|
From: | Tim Daly |
Subject: | Re: [Axiom-developer] Examples on functions |
Date: | Fri, 30 Oct 2009 17:42:15 -0500 |
User-agent: | Thunderbird 2.0.0.21 (Windows/20090302) |
Bertfried Fauser wrote:
Dear Tim,On another topic, Axiom algebra is integrating the regression test files with the algebra sources. If you generate new code I beg you to write as many tests as you can, especially including a test for each function.I try to comply with this rule (ralf also asked for documentation) for the symmetric functions and hope they make their way into AXIOM(TM) too. However a fully documeted code is out of my working power, there is just far too much to say about symmetric functions. You saw possibly the mail of Martin on Clifford algebras. I have tried to test some Clifford issues in FriCAS (no AXIOM(TM) on my computer currently) but that should be identical. FriCAS is assuming that the quadratic form is diagonal, that is very bad. Perhaps I will find time to improve that. I tried to put up a sandbox on newsynthesis, but I possibyl forgot my user paseword there, since the system didn't let me create a new sandbox Ciao BF.
Given working tests we have some hope of figuring out what a function is supposed to do. Without a working test we can only guess whether the answer is right or not. You certainly have to construct these simple test cases in order to debug your code. Save the tests you do. As for documentation, it helps to include a small paragraph that contains enough so that we can find the rest in the literature. Even something as simple as a link to Eric's mathworld or wikipedia or a book reference would be a start. When I worked on the special function E1 at least there was a reference to Luke's book. Nobody can really expect you to write a tutorial or a book chapter but references help a lot. I'm trying to avoid just a straight "code dump" because then we have to guess. The main reason I started added example output to the ")display operation" command is that the new user would have at least one working example of every function. Sometimes it is very hard to construct the Axiom type even if you know what the input should look like, which I'm sure you have experienced. Tim
[Prev in Thread] | Current Thread | [Next in Thread] |