/usr/src/castle-game-engine-5.2.0/3d/castlequaternions.pas is in castle-game-engine-src 5.2.0-3.
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 | {
Copyright 2003-2014 Michalis Kamburelis.
This file is part of "Castle Game Engine".
"Castle Game Engine" is free software; see the file COPYING.txt,
included in this distribution, for details about the copyright.
"Castle Game Engine" is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
----------------------------------------------------------------------------
}
{ Quaternions (in particular using them to express 3D rotations).
@noAutoLinkHere }
unit CastleQuaternions;
interface
uses CastleVectors;
type
TQuaternion = object
Data: packed record
case Integer of
0: (Vector: TVector3Single;
Real: Single);
1: ({ Alternative, sometimes comfortable, view on quaternion as one
4-element vector. }
Vector4: TVector4Single);
end;
{ Calculate axis (will be normalized) and angle (will be in radians)
of rotation encoded in unit quaternion Q.
This is the reverse of QuatFromAxisAngle. }
procedure ToAxisAngle(out Axis: TVector3Single; out AngleRad: Single);
{ Convert quaternion to a rotation axis and angle encoded in 4D vector.
Axis is normalized if quaternion was also normalized
(which is true if working with rotation quaternions).
Angle is in radians. }
function ToAxisAngle: TVector4Single;
{ Calculate matrix doing rotation described by unit quaternion. }
function ToRotationMatrix: TMatrix4Single;
{ Rotate by unit quaternion.
You can pass here TVector4Single, which is then understood to be a 3D
position in homogeneous coordinates.
@groupBegin }
function Rotate(const Point: TVector4Single): TVector4Single; overload;
function Rotate(const Point: TVector3Single): TVector3Single; overload;
{ @groupEnd }
{ Quaternion conjugation. This is just a fancy name for negating Q.Vector.
@groupBegin }
function Conjugate: TQuaternion;
procedure ConjugateMe;
{ @groupEnd }
procedure Normalize;
{ Perform normalization, only if the quaternion is detected to be
"significantly unnormalized". It checks if the quaternion needs
normalization using fast VectorLenSqr, that is quaternion length
is not needed for the check (sqrt not needed). Only if it's significantly
different that 1.0, sqrt is done and quaternion is normalized.
This may be useful if you fear of eventual
errors because of floating-point error cumulations, e.g. when you
repeatedly multiply one quaternion by another, and yet another, and yet
another etc. Calling this will trigger normalization from time to time
(although will tolerate very small, epsilon-like, differences that are
normal). Thus it prevents the quaternion from getting "too unnormalized".
Generally, this is not needed, as quaternions are nicely numerically
stable (which means that quaternion "very slightly unnormalized" will
only generate "very slightly wrong" results, so it's not that bad).
And no, I didn't actually observe the need for this in my programs.
But you can see it actually called when you use TMatrixExaminer and
deliberately cause spinning by very very large value (e.g. run
view3dscene and press and hold right key, this will cause model
spinning very fast, which causes quat multiplication every frame).
So possibly this would trigger incorrect quaternions at some point.
Anyway, this remains mostly a paranoid correctness measure. }
procedure LazyNormalize;
end;
const
QuatIdentityRot: TQuaternion = (Data: (Vector: (0, 0, 0); Real: 1));
{ Calculate unit quaternion representing rotation around Axis
by AngleRad angle (in radians).
Axis must be normalized, or you have to pass NormalizeAxis = true
(then we'll normalize it ourselves inside). Otherwise you will
get non-normalized quaternion that doesn't represent rotation,
and is usually useless for us. }
function QuatFromAxisAngle(const Axis: TVector3Single;
const AngleRad: Single; const NormalizeAxis: boolean = false): TQuaternion;
function QuatFromAxisAngle(const AxisAngle: TVector4Single;
const NormalizeAxis: boolean = false): TQuaternion;
{ Multiply two quaternions.
Geometric interpretation: If these are unit quaternions representing
rotations, multiplying them calculates one rotation that has the same
effect as rotating by Q2 and then by Q1.
Normal of result is equal to norm of Q1 * norm of Q2 (in particular,
multiplying unit quaternions (used for rotations) yields another unit
quaternion for sure). }
operator* (const Q1, Q2: TQuaternion): TQuaternion;
{ Interpolate between two rotations, along the shortest path on the unit sphere,
with constant speed.
The overloaded version that works with TVector4Single takes
a rotation (not a quaternion) expressed as an axis
(first 3 elements) and angle (in radians, 4th element).
Axis does not have to be normalized (we'll normalize it).
This is nice e.g. to interpolate VRML/X3D rotations.
@groupBegin }
function SLerp(const A: Single; const Q1, Q2: TQuaternion): TQuaternion;
function SLerp(const A: Single; const Rot1, Rot2: TVector4Single): TVector4Single;
{ @groupEnd }
{ Interpolate between two rotations, along the straightest path on the unit sphere.
This is faster than SLerp, but does not guarantee the interpolated
result travels with constant speed.
Often it's not a noticeable / important problem (see
http://number-none.com/product/Understanding%20Slerp,%20Then%20Not%20Using%20It/)
When ForceShortestPath = @false, this doesn't guarantee choosing
the shortest path. Although it goes through the @italic(straightest) path,
there are two such paths, it may go through the shorter or longer one.
Use ForceShortestPath = @true if you want to interpolate through the
shortest.
The overloaded version that works with TVector4Single takes
a rotation (not a quaternion) expressed as an axis
(first 3 elements) and angle (in radians, 4th element).
Axis does not have to be normalized (we'll normalize it).
This is nice e.g. to interpolate VRML/X3D rotations.
@groupBegin }
function NLerp(const A: Single; const Q1, Q2: TQuaternion;
const ForceShortestPath: boolean = true): TQuaternion;
function NLerp(const A: Single; const Rot1, Rot2: TVector4Single;
const ForceShortestPath: boolean = true): TVector4Single;
{ @groupEnd }
implementation
uses Math, CastleUtils;
{ TQuaternion ---------------------------------------------------------------- }
procedure TQuaternion.ToAxisAngle(out Axis: TVector3Single;
out AngleRad: Single);
{ Data is a normalized quaternion, so
Data.Vector = Sin(AngleRad / 2) * Axis
Data.Real = Cos(AngleRad / 2)
}
var
HalfAngle, SinHalfAngle: Single;
begin
{ Use Clamped to secure against cosinus being slightly outside [-1,1], like in SLerp. }
HalfAngle := ArcCos(Clamped(Data.Real, -1, 1));
SinHalfAngle := Sin(HalfAngle);
AngleRad := HalfAngle * 2;
if Zero(SinHalfAngle) then
begin
{ Then Data.Vector must be zero also... How could this happen?
SinHalfAngle = 0 means that HalfAngle = Pi * K (e.g. 0).
So Angle = 2 * Pi * K so there's no rotation actually happening...
Which means that any Axis is Ok (but return anything normalized,
to keep assertion that returned Axis is normalized). }
Axis := Vector3Single(0, 0, 1);
end else
Axis := VectorScale(Data.Vector, 1 / SinHalfAngle);
end;
function TQuaternion.ToAxisAngle: TVector4Single;
var
Axis: TVector3Single absolute Result;
begin
ToAxisAngle(Axis, Result[3]);
end;
function TQuaternion.Conjugate: TQuaternion;
begin
Result.Data.Vector := VectorNegate(Data.Vector);
Result.Data.Real := Data.Real;
end;
procedure TQuaternion.ConjugateMe;
begin
VectorNegateTo1st(Data.Vector);
end;
function TQuaternion.Rotate(const Point: TVector4Single): TVector4Single;
begin
Result := (Self * TQuaternion(Point)).Data.Vector4;
Result := (TQuaternion(Result) * Conjugate).Data.Vector4;
end;
function TQuaternion.Rotate(const Point: TVector3Single): TVector3Single;
var
P4: TVector4Single;
begin
P4 := Vector4Single(Point, 0);
P4 := Rotate(P4);
Result := Vector3SingleCut(P4);
end;
function QuatToRotationMatrix(const X, Y, Z, W: Single): TMatrix4Single;
var
XX, YY, ZZ: Single;
begin
XX := Sqr(X);
YY := Sqr(Y);
ZZ := Sqr(Z);
{ row 0 }
Result[0, 0] := 1 - 2 * (YY + ZZ);
Result[1, 0] := 2 * ( X * Y - W * Z );
Result[2, 0] := 2 * ( X * Z + W * Y );
Result[3, 0] := 0;
{ row 1 }
Result[0, 1] := 2 * ( X * Y + W * Z );
Result[1, 1] := 1 - 2 * (XX + ZZ);
Result[2, 1] := 2 * ( Y * Z - W * X );
Result[3, 1] := 0;
{ row 2 }
Result[0, 2] := 2 * ( X * Z - W * Y );
Result[1, 2] := 2 * ( Y * Z + W * X );
Result[2, 2] := 1 - 2 * (XX + YY);
Result[3, 2] := 0;
{ row 3 - like in identity matrix,
only 3x3 matrix is interesting in rotations. }
Result[0, 3] := 0;
Result[1, 3] := 0;
Result[2, 3] := 0;
Result[3, 3] := 1;
end;
function TQuaternion.ToRotationMatrix: TMatrix4Single;
begin
Result := QuatToRotationMatrix(Data.Vector[0], Data.Vector[1], Data.Vector[2], Data.Real);
end;
procedure TQuaternion.Normalize;
var
Len: Single;
begin
Len := VectorLen(Data.Vector4);
if Len <> 0 then
begin
Len := 1/Len;
Data.Vector[0] *= Len;
Data.Vector[1] *= Len;
Data.Vector[2] *= Len;
Data.Real *= Len;
end;
end;
procedure TQuaternion.LazyNormalize;
var
Len: Single;
begin
Len := VectorLenSqr(Data.Vector4);
if (Len - 1) > 0.001 then
begin
{ tests: Writeln('quat lazily normed'); }
Len := Sqrt(Len);
if Len <> 0 then
begin
Len := 1/Len;
Data.Vector[0] *= Len;
Data.Vector[1] *= Len;
Data.Vector[2] *= Len;
Data.Real *= Len;
end;
end;
end;
{ routines ------------------------------------------------------------------- }
function QuatFromAxisAngle(const Axis: TVector3Single;
const AngleRad: Single; const NormalizeAxis: boolean): TQuaternion;
var
SinHalfAngle, CosHalfAngle: Float;
begin
{ The quaternion requires half angles. }
SinCos(AngleRad / 2, SinHalfAngle, CosHalfAngle);
if NormalizeAxis then
SinHalfAngle /= VectorLen(Axis);
Result.Data.Vector := VectorScale(Axis, SinHalfAngle);
Result.Data.Real := CosHalfAngle;
end;
function QuatFromAxisAngle(const AxisAngle: TVector4Single;
const NormalizeAxis: boolean): TQuaternion;
var
Axis: TVector3Single absolute AxisAngle;
begin
Result := QuatFromAxisAngle(Axis, AxisAngle[3], NormalizeAxis);
end;
operator* (const Q1, Q2: TQuaternion): TQuaternion;
begin
Result.Data.Vector := VectorProduct(Q1.Data.Vector, Q2.Data.Vector);
VectorAddTo1st(Result.Data.Vector, VectorScale(Q1.Data.Vector, Q2.Data.Real));
VectorAddTo1st(Result.Data.Vector, VectorScale(Q2.Data.Vector, Q1.Data.Real));
Result.Data.Real := Q1.Data.Real * Q2.Data.Real - VectorDotProduct(Q1.Data.Vector, Q2.Data.Vector);
end;
{ For SLerp and NLerp implementations, see
http://www.3dkingdoms.com/weekly/weekly.php?a=36
http://www.3dkingdoms.com/weekly/quat.h
http://number-none.com/product/Understanding%20Slerp,%20Then%20Not%20Using%20It/
http://en.wikipedia.org/wiki/Slerp
}
function SLerp(const A: Single; const Q1, Q2: TQuaternion): TQuaternion;
var
W1, W2, NegateOneQuaternion: Single;
CosTheta, Theta: Float;
SinTheta: Single;
begin
CosTheta := VectorDotProduct(Q1.Data.Vector4, Q2.Data.Vector4);
{ Following wikipedia:
Long paths can be prevented by negating one end if the dot product,
CosTheta, is negative. See also
http://www.euclideanspace.com/maths/algebra/realNormedAlgebra/quaternions/slerp/index.htm
http://www.shatters.net/forum/viewtopic.php?f=4&t=10955&p=86203
We do it by NegateOneQuaternion, either 1 (not negate) or -1 (negate).
Later W1 will be multiplied by this.
This way the actual negation of Q1 happens when it's multiplied by W1,
so at ~zero cost. }
if CosTheta < 0 then
begin
NegateOneQuaternion := -1;
CosTheta := -CosTheta;
end else
NegateOneQuaternion := 1;
{ Sometimes CosTheta may get slightly > 1, and then ArcCos fails with
EInvalidArgument. Testcase: demo_models/x3d/orientation_cos_1.x3d
with view3dscene. }
MinTo1st(CosTheta, 1);
Theta := ArcCos(CosTheta);
SinTheta := Sin(Theta);
if SinTheta > 0.001 then
begin
W1 := NegateOneQuaternion * Sin( (1-A) * Theta ) / SinTheta;
W2 := Sin( A * Theta ) / SinTheta;
end else
begin
{ Theta ~= 0, so both rotations equal (or opposite, in which case
result in undefined anyway). }
W1 := 1 - A;
W2 := A;
end;
Result.Data.Vector4 := VectorAdd(VectorScale(Q1.Data.Vector4, W1), VectorScale(Q2.Data.Vector4, W2));
end;
function SLerp(const A: Single; const Rot1, Rot2: TVector4Single): TVector4Single;
begin
Result := SLerp(A,
QuatFromAxisAngle(Rot1, true),
QuatFromAxisAngle(Rot2, true)).ToAxisAngle;
end;
function NLerp(const A: Single; const Q1, Q2: TQuaternion;
const ForceShortestPath: boolean): TQuaternion;
begin
if ForceShortestPath and (VectorDotProduct(Q1.Data.Vector4, Q2.Data.Vector4) < 0) then
begin
{ negate one quaternion }
Result.Data.Vector4 := Lerp(A, VectorNegate(Q1.Data.Vector4), Q2.Data.Vector4);
end else
Result.Data.Vector4 := Lerp(A, Q1.Data.Vector4, Q2.Data.Vector4);
Result.Normalize;
end;
function NLerp(const A: Single; const Rot1, Rot2: TVector4Single;
const ForceShortestPath: boolean): TVector4Single;
begin
Result := NLerp(A,
QuatFromAxisAngle(Rot1, true),
QuatFromAxisAngle(Rot2, true),
ForceShortestPath).ToAxisAngle;
end;
end.
|