Hi Tim, On Mon, Aug 30, 2010 at 1:47 PM, Tim Daly <d...@axiom-developer.org> wrote: > Tim (somebody, I can't find his name) was going to try the > CATS integration test suite in Sage. It should be rather easy > as Sage uses Maxima and the syntax is very close. I don't > know what became of that effort.
I don't recall reading about any such effort. > If you do a test suite on graphs I'd be happy to try to > make it work in Axiom. This would likely involve writing new > algebra code so it would be helpful if you included citations > of algorithms (if you know them) as you develop it. It looks > like Gross and Yellin includes high level algorithms so that > would be exceptionally helpful. I strongly favor basing test > suites on books containing algorithms. As a start, one could gather known databases of graphs such as the following page of Chris Godsil linking to various graph databases: http://quoll.uwaterloo.ca/mine/theirs.html or the book (Read & Wilson 2005) "An Atlas of Graphs". For each graph and its associated properties as contained in the database, we list that graph in some type of representation, e.g. graph6 or sparse6, and the associated properties as extracted from the database. We then read in that graph format and invoke graph theoretic functions to confirm that the output of the system in question agrees with the stated properties. That is, we're testing to ensure that properties as contained in the database match the output of the system. One could also go beyond verifying properties contained in a database and consider verifying mathematical properties of a graph itself. Numerous known properties of graphs exist that could make valuable test cases. The above outline could be a start. This has been on my mind for months now. I just haven't the luxury of time to devote to this task. > For the CATS test suite each file contains the problem, the > expected answer, and Axiom's answer. The format of the files > makes it possible to run them automatically which I do with > every system build. Does Sage have such a file format? I don't think so. One way to write test cases at the moment is to write them in a source file of the Sage library or contribute test cases to a relevant upstream project whose source code makes up a standard Sage package. There's also the option to write a list of test cases, embed them in the docstring of a Python file that only has those test cases (and no Python code whatsoever), and integrate that file of test cases into the Sage library. When generating the Sage standard documentation you have the choice of which files in the Sage library to extract documentation from and pretty display the doc in the reference manual. A few months ago I experimented with writing test cases in LaTeX as part of a detailed discussion of underlying relevant theory. So for example you want to write some test cases for graph theory. You start by writing a short document explaining any relevant theory and runtime complexity of algorithms you want to cover, with references to any literature where appropriate. Next write down properties you're interested in together with any expected results, followed by a transcript of a Sage session that verify the stated properties. All of these are written up as a LaTeX document just like your normal LaTeX document. To run the test cases through Sage, I used a little tool [1] to extract the test cases and run them through the Sage doctesting framework. As an example of what I mean, refer to this document [2] which is a work in progress. The idea described above is nothing new. One of my original intentions was to write test cases with accompanying explanatory theory so any future maintainer would have some documentation to help them update the test suite. Of course this means that it would require more time and effort than simply writing test cases. What you end up resembles more of an expository document than a test suite. [1] http://bitbucket.org/mvngu/sageutil/src/tip/textract.py [2] https://code.google.com/p/factor-book/ -- Regards Minh Van Nguyen -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
