Computing test methods in unittest.TestCa se

[Jan Decaluwe]

I'm working on a unit test for a finite state machine (FSM). The FSM
behavior is specified in a dictionary called transitionTable . It has a
key per state with a tuple of possible transitions as the corresponding
value. A transition is defined as a number of input values, a next
state, and a documentation string.

I want to test each possible transition in a separate test method. For
a particular transition, a test method could look as follows:

class FSMTest(unittes t.TestCase):
....
def test_START_2(se lf):
"""Check state START - Switched channels before data started"""
state, transition = START, transitionTable[START][2]
sim = Simulation(self .bench(state, transition))
sim.run()

As the number of transitions can be large, I want to
"compute" the test methods, based on the transitionTable info.
Currently, I have the following:

class FSMTest(unittes t.TestCase):
....
for st, trs in transitionTable .items():
for i, tr in enumerate(trs):
def _tmpfunc(self, st=st, tr=tr):
sim = Simulation(self .bench(st, tr))
sim.run()
_tmpfunc.func_d oc = "Check state %s - %s" % (st, getDoc(tr))
exec "test_%s_%s = _tmpfunc" % (st, i)

This works, but uses some "ugly" tricks:

* default arguments to pass context info. As the code is executed in
class context, not function context, I cannot use free variables.
* The use of 'exec'. unittest looks for methods with name prefix
'test_' in the class namespace, and I didn't find another way
to achieve that.

Anyone with better ideas?

--
Jan Decaluwe - Resources bvba - http://jandecaluwe.com
Losbergenlaan 16, B-3010 Leuven, Belgium
Python is fun, and now you can design hardware with it:
http://jandecaluwe.com/Tools/MyHDL/Overview.html

[James Kew]

"Jan Decaluwe" <ja*@jandecaluw e.com> wrote in message
news:40******** ******@jandecal uwe.com...

class FSMTest(unittes t.TestCase):

....
for st, trs in transitionTable .items():
for i, tr in enumerate(trs):
def _tmpfunc(self, st=st, tr=tr):
sim = Simulation(self .bench(st, tr))
sim.run()
_tmpfunc.func_d oc = "Check state %s - %s" % (st, getDoc(tr))
exec "test_%s_%s = _tmpfunc" % (st, i)

This works, but uses some "ugly" tricks:

* default arguments to pass context info. As the code is executed in
class context, not function context, I cannot use free variables.
* The use of 'exec'. unittest looks for methods with name prefix
'test_' in the class namespace, and I didn't find another way
to achieve that.

Anyone with better ideas?

Define the test methods after defining the class, and inject them into the
class with setattr:

class testClass:
pass

def makeTest(param) :
def test(self):
print "test: param=%s" % param
return test

setattr(testCla ss, "test1", makeTest(1))
setattr(testCla ss, "testHello" , makeTest("Hello "))

t = testClass()
t.test1() test: param=1 t.testHello()

test: param=Hello

James

[Peter Otten]

Jan Decaluwe wrote:

I'm working on a unit test for a finite state machine (FSM). The FSM
behavior is specified in a dictionary called transitionTable . It has a
key per state with a tuple of possible transitions as the corresponding
value. A transition is defined as a number of input values, a next
state, and a documentation string.

I want to test each possible transition in a separate test method. For
[...]
As the number of transitions can be large, I want to
"compute" the test methods, based on the transitionTable info.
[...]
Anyone with better ideas?

Use a test suite instead and turn each FSMTest.testXXX () method into a
TestCase instance:

class TransitionTest( unittest.TestCa se):
def __init__(self, state, transition):
unittest.TestCa se.__init__(sel f)
self.state = state
self.transition = transition

def bench(self, state, transition):
pass

def runTest(self):
Simulation(self .bench(self.sta te, self.transition )).run()

def __str__(self):
return "Check state %s - %s" % (self.state, getDoc(self.tra nsition))

def makeSuite():
suite = unittest.TestSu ite()
for st, trs in transitionTable .iteritems():
for tr in trs:
suite.addTest(T ransitionTest(s t, tr))
return suite

if __name__ == "__main__":
unittest.main(d efaultTest="mak eSuite")

Peter

[Jan Decaluwe]

James Kew wrote:

"Jan Decaluwe" <ja*@jandecaluw e.com> wrote in message
news:40******** ******@jandecal uwe.com...

class FSMTest(unittes t.TestCase):

....
for st, trs in transitionTable .items():
for i, tr in enumerate(trs):
def _tmpfunc(self, st=st, tr=tr):
sim = Simulation(self .bench(st, tr))
sim.run()
_tmpfunc.func_d oc = "Check state %s - %s" % (st, getDoc(tr))
exec "test_%s_%s = _tmpfunc" % (st, i)

This works, but uses some "ugly" tricks:

* default arguments to pass context info. As the code is executed in
class context, not function context, I cannot use free variables.
* The use of 'exec'. unittest looks for methods with name prefix
'test_' in the class namespace, and I didn't find another way
to achieve that.

Anyone with better ideas?

Define the test methods after defining the class, and inject them into the
class with setattr:

class testClass:
pass

def makeTest(param) :
def test(self):
print "test: param=%s" % param
return test

setattr(testCla ss, "test1", makeTest(1))
setattr(testCla ss, "testHello" , makeTest("Hello "))

Thanks - this is a nice way to get rid of the exec. However, I found
out that I still need the default arguments in my case. Free variables
do behave differently as shown in the following example:

def funcs():
flist = []
glist = []
for i in range(3):
def f():
print "f: " + str(i)
def g(i=i):
print "g: " + str(i)
flist.append(f)
glist.append(g)
for f, g in zip(flist, glist):
f(); g()

funcs()

f: 2
g: 0
f: 2
g: 1
f: 2
g: 2

When generating such functions (with free variables) in loops,
it's apparently very easy to get it wrong ...

Regards, Jan

--
Jan Decaluwe - Resources bvba - http://jandecaluwe.com
Losbergenlaan 16, B-3010 Leuven, Belgium
Python is fun, and now you can design hardware with it:
http://jandecaluwe.com/Tools/MyHDL/Overview.html