Using an SQL coverage measurement for testing database application

USING AN SQL COVERAGE MEASUREMENT FOR
TESTING DATABASE APPLICATION
PUBLISHED BY: PRESENTED BY:
MARÍA JOSÉ SUÁREZ-CABAL KHALID KHAN
JAVIER TUYA

INTRODUCTION
• Testing is very expensive process.
 Around 30% of resources are committed to this.
 Not adequate tool available for testing databases
• Software applications use database for accessing the content they offer.
• It uses SQL queries to access the database.
• The automation of software testing is only for programs written in imperative and structured
languages.
• It is not applied to SQL queries embedded in application code.
• The main focus of this research paper is to test the SQL queries and the database.

OBJECTIVES
The main objectives of the research are:
• To define a measurement of coverage for SQL SELECT queries in database loaded
with test data
• To present an algorithm that automates the calculation of coverage
• To guide the expert with respect to how the database tuples might be changed to
increase the obtained coverage value, if possible.

RELATED WORK
- RESEARCH
Microsoft Research (1998) :. The aim was
to evaluate DBMS with SQL statements
and compare the outputs.
1998
Hong Kong University: White box testing
is used on application which uses DMBS.
SQL queries were translated into
imperative languages.
1999
At first international workshop of
Automated Program Analysis, Test data is
generated, SQL queries are executed and
state of database is verified.
2000

RELATED WORK - COMMERCIAL
• In commercial area, there are small number of generators of database instances, e.g. TestByte,
TestBase, DataTect which generate random information depending on the type of fields.
 The problem is that user must know the database structure in order to use these tools.
• SQLUnit is another tool that facilitates testing of stored procedure and SQL statements.

SOFTWARE TESTING WITH DATABASES
There are several problems attached with testing of software application which uses DBMS
through SQL queries, which makes this testing difficult for the following reasons:
• Selection of initial instances of the test database is one of the most important steps
• Combination of tuples must be taken into account to cover all the SQL query situations
• Test data should be useful for the greatest possible number of queries
• As in imperative languages, SQL statements may be parameterized by variables and
constants, so these inputs must be considered.

EXAMPLE
sup_code sup_name
001 sup1
002 sup2
SUPPLIER ORDER
ord_num sup_code
1 001
2 001
3 002
ord_num sup_name
1 sup1
2 sup1
3 sup2
SELECT sup_name, ord_num
FROM Supplier s LEFT JOIN Order o
ON (s.sup_code = o.sup_code)

DETECTING FAULTS IN SELECT QUERIES
The problems detected are:
 There may exist a supplier with no orders, but this situation is not represented by the test
data.
 Impossible to know whether the query always return the information required.
Hence, we can affirm that an adequate SELECT coverage is not obtained with these test data.

DETECTING FAULTS IN SELECT QUERIES
sup_code sup_name
001 sup1
002 sup2
003 sup3
SUPPLIER ORDER
ord_num sup_code
1 001
2 001
3 002
ord_num sup_name
1 sup1
2 sup1
3 sup2
sup3
SELECT sup_name, ord_num
FROM Supplier s LEFT JOIN Order o
ON (s.sup_code = o.sup_code)

COVERAGE OF SELECT QUERIES
The solution for automated search of SQL query situations covered with the data stored in the
database is to evaluate the conditions of SELECT queries:
 When they are in FROM clause
 When they include JOIN clause
 When they are in WHERE clause
Also, the null values of field should be verified at the same time as the condition is evaluated.

COVERAGE OF SELECT QUERIES - EVALUATION OF CONDITIONS
Conditions are evaluated between sets of values, information in each field corresponds to a
column from a table to several rows in the database. During an evaluation of a condition,
 Each value in the first field must be compared with each one in the second field
 Each value in the second field must be compared with each one in the first field

COVERAGE OF SELECT QUERIES - COVERAGE TREE
A tree structure, called coverage tree, is created prior to coverage evaluation, in which each level
of represents a condition of the query:
 Beginning with the condition of JOIN clause (if it exists)
 Those with the WHERE clause
in the same order, in which they are found in the query.
Each node of the tree will store:
 Whether the condition is true for values of the fields,
 Whether the condition is false for values of the fields,
 Whether there are null values in condition field,

EVALUATION OF CONDITIONS (CONT’D)
The evaluation of a condition:
• The grammar:
• A condition will be true (T) if it is verified for a pair of values from the field to compare.
• A condition will be false from left to right (Fl), if none of the values from second field verifies
the condition with a value from the first field.

EVALUATION OF CONDITIONS (CONT’D)
• A condition will be false from right to left (Fr), if none of the
values from first field verifies the condition with a value from
the second field.
• A condition will have null values (Nl) when a value from the first field is null.
• A condition will have null values (Nr) when a value from the second field is null.
• A condition will have null values (Nb) when a value from the first and second fields are null.

COVERAGE TREE
Nl Nb Nr
N N N
s.sup_code = o.sup_code
Fl T Fr
N Y N

EVALUATION OF COVERAGE TREE
The complete evaluation of the query is carried out by crossing over the tuples of the tables
participating in the condition at each level of the coverage tree.
 The evaluation finishes when the coverage is 100% or there are no more values for
comparing.
After evaluating the coverage tree, two different coverage measures are established:
 Theoretical Coverage: It takes into account every possible situation at every node.
 Schema Coverage: It takes into account the database schema constraint by excluding impossible
situation due to these constraints.

THEORETICAL COVERAGE
Problem: It is not usually possible to reach 100% theoretical coverage, because of forbidden
null values or referential integrity constraint.
Solution: Schema Coverage

SCHEMA COVERAGE
 It takes into account the database schema constraint by excluding impossible situation due to
these constraints
 Due to referential integrity, an order always had a supplier, Fr, Nb and Nr will never be
verified (X)
Nl Nb Nr
N X X
s.sup_code = o.sup_code
Fl T Fr
N Y X

THE TOOL
The schema of the tool developed for measuring the coverage of SELECT query.

CONCLUSION AND FUTURE WORK
 The most important aim considered: two different coverage measures for the coverage of SQL queries
have been established, i.e. Theoretical and Schema Coverage.
 The number of tuples in the database may be simplified by tracing the coverage tree. This is useful for
creating smaller databases and for detecting information that would have to be inserted to make them
complete.
For future work,
 SQL Queries, which were analyzed, were simple and isolated. There is a plan to increase the complexities
of queries by including the parameters in it.
 It will be necessary to integrate SQL coverage criteria with other criteria for imperative languages for
stored procedures to be tested.

Using an SQL coverage measurement for testing database application

More Related Content

What's hot (20)

Similar to Using an SQL coverage measurement for testing database application (20)

Recently uploaded (20)

Using an SQL coverage measurement for testing database application