| /* |
| * Copyright (C) 2005, 2006 Apple Computer, Inc. All rights reserved. |
| * |
| * Redistribution and use in source and binary forms, with or without |
| * modification, are permitted provided that the following conditions |
| * are met: |
| * 1. Redistributions of source code must retain the above copyright |
| * notice, this list of conditions and the following disclaimer. |
| * 2. Redistributions in binary form must reproduce the above copyright |
| * notice, this list of conditions and the following disclaimer in the |
| * documentation and/or other materials provided with the distribution. |
| * |
| * THIS SOFTWARE IS PROVIDED BY APPLE COMPUTER, INC. ``AS IS'' AND ANY |
| * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
| * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR |
| * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL APPLE COMPUTER, INC. OR |
| * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, |
| * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, |
| * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR |
| * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY |
| * OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT |
| * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE |
| * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
| */ |
| |
| #ifndef THIRD_PARTY_BLINK_RENDERER_PLATFORM_TRANSFORMS_TRANSFORMATION_MATRIX_H_ |
| #define THIRD_PARTY_BLINK_RENDERER_PLATFORM_TRANSFORMS_TRANSFORMATION_MATRIX_H_ |
| |
| #include <string.h> // for memcpy |
| |
| #include <cmath> |
| #include <limits> |
| #include <memory> |
| |
| #include "build/build_config.h" |
| #include "third_party/blink/renderer/platform/geometry/float_point.h" |
| #include "third_party/blink/renderer/platform/geometry/float_point_3d.h" |
| #include "third_party/blink/renderer/platform/wtf/allocator/allocator.h" |
| #include "third_party/skia/include/core/SkMatrix44.h" |
| |
| namespace gfx { |
| class Transform; |
| } |
| |
| namespace blink { |
| |
| class AffineTransform; |
| class IntRect; |
| class LayoutRect; |
| class FloatRect; |
| class FloatQuad; |
| class FloatBox; |
| class JSONArray; |
| struct Rotation; |
| #if defined(ARCH_CPU_X86_64) |
| #define TRANSFORMATION_MATRIX_USE_X86_64_SSE2 |
| #define ALIGNAS_TRANSFORMATION_MATRIX alignas(16) |
| #else |
| #define ALIGNAS_TRANSFORMATION_MATRIX |
| #endif |
| |
| class PLATFORM_EXPORT TransformationMatrix { |
| // TransformationMatrix must not be allocated on Oilpan's heap since |
| // Oilpan doesn't (yet) have an ability to allocate the TransformationMatrix |
| // with 16-byte alignment. PartitionAlloc has the ability. |
| USING_FAST_MALLOC(TransformationMatrix); |
| |
| public: |
| // Throughout this class, we will be speaking in column vector convention. |
| // i.e. Applying a transform T to point P is T * P. |
| // The elements of the matrix and the vector looks like: |
| // | scale_x skew_y_x skew_z_x translate_x | | x | |
| // | skew_x_y scale_y skew_z_y translate_y | * | y | |
| // | skew_x_z skew_y_z scale_z translate_z | | z | |
| // | persp_x persp_y persp_z persp_w | | w | |
| // Internally the matrix is stored as a 2-dimensional array in col-major |
| // order. In other words, this is the layout of the matrix: |
| // | matrix_[0][0] matrix_[1][0] matrix_[2][0] matrix_[3][0] | |
| // | matrix_[0][1] matrix_[1][1] matrix_[2][1] matrix_[3][1] | |
| // | matrix_[0][2] matrix_[1][2] matrix_[2][2] matrix_[3][2] | |
| // | matrix_[0][3] matrix_[1][3] matrix_[2][3] matrix_[3][3] | |
| struct ALIGNAS_TRANSFORMATION_MATRIX Matrix4 { |
| using Column = double[4]; |
| Column& operator[](size_t i) { return columns[i]; } |
| const Column& operator[](size_t i) const { return columns[i]; } |
| Column columns[4]; |
| }; |
| |
| TransformationMatrix() { |
| CheckAlignment(); |
| MakeIdentity(); |
| } |
| TransformationMatrix(const AffineTransform&); |
| TransformationMatrix(const TransformationMatrix& t) { |
| CheckAlignment(); |
| *this = t; |
| } |
| TransformationMatrix(double a, |
| double b, |
| double c, |
| double d, |
| double e, |
| double f) { |
| CheckAlignment(); |
| SetMatrix(a, b, c, d, e, f); |
| } |
| TransformationMatrix(double m11, |
| double m12, |
| double m13, |
| double m14, |
| double m21, |
| double m22, |
| double m23, |
| double m24, |
| double m31, |
| double m32, |
| double m33, |
| double m34, |
| double m41, |
| double m42, |
| double m43, |
| double m44) { |
| CheckAlignment(); |
| SetMatrix(m11, m12, m13, m14, m21, m22, m23, m24, m31, m32, m33, m34, m41, |
| m42, m43, m44); |
| } |
| TransformationMatrix(const SkMatrix44& matrix) { |
| CheckAlignment(); |
| SetMatrix( |
| matrix.get(0, 0), matrix.get(1, 0), matrix.get(2, 0), matrix.get(3, 0), |
| matrix.get(0, 1), matrix.get(1, 1), matrix.get(2, 1), matrix.get(3, 1), |
| matrix.get(0, 2), matrix.get(1, 2), matrix.get(2, 2), matrix.get(3, 2), |
| matrix.get(0, 3), matrix.get(1, 3), matrix.get(2, 3), matrix.get(3, 3)); |
| } |
| |
| void SetMatrix(double a, double b, double c, double d, double e, double f) { |
| matrix_[0][0] = a; |
| matrix_[0][1] = b; |
| matrix_[0][2] = 0; |
| matrix_[0][3] = 0; |
| matrix_[1][0] = c; |
| matrix_[1][1] = d; |
| matrix_[1][2] = 0; |
| matrix_[1][3] = 0; |
| matrix_[2][0] = 0; |
| matrix_[2][1] = 0; |
| matrix_[2][2] = 1; |
| matrix_[2][3] = 0; |
| matrix_[3][0] = e; |
| matrix_[3][1] = f; |
| matrix_[3][2] = 0; |
| matrix_[3][3] = 1; |
| } |
| |
| void SetMatrix(double m11, |
| double m12, |
| double m13, |
| double m14, |
| double m21, |
| double m22, |
| double m23, |
| double m24, |
| double m31, |
| double m32, |
| double m33, |
| double m34, |
| double m41, |
| double m42, |
| double m43, |
| double m44) { |
| matrix_[0][0] = m11; |
| matrix_[0][1] = m12; |
| matrix_[0][2] = m13; |
| matrix_[0][3] = m14; |
| matrix_[1][0] = m21; |
| matrix_[1][1] = m22; |
| matrix_[1][2] = m23; |
| matrix_[1][3] = m24; |
| matrix_[2][0] = m31; |
| matrix_[2][1] = m32; |
| matrix_[2][2] = m33; |
| matrix_[2][3] = m34; |
| matrix_[3][0] = m41; |
| matrix_[3][1] = m42; |
| matrix_[3][2] = m43; |
| matrix_[3][3] = m44; |
| } |
| |
| TransformationMatrix& operator=(const TransformationMatrix& t) { |
| SetMatrix(t.matrix_); |
| return *this; |
| } |
| |
| TransformationMatrix& MakeIdentity() { |
| SetMatrix(1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1); |
| return *this; |
| } |
| |
| bool IsIdentity() const { |
| return matrix_[0][0] == 1 && matrix_[0][1] == 0 && matrix_[0][2] == 0 && |
| matrix_[0][3] == 0 && matrix_[1][0] == 0 && matrix_[1][1] == 1 && |
| matrix_[1][2] == 0 && matrix_[1][3] == 0 && matrix_[2][0] == 0 && |
| matrix_[2][1] == 0 && matrix_[2][2] == 1 && matrix_[2][3] == 0 && |
| matrix_[3][0] == 0 && matrix_[3][1] == 0 && matrix_[3][2] == 0 && |
| matrix_[3][3] == 1; |
| } |
| |
| // Map a 3D point through the transform, returning a 3D point. |
| FloatPoint3D MapPoint(const FloatPoint3D&) const; |
| |
| // Map a 2D point through the transform, returning a 2D point. |
| // Note that this ignores the z component, effectively projecting the point |
| // into the z=0 plane. |
| FloatPoint MapPoint(const FloatPoint&) const; |
| |
| // If the matrix has 3D components, the z component of the result is |
| // dropped, effectively projecting the rect into the z=0 plane |
| FloatRect MapRect(const FloatRect&) const; |
| |
| // Rounds the resulting mapped rectangle out. This is helpful for bounding |
| // box computations but may not be what is wanted in other contexts. |
| IntRect MapRect(const IntRect&) const; |
| LayoutRect MapRect(const LayoutRect&) const; |
| |
| // If the matrix has 3D components, the z component of the result is |
| // dropped, effectively projecting the quad into the z=0 plane |
| FloatQuad MapQuad(const FloatQuad&) const; |
| |
| // Map a point on the z=0 plane into a point on |
| // the plane with with the transform applied, by extending |
| // a ray perpendicular to the source plane and computing |
| // the local x,y position of the point where that ray intersects |
| // with the destination plane. |
| FloatPoint ProjectPoint(const FloatPoint&, bool* clamped = nullptr) const; |
| // Projects the four corners of the quad |
| FloatQuad ProjectQuad(const FloatQuad&, bool* clamped = nullptr) const; |
| // Projects the four corners of the quad and takes a bounding box, |
| // while sanitizing values created when the w component is negative. |
| LayoutRect ClampedBoundsOfProjectedQuad(const FloatQuad&) const; |
| |
| void TransformBox(FloatBox&) const; |
| |
| // Important: These indices are spoken in col-major order. i.e.: |
| // | M11() M21() M31() M41() | |
| // | M12() M22() M32() M42() | |
| // | M13() M23() M33() M43() | |
| // | M14() M24() M34() M44() | |
| double M11() const { return matrix_[0][0]; } |
| void SetM11(double f) { matrix_[0][0] = f; } |
| double M12() const { return matrix_[0][1]; } |
| void SetM12(double f) { matrix_[0][1] = f; } |
| double M13() const { return matrix_[0][2]; } |
| void SetM13(double f) { matrix_[0][2] = f; } |
| double M14() const { return matrix_[0][3]; } |
| void SetM14(double f) { matrix_[0][3] = f; } |
| double M21() const { return matrix_[1][0]; } |
| void SetM21(double f) { matrix_[1][0] = f; } |
| double M22() const { return matrix_[1][1]; } |
| void SetM22(double f) { matrix_[1][1] = f; } |
| double M23() const { return matrix_[1][2]; } |
| void SetM23(double f) { matrix_[1][2] = f; } |
| double M24() const { return matrix_[1][3]; } |
| void SetM24(double f) { matrix_[1][3] = f; } |
| double M31() const { return matrix_[2][0]; } |
| void SetM31(double f) { matrix_[2][0] = f; } |
| double M32() const { return matrix_[2][1]; } |
| void SetM32(double f) { matrix_[2][1] = f; } |
| double M33() const { return matrix_[2][2]; } |
| void SetM33(double f) { matrix_[2][2] = f; } |
| double M34() const { return matrix_[2][3]; } |
| void SetM34(double f) { matrix_[2][3] = f; } |
| double M41() const { return matrix_[3][0]; } |
| void SetM41(double f) { matrix_[3][0] = f; } |
| double M42() const { return matrix_[3][1]; } |
| void SetM42(double f) { matrix_[3][1] = f; } |
| double M43() const { return matrix_[3][2]; } |
| void SetM43(double f) { matrix_[3][2] = f; } |
| double M44() const { return matrix_[3][3]; } |
| void SetM44(double f) { matrix_[3][3] = f; } |
| |
| double A() const { return matrix_[0][0]; } |
| void SetA(double a) { matrix_[0][0] = a; } |
| |
| double B() const { return matrix_[0][1]; } |
| void SetB(double b) { matrix_[0][1] = b; } |
| |
| double C() const { return matrix_[1][0]; } |
| void SetC(double c) { matrix_[1][0] = c; } |
| |
| double D() const { return matrix_[1][1]; } |
| void SetD(double d) { matrix_[1][1] = d; } |
| |
| double E() const { return matrix_[3][0]; } |
| void SetE(double e) { matrix_[3][0] = e; } |
| |
| double F() const { return matrix_[3][1]; } |
| void SetF(double f) { matrix_[3][1] = f; } |
| |
| // *this = *this * mat. |
| TransformationMatrix& Multiply(const TransformationMatrix&); |
| |
| TransformationMatrix& Scale(double); |
| TransformationMatrix& ScaleNonUniform(double sx, double sy); |
| TransformationMatrix& Scale3d(double sx, double sy, double sz); |
| |
| TransformationMatrix& Rotate(double d) { return Rotate3d(0, 0, d); } |
| // Angles are in degrees. |
| TransformationMatrix& Rotate3d(double rx, double ry, double rz); |
| TransformationMatrix& Rotate3d(const Rotation&); |
| |
| // The vector (x,y,z) is normalized if it's not already. A vector of |
| // (0,0,0) uses a vector of (0,0,1). |
| TransformationMatrix& Rotate3d(double x, double y, double z, double angle); |
| |
| TransformationMatrix& Translate(double tx, double ty); |
| TransformationMatrix& Translate3d(double tx, double ty, double tz); |
| |
| // Append translation after existing operations. i.e. |
| // TransformationMatrix t2 = t1; |
| // t2.PostTranslate(x, y); |
| // t2.MapPoint(p) == t1.MapPoint(p) + FloatPoint(x, y) |
| TransformationMatrix& PostTranslate(double tx, double ty); |
| TransformationMatrix& PostTranslate3d(double tx, double ty, double tz); |
| |
| TransformationMatrix& Skew(double angle_x, double angle_y); |
| TransformationMatrix& SkewX(double angle) { return Skew(angle, 0); } |
| TransformationMatrix& SkewY(double angle) { return Skew(0, angle); } |
| |
| TransformationMatrix& ApplyPerspective(double p); |
| |
| // Changes the transform to apply as if the origin were at (x, y, z). |
| TransformationMatrix& ApplyTransformOrigin(double x, double y, double z); |
| TransformationMatrix& ApplyTransformOrigin(const FloatPoint3D& origin) { |
| return ApplyTransformOrigin(origin.X(), origin.Y(), origin.Z()); |
| } |
| |
| // Changes the transform to: |
| // |
| // scale3d(z, z, z) * mat * scale3d(1/z, 1/z, 1/z) |
| // |
| // Useful for mapping zoomed points to their zoomed transformed result: |
| // |
| // new_mat * (scale3d(z, z, z) * x) == scale3d(z, z, z) * (mat * x) |
| // |
| TransformationMatrix& Zoom(double zoom_factor); |
| |
| bool IsInvertible() const; |
| |
| // This method returns the identity matrix if it is not invertible. |
| // Use isInvertible() before calling this if you need to know. |
| TransformationMatrix Inverse() const; |
| |
| // decompose the matrix into its component parts |
| typedef struct { |
| double scale_x, scale_y, scale_z; |
| double skew_xy, skew_xz, skew_yz; |
| double quaternion_x, quaternion_y, quaternion_z, quaternion_w; |
| double translate_x, translate_y, translate_z; |
| double perspective_x, perspective_y, perspective_z, perspective_w; |
| } DecomposedType; |
| |
| // Decompose 2-D transform matrix into its component parts. |
| typedef struct { |
| double scale_x, scale_y; |
| double skew_xy; |
| double translate_x, translate_y; |
| double angle; |
| } Decomposed2dType; |
| |
| WARN_UNUSED_RESULT bool Decompose(DecomposedType&) const; |
| WARN_UNUSED_RESULT bool Decompose2D(Decomposed2dType&) const; |
| void Recompose(const DecomposedType&); |
| void Recompose2D(const Decomposed2dType&); |
| void Blend(const TransformationMatrix& from, double progress); |
| void Blend2D(const TransformationMatrix& from, double progress); |
| |
| bool IsAffine() const { |
| return M13() == 0 && M14() == 0 && M23() == 0 && M24() == 0 && M31() == 0 && |
| M32() == 0 && M33() == 1 && M34() == 0 && M43() == 0 && M44() == 1; |
| } |
| |
| // Throw away the non-affine parts of the matrix (lossy!) |
| void MakeAffine(); |
| |
| AffineTransform ToAffineTransform() const; |
| |
| // Flatten into a 2-D transformation (non-invertable). |
| // Same as gfx::Transform::FlattenTo2d(); see the docs for that function for |
| // details and discussion. |
| void FlattenTo2d(); |
| |
| bool operator==(const TransformationMatrix& m2) const { |
| return matrix_[0][0] == m2.matrix_[0][0] && |
| matrix_[0][1] == m2.matrix_[0][1] && |
| matrix_[0][2] == m2.matrix_[0][2] && |
| matrix_[0][3] == m2.matrix_[0][3] && |
| matrix_[1][0] == m2.matrix_[1][0] && |
| matrix_[1][1] == m2.matrix_[1][1] && |
| matrix_[1][2] == m2.matrix_[1][2] && |
| matrix_[1][3] == m2.matrix_[1][3] && |
| matrix_[2][0] == m2.matrix_[2][0] && |
| matrix_[2][1] == m2.matrix_[2][1] && |
| matrix_[2][2] == m2.matrix_[2][2] && |
| matrix_[2][3] == m2.matrix_[2][3] && |
| matrix_[3][0] == m2.matrix_[3][0] && |
| matrix_[3][1] == m2.matrix_[3][1] && |
| matrix_[3][2] == m2.matrix_[3][2] && |
| matrix_[3][3] == m2.matrix_[3][3]; |
| } |
| |
| bool operator!=(const TransformationMatrix& other) const { |
| return !(*this == other); |
| } |
| |
| // *this = *this * t |
| TransformationMatrix& operator*=(const TransformationMatrix& t) { |
| return Multiply(t); |
| } |
| |
| // result = *this * t |
| TransformationMatrix operator*(const TransformationMatrix& t) const { |
| TransformationMatrix result = *this; |
| result.Multiply(t); |
| return result; |
| } |
| |
| bool IsFlat() const { |
| return matrix_[0][2] == 0.f && matrix_[1][2] == 0.f && |
| matrix_[2][0] == 0.f && matrix_[2][1] == 0.f && |
| matrix_[2][2] == 1.f && matrix_[2][3] == 0.f && matrix_[3][2] == 0.f; |
| } |
| |
| bool IsIdentityOrTranslation() const { |
| return matrix_[0][0] == 1 && matrix_[0][1] == 0 && matrix_[0][2] == 0 && |
| matrix_[0][3] == 0 && matrix_[1][0] == 0 && matrix_[1][1] == 1 && |
| matrix_[1][2] == 0 && matrix_[1][3] == 0 && matrix_[2][0] == 0 && |
| matrix_[2][1] == 0 && matrix_[2][2] == 1 && matrix_[2][3] == 0 && |
| matrix_[3][3] == 1; |
| } |
| |
| bool IsIdentityOr2DTranslation() const { |
| return IsIdentityOrTranslation() && matrix_[3][2] == 0; |
| } |
| |
| bool Is2DProportionalUpscaleAndOr2DTranslation() const { |
| if (matrix_[0][0] < 1 || matrix_[0][0] != matrix_[1][1]) |
| return false; |
| return matrix_[0][1] == 0 && matrix_[0][2] == 0 && matrix_[0][3] == 0 && |
| matrix_[1][0] == 0 && matrix_[1][2] == 0 && matrix_[1][3] == 0 && |
| matrix_[2][0] == 0 && matrix_[2][1] == 0 && matrix_[2][2] == 1 && |
| matrix_[2][3] == 0 && matrix_[3][2] == 0 && matrix_[3][3] == 1; |
| } |
| |
| bool Is2dTransform() const; |
| |
| bool IsIntegerTranslation() const; |
| |
| // Returns true if axis-aligned 2d rects will remain axis-aligned after being |
| // transformed by this matrix. |
| bool Preserves2dAxisAlignment() const; |
| |
| // If this transformation is identity or 2D translation, returns the |
| // translation. |
| FloatSize To2DTranslation() const { |
| DCHECK(IsIdentityOr2DTranslation()); |
| return FloatSize(matrix_[3][0], matrix_[3][1]); |
| } |
| |
| typedef float FloatMatrix4[16]; |
| void ToColumnMajorFloatArray(FloatMatrix4& result) const; |
| |
| static SkMatrix44 ToSkMatrix44(const TransformationMatrix&); |
| static gfx::Transform ToTransform(const TransformationMatrix&); |
| |
| // If |asMatrix|, return the matrix in row-major order. Otherwise, return |
| // the transform's decomposition which shows the translation, scale, etc. |
| String ToString(bool as_matrix = false) const; |
| |
| private: |
| // multiply passed 2D point by matrix (assume z=0) |
| void MultVecMatrix(double x, double y, double& dst_x, double& dst_y) const; |
| FloatPoint InternalMapPoint(const FloatPoint& source_point) const { |
| double result_x; |
| double result_y; |
| MultVecMatrix(source_point.X(), source_point.Y(), result_x, result_y); |
| return FloatPoint(static_cast<float>(result_x), |
| static_cast<float>(result_y)); |
| } |
| |
| // multiply passed 3D point by matrix |
| void MultVecMatrix(double x, |
| double y, |
| double z, |
| double& dst_x, |
| double& dst_y, |
| double& dst_z) const; |
| FloatPoint3D InternalMapPoint(const FloatPoint3D& source_point) const { |
| double result_x; |
| double result_y; |
| double result_z; |
| MultVecMatrix(source_point.X(), source_point.Y(), source_point.Z(), |
| result_x, result_y, result_z); |
| return FloatPoint3D(static_cast<float>(result_x), |
| static_cast<float>(result_y), |
| static_cast<float>(result_z)); |
| } |
| |
| void SetMatrix(const Matrix4& m) { memcpy(&matrix_, &m, sizeof(Matrix4)); } |
| |
| void CheckAlignment() { |
| #if defined(TRANSFORMATION_MATRIX_USE_X86_64_SSE2) |
| // m_matrix can cause this class to require higher than usual alignment. |
| // Make sure the allocator handles this. |
| DCHECK_EQ((reinterpret_cast<uintptr_t>(this) & |
| (alignof(TransformationMatrix) - 1)), |
| 0UL); |
| #endif |
| } |
| |
| Matrix4 matrix_; |
| }; |
| |
| PLATFORM_EXPORT std::ostream& operator<<(std::ostream&, |
| const TransformationMatrix&); |
| PLATFORM_EXPORT std::unique_ptr<JSONArray> TransformAsJSONArray( |
| const TransformationMatrix&); |
| |
| } // namespace blink |
| |
| #endif // THIRD_PARTY_BLINK_RENDERER_PLATFORM_TRANSFORMS_TRANSFORMATION_MATRIX_H_ |