The SQL Standard

•  SQL – Structured Query Language

An international standard (ANSI, ISO) that specifies how
§  a relational schema is created
§  data is inserted / updated in the relations
§  data is queried
§  transactions are started and stopped
§  programs access data in the relations
§  and a host of other things are done

•  Every relational database management system (RDBMS) is

required to support / implement the SQL standard.
•  RDBMS vendors may give additional features
•  Downside of using vendor-specific features - portability

Prof P Sreenivasa Kumar 1

Department of CS&E, IITM
 History of SQL
SEQUEL
§  developed by IBM in early 70’s
§  relational query language as part of System-R project at
IBM San Jose Research Lab.
§  the earliest version of SQL

SQL evolution
§  SQL- 86/89
§  SQL- 92 - SQL2
§  SQL- 99/03 - SQL3
(includes object relational features)
And the evolution continues .....

Disclaimer: This module covers only important principles of SQL

Prof P Sreenivasa Kumar 2
Department of CS&E, IITM
 Components of SQL Standard(1/2)
§  Data Definition Language (DDL)
Specifies constructs for schema definition, relation definition,
integrity constraints, views and schema modification.

§  Data Manipulation Language (DML)

Specifies constructs for inserting, updating and querying the
data in the relational instances ( or tables ).

§  Embedded SQL and Dynamic SQL

Specifies how SQL commands can be embedded in a high-level
host language such as C, C++ or Java for programmatic access
to the data.

Prof P Sreenivasa Kumar 3

Department of CS&E, IITM
 Components of SQL Standard(2/2)
§  Transaction Control
Specifies how transactions can be started / stopped, how a set
of concurrently executing transactions can be managed.

§  Authorization
Specifies how to restrict a user / set of users to access only
certain parts of data, perform only certain types of queries etc.

Prof P Sreenivasa Kumar 4

Department of CS&E, IITM
 Data Definition in SQL
Defining the schema of a relation
create table r ( attributeDefinition-1, attributeDefinition-2,…,
name of the attributeDefinition-n, [integrityConstraints-1],
relation
[integrityConstraints-2],…,[integrityConstraints-m])
Attribute Definition –
attribute-name domain-type [NOT NULL] [DEFAULT v]
E.g.:
create table example1 ( A char(6) not null default “000000”,
B int, C char(1) default “F” );

Prof P Sreenivasa Kumar 5

Department of CS&E, IITM
 Domain Types in SQL-92 (1/2)
§  Numeric data types
•  integers of various sizes – INT, SMALLINT
•  real numbers of various precision – REAL, FLOAT,
DOUBLE PRECISION
•  formatted numbers – DECIMAL ( i, j ) or NUMERIC ( i, j )
i – total number of digits ( precision )
j – number of digits after the decimal point ( scale )
§  Character string data types
•  fixed length – CHAR(n) – n: no. of characters
•  varying length – VARCHAR(n) – n: max.no. of characters
§  Bit string data types
•  fixed length – BIT(n)
•  varying length – BIT VARYING(n)

Prof P Sreenivasa Kumar 6

Department of CS&E, IITM
 Domain Types in SQL-92 (2/2)
§  Date data type
DATE type has 10 position format – YYYY-MM-DD
§  Time data type
TIME type has 8 position format – HH : MM : SS
§  Others
There are several more data types whose details are
available in SQL reference books

Prof P Sreenivasa Kumar 7

Department of CS&E, IITM
Specifying Integrity Constraints in SQL
Also called Table Constraints
Included in the definition of a table
Key constraints
PRIMARY KEY (A1,A2,…,Ak)
specifies that {A1,A2,…,Ak} is the primary key of the table

UNIQUE (B1,B2,…,Bk)
specifies that {B1,B2,…,Bk} is a candidate key for the table

There can be more than one UNIQUE constraint but only one
PRIMARY KEY constraint for a table.

Prof P Sreenivasa Kumar 8

Department of CS&E, IITM
 Specifying Referential Integrity Constraints
FOREIGN KEY (A1) REFERENCES r2 (B1)

§  specifies that attribute A1 of the table being defined, say r1, is a

foreign key referring to attribute B1 of table r2

§  recall that this means:

each value of column A1 is either null or is one of the
values appearing in column B1 of r2

Prof P Sreenivasa Kumar 9

Department of CS&E, IITM
Specifying What to Do if RIC Violation Occurs
RIC violation
§  can occur if a referenced tuple is deleted or modified
§  action can be specified for each case using qualifiers
ON DELETE or ON UPDATE
Actions
§  three possibilities can be specified
SET NULL, SET DEFAULT, CASCADE
§  these are actions to be taken on the referencing tuple
§  SET NULL – foreign key attribute value to be set null
§  SET DEFAULT – foreign key attribute value to be set to its
default value
§  CASCADE – delete the referencing tuple if the referenced
tuple is deleted or update the FK attribute if the
referenced tuple is updated
Prof P Sreenivasa Kumar 10
Department of CS&E, IITM
Table Definition Example
create table students (
rollNo char(8) not null,
name varchar(15) not null,
degree char(5),
year smallint,
sex char not null,
deptNo smallint,
advisor char(6),
primary key(rollNo),
foreign key(deptNo) references
department(deptId)
on delete set null on update cascade,
foreign key(advisor) references
professor(empId)
on delete set null on update cascade
);

Prof P Sreenivasa Kumar 11

Department of CS&E, IITM
Modifying a Defined Schema
ALTER TABLE command can be used to modify a schema
Adding a new attribute
ALTER table student ADD address varchar(30);
Deleting an attribute
§  need to specify what needs to be done about views or
constraints that refer to the attribute being dropped
§  two possibilities
CASCADE – delete the views/constraints also
RESTRICT – do not delete the attributes if there are some
views/constraints that refer to it.
§  ALTER TABLE student DROP degree RESTRICT
Similarly, an entire table definition can be deleted

Prof P Sreenivasa Kumar 12

Department of CS&E, IITM
 Data Manipulation in SQL
Basic query syntax
select A1,A2,…,Am a set of attributes
from relations R1,…,Rp that are
from R1,R2,…,Rp required in the output table.
the set of tables that
where θ contain the relevant
tuples to answer the query.
a boolean predicate that
specifies when a combined
tuple of R1,…,Rp contributes
to the output.
Equivalent to: Assuming that each attribute
π A1 ,A2 ,....An (σθ (R1×R 2 ×.....×R p )) name appears exactly once
in the table.
Prof P Sreenivasa Kumar 13
Department of CS&E, IITM
 Meaning of the Basic Query Block
§  The cross product M of the tables in the from clause
would be considered.
Tuples in M that satisfy the condition θ are selected.
For each such tuple, values for the attributes A1,A2,….,Am
( mentioned in the select clause) are projected.

§  This is a conceptual description

- in practice more efficient methods are employed for
evaluation.

§  The word select in SQL should not be confused with select

operation of relational algebra.

Prof P Sreenivasa Kumar 14

Department of CS&E, IITM
SQL Query Result
The result of any SQL query
§  a table with select clause attributes as column names.
§  duplicate rows may be present.
- differs from the definition of a relation.
§  duplicate rows can be eliminated by specifying DISTINCT
keyword in the select clause, if necessary.
SELECT DISTINCT name
FROM student WHERE …

§  duplicate rows are essential while computing aggregate

functions ( average, sum etc ).
§  removing duplicate rows involves additional effort and is
done only when necessary.
Prof P Sreenivasa Kumar 15
Department of CS&E, IITM
Example Relational Scheme with RIC’s shown
student (rollNo, name, degree, year, sex, deptNo, advisor)

department (deptId, name, hod, phone)

professor (empId, name, sex, startYear, deptNo, phone)

course (courseId, cname, credits, deptNo)

enrollment (rollNo, courseId, sem, year, grade)

teaching (empId, courseId, sem, year, classRoom)

preRequisite (preReqCourse, courseID)

Prof P Sreenivasa Kumar 16

Department of CS&E, IITM
Example Queries Involving a Single Table
Get the rollNo, name of all women students in the
dept no. 5.

select rollNo, name

from student
where sex = ‘F’ and deptNo = 5;

Get the employee Id, name and phone number of

professors in the CS dept (deptNo = 3) who have
joined after 1999.

select empId, name, phone

from professor
where deptNo = 3 and startYear > 1999;

Prof P Sreenivasa Kumar 17

Department of CS&E, IITM
Examples Involving Two or More Relations (1/2)
Get the rollNo, name of students in the CSE
dept (deptNo = 3)along with their advisor’s
name and phone number.

select rollNo, s.name, f.name as advisorName,

phone as advisorPhone attribute
from student as s, professor as f renaming in
where s.advisor = f.empId and the output
s.deptNo = 3;
table aliases are required
if an attribute name
table aliases are appears in more than
used to disambiguate one table.
the common attributes Also when same relation
appears twice in the from
clause.

Prof P Sreenivasa Kumar 18

Department of CS&E, IITM
Examples Involving Two or More Relations (2/2)
Get the names, employee ID’s, phone numbers
of professors in CSE dept who joined
before 1995.

select empId, f.name, f.phone

from professor as f, department as d
where f.deptNo = d.deptId and
d.name = ‘CSE’ and
f.startYear < 1995

Prof P Sreenivasa Kumar 19

Department of CS&E, IITM
Nested Queries or Subqueries
While dealing with certain complex queries
§  beneficial to specify part of the computation as a
separate query & make use of its result to formulate
the main query.
§  such queries – nested / subqueries.

Using subqueries
§  makes the main query easy to understand / formulate
§  sometimes makes it more efficient also
•  sub query result can be computed once and
used many times.
•  not the case with all subqueries.

Prof P Sreenivasa Kumar 20

Department of CS&E, IITM
Nested Query Example
Get the rollNo, name of students who have
a lady professor as their advisor.
IN Operator: One of the
ways of making use of
the subquery result
select s.rollNo, s.name Subquery computes
from student s the empId’s of
where s.advisor IN lady professors
(select empId
from professor
where sex = ‘F’);

NOT IN can be used in the above query to get details of students

who don’t have a lady professor as their advisor.

Prof P Sreenivasa Kumar 21

Department of CS&E, IITM
Set Comparison Operators
SQL supports several operators to deal with subquery results or
in general with collection of tuples.

Combination of { =, <, ≤, ≥, >, < > } with keywords

{ ANY, ALL } can be used as set comparison operators.

Get the empId, name of the senior-most Professor(s):

select p.empId, p.name

from professors p
where p.startYear <= ALL ( select distinct startYear
from professor );

Prof P Sreenivasa Kumar 22

Department of CS&E, IITM
Semantics of Set Comparison Operators
op is one of <, ≤, >, ≥, =, < >
§  v op ANY S
true if for some member x of S, v op x is true
false if for no member x of S, v op x is true
S is a subquery
§  v op ALL S
true if for every member x of S, v op x is true
false if for some member x of S, v op x is not true

§  IN is equivalent to = ANY
NOT IN is equivalent to
< > ALL

§  v is normally a single attribute, but while using IN or

NOT IN it can be a tuple of attributes
Prof P Sreenivasa Kumar 23
Department of CS&E, IITM
Correlated and Uncorrelated Nested Queries
If the nested query result is independent of the current tuple
being examined in the outer query,
nested query is called uncorrelated,
otherwise, nested query is called correlated.

Uncorrelated nested query

§  nested query needs to be computed only once.

Correlated nested query

§  nested query needs to be re-computed for each row
examined in the outer query.

Prof P Sreenivasa Kumar 24

Department of CS&E, IITM
Example of a Correlated Subquery
Get the roll number and name of students
whose gender is same as their advisor’s.

select s.rollNo, s.name

from student s
where s.sex = ALL ( select f.sex
from professor f
where f.empId = s.advisor );

Prof P Sreenivasa Kumar 25

Department of CS&E, IITM
The EXISTS Operator
Using EXISTS, we can check if a subquery result is non-empty

EXISTS( S ) is true if S has at least one tuple / member

is false if S contain no tuples

Get the employee Id and name of professors

who advise at least one women student.
a correlated
select f.empId, f.name subquery
from professors f
where EXISTS ( select s.rollNo
from student s
where s.advisor = f.empId and
s.sex = ‘F’ );

SQL does not have an operator for universal quantification.

Prof P Sreenivasa Kumar 26
Department of CS&E, IITM
The NOT EXISTS Operator
Obtain the department Id and name of
departments that do not offer any 4 credit courses.
a correlated
select d.deptId, d.name subquery
from department d
where NOT EXISTS ( select courseId
from course c
where c.deptNo = d.deptId and
c.credits = 4 );

Queries with existentially quantified predicates can be easily

specified using EXISTS operator.

Queries with universally quantified predicates can only be

specified after translating them to use existential quantifiers.
Prof P Sreenivasa Kumar 27
Department of CS&E, IITM
 Example Involving Universal Quantifier
Determine the students who are enrolled for every
course taught by Prof Ramanujam. Assume that Prof
Ramanujam teaches at least one course.

As SQL does not have universal quantifier, we will rewrite the query this way:

Determine the student(s) who are st there does not

exist a course taught by Prof Ramanujam which is not
enrolled by the student.

Prof P Sreenivasa Kumar 28

Department of CS&E, IITM
Same query expressed in TRC
Determine the students who are enrolled for every
course taught by Prof Ramanujam. Assume that Prof
Ramanujam teaches at least one course.

1.  {s.rollNo | student (s) ^

2.  (∀c)(course (c) ^
3.  ((∃t),(∃p)( teaching(t) ^ professor(p) ^
4.  t.courseId = c.courseId ^
5.  p.name = “Ramanujam” ^
6.  p.empId = t.empId )) →
7.  (∃e) (enrollment(e) ^
8.  e.courseId = c.courseId ^
9.  e.rollNo = s.rollNo)
10.  )
11.  }

Prof P Sreenivasa Kumar, 29

Department of CS&E, IITM.
An Example Involving the Universal Quantifier
Determine the students who are enrolled for every
course taught by Prof Ramanujam. Assume that Prof
Ramanujam teaches at least one course.
select s.rollNo, s.name
from student s
where NOT EXISTS ( select t.*
from teaching t, professor p
where t.empId = p.empId and
p.name = “Ramanujam” and
NOT EXISTS
(select e.*
from enrollment e
where e.courseId = t.courseId
and e.rollNo = s.rollNo)
);

Prof P Sreenivasa Kumar 30

Department of CS&E, IITM
Another Example Involving the Universal Quantifier
Determine the students who have obtained either S or A
grade in all the pre-requisite courses of the course
CS7890. It is known that CS7890 has at least one pre-
requisite.

select s.rollNo, s.name

from student s
where NOT EXISTS(select *
from preRequisite p
where p.courseId = “CS7890” and
NOT EXISTS
(select *
from enrollment e
where e.courseId = p.preReqcourse
and e.rollNo = s.rollNo and
e.grade = “S” or e.grade = “A”)
);
Prof P Sreenivasa Kumar 31
Department of CS&E, IITM
Missing where Clause
If the where clause in an SQL query is not specified, it is treated
as - the where condition is true for all tuple combinations.

§  Essentially no filtering is done on the cross product of from

clause tables.

Get the name and contact phone of all Departments.

select name, phone

from department

Prof P Sreenivasa Kumar 32

Department of CS&E, IITM
 Union, Intersection and Difference Operations

§  In SQL, using operators UNION, INTERSECT and EXCEPT,

one can perform set union, intersection and difference
respectively.

§  Results of these operators are sets –

i.e duplicates are automatically removed.

§  Operands need to be union compatible and also have same

attribute names in the same order.

Prof P Sreenivasa Kumar 33

Department of CS&E, IITM
Example using UNION
Obtain the roll numbers of students who are currently
enrolled for either CS2300 or CS2320 courses.

(SELECT rollNo
FROM enrollment
WHERE courseId = ‘CS2300’ and
sem = ‘odd’ and year = ‘2019’ ) UNION
(SELECT rollNo
FROM enrollment
WHERE courseId = ‘CS2320’ and
sem = ‘odd’ and year = ‘2019’ );

Equivalent to:
(SELECT rollNo
FROM enrollment
WHERE (courseId = ‘CS2300’ or courseID = ‘CS2320’)
and sem = ‘odd’ and year = ‘2019’ )
Prof P Sreenivasa Kumar 34
Department of CS&E, IITM
Example using INTERSECTION
Obtain the roll numbers of students who are currently
enrolled for both CS230 and CS232 Courses.

(select rollNo
from enrollment
where courseId = ‘CS2300’ and
sem = ‘odd’ and year = ‘2019’ )

INTERSECT

(select rollNo
from enrollment
where courseId = ‘CS2320’ and
sem = ‘odd’ and year = ‘2019’;

Prof P Sreenivasa Kumar 35

Department of CS&E, IITM
Example using EXCEPT
Obtain the roll numbers of students who are
currently not enrolled for CS2300 course.

(SELECT rollNo
FROM enrollment
WHERE sem = ‘odd’ and year = ‘2019’)
EXCEPT
(SELECT rollNo
FROM enrollment
WHERE courseId = ‘CS2300’ and
sem = ‘odd’ and year = ‘2019’);

Prof P Sreenivasa Kumar 36

Department of CS&E, IITM
Aggregation of Data

Data analysis
§  to get info on summary and trends in certain attributes
§  need for computing aggregate values for data
§  total value, average value etc

Aggregate functions in SQL

§  five aggregate function are provided in SQL
§  AVG, SUM, COUNT, MAX, MIN
§  can be applied to any column of a table
§  can be used in the select clause of SQL queries

Prof P Sreenivasa Kumar 37

Department of CS&E, IITM
 Aggregate functions Optional
keyword
§  AVG ( [DISTINCT]A):
computes the average of (distinct) values in column A

§  SUM ( [DISTINCT]A):
computes the sum of (distinct) values in column A

§  COUNT ( [DISTINCT]A):
computes the number of (distinct) values in column A or no.
of tuples in result

§  MAX (A): computes the maximum of values in column A

§  MIN (A): computes the minimum of values in column A

Prof P Sreenivasa Kumar 38

Department of CS&E, IITM
Examples involving aggregate functions (1/2)
Suppose data about GATE exam in a particular year is available as
a table with schema
gateMarks(regNo,name,sex,branch,city,state,marks)

Obtain the total number of students who have taken

GATE in CS and their average marks
Select count(regNo) as CsTotal, avg(marks) as CsAvg
from gateMarks Output
where branch = ‘CS’ CStotal CSavg

Get the maximum, minimum and average marks obtained

by Students from the city of Hyderabad
Select max(marks), min(marks), avg(marks)
from gateMarks
where city = ‘Hyderabad’;

Prof P Sreenivasa Kumar 39

Department of CS&E, IITM
Examples involving aggregate functions (2/2)
Get the names of students who obtained the
maximum marks in the branch of EC
Select name, max(marks)
from gateMarks Will not work
where branch = ‘EC’

Only aggregate functions can be specified here. It does not

make sense to include normal attributes ! (unless they are
grouping attributes – to be seen later)
Select regNo, name, marks Correct way of
from gateMarks specifying the query
where branch = ‘EC’ and marks = ANY
(select max(marks)
from gateMarks
where branch = ‘EC’);
Prof P Sreenivasa Kumar 40
Department of CS&E, IITM
 Date Aggregation and Grouping
Grouping
§  Partition the set of tuples in a relation into groups based on
certain criteria and compute aggregate functions for each group
§  All tuples that agree on a set of attributes (i.e have the same
value for each of these attributes ) are put into a group
Called the grouping
attributes

§  The specified aggregate functions are computed for each group

§  Each group contributes one tuple to the output
§  All the grouping attributes must also appear in the select clause
§  the result tuple of the group is listed along with the values of the
grouping attributes of the group

Prof P Sreenivasa Kumar 41

Department of CS&E, IITM
 Examples involving grouping(1/2)
Determine the maximum of the GATE CS marks obtained by
students in each city, for all Cities. Assume 4 cities
exist - Hyderabad, Chennai, Mysore and Bangalore.

Select city, max(marks) as maxMarks

from gateMarks
Grouping attributes
where branch = ‘CS’ must appear in the
group by city; Grouping select clause
attribute
Result:
City maxMarks
Hyderabad 87
Chennai 88
Mysore 90
Bangalore 86

Prof P Sreenivasa Kumar 42

Department of CS&E, IITM
Examples involving grouping(2/2)
In the University database, for each department,
obtain the name, deptId and the total number of four
credit courses offered by the department

Select deptId, name, count(*) as totalCourses

from department, course
where deptId = deptNo and credits = 4
group by deptId, name;

Prof P Sreenivasa Kumar 43

Department of CS&E, IITM
Having clause
After performing grouping, is it possible to report information
about only a subset of the groups ?
§  Yes, with the help of having clause which is always used in
conjunction with Group By clause

Report the total enrollment in each course in the

even semester of 2014; include only the courses with
a minimum enrollment of 10.

Select courseId, count(rollNo) as Enrollment

from enrollment
where sem = even and year = 2014
group by courseId
having count(rollNo) ≥ 10;

Prof P Sreenivasa Kumar 44

Department of CS&E, IITM
Where clause versus Having clause
•  Where clause
•  Performs tests on rows and eliminates rows not satisfying
the specified condition
•  Performed before any grouping of rows is done
•  Having clause
•  Always performed after grouping
•  Performs tests on groups and eliminates groups not
satisfying the specified condition
•  Tests can only involve grouping attributes and aggregate
functions
Select courseId, count(rollNo) as Enrollment
from enrollment
where sem = 2 and year = 2014
group by courseId
having count(rollNo) ≥ 10;

Prof P Sreenivasa Kumar 45

Department of CS&E, IITM
String Operators in SQL
§  Specify strings by enclosing them in single quotes
e.g., ‘Chennai’

Common operations on strings –

•  Pattern matching – using ‘LIKE’ comparison operator
•  specify patterns using special characters –
•  Character ‘%’ (percent) matches any Substring
e.g., ‘Ram%’ matches any string starting with “Ram”
•  Character ‘_’ (underscore) matches any single character
e.g., (a) ‘_ _ _ nagar’ matches with any string ending
with “nagar”, with any 3 characters before that.
(b) ‘_ _ _ _’ matches any string with exactly four
characters

Prof P Sreenivasa Kumar 46

Department of CS&E, IITM
Using the ‘LIKE’ operator
Obtain roll numbers and names of all students
whose names end with ‘Mohan’

Select rollNo, name

from student
where name like ‘%Mohan’;

§  Patterns are case sensitive.

§  Special characters (percent, underscore) can be included in
patterns using an escape character ‘\’ (backslash)

Prof P Sreenivasa Kumar 47

Department of CS&E, IITM
Join Operation
In SQL, usually joining of tuples from different relations is
implicitly specified in the ‘where’ clause

Get the names of professors working in CSE dept.

Select f.name
from professor as f, department as d
where f.deptNo = d.deptId and
d.name = ‘CSE’;

The above query specifies joining of professor and department

relations on condition f.deptNo = d.deptId and selection on
department relation using d.name = ‘CSE’

Prof P Sreenivasa Kumar 48

Department of CS&E, IITM
Explicit Specification of Joining in ‘From’ Clause
select f.name
from (professor as f join department as d on
f.deptNo = d.deptId)
where d.name = ‘CSE’;

Join types:
1.  inner join (default):
from (r1 inner join r2 on <predicate>)
use of just ‘join’ is equivalent to ‘inner join’
2.  left outer join:
from (r1 left outer join r2 on <predicate>)
3.  right outer join:
from (r1 right outer join r2 on <predicate>)
4.  full outer join:
from (r1 full outer join r2 on <predicate>)
Prof P Sreenivasa Kumar 49
Department of CS&E, IITM
Natural join
The adjective ‘natural’ can be used with any of the join types to
specify natural join.

FROM (r1 NATURAL <join type> r2 [USING <attr. list>])

•  natural join by default considers all common attributes

•  a subset of common attributes can be specified in an
optional USING <attr. list> phrase

REMARKS
•  Specifying join operation explicitly goes against the spirit of
declarative style of query specification
•  But the queries may be easier to understand
•  The feature is to be used judiciously

Prof P Sreenivasa Kumar 50

Department of CS&E, IITM
Views ( or Virtual Tables)
§  Views provide virtual relations which may contain data spread
across different tables. Used by applications.
•  simplified query formulations
•  data hiding
•  a view of frequently used data – efficient query answering
§  Once created, a view is always kept up-to-date by the RDBMS
§  View is not part of conceptual schema
•  created to give a user group, concerned with a certain aspect
of the information system, their view of the system
§  View implementation
•  Views need not be stored as permanent tables
•  They can be created on-the-fly whenever needed or
•  They can also be materialized and kept up-to-date
§  Tables involved in the view definition – base tables
Prof P Sreenivasa Kumar 51
Department of CS&E, IITM
Creating Views
CREATE VIEW v AS <query expr>
creates a view ‘v’, with structure and data defined by the
outcome of the query expression
Create a view which contains name, employee Id and
phone number of professors who joined CSE dept
in or after the year 2000. name of the view
create view profAft2K as
(Select f.name, empId, phone
from professor as f, department as d
where f.depNo = d.deptId and
d.name = ‘CSE’ and
f.startYear >= 2000);

If the details of a new CSE professor are entered into professor table,
the above view gets updated automatically
Prof P Sreenivasa Kumar 52
Department of CS&E, IITM
Queries on Views
Once created a view can be used in queries just like any other
table.

e.g. Obtain names of professors in CSE dept,

who joined after 2000 and whose name
starts with ‘Ram’

select name
from profAft2K
where name like ‘Ram%’;

The definition of the view is stored in DBMS, and executed to

create the temporary table (view), when encountered in query

Prof P Sreenivasa Kumar 53

Department of CS&E, IITM
 Operations on Views

§  Querying is allowed

§  Update operations are usually restricted

because – to update a view, we may require
to modify many base tables
– there may not be a unique way of updating the
base tables to reflect the update on view
– views may contain some aggregate values
– ambiguity where primary key of a base table is not
included in view definition.

Prof P Sreenivasa Kumar 54

Department of CS&E, IITM
 Restrictions on Updating Views
§  Updates on views defined on joining of more than one table
are not allowed
§  For example, updates on the following view are not allowed
§  Note that we are not keeping information about when a student has
completed the course in the view
create a view StudentGrades with rollNo, name, courseID
and grade

create view studentGrade(rollNo,name,courseId,grade) as

(select s.rollNo, s.name, e.courseId, e.grade
from student s, enrollment e
where s.rollNo = e.rollNo);
§  Suppose we want to update grade in the view from “U” to “D” for
one particular course for a student, there will be ambiguity in doing
the update on base tables.
Prof P Sreenivasa Kumar 55
Department of CS&E, IITM
 Restrictions on Updating Views
§  Updates on views defined with ‘group by’ clause and aggregate
functions is not permitted, as a tuple in view will not have a
corresponding tuple in base relation.

§  For example, updates on the following view are not allowed

Create a view deptAvgCredits which contains the

average credits of courses offered by a dept.

create view deptAvgCredits(deptNo,avgCredits)

as select deptNo, avg(credits)
from course
group by deptNo;

Prof P Sreenivasa Kumar 56

Department of CS&E, IITM
 Restrictions on Updating Views
§  Updates on views which do not include Primary Key of
base table, are also not permitted
§  For example, updates on the following view are not allowed

Create a view StudentPhone with Student name and

phone number.

create view StudentPhone (sname,sphone) as

(select name, phone
from student);

View StudentPhone does not include Primary key of the

base table.

Prof P Sreenivasa Kumar 57

Department of CS&E, IITM
Allowed Updates on Views
Updates to views are allowed only if

§  defined on single base table

§  not defined using ‘group by’ clause and aggregate functions

§  view includes Primary Key of base table

Prof P Sreenivasa Kumar 58

Department of CS&E, IITM
 Inserting data into a table
§  Specify a tuple(or tuples) to be inserted
INSERT INTO student VALUES
(‘CSO5D014’,‘Mohan’,‘PhD’,2005,‘M’,3,‘FCS008’),
(‘CSO5S031’,‘Madhav’,‘MS’,2005,‘M’,4,‘FCE009’);

§  Specify the result of query to be inserted

INSERT INTO r1 SELECT … FROM … WHERE …

§  Specify that a sub-tuple be inserted

INSERT INTO student(rollNo, name, sex)
VALUES (CS05M022, ‘Rajasri’, ‘F’),
(CS05B033, ‘Kalyan’, ‘M’);

§  the attributes that can be NULL or have

declared default values can be left-out to be
updated later
Prof P Sreenivasa Kumar 59
Department of CS&E, IITM
 Deleting rows from a table
§  Deletion of tuples is possible ; deleting only part of a tuple is
not possible
§  Deletion of tuples can be done only from one relation at a time
§  Deleting a tuple might trigger further deletions due to
referentially triggered actions specified as part of RIC’s
§  Generic form: delete from r where <predicate>;

Delete tuples from professor relation with start year

as 1982.
delete from professor
where startYear = 1982;

§  If ‘where’ clause is not specified, then all the tuples of that

relation are deleted ( Be careful !)
Prof P Sreenivasa Kumar 60
Department of CS&E, IITM
 A Remark on Deletion
§  The where predicate is evaluated for each of the tuples in the
relation to mark them as qualified for deletion before any
tuple is actually deleted from the relation
§  Note that the result may be different if tuples are deleted as and
when we find that they satisfy the where condition!
§  An example:
Delete all tuples of students that scored the least marks in the
CS branch:
DELETE
FROM gateMarks
WHERE branch = “CS” and
marks = ANY ( SELECT MIN(marks)
FROM gateMarks
WHERE branch = “CS”)
Prof P Sreenivasa Kumar 61
Department of CS&E, IITM
Updating tuples in a relation
update r
set <<attr = newValue> list>
where <predicates>;

Change phone number of all professors working in CSE

dept to “94445 22605”

update professors
set phone = ‘9444422605’
where deptNo = (select deptId
from department
where name = ‘CSE’);

If ‘where’ clause is not specified, values for the specified

attributes in all tuples is changed.

Prof P Sreenivasa Kumar 62

Department of CS&E, IITM
 Miscellaneous features in SQL (1/3)
§  Ordering of result tuples can be done using ‘order by’ clause
e.g., List the names of professors who joined
after 1980, in alphabetic order.
select name
from professor
where startYear > 1980
order by name;

§  Use of ‘null’ to test for a null value, if the attribute can take null
e.g., Obtain roll numbers of students who
don’t have phone numbers
select rollNo
from student
where phoneNumber is null;

Prof P Sreenivasa Kumar 63

Department of CS&E, IITM
 Miscellaneous features in SQL (2/3)
§  Use of ‘between and’ to test the range of a value
e.g., Obtain names of professors who have
joined between 1980 and 1990

select name
from professor
where startYear between 1980 and 1990;

§  Change the column name in result relation

e.g.,
select name as studentName, rollNo as studentNo
from student;

Prof P Sreenivasa Kumar 64

Department of CS&E, IITM
 Miscellaneous features in SQL (3/3)
§  Use of ‘distinct’ key word in ‘select’ clause to determine
duplicate tuples in result.
Obtain all distinct branches of study for students
select distinct d.name
from student as s, department as d
where s.deptNo = d.deptId;

§  Use of asterisk (*) to retrieve all the attribute values of

selected tuples.
Obtain details of professors along with their
department details.
select *
from professor as f, department as d
where f.deptNo = d.deptId;

Prof P Sreenivasa Kumar 65

Department of CS&E, IITM
Database System Architectures
Centralized Architecture – used long ago, before PCs were born
Complete DB functionality – storage, running application
programs, transaction processing etc – is on one system - Server
Access systems are just display devices - terminals

Client/Server Architecture – two tier systems

Client – powerful enough to do local processing
– runs graphical user interface and application programs
– sends Database queries/updates to Server
Server – provides rest of the DB System functionality

Three Tier System Architectures – also possible – details left out here

Prof P Sreenivasa Kumar 66

Department of CS&E, IITM
Application Development Process: 2-tier Systems
Host language (HL) – the high-level programming language in
which the application is developed (e.g., C, C++, Java etc.)

Managing Database Access – several approaches are available

•  Embedded SQL approach – SQL commands are embedded
in the HL programs
•  A static approach - SQL commands can’t be given at runtime
•  Dynamic SQL
•  Call Level Interface SQL/CLI – an API based approach
•  JDBC – Java DB connectivity – an API based approach
•  Use a Database programming language – Oracle’s PL/SQL

Embedded SQL, Dynamic SQL – we will study in some detail

Other approaches – to be studied by students
Prof P Sreenivasa Kumar 67
Department of CS&E, IITM
Impedance Mismatch
§  Impedance Mismatch:
Problems due to difference in HL data model vs DB data model
- Data types of HL vs those in DB
- HL languages do not support set-of-records as supported by SQL
§  Handling Data types
§  For each SQL attribute data type – corresponding HL data type
§  Specified as language binding
§  To be done for each host language
§  Handling SQL query results
§  Results are either sets or multi-sets of tuples
§  A data structure to hold results and an iterator are needed
§  Does not arise in case of dedicated DB programming languages
Prof P Sreenivasa Kumar 68
Department of CS&E, IITM
Embedded SQL Approach
Host language (HL) – the high-level programming language in
which the application is developed (e.g., C, C++, Java etc.)

Embedded SQL approach:

•  SQL statements are interspersed in HL program for application
development
•  Pre-compiler replaces these with suitable library calls
•  Library is supplied by the RDBMS vendor
•  SQL commands – identified by special reserved words –
EXEC SQL

Data transfer –
takes place through specially declared HL variables

Prof P Sreenivasa Kumar 69

Department of CS&E, IITM
Declaring Variables
Variables that need to be used in SQL statements are declared in a
special section. These are called shared variables.

EXEC SQL BEGIN DECLARE SECTION

char rollNo[9]; //HL is C language
char studName[20], degree[6];
int year; char sex;
int deptNo; char advisor[9];
EXEC SQL END DECLARE SECTION

Note that schema for student relation is

student(rollNo, name, degree, year, sex, deptNo, advisor)

Use in SQL statements: variable name is prefixed with a colon(:)

e.g., :ROLLNO in an SQL statement refers to rollNo variable
In HL program, shared variables can be used directly w/o colon.
Prof P Sreenivasa Kumar 70
Department of CS&E, IITM
Handling Error Conditions
The HL program needs to know if an SQL statement has
executed successfully or otherwise

Special variable called SQLSTATE is used for this purpose

It is a string of 6 characters.

§  SQLSTATE is set to appropriate value by the RDBMS

run-time system after executing each SQL statement
§  Non-zero values indicate errors in execution
•  Different values indicate different types of error situations
§  SQLSTATE variable must be declared in the HL program
§  HL program needs to check for error situations and handle them
appropriately.
Prof P Sreenivasa Kumar 71
Department of CS&E, IITM
Database Connections in Embedded SQL Approach
DB connection
-  Needs to be established before accessing the DB in the app pgm
-  Specify the particular server and authenticate the application
-  Several connections - to access 2 or more DB servers
-  Only one connection can be active at any time

SQL Commands
CONNECT TO <serverName> AS <connName>
AUTHORIZATION <uName, passWd>

To change to a different server

SET CONNECTION <connName>

DISCONNECT <connName>
Prof P Sreenivasa Kumar 72
Department of CS&E, IITM
Embedded SQL Statements – An example
Suppose we collect data through user interface into variables
rollNo, studName, degree, year, sex, deptNo, advisor

A row into the student table can be inserted as follows:

EXEC SQL INSERT INTO STUDENT

VALUES (:rollNo,:studName,:degree,
:year,:sex,:deptNo,:advisor);

Prof P Sreenivasa Kumar 73

Department of CS&E, IITM
 Query result handling and Cursors
§  HL languages do not support set-of-records as supported by SQL
§  A cursor is a mechanism which allows us to retrieve one row
at a time from the result of a query
§  We can declare a cursor on any SQL query
§  Once declared, we use open, fetch, move and close commands
to work with cursors
§  We usually need a cursor when embedded statement is
a SELECT query
§  INSERT, DELETE and UPDATE do not need a cursor.

Prof P Sreenivasa Kumar 74

Department of CS&E, IITM
Embedded SQL - Cursors (1/2)
We do not need a cursor if the query results in a single row.

e.g., EXEC SQL SELECT s.name, s.sex

INTO :name, :sex
FROM student s
WHERE s.rollNo = :rollNo;

§  Result row values name and phone are assigned to HL variables

:name and :phone, using INTO clause

§  Cursor is not required as the result always contains only

one row ( rollNo is a key for student relation)

Prof P Sreenivasa Kumar 75

Department of CS&E, IITM
Embedded SQL - Cursors (2/2)
§  If the result contains more than one row, cursor declaration
is needed

e.g., select s.name, s.degree

from student s
where s.sex = ‘F’;

§  Query results in a collection of rows

§  HL program has to deal with set of records.
§  The use of ‘INTO’ will not work here
§  We can solve this problem by using a cursor.

Prof P Sreenivasa Kumar 76

Department of CS&E, IITM
Declaring a cursor on a query
Cursor name
declare studInfo cursor for
select name, degree
from student
where sex = ‘F’;
§  Command OPEN studInfo; opens the cursor and makes it point
to first record
§  To read current row of values into HL variables:
FETCH studInfo INTO :name, :degree;
§  After executing FETCH statement cursor is pointed to next
row by default
§  Cursor movement can be optionally controlled by the
programmer
§  After reading all records we close the cursor using the
CLOSE studInfo command.
Prof P Sreenivasa Kumar 77
Department of CS&E, IITM
Dynamic SQL

§  Useful for applications to generate and run SQL statements,

based on user inputs
§  Queries may not be known in advance

e.g., char sqlString[ ] = {“select * from student”};

EXEC SQL PREPARE runQ FROM sqlString;
EXEC SQL EXECUTE runQ;

§  sqlString is a C variable that holds user submitted query

§  Typically built by previous statements in the HL program
using end-user inputs.
§  runQ is an SQL variable that holds the SQL statement.

Prof P Sreenivasa Kumar 78

Department of CS&E, IITM
Connecting to Database from HL – Other Approaches
ODBC (Open Database Connectivity), SQL/CLI and JDBC

§  accessing database and data is through an API

§  many DBMSs can be accessed
§  no restriction on number of active connections
§  appropriate drivers are required
§  steps in accessing data from a HL program
•  select the data source
•  load the appropriate driver dynamically
•  establish the connection
•  authenticate the client
•  work with database
•  close the connection.

Prof P Sreenivasa Kumar 79

Department of CS&E, IITM
A comparison of the Approaches
Embedded SQL Approach
+ queries are part of source code, syntax-check at compile time
+ application programs are easy to understand, readable
+ development is eaiser
- Any changes to queries : recompilation required
- Complex applications requiring runtime query creation are difficult
API based Approach
+ More flexibility in programming
+ Complex applications can be developed
- Application development is complex, error-prone

DB Language Approach
+ No impedance mismatch
- Programmers need to learn a new language; apps not portable
Prof P Sreenivasa Kumar 80
Department of CS&E, IITM