Software Testing with Python
Magnus Lyckå Thinkware AB www.thinkware.se
EuroPython Conference 2004 Chalmers, Göteborg, Sweden
In the next 30 minutes you should...
● Learn about different aspects of software testing
such as unit tests and acceptance tests.
● See how the standard modules for unit tests are
used, learn about some alternatives and have a grasp of their respective pros and cons.
● Know about some options for acceptance tests. ● Know about some standard Python modules and
third party tools that are helpful in software testing and related activities.
What is software testing?
● “Testing means verifying that your code is
running correctly by exercising the code under known conditions and checking that the results are as expected.” – Alex Martelli, Python in a Nutshell
● If we are serious with our requirements on a
software system, we should make sure that we
verify them, and testing is the most common kind of verification for software.
Software testing in a context
Software Quality Assurance
Software Verification
Automating tests
● Writing automated tests is usually more work
than testing things manually once, but they make it much easier to...
– Work in a repeatable and predicable way – Run tests more often
– Run tests unattended – Find regression bugs
– Test more operating systems
Example of requirements and
tests on different levels
System Specification Acceptance Test
Detailed Design API Specification
Unit Test
Requirements and tests on different
levels in Extreme Programming (XP)
System specified through customer tests
Design by writing API validated through
programmer tests
Consequences of the XP approach
● Requirements are written in a verifiable way.
● No discrepancies between requirements and tests. ● We get unambiguous and repeatable ways of
verifying requirements.
● Continuous picture of project progress
But...
● Customers must be able to understand the tests. ● We need test automation frameworks.
Software Testing with Python
Unit tests in Python
● Unit tests, or programmer tests, are tests the
programmer writes to make sure that the code does what the programmer intended it to do.
● Unit tests can also help document how a piece of
code (e.g. a class) is supposed to be used.
● Python has two standard modules for unit testing,
unittest and doctest.
● There are several third party modules that aim to
The
unittest
module
● The Python unittest module is modeled after
the unit test modules developed within the XP community by Kent Beck and Erich Gamma.
● It's intended for a test-first approach, where tests
are written before the actual code that it tests.
● Always writing tests first is a bit like always
eating your meat before you eat your dessert...
● When your tests pass, you're done. (XPers
Concepts in
unittest
● A test fixture consists of the actions needed to
setup for a test, and clean up afterwards.
● A test case is the smallest unit of testing. It might
for instance call a function and check the results.
● A test suite is a collection of test cases and/or test
suites that should be executed together.
● A test runner runs test suites (or individual test
cases), collects the results and presents it as text or graphically to the user.
The
TestCase
class
● Test cases sharing a common fixture will be
implemented as methods (whose names start with
test) in a sub class of unittest.TestCase.
● The test runner will create one instance object per
test case, and run SetUp, followed by the test case method and finally tearDown.
● Checks are made with the methods fail, failIf,
assert_/failUnless, assertEqual/failUnlessEqual, assertNotEqual/failIfEqual and
The test runner
● In many cases, running tests is as simple as
executing unitest.main() in the file containing the test cases.
● Testing progress will be reported, and the test
runner makes a distinction between FAILURE and ERROR.
– FAILURE means that the check didn't produce the
expected result.
– ERROR means that something else in the test went
Unittest template
import unittest import MyModule
class MyModuleTest(unittest.TestCase):
def testOneCase(self):
pass
if __name__ == '__main__':
unittest.main()
Unittest examples
● divtest.py
– Trivial example to demonstrate features of unittest.
● container_ut.py
– Real world example from SystemSpecifyer (see
The
doctest
module
● The Python doctest module was written by
Tim Peters to check that coding examples in doc-strings were correct, by testing them.
● It has also found use as a more general unit
testing module, and from Python 2.3, this has been made more convenient.
● Using doctest helps you combine tests and
documentation in a more readable way than
Doctest example
def divide(a,b):"""
Return a divided with b.
Divide will return the result of an integer division >>> divide(6,3)
2
Division by zero is handled as expected... >>> divide(1,0)
Traceback (most recent call last): ...
ZeroDivisionError: integer division or modulo by zero """
result = a / b return result
if __name__ == '__main__': import doctest, sys
Unittest vs doctest
● Unittest mantra: Test a little, code a little
– The logical choice if you use extreme programming
or test-driven development of some other kind.
– Maybe better for complex tests?
● Doctest mantra: Code a little, test a little
– The natural companion if you prefer to experiment
interactively with your classes and functions in the Python interactive environment.
– Probably better for adding tests to already written
Unittest vs doctest on the web
● Charming Python: Testing frameworks in Python
– http://www-106.ibm.com/developerworks/linux/library/l-cptest.html
● Literate Testing: Automated Testing with doctest
– http://www.python.org/pycon/dc2004/papers/4/
● DocTest at WikiWiki
Other contenders...
● Sancho – a unit testing framework from MEMS
Exchange. Adds coverage analysis etc.
– http://www.mems-exchange.org/software/sancho/
● Peckcheck by Darius Bacon.
– http://www.accesscom.com/~darius/software/clickcheck.html
● The “second standard library” std.
Unittest complements
● Pester – the Python version of Jester.
– It finds code that is not covered by tests, makes some
change to your code, runs your tests, and if the tests pass it displays a message saying what it changed.
– http://jester.sourceforge.net/
● Test coverage support
– E.g. http://www.garethrees.org/2001/12/04/python-coverage/
● Mock objects: Here, 9:30
Software Testing with Python
Python Integration Test
● No static linking. We must run code to test
interfaces! (On the other hand, compilers and linkers fail to see lots of integration problems...)
● Since building/linking isn't an issue in a pure
Python project, we can use either unit test tools or acceptance test tools for integration tests.
● Python is good at gluing things together, and thus
helpful in all sorts of integration work. See e.g.
Software Testing with Python
Test Frameworks using Python
● QMTest – Uses Python for test expressions.
– http://www.codesourcery.com/qmtest/
● PyFIT – Python clone of Ward Cunningham's
Framework For Integrated Testing (FIT).
– http://www.xprogramming.com/software.htm
● Software Testing Automation Framework (STAF)
Big framework from IBM with Python API.
– http://staf.sourceforge.net/
Software Testing with Python
The usual suspects...
● Python is excellent for analysis and manipulation
of data. A great tool for test related work.
● The re library is useful but use special tools
when available, such as for parsing XML files.
● For dealing with files you might use the modules
gzip, zipfile, codecs, filecmp, struct, sgmllib, xml.* etc
● Python interfaces well with internet services,
difflib
● More or less like the unix diff utility, but as a
Python module. (New in version 2.1.)
● Useful when we want a more detailed response
than FAILED from a test – particularly if we are trying to spot small changes in big amounts of data.
● Great for spotting differences in configurations,
Difflib
example
>>> import difflib; d=difflib.Differ()>>> diff =d.compare(['Hello World', "This is the same.",
"Time flies like an arrow. Isn't that great?"], ['Hello World!', "This is the same.",
"Fruit flies like a banana. Isn't that great?"]) >>> print "\n".join(diff)
- Hello World + Hello World! ? +
This is the same.
- Time flies like an arrow. Isn't that great? ? ^ ^^ - ^^^^
+ Fruit flies like a banana. Isn't that great? ? ^^^ ^^ +++ ^^
AT&T Graphviz
● Tools to generate graphs from C-like text files.
– Dot – for directed graphs.
– Neato – for undirected graphs
– Fairly clever algorithms for decent looking layout – Generates graphs in many file formats
– Suitable for automatic generation of graphs – Some control over placement is possible
● Great for dependency analysis!
Remember this?
System Specification Acceptance Test
Detailed Design API Specification
Unit Test
AT&T Graphviz example 1
digraph G {
label = "V-model"; sysspec
[label="System Specification"]; systest [label="Acceptance Test"]; apispec [label="API Specification"]; apitest [label="Integration Test"]; unitspec [label="Detailed Design"]; unittest [label="Unit Test"];
sysspec -> apispec -> unitspec; unittest -> apitest -> systest; sysspec -> systest;
apispec -> apitest; unitspec -> unittest; }
AT&T Graphviz example 2
digraph G {
label = "V-model";
node [shape=box, style=filled, color="#CCCCFF"];
sysspec [label=
"System Specification"];
apispec [label="API Specification"]; unitspec [label="Detailed Design"]; node [shape=box, style=filled,
color="#FFFF99"];
systest [label="Acceptance Test"]; apitest [label="Integration Test"]; unittest [label="Unit Test"];
{rank = same; sysspec; systest}; {rank = same; apispec; apitest}; {rank = same; unitspec; unittest}; ...
Useful Books
● Python in a Nutshell, by Alex Martelli
● Text Processing in Python, by David Mertz
● Software Test Automation, by Fewster & Graham ● Just Enough Software Test Automation, by
Daniel J. Mosley & Bruce A. Posey
● Testing Extreme Programming, by Lisa Crispin
and Tip House
● Test-Driven Development By Example, by Kent