On Sun, Sep 6, 2009 at 2:08 PM, Moritz Lenz<[email protected]> wrote:
> Currently all the tests which are not fudged in trig.t pass here Same here. >> 3) In general, I'd love it if a few people could look over the new >> tests for sine (cosine duplicates them). I'm not 100% comfortable with >> how repetitive they are. > > I know that feeling, but I stopped bothering about it some time ago. > Testing the basics just requires lots of repetition. Especially > numerical stuff. When I was working on S03-operators/autoincrement-range.t, I wound up writing some data-driven Perl 5 that would output the many static tests I wanted. I wish I'd kept that. Once I'd automated what I could, I took its output and modified it for oddities. >> (BTW, these >> changes I've made add over 1000 tests in trig.t, and they are only the >> tip of the iceberg.) I'd suggest breaking that into multiple files. To me, the most obvious divisions are between functions (sin, cos, etc.), but someone else might think it makes sense to break along units (deg, rad, etc.). If you're really aggressive, break along both (sin-deg, sin-rad, cos-dec, cos-rad, etc.) On the other hand, there's nothing really wrong with 10K tests in one file. I only bring it up because when I have my nicer machine run five files at once, trig.t is always the one it ends up waiting on to finish after all the others are done. Kyle.
