This file is indexed.

/usr/include/opencascade/BSplCLib.hxx is in libopencascade-foundation-dev 6.5.0.dfsg-2build1.

This file is owned by root:root, with mode 0o644.

The actual contents of the file can be viewed below.

   1
   2
   3
   4
   5
   6
   7
   8
   9
  10
  11
  12
  13
  14
  15
  16
  17
  18
  19
  20
  21
  22
  23
  24
  25
  26
  27
  28
  29
  30
  31
  32
  33
  34
  35
  36
  37
  38
  39
  40
  41
  42
  43
  44
  45
  46
  47
  48
  49
  50
  51
  52
  53
  54
  55
  56
  57
  58
  59
  60
  61
  62
  63
  64
  65
  66
  67
  68
  69
  70
  71
  72
  73
  74
  75
  76
  77
  78
  79
  80
  81
  82
  83
  84
  85
  86
  87
  88
  89
  90
  91
  92
  93
  94
  95
  96
  97
  98
  99
 100
 101
 102
 103
 104
 105
 106
 107
 108
 109
 110
 111
 112
 113
 114
 115
 116
 117
 118
 119
 120
 121
 122
 123
 124
 125
 126
 127
 128
 129
 130
 131
 132
 133
 134
 135
 136
 137
 138
 139
 140
 141
 142
 143
 144
 145
 146
 147
 148
 149
 150
 151
 152
 153
 154
 155
 156
 157
 158
 159
 160
 161
 162
 163
 164
 165
 166
 167
 168
 169
 170
 171
 172
 173
 174
 175
 176
 177
 178
 179
 180
 181
 182
 183
 184
 185
 186
 187
 188
 189
 190
 191
 192
 193
 194
 195
 196
 197
 198
 199
 200
 201
 202
 203
 204
 205
 206
 207
 208
 209
 210
 211
 212
 213
 214
 215
 216
 217
 218
 219
 220
 221
 222
 223
 224
 225
 226
 227
 228
 229
 230
 231
 232
 233
 234
 235
 236
 237
 238
 239
 240
 241
 242
 243
 244
 245
 246
 247
 248
 249
 250
 251
 252
 253
 254
 255
 256
 257
 258
 259
 260
 261
 262
 263
 264
 265
 266
 267
 268
 269
 270
 271
 272
 273
 274
 275
 276
 277
 278
 279
 280
 281
 282
 283
 284
 285
 286
 287
 288
 289
 290
 291
 292
 293
 294
 295
 296
 297
 298
 299
 300
 301
 302
 303
 304
 305
 306
 307
 308
 309
 310
 311
 312
 313
 314
 315
 316
 317
 318
 319
 320
 321
 322
 323
 324
 325
 326
 327
 328
 329
 330
 331
 332
 333
 334
 335
 336
 337
 338
 339
 340
 341
 342
 343
 344
 345
 346
 347
 348
 349
 350
 351
 352
 353
 354
 355
 356
 357
 358
 359
 360
 361
 362
 363
 364
 365
 366
 367
 368
 369
 370
 371
 372
 373
 374
 375
 376
 377
 378
 379
 380
 381
 382
 383
 384
 385
 386
 387
 388
 389
 390
 391
 392
 393
 394
 395
 396
 397
 398
 399
 400
 401
 402
 403
 404
 405
 406
 407
 408
 409
 410
 411
 412
 413
 414
 415
 416
 417
 418
 419
 420
 421
 422
 423
 424
 425
 426
 427
 428
 429
 430
 431
 432
 433
 434
 435
 436
 437
 438
 439
 440
 441
 442
 443
 444
 445
 446
 447
 448
 449
 450
 451
 452
 453
 454
 455
 456
 457
 458
 459
 460
 461
 462
 463
 464
 465
 466
 467
 468
 469
 470
 471
 472
 473
 474
 475
 476
 477
 478
 479
 480
 481
 482
 483
 484
 485
 486
 487
 488
 489
 490
 491
 492
 493
 494
 495
 496
 497
 498
 499
 500
 501
 502
 503
 504
 505
 506
 507
 508
 509
 510
 511
 512
 513
 514
 515
 516
 517
 518
 519
 520
 521
 522
 523
 524
 525
 526
 527
 528
 529
 530
 531
 532
 533
 534
 535
 536
 537
 538
 539
 540
 541
 542
 543
 544
 545
 546
 547
 548
 549
 550
 551
 552
 553
 554
 555
 556
 557
 558
 559
 560
 561
 562
 563
 564
 565
 566
 567
 568
 569
 570
 571
 572
 573
 574
 575
 576
 577
 578
 579
 580
 581
 582
 583
 584
 585
 586
 587
 588
 589
 590
 591
 592
 593
 594
 595
 596
 597
 598
 599
 600
 601
 602
 603
 604
 605
 606
 607
 608
 609
 610
 611
 612
 613
 614
 615
 616
 617
 618
 619
 620
 621
 622
 623
 624
 625
 626
 627
 628
 629
 630
 631
 632
 633
 634
 635
 636
 637
 638
 639
 640
 641
 642
 643
 644
 645
 646
 647
 648
 649
 650
 651
 652
 653
 654
 655
 656
 657
 658
 659
 660
 661
 662
 663
 664
 665
 666
 667
 668
 669
 670
 671
 672
 673
 674
 675
 676
 677
 678
 679
 680
 681
 682
 683
 684
 685
 686
 687
 688
 689
 690
 691
 692
 693
 694
 695
 696
 697
 698
 699
 700
 701
 702
 703
 704
 705
 706
 707
 708
 709
 710
 711
 712
 713
 714
 715
 716
 717
 718
 719
 720
 721
 722
 723
 724
 725
 726
 727
 728
 729
 730
 731
 732
 733
 734
 735
 736
 737
 738
 739
 740
 741
 742
 743
 744
 745
 746
 747
 748
 749
 750
 751
 752
 753
 754
 755
 756
 757
 758
 759
 760
 761
 762
 763
 764
 765
 766
 767
 768
 769
 770
 771
 772
 773
 774
 775
 776
 777
 778
 779
 780
 781
 782
 783
 784
 785
 786
 787
 788
 789
 790
 791
 792
 793
 794
 795
 796
 797
 798
 799
 800
 801
 802
 803
 804
 805
 806
 807
 808
 809
 810
 811
 812
 813
 814
 815
 816
 817
 818
 819
 820
 821
 822
 823
 824
 825
 826
 827
 828
 829
 830
 831
 832
 833
 834
 835
 836
 837
 838
 839
 840
 841
 842
 843
 844
 845
 846
 847
 848
 849
 850
 851
 852
 853
 854
 855
 856
 857
 858
 859
 860
 861
 862
 863
 864
 865
 866
 867
 868
 869
 870
 871
 872
 873
 874
 875
 876
 877
 878
 879
 880
 881
 882
 883
 884
 885
 886
 887
 888
 889
 890
 891
 892
 893
 894
 895
 896
 897
 898
 899
 900
 901
 902
 903
 904
 905
 906
 907
 908
 909
 910
 911
 912
 913
 914
 915
 916
 917
 918
 919
 920
 921
 922
 923
 924
 925
 926
 927
 928
 929
 930
 931
 932
 933
 934
 935
 936
 937
 938
 939
 940
 941
 942
 943
 944
 945
 946
 947
 948
 949
 950
 951
 952
 953
 954
 955
 956
 957
 958
 959
 960
 961
 962
 963
 964
 965
 966
 967
 968
 969
 970
 971
 972
 973
 974
 975
 976
 977
 978
 979
 980
 981
 982
 983
 984
 985
 986
 987
 988
 989
 990
 991
 992
 993
 994
 995
 996
 997
 998
 999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
// This file is generated by WOK (CPPExt).
// Please do not edit this file; modify original file instead.
// The copyright and license terms as defined for the original file apply to 
// this header file considered to be the "object code" form of the original source.

#ifndef _BSplCLib_HeaderFile
#define _BSplCLib_HeaderFile

#ifndef _Standard_HeaderFile
#include <Standard.hxx>
#endif
#ifndef _Standard_Macro_HeaderFile
#include <Standard_Macro.hxx>
#endif

#ifndef _Standard_Real_HeaderFile
#include <Standard_Real.hxx>
#endif
#ifndef _Standard_Integer_HeaderFile
#include <Standard_Integer.hxx>
#endif
#ifndef _Standard_Boolean_HeaderFile
#include <Standard_Boolean.hxx>
#endif
#ifndef _BSplCLib_KnotDistribution_HeaderFile
#include <BSplCLib_KnotDistribution.hxx>
#endif
#ifndef _BSplCLib_MultDistribution_HeaderFile
#include <BSplCLib_MultDistribution.hxx>
#endif
#ifndef _Handle_TColStd_HArray1OfReal_HeaderFile
#include <Handle_TColStd_HArray1OfReal.hxx>
#endif
#ifndef _Handle_TColStd_HArray1OfInteger_HeaderFile
#include <Handle_TColStd_HArray1OfInteger.hxx>
#endif
#ifndef _BSplCLib_EvaluatorFunction_HeaderFile
#include <BSplCLib_EvaluatorFunction.hxx>
#endif
class TColStd_Array1OfReal;
class TColStd_Array1OfInteger;
class TColgp_Array1OfPnt;
class TColgp_Array1OfPnt2d;
class gp_Pnt;
class gp_Pnt2d;
class gp_Vec;
class gp_Vec2d;
class math_Matrix;
class TColStd_HArray1OfReal;
class TColStd_HArray1OfInteger;


//!  BSplCLib   B-spline curve Library. <br>
//! <br>
//!  The BSplCLib package is  a basic library  for BSplines. It <br>
//!  provides three categories of functions. <br>
//! <br>
//!  * Management methods to  process knots and multiplicities. <br>
//! <br>
//!  * Multi-Dimensions  spline methods.  BSpline methods where <br>
//!  poles have an arbitrary number of dimensions. They divides <br>
//!  in two groups : <br>
//! <br>
//!    - Global methods modifying the  whole set of  poles. The <br>
//!    poles are    described   by an array   of   Reals and  a <br>
//!    Dimension. Example : Inserting knots. <br>
//! <br>
//!    - Local methods  computing  points and derivatives.  The <br>
//!    poles  are described by a pointer  on  a local array  of <br>
//!    Reals and a Dimension. The local array is modified. <br>
//! <br>
//!  *  2D  and 3D spline   curves  methods. <br>
//! <br>
//!    Methods  for 2d and 3d BSplines  curves  rational or not <br>
//!    rational. <br>
//! <br>
//!    Those methods have the following structure : <br>
//! <br>
//!     - They extract the pole informations in a working array. <br>
//! <br>
//!     -  They      process the  working   array    with   the <br>
//!     multi-dimension  methods. (for example  a  3d  rational <br>
//!     curve is processed as a 4 dimension curve). <br>
//! <br>
//!     - They get back the result in the original dimension. <br>
//! <br>
//!     Note that the  bspline   surface methods found   in the <br>
//!     package BSplSLib  uses  the same  structure and rely on <br>
//!     BSplCLib. <br>
//! <br>
//!     In the following list  of methods the  2d and 3d  curve <br>
//!     methods   will be  described   with  the  corresponding <br>
//!     multi-dimension method. <br>
//! <br>
//!  The 3d or 2d B-spline curve is defined with : <br>
//! <br>
//!  . its control points : TColgp_Array1OfPnt(2d)        Poles <br>
//!  . its weights        : TColStd_Array1OfReal          Weights <br>
//!  . its knots          : TColStd_Array1OfReal          Knots <br>
//!  . its multiplicities : TColStd_Array1OfInteger       Mults <br>
//!  . its degree         : Standard_Integer              Degree <br>
//!  . its periodicity    : Standard_Boolean              Periodic <br>
class BSplCLib  {
public:

  void* operator new(size_t,void* anAddress) 
  {
    return anAddress;
  }
  void* operator new(size_t size) 
  {
    return Standard::Allocate(size); 
  }
  void  operator delete(void *anAddress) 
  {
    if (anAddress) Standard::Free((Standard_Address&)anAddress); 
  }

  //!  This routine searches the position of the real <br>
//!         value X in  the ordered set of  real  values XX. <br>
//! <br>
//!         The  elements   in   the  table    XX  are   either <br>
//!         monotonically    increasing     or    monotonically <br>
//!         decreasing. <br>
//! <br>
//!         The input   value Iloc is    used to initialize the <br>
//!         algorithm  :  if  Iloc  is outside  of   the bounds <br>
//!         [XX.Lower(), -- XX.Upper()] the bisection algorithm <br>
//!         is used else  the routine searches from  a previous <br>
//!         known position  by increasing steps  then converges <br>
//!         by bisection. <br>
//! <br>
//!         This  routine is used to  locate a  knot value in a <br>
//!         set of knots. <br>
//! <br>
  Standard_EXPORT   static  void Hunt(const TColStd_Array1OfReal& XX,const Standard_Real X,Standard_Integer& Iloc) ;
  //! Computes the index of the knots value which gives <br>
//!         the start point of the curve. <br>
  Standard_EXPORT   static  Standard_Integer FirstUKnotIndex(const Standard_Integer Degree,const TColStd_Array1OfInteger& Mults) ;
  //! Computes the index of the knots value which gives <br>
//!         the end point of the curve. <br>
  Standard_EXPORT   static  Standard_Integer LastUKnotIndex(const Standard_Integer Degree,const TColStd_Array1OfInteger& Mults) ;
  //! Computes the index  of  the  flats knots  sequence <br>
//!          corresponding  to  <Index> in  the  knots sequence <br>
//!          which multiplicities are <Mults>. <br>
  Standard_EXPORT   static  Standard_Integer FlatIndex(const Standard_Integer Degree,const Standard_Integer Index,const TColStd_Array1OfInteger& Mults,const Standard_Boolean Periodic) ;
  //!  Locates  the parametric value    U  in the knots <br>
//!         sequence  between  the  knot K1   and the knot  K2. <br>
//!         The value return in Index verifies. <br>
//! <br>
//!          Knots(Index) <= U < Knots(Index + 1) <br>
//!          if U <= Knots (K1) then Index = K1 <br>
//!          if U >= Knots (K2) then Index = K2 - 1 <br>
//! <br>
//!          If Periodic is True U  may be  modified  to fit in <br>
//!          the range  Knots(K1), Knots(K2).  In any case  the <br>
//!          correct value is returned in NewU. <br>
//! <br>
//!  Warnings :Index is used  as input   data to initialize  the <br>
//!          searching  function. <br>
//!  Warning: Knots have to be "withe repetitions" <br>
  Standard_EXPORT   static  void LocateParameter(const Standard_Integer Degree,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,const Standard_Real U,const Standard_Boolean IsPeriodic,const Standard_Integer FromK1,const Standard_Integer ToK2,Standard_Integer& KnotIndex,Standard_Real& NewU) ;
  //!  Locates  the parametric value    U  in the knots <br>
//!         sequence  between  the  knot K1   and the knot  K2. <br>
//!         The value return in Index verifies. <br>
//! <br>
//!          Knots(Index) <= U < Knots(Index + 1) <br>
//!          if U <= Knots (K1) then Index = K1 <br>
//!          if U >= Knots (K2) then Index = K2 - 1 <br>
//! <br>
//!          If Periodic is True U  may be  modified  to fit in <br>
//!          the range  Knots(K1), Knots(K2).  In any case  the <br>
//!          correct value is returned in NewU. <br>
//! <br>
//!  Warnings :Index is used  as input   data to initialize  the <br>
//!          searching  function. <br>
//!  Warning: Knots have to be "flat" <br>
  Standard_EXPORT   static  void LocateParameter(const Standard_Integer Degree,const TColStd_Array1OfReal& Knots,const Standard_Real U,const Standard_Boolean IsPeriodic,const Standard_Integer FromK1,const Standard_Integer ToK2,Standard_Integer& KnotIndex,Standard_Real& NewU) ;
  
  Standard_EXPORT   static  void LocateParameter(const Standard_Integer Degree,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,const Standard_Real U,const Standard_Boolean Periodic,Standard_Integer& Index,Standard_Real& NewU) ;
  //! Finds the greatest multiplicity in a set of knots <br>
//!         between  K1  and K2.   Mults  is  the  multiplicity <br>
//!         associated with each knot value. <br>
  Standard_EXPORT   static  Standard_Integer MaxKnotMult(const TColStd_Array1OfInteger& Mults,const Standard_Integer K1,const Standard_Integer K2) ;
  //! Finds the lowest multiplicity in  a  set of knots <br>
//!         between   K1  and K2.   Mults is  the  multiplicity <br>
//!         associated with each knot value. <br>
  Standard_EXPORT   static  Standard_Integer MinKnotMult(const TColStd_Array1OfInteger& Mults,const Standard_Integer K1,const Standard_Integer K2) ;
  //! Returns the number of poles of the curve. Returns 0 if <br>
//!          one of the multiplicities is incorrect. <br>
//! <br>
//!          * Non positive. <br>
//! <br>
//!          * Greater than Degree,  or  Degree+1  at the first and <br>
//!          last knot of a non periodic curve. <br>
//! <br>
//!          *  The  last periodicity  on  a periodic  curve is not <br>
//!          equal to the first. <br>
  Standard_EXPORT   static  Standard_Integer NbPoles(const Standard_Integer Degree,const Standard_Boolean Periodic,const TColStd_Array1OfInteger& Mults) ;
  //! Returns the length  of the sequence  of knots with <br>
//!          repetition. <br>
//! <br>
//!          Periodic : <br>
//! <br>
//!          Sum(Mults(i), i = Mults.Lower(); i <= Mults.Upper()); <br>
//! <br>
//!          Non Periodic : <br>
//! <br>
//!          Sum(Mults(i); i = Mults.Lower(); i < Mults.Upper()) <br>
//!          + 2 * Degree <br>
  Standard_EXPORT   static  Standard_Integer KnotSequenceLength(const TColStd_Array1OfInteger& Mults,const Standard_Integer Degree,const Standard_Boolean Periodic) ;
  
  Standard_EXPORT   static  void KnotSequence(const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,TColStd_Array1OfReal& KnotSeq) ;
  //! Computes  the  sequence   of knots KnotSeq  with <br>
//!         repetition  of the  knots  of multiplicity  greater <br>
//!         than 1. <br>
//! <br>
//!  Length of KnotSeq must be KnotSequenceLength(Mults,Degree,Periodic) <br>
  Standard_EXPORT   static  void KnotSequence(const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,const Standard_Integer Degree,const Standard_Boolean Periodic,TColStd_Array1OfReal& KnotSeq) ;
  //! Returns the  length  of the   sequence of  knots  (and <br>
//!          Mults)  without repetition. <br>
  Standard_EXPORT   static  Standard_Integer KnotsLength(const TColStd_Array1OfReal& KnotSeq,const Standard_Boolean Periodic = Standard_False) ;
  //!  Computes  the  sequence   of knots Knots  without <br>
//!         repetition  of the  knots  of multiplicity  greater <br>
//!         than 1. <br>
//! <br>
//!       Length  of <Knots> and  <Mults> must be <br>
//!    KnotsLength(KnotSequence,Periodic) <br>
  Standard_EXPORT   static  void Knots(const TColStd_Array1OfReal& KnotSeq,TColStd_Array1OfReal& Knots,TColStd_Array1OfInteger& Mults,const Standard_Boolean Periodic = Standard_False) ;
  //! Analyses if the  knots distribution is "Uniform" <br>
//!         or  "NonUniform" between  the  knot  FromK1 and the <br>
//!         knot ToK2.  There is  no repetition of  knot in the <br>
//!         knots'sequence <Knots>. <br>
  Standard_EXPORT   static  BSplCLib_KnotDistribution KnotForm(const TColStd_Array1OfReal& Knots,const Standard_Integer FromK1,const Standard_Integer ToK2) ;
  
//!  Analyses the distribution of multiplicities between <br>
//!  the knot FromK1 and the Knot ToK2. <br>
  Standard_EXPORT   static  BSplCLib_MultDistribution MultForm(const TColStd_Array1OfInteger& Mults,const Standard_Integer FromK1,const Standard_Integer ToK2) ;
  
//!  Reparametrizes a B-spline curve to [U1, U2]. <br>
//!  The knot values are recomputed such that Knots (Lower) = U1 <br>
//!  and Knots (Upper) = U2   but the knot form is not modified. <br>
//!   Warnings : <br>
//!  In the array Knots the values must be in ascending order. <br>
//!  U1 must not be equal to U2 to avoid division by zero. <br>
  Standard_EXPORT   static  void Reparametrize(const Standard_Real U1,const Standard_Real U2,TColStd_Array1OfReal& Knots) ;
  //! Reverses  the  array   knots  to  become  the knots <br>
//!          sequence of the reversed curve. <br>
  Standard_EXPORT   static  void Reverse(TColStd_Array1OfReal& Knots) ;
  //! Reverses  the  array of multiplicities. <br>
  Standard_EXPORT   static  void Reverse(TColStd_Array1OfInteger& Mults) ;
  //! Reverses the array of poles. Last is the  index of <br>
//!          the new first pole. On  a  non periodic curve last <br>
//!          is Poles.Upper(). On a periodic curve last is <br>
//! <br>
//!          (number of flat knots - degree - 1) <br>
//! <br>
//!          or <br>
//! <br>
//!          (sum of multiplicities(but  for the last) + degree <br>
//!          - 1) <br>
  Standard_EXPORT   static  void Reverse(TColgp_Array1OfPnt& Poles,const Standard_Integer Last) ;
  //! Reverses the array of poles. <br>
  Standard_EXPORT   static  void Reverse(TColgp_Array1OfPnt2d& Poles,const Standard_Integer Last) ;
  //! Reverses the array of poles. <br>
  Standard_EXPORT   static  void Reverse(TColStd_Array1OfReal& Weights,const Standard_Integer Last) ;
  
//!   Returns False if all the weights  of the  array <Weights> <br>
//!   between   I1 an I2   are  identic.   Epsilon  is used for <br>
//!   comparing  weights. If Epsilon  is 0. the  Epsilon of the <br>
//!   first weight is used. <br>
  Standard_EXPORT   static  Standard_Boolean IsRational(const TColStd_Array1OfReal& Weights,const Standard_Integer I1,const Standard_Integer I2,const Standard_Real Epsilon = 0.0) ;
  //! returns the degree maxima for a BSplineCurve. <br>
      static  Standard_Integer MaxDegree() ;
  //! Perform the Boor  algorithm  to  evaluate a point at <br>
//!          parameter <U>, with <Degree> and <Dimension>. <br>
//! <br>
//!          Poles is  an array of  Reals of size <br>
//! <br>
//!          <Dimension> *  <Degree>+1 <br>
//! <br>
//!          Containing  the poles.  At  the end <Poles> contains <br>
//!          the current point. <br>
  Standard_EXPORT   static  void Eval(const Standard_Real U,const Standard_Integer Degree,Standard_Real& Knots,const Standard_Integer Dimension,Standard_Real& Poles) ;
  //! Performs the  Boor Algorithm  at  parameter <U> with <br>
//!          the given <Degree> and the  array of <Knots> on  the <br>
//!          poles <Poles> of dimension  <Dimension>.  The schema <br>
//!          is  computed  until  level  <Depth>  on a   basis of <br>
//!          <Length+1> poles. <br>
//! <br>
//!          * Knots is an array of reals of length : <br>
//! <br>
//!            <Length> + <Degree> <br>
//! <br>
//!          * Poles is an array of reals of length : <br>
//! <br>
//!           (2 * <Length> + 1) * <Dimension> <br>
//! <br>
//!           The poles values  must be  set  in the array at the <br>
//!           positions. <br>
//! <br>
//!            0..Dimension, <br>
//! <br>
//!            2 * Dimension .. <br>
//!            3 * Dimension <br>
//! <br>
//!            4  * Dimension .. <br>
//!            5  * Dimension <br>
//! <br>
//!            ... <br>
//! <br>
//!            The results are found in the array poles depending <br>
//!            on the Depth. (See the method GetPole). <br>
  Standard_EXPORT   static  void BoorScheme(const Standard_Real U,const Standard_Integer Degree,Standard_Real& Knots,const Standard_Integer Dimension,Standard_Real& Poles,const Standard_Integer Depth,const Standard_Integer Length) ;
  //! Compute  the content of  Pole before the BoorScheme. <br>
//!          This method is used to remove poles. <br>
//! <br>
//!          U is the poles to  remove, Knots should contains the <br>
//!          knots of the curve after knot removal. <br>
//! <br>
//!          The first  and last poles  do not  change, the other <br>
//!          poles are computed by averaging two possible values. <br>
//!          The distance between  the  two   possible  poles  is <br>
//!          computed, if it  is higher than <Tolerance> False is <br>
//!          returned. <br>
  Standard_EXPORT   static  Standard_Boolean AntiBoorScheme(const Standard_Real U,const Standard_Integer Degree,Standard_Real& Knots,const Standard_Integer Dimension,Standard_Real& Poles,const Standard_Integer Depth,const Standard_Integer Length,const Standard_Real Tolerance) ;
  //! Computes   the   poles of  the    BSpline  giving the <br>
//!          derivatives of order <Order>. <br>
//! <br>
//!          The formula for the first order is <br>
//! <br>
//!          Pole(i) = Degree * (Pole(i+1) - Pole(i)) / <br>
//!                             (Knots(i+Degree+1) - Knots(i+1)) <br>
//! <br>
//!          This formula  is repeated  (Degree  is decremented at <br>
//!          each step). <br>
  Standard_EXPORT   static  void Derivative(const Standard_Integer Degree,Standard_Real& Knots,const Standard_Integer Dimension,const Standard_Integer Length,const Standard_Integer Order,Standard_Real& Poles) ;
  //! Performs the Bohm  Algorithm at  parameter <U>. This <br>
//!          algorithm computes the value and all the derivatives <br>
//!          up to order N (N <= Degree). <br>
//! <br>
//!          <Poles> is the original array of poles. <br>
//! <br>
//!          The   result in  <Poles>  is    the value and    the <br>
//!          derivatives.  Poles[0] is  the value,  Poles[Degree] <br>
//!          is the last  derivative. <br>
  Standard_EXPORT   static  void Bohm(const Standard_Real U,const Standard_Integer Degree,const Standard_Integer N,Standard_Real& Knots,const Standard_Integer Dimension,Standard_Real& Poles) ;
  //! Used as argument for a non rational curve. <br>
//! <br>
      static  TColStd_Array1OfReal& NoWeights() ;
  //! Used as argument for a flatknots evaluation. <br>
//! <br>
      static  TColStd_Array1OfInteger& NoMults() ;
  //! Stores in LK  the usefull knots  for the BoorSchem <br>
//!          on the span Knots(Index) - Knots(Index+1) <br>
  Standard_EXPORT   static  void BuildKnots(const Standard_Integer Degree,const Standard_Integer Index,const Standard_Boolean Periodic,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,Standard_Real& LK) ;
  //! Return the index of the  first Pole to  use on the <br>
//!          span  Mults(Index)  - Mults(Index+1).  This  index <br>
//!          must be added to Poles.Lower(). <br>
  Standard_EXPORT   static  Standard_Integer PoleIndex(const Standard_Integer Degree,const Standard_Integer Index,const Standard_Boolean Periodic,const TColStd_Array1OfInteger& Mults) ;
  
  Standard_EXPORT   static  void BuildEval(const Standard_Integer Degree,const Standard_Integer Index,const TColStd_Array1OfReal& Poles,const TColStd_Array1OfReal& Weights,Standard_Real& LP) ;
  
  Standard_EXPORT   static  void BuildEval(const Standard_Integer Degree,const Standard_Integer Index,const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& Weights,Standard_Real& LP) ;
  //! Copy in <LP>  the poles and  weights for  the Eval <br>
//!          scheme. starting from  Poles(Poles.Lower()+Index) <br>
  Standard_EXPORT   static  void BuildEval(const Standard_Integer Degree,const Standard_Integer Index,const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& Weights,Standard_Real& LP) ;
  //! Copy in <LP>  poles for <Dimension>  Boor  scheme. <br>
//!          Starting  from    <Index>     *  <Dimension>, copy <br>
//!          <Length+1> poles. <br>
  Standard_EXPORT   static  void BuildBoor(const Standard_Integer Index,const Standard_Integer Length,const Standard_Integer Dimension,const TColStd_Array1OfReal& Poles,Standard_Real& LP) ;
  //! Returns the index in  the Boor result array of the <br>
//!          poles <Index>. If  the Boor  algorithm was perform <br>
//!          with <Length> and <Depth>. <br>
  Standard_EXPORT   static  Standard_Integer BoorIndex(const Standard_Integer Index,const Standard_Integer Length,const Standard_Integer Depth) ;
  //! Copy  the  pole at  position  <Index>  in  the Boor <br>
//!          scheme of   dimension <Dimension> to  <Position> in <br>
//!          the array <Pole>. <Position> is updated. <br>
  Standard_EXPORT   static  void GetPole(const Standard_Integer Index,const Standard_Integer Length,const Standard_Integer Depth,const Standard_Integer Dimension,Standard_Real& LocPoles,Standard_Integer& Position,TColStd_Array1OfReal& Pole) ;
  //! Returns in <NbPoles, NbKnots> the  new number of poles <br>
//!          and  knots    if  the  sequence   of  knots <AddKnots, <br>
//!          AddMults> is inserted in the sequence <Knots, Mults>. <br>
//! <br>
//!          Epsilon is used to compare knots for equality. <br>
//! <br>
//!          If Add is True  the multiplicities on  equal knots are <br>
//!          added. <br>
//! <br>
//!          If Add is False the max value of the multiplicities is <br>
//!          kept. <br>
//! <br>
//!          Return False if : <br>
//!            The knew knots are knot increasing. <br>
//!            The new knots are not in the range. <br>
  Standard_EXPORT   static  Standard_Boolean PrepareInsertKnots(const Standard_Integer Degree,const Standard_Boolean Periodic,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,const TColStd_Array1OfReal& AddKnots,const TColStd_Array1OfInteger& AddMults,Standard_Integer& NbPoles,Standard_Integer& NbKnots,const Standard_Real Epsilon,const Standard_Boolean Add = Standard_True) ;
  
  Standard_EXPORT   static  void InsertKnots(const Standard_Integer Degree,const Standard_Boolean Periodic,const Standard_Integer Dimension,const TColStd_Array1OfReal& Poles,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,const TColStd_Array1OfReal& AddKnots,const TColStd_Array1OfInteger& AddMults,TColStd_Array1OfReal& NewPoles,TColStd_Array1OfReal& NewKnots,TColStd_Array1OfInteger& NewMults,const Standard_Real Epsilon,const Standard_Boolean Add = Standard_True) ;
  
  Standard_EXPORT   static  void InsertKnots(const Standard_Integer Degree,const Standard_Boolean Periodic,const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& Weights,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,const TColStd_Array1OfReal& AddKnots,const TColStd_Array1OfInteger& AddMults,TColgp_Array1OfPnt& NewPoles,TColStd_Array1OfReal& NewWeights,TColStd_Array1OfReal& NewKnots,TColStd_Array1OfInteger& NewMults,const Standard_Real Epsilon,const Standard_Boolean Add = Standard_True) ;
  //! Insert   a  sequence  of  knots <AddKnots> with <br>
//!  multiplicities   <AddMults>. <AddKnots>   must  be a   non <br>
//!  decreasing sequence and verifies : <br>
//! <br>
//!  Knots(Knots.Lower()) <= AddKnots(AddKnots.Lower()) <br>
//!  Knots(Knots.Upper()) >= AddKnots(AddKnots.Upper()) <br>
//! <br>
//!  The NewPoles and NewWeights arrays must have a length : <br>
//!    Poles.Length() + Sum(AddMults()) <br>
//! <br>
//!  When a knot  to insert is identic  to an existing knot the <br>
//!  multiplicities   are added. <br>
//! <br>
//!  Epsilon is used to test knots for equality. <br>
//! <br>
//!  When AddMult is negative or null the knot is not inserted. <br>
//!  No multiplicity will becomes higher than the degree. <br>
//! <br>
//!  The new Knots and Multiplicities  are copied in <NewKnots> <br>
//!  and  <NewMults>. <br>
//! <br>
//!  All the New arrays should be correctly dimensioned. <br>
//! <br>
//!  When all  the new knots  are existing knots, i.e. only the <br>
//!  multiplicities  will  change it is   safe to  use the same <br>
//!  arrays as input and output. <br>
  Standard_EXPORT   static  void InsertKnots(const Standard_Integer Degree,const Standard_Boolean Periodic,const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& Weights,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,const TColStd_Array1OfReal& AddKnots,const TColStd_Array1OfInteger& AddMults,TColgp_Array1OfPnt2d& NewPoles,TColStd_Array1OfReal& NewWeights,TColStd_Array1OfReal& NewKnots,TColStd_Array1OfInteger& NewMults,const Standard_Real Epsilon,const Standard_Boolean Add = Standard_True) ;
  
  Standard_EXPORT   static  void InsertKnot(const Standard_Integer UIndex,const Standard_Real U,const Standard_Integer UMult,const Standard_Integer Degree,const Standard_Boolean Periodic,const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& Weights,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,TColgp_Array1OfPnt& NewPoles,TColStd_Array1OfReal& NewWeights) ;
  //! Insert a new knot U of multiplicity UMult in the <br>
//!  knot sequence. <br>
//! <br>
//!  The  location of the new Knot  should be given as an input <br>
//!  data.  UIndex locates the new knot U  in the knot sequence <br>
//!  and Knots (UIndex) < U < Knots (UIndex + 1). <br>
//! <br>
//!  The new control points corresponding to this insertion are <br>
//!  returned. Knots and Mults are not updated. <br>
  Standard_EXPORT   static  void InsertKnot(const Standard_Integer UIndex,const Standard_Real U,const Standard_Integer UMult,const Standard_Integer Degree,const Standard_Boolean Periodic,const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& Weights,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,TColgp_Array1OfPnt2d& NewPoles,TColStd_Array1OfReal& NewWeights) ;
  
  Standard_EXPORT   static  void RaiseMultiplicity(const Standard_Integer KnotIndex,const Standard_Integer Mult,const Standard_Integer Degree,const Standard_Boolean Periodic,const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& Weights,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,TColgp_Array1OfPnt& NewPoles,TColStd_Array1OfReal& NewWeights) ;
  //! Raise the multiplicity of knot to <UMult>. <br>
//! <br>
//!  The new control points  are  returned. Knots and Mults are <br>
//!  not updated. <br>
  Standard_EXPORT   static  void RaiseMultiplicity(const Standard_Integer KnotIndex,const Standard_Integer Mult,const Standard_Integer Degree,const Standard_Boolean Periodic,const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& Weights,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,TColgp_Array1OfPnt2d& NewPoles,TColStd_Array1OfReal& NewWeights) ;
  
  Standard_EXPORT   static  Standard_Boolean RemoveKnot(const Standard_Integer Index,const Standard_Integer Mult,const Standard_Integer Degree,const Standard_Boolean Periodic,const Standard_Integer Dimension,const TColStd_Array1OfReal& Poles,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,TColStd_Array1OfReal& NewPoles,TColStd_Array1OfReal& NewKnots,TColStd_Array1OfInteger& NewMults,const Standard_Real Tolerance) ;
  
  Standard_EXPORT   static  Standard_Boolean RemoveKnot(const Standard_Integer Index,const Standard_Integer Mult,const Standard_Integer Degree,const Standard_Boolean Periodic,const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& Weights,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,TColgp_Array1OfPnt& NewPoles,TColStd_Array1OfReal& NewWeights,TColStd_Array1OfReal& NewKnots,TColStd_Array1OfInteger& NewMults,const Standard_Real Tolerance) ;
  //! Decrement the  multiplicity  of <Knots(Index)> <br>
//!          to <Mult>. If <Mult>   is  null the   knot  is <br>
//!          removed. <br>
//! <br>
//!          As there are two ways to compute the new poles <br>
//!          the midlle   will  be used  as  long    as the <br>
//!          distance is lower than Tolerance. <br>
//! <br>
//!          If a  distance is  bigger  than  tolerance the <br>
//!          methods returns False  and  the new arrays are <br>
//!          not modified. <br>
//! <br>
//!          A low  tolerance can be  used  to test  if the <br>
//!          knot  can be  removed  without  modifying  the <br>
//!          curve. <br>
//! <br>
//!          A high tolerance  can be used  to "smooth" the <br>
//!          curve. <br>
  Standard_EXPORT   static  Standard_Boolean RemoveKnot(const Standard_Integer Index,const Standard_Integer Mult,const Standard_Integer Degree,const Standard_Boolean Periodic,const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& Weights,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,TColgp_Array1OfPnt2d& NewPoles,TColStd_Array1OfReal& NewWeights,TColStd_Array1OfReal& NewKnots,TColStd_Array1OfInteger& NewMults,const Standard_Real Tolerance) ;
  //! Returns the   number   of  knots   of  a  curve   with <br>
//!          multiplicities <Mults> after elevating the degree from <br>
//!          <Degree> to <NewDegree>. See the IncreaseDegree method <br>
//!          for more comments. <br>
  Standard_EXPORT   static  Standard_Integer IncreaseDegreeCountKnots(const Standard_Integer Degree,const Standard_Integer NewDegree,const Standard_Boolean Periodic,const TColStd_Array1OfInteger& Mults) ;
  
  Standard_EXPORT   static  void IncreaseDegree(const Standard_Integer Degree,const Standard_Integer NewDegree,const Standard_Boolean Periodic,const Standard_Integer Dimension,const TColStd_Array1OfReal& Poles,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,TColStd_Array1OfReal& NewPoles,TColStd_Array1OfReal& NewKnots,TColStd_Array1OfInteger& NewMults) ;
  
  Standard_EXPORT   static  void IncreaseDegree(const Standard_Integer Degree,const Standard_Integer NewDegree,const Standard_Boolean Periodic,const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& Weights,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,TColgp_Array1OfPnt& NewPoles,TColStd_Array1OfReal& NewWeights,TColStd_Array1OfReal& NewKnots,TColStd_Array1OfInteger& NewMults) ;
  
  Standard_EXPORT   static  void IncreaseDegree(const Standard_Integer Degree,const Standard_Integer NewDegree,const Standard_Boolean Periodic,const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& Weights,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,TColgp_Array1OfPnt2d& NewPoles,TColStd_Array1OfReal& NewWeights,TColStd_Array1OfReal& NewKnots,TColStd_Array1OfInteger& NewMults) ;
  
  Standard_EXPORT   static  void IncreaseDegree(const Standard_Integer NewDegree,const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& Weights,TColgp_Array1OfPnt& NewPoles,TColStd_Array1OfReal& NewWeights) ;
  //! Increase the degree of a bspline (or bezier) curve <br>
//!          of   dimension  <Dimension>  form <Degree>      to <br>
//!          <NewDegree>. <br>
//! <br>
//!          The number of poles in the new curve is : <br>
//! <br>
//!           Poles.Length() + (NewDegree - Degree) * Number of spans <br>
//! <br>
//!          Where the number of spans is : <br>
//! <br>
//!           LastUKnotIndex(Mults) - FirstUKnotIndex(Mults) + 1 <br>
//! <br>
//!           for a non-periodic curve <br>
//! <br>
//!           And Knots.Length() - 1 for a periodic curve. <br>
//! <br>
//!          The multiplicities of all  knots  are increased by <br>
//!          the degree elevation. <br>
//! <br>
//!          The new knots are usually  the same knots with the <br>
//!          exception of  a non-periodic curve with  the first <br>
//!          and last multiplicity not  equal to Degree+1 where <br>
//!          knots are removed  form the start  and the  bottom <br>
//!          untils the sum of the  multiplicities is  equal to <br>
//!          NewDegree+1  at the  knots   corresponding  to the <br>
//!          first and last parameters of the curve. <br>
//! <br>
//!          Example  :  Suppose a  curve  of degree 3 starting <br>
//!          with following knots and multiplicities : <br>
//! <br>
//!          knot : 0.  1.  2. <br>
//!          mult : 1   2   1 <br>
//! <br>
//!          The  FirstUKnot is  2.     because the   sum    of <br>
//!          multiplicities is Degree+1 : 1 + 2 + 1 = 4 = 3 + 1 <br>
//! <br>
//!          i.e. the first parameter  of the  curve is  2. and <br>
//!          will still be   2.  after degree  elevation.   Let <br>
//!          raises this curve to degree 4.  The multiplicities <br>
//!          are increased by 2. <br>
//! <br>
//!          They   become 2 3  2.   But     we need a   sum of <br>
//!          multiplicities  of 5 at knot  2. So the first knot <br>
//!          is removed and the new knots are : <br>
//! <br>
//!          knot : 1.  2. <br>
//!          mult : 3   2 <br>
//! <br>
//!          The multipicity   of the first  knot may   also be <br>
//!          reduced if the sum is still to big. <br>
//! <br>
//!          In the  most common situations (periodic  curve or <br>
//!          curve with first and last multiplicities equals to <br>
//!          Degree+1) the knots are knot changes. <br>
//! <br>
//!          The method IncreaseDegreeCountKnots can be used to <br>
//!          compute the new number of knots.\ <br>
  Standard_EXPORT   static  void IncreaseDegree(const Standard_Integer NewDegree,const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& Weights,TColgp_Array1OfPnt2d& NewPoles,TColStd_Array1OfReal& NewWeights) ;
  //! Set in <NbKnots> and <NbPolesToAdd> the number of Knots and <br>
//!          Poles   of  the NotPeriodic  Curve   identical  at the <br>
//!          periodic     curve with    a  degree    <Degree>  ,  a <br>
//!          knots-distribution with Multiplicities <Mults>. <br>
  Standard_EXPORT   static  void PrepareUnperiodize(const Standard_Integer Degree,const TColStd_Array1OfInteger& Mults,Standard_Integer& NbKnots,Standard_Integer& NbPoles) ;
  
  Standard_EXPORT   static  void Unperiodize(const Standard_Integer Degree,const Standard_Integer Dimension,const TColStd_Array1OfInteger& Mults,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfReal& Poles,TColStd_Array1OfInteger& NewMults,TColStd_Array1OfReal& NewKnots,TColStd_Array1OfReal& NewPoles) ;
  
  Standard_EXPORT   static  void Unperiodize(const Standard_Integer Degree,const TColStd_Array1OfInteger& Mults,const TColStd_Array1OfReal& Knots,const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& Weights,TColStd_Array1OfInteger& NewMults,TColStd_Array1OfReal& NewKnots,TColgp_Array1OfPnt& NewPoles,TColStd_Array1OfReal& NewWeights) ;
  
  Standard_EXPORT   static  void Unperiodize(const Standard_Integer Degree,const TColStd_Array1OfInteger& Mults,const TColStd_Array1OfReal& Knots,const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& Weights,TColStd_Array1OfInteger& NewMults,TColStd_Array1OfReal& NewKnots,TColgp_Array1OfPnt2d& NewPoles,TColStd_Array1OfReal& NewWeights) ;
  //! Set in <NbKnots> and <NbPoles> the number of Knots and <br>
//!          Poles of the  curve resulting  of  the trimming of the <br>
//!          BSplinecurve definded with <degree>, <knots>, <mults> <br>
  Standard_EXPORT   static  void PrepareTrimming(const Standard_Integer Degree,const Standard_Boolean Periodic,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,const Standard_Real U1,const Standard_Real U2,Standard_Integer& NbKnots,Standard_Integer& NbPoles) ;
  
  Standard_EXPORT   static  void Trimming(const Standard_Integer Degree,const Standard_Boolean Periodic,const Standard_Integer Dimension,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,const TColStd_Array1OfReal& Poles,const Standard_Real U1,const Standard_Real U2,TColStd_Array1OfReal& NewKnots,TColStd_Array1OfInteger& NewMults,TColStd_Array1OfReal& NewPoles) ;
  
  Standard_EXPORT   static  void Trimming(const Standard_Integer Degree,const Standard_Boolean Periodic,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& Weights,const Standard_Real U1,const Standard_Real U2,TColStd_Array1OfReal& NewKnots,TColStd_Array1OfInteger& NewMults,TColgp_Array1OfPnt& NewPoles,TColStd_Array1OfReal& NewWeights) ;
  
  Standard_EXPORT   static  void Trimming(const Standard_Integer Degree,const Standard_Boolean Periodic,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& Weights,const Standard_Real U1,const Standard_Real U2,TColStd_Array1OfReal& NewKnots,TColStd_Array1OfInteger& NewMults,TColgp_Array1OfPnt2d& NewPoles,TColStd_Array1OfReal& NewWeights) ;
  
  Standard_EXPORT   static  void D0(const Standard_Real U,const Standard_Integer Index,const Standard_Integer Degree,const Standard_Boolean Periodic,const TColStd_Array1OfReal& Poles,const TColStd_Array1OfReal& Weights,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,Standard_Real& P) ;
  
  Standard_EXPORT   static  void D0(const Standard_Real U,const Standard_Integer Index,const Standard_Integer Degree,const Standard_Boolean Periodic,const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& Weights,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,gp_Pnt& P) ;
  
  Standard_EXPORT   static  void D0(const Standard_Real U,const Standard_Integer UIndex,const Standard_Integer Degree,const Standard_Boolean Periodic,const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& Weights,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,gp_Pnt2d& P) ;
  
  Standard_EXPORT   static  void D0(const Standard_Real U,const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& Weights,gp_Pnt& P) ;
  
  Standard_EXPORT   static  void D0(const Standard_Real U,const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& Weights,gp_Pnt2d& P) ;
  
  Standard_EXPORT   static  void D1(const Standard_Real U,const Standard_Integer Index,const Standard_Integer Degree,const Standard_Boolean Periodic,const TColStd_Array1OfReal& Poles,const TColStd_Array1OfReal& Weights,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,Standard_Real& P,Standard_Real& V) ;
  
  Standard_EXPORT   static  void D1(const Standard_Real U,const Standard_Integer Index,const Standard_Integer Degree,const Standard_Boolean Periodic,const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& Weights,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,gp_Pnt& P,gp_Vec& V) ;
  
  Standard_EXPORT   static  void D1(const Standard_Real U,const Standard_Integer UIndex,const Standard_Integer Degree,const Standard_Boolean Periodic,const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& Weights,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,gp_Pnt2d& P,gp_Vec2d& V) ;
  
  Standard_EXPORT   static  void D1(const Standard_Real U,const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& Weights,gp_Pnt& P,gp_Vec& V) ;
  
  Standard_EXPORT   static  void D1(const Standard_Real U,const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& Weights,gp_Pnt2d& P,gp_Vec2d& V) ;
  
  Standard_EXPORT   static  void D2(const Standard_Real U,const Standard_Integer Index,const Standard_Integer Degree,const Standard_Boolean Periodic,const TColStd_Array1OfReal& Poles,const TColStd_Array1OfReal& Weights,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,Standard_Real& P,Standard_Real& V1,Standard_Real& V2) ;
  
  Standard_EXPORT   static  void D2(const Standard_Real U,const Standard_Integer Index,const Standard_Integer Degree,const Standard_Boolean Periodic,const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& Weights,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,gp_Pnt& P,gp_Vec& V1,gp_Vec& V2) ;
  
  Standard_EXPORT   static  void D2(const Standard_Real U,const Standard_Integer UIndex,const Standard_Integer Degree,const Standard_Boolean Periodic,const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& Weights,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,gp_Pnt2d& P,gp_Vec2d& V1,gp_Vec2d& V2) ;
  
  Standard_EXPORT   static  void D2(const Standard_Real U,const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& Weights,gp_Pnt& P,gp_Vec& V1,gp_Vec& V2) ;
  
  Standard_EXPORT   static  void D2(const Standard_Real U,const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& Weights,gp_Pnt2d& P,gp_Vec2d& V1,gp_Vec2d& V2) ;
  
  Standard_EXPORT   static  void D3(const Standard_Real U,const Standard_Integer Index,const Standard_Integer Degree,const Standard_Boolean Periodic,const TColStd_Array1OfReal& Poles,const TColStd_Array1OfReal& Weights,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,Standard_Real& P,Standard_Real& V1,Standard_Real& V2,Standard_Real& V3) ;
  
  Standard_EXPORT   static  void D3(const Standard_Real U,const Standard_Integer Index,const Standard_Integer Degree,const Standard_Boolean Periodic,const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& Weights,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,gp_Pnt& P,gp_Vec& V1,gp_Vec& V2,gp_Vec& V3) ;
  
  Standard_EXPORT   static  void D3(const Standard_Real U,const Standard_Integer UIndex,const Standard_Integer Degree,const Standard_Boolean Periodic,const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& Weights,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,gp_Pnt2d& P,gp_Vec2d& V1,gp_Vec2d& V2,gp_Vec2d& V3) ;
  
  Standard_EXPORT   static  void D3(const Standard_Real U,const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& Weights,gp_Pnt& P,gp_Vec& V1,gp_Vec& V2,gp_Vec& V3) ;
  
  Standard_EXPORT   static  void D3(const Standard_Real U,const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& Weights,gp_Pnt2d& P,gp_Vec2d& V1,gp_Vec2d& V2,gp_Vec2d& V3) ;
  
  Standard_EXPORT   static  void DN(const Standard_Real U,const Standard_Integer N,const Standard_Integer Index,const Standard_Integer Degree,const Standard_Boolean Periodic,const TColStd_Array1OfReal& Poles,const TColStd_Array1OfReal& Weights,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,Standard_Real& VN) ;
  
  Standard_EXPORT   static  void DN(const Standard_Real U,const Standard_Integer N,const Standard_Integer Index,const Standard_Integer Degree,const Standard_Boolean Periodic,const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& Weights,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,gp_Vec& VN) ;
  
  Standard_EXPORT   static  void DN(const Standard_Real U,const Standard_Integer N,const Standard_Integer UIndex,const Standard_Integer Degree,const Standard_Boolean Periodic,const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& Weights,const TColStd_Array1OfReal& Knots,const TColStd_Array1OfInteger& Mults,gp_Vec2d& V) ;
  
  Standard_EXPORT   static  void DN(const Standard_Real U,const Standard_Integer N,const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& Weights,gp_Pnt& P,gp_Vec& VN) ;
  //!  The  above  functions  compute   values and <br>
//!         derivatives in the following situations : <br>
//! <br>
//!         * 3D, 2D and 1D <br>
//! <br>
//!         * Rational or not Rational. <br>
//! <br>
//!         * Knots  and multiplicities or "flat knots" without <br>
//!         multiplicities. <br>
//! <br>
//!         * The  <Index>  is   the the  localization  of  the <br>
//!         parameter in the knot sequence.  If <Index> is  out <br>
//!         of range the correct value will be searched. <br>
//! <br>
//! <br>
//!         VERY IMPORTANT!!! <br>
//!         USE  BSplCLib::NoWeights()  as Weights argument for non <br>
//!         rational curves computations. <br>
  Standard_EXPORT   static  void DN(const Standard_Real U,const Standard_Integer N,const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& Weights,gp_Pnt2d& P,gp_Vec2d& VN) ;
  //! This  evaluates  the Bspline  Basis  at  a <br>
//!          given  parameter  Parameter   up   to  the <br>
//!          requested   DerivativeOrder  and store the <br>
//!          result  in the  array BsplineBasis  in the <br>
//!          following   fashion <br>
//!          BSplineBasis(1,1)   = <br>
//!          value of first non vanishing <br>
//!          Bspline function which has Index FirstNonZeroBsplineIndex <br>
//!            BsplineBasis(1,2)   = <br>
//!            value of second non vanishing <br>
//!          Bspline   function which  has   Index <br>
//!          FirstNonZeroBsplineIndex + 1 <br>
//!          BsplineBasis(1,n)   = <br>
//!            value of second non vanishing non vanishing <br>
//!          Bspline   function which  has   Index <br>
//!          FirstNonZeroBsplineIndex + n (n <= Order) <br>
//!          BSplineBasis(2,1)   = <br>
//!          value of derivative of first non vanishing <br>
//!          Bspline function which has Index FirstNonZeroBsplineIndex <br>
//!            BSplineBasis(N,1)   = <br>
//!          value of Nth derivative of first non vanishing <br>
//!          Bspline function which has Index FirstNonZeroBsplineIndex <br>
//!          if N <= DerivativeOrder + 1 <br>
  Standard_EXPORT   static  Standard_Integer EvalBsplineBasis(const Standard_Integer Side,const Standard_Integer DerivativeOrder,const Standard_Integer Order,const TColStd_Array1OfReal& FlatKnots,const Standard_Real Parameter,Standard_Integer& FirstNonZeroBsplineIndex,math_Matrix& BsplineBasis) ;
  //! This Builds   a fully  blown   Matrix of <br>
//!            (ni) <br>
//!          Bi    (tj) <br>
//! <br>
//!          with i  and j within 1..Order + NumPoles <br>
//!          The  integer ni is   the ith slot of the <br>
//!          array OrderArray, tj is the jth slot of <br>
//!          the array Parameters <br>
  Standard_EXPORT   static  Standard_Integer BuildBSpMatrix(const TColStd_Array1OfReal& Parameters,const TColStd_Array1OfInteger& OrderArray,const TColStd_Array1OfReal& FlatKnots,const Standard_Integer Degree,math_Matrix& Matrix,Standard_Integer& UpperBandWidth,Standard_Integer& LowerBandWidth) ;
  //! this  factors  the Banded Matrix in <br>
//!         the LU form with a Banded storage of <br>
//!         components of the L matrix <br>
//!         WARNING : do not use if the Matrix is <br>
//!         totally positive (It is the case for <br>
//!         Bspline matrices build as above with <br>
//!         parameters being the Schoenberg points <br>
  Standard_EXPORT   static  Standard_Integer FactorBandedMatrix(math_Matrix& Matrix,const Standard_Integer UpperBandWidth,const Standard_Integer LowerBandWidth,Standard_Integer& PivotIndexProblem) ;
  //! This solves  the system Matrix.X =  B <br>
//!         with when Matrix is factored in LU form <br>
//!         The  Array   is    an   seen   as    an <br>
//!         Array[1..N][1..ArrayDimension] with N = <br>
//!         the  rank  of the  matrix  Matrix.  The <br>
//!         result is stored   in Array  when  each <br>
//!         coordinate is  solved that is  B is the <br>
//!         array whose values are <br>
//!         B[i] = Array[i][p] for each p in 1..ArrayDimension <br>
  Standard_EXPORT   static  Standard_Integer SolveBandedSystem(const math_Matrix& Matrix,const Standard_Integer UpperBandWidth,const Standard_Integer LowerBandWidth,const Standard_Integer ArrayDimension,Standard_Real& Array) ;
  //! This solves  the system Matrix.X =  B <br>
//!         with when Matrix is factored in LU form <br>
//!         The  Array   has the length of <br>
//!         the  rank  of the  matrix  Matrix.  The <br>
//!         result is stored   in Array  when  each <br>
//!         coordinate is  solved that is  B is the <br>
//!         array whose values are <br>
//!         B[i] = Array[i][p] for each p in 1..ArrayDimension <br>
  Standard_EXPORT   static  Standard_Integer SolveBandedSystem(const math_Matrix& Matrix,const Standard_Integer UpperBandWidth,const Standard_Integer LowerBandWidth,TColgp_Array1OfPnt2d& Array) ;
  //! This solves  the system Matrix.X =  B <br>
//!         with when Matrix is factored in LU form <br>
//!         The  Array   has the length of <br>
//!         the  rank  of the  matrix  Matrix.  The <br>
//!         result is stored   in Array  when  each <br>
//!         coordinate is  solved that is  B is the <br>
//!         array whose values are <br>
//!         B[i] = Array[i][p] for each p in 1..ArrayDimension <br>
  Standard_EXPORT   static  Standard_Integer SolveBandedSystem(const math_Matrix& Matrix,const Standard_Integer UpperBandWidth,const Standard_Integer LowerBandWidth,TColgp_Array1OfPnt& Array) ;
  
  Standard_EXPORT   static  Standard_Integer SolveBandedSystem(const math_Matrix& Matrix,const Standard_Integer UpperBandWidth,const Standard_Integer LowerBandWidth,const Standard_Boolean HomogenousFlag,const Standard_Integer ArrayDimension,Standard_Real& Array,Standard_Real& Weights) ;
  //! This solves the  system Matrix.X =  B <br>
//!         with when Matrix is factored in LU form <br>
//!         The    Array   is    an   seen  as   an <br>
//!         Array[1..N][1..ArrayDimension] with N = <br>
//!         the  rank  of  the  matrix Matrix.  The <br>
//!         result is  stored   in Array when  each <br>
//!         coordinate is  solved that is B  is the <br>
//!         array  whose   values     are   B[i]  = <br>
//!         Array[i][p]       for     each  p    in <br>
//!         1..ArrayDimension. If  HomogeneousFlag == <br>
//!         0  the  Poles  are  multiplied by   the <br>
//!         Weights   uppon   Entry   and      once <br>
//!         interpolation   is    carried  over the <br>
//!         result of the  poles are divided by the <br>
//!         result of   the   interpolation of  the <br>
//!         weights. Otherwise if HomogenousFlag == 1 <br>
//!         the Poles and Weigths are treated homogenously <br>
//!         that is that those are interpolated as they <br>
//!         are and result is returned without division <br>
//!         by the interpolated weigths. <br>
  Standard_EXPORT   static  Standard_Integer SolveBandedSystem(const math_Matrix& Matrix,const Standard_Integer UpperBandWidth,const Standard_Integer LowerBandWidth,const Standard_Boolean HomogenousFlag,TColgp_Array1OfPnt2d& Array,TColStd_Array1OfReal& Weights) ;
  //! This solves  the system Matrix.X =  B <br>
//!         with when Matrix is factored in LU form <br>
//!         The  Array   is    an   seen   as    an <br>
//!         Array[1..N][1..ArrayDimension] with N = <br>
//!         the  rank  of the  matrix  Matrix.  The <br>
//!         result is stored   in Array  when  each <br>
//!         coordinate is  solved that is  B is the <br>
//!         array whose values are <br>
//!         B[i] = Array[i][p] for each p in 1..ArrayDimension <br>
//!         If  HomogeneousFlag == <br>
//!         0  the  Poles  are  multiplied by   the <br>
//!         Weights   uppon   Entry   and      once <br>
//!         interpolation   is    carried  over the <br>
//!         result of the  poles are divided by the <br>
//!         result of   the   interpolation of  the <br>
//!         weights. Otherwise if HomogenousFlag == 1 <br>
//!         the Poles and Weigths are treated homogenously <br>
//!         that is that those are interpolated as they <br>
//!         are and result is returned without division <br>
//!         by the interpolated weigths. <br>
  Standard_EXPORT   static  Standard_Integer SolveBandedSystem(const math_Matrix& Matrix,const Standard_Integer UpperBandWidth,const Standard_Integer LowerBandWidth,const Standard_Boolean HomogeneousFlag,TColgp_Array1OfPnt& Array,TColStd_Array1OfReal& Weights) ;
  //!  Merges  two knot vector by   setting the starting and <br>
//!          ending values to StartValue and EndValue <br>
  Standard_EXPORT   static  void MergeBSplineKnots(const Standard_Real Tolerance,const Standard_Real StartValue,const Standard_Real EndValue,const Standard_Integer Degree1,const TColStd_Array1OfReal& Knots1,const TColStd_Array1OfInteger& Mults1,const Standard_Integer Degree2,const TColStd_Array1OfReal& Knots2,const TColStd_Array1OfInteger& Mults2,Standard_Integer& NumPoles,Handle(TColStd_HArray1OfReal)& NewKnots,Handle(TColStd_HArray1OfInteger)& NewMults) ;
  //! This function will compose  a given Vectorial BSpline F(t) <br>
//!          defined  by its  BSplineDegree and BSplineFlatKnotsl, <br>
//!          its Poles  array which are coded as  an array of Real <br>
//!          of  the  form  [1..NumPoles][1..PolesDimension] with  a <br>
//!          function     a(t) which is   assumed to   satisfy the <br>
//!          following: <br>
  Standard_EXPORT   static  void FunctionReparameterise(const BSplCLib_EvaluatorFunction& Function,const Standard_Integer BSplineDegree,const TColStd_Array1OfReal& BSplineFlatKnots,const Standard_Integer PolesDimension,Standard_Real& Poles,const TColStd_Array1OfReal& FlatKnots,const Standard_Integer NewDegree,Standard_Real& NewPoles,Standard_Integer& Status) ;
  //! This function will compose  a given Vectorial BSpline F(t) <br>
//!          defined  by its  BSplineDegree and BSplineFlatKnotsl, <br>
//!          its Poles  array which are coded as  an array of Real <br>
//!          of  the  form  [1..NumPoles][1..PolesDimension] with  a <br>
//!          function     a(t) which is   assumed to   satisfy the <br>
//!          following: <br>
  Standard_EXPORT   static  void FunctionReparameterise(const BSplCLib_EvaluatorFunction& Function,const Standard_Integer BSplineDegree,const TColStd_Array1OfReal& BSplineFlatKnots,const TColStd_Array1OfReal& Poles,const TColStd_Array1OfReal& FlatKnots,const Standard_Integer NewDegree,TColStd_Array1OfReal& NewPoles,Standard_Integer& Status) ;
  //! this will compose  a given Vectorial BSpline F(t) <br>
//!          defined  by its  BSplineDegree and BSplineFlatKnotsl, <br>
//!          its Poles  array which are coded as  an array of Real <br>
//!          of  the  form  [1..NumPoles][1..PolesDimension] with  a <br>
//!          function     a(t) which is   assumed to   satisfy the <br>
//!          following  : 1. F(a(t))  is a polynomial BSpline <br>
//!          that can be expressed  exactly as a BSpline of degree <br>
//!          NewDegree on the knots FlatKnots <br>
//!                       2. a(t) defines a differentiable <br>
//!          isomorphism between the range of FlatKnots to the range <br>
//!          of BSplineFlatKnots which is the <br>
//!          same as the  range of F(t) <br>
//!  Warning: it is <br>
//!          the caller's responsability to insure that conditions <br>
//!          1. and  2. above are  satisfied : no check whatsoever <br>
//!          is made in this method <br>
//! Status will return 0 if OK else it will return the pivot index <br>
//!        of the matrix that was inverted to compute the multiplied <br>
//!        BSpline : the method used is interpolation at Schoenenberg <br>
//!        points of F(a(t)) <br>
  Standard_EXPORT   static  void FunctionReparameterise(const BSplCLib_EvaluatorFunction& Function,const Standard_Integer BSplineDegree,const TColStd_Array1OfReal& BSplineFlatKnots,const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& FlatKnots,const Standard_Integer NewDegree,TColgp_Array1OfPnt& NewPoles,Standard_Integer& Status) ;
  //! this will compose  a given Vectorial BSpline F(t) <br>
//!          defined  by its  BSplineDegree and BSplineFlatKnotsl, <br>
//!          its Poles  array which are coded as  an array of Real <br>
//!          of  the  form  [1..NumPoles][1..PolesDimension] with  a <br>
//!          function     a(t) which is   assumed to   satisfy the <br>
//!          following  : 1. F(a(t))  is a polynomial BSpline <br>
//!          that can be expressed  exactly as a BSpline of degree <br>
//!          NewDegree on the knots FlatKnots <br>
//!                       2. a(t) defines a differentiable <br>
//!          isomorphism between the range of FlatKnots to the range <br>
//!          of BSplineFlatKnots which is the <br>
//!          same as the  range of F(t) <br>
//!  Warning: it is <br>
//!          the caller's responsability to insure that conditions <br>
//!          1. and  2. above are  satisfied : no check whatsoever <br>
//!          is made in this method <br>
//! Status will return 0 if OK else it will return the pivot index <br>
//!        of the matrix that was inverted to compute the multiplied <br>
//!        BSpline : the method used is interpolation at Schoenenberg <br>
//!        points of F(a(t)) <br>
  Standard_EXPORT   static  void FunctionReparameterise(const BSplCLib_EvaluatorFunction& Function,const Standard_Integer BSplineDegree,const TColStd_Array1OfReal& BSplineFlatKnots,const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& FlatKnots,const Standard_Integer NewDegree,TColgp_Array1OfPnt2d& NewPoles,Standard_Integer& Status) ;
  //! this will  multiply a given Vectorial BSpline F(t) <br>
//!          defined  by its  BSplineDegree and BSplineFlatKnotsl, <br>
//!          its Poles  array which are coded as  an array of Real <br>
//!          of  the  form  [1..NumPoles][1..PolesDimension] by  a <br>
//!          function     a(t) which is   assumed to   satisfy the <br>
//!          following  : 1. a(t)  * F(t)  is a polynomial BSpline <br>
//!          that can be expressed  exactly as a BSpline of degree <br>
//!          NewDegree on the knots FlatKnots 2. the range of a(t) <br>
//!           is the same as the  range of F(t) <br>
//!  Warning: it is <br>
//!          the caller's responsability to insure that conditions <br>
//!          1. and  2. above are  satisfied : no check whatsoever <br>
//!          is made in this method <br>
//! Status will return 0 if OK else it will return the pivot index <br>
//!        of the matrix that was inverted to compute the multiplied <br>
//!        BSpline : the method used is interpolation at Schoenenberg <br>
//!        points of a(t)*F(t) <br>
  Standard_EXPORT   static  void FunctionMultiply(const BSplCLib_EvaluatorFunction& Function,const Standard_Integer BSplineDegree,const TColStd_Array1OfReal& BSplineFlatKnots,const Standard_Integer PolesDimension,Standard_Real& Poles,const TColStd_Array1OfReal& FlatKnots,const Standard_Integer NewDegree,Standard_Real& NewPoles,Standard_Integer& Status) ;
  //! this will  multiply a given Vectorial BSpline F(t) <br>
//!          defined  by its  BSplineDegree and BSplineFlatKnotsl, <br>
//!          its Poles  array which are coded as  an array of Real <br>
//!          of  the  form  [1..NumPoles][1..PolesDimension] by  a <br>
//!          function     a(t) which is   assumed to   satisfy the <br>
//!          following  : 1. a(t)  * F(t)  is a polynomial BSpline <br>
//!          that can be expressed  exactly as a BSpline of degree <br>
//!          NewDegree on the knots FlatKnots 2. the range of a(t) <br>
//!           is the same as the  range of F(t) <br>
//!  Warning: it is <br>
//!          the caller's responsability to insure that conditions <br>
//!          1. and  2. above are  satisfied : no check whatsoever <br>
//!          is made in this method <br>
//! Status will return 0 if OK else it will return the pivot index <br>
//!        of the matrix that was inverted to compute the multiplied <br>
//!        BSpline : the method used is interpolation at Schoenenberg <br>
//!        points of a(t)*F(t) <br>
  Standard_EXPORT   static  void FunctionMultiply(const BSplCLib_EvaluatorFunction& Function,const Standard_Integer BSplineDegree,const TColStd_Array1OfReal& BSplineFlatKnots,const TColStd_Array1OfReal& Poles,const TColStd_Array1OfReal& FlatKnots,const Standard_Integer NewDegree,TColStd_Array1OfReal& NewPoles,Standard_Integer& Status) ;
  //! this will  multiply a given Vectorial BSpline F(t) <br>
//!          defined  by its  BSplineDegree and BSplineFlatKnotsl, <br>
//!          its Poles  array which are coded as  an array of Real <br>
//!          of  the  form  [1..NumPoles][1..PolesDimension] by  a <br>
//!          function     a(t) which is   assumed to   satisfy the <br>
//!          following  : 1. a(t)  * F(t)  is a polynomial BSpline <br>
//!          that can be expressed  exactly as a BSpline of degree <br>
//!          NewDegree on the knots FlatKnots 2. the range of a(t) <br>
//!           is the same as the  range of F(t) <br>
//!  Warning: it is <br>
//!          the caller's responsability to insure that conditions <br>
//!          1. and  2. above are  satisfied : no check whatsoever <br>
//!          is made in this method <br>
//! Status will return 0 if OK else it will return the pivot index <br>
//!        of the matrix that was inverted to compute the multiplied <br>
//!        BSpline : the method used is interpolation at Schoenenberg <br>
//!        points of a(t)*F(t) <br>
  Standard_EXPORT   static  void FunctionMultiply(const BSplCLib_EvaluatorFunction& Function,const Standard_Integer BSplineDegree,const TColStd_Array1OfReal& BSplineFlatKnots,const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& FlatKnots,const Standard_Integer NewDegree,TColgp_Array1OfPnt2d& NewPoles,Standard_Integer& Status) ;
  //! this will  multiply a given Vectorial BSpline F(t) <br>
//!          defined  by its  BSplineDegree and BSplineFlatKnotsl, <br>
//!          its Poles  array which are coded as  an array of Real <br>
//!          of  the  form  [1..NumPoles][1..PolesDimension] by  a <br>
//!          function     a(t) which is   assumed to   satisfy the <br>
//!          following  : 1. a(t)  * F(t)  is a polynomial BSpline <br>
//!          that can be expressed  exactly as a BSpline of degree <br>
//!          NewDegree on the knots FlatKnots 2. the range of a(t) <br>
//!           is the same as the  range of F(t) <br>
//!  Warning: it is <br>
//!          the caller's responsability to insure that conditions <br>
//!          1. and  2. above are  satisfied : no check whatsoever <br>
//!          is made in this method <br>
//! Status will return 0 if OK else it will return the pivot index <br>
//!        of the matrix that was inverted to compute the multiplied <br>
//!        BSpline : the method used is interpolation at Schoenenberg <br>
//!        points of a(t)*F(t) <br>
  Standard_EXPORT   static  void FunctionMultiply(const BSplCLib_EvaluatorFunction& Function,const Standard_Integer BSplineDegree,const TColStd_Array1OfReal& BSplineFlatKnots,const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& FlatKnots,const Standard_Integer NewDegree,TColgp_Array1OfPnt& NewPoles,Standard_Integer& Status) ;
  //! Perform the De Boor   algorithm  to  evaluate a point at <br>
//!          parameter <U>, with <Degree> and <Dimension>. <br>
//! <br>
//!          Poles is  an array of  Reals of size <br>
//! <br>
//!          <Dimension> *  <Degree>+1 <br>
//! <br>
//!          Containing the  poles.  At  the end <Poles> contains <br>
//!          the current point.   Poles Contain all  the poles of <br>
//!          the BsplineCurve, Knots  also Contains all the knots <br>
//!          of the BsplineCurve.  ExtrapMode has two slots [0] = <br>
//!          Degree used to extrapolate before the first knot [1] <br>
//!          = Degre used to  extrapolate after the last knot has <br>
//!          to be between 1 and  Degree <br>
  Standard_EXPORT   static  void Eval(const Standard_Real U,const Standard_Boolean PeriodicFlag,const Standard_Integer DerivativeRequest,Standard_Integer& ExtrapMode,const Standard_Integer Degree,const TColStd_Array1OfReal& FlatKnots,const Standard_Integer ArrayDimension,Standard_Real& Poles,Standard_Real& Result) ;
  //! Perform the  De Boor algorithm  to evaluate a point at <br>
//!          parameter   <U>,  with   <Degree>    and  <Dimension>. <br>
//!          Evaluates by multiplying the  Poles by the Weights and <br>
//!          gives  the homogeneous  result  in PolesResult that is <br>
//!          the results of the evaluation of the numerator once it <br>
//!          has     been  multiplied   by  the     weights and  in <br>
//!          WeightsResult one has  the result of the evaluation of <br>
//!          the denominator <br>
//! <br>
//!  Warning:   <PolesResult> and <WeightsResult>  must be   dimensionned <br>
//!          properly. <br>
  Standard_EXPORT   static  void Eval(const Standard_Real U,const Standard_Boolean PeriodicFlag,const Standard_Integer DerivativeRequest,Standard_Integer& ExtrapMode,const Standard_Integer Degree,const TColStd_Array1OfReal& FlatKnots,const Standard_Integer ArrayDimension,Standard_Real& Poles,Standard_Real& Weights,Standard_Real& PolesResult,Standard_Real& WeightsResult) ;
  //! Perform the evaluation of the Bspline Basis <br>
//!          and then multiplies by the weights <br>
//!          this just evaluates the current point <br>
  Standard_EXPORT   static  void Eval(const Standard_Real U,const Standard_Boolean PeriodicFlag,const Standard_Boolean HomogeneousFlag,Standard_Integer& ExtrapMode,const Standard_Integer Degree,const TColStd_Array1OfReal& FlatKnots,const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& Weights,gp_Pnt& Point,Standard_Real& Weight) ;
  //! Perform the evaluation of the Bspline Basis <br>
//!          and then multiplies by the weights <br>
//!          this just evaluates the current point <br>
//! <br>
  Standard_EXPORT   static  void Eval(const Standard_Real U,const Standard_Boolean PeriodicFlag,const Standard_Boolean HomogeneousFlag,Standard_Integer& ExtrapMode,const Standard_Integer Degree,const TColStd_Array1OfReal& FlatKnots,const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& Weights,gp_Pnt2d& Point,Standard_Real& Weight) ;
  //! Extend a BSpline nD using the tangency map <br>
//!          <C1Coefficient> is the coefficient of reparametrisation <br>
//!          <Continuity> must be equal to 1, 2 or 3. <br>
//!          <Degree> must be greater or equal than <Continuity> + 1. <br>
//! <br>
//!  Warning:   <KnotsResult> and <PolesResult>  must be   dimensionned <br>
//!          properly. <br>
  Standard_EXPORT   static  void TangExtendToConstraint(const TColStd_Array1OfReal& FlatKnots,const Standard_Real C1Coefficient,const Standard_Integer NumPoles,Standard_Real& Poles,const Standard_Integer Dimension,const Standard_Integer Degree,const TColStd_Array1OfReal& ConstraintPoint,const Standard_Integer Continuity,const Standard_Boolean After,Standard_Integer& NbPolesResult,Standard_Integer& NbKnotsRsult,Standard_Real& KnotsResult,Standard_Real& PolesResult) ;
  //! Perform the evaluation of the of the cache <br>
//!          the parameter must be normalized between <br>
//!          the 0 and 1 for the span. <br>
//!          The Cache must be valid when calling this <br>
//!          routine. Geom Package will insure that. <br>
//!          and then multiplies by the weights <br>
//!          this just evaluates the current point <br>
//!          the CacheParameter is where the Cache was <br>
//!          constructed the SpanLength is to normalize <br>
//!          the polynomial in the cache to avoid bad conditioning <br>
//!          effects <br>
  Standard_EXPORT   static  void CacheD0(const Standard_Real U,const Standard_Integer Degree,const Standard_Real CacheParameter,const Standard_Real SpanLenght,const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& Weights,gp_Pnt& Point) ;
  //! Perform the evaluation of the Bspline Basis <br>
//!          and then multiplies by the weights <br>
//!          this just evaluates the current point <br>
//!          the parameter must be normalized between <br>
//!          the 0 and 1 for the span. <br>
//!          The Cache must be valid when calling this <br>
//!          routine. Geom Package will insure that. <br>
//!          and then multiplies by the weights <br>
//!          ththe CacheParameter is where the Cache was <br>
//!          constructed the SpanLength is to normalize <br>
//!          the polynomial in the cache to avoid bad conditioning <br>
//!          effectsis just evaluates the current point <br>
  Standard_EXPORT   static  void CacheD0(const Standard_Real U,const Standard_Integer Degree,const Standard_Real CacheParameter,const Standard_Real SpanLenght,const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& Weights,gp_Pnt2d& Point) ;
  //! Calls CacheD0 for Bezier  Curves Arrays computed with <br>
//!          the method PolesCoefficients. <br>
//!  Warning: To be used for Beziercurves ONLY!!! <br>
      static  void CoefsD0(const Standard_Real U,const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& Weights,gp_Pnt& Point) ;
  //! Calls CacheD0 for Bezier  Curves Arrays computed with <br>
//!          the method PolesCoefficients. <br>
//!  Warning: To be used for Beziercurves ONLY!!! <br>
      static  void CoefsD0(const Standard_Real U,const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& Weights,gp_Pnt2d& Point) ;
  //! Perform the evaluation of the of the cache <br>
//!          the parameter must be normalized between <br>
//!          the 0 and 1 for the span. <br>
//!          The Cache must be valid when calling this <br>
//!          routine. Geom Package will insure that. <br>
//!          and then multiplies by the weights <br>
//!          this just evaluates the current point <br>
//!          the CacheParameter is where the Cache was <br>
//!          constructed the SpanLength is to normalize <br>
//!          the polynomial in the cache to avoid bad conditioning <br>
//!          effects <br>
  Standard_EXPORT   static  void CacheD1(const Standard_Real U,const Standard_Integer Degree,const Standard_Real CacheParameter,const Standard_Real SpanLenght,const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& Weights,gp_Pnt& Point,gp_Vec& Vec) ;
  //! Perform the evaluation of the Bspline Basis <br>
//!          and then multiplies by the weights <br>
//!          this just evaluates the current point <br>
//!          the parameter must be normalized between <br>
//!          the 0 and 1 for the span. <br>
//!          The Cache must be valid when calling this <br>
//!          routine. Geom Package will insure that. <br>
//!          and then multiplies by the weights <br>
//!          ththe CacheParameter is where the Cache was <br>
//!          constructed the SpanLength is to normalize <br>
//!          the polynomial in the cache to avoid bad conditioning <br>
//!          effectsis just evaluates the current point <br>
  Standard_EXPORT   static  void CacheD1(const Standard_Real U,const Standard_Integer Degree,const Standard_Real CacheParameter,const Standard_Real SpanLenght,const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& Weights,gp_Pnt2d& Point,gp_Vec2d& Vec) ;
  //! Calls CacheD1 for Bezier  Curves Arrays computed with <br>
//!          the method PolesCoefficients. <br>
//!  Warning: To be used for Beziercurves ONLY!!! <br>
      static  void CoefsD1(const Standard_Real U,const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& Weights,gp_Pnt& Point,gp_Vec& Vec) ;
  //! Calls CacheD1 for Bezier  Curves Arrays computed with <br>
//!          the method PolesCoefficients. <br>
//!  Warning: To be used for Beziercurves ONLY!!! <br>
      static  void CoefsD1(const Standard_Real U,const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& Weights,gp_Pnt2d& Point,gp_Vec2d& Vec) ;
  //! Perform the evaluation of the of the cache <br>
//!          the parameter must be normalized between <br>
//!          the 0 and 1 for the span. <br>
//!          The Cache must be valid when calling this <br>
//!          routine. Geom Package will insure that. <br>
//!          and then multiplies by the weights <br>
//!          this just evaluates the current point <br>
//!          the CacheParameter is where the Cache was <br>
//!          constructed the SpanLength is to normalize <br>
//!          the polynomial in the cache to avoid bad conditioning <br>
//!          effects <br>
  Standard_EXPORT   static  void CacheD2(const Standard_Real U,const Standard_Integer Degree,const Standard_Real CacheParameter,const Standard_Real SpanLenght,const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& Weights,gp_Pnt& Point,gp_Vec& Vec1,gp_Vec& Vec2) ;
  //! Perform the evaluation of the Bspline Basis <br>
//!          and then multiplies by the weights <br>
//!          this just evaluates the current point <br>
//!          the parameter must be normalized between <br>
//!          the 0 and 1 for the span. <br>
//!          The Cache must be valid when calling this <br>
//!          routine. Geom Package will insure that. <br>
//!          and then multiplies by the weights <br>
//!          ththe CacheParameter is where the Cache was <br>
//!          constructed the SpanLength is to normalize <br>
//!          the polynomial in the cache to avoid bad conditioning <br>
//!          effectsis just evaluates the current point <br>
  Standard_EXPORT   static  void CacheD2(const Standard_Real U,const Standard_Integer Degree,const Standard_Real CacheParameter,const Standard_Real SpanLenght,const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& Weights,gp_Pnt2d& Point,gp_Vec2d& Vec1,gp_Vec2d& Vec2) ;
  //! Calls CacheD1 for Bezier  Curves Arrays computed with <br>
//!          the method PolesCoefficients. <br>
//!  Warning: To be used for Beziercurves ONLY!!! <br>
      static  void CoefsD2(const Standard_Real U,const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& Weights,gp_Pnt& Point,gp_Vec& Vec1,gp_Vec& Vec2) ;
  //! Calls CacheD1 for Bezier  Curves Arrays computed with <br>
//!          the method PolesCoefficients. <br>
//!  Warning: To be used for Beziercurves ONLY!!! <br>
      static  void CoefsD2(const Standard_Real U,const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& Weights,gp_Pnt2d& Point,gp_Vec2d& Vec1,gp_Vec2d& Vec2) ;
  //! Perform the evaluation of the of the cache <br>
//!          the parameter must be normalized between <br>
//!          the 0 and 1 for the span. <br>
//!          The Cache must be valid when calling this <br>
//!          routine. Geom Package will insure that. <br>
//!          and then multiplies by the weights <br>
//!          this just evaluates the current point <br>
//!          the CacheParameter is where the Cache was <br>
//!          constructed the SpanLength is to normalize <br>
//!          the polynomial in the cache to avoid bad conditioning <br>
//!          effects <br>
  Standard_EXPORT   static  void CacheD3(const Standard_Real U,const Standard_Integer Degree,const Standard_Real CacheParameter,const Standard_Real SpanLenght,const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& Weights,gp_Pnt& Point,gp_Vec& Vec1,gp_Vec& Vec2,gp_Vec& Vec3) ;
  //! Perform the evaluation of the Bspline Basis <br>
//!          and then multiplies by the weights <br>
//!          this just evaluates the current point <br>
//!          the parameter must be normalized between <br>
//!          the 0 and 1 for the span. <br>
//!          The Cache must be valid when calling this <br>
//!          routine. Geom Package will insure that. <br>
//!          and then multiplies by the weights <br>
//!          ththe CacheParameter is where the Cache was <br>
//!          constructed the SpanLength is to normalize <br>
//!          the polynomial in the cache to avoid bad conditioning <br>
//!          effectsis just evaluates the current point <br>
  Standard_EXPORT   static  void CacheD3(const Standard_Real U,const Standard_Integer Degree,const Standard_Real CacheParameter,const Standard_Real SpanLenght,const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& Weights,gp_Pnt2d& Point,gp_Vec2d& Vec1,gp_Vec2d& Vec2,gp_Vec2d& Vec3) ;
  //! Calls CacheD1 for Bezier  Curves Arrays computed with <br>
//!          the method PolesCoefficients. <br>
//!  Warning: To be used for Beziercurves ONLY!!! <br>
      static  void CoefsD3(const Standard_Real U,const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& Weights,gp_Pnt& Point,gp_Vec& Vec1,gp_Vec& Vec2,gp_Vec& Vec3) ;
  //! Calls CacheD1 for Bezier  Curves Arrays computed with <br>
//!          the method PolesCoefficients. <br>
//!  Warning: To be used for Beziercurves ONLY!!! <br>
      static  void CoefsD3(const Standard_Real U,const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& Weights,gp_Pnt2d& Point,gp_Vec2d& Vec1,gp_Vec2d& Vec2,gp_Vec2d& Vec3) ;
  //! Perform the evaluation of the Taylor expansion <br>
//!          of the Bspline normalized between 0 and 1. <br>
//!          If rational computes the homogeneous Taylor expension <br>
//!          for the numerator and stores it in CachePoles <br>
  Standard_EXPORT   static  void BuildCache(const Standard_Real U,const Standard_Real InverseOfSpanDomain,const Standard_Boolean PeriodicFlag,const Standard_Integer Degree,const TColStd_Array1OfReal& FlatKnots,const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& Weights,TColgp_Array1OfPnt& CachePoles,TColStd_Array1OfReal& CacheWeights) ;
  //! Perform the evaluation of the Taylor expansion <br>
//!          of the Bspline normalized between 0 and 1. <br>
//!          If rational computes the homogeneous Taylor expension <br>
//!          for the numerator and stores it in CachePoles <br>
  Standard_EXPORT   static  void BuildCache(const Standard_Real U,const Standard_Real InverseOfSpanDomain,const Standard_Boolean PeriodicFlag,const Standard_Integer Degree,const TColStd_Array1OfReal& FlatKnots,const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& Weights,TColgp_Array1OfPnt2d& CachePoles,TColStd_Array1OfReal& CacheWeights) ;
  
      static  void PolesCoefficients(const TColgp_Array1OfPnt2d& Poles,TColgp_Array1OfPnt2d& CachePoles) ;
  
  Standard_EXPORT   static  void PolesCoefficients(const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& Weights,TColgp_Array1OfPnt2d& CachePoles,TColStd_Array1OfReal& CacheWeights) ;
  
      static  void PolesCoefficients(const TColgp_Array1OfPnt& Poles,TColgp_Array1OfPnt& CachePoles) ;
  //! Encapsulation   of  BuildCache    to   perform   the <br>
//!          evaluation  of the Taylor expansion for beziercurves <br>
//!          at parameter 0. <br>
//!  Warning: To be used for Beziercurves ONLY!!! <br>
  Standard_EXPORT   static  void PolesCoefficients(const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& Weights,TColgp_Array1OfPnt& CachePoles,TColStd_Array1OfReal& CacheWeights) ;
  //! Returns pointer to statically allocated array representing <br>
//!          flat knots for bezier curve of the specified degree. <br>
//!          Raises OutOfRange if Degree > MaxDegree() <br>
  Standard_EXPORT   static const Standard_Real& FlatBezierKnots(const Standard_Integer Degree) ;
  //! builds the Schoenberg points from the flat knot <br>
//!          used to interpolate a BSpline since the <br>
//!          BSpline matrix is invertible. <br>
  Standard_EXPORT   static  void BuildSchoenbergPoints(const Standard_Integer Degree,const TColStd_Array1OfReal& FlatKnots,TColStd_Array1OfReal& Parameters) ;
  //! Performs the interpolation of  the data given in <br>
//!          the Poles  array  according  to the  requests in <br>
//!          ContactOrderArray    that is      :           if <br>
//!          ContactOrderArray(i) has value  d it means  that <br>
//!          Poles(i)   containes the dth  derivative of  the <br>
//!          function to be interpolated. The length L of the <br>
//!          following arrays must be the same : <br>
//!          Parameters, ContactOrderArray, Poles, <br>
//!          The length of FlatKnots is Degree + L + 1 <br>
//!  Warning: <br>
//!          the method used to do that interpolation is <br>
//!          gauss elimination WITHOUT pivoting. Thus if the <br>
//!          diagonal is not dominant there is no guarantee <br>
//!          that the algorithm will work. Nevertheless for <br>
//!          Cubic interpolation or interpolation at Scheonberg <br>
//!          points the method will work <br>
//!          The InversionProblem will report 0 if there was no <br>
//!          problem else it will give the index of the faulty <br>
//!          pivot <br>
  Standard_EXPORT   static  void Interpolate(const Standard_Integer Degree,const TColStd_Array1OfReal& FlatKnots,const TColStd_Array1OfReal& Parameters,const TColStd_Array1OfInteger& ContactOrderArray,TColgp_Array1OfPnt& Poles,Standard_Integer& InversionProblem) ;
  //! Performs the interpolation of  the data given in <br>
//!          the Poles  array  according  to the  requests in <br>
//!          ContactOrderArray    that is      :           if <br>
//!          ContactOrderArray(i) has value  d it means  that <br>
//!          Poles(i)   containes the dth  derivative of  the <br>
//!          function to be interpolated. The length L of the <br>
//!          following arrays must be the same : <br>
//!          Parameters, ContactOrderArray, Poles, <br>
//!          The length of FlatKnots is Degree + L + 1 <br>
//!  Warning: <br>
//!          the method used to do that interpolation is <br>
//!          gauss elimination WITHOUT pivoting. Thus if the <br>
//!          diagonal is not dominant there is no guarantee <br>
//!          that the algorithm will work. Nevertheless for <br>
//!          Cubic interpolation at knots or interpolation at Scheonberg <br>
//!          points the method will work. <br>
//!          The InversionProblem w <br>
//!          ll report 0 if there was no <br>
//!          problem else it will give the index of the faulty <br>
//!          pivot <br>
  Standard_EXPORT   static  void Interpolate(const Standard_Integer Degree,const TColStd_Array1OfReal& FlatKnots,const TColStd_Array1OfReal& Parameters,const TColStd_Array1OfInteger& ContactOrderArray,TColgp_Array1OfPnt2d& Poles,Standard_Integer& InversionProblem) ;
  //! Performs the interpolation of  the data given in <br>
//!          the Poles  array  according  to the  requests in <br>
//!          ContactOrderArray    that is      :           if <br>
//!          ContactOrderArray(i) has value  d it means  that <br>
//!          Poles(i)   containes the dth  derivative of  the <br>
//!          function to be interpolated. The length L of the <br>
//!          following arrays must be the same : <br>
//!          Parameters, ContactOrderArray, Poles, <br>
//!          The length of FlatKnots is Degree + L + 1 <br>
//!  Warning: <br>
//!          the method used to do that interpolation is <br>
//!          gauss elimination WITHOUT pivoting. Thus if the <br>
//!          diagonal is not dominant there is no guarantee <br>
//!          that the algorithm will work. Nevertheless for <br>
//!          Cubic interpolation at knots or interpolation at Scheonberg <br>
//!          points the method will work. <br>
//!          The InversionProblem will report 0 if there was no <br>
//!          problem else it will give the index of the faulty <br>
//!          pivot <br>
//! <br>
//! <br>
  Standard_EXPORT   static  void Interpolate(const Standard_Integer Degree,const TColStd_Array1OfReal& FlatKnots,const TColStd_Array1OfReal& Parameters,const TColStd_Array1OfInteger& ContactOrderArray,TColgp_Array1OfPnt& Poles,TColStd_Array1OfReal& Weights,Standard_Integer& InversionProblem) ;
  //! Performs the interpolation of  the data given in <br>
//!          the Poles  array  according  to the  requests in <br>
//!          ContactOrderArray    that is      :           if <br>
//!          ContactOrderArray(i) has value  d it means  that <br>
//!          Poles(i)   containes the dth  derivative of  the <br>
//!          function to be interpolated. The length L of the <br>
//!          following arrays must be the same : <br>
//!          Parameters, ContactOrderArray, Poles, <br>
//!          The length of FlatKnots is Degree + L + 1 <br>
//!  Warning: <br>
//!          the method used to do that interpolation is <br>
//!          gauss elimination WITHOUT pivoting. Thus if the <br>
//!          diagonal is not dominant there is no guarantee <br>
//!          that the algorithm will work. Nevertheless for <br>
//!          Cubic interpolation at knots or interpolation at Scheonberg <br>
//!          points the method will work. <br>
//!          The InversionProblem w <br>
//!          ll report 0 if there was no <br>
//!          problem else it will give the i <br>
  Standard_EXPORT   static  void Interpolate(const Standard_Integer Degree,const TColStd_Array1OfReal& FlatKnots,const TColStd_Array1OfReal& Parameters,const TColStd_Array1OfInteger& ContactOrderArray,TColgp_Array1OfPnt2d& Poles,TColStd_Array1OfReal& Weights,Standard_Integer& InversionProblem) ;
  //! Performs the interpolation of  the data given in <br>
//!          the Poles  array  according  to the  requests in <br>
//!          ContactOrderArray    that is      :           if <br>
//!          ContactOrderArray(i) has value  d it means  that <br>
//!          Poles(i)   containes the dth  derivative of  the <br>
//!          function to be interpolated. The length L of the <br>
//!          following arrays must be the same : <br>
//!          Parameters, ContactOrderArray <br>
//!          The length of FlatKnots is Degree + L + 1 <br>
//!          The  PolesArray   is    an   seen   as    an <br>
//!          Array[1..N][1..ArrayDimension] with N = tge length <br>
//!          of the parameters array <br>
//!  Warning: <br>
//!          the method used to do that interpolation is <br>
//!          gauss elimination WITHOUT pivoting. Thus if the <br>
//!          diagonal is not dominant there is no guarantee <br>
//!          that the algorithm will work. Nevertheless for <br>
//!          Cubic interpolation or interpolation at Scheonberg <br>
//!          points the method will work <br>
//!          The InversionProblem will report 0 if there was no <br>
//!          problem else it will give the index of the faulty <br>
//!          pivot <br>
//! <br>
  Standard_EXPORT   static  void Interpolate(const Standard_Integer Degree,const TColStd_Array1OfReal& FlatKnots,const TColStd_Array1OfReal& Parameters,const TColStd_Array1OfInteger& ContactOrderArray,const Standard_Integer ArrayDimension,Standard_Real& Poles,Standard_Integer& InversionProblem) ;
  
  Standard_EXPORT   static  void Interpolate(const Standard_Integer Degree,const TColStd_Array1OfReal& FlatKnots,const TColStd_Array1OfReal& Parameters,const TColStd_Array1OfInteger& ContactOrderArray,const Standard_Integer ArrayDimension,Standard_Real& Poles,Standard_Real& Weights,Standard_Integer& InversionProblem) ;
  //! Find the new poles which allows  an old point (with a <br>
//!          given  u as parameter) to reach a new position <br>
//!          Index1 and Index2 indicate the range of poles we can move <br>
//!          (1, NbPoles-1) or (2, NbPoles) -> no constraint for one side <br>
//!                     don't enter (1,NbPoles) -> error: rigid move <br>
//!          (2, NbPoles-1) -> the ends are enforced <br>
//!          (3, NbPoles-2) -> the ends and the tangency are enforced <br>
//!    if Problem in BSplineBasis calculation, no change for the curve <br>
//!    and FirstIndex, LastIndex = 0 <br>
  Standard_EXPORT   static  void MovePoint(const Standard_Real U,const gp_Vec2d& Displ,const Standard_Integer Index1,const Standard_Integer Index2,const Standard_Integer Degree,const Standard_Boolean Rational,const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& Weights,const TColStd_Array1OfReal& FlatKnots,Standard_Integer& FirstIndex,Standard_Integer& LastIndex,TColgp_Array1OfPnt2d& NewPoles) ;
  //! Find the new poles which allows  an old point (with a <br>
//!          given  u as parameter) to reach a new position <br>
//!          Index1 and Index2 indicate the range of poles we can move <br>
//!          (1, NbPoles-1) or (2, NbPoles) -> no constraint for one side <br>
//!                     don't enter (1,NbPoles) -> error: rigid move <br>
//!          (2, NbPoles-1) -> the ends are enforced <br>
//!          (3, NbPoles-2) -> the ends and the tangency are enforced <br>
//!    if Problem in BSplineBasis calculation, no change for the curve <br>
//!    and FirstIndex, LastIndex = 0 <br>
  Standard_EXPORT   static  void MovePoint(const Standard_Real U,const gp_Vec& Displ,const Standard_Integer Index1,const Standard_Integer Index2,const Standard_Integer Degree,const Standard_Boolean Rational,const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& Weights,const TColStd_Array1OfReal& FlatKnots,Standard_Integer& FirstIndex,Standard_Integer& LastIndex,TColgp_Array1OfPnt& NewPoles) ;
  //! This is the dimension free version of the utility <br>
//! U is the parameter  must be within the  first FlatKnots and the <br>
//! last FlatKnots  Delta is the amount the  curve has  to be moved <br>
//! DeltaDerivative is the  amount the derivative  has to be moved. <br>
//! Delta  and   DeltaDerivative   must be    array   of  dimension <br>
//! ArrayDimension  Degree  is the degree  of   the BSpline and the <br>
//! FlatKnots are the knots of the BSpline  Starting Condition if = <br>
//! -1 means the starting point of the curve can move <br>
//! = 0 means the <br>
//! starting  point  of the cuve  cannot  move but  tangen  starting <br>
//! point of the curve cannot move <br>
//! = 1 means the starting point and tangents cannot move <br>
//! = 2 means the starting point tangent and curvature cannot move <br>
//! = ... <br>
//! Same holds for EndingCondition <br>
//! Poles are the poles of the curve <br>
//! Weights are the weights of the curve if Rational = Standard_True <br>
//! NewPoles are the poles of the deformed curve <br>
//! ErrorStatus will be 0 if no error happened <br>
//!                     1 if there are not enough knots/poles <br>
//!                          the imposed conditions <br>
//! The way to solve this problem is to add knots to the BSpline <br>
//! If StartCondition = 1 and EndCondition = 1 then you need at least <br>
//! 4 + 2 = 6 poles so for example to have a C1 cubic you will need <br>
//! have at least 2 internal knots. <br>
  Standard_EXPORT   static  void MovePointAndTangent(const Standard_Real U,const Standard_Integer ArrayDimension,Standard_Real& Delta,Standard_Real& DeltaDerivative,const Standard_Real Tolerance,const Standard_Integer Degree,const Standard_Boolean Rational,const Standard_Integer StartingCondition,const Standard_Integer EndingCondition,Standard_Real& Poles,const TColStd_Array1OfReal& Weights,const TColStd_Array1OfReal& FlatKnots,Standard_Real& NewPoles,Standard_Integer& ErrorStatus) ;
  //! This is the dimension free version of the utility <br>
//! U is the parameter  must be within the  first FlatKnots and the <br>
//! last FlatKnots  Delta is the amount the  curve has  to be moved <br>
//! DeltaDerivative is the  amount the derivative  has to be moved. <br>
//! Delta  and   DeltaDerivative   must be    array   of  dimension <br>
//! ArrayDimension  Degree  is the degree  of   the BSpline and the <br>
//! FlatKnots are the knots of the BSpline  Starting Condition if = <br>
//! -1 means the starting point of the curve can move <br>
//! = 0 means the <br>
//! starting  point  of the cuve  cannot  move but  tangen  starting <br>
//! point of the curve cannot move <br>
//! = 1 means the starting point and tangents cannot move <br>
//! = 2 means the starting point tangent and curvature cannot move <br>
//! = ... <br>
//! Same holds for EndingCondition <br>
//! Poles are the poles of the curve <br>
//! Weights are the weights of the curve if Rational = Standard_True <br>
//! NewPoles are the poles of the deformed curve <br>
//! ErrorStatus will be 0 if no error happened <br>
//!                     1 if there are not enough knots/poles <br>
//!                          the imposed conditions <br>
//! The way to solve this problem is to add knots to the BSpline <br>
//! If StartCondition = 1 and EndCondition = 1 then you need at least <br>
//! 4 + 2 = 6 poles so for example to have a C1 cubic you will need <br>
//! have at least 2 internal knots. <br>
  Standard_EXPORT   static  void MovePointAndTangent(const Standard_Real U,const gp_Vec& Delta,const gp_Vec& DeltaDerivative,const Standard_Real Tolerance,const Standard_Integer Degree,const Standard_Boolean Rational,const Standard_Integer StartingCondition,const Standard_Integer EndingCondition,const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& Weights,const TColStd_Array1OfReal& FlatKnots,TColgp_Array1OfPnt& NewPoles,Standard_Integer& ErrorStatus) ;
  //! This is the dimension free version of the utility <br>
//! U is the parameter  must be within the  first FlatKnots and the <br>
//! last FlatKnots  Delta is the amount the  curve has  to be moved <br>
//! DeltaDerivative is the  amount the derivative  has to be moved. <br>
//! Delta  and   DeltaDerivative   must be    array   of  dimension <br>
//! ArrayDimension  Degree  is the degree  of   the BSpline and the <br>
//! FlatKnots are the knots of the BSpline  Starting Condition if = <br>
//! -1 means the starting point of the curve can move <br>
//! = 0 means the <br>
//! starting  point  of the cuve  cannot  move but  tangen  starting <br>
//! point of the curve cannot move <br>
//! = 1 means the starting point and tangents cannot move <br>
//! = 2 means the starting point tangent and curvature cannot move <br>
//! = ... <br>
//! Same holds for EndingCondition <br>
//! Poles are the poles of the curve <br>
//! Weights are the weights of the curve if Rational = Standard_True <br>
//! NewPoles are the poles of the deformed curve <br>
//! ErrorStatus will be 0 if no error happened <br>
//!                     1 if there are not enough knots/poles <br>
//!                          the imposed conditions <br>
//! The way to solve this problem is to add knots to the BSpline <br>
//! If StartCondition = 1 and EndCondition = 1 then you need at least <br>
//! 4 + 2 = 6 poles so for example to have a C1 cubic you will need <br>
//! have at least 2 internal knots. <br>
  Standard_EXPORT   static  void MovePointAndTangent(const Standard_Real U,const gp_Vec2d& Delta,const gp_Vec2d& DeltaDerivative,const Standard_Real Tolerance,const Standard_Integer Degree,const Standard_Boolean Rational,const Standard_Integer StartingCondition,const Standard_Integer EndingCondition,const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& Weights,const TColStd_Array1OfReal& FlatKnots,TColgp_Array1OfPnt2d& NewPoles,Standard_Integer& ErrorStatus) ;
  
//!          given a tolerance in 3D space returns a <br>
//!          tolerance    in U parameter space such that <br>
//!          all u1 and u0 in the domain of the curve f(u) <br>
//!          | u1 - u0 | < UTolerance and <br>
//!          we have |f (u1) - f (u0)| < Tolerance3D <br>
  Standard_EXPORT   static  void Resolution(Standard_Real& PolesArray,const Standard_Integer ArrayDimension,const Standard_Integer NumPoles,const TColStd_Array1OfReal& Weights,const TColStd_Array1OfReal& FlatKnots,const Standard_Integer Degree,const Standard_Real Tolerance3D,Standard_Real& UTolerance) ;
  
//!          given a tolerance in 3D space returns a <br>
//!          tolerance    in U parameter space such that <br>
//!          all u1 and u0 in the domain of the curve f(u) <br>
//!          | u1 - u0 | < UTolerance and <br>
//!          we have |f (u1) - f (u0)| < Tolerance3D <br>
  Standard_EXPORT   static  void Resolution(const TColgp_Array1OfPnt& Poles,const TColStd_Array1OfReal& Weights,const Standard_Integer NumPoles,const TColStd_Array1OfReal& FlatKnots,const Standard_Integer Degree,const Standard_Real Tolerance3D,Standard_Real& UTolerance) ;
  
//!          given a tolerance in 3D space returns a <br>
//!          tolerance    in U parameter space such that <br>
//!          all u1 and u0 in the domain of the curve f(u) <br>
//!          | u1 - u0 | < UTolerance and <br>
//!          we have |f (u1) - f (u0)| < Tolerance3D <br>
  Standard_EXPORT   static  void Resolution(const TColgp_Array1OfPnt2d& Poles,const TColStd_Array1OfReal& Weights,const Standard_Integer NumPoles,const TColStd_Array1OfReal& FlatKnots,const Standard_Integer Degree,const Standard_Real Tolerance3D,Standard_Real& UTolerance) ;





protected:





private:

  
  Standard_EXPORT   static  void LocateParameter(const TColStd_Array1OfReal& Knots,const Standard_Real U,const Standard_Boolean Periodic,const Standard_Integer K1,const Standard_Integer K2,Standard_Integer& Index,Standard_Real& NewU,const Standard_Real Uf,const Standard_Real Ue) ;




};


#include <BSplCLib.lxx>



// other Inline functions and methods (like "C++: function call" methods)


#endif