SQL QUERIES

CS121: Relational Databases

Fall 2017 – Lecture 5
 SQL Queries
2

¨ SQL queries use the SELECT statement

¨ General form is:
SELECT A1, A2, ...
FROM r1, r2, ...
WHERE P;
¤ riare the relations (tables)
¤ Ai are attributes (columns)

¤ P is the selection predicate

¨ Equivalent to: PA1, A2, …(sP(r1 ´ r2 ´ …))

 Ordered Results
3

¨ SQL query results can be ordered by particular

attributes
¨ Two main categories of query results:
¤ “Not ordered by anything”
n Tuples can appear in any order
¤ “Ordered by attributes A1, A2, …”
n Tuples are sorted by specified attributes
n Results are sorted by A1 first
n Within each value of A1, results are sorted by A2
n etc.
¨ Specify an ORDER BY clause at end of SELECT
statement
 Ordered Results (2)
4

¨ Find bank accounts with a balance under $700:

SELECT account_number, balance
+----------------+---------+
FROM account | account_number | balance |
+----------------+---------+
WHERE balance < 700; | A-102 | 400.00 |
| A-101 | 500.00 |
| A-444 | 625.00 |
| A-305 | 350.00 |
¨ Order results in increasing +----------------+---------+

order of bank balance:

SELECT account_number, balance +----------------+---------+
FROM account | account_number | balance |
+----------------+---------+
WHERE balance < 700 | A-305 | 350.00 |
| A-102 | 400.00 |
ORDER BY balance; | A-101 | 500.00 |
| A-444 | 625.00 |
¤ Default order is ascending order +----------------+---------+
 Ordered Results (3)
5

¨ Say ASC or DESC after attribute name to specify order

¤ ASC is redundant, but can improve readability in some cases
¨ Can list multiple attributes, each with its own order
“Retrieve a list of all bank branch details, ordered by branch city, with
each city’s branches listed in reverse order of holdings.”
SELECT * FROM branch
ORDER BY branch_city ASC, assets DESC;

+-------------+-------------+------------+
| branch_name | branch_city | assets |
+-------------+-------------+------------+
| Pownal | Bennington | 400000.00 |
| Brighton | Brooklyn | 7000000.00 |
| Downtown | Brooklyn | 900000.00 |
| Round Hill | Horseneck | 8000000.00 |
| Perryridge | Horseneck | 1700000.00 |
| Mianus | Horseneck | 400200.00 |
| Redwood | Palo Alto | 2100000.00 |
| ... | ... | ... |
 Aggregate Functions in SQL
6

¨ SQL provides grouping and aggregate operations,

just like relational algebra
¨ Aggregate functions:
SUM sums the values in the collection
AVG computes average of values in the collection
COUNT counts number of elements in the collection
MIN returns minimum value in the collection
MAX returns maximum value in the collection
¨ SUM and AVG require numeric inputs (obvious)
 Aggregate Examples
7

¨ Find average balance of accounts at Perryridge

branch
SELECT AVG(balance) FROM account
WHERE branch_name = 'Perryridge';
+--------------+
| AVG(balance) |
+--------------+
| 650.000000 |
+--------------+

¨ Find maximum amount of any loan in the bank

SELECT MAX(amount) AS max_amt FROM loan;
¤ Can name computed values, like usual +---------+
| max_amt |
+---------+
| 7500.00 |
+---------+
 Aggregate Examples (2)
8

¨ This query produces an error:

SELECT branch_name,
MAX(amount) AS max_amt
FROM loan;
¨ Aggregate functions compute a single value from a
multiset of inputs
¤ Doesn’t
make sense to combine individual attributes and
aggregate functions like this
¨ This does work:
SELECT MIN(amount) AS min_amt,
MAX(amount) AS max_amt
+---------+---------+
FROM loan; | min_amt | max_amt |
+---------+---------+
| 500.00 | 7500.00 |
+---------+---------+
 Eliminating Duplicates
9

¨ Sometimes need to eliminate duplicates in SQL

queries
¤ Can use DISTINCT keyword to eliminate duplicates
¨ Example:
“Find the number of branches that currently have loans.”
SELECT COUNT(branch_name) FROM loan;
¤ Doesn’t work, because branches may have multiple loans
¤ Instead, do this:
SELECT COUNT(DISTINCT branch_name) FROM loan;
¤ Duplicates are eliminated from input multiset before
aggregate function is applied
 Computing Counts
10

¨ Can count individual attribute values

COUNT(branch_name)
COUNT(DISTINCT branch_name)
¨ Can also count the total number of tuples
COUNT(*)
¤ If used with grouping, counts total number of tuples in each group
¤ If used without grouping, counts total number of tuples

¨ Counting a specific attribute is useful when:

¤ Need to count (possibly distinct) values of a particular attribute
¤ Cases where some values in input multiset may be NULL
n As before, COUNT ignores NULL values (more on this next week)
 Grouping and Aggregates
11

¨ Can also perform grouping on a relation before

computing aggregates
¤ Specify a GROUP BY A1,A2,... clause at end of query
¨ Example:
“Find the average loan amount for each branch.”
SELECT branch_name, AVG(amount) AS avg_amt
FROM loan GROUP BY branch_name;
¤ First, tuples in loan are +-------------+-------------+
| branch_name | avg_amt |
grouped by branch_name +-------------+-------------+
| Central | 570.000000 |
¤ Then, aggregate functions | Downtown | 1250.000000 |
| Mianus | 500.000000 |
are applied to each group | North Town | 7500.000000 |
| Perryridge | 1400.000000 |
| Redwood | 2000.000000 |
| Round Hill | 900.000000 |
+-------------+-------------+
 Grouping and Aggregates (2)
12

¨ Can group on multiple attributes

¤ Each group has unique values for the entire set of grouping
attributes
¨ Example:
“How many accounts does each customer have at each branch?”
¤ Group by both customer name and branch name

¤ Compute count of tuples in each group

¤ Can write the SQL statement yourself, and try it out

 Grouping and Aggregates (3)
13

¨ Note the difference between relational algebra

notation and SQL syntax
¨ Relational algebra syntax:
G1, G2, …, GnGF1(A1), F2(A2), …, Fm(Am)(E)
¤ Grouping attributes only appear on left of G
¨ SQL syntax:
SELECT G1,G2,..., F1(A1),F2(A2),...
FROM r1,r2,... WHERE P
GROUP BY G1,G2,...
¤ Frequently, grouping attributes are specified in both the
SELECT clause and GROUP BY clause
 Grouping and Aggregates (4)
14

¨ SQL doesn’t require that you specify the grouping

attributes in the SELECT clause
¤ Only requirement is that the grouping attributes are
specified in the GROUP BY clause
¤ e.g. if you only want the aggregated results, could do this:
SELECT F1(A1),F2(A2),...
FROM r1,r2,... WHERE P
GROUP BY G1,G2,...
¨ Also, can use expressions for grouping and aggregates
¤ Example (very uncommon, but also valid):
SELECT MIN(a + b) – MAX(c)
FROM t GROUP BY d * e;
 Filtering Tuples
15

¨ The WHERE clause is applied before any grouping

occurs
SELECT G1,G2,..., F1(A1),F2(A2),...
FROM r1,r2,... WHERE P
GROUP BY G1,G2,...
¤ Translates into relational algebra expression:
P…( G1, G2, …GF1(A1), F2(A2), … (sP(r1 ´ r2 ´ …)))
¤A WHERE clause constrains the set of tuples that
grouping and aggregation are applied to
 Filtering Results
16

¨ To apply filtering to the results of grouping and

aggregation, use a HAVING clause
¤ Exactly like WHERE clause, except applied after
grouping and aggregation
SELECT G1,G2,..., F1(A1),F2(A2),...
FROM r1,r2,... WHERE PW
GROUP BY G1,G2,...
HAVING PH
¤ Translates into:
P…(sPH (G1, G2, …GF1(A1), F2(A2), … (sPW (r1 ´ r2 ´ …))))
 The HAVING Clause
17

¨ The HAVING clause can use aggregate functions in its

predicate
¤ It’s applied after grouping/aggregation, so those values
are available
¤ The WHERE clause cannot do this, of course

¨ Example:
“Find all customers with more than one loan.”
SELECT customer_name, COUNT(*) AS num_loans
FROM borrower GROUP BY customer_name
HAVING COUNT(*) > 1;
+---------------+-----------+
| customer_name | num_loans |
+---------------+-----------+
| Smith | 3 |
+---------------+-----------+
 Nested Subqueries
18

¨ SQL provides broad support for nested subqueries

¤ A SQL query is a “select-from-where” expression
¤ Nested subqueries are “select-from-where” expressions
embedded within another query
¨ Can embed queries in WHERE clauses
¤ Sophisticated selection tests
¨ Can embed queries in FROM clauses
¤ Issuing a query against a derived relation
¨ Can even embed queries in SELECT clauses!
¤ Appeared in SQL:2003 standard; many DBs support this
¤ Makes many queries easier to write, but can be slow too
 Kinds of Subqueries
19

¨ Some subqueries produce only a single result

SELECT MAX(assets) FROM branch;
¤ Called a scalar subquery
¤ Still a relation, just with one attribute and one tuple

¨ Most subqueries produce a relation containing

multiple tuples
¤ Nested queries often produce relation with single attribute
n Very common for subqueries in WHERE clause
¤ Nested queries can also produce multiple-attribute relation
n Very common for subqueries in FROM clause
n Can also be used in the WHERE clause in some cases
 Subqueries in WHERE Clause
20

¨ Widely used:
¤ Directcomparison with scalar-subquery results
¤ Set-membership tests: IN, NOT IN

¤ Empty-set tests: EXISTS, NOT EXISTS

¨ Less frequently used:

¤ Set-comparison tests: ANY, SOME, ALL
¤ Uniqueness tests: UNIQUE, NOT UNIQUE

¨ (Can also use these in the HAVING clause)

 Comparison with Subquery Result
21

¨ Can use scalar subqueries in WHERE clause

comparisons
¨ Example:
¤ Want to find the name of the branch with the smallest
number of assets.
¤ Can easily find the smallest number of assets:
SELECT MIN(assets) FROM branch;
¤ This is a scalar subquery; can use it in WHERE clause:
SELECT branch_name FROM branch
WHERE assets = (SELECT MIN(assets) FROM branch);
+-------------+
| branch_name |
+-------------+
| Pownal |
+-------------+
 Set Membership Tests
22

¨ Can use IN (...) and NOT IN (...) for set

membership tests
¨ Example:
¤ Find customers with both an account and a loan.
¤ Before, did this with a INTERSECT operation
¤ Can also use a set-membership test:
“Select all customer names from depositor relation, that also
appear somewhere in borrower relation.”
SELECT DISTINCT customer_name FROM depositor
WHERE customer_name IN (
SELECT customer_name FROM borrower)
¤ DISTINCT necessary because a customer might appear
multiple times in depositor
 Set Membership Tests (2)
23

¨ IN (...) and NOT IN (...) support subqueries

that return multiple columns (!!!)
¨ Example: “Find the ID of the largest loan at each
branch, including the branch name and the amount of
the loan.”
¤ First, need to find the largest loan at each branch
SELECT branch_name, MAX(amount)
FROM loan GROUP BY branch_name
¤ Use this result to identify the rest of the loan details
SELECT * FROM loan
WHERE (branch_name, amount) IN (
SELECT branch_name, MAX(amount)
FROM loan GROUP BY branch_name);
 Empty-Set Tests
24

¨ Can test whether or not a subquery generates any

results at all
n EXISTS (...)
n NOT EXISTS (...)
¨ Example:
“Find customers with an account but not a loan.”
SELECT DISTINCT customer_name FROM depositor d
WHERE NOT EXISTS (
SELECT * FROM borrower b
WHERE b.customer_name = d.customer_name);
¤ Result includes every customer that appears in depositor
table, that doesn’t also appear in the borrower table.
 Empty-Set Tests (2)
25

“Find customers with an account but not a loan.”

SELECT DISTINCT customer_name FROM depositor d
WHERE NOT EXISTS (
SELECT * FROM borrower b
WHERE b.customer_name = d.customer_name);
¤ Inner query refers to an attribute in outer query’s relation

¨ In general, nested subqueries can refer to enclosing

queries’ relations.
¨ However, enclosing queries cannot refer to the nested
queries’ relations.
 Correlated Subqueries
26

“Find customers with an account but not a loan.”

SELECT DISTINCT customer_name FROM depositor d
WHERE NOT EXISTS (
SELECT * FROM borrower b
WHERE b.customer_name = d.customer_name);

¨ When a nested query refers to an enclosing query’s

attributes, it is a correlated subquery
¤ The inner query must be evaluated once for each tuple
considered by the enclosing query
¤ Generally to be avoided! Very slow.
 Correlated Subqueries (2)
27

¨ Many correlated subqueries can be restated using a join

or a Cartesian product
¤ Often the join operation will be much faster
¤ More advanced DBMSes will automatically decorrelate such
queries, but some can’t…
¨ Certain conditions, e.g. EXISTS/NOT EXISTS,
usually indicate presence of a correlated subquery
¨ If it’s easy to decorrelate the subquery, do that! J
¨ If not, test the query for its performance.
¤ If the database can decorrelate it, you’re done!
¤ If the database can’t decorrelate it, may need to come up
with an alternate formulation.
 Set Comparison Tests
28

¨ Can compare a value to a set of values

¤ Isa value larger/smaller/etc. than some value in the
set?
¨ Example:
“Find all branches with assets greater than at least one
branch in Brooklyn.”
SELECT branch_name FROM branch
WHERE assets > SOME (
SELECT assets FROM branch
WHERE branch_name='Brooklyn');
 Set Comparison Tests (2)
29

¨ General form of test:

attr compare_op SOME ( subquery )
¤ Can use any comparison operation
= SOME is same as IN
¤ ANY is a synonym for SOME
¨ Can also compare a value with all values in a set
¤ Use ALL instead of SOME
<> ALL is same as NOT IN
 Set Comparison Tests (3)
30

¨ Example:
“Find branches with assets greater than all branches in
Brooklyn.”
SELECT branch_name FROM branch
WHERE assets > ALL (
SELECT assets FROM branch
WHERE branch_name='Brooklyn');
¤ Could also write this with a scalar subquery
SELECT branch_name FROM branch
WHERE assets >
(SELECT MAX(assets) FROM branch
WHERE branch_name='Brooklyn');
 Uniqueness Tests
31

¨ Can test whether a nested query generates any

duplicate tuples
¤ UNIQUE (...)
¤ NOT UNIQUE (...)

¨ Not widely implemented

¤ Expensive operation!
¨ Can emulate in a number of ways
¤ GROUP BY ... HAVING COUNT(*) = 1 or
GROUP BY ... HAVING COUNT(*) > 1 is one
approach
 Subqueries in FROM Clause
32

¨ Often need to compute a result in multiple steps

¨ Can query against a subquery’s results
¤ Called a derived relation
¨ A trivial example:
¤A HAVING clause can be implemented as a nested
query in the FROM clause
 HAVING vs. Nested Query
33

“Find all cities with more than two customers living in the city.”
SELECT customer_city, COUNT(*) AS num_customers
FROM customer GROUP BY customer_city
HAVING COUNT(*) > 2;
¨ Or, can write:
SELECT customer_city, num_customers
FROM (SELECT customer_city, COUNT(*)
FROM customer GROUP BY customer_city)
AS counts (customer_city, num_customers)
WHERE num_customers > 2;
¤ Grouping and aggregation is computed by inner query
¤ Outer query selects desired results generated by inner query
 Derived Relation Syntax
34

¨ Subquery in FROM clause must be given a name

¤ Many DBMSes also require attributes to be named
SELECT customer_city, num_customers
FROM (SELECT customer_city, COUNT(*)
FROM customer GROUP BY customer_city)
AS counts (customer_city, num_customers)
WHERE num_customers > 2;
¤ Nested query is called counts, and specifies two
attributes
¤ Syntax varies from DBMS to DBMS…
n MySQL requires a name for derived relations, but doesn’t allow
attribute names to be specified.
 Using Derived Relations
35

¨ More typical is a query against aggregate values

¨ Example:
“Find the largest total account balance of any branch.”
¤ Need to compute total account balance for each branch first.
SELECT branch_name, SUM(balance) AS total_bal
FROM account GROUP BY branch_name;
¤ Then we can easily find the answer:
SELECT MAX(total_bal) AS largest_total
FROM (SELECT branch_name,
SUM(balance) AS total_bal
FROM account GROUP BY branch_name)
AS totals (branch_name, tot_bal);
 Aggregates of Aggregates
36

¨ Always take note when computing aggregates of

aggregates!
“Find the largest total account balance of any branch.”
¤ Two nested aggregates: max of sums
¨ A very common mistake:
SELECT branch_name, SUM(balance) AS tot_bal
FROM account GROUP BY branch_name
HAVING tot_bal = MAX(tot_bal)
¤ A SELECT query can only perform one level of
aggregation
¤ Need a second SELECT to find the maximum total
¤ Unfortunately, MySQL accepts this and returns bogus result
 More Data Manipulation Operations
37

¨ SQL provides many other options for inserting,

updating, and deleting tuples
¨ All commands support SELECT-style syntax
¨ Can insert individual tuples into a table:
INSERT INTO table VALUES (1, 'foo', 50);
¨ Can also insert the result of a query into a table:
INSERT INTO table SELECT ...;
¤ Only constraint is that generated results must have a
compatible schema
 Deleting Tuples
38

¨ SQL DELETE command can use a WHERE clause

DELETE FROM table;
¤ Deletes all rows in the table
DELETE FROM table WHERE ...;
¤ Only deletes rows that satisfy the conditions
¤ The WHERE clause can use anything that SELECT’s
WHERE clause supports
n Nested queries, in particular!
 Updating Tables
39

¨ SQL also has an UPDATE command for modifying

existing tuples in a table
¨ General form:
UPDATE table
SET attr1=val1, attr2=val2, ...
WHERE condition;
¤ Must specify the attributes to update
¤ Attributes being modified must appear in table being
updated (obvious)
¤ The WHERE clause is optional! If unspecified, all rows are
updated.
¤ WHERE condition can contain nested queries, etc.
 Updating Tables (2)
40

¨ Values in UPDATE can be arithmetic expressions

¤ Can refer to any attribute in table being updated
¨ Example:
¤ Add 2% interest to all bank account balances with a
balance of $500 or less.
UPDATE account
SET balance = balance * 1.02
WHERE balance <= 500;
 Review
41

¨ SQL query syntax is very rich

¤ Canstate a wide range of complex queries
¤ Many ways to state a particular query

¨ SQL supports nested queries

¤ Often essential for computing particular results
¤ Can sometimes be very inefficient

¨ SQL also provides similar capability for inserting,

deleting, and updating tables
 Next Time
42

¨ NULL values in SQL

¨ Additional SQL join operations
¤ Natural join
¤ Outer joins

¨ SQL views