This is an extension to Kent Beck's SUnit framework for Smalltalk unit testing,
described at http://www.armaties.com/testfram.htm.

Beck's framework requires a test class to encapsulate each set of unit
tests.  For example, to test a hypothetical class Date, you'd subclass
TestCase, and call it, say, DateTestCase.  You'd instantiate an instance of
this class for each test case.  For more details, see Kent Beck's
documentation.

This seemed bulky to me, so I wrote a subclass of TestCase called
DelegatingTestCase.  It's constructed with a selector (a symbol), and a
target object that the selector is delegated to.  Then you can write your
test cases on the class that's being tested, and construct a test case that
calls the test method via: DelegatingTestCase target: aClass selector:
aSymbol So, for example, to create a test case for Date, create a class
method called #testDateAddition, and evaluate 'DelegatingTestCase target:
Date selector: #testDateAddition' to create a test case that, when run,
will call that method.  The test case can then be sent the #run message, or
added to a TestSuite -- again, see Kent Beck's documentation for more
details on how to use TestCases.

'TestSuite forClass: aClass' creates a TestSuite that has one test case for
each class method in the 'test cases' protocol on aClass.  So, in the above
example, if #testDateAddition were in the 'test cases' protocol of the Date
class, then evaluating '(TestSuite forClass: Date) run' would run Date
class>>testDateAddition -- as well as whatever other class methods were in
the "test cases" protocol of Date -- and return a string summarizing the
results.

This code is released to the public domain December 12, 1999, by Oliver
Steele.