[Axiom-developer] Some thoughts on CATS

[root]

Ayal,

My attack on the problem involves literate programming, which should
come as no surprise. Each test case involves a tex explanation of the
mathematics of the problem and the expected result. This is followed
by code chunks for each system. Developers of the different systems
will have to write the specific test code to implement the test.  The
developer has the option to skip or include any test and have any code
executed for their specific system.

There is no common method shared by any two systems which verifies
results.  Axiom, for instance, won't accept lisp expressions as an
input form as there is no general conversion mechanism (though there
probably should be) from s-expressions to internal data structures and
back. Defining VERIFY-EQUAL functions for the Axiom domains would be
a large development effort.

Yacas can, as I see from your examples, not only runs the extracted
code but can wrap the test in a verification function. Other systems
don't have such a verification function and the expected results will
have to be verified in a different manner. In Axiom we simply diff the
resulting output files and check for changes.

Tests for yacas can be extracted by writing:

    notangle -Ryacas CATS.pamphlet

Documentation for the tests is available by writing:

    noweave CATS.pamphlet

I'm going to work up a first draft example this week so people have
something specific to complain about. I hope to post it by next week.

Tim