From: Ian Jackson <ijackson@chiark.greenend.org.uk>
Date: Sat, 7 Apr 2018 22:54:55 +0000 (+0100)
Subject: curveopt: wip, observe a bug
X-Git-Url: http://www.chiark.greenend.org.uk/ucgi/~ian/git?a=commitdiff_plain;h=ece586e7d6e8f68236020935fdb7439c54dfa382;p=moebius3.git

curveopt: wip, observe a bug

This algorithm encourages a segment with a curve either side of it to
be shorter than its neighbours.  In this test case there is another
bug: the cp[1] (F) and cp[2] (G) vectors are wrong, exaggerating the
effect of the bug.

To fix the wrong encouragement bug, I think we will have to split the
penalty for curvature from the penalty for density variation.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>
---

diff --git a/curveopt.py b/curveopt.py
index 5ebe166..12d9625 100644
--- a/curveopt.py
+++ b/curveopt.py
@@ -29,8 +29,8 @@ class OptimisedCurve():
     for end in (False,True):
       ei = nt if end else 0
       fi = nt-1 if end else 1
-      cp0i = 0 if end else 3
-      cp1i = 1 if end else 2
+      cp0i = 3 if end else 0
+      cp1i = 2 if end else 1
       e = np.array(cp[cp0i])
       ef_dirn = unit_v(cp[cp1i] - cp[cp0i])
       ef_len = np.linalg.norm(np.array(fc_input[fi]) - np.array(fc_input[ei]))