Remove .gitignore do nothing is ignored.
[synfig.git] / synfig-studio / trunk / src / synfigapp / blineconvert.cpp
index d242e9f..42cb57a 100644 (file)
@@ -2,10 +2,11 @@
 /*!    \file blineconvert.cpp
 **     \brief Template File
 **
-**     $Id: blineconvert.cpp,v 1.1.1.1 2005/01/07 03:34:37 darco Exp $
+**     $Id$
 **
 **     \legal
 **     Copyright (c) 2002-2005 Robert B. Quattlebaum Jr., Adrian Bentley
+**     Copyright (c) 2007 Chris Moore
 **
 **     This package is free software; you can redistribute it and/or
 **     modify it under the terms of the GNU General Public License as
@@ -39,7 +40,7 @@
 #include <synfig/general.h>
 #include <cassert>
 
-
+#include "general.h"
 
 #endif
 
@@ -66,127 +67,114 @@ using namespace synfig;
 template < class T >
 inline void FivePointdt(T &df, const T &f1, const T &f2, const T &f3, const T &f4, const T &f5, int bias)
 {
-       if(bias == 0)
-       {
-               //middle
+       if (bias == 0)                          // middle
                df = (f1 - f2*8 + f4*8 - f5)*(1/12.0f);
-       }else if(bias < 0)
-       {
-               //left
+       else if (bias < 0)                      // left
                df = (-f1*25 + f2*48 - f3*36 + f4*16 - f5*3)*(1/12.0f);
-       }else
-       {
-               //right
+       else                                            // right
                df = (f1*3 - f2*16 + f3*36 - f4*48 + f5*25)*(1/12.0f);
-       }
 }
 
 template < class T >
 inline void ThreePointdt(T &df, const T &f1, const T &f2, const T &f3, int bias)
 {
-       if(bias == 0)
-       {
-               //middle
+       if (bias == 0)                          // middle
                df = (-f1 + f3)*(1/2.0f);
-       }else if(bias < 0)
-       {
-               //left
+       else if (bias < 0)                      // left
                df = (-f1*3 + f2*4 - f3)*(1/2.0f);
-       }else
-       {
-               //right
+       else                                            // right
                df = (f1 - f2*4 + f3*3)*(1/2.0f);
-       }
-}
-
-template < class T >
-inline void ThreePointddt(T &df, const T &f1, const T &f2, const T &f3, int bias)
-{
-       //a 3 point approximation pretends to have constant acceleration, so only one algorithm needed for left, middle, or right
-       df = (f1 -f2*2 + f3)*(1/2.0f);
 }
 
-// WARNING -- totaly broken
-template < class T >
-inline void FivePointddt(T &df, const T &f1, const T &f2, const T &f3, int bias)
-{
-       if(bias == 0)
-       {
-               assert(0); // !?
-               //middle
-               //df = (- f1 + f2*16 - f3*30 +  f4*16 - f5)*(1/12.0f);
-       }/*else if(bias < 0)
-       {
-               //left
-               df = (f1*7 - f2*26*4 + f3*19*6 - f4*14*4 + f5*11)*(1/12.0f);
-       }else
-       {
-               //right
-               df = (f1*3 - f2*16 + f3*36 - f4*48 + f5*25)*(1/12.0f);
-       }*/
-       //side ones don't work, use 3 point
-}
-
-//implement an arbitrary derivative
-//dumb algorithm
-template < class T >
-void DerivativeApprox(T &df, const T f[], const Real t[], int npoints, int indexval)
-{
-       /*
-       Lj(x) = PI_i!=j (x - xi) / PI_i!=j (xj - xi)
-       
-       so Lj'(x) = SUM_k PI_i!=j|k (x - xi) / PI_i!=j (xj - xi)
-       */
-       
-       unsigned int i,j,k,i0,i1;
-       
-       Real Lpj,mult,div,tj;
-       Real tval = t[indexval];
-                               
-       //sum k 
-       for(j=0;j<npoints;++j)
-       {
-               Lpj = 0;
-               div = 1;
-               tj = t[j];
-               
-               for(k=0;k<npoints;++k)
-               {
-                       if(k != j) //because there is no summand for k == j, since that term is missing from the original equation
-                       {
-                               //summation for k
-                               for(i=0;i<npoints;++i)
-                               {
-                                       if(i != k)
-                                       {
-                                               mult *= tval - t[i];
-                                       }
-                               }
-                               
-                               Lpj += mult; //add into the summation
-                               
-                               //since the ks follow the exact patern we need for the divisor (use that too)
-                               div *= tj - t[k];
-                       }
-               }
-               
-               //get the actual coefficient
-               Lpj /= div;
-               
-               //add it in to the equation
-               df += f[j]*Lpj;
-       }
-}
+// template < class T >
+// inline void ThreePointddt(T &df, const T &f1, const T &f2, const T &f3, int bias)
+// {
+//     // a 3 point approximation pretends to have constant acceleration,
+//     // so only one algorithm needed for left, middle, or right
+//     df = (f1 -f2*2 + f3)*(1/2.0f);
+// }
+//
+// // WARNING -- totally broken
+// template < class T >
+// inline void FivePointddt(T &df, const T &f1, const T &f2, const T &f3, int bias)
+// {
+//     if(bias == 0)
+//     {
+//             assert(0); // !?
+//             //middle
+//             //df = (- f1 + f2*16 - f3*30 +  f4*16 - f5)*(1/12.0f);
+//     }/*else if(bias < 0)
+//     {
+//             //left
+//             df = (f1*7 - f2*26*4 + f3*19*6 - f4*14*4 + f5*11)*(1/12.0f);
+//     }else
+//     {
+//             //right
+//             df = (f1*3 - f2*16 + f3*36 - f4*48 + f5*25)*(1/12.0f);
+//     }*/
+//     //side ones don't work, use 3 point
+// }
+//
+// //implement an arbitrary derivative
+// //dumb algorithm
+// template < class T >
+// void DerivativeApprox(T &df, const T f[], const Real t[], int npoints, int indexval)
+// {
+//     /*
+//     Lj(x) = PI_i!=j (x - xi) / PI_i!=j (xj - xi)
+//
+//     so Lj'(x) = SUM_k PI_i!=j|k (x - xi) / PI_i!=j (xj - xi)
+//     */
+//
+//     unsigned int i,j,k,i0,i1;
+//
+//     Real Lpj,mult,div,tj;
+//     Real tval = t[indexval];
+//
+//     //sum k
+//     for(j=0;j<npoints;++j)
+//     {
+//             Lpj = 0;
+//             div = 1;
+//             tj = t[j];
+//
+//             for(k=0;k<npoints;++k)
+//             {
+//                     if(k != j) //because there is no summand for k == j, since that term is missing from the original equation
+//                     {
+//                             //summation for k
+//                             for(i=0;i<npoints;++i)
+//                             {
+//                                     if(i != k)
+//                                     {
+//                                             mult *= tval - t[i];
+//                                     }
+//                             }
+//
+//                             Lpj += mult; //add into the summation
+//
+//                             //since the ks follow the exact pattern we need for the divisor (use that too)
+//                             div *= tj - t[k];
+//                     }
+//             }
+//
+//             //get the actual coefficient
+//             Lpj /= div;
+//
+//             //add it in to the equation
+//             df += f[j]*Lpj;
+//     }
+// }
 
 //END numerical derivatives
 
-template < class T >
-inline int sign(T f, T tol)
-{
-       if(f < -tol) return -1;
-       if(f > tol) return 1;
-       return 0;
-}
+// template < class T >
+// inline int sign(T f, T tol)
+// {
+//     if(f < -tol) return -1;
+//     if(f > tol) return 1;
+//     return 0;
+// }
 
 void GetFirstDerivatives(const std::vector<synfig::Point> &f, unsigned int left, unsigned int right, char *out, unsigned int dfstride)
 {
@@ -194,10 +182,10 @@ void GetFirstDerivatives(const std::vector<synfig::Point> &f, unsigned int left,
 
        if(right - left < 2)
                return;
-       else if(right - left < 3)
+       else if(right - left == 2)
        {
                synfig::Vector v = f[left+1] - f[left];
-               
+
                //set both to the one we want
                *(synfig::Vector*)out = v;
                out += dfstride;
@@ -208,57 +196,53 @@ void GetFirstDerivatives(const std::vector<synfig::Point> &f, unsigned int left,
        {
                //left then middle then right
                ThreePointdt(*(synfig::Vector*)out,f[left+0], f[left+1], f[left+2], -1);
-               current += 1;
+               current++;
                out += dfstride;
-               
+
                for(;current < right-1; current++, out += dfstride)
-               {
                        ThreePointdt(*(synfig::Vector*)out,f[current-1], f[current], f[current+1], 0);
-               }
 
                ThreePointdt(*(synfig::Vector*)out,f[right-3], f[right-2], f[right-1], 1);
                current++;
                out += dfstride;
-               
+
        }else //can use 5 point
        {
                //left 2 then middle bunch then right two
                //may want to use 3 point for inner edge ones
-               
+
                FivePointdt(*(synfig::Vector*)out,f[left+0], f[left+1], f[left+2], f[left+3], f[left+4], -2);
                out += dfstride;
                FivePointdt(*(synfig::Vector*)out,f[left+1], f[left+2], f[left+3], f[left+4], f[left+5], -1);
                out += dfstride;
                current += 2;
-               
+
                for(;current < right-2; current++, out += dfstride)
-               {
                        FivePointdt(*(synfig::Vector*)out,f[current-2], f[current-1], f[current], f[current+1], f[current+2], 0);
-               }
 
-               FivePointdt(*(synfig::Vector*)out,f[right-5], f[right-4], f[right-3], f[right-2], f[right-1], 1);
+               FivePointdt(*(synfig::Vector*)out,f[right-6], f[right-5], f[right-4], f[right-3], f[right-2], 2);
                out += dfstride;
-               FivePointdt(*(synfig::Vector*)out,f[right-6], f[right-5], f[right-4], f[right-3], f[right-2], 2);               
+               FivePointdt(*(synfig::Vector*)out,f[right-5], f[right-4], f[right-3], f[right-2], f[right-1], 1);
                out += dfstride;
                current += 2;
        }
 }
 
-void GetSimpleDerivatives(const std::vector<synfig::Point> &f, int left, int right, 
+void GetSimpleDerivatives(const std::vector<synfig::Point> &f, int left, int right,
                                                        std::vector<synfig::Point> &df, int outleft,
-                                                       const std::vector<synfig::Real> &di)
+                                                       const std::vector<synfig::Real> &/*di*/)
 {
        int i1,i2,i;
        int offset = 2; //df = 1/2 (f[i+o]-f[i-o])
-       
+
        assert((int)df.size() >= right-left+outleft); //must be big enough
-       
+
        for(i = left; i < right; ++i)
        {
                //right now indices (figure out distance later)
                i1 = std::max(left,i-offset);
                i2 = std::max(left,i+offset);
-               
+
                df[outleft++] = (f[i2] - f[i1])*0.5f;
        }
 }
@@ -267,115 +251,119 @@ void GetSimpleDerivatives(const std::vector<synfig::Point> &f, int left, int rig
 Real CurveError(const synfig::Point *pts, unsigned int n, std::vector<synfig::Point> &work, int left, int right)
 {
        if(right-left < 2) return -1;
-               
+
        int i,j;
-       
+
        //get distances to each point
        Real d,dtemp,dsum;
        //synfig::Vector v,vt;
        //synfig::Point p1,p2;
        synfig::Point pi;
        std::vector<synfig::Point>::const_iterator it;//,end = work.begin()+right;
-       
+
        //unsigned int size = work.size();
-       
+
        //for each line, get distance
        d = 0; //starts at 0
        for(i = 0; i < (int)n; ++i)
-       {               
+       {
                pi = pts[i];
-               
+
                dsum = FLT_MAX;
-               
+
                it = work.begin()+left;
                //p2 = *it++; //put it at left+1
                for(j = left/*+1*/; j < right; ++j,++it)
                {
                        /*p1 = p2;
                        p2 = *it;
-                       
-                       v = p2 - p1;                    
+
+                       v = p2 - p1;
                        vt = pi - p1;
-                       
+
                        dtemp = v.mag_squared() > 1e-12 ? (vt*v)/v.mag_squared() : 0; //get the projected time value for the current line
-                       
-                       //get distance to line segment with the time value clamped 0-1                  
+
+                       //get distance to line segment with the time value clamped 0-1
                        if(dtemp >= 1)  //use p+v
                        {
-                               vt += v; //makes it pp - (p+v)  
+                               vt += v; //makes it pp - (p+v)
                        }else if(dtemp > 0)     //use vt-proj
                        {
                                vt -= v*dtemp; // vt - proj_v(vt)       //must normalize the projection vector to work
                        }
-                       
+
                        //else use p
                        dtemp = vt.mag_squared();*/
-                       
-                       dtemp = (pi - *it).mag_squared();                       
+
+                       dtemp = (pi - *it).mag_squared();
                        if(dtemp < dsum)
                                dsum = dtemp;
                }
-               
+
                //accumulate the points' min distance from the curve
                d += sqrt(dsum);
        }
-       
+
        return d;
 }
 
 typedef synfigapp::BLineConverter::cpindex cpindex;
 
 //has the index data and the tangent scale data (relevant as it may be)
-int tesselate_curves(const std::vector<cpindex> &inds, const std::vector<Point> &f, const std::vector<synfig::Vector> &df, std::vector<Point> &work)
+int tessellate_curves(const std::vector<cpindex> &inds, const std::vector<Point> &f, const std::vector<synfig::Vector> &df, std::vector<Point> &work)
 {
        if(inds.size() < 2)
                return 0;
-       
+
        etl::hermite<Point>     curve;
        int ntess = 0;
-       
+
        std::vector<cpindex>::const_iterator j = inds.begin(),j2, end = inds.end();
-       
+
        unsigned int ibase = inds[0].curind;
-               
+
        j2 = j++;
        for(; j != end; j2 = j++)
        {
-               //if this curve has invalid error (in j) then retesselate it's work points (requires reparametrization, etc.)
+               //if this curve has invalid error (in j) then retessellate its work points (requires reparametrization, etc.)
                if(j->error < 0)
                {
                        //get the stepsize etc. for the number of points in here
                        unsigned int n = j->curind - j2->curind + 1; //thats the number of points in the span
                        unsigned int k, kend, i0, i3;
                        //so reset the right chunk
-                       
+
                        Real t, dt = 1/(Real)(n*2-2); //assuming that they own only n points
-                       
+
                        //start at first intermediate
                        t = 0;
 
-                       i0 = j2->curind; i3 = j->curind;                        
+                       i0 = j2->curind; i3 = j->curind;
                        k = (i0-ibase)*2; //start on first intermediary point (2x+1)
                        kend = (i3-ibase)*2; //last point to set (not intermediary)
-                       
+
                        //build hermite curve, it's easier
                        curve.p1() = f[i0];
                        curve.p2() = f[i3];
-                       curve.t1() = df[i0]*(df[i0].mag_squared() > 1e-4 ? j2->tangentscale/df[i0].mag() : j2->tangentscale);
-                       curve.t2() = df[i3]*(df[i3].mag_squared() > 1e-4 ? j->tangentscale/df[i3].mag() : j->tangentscale);
+                       curve.t1() = df[i0-ibase] * (df[i0-ibase].mag_squared() > 1e-4
+                                                                                ? j2->tangentscale/df[i0-ibase].mag()
+                                                                                : j2->tangentscale);
+                       curve.t2() = df[i3-ibase] * (df[i3-ibase].mag_squared() > 1e-4
+                                                                                ? j->tangentscale/df[i3-ibase].mag()
+                                                                                : j->tangentscale);
                        curve.sync();
-                                               
+
                        //MUST include the end point (since we are ignoring left one)
                        for(; k < kend; ++k, t += dt)
                        {
                                work[k] = curve(t);
                        }
-                       
+
                        work[k] = curve(1); //k == kend, t == 1 -> c(t) == p2
                        ++ntess;
                }
        }
-       
+
        return ntess;
 }
 
@@ -386,24 +374,26 @@ synfigapp::BLineConverter::BLineConverter()
        width = 0;
 };
 
-void 
+void
 synfigapp::BLineConverter::clear()
 {
-       f.clear();
-       f_w.clear();
+       point_cache.clear();
+       width_cache.clear();
        ftemp.clear();
-       df.clear();
-       cvt.clear();
-       brk.clear();
-       di.clear();
-       d_i.clear();
+       deriv.clear();
+       curvature.clear();
+       break_tangents.clear();
+       cum_dist.clear();
+       this_dist.clear();
        work.clear();
        curind.clear();
 }
 
 void
-synfigapp::BLineConverter::operator () (std::list<synfig::BLinePoint> &out, const std::list<synfig::Point> &in,const std::list<synfig::Real> &in_w)
-{      
+synfigapp::BLineConverter::operator()(std::list<synfig::BLinePoint>  &blinepoints_out,
+                                                                         const std::list<synfig::Point> &points_in,
+                                                                         const std::list<synfig::Real>  &widths_in)
+{
        //Profiling information
        /*etl::clock::value_type initialprocess=0, curveval=0, breakeval=0, disteval=0;
        etl::clock::value_type preproceval=0, tesseval=0, erroreval=0, spliteval=0;
@@ -411,292 +401,309 @@ synfigapp::BLineConverter::operator () (std::list<synfig::BLinePoint> &out, cons
        etl::clock_realtime timer,total;*/
 
        //total.reset();
-       if(in.size()<=1)
+       if (points_in.size() < 2)
                return;
 
        clear();
-       
+
        //removing digitization error harder than expected
-       
+
        //intended to fix little pen errors caused by the way people draw (tiny juts in opposite direction)
        //Different solutions
        //      Average at both end points (will probably eliminate many points at each end of the samples)
        //      Average after the break points are found (weird points would still affect the curve)
        //      Just always get rid of breaks at the beginning and end if they are a certain distance apart
        //              This is will be current approach so all we do now is try to remove duplicate points
-       
+
        //remove duplicate points - very bad for fitting
-       
+
        //timer.reset();
-       
+
        {
-               std::list<synfig::Point>::const_iterator i = in.begin(), end = in.end();
-               std::list<synfig::Real>::const_iterator iw = in_w.begin();
-               synfig::Point   c;
-               
-               if(in.size() == in_w.size())
-               {
-                       for(;i != end; ++i,++iw)
-                       {       
-                               //eliminate duplicate points
-                               if(*i != c)
-                               {
-                                       f.push_back(c = *i);
-                                       f_w.push_back(*iw);
-                               }
-                       }
-               }else
+               std::list<synfig::Point>::const_iterator point_iter = points_in.begin(), end = points_in.end();
+               std::list<synfig::Real>::const_iterator width_iter = widths_in.begin();
+               synfig::Point c;
+
+               if (points_in.size() == widths_in.size())
                {
-                       for(;i != end; ++i)
-                       {       
-                               //eliminate duplicate points
-                               if(*i != c)
+                       for(bool first = true; point_iter != end; ++point_iter,++width_iter)
+                               if (first || *point_iter != c)          // eliminate duplicate points
                                {
-                                       f.push_back(c = *i);
+                                       first = false;
+                                       point_cache.push_back(c = *point_iter);
+                                       width_cache.push_back(*width_iter);
                                }
-                       }
                }
+               else
+                       for(;point_iter != end; ++point_iter)
+                               if(*point_iter != c)            // eliminate duplicate points
+                                       point_cache.push_back(c = *point_iter);
        }
        //initialprocess = timer();
-       
-       if(f.size()<=6)
+
+       if (point_cache.size() < 7)
+       {
+               info("only %d unique points - giving up", point_cache.size());
                return;
-       
+       }
+
        //get curvature information
        //timer.reset();
-       
+
        {
-               int i,i0,i1;
-               synfig::Vector v1,v2;
-               
-               cvt.resize(f.size());
-               
-               cvt.front() = 1;
-               cvt.back() = 1;
-               
-               for(i = 1; i < (int)f.size()-1; ++i)
+               int i_this, i_prev, i_next;
+               synfig::Vector v_prev, v_next;
+
+               curvature.resize(point_cache.size());
+               curvature.front() = curvature.back() = 1;
+
+               for (i_this = 1; i_this < (int)point_cache.size()-1; i_this++)
                {
-                       i0 = std::max(0,i - 2);
-                       i1 = std::min((int)(f.size()-1),i + 2);
-                       
-                       v1 = f[i] - f[i0];
-                       v2 = f[i1] - f[i];
-       
-                       cvt[i] = (v1*v2)/(v1.mag()*v2.mag());
+                       i_prev = std::max(0, i_this-2);
+                       i_next = std::min((int)(point_cache.size()-1), i_this+2);
+
+                       v_prev = point_cache[i_this] - point_cache[i_prev];
+                       v_next = point_cache[i_next] - point_cache[i_this];
+
+                       curvature[i_this] = (v_prev*v_next) / (v_prev.mag()*v_next.mag());
                }
        }
-       
+
        //curveval = timer();
        //synfig::info("calculated curvature");
-       
+
        //find corner points and interpolate inside those
        //timer.reset();
-       {               
+       {
                //break at sharp derivative points
                //TODO tolerance should be set based upon digitization resolution (length dependent index selection)
                Real    tol = 0;                //break tolerance, for the cosine of the change in angle (really high curvature or something)
-               Real    fixdistsq = 4*width*width; //the distance to ignore breaks at the end points (for fixing stuff)
                unsigned int i = 0;
-               
-               int             maxi = -1, last=0;
-               Real    minc = 1;
-               
-               brk.push_back(0);
-               
-               for(i = 1; i < cvt.size()-1; ++i)
-               {                       
-                       //insert if too sharp (we need to break the tangents to insert onto the break list)
-                       
-                       if(cvt[i] < tol)
+
+               int             sharpest_i=-1;
+               int             last=0;
+               Real    sharpest_curvature = 1;
+
+               break_tangents.push_back(0);
+
+               // loop through the curvatures; in each continuous run of
+               // curvatures that exceed the tolerence, find the one with the
+               // sharpest curvature and add its index to the list of indices
+               // at which to split tangents
+               for (i = 1; i < curvature.size()-1; ++i)
+               {
+                       if (curvature[i] < tol)
                        {
-                               if(cvt[i] < minc)
+                               if(curvature[i] < sharpest_curvature)
                                {
-                                       minc = cvt[i];
-                                       maxi = i;
+                                       sharpest_curvature = curvature[i];
+                                       sharpest_i = i;
                                }
-                       }else if(maxi >= 0)
+                       }
+                       else if (sharpest_i > 0)
                        {
-                               if(maxi >= last + 8)
+                               // don't have 2 corners too close to each other
+                               if (sharpest_i >= last + 8) //! \todo make this configurable
                                {
-                                       //synfig::info("break: %d-%d",maxi+1,cvt.size());                                               
-                                       brk.push_back(maxi);
-                                       last = maxi;
+                                       //synfig::info("break: %d-%d",sharpest_i+1,curvature.size());
+                                       break_tangents.push_back(sharpest_i);
+                                       last = sharpest_i;
                                }
-                               maxi = -1;
-                               minc = 1;
+                               sharpest_i = -1;
+                               sharpest_curvature = 1;
                        }
                }
-               
-               brk.push_back(i);
-               
-               //postprocess for breaks too close to eachother
+
+               break_tangents.push_back(i);
+
+// this section causes bug 1892566 if enabled
+#if 1
+               //postprocess for breaks too close to each other
+               Real    fixdistsq = 4*width*width; //the distance to ignore breaks at the end points (for fixing stuff)
                Real d = 0;
-               Point p = f[brk.front()];
-               
+               Point p = point_cache[break_tangents.front()];
+
                //first set
-               for(i = 1; i < brk.size()-1; ++i) //do not want to include end point...
+               for (i = 1; i < break_tangents.size()-1; ++i) //do not want to include end point...
                {
-                       d = (f[brk[i]] - p).mag_squared();
-                       if(d > fixdistsq) break; //don't want to group breaks if we ever get over the dist... 
+                       d = (point_cache[break_tangents[i]] - p).mag_squared();
+                       if(d > fixdistsq) break; //don't want to group breaks if we ever get over the dist...
                }
                //want to erase all points before...
                if(i != 1)
-                       brk.erase(brk.begin(),brk.begin()+i-1); 
-               
+                       break_tangents.erase(break_tangents.begin(),break_tangents.begin()+i-1);
+
                //end set
-               p = f[brk.back()];
-               for(i = brk.size()-2; i > 0; --i) //start at one in from the end
+               p = point_cache[break_tangents.back()];
+               for(i = break_tangents.size()-2; i > 0; --i) //start at one in from the end
                {
-                       d = (f[brk[i]] - p).mag_squared();
+                       d = (point_cache[break_tangents[i]] - p).mag_squared();
                        if(d > fixdistsq) break; //don't want to group breaks if we ever get over the dist
                }
-               if(i != brk.size()-2)
-                       brk.erase(brk.begin()+i+2,brk.end()); //erase all points that we found... found none if i has not advanced
+               if(i != break_tangents.size()-2)
+                       break_tangents.erase(break_tangents.begin()+i+2,break_tangents.end()); //erase all points that we found... found none if i has not advanced
                //must not include the one we ended up on
+#endif
        }
        //breakeval = timer();
-       //synfig::info("found break points: %d",brk.size());
-       
+       //synfig::info("found break points: %d",break_tangents.size());
+
        //get the distance calculation of the entire curve (for tangent scaling)
 
        //timer.reset();
        {
                synfig::Point p1,p2;
-               
-               p1=p2=f[0];
-               
-               di.resize(f.size()); d_i.resize(f.size());
+
+               p1=p2=point_cache[0];
+
+               cum_dist.resize(point_cache.size()); this_dist.resize(point_cache.size());
                Real d = 0;
-               for(unsigned int i = 0; i < f.size();)
+               for(unsigned int i = 0; i < point_cache.size();)
                {
-                       d += (d_i[i] = (p2-p1).mag());
-                       di[i] = d;
-                       
+                       d += (this_dist[i] = (p2-p1).mag());
+                       cum_dist[i] = d;
+
                        p1=p2;
-                       p2=f[++i];
+                       //! \todo is this legal?  it reads off the end of the vector
+                       p2=point_cache[++i];
                }
        }
        //disteval = timer();
        //synfig::info("calculated distance");
-               
+
        //now break at every point - calculate new derivatives each time
-       
+
        //TODO
        //must be sure that the break points are 3 or more apart
        //then must also store the breaks which are not smooth, etc.
        //and figure out tangents between there
-       
+
        //for each pair of break points (as long as they are far enough apart) recursively subdivide stuff
        //ignore the detected intermediate points
        {
                unsigned int i0=0,i3=0,is=0;
                int i=0,j=0;
-               
+
                bool done = false;
-               
-               Real errortol = smoothness*pixelwidth; //???? what the hell should this value be
-               
+
+               Real errortol = smoothness*pixelwidth; //???? what should this value be
+
                BLinePoint a;
                synfig::Vector v;
-               
+
                //intemp = f; //don't want to smooth out the corners
-               
+
                bool breaktan = false, setwidth;
                a.set_split_tangent_flag(false);
                //a.set_width(width);
                a.set_width(1.0f);
-               
-               setwidth = (f.size() == f_w.size());
-               
-               for(j = 0; j < (int)brk.size() - 1; ++j)
+
+               setwidth = (point_cache.size() == width_cache.size());
+
+               for(j = 0; j < (int)break_tangents.size() - 1; ++j)
                {
                        //for b[j] to b[j+1] subdivide and stuff
-                       i0 = brk[j];
-                       i3 = brk[j+1];
-                       
+                       i0 = break_tangents[j];
+                       i3 = break_tangents[j+1];
+
                        unsigned int size = i3-i0+1; //must include the end points
-                       
+
                        //new derivatives
                        //timer.reset();
-                       ftemp.assign(f.begin()+i0, f.begin()+i3+1);
+                       ftemp.assign(point_cache.begin()+i0, point_cache.begin()+i3+1);
                        for(i=0;i<20;++i)
                                gaussian_blur_3(ftemp.begin(),ftemp.end(),false);
-                       
-                       df.resize(size);
-                       GetFirstDerivatives(ftemp,0,size,(char*)&df[0],sizeof(df[0]));
-                       //GetSimpleDerivatives(ftemp,0,size,df,0,di); 
-                       //< don't have to worry about indexing stuff as it is all being taken car of right now
+
+                       deriv.resize(size);
+
+                       // Wondering whether the modification of the deriv vector
+                       // using a char* pointer and pointer arithmetric was safe,
+                       // I looked it up...
+                       //
+                       // http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2369.pdf tells me:
+                       //
+                       //      23.2.5  Class template vector [vector]
+                       //
+                       //      [...] The elements of a vector are stored contiguously,
+                       //      meaning that if v is a vector<T,Allocator> where T is
+                       //      some type other than bool, then it obeys the identity
+                       //      &v[n] == &v[0] + n for all 0 <= n < v.size().
+                       //
+                       GetFirstDerivatives(ftemp,0,size,(char*)&deriv[0],sizeof(deriv[0]));
+
+                       //GetSimpleDerivatives(ftemp,0,size,deriv,0,cum_dist);
+                       //< don't have to worry about indexing stuff as it is all being taken care of right now
                        //preproceval += timer();
                        //numpre++;
-                       
-                       work.resize(size*2-1); //guarantee that all points will be tesselated correctly (one point inbetween every 2 adjacent points)
-                       
+
+                       work.resize(size*2-1); //guarantee that all points will be tessellated correctly (one point in between every 2 adjacent points)
+
                        //if size of work is size*2-1, the step size should be 1/(size*2 - 2)
                        //Real step = 1/(Real)(size*2 - 1);
-                       
+
                        //start off with break points as indices
                        curind.clear();
-                       curind.push_back(cpindex(i0,di[i3]-di[i0],0)); //0 error because no curve on the left
-                       curind.push_back(cpindex(i3,di[i3]-di[i0],-1)); //error needs to be reevaluated
+                       curind.push_back(cpindex(i0,cum_dist[i3]-cum_dist[i0],0)); //0 error because no curve on the left
+                       curind.push_back(cpindex(i3,cum_dist[i3]-cum_dist[i0],-1)); //error needs to be reevaluated
                        done = false; //we want to loop
-                       
+
                        unsigned int dcount = 0;
-                       
-                       //while there are still enough points between us, and the error is too high subdivide (and invalidate neighbors that share tangents)            
+
+                       //while there are still enough points between us, and the error is too high subdivide (and invalidate neighbors that share tangents)
                        while(!done)
-                       {                                       
-                               //tesselate all curves with invalid error values
-                               work[0] = f[i0];
-                               
+                       {
+                               //tessellate all curves with invalid error values
+                               work[0] = point_cache[i0];
+
                                //timer.reset();
-                               /*numtess += */tesselate_curves(curind,f,df,work);
+                               /*numtess += */tessellate_curves(curind,point_cache,deriv,work);
                                //tesseval += timer();
-                               
+
                                //now get all error values
                                //timer.reset();
                                for(i = 1; i < (int)curind.size(); ++i)
                                {
-                                       if(curind[i].error < 0) //must have been retesselated, so now recalculate error value
+                                       if(curind[i].error < 0) //must have been retessellated, so now recalculate error value
                                        {
                                                //evaluate error from points (starting at current index)
                                                int size = curind[i].curind - curind[i-1].curind + 1;
-                                               curind[i].error = CurveError(&f[curind[i-1].curind], size,
+                                               curind[i].error = CurveError(&point_cache[curind[i-1].curind], size,
                                                                                                         work,(curind[i-1].curind - i0)*2,(curind[i].curind - i0)*2+1);
-                                               
+
                                                /*if(curind[i].error > 1.0e5)
                                                {
                                                        synfig::info("Holy crap %d-%d error %f",curind[i-1].curind,curind[i].curind,curind[i].error);
                                                        curind[i].error = -1;
-                                                       numtess += tesselate_curves(curind,f,df,work);
-                                                       curind[i].error = CurveError(&f[curind[i-1].curind], size,
+                                                       numtess += tessellate_curves(curind,f,deriv,work);
+                                                       curind[i].error = CurveError(&point_cache[curind[i-1].curind], size,
                                                                                                         work,0,work.size());//(curind[i-1].curind - i0)*2,(curind[i].curind - i0)*2+1);
                                                }*/
                                                //numerror++;
                                        }
                                }
                                //erroreval += timer();
-                               
+
                                //assume we're done
                                done = true;
-                               
+
                                //check each error to see if it's too big, if so, then subdivide etc.
                                int indsize = (int)curind.size();
                                Real maxrelerror = 0;
                                int maxi = -1;//, numpoints;
-                               
+
                                //timer.reset();
                                //get the maximum error and split there
                                for(i = 1; i < indsize; ++i)
                                {
                                        //numpoints = curind[i].curind - curind[i-1].curind + 1;
-                                       
+
                                        if(curind[i].error > maxrelerror && curind[i].curind - curind[i-1].curind > 2) //only accept if it's valid
                                        {
                                                maxrelerror = curind[i].error;
                                                maxi = i;
                                        }
                                }
-                               
+
                                //split if error is too great
                                if(maxrelerror > errortol)
                                {
@@ -704,11 +711,11 @@ synfigapp::BLineConverter::operator () (std::list<synfig::BLinePoint> &out, cons
                                        unsigned int    ibase = curind[maxi-1].curind, itop = curind[maxi].curind,
                                                                        ibreak = (ibase + itop)/2;
                                        Real scale, scale2;
-                                       
-                                       assert(ibreak < f.size());
-                                       
+
+                                       assert(ibreak < point_cache.size());
+
                                        //synfig::info("Split %d -%d- %d, error: %f", ibase,ibreak,itop,maxrelerror);
-                                       
+
                                        if(ibase != itop)
                                        {
                                                //invalidate current error of the changed tangents and add an extra segment
@@ -716,89 +723,89 @@ synfigapp::BLineConverter::operator () (std::list<synfig::BLinePoint> &out, cons
                                                curind[maxi].error = -1;
                                                curind[maxi-1].error = -1;
                                                if(maxi+1 < indsize) curind[maxi+1].error = -1; //if there is a curve segment beyond this it will be effected as well
-                                               
-                                               scale = di[itop] - di[ibreak];
-                                               scale2 = maxi+1 < indsize ? di[curind[maxi+1].curind] - di[itop] : scale; //to the right valid?
+
+                                               scale = cum_dist[itop] - cum_dist[ibreak];
+                                               scale2 = maxi+1 < indsize ? cum_dist[curind[maxi+1].curind] - cum_dist[itop] : scale; //to the right valid?
                                                curind[maxi].tangentscale = std::min(scale, scale2);
-                                                                                               
-                                               scale = di[ibreak] - di[ibase];
-                                               scale2 = maxi >= 2 ? di[ibase] - di[curind[maxi-2].curind] : scale; // to the left valid -2 ?
+
+                                               scale = cum_dist[ibreak] - cum_dist[ibase];
+                                               scale2 = maxi >= 2 ? cum_dist[ibase] - cum_dist[curind[maxi-2].curind] : scale; // to the left valid -2 ?
                                                curind[maxi-1].tangentscale = std::min(scale, scale2);
-                                               
-                                               scale = std::min(di[ibreak] - di[ibase], di[itop] - di[ibreak]);
-                                               
+
+                                               scale = std::min(cum_dist[ibreak] - cum_dist[ibase], cum_dist[itop] - cum_dist[ibreak]);
+
                                                curind.insert(curind.begin()+maxi,cpindex(ibreak, scale, -1));
                                                //curind.push_back(cpindex(ibreak, scale, -1));
                                                //std::sort(curind.begin(), curind.end());
-                                               
+
                                                done = false;
                                                //numsplit++;
                                        }
                                }
                                //spliteval += timer();
-                               
+
                                dcount++;
                        }
-       
-                       //insert the last point too (just set tangent for now                   
+
+                       //insert the last point too (just set tangent for now
                        is = curind[0].curind;
-                       
-                       //first point inherits current tangent status                   
-                       v = df[is - i0];
+
+                       //first point inherits current tangent status
+                       v = deriv[is - i0];
                        if(v.mag_squared() > EPSILON)
                                v *= (curind[0].tangentscale/v.mag());
-                                                       
+
                        if(!breaktan)
                                a.set_tangent(v);
                        else a.set_tangent2(v);
-                       
-                       a.set_vertex(f[is]);
-                       if(setwidth)a.set_width(f_w[is]);
-                       
-                       out.push_back(a);
+
+                       a.set_vertex(point_cache[is]);
+                       if(setwidth)a.set_width(width_cache[is]);
+
+                       blinepoints_out.push_back(a);
                        a.set_split_tangent_flag(false); //won't need to break anymore
                        breaktan = false;
-                       
+
                        for(i = 1; i < (int)curind.size()-1; ++i)
                        {
                                is = curind[i].curind;
-                               
+
                                //first point inherits current tangent status
-                               v = df[is-i0];
+                               v = deriv[is-i0];
                                if(v.mag_squared() > EPSILON)
                                        v *= (curind[i].tangentscale/v.mag());
-                                                               
+
                                a.set_tangent(v); // always inside, so guaranteed to be smooth
-                               a.set_vertex(f[is]);
-                               if(setwidth)a.set_width(f_w[is]);
-                               
-                               out.push_back(a);
+                               a.set_vertex(point_cache[is]);
+                               if(setwidth)a.set_width(width_cache[is]);
+
+                               blinepoints_out.push_back(a);
                        }
-                       
+
                        //set the last point's data
                        is = curind.back().curind; //should already be this
-                       
-                       v = df[is-i0];
+
+                       v = deriv[is-i0];
                        if(v.mag_squared() > EPSILON)
                                v *= (curind.back().tangentscale/v.mag());
-                       
+
                        a.set_tangent1(v);
                        a.set_split_tangent_flag(true);
                        breaktan = true;
-                       
+
                        //will get the vertex and tangent 2 from next round
                }
-               
-               a.set_vertex(f[i3]);
+
+               a.set_vertex(point_cache[i3]);
                a.set_split_tangent_flag(false);
                if(setwidth)
-                       a.set_width(f_w[i3]);
-               out.push_back(a);
-               
+                       a.set_width(width_cache[i3]);
+               blinepoints_out.push_back(a);
+
                /*etl::clock::value_type totaltime = total(),
                                                           misctime = totaltime - initialprocess - curveval - breakeval - disteval
                                                                          - preproceval - tesseval - erroreval - spliteval;
-               
+
                synfig::info(
                        "Curve Convert Profile:\n"
                        "\tInitial Preprocess:    %f\n"
@@ -807,16 +814,16 @@ synfigapp::BLineConverter::operator () (std::list<synfig::BLinePoint> &out, cons
                        "\tDistance Calculation:  %f\n"
                        "  Algorithm: (numtimes,totaltime)\n"
                        "\tPreprocess step:      (%d,%f)\n"
-                       "\tTesselation step:     (%d,%f)\n"
+                       "\tTessellation step:    (%d,%f)\n"
                        "\tError step:           (%d,%f)\n"
                        "\tSplit step:           (%d,%f)\n"
                        "  Num Input: %d, Num Output: %d\n"
                        "  Total time: %f, Misc time: %f\n",
                        initialprocess, curveval,breakeval,disteval,
                        numpre,preproceval,numtess,tesseval,numerror,erroreval,numsplit,spliteval,
-                       in.size(),out.size(),
+                       points_in.size(),blinepoints_out.size(),
                        totaltime,misctime);*/
-               
+
                return;
        }
 }
@@ -825,12 +832,8 @@ void synfigapp::BLineConverter::EnforceMinWidth(std::list<synfig::BLinePoint> &b
 {
        std::list<synfig::BLinePoint>::iterator i = bline.begin(),
                                                                                        end = bline.end();
-       
+
        for(i = bline.begin(); i != end; ++i)
-       {
                if(i->get_width() < min_pressure)
-               {
                        i->set_width(min_pressure);
-               }
-       }
 }