From: Egon W. <ego...@gm...> - 2007-10-25 14:50:31
|
On 10/25/07, Rajarshi Guha <rg...@in...> wrote: > On Oct 25, 2007, at 9:59 AM, Egon Willighagen wrote: > > On 10/25/07, raj...@us... > > <raj...@us...> wrote: > >> Revision: 9211 > >> http://cdk.svn.sourceforge.net/cdk/?rev=9211&view=rev > >> ----------- > >> Fixed some PMD (idiotic) warnings > > > > I think the choices we made are quite reasonable. If you would like to > > discuss them, that's fair :) > > > > The PMD default are already tuned to our needs. Please see the > > cdk/pmd/ subdirectory for CDK's customizations. > > As I noted in the bug report update, in a small function, where it is > very clear what the variable 'n' is doing, I don't see the need to > make that a longer, more 'intelligable' name. Yeah, it would be nice if that test would include the method length... if <10 lines, then single/two char var names would be fine... you could suggest that with PMD. > Also I woke up with a headache today - which might explain terse > comments :( Yeah, same here... > Also, one thing that has seemed quite kludgy to me is the way we are > identifying missing unit tests - IIRC, you have to construct the unit > test name from the method name and its argument types. Yes, that's the current coding convention. > Wouldn't it be better to start using annotations on the main source > code to indicate whether something has been tested (maybe associate > the annotation with the test class). That way there is no restriction > on the test class name, a given class can have multiple tests etc Yes, that would be nice, now that we can use annotations... I do not have a clue how to do this. Suggestions? We should likely have a close look at: http://www.junitext.org/ Another thing we could do then, is indicate which tests depend on each other... that is, if test X fails, then Y will surely fail too... (think about tests using another algorithm...) Egon -- ---- http://chem-bla-ics.blogspot.com/ |