
Single-Table DELETE Syntax
Solutions:
-- Delete all ratings.
DELETE FROM Ratings;
-- Delete all ratings by semory.
DELETE FROM Ratings
WHERE AccountID = (SELECT AccountID
FROM Accounts WHERE Username = 'semory');
Multi-Table DELETE Syntax
Deletes rows from multiple tables
You must be very cautious or else you may delete
something you didn’t want to delete
Syntax:
DELETE T1, T2, ..., Tn
FROM T1 JOIN T2 JOIN ... JOIN Tn
WHERE conditions;
Note: If you use table alias in the FROM clause, you
must use the alias in the DELETE clause as well (see
examples later on).
Multi-Table DELETE Syntax
Examples:
-- Delete all ratings by semory (use multi-table delete
syntax instead of using single-table delete syntax with
a subquery).
-- Delete all directors from the database (from both
the People and XRefDirectorsMovies tables).
Multi-Table DELETE Syntax
Examples:
-- Delete all ratings by semory.
DELETE Ratings FROM Accounts JOIN Ratings
USING(AccountID) WHERE Username = 'semory';
-- Delete all ratings by semory (alternate).
DELETE R
FROM Accounts A JOIN Ratings R
USING(AccountID) WHERE Username = 'semory';
-- Delete all directors from the database.
DELETE P, D
FROM People P JOIN XRefDirectorsMovies D
ON P.PersonID = D.DirectorID;
Multi-Table DELETE Syntax
There is a big problem in the last example.
Multi-table delete can lead to orphaned records if
misused (a foreign key with no primary key).
DELETE P, D
FROM People P JOIN XRefDirectorsMovies D
ON P.PersonID = D.DirectorID;
Jonathan Frakes is both an actor and a director.
If we delete him from the People and
XRefDirectorsMovies tables, we orphan him in the
XRefActorsMovies and Spouses table!
Multi-Table DELETE Syntax
Therefore, be careful what you delete!
If you delete a record, and that record is referenced
somewhere else, you have an orphaned record!
No real good solution exists in MySQL.
Best solution is to only delete records from the
XRefDirectorsMovies table and leave the People
table alone since these people may be referenced
somewhere else.
DELETE FROM XRefDirectorsMovies;
Updating Records
To update existing records, you may use one of
the following syntaxes:
Single-table syntax.
Multi-table equi-join syntax.
Only equi-join is supported.
You may not use any other join syntax (JOIN ON, JOIN
USING, etc.)
Updating Records
To update existing records:
-- Single-table syntax.
UPDATE [IGNORE] tablename
SET field1 = value1, field2 = value2, ...
WHERE conditions;
-- Multi-table equi-join syntax.
UPDATE [IGNORE] tablename1, tablename2, ...
SET field1 = value1, field2 = value2, ...
WHERE conditions;
-- Multi-table subquery syntax.
UPDATE [IGNORE] tablename
SET field1 = subquery1, field2 = subquery2, ...
WHERE conditions;
Updating Records
You may use the IGNORE keyword immediately
after UPDATE to ignore errors when an update
produces duplicate primary keys.
Example:
-- ERROR! BEEP!
UPDATE Movies
SET MovieID = 1
WHERE MovieID = 2;
-- Error will be ignored!
UPDATE IGNORE Movies
SET MovieID = 1
WHERE MovieID = 2;
Updating Records
Examples:
-- The user ojisan has decided to change his
username to uncle_steve. Update the database to
reflect this change.
-- The user colderstone has decided to change his
username and password to jackstone and
slsev0812z3, respectively. Update the database to
reflect this change.
-- Dreamworks Pictures went bankrupt and was
bought out by Paramount Pictures. Modify all
CompanyIDs in the Movies table to reflect this
change.
Updating Records
Examples:
-- The user ojisan has decided to change his
username to uncle_steve. Update the database to
reflect this change.
UPDATE Accounts
SET Username = 'uncle_steve'
WHERE Username = 'ojisan';
Updating Records
Examples:
-- The user colderstone has decided to change his
username and password to jackstone and slsev0812z3,
respectively. Update the database to reflect this change.
UPDATE Accounts
SET Username = 'jackstone', Password = 'slsev0812z3'
WHERE Username = 'colderstone';
Updating Records
Examples:
-- Dreamworks Pictures went bankrupt and was
bought out by Paramount Pictures. Modify all
CompanyIDs in the Movies table to reflect this
change.
UPDATE Movies
SET CompanyID = (SELECT CompanyID FROM
Companies WHERE Name = 'Paramount Pictures')
WHERE CompanyID = (SELECT CompanyID
FROM Companies WHERE Name = 'Dreamworks
Pictures');