2018-04-22 19:13 GMT+02:00 Torsten Hämmerle <torsten.haemme...@web.de>: > dak wrote >> Well, yeah. (* (/ ... h_inf) h_inf) does not make a whole lot of sense >> either way. > > @Harm > > Function F0_1(w) basically is supposed to return 0 for w = 0 and > asymptotically approach a limit of 1 as w grows wider. > Because atan(x) -> π/2 (i.e. 90°) as x -> infinity, the result is multiplied > by 2/π to get 1 as a limit. > > h_inf is the maximum height the tie can reach. > > Dividing by h_inf within the function just influences the shape of the > curve, i.e. how exactly F0_1 it will change from 0 to 1. > Finally, the result (!) of F0_1 is multiplied by h_inf. > > In the current scheme code effectively (/ h_inf h_inf) gives a factor of 1 > and has no effect at all. > > The coding will work anyway, but the form of the curve will be different and > the maximum height will always be 1, no matter how big h_inf actually is. > > *Therefore, your solution* > (* (F0_1 (/ (* w r_0) h_inf)) h_inf))) > *exactly matches the C implementation:* > > Apart from that, either code will "work". As an example, I've plotted three > functions: > > <http://lilypond.1069038.n5.nabble.com/file/t3887/bezier-tie-f0_1.png> > red: the "wrong" calculation with h_inf/h_inf = 1 cancelling out, no matter > what value you pass as h_inf > green: h_inf = 2, so that a maximum height of 2 is being approached. > green: h_inf = 3, so that a maximum height of 3 is being approached. > > The "wrong" coding will always result in the red curve, as if h_inf was 1. > So if (incidentally) h_inf = 1, you won't see a difference. > > All the best, > Torsten
David, Torsten Thanks for your comments and hints. Patch is up: https://sourceforge.net/p/testlilyissues/issues/5311/ http://codereview.appspot.com/342030043 I think Jan intended to do the cc-definition exactly in scheme, though with a little less height-limit (h_inf) than Tie-default. It was a simple blunder with those brackets. While implementing the different tie-markup it slipped me as well. After this patch make-tie-stencil is better customizable and working as it should. Probably the tie-markups will show a very liitle difference, but I'm not sure whether patchy will catch them, at least my local tests didn't. Thanks, Harm _______________________________________________ lilypond-devel mailing list lilypond-devel@gnu.org https://lists.gnu.org/mailman/listinfo/lilypond-devel
