synfig-studio/trunk/src/synfigapp/blineconvert.cpp

   1 /* === S Y N F I G ========================================================= */
   2 /*!     \file blineconvert.cpp
   3 **      \brief Template File
   4 **
   5 **      $Id$
   6 **
   7 **      \legal
   8 **      Copyright (c) 2002-2005 Robert B. Quattlebaum Jr., Adrian Bentley
   9 **
  10 **      This package is free software; you can redistribute it and/or
  11 **      modify it under the terms of the GNU General Public License as
  12 **      published by the Free Software Foundation; either version 2 of
  13 **      the License, or (at your option) any later version.
  14 **
  15 **      This package is distributed in the hope that it will be useful,
  16 **      but WITHOUT ANY WARRANTY; without even the implied warranty of
  17 **      MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  18 **      General Public License for more details.
  19 **      \endlegal
  20 */
  21 /* ========================================================================= */
  22
  23 /* === H E A D E R S ======================================================= */
  24
  25 #ifdef USING_PCH
  26 #       include "pch.h"
  27 #else
  28 #ifdef HAVE_CONFIG_H
  29 #       include <config.h>
  30 #endif
  31
  32 #include "blineconvert.h"
  33 #include <vector>
  34 #include <ETL/gaussian>
  35 #include <ETL/hermite>
  36 #include <ETL/clock>
  37 #include <float.h>
  38 #include <algorithm>
  39 #include <synfig/general.h>
  40 #include <cassert>
  41
  42
  43
  44 #endif
  45
  46 /* === U S I N G =========================================================== */
  47
  48 using namespace std;
  49 using namespace etl;
  50 using namespace synfig;
  51
  52 /* === M A C R O S ========================================================= */
  53
  54 #define EPSILON         (1e-10)
  55
  56 /* === G L O B A L S ======================================================= */
  57
  58 /* === P R O C E D U R E S ================================================= */
  59
  60 /* === M E T H O D S ======================================================= */
  61
  62
  63 //Derivative Functions for numerical approximation
  64
  65 //bias == 0 will get F' at f3, bias < 0 will get F' at f1, and bias > 0 will get F' at f5
  66 template < class T >
  67 inline void FivePointdt(T &df, const T &f1, const T &f2, const T &f3, const T &f4, const T &f5, int bias)
  68 {
  69         if(bias == 0)
  70         {
  71                 //middle
  72                 df = (f1 - f2*8 + f4*8 - f5)*(1/12.0f);
  73         }else if(bias < 0)
  74         {
  75                 //left
  76                 df = (-f1*25 + f2*48 - f3*36 + f4*16 - f5*3)*(1/12.0f);
  77         }else
  78         {
  79                 //right
  80                 df = (f1*3 - f2*16 + f3*36 - f4*48 + f5*25)*(1/12.0f);
  81         }
  82 }
  83
  84 template < class T >
  85 inline void ThreePointdt(T &df, const T &f1, const T &f2, const T &f3, int bias)
  86 {
  87         if(bias == 0)
  88         {
  89                 //middle
  90                 df = (-f1 + f3)*(1/2.0f);
  91         }else if(bias < 0)
  92         {
  93                 //left
  94                 df = (-f1*3 + f2*4 - f3)*(1/2.0f);
  95         }else
  96         {
  97                 //right
  98                 df = (f1 - f2*4 + f3*3)*(1/2.0f);
  99         }
 100 }
 101
 102 // template < class T >
 103 // inline void ThreePointddt(T &df, const T &f1, const T &f2, const T &f3, int bias)
 104 // {
 105 //      // a 3 point approximation pretends to have constant acceleration,
 106 //      // so only one algorithm needed for left, middle, or right
 107 //      df = (f1 -f2*2 + f3)*(1/2.0f);
 108 // }
 109 //
 110 // // WARNING -- totally broken
 111 // template < class T >
 112 // inline void FivePointddt(T &df, const T &f1, const T &f2, const T &f3, int bias)
 113 // {
 114 //      if(bias == 0)
 115 //      {
 116 //              assert(0); // !?
 117 //              //middle
 118 //              //df = (- f1 + f2*16 - f3*30 +  f4*16 - f5)*(1/12.0f);
 119 //      }/*else if(bias < 0)
 120 //      {
 121 //              //left
 122 //              df = (f1*7 - f2*26*4 + f3*19*6 - f4*14*4 + f5*11)*(1/12.0f);
 123 //      }else
 124 //      {
 125 //              //right
 126 //              df = (f1*3 - f2*16 + f3*36 - f4*48 + f5*25)*(1/12.0f);
 127 //      }*/
 128 //      //side ones don't work, use 3 point
 129 // }
 130 //
 131 // //implement an arbitrary derivative
 132 // //dumb algorithm
 133 // template < class T >
 134 // void DerivativeApprox(T &df, const T f[], const Real t[], int npoints, int indexval)
 135 // {
 136 //      /*
 137 //      Lj(x) = PI_i!=j (x - xi) / PI_i!=j (xj - xi)
 138 //
 139 //      so Lj'(x) = SUM_k PI_i!=j|k (x - xi) / PI_i!=j (xj - xi)
 140 //      */
 141 //
 142 //      unsigned int i,j,k,i0,i1;
 143 //
 144 //      Real Lpj,mult,div,tj;
 145 //      Real tval = t[indexval];
 146 //
 147 //      //sum k
 148 //      for(j=0;j<npoints;++j)
 149 //      {
 150 //              Lpj = 0;
 151 //              div = 1;
 152 //              tj = t[j];
 153 //
 154 //              for(k=0;k<npoints;++k)
 155 //              {
 156 //                      if(k != j) //because there is no summand for k == j, since that term is missing from the original equation
 157 //                      {
 158 //                              //summation for k
 159 //                              for(i=0;i<npoints;++i)
 160 //                              {
 161 //                                      if(i != k)
 162 //                                      {
 163 //                                              mult *= tval - t[i];
 164 //                                      }
 165 //                              }
 166 //
 167 //                              Lpj += mult; //add into the summation
 168 //
 169 //                              //since the ks follow the exact pattern we need for the divisor (use that too)
 170 //                              div *= tj - t[k];
 171 //                      }
 172 //              }
 173 //
 174 //              //get the actual coefficient
 175 //              Lpj /= div;
 176 //
 177 //              //add it in to the equation
 178 //              df += f[j]*Lpj;
 179 //      }
 180 // }
 181
 182 //END numerical derivatives
 183
 184 // template < class T >
 185 // inline int sign(T f, T tol)
 186 // {
 187 //      if(f < -tol) return -1;
 188 //      if(f > tol) return 1;
 189 //      return 0;
 190 // }
 191
 192 void GetFirstDerivatives(const std::vector<synfig::Point> &f, unsigned int left, unsigned int right, char *out, unsigned int dfstride)
 193 {
 194         unsigned int current = left;
 195
 196         if(right - left < 2)
 197                 return;
 198         else if(right - left < 3)
 199         {
 200                 synfig::Vector v = f[left+1] - f[left];
 201
 202                 //set both to the one we want
 203                 *(synfig::Vector*)out = v;
 204                 out += dfstride;
 205                 *(synfig::Vector*)out = v;
 206                 out += dfstride;
 207         }
 208         else if(right - left < 6/*5*/) //should use 3 point
 209         {
 210                 //left then middle then right
 211                 ThreePointdt(*(synfig::Vector*)out,f[left+0], f[left+1], f[left+2], -1);
 212                 current += 1;
 213                 out += dfstride;
 214
 215                 for(;current < right-1; current++, out += dfstride)
 216                 {
 217                         ThreePointdt(*(synfig::Vector*)out,f[current-1], f[current], f[current+1], 0);
 218                 }
 219
 220                 ThreePointdt(*(synfig::Vector*)out,f[right-3], f[right-2], f[right-1], 1);
 221                 current++;
 222                 out += dfstride;
 223
 224         }else //can use 5 point
 225         {
 226                 //left 2 then middle bunch then right two
 227                 //may want to use 3 point for inner edge ones
 228
 229                 FivePointdt(*(synfig::Vector*)out,f[left+0], f[left+1], f[left+2], f[left+3], f[left+4], -2);
 230                 out += dfstride;
 231                 FivePointdt(*(synfig::Vector*)out,f[left+1], f[left+2], f[left+3], f[left+4], f[left+5], -1);
 232                 out += dfstride;
 233                 current += 2;
 234
 235                 for(;current < right-2; current++, out += dfstride)
 236                 {
 237                         FivePointdt(*(synfig::Vector*)out,f[current-2], f[current-1], f[current], f[current+1], f[current+2], 0);
 238                 }
 239
 240                 FivePointdt(*(synfig::Vector*)out,f[right-5], f[right-4], f[right-3], f[right-2], f[right-1], 1);
 241                 out += dfstride;
 242                 FivePointdt(*(synfig::Vector*)out,f[right-6], f[right-5], f[right-4], f[right-3], f[right-2], 2);
 243                 out += dfstride;
 244                 current += 2;
 245         }
 246 }
 247
 248 void GetSimpleDerivatives(const std::vector<synfig::Point> &f, int left, int right,
 249                                                         std::vector<synfig::Point> &df, int outleft,
 250                                                         const std::vector<synfig::Real> &/*di*/)
 251 {
 252         int i1,i2,i;
 253         int offset = 2; //df = 1/2 (f[i+o]-f[i-o])
 254
 255         assert((int)df.size() >= right-left+outleft); //must be big enough
 256
 257         for(i = left; i < right; ++i)
 258         {
 259                 //right now indices (figure out distance later)
 260                 i1 = std::max(left,i-offset);
 261                 i2 = std::max(left,i+offset);
 262
 263                 df[outleft++] = (f[i2] - f[i1])*0.5f;
 264         }
 265 }
 266
 267 //get the curve error from the double sample list of work points (hopefully that's enough)
 268 Real CurveError(const synfig::Point *pts, unsigned int n, std::vector<synfig::Point> &work, int left, int right)
 269 {
 270         if(right-left < 2) return -1;
 271
 272         int i,j;
 273
 274         //get distances to each point
 275         Real d,dtemp,dsum;
 276         //synfig::Vector v,vt;
 277         //synfig::Point p1,p2;
 278         synfig::Point pi;
 279         std::vector<synfig::Point>::const_iterator it;//,end = work.begin()+right;
 280
 281         //unsigned int size = work.size();
 282
 283         //for each line, get distance
 284         d = 0; //starts at 0
 285         for(i = 0; i < (int)n; ++i)
 286         {
 287                 pi = pts[i];
 288
 289                 dsum = FLT_MAX;
 290
 291                 it = work.begin()+left;
 292                 //p2 = *it++; //put it at left+1
 293                 for(j = left/*+1*/; j < right; ++j,++it)
 294                 {
 295                         /*p1 = p2;
 296                         p2 = *it;
 297
 298                         v = p2 - p1;
 299                         vt = pi - p1;
 300
 301                         dtemp = v.mag_squared() > 1e-12 ? (vt*v)/v.mag_squared() : 0; //get the projected time value for the current line
 302
 303                         //get distance to line segment with the time value clamped 0-1
 304                         if(dtemp >= 1)  //use p+v
 305                         {
 306                                 vt += v; //makes it pp - (p+v)
 307                         }else if(dtemp > 0)     //use vt-proj
 308                         {
 309                                 vt -= v*dtemp; // vt - proj_v(vt)       //must normalize the projection vector to work
 310                         }
 311
 312                         //else use p
 313                         dtemp = vt.mag_squared();*/
 314
 315                         dtemp = (pi - *it).mag_squared();
 316                         if(dtemp < dsum)
 317                                 dsum = dtemp;
 318                 }
 319
 320                 //accumulate the points' min distance from the curve
 321                 d += sqrt(dsum);
 322         }
 323
 324         return d;
 325 }
 326
 327 typedef synfigapp::BLineConverter::cpindex cpindex;
 328
 329 //has the index data and the tangent scale data (relevant as it may be)
 330 int tessellate_curves(const std::vector<cpindex> &inds, const std::vector<Point> &f, const std::vector<synfig::Vector> &df, std::vector<Point> &work)
 331 {
 332         if(inds.size() < 2)
 333                 return 0;
 334
 335         etl::hermite<Point>     curve;
 336         int ntess = 0;
 337
 338         std::vector<cpindex>::const_iterator j = inds.begin(),j2, end = inds.end();
 339
 340         unsigned int ibase = inds[0].curind;
 341
 342         j2 = j++;
 343         for(; j != end; j2 = j++)
 344         {
 345                 //if this curve has invalid error (in j) then retessellate its work points (requires reparametrization, etc.)
 346                 if(j->error < 0)
 347                 {
 348                         //get the stepsize etc. for the number of points in here
 349                         unsigned int n = j->curind - j2->curind + 1; //thats the number of points in the span
 350                         unsigned int k, kend, i0, i3;
 351                         //so reset the right chunk
 352
 353                         Real t, dt = 1/(Real)(n*2-2); //assuming that they own only n points
 354
 355                         //start at first intermediate
 356                         t = 0;
 357
 358                         i0 = j2->curind; i3 = j->curind;
 359                         k = (i0-ibase)*2; //start on first intermediary point (2x+1)
 360                         kend = (i3-ibase)*2; //last point to set (not intermediary)
 361
 362                         //build hermite curve, it's easier
 363                         curve.p1() = f[i0];
 364                         curve.p2() = f[i3];
 365                         curve.t1() = df[i0-ibase] * (df[i0-ibase].mag_squared() > 1e-4
 366                                                                                  ? j2->tangentscale/df[i0-ibase].mag()
 367                                                                                  : j2->tangentscale);
 368                         curve.t2() = df[i3-ibase] * (df[i3-ibase].mag_squared() > 1e-4
 369                                                                                  ? j->tangentscale/df[i3-ibase].mag()
 370                                                                                  : j->tangentscale);
 371                         curve.sync();
 372
 373                         //MUST include the end point (since we are ignoring left one)
 374                         for(; k < kend; ++k, t += dt)
 375                         {
 376                                 work[k] = curve(t);
 377                         }
 378
 379                         work[k] = curve(1); //k == kend, t == 1 -> c(t) == p2
 380                         ++ntess;
 381                 }
 382         }
 383
 384         return ntess;
 385 }
 386
 387 synfigapp::BLineConverter::BLineConverter()
 388 {
 389         pixelwidth = 1;
 390         smoothness = 0.70f;
 391         width = 0;
 392 };
 393
 394 void
 395 synfigapp::BLineConverter::clear()
 396 {
 397         f.clear();
 398         f_w.clear();
 399         ftemp.clear();
 400         df.clear();
 401         cvt.clear();
 402         brk.clear();
 403         di.clear();
 404         d_i.clear();
 405         work.clear();
 406         curind.clear();
 407 }
 408
 409 void
 410 synfigapp::BLineConverter::operator () (std::list<synfig::BLinePoint> &out, const std::list<synfig::Point> &in,const std::list<synfig::Real> &in_w)
 411 {
 412         //Profiling information
 413         /*etl::clock::value_type initialprocess=0, curveval=0, breakeval=0, disteval=0;
 414         etl::clock::value_type preproceval=0, tesseval=0, erroreval=0, spliteval=0;
 415         unsigned int                    numpre=0, numtess=0, numerror=0, numsplit=0;
 416         etl::clock_realtime timer,total;*/
 417
 418         //total.reset();
 419         if(in.size()<=1)
 420                 return;
 421
 422         clear();
 423
 424         //removing digitization error harder than expected
 425
 426         //intended to fix little pen errors caused by the way people draw (tiny juts in opposite direction)
 427         //Different solutions
 428         //      Average at both end points (will probably eliminate many points at each end of the samples)
 429         //      Average after the break points are found (weird points would still affect the curve)
 430         //      Just always get rid of breaks at the beginning and end if they are a certain distance apart
 431         //              This is will be current approach so all we do now is try to remove duplicate points
 432
 433         //remove duplicate points - very bad for fitting
 434
 435         //timer.reset();
 436
 437         {
 438                 std::list<synfig::Point>::const_iterator i = in.begin(), end = in.end();
 439                 std::list<synfig::Real>::const_iterator iw = in_w.begin();
 440                 synfig::Point   c;
 441
 442                 if(in.size() == in_w.size())
 443                 {
 444                         for(;i != end; ++i,++iw)
 445                         {
 446                                 //eliminate duplicate points
 447                                 if(*i != c)
 448                                 {
 449                                         f.push_back(c = *i);
 450                                         f_w.push_back(*iw);
 451                                 }
 452                         }
 453                 }else
 454                 {
 455                         for(;i != end; ++i)
 456                         {
 457                                 //eliminate duplicate points
 458                                 if(*i != c)
 459                                 {
 460                                         f.push_back(c = *i);
 461                                 }
 462                         }
 463                 }
 464         }
 465         //initialprocess = timer();
 466
 467         if(f.size()<=6)
 468                 return;
 469
 470         //get curvature information
 471         //timer.reset();
 472
 473         {
 474                 int i,i0,i1;
 475                 synfig::Vector v1,v2;
 476
 477                 cvt.resize(f.size());
 478
 479                 cvt.front() = 1;
 480                 cvt.back() = 1;
 481
 482                 for(i = 1; i < (int)f.size()-1; ++i)
 483                 {
 484                         i0 = std::max(0,i - 2);
 485                         i1 = std::min((int)(f.size()-1),i + 2);
 486
 487                         v1 = f[i] - f[i0];
 488                         v2 = f[i1] - f[i];
 489
 490                         cvt[i] = (v1*v2)/(v1.mag()*v2.mag());
 491                 }
 492         }
 493
 494         //curveval = timer();
 495         //synfig::info("calculated curvature");
 496
 497         //find corner points and interpolate inside those
 498         //timer.reset();
 499         {
 500                 //break at sharp derivative points
 501                 //TODO tolerance should be set based upon digitization resolution (length dependent index selection)
 502                 Real    tol = 0;                //break tolerance, for the cosine of the change in angle (really high curvature or something)
 503                 Real    fixdistsq = 4*width*width; //the distance to ignore breaks at the end points (for fixing stuff)
 504                 unsigned int i = 0;
 505
 506                 int             maxi = -1, last=0;
 507                 Real    minc = 1;
 508
 509                 brk.push_back(0);
 510
 511                 for(i = 1; i < cvt.size()-1; ++i)
 512                 {
 513                         //insert if too sharp (we need to break the tangents to insert onto the break list)
 514
 515                         if(cvt[i] < tol)
 516                         {
 517                                 if(cvt[i] < minc)
 518                                 {
 519                                         minc = cvt[i];
 520                                         maxi = i;
 521                                 }
 522                         }else if(maxi >= 0)
 523                         {
 524                                 if(maxi >= last + 8)
 525                                 {
 526                                         //synfig::info("break: %d-%d",maxi+1,cvt.size());
 527                                         brk.push_back(maxi);
 528                                         last = maxi;
 529                                 }
 530                                 maxi = -1;
 531                                 minc = 1;
 532                         }
 533                 }
 534
 535                 brk.push_back(i);
 536
 537                 //postprocess for breaks too close to each other
 538                 Real d = 0;
 539                 Point p = f[brk.front()];
 540
 541                 //first set
 542                 for(i = 1; i < brk.size()-1; ++i) //do not want to include end point...
 543                 {
 544                         d = (f[brk[i]] - p).mag_squared();
 545                         if(d > fixdistsq) break; //don't want to group breaks if we ever get over the dist...
 546                 }
 547                 //want to erase all points before...
 548                 if(i != 1)
 549                         brk.erase(brk.begin(),brk.begin()+i-1);
 550
 551                 //end set
 552                 p = f[brk.back()];
 553                 for(i = brk.size()-2; i > 0; --i) //start at one in from the end
 554                 {
 555                         d = (f[brk[i]] - p).mag_squared();
 556                         if(d > fixdistsq) break; //don't want to group breaks if we ever get over the dist
 557                 }
 558                 if(i != brk.size()-2)
 559                         brk.erase(brk.begin()+i+2,brk.end()); //erase all points that we found... found none if i has not advanced
 560                 //must not include the one we ended up on
 561         }
 562         //breakeval = timer();
 563         //synfig::info("found break points: %d",brk.size());
 564
 565         //get the distance calculation of the entire curve (for tangent scaling)
 566
 567         //timer.reset();
 568         {
 569                 synfig::Point p1,p2;
 570
 571                 p1=p2=f[0];
 572
 573                 di.resize(f.size()); d_i.resize(f.size());
 574                 Real d = 0;
 575                 for(unsigned int i = 0; i < f.size();)
 576                 {
 577                         d += (d_i[i] = (p2-p1).mag());
 578                         di[i] = d;
 579
 580                         p1=p2;
 581                         p2=f[++i];
 582                 }
 583         }
 584         //disteval = timer();
 585         //synfig::info("calculated distance");
 586
 587         //now break at every point - calculate new derivatives each time
 588
 589         //TODO
 590         //must be sure that the break points are 3 or more apart
 591         //then must also store the breaks which are not smooth, etc.
 592         //and figure out tangents between there
 593
 594         //for each pair of break points (as long as they are far enough apart) recursively subdivide stuff
 595         //ignore the detected intermediate points
 596         {
 597                 unsigned int i0=0,i3=0,is=0;
 598                 int i=0,j=0;
 599
 600                 bool done = false;
 601
 602                 Real errortol = smoothness*pixelwidth; //???? what the hell should this value be
 603
 604                 BLinePoint a;
 605                 synfig::Vector v;
 606
 607                 //intemp = f; //don't want to smooth out the corners
 608
 609                 bool breaktan = false, setwidth;
 610                 a.set_split_tangent_flag(false);
 611                 //a.set_width(width);
 612                 a.set_width(1.0f);
 613
 614                 setwidth = (f.size() == f_w.size());
 615
 616                 for(j = 0; j < (int)brk.size() - 1; ++j)
 617                 {
 618                         //for b[j] to b[j+1] subdivide and stuff
 619                         i0 = brk[j];
 620                         i3 = brk[j+1];
 621
 622                         unsigned int size = i3-i0+1; //must include the end points
 623
 624                         //new derivatives
 625                         //timer.reset();
 626                         ftemp.assign(f.begin()+i0, f.begin()+i3+1);
 627                         for(i=0;i<20;++i)
 628                                 gaussian_blur_3(ftemp.begin(),ftemp.end(),false);
 629
 630                         df.resize(size);
 631                         GetFirstDerivatives(ftemp,0,size,(char*)&df[0],sizeof(df[0]));
 632                         //GetSimpleDerivatives(ftemp,0,size,df,0,di);
 633                         //< don't have to worry about indexing stuff as it is all being taken car of right now
 634                         //preproceval += timer();
 635                         //numpre++;
 636
 637                         work.resize(size*2-1); //guarantee that all points will be tessellated correctly (one point in between every 2 adjacent points)
 638
 639                         //if size of work is size*2-1, the step size should be 1/(size*2 - 2)
 640                         //Real step = 1/(Real)(size*2 - 1);
 641
 642                         //start off with break points as indices
 643                         curind.clear();
 644                         curind.push_back(cpindex(i0,di[i3]-di[i0],0)); //0 error because no curve on the left
 645                         curind.push_back(cpindex(i3,di[i3]-di[i0],-1)); //error needs to be reevaluated
 646                         done = false; //we want to loop
 647
 648                         unsigned int dcount = 0;
 649
 650                         //while there are still enough points between us, and the error is too high subdivide (and invalidate neighbors that share tangents)
 651                         while(!done)
 652                         {
 653                                 //tessellate all curves with invalid error values
 654                                 work[0] = f[i0];
 655
 656                                 //timer.reset();
 657                                 /*numtess += */tessellate_curves(curind,f,df,work);
 658                                 //tesseval += timer();
 659
 660                                 //now get all error values
 661                                 //timer.reset();
 662                                 for(i = 1; i < (int)curind.size(); ++i)
 663                                 {
 664                                         if(curind[i].error < 0) //must have been retessellated, so now recalculate error value
 665                                         {
 666                                                 //evaluate error from points (starting at current index)
 667                                                 int size = curind[i].curind - curind[i-1].curind + 1;
 668                                                 curind[i].error = CurveError(&f[curind[i-1].curind], size,
 669                                                                                                          work,(curind[i-1].curind - i0)*2,(curind[i].curind - i0)*2+1);
 670
 671                                                 /*if(curind[i].error > 1.0e5)
 672                                                 {
 673                                                         synfig::info("Holy crap %d-%d error %f",curind[i-1].curind,curind[i].curind,curind[i].error);
 674                                                         curind[i].error = -1;
 675                                                         numtess += tessellate_curves(curind,f,df,work);
 676                                                         curind[i].error = CurveError(&f[curind[i-1].curind], size,
 677                                                                                                          work,0,work.size());//(curind[i-1].curind - i0)*2,(curind[i].curind - i0)*2+1);
 678                                                 }*/
 679                                                 //numerror++;
 680                                         }
 681                                 }
 682                                 //erroreval += timer();
 683
 684                                 //assume we're done
 685                                 done = true;
 686
 687                                 //check each error to see if it's too big, if so, then subdivide etc.
 688                                 int indsize = (int)curind.size();
 689                                 Real maxrelerror = 0;
 690                                 int maxi = -1;//, numpoints;
 691
 692                                 //timer.reset();
 693                                 //get the maximum error and split there
 694                                 for(i = 1; i < indsize; ++i)
 695                                 {
 696                                         //numpoints = curind[i].curind - curind[i-1].curind + 1;
 697
 698                                         if(curind[i].error > maxrelerror && curind[i].curind - curind[i-1].curind > 2) //only accept if it's valid
 699                                         {
 700                                                 maxrelerror = curind[i].error;
 701                                                 maxi = i;
 702                                         }
 703                                 }
 704
 705                                 //split if error is too great
 706                                 if(maxrelerror > errortol)
 707                                 {
 708                                         //add one to the left etc
 709                                         unsigned int    ibase = curind[maxi-1].curind, itop = curind[maxi].curind,
 710                                                                         ibreak = (ibase + itop)/2;
 711                                         Real scale, scale2;
 712
 713                                         assert(ibreak < f.size());
 714
 715                                         //synfig::info("Split %d -%d- %d, error: %f", ibase,ibreak,itop,maxrelerror);
 716
 717                                         if(ibase != itop)
 718                                         {
 719                                                 //invalidate current error of the changed tangents and add an extra segment
 720                                                 //enforce minimum tangents property
 721                                                 curind[maxi].error = -1;
 722                                                 curind[maxi-1].error = -1;
 723                                                 if(maxi+1 < indsize) curind[maxi+1].error = -1; //if there is a curve segment beyond this it will be effected as well
 724
 725                                                 scale = di[itop] - di[ibreak];
 726                                                 scale2 = maxi+1 < indsize ? di[curind[maxi+1].curind] - di[itop] : scale; //to the right valid?
 727                                                 curind[maxi].tangentscale = std::min(scale, scale2);
 728
 729                                                 scale = di[ibreak] - di[ibase];
 730                                                 scale2 = maxi >= 2 ? di[ibase] - di[curind[maxi-2].curind] : scale; // to the left valid -2 ?
 731                                                 curind[maxi-1].tangentscale = std::min(scale, scale2);
 732
 733                                                 scale = std::min(di[ibreak] - di[ibase], di[itop] - di[ibreak]);
 734
 735                                                 curind.insert(curind.begin()+maxi,cpindex(ibreak, scale, -1));
 736                                                 //curind.push_back(cpindex(ibreak, scale, -1));
 737                                                 //std::sort(curind.begin(), curind.end());
 738
 739                                                 done = false;
 740                                                 //numsplit++;
 741                                         }
 742                                 }
 743                                 //spliteval += timer();
 744
 745                                 dcount++;
 746                         }
 747
 748                         //insert the last point too (just set tangent for now
 749                         is = curind[0].curind;
 750
 751                         //first point inherits current tangent status
 752                         v = df[is - i0];
 753                         if(v.mag_squared() > EPSILON)
 754                                 v *= (curind[0].tangentscale/v.mag());
 755
 756                         if(!breaktan)
 757                                 a.set_tangent(v);
 758                         else a.set_tangent2(v);
 759
 760                         a.set_vertex(f[is]);
 761                         if(setwidth)a.set_width(f_w[is]);
 762
 763                         out.push_back(a);
 764                         a.set_split_tangent_flag(false); //won't need to break anymore
 765                         breaktan = false;
 766
 767                         for(i = 1; i < (int)curind.size()-1; ++i)
 768                         {
 769                                 is = curind[i].curind;
 770
 771                                 //first point inherits current tangent status
 772                                 v = df[is-i0];
 773                                 if(v.mag_squared() > EPSILON)
 774                                         v *= (curind[i].tangentscale/v.mag());
 775
 776                                 a.set_tangent(v); // always inside, so guaranteed to be smooth
 777                                 a.set_vertex(f[is]);
 778                                 if(setwidth)a.set_width(f_w[is]);
 779
 780                                 out.push_back(a);
 781                         }
 782
 783                         //set the last point's data
 784                         is = curind.back().curind; //should already be this
 785
 786                         v = df[is-i0];
 787                         if(v.mag_squared() > EPSILON)
 788                                 v *= (curind.back().tangentscale/v.mag());
 789
 790                         a.set_tangent1(v);
 791                         a.set_split_tangent_flag(true);
 792                         breaktan = true;
 793
 794                         //will get the vertex and tangent 2 from next round
 795                 }
 796
 797                 a.set_vertex(f[i3]);
 798                 a.set_split_tangent_flag(false);
 799                 if(setwidth)
 800                         a.set_width(f_w[i3]);
 801                 out.push_back(a);
 802
 803                 /*etl::clock::value_type totaltime = total(),
 804                                                            misctime = totaltime - initialprocess - curveval - breakeval - disteval
 805                                                                           - preproceval - tesseval - erroreval - spliteval;
 806
 807                 synfig::info(
 808                         "Curve Convert Profile:\n"
 809                         "\tInitial Preprocess:    %f\n"
 810                         "\tCurvature Calculation: %f\n"
 811                         "\tBreak Calculation:     %f\n"
 812                         "\tDistance Calculation:  %f\n"
 813                         "  Algorithm: (numtimes,totaltime)\n"
 814                         "\tPreprocess step:      (%d,%f)\n"
 815                         "\tTessellation step:    (%d,%f)\n"
 816                         "\tError step:           (%d,%f)\n"
 817                         "\tSplit step:           (%d,%f)\n"
 818                         "  Num Input: %d, Num Output: %d\n"
 819                         "  Total time: %f, Misc time: %f\n",
 820                         initialprocess, curveval,breakeval,disteval,
 821                         numpre,preproceval,numtess,tesseval,numerror,erroreval,numsplit,spliteval,
 822                         in.size(),out.size(),
 823                         totaltime,misctime);*/
 824
 825                 return;
 826         }
 827 }
 828
 829 void synfigapp::BLineConverter::EnforceMinWidth(std::list<synfig::BLinePoint> &bline, synfig::Real min_pressure)
 830 {
 831         std::list<synfig::BLinePoint>::iterator i = bline.begin(),
 832                                                                                         end = bline.end();
 833
 834         for(i = bline.begin(); i != end; ++i)
 835         {
 836                 if(i->get_width() < min_pressure)
 837                 {
 838                         i->set_width(min_pressure);
 839                 }
 840         }
 841 }