gecko-dev/dom/base/DOMMatrix.h

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

334 строки
13 KiB
C
Исходник Обычный вид История

/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
/* vim: set ts=8 sts=2 et sw=2 tw=80: */
/* This Source Code Form is subject to the terms of the Mozilla Public
* License, v. 2.0. If a copy of the MPL was not distributed with this
* file, You can obtain one at http://mozilla.org/MPL/2.0/. */
#ifndef MOZILLA_DOM_DOMMATRIX_H_
#define MOZILLA_DOM_DOMMATRIX_H_
#include "js/StructuredClone.h"
#include "nsWrapperCache.h"
#include "nsISupports.h"
#include "nsCycleCollectionParticipant.h"
#include "mozilla/Attributes.h"
#include "mozilla/ErrorResult.h"
#include "mozilla/UniquePtr.h"
#include "nsCOMPtr.h"
#include "mozilla/dom/BindingDeclarations.h"
#include "mozilla/dom/TypedArray.h"
#include "mozilla/gfx/Matrix.h" // for Matrix4x4Double
class nsIGlobalObject;
namespace mozilla {
namespace dom {
class GlobalObject;
class DOMMatrix;
class DOMPoint;
class UTF8StringOrUnrestrictedDoubleSequenceOrDOMMatrixReadOnly;
struct DOMPointInit;
struct DOMMatrixInit;
struct DOMMatrix2DInit;
class DOMMatrixReadOnly : public nsWrapperCache {
public:
explicit DOMMatrixReadOnly(nsISupports* aParent)
: mParent(aParent), mMatrix2D(new gfx::MatrixDouble()) {}
DOMMatrixReadOnly(nsISupports* aParent, const DOMMatrixReadOnly& other)
: mParent(aParent) {
if (other.mMatrix2D) {
mMatrix2D = MakeUnique<gfx::MatrixDouble>(*other.mMatrix2D);
} else {
mMatrix3D = MakeUnique<gfx::Matrix4x4Double>(*other.mMatrix3D);
}
}
DOMMatrixReadOnly(nsISupports* aParent, const gfx::Matrix4x4& aMatrix)
: mParent(aParent) {
mMatrix3D = MakeUnique<gfx::Matrix4x4Double>(aMatrix);
}
DOMMatrixReadOnly(nsISupports* aParent, const gfx::Matrix& aMatrix)
: mParent(aParent) {
mMatrix2D = MakeUnique<gfx::MatrixDouble>(aMatrix);
}
NS_INLINE_DECL_CYCLE_COLLECTING_NATIVE_REFCOUNTING(DOMMatrixReadOnly)
NS_DECL_CYCLE_COLLECTION_SCRIPT_HOLDER_NATIVE_CLASS(DOMMatrixReadOnly)
Bug 1186265 - Partially update DOMPoint, DOMQuad, DOMRect, DOMMatrix. r=bz Some notes: this does not fully bring us to compliance to the current spec. Instead, these are the fixes that I needed to make in order to make css/geometry/interfaces.html pass with the DOMPoint changes in the previous patches. I don't fully understand why that patch caused the test to fail the way it did, but it ended up being easier to fix our code than understand why the harness was falling over. The DOMQuad::QuadBounds class was the source of some confusion for me. Now that DOMRectReadOnly is a concrete class with members, I wanted to avoid wasting them. However, the spec is unclear as to whether a DOMQuad's bound's should be live -- that is because DOMQuad exposes DOMPoint, we can set its points after retrieving a QuadBounds object. Our current code is live, setting the points changes the QuadBounds. Chromium's current behavior is to never update the QuadBounds object. I've left our behavior untouched in this patch (and waste 4 doubles per QuadBounds object), but I am intending to file a bug to understand what the intent of the spec is. I wonder if the author intended the points to be DOMPointReadOnly instead. If so, we could simplify the DOMRectReadOnly code and get rid of the virtual getters, which would be nice. I also wasn't thrilled to put the DOMMatrix setters on the DOMMatrixReadOnly class, but for brevity and simplicity of implementation, I've made them public. I briefly considered making the setters protected on the ReadOnly version of the class, but I'm not convinced that having to explicitly make them public on the derived class is worth the extra copies of the names. MozReview-Commit-ID: CjdW4Nbnc6A --HG-- extra : rebase_source : 44489693afebff571a415b487e29fa6153288421
2018-03-30 02:19:31 +03:00
nsISupports* GetParentObject() const { return mParent; }
virtual JSObject* WrapObject(JSContext* cx,
JS::Handle<JSObject*> aGivenProto) override;
static already_AddRefed<DOMMatrixReadOnly> FromMatrix(
nsISupports* aParent, const DOMMatrix2DInit& aMatrixInit,
ErrorResult& aRv);
static already_AddRefed<DOMMatrixReadOnly> FromMatrix(
nsISupports* aParent, const DOMMatrixInit& aMatrixInit, ErrorResult& aRv);
static already_AddRefed<DOMMatrixReadOnly> FromMatrix(
const GlobalObject& aGlobal, const DOMMatrixInit& aMatrixInit,
ErrorResult& aRv);
static already_AddRefed<DOMMatrixReadOnly> FromFloat32Array(
const GlobalObject& aGlobal, const Float32Array& aArray32,
ErrorResult& aRv);
static already_AddRefed<DOMMatrixReadOnly> FromFloat64Array(
const GlobalObject& aGlobal, const Float64Array& aArray64,
ErrorResult& aRv);
Bug 1186265 - Partially update DOMPoint, DOMQuad, DOMRect, DOMMatrix. r=bz Some notes: this does not fully bring us to compliance to the current spec. Instead, these are the fixes that I needed to make in order to make css/geometry/interfaces.html pass with the DOMPoint changes in the previous patches. I don't fully understand why that patch caused the test to fail the way it did, but it ended up being easier to fix our code than understand why the harness was falling over. The DOMQuad::QuadBounds class was the source of some confusion for me. Now that DOMRectReadOnly is a concrete class with members, I wanted to avoid wasting them. However, the spec is unclear as to whether a DOMQuad's bound's should be live -- that is because DOMQuad exposes DOMPoint, we can set its points after retrieving a QuadBounds object. Our current code is live, setting the points changes the QuadBounds. Chromium's current behavior is to never update the QuadBounds object. I've left our behavior untouched in this patch (and waste 4 doubles per QuadBounds object), but I am intending to file a bug to understand what the intent of the spec is. I wonder if the author intended the points to be DOMPointReadOnly instead. If so, we could simplify the DOMRectReadOnly code and get rid of the virtual getters, which would be nice. I also wasn't thrilled to put the DOMMatrix setters on the DOMMatrixReadOnly class, but for brevity and simplicity of implementation, I've made them public. I briefly considered making the setters protected on the ReadOnly version of the class, but I'm not convinced that having to explicitly make them public on the derived class is worth the extra copies of the names. MozReview-Commit-ID: CjdW4Nbnc6A --HG-- extra : rebase_source : 44489693afebff571a415b487e29fa6153288421
2018-03-30 02:19:31 +03:00
static already_AddRefed<DOMMatrixReadOnly> Constructor(
const GlobalObject& aGlobal,
const Optional<UTF8StringOrUnrestrictedDoubleSequenceOrDOMMatrixReadOnly>&
aArg,
Bug 1186265 - Partially update DOMPoint, DOMQuad, DOMRect, DOMMatrix. r=bz Some notes: this does not fully bring us to compliance to the current spec. Instead, these are the fixes that I needed to make in order to make css/geometry/interfaces.html pass with the DOMPoint changes in the previous patches. I don't fully understand why that patch caused the test to fail the way it did, but it ended up being easier to fix our code than understand why the harness was falling over. The DOMQuad::QuadBounds class was the source of some confusion for me. Now that DOMRectReadOnly is a concrete class with members, I wanted to avoid wasting them. However, the spec is unclear as to whether a DOMQuad's bound's should be live -- that is because DOMQuad exposes DOMPoint, we can set its points after retrieving a QuadBounds object. Our current code is live, setting the points changes the QuadBounds. Chromium's current behavior is to never update the QuadBounds object. I've left our behavior untouched in this patch (and waste 4 doubles per QuadBounds object), but I am intending to file a bug to understand what the intent of the spec is. I wonder if the author intended the points to be DOMPointReadOnly instead. If so, we could simplify the DOMRectReadOnly code and get rid of the virtual getters, which would be nice. I also wasn't thrilled to put the DOMMatrix setters on the DOMMatrixReadOnly class, but for brevity and simplicity of implementation, I've made them public. I briefly considered making the setters protected on the ReadOnly version of the class, but I'm not convinced that having to explicitly make them public on the derived class is worth the extra copies of the names. MozReview-Commit-ID: CjdW4Nbnc6A --HG-- extra : rebase_source : 44489693afebff571a415b487e29fa6153288421
2018-03-30 02:19:31 +03:00
ErrorResult& aRv);
static already_AddRefed<DOMMatrixReadOnly> ReadStructuredClone(
JSContext* aCx, nsIGlobalObject* aGlobal,
JSStructuredCloneReader* aReader);
// clang-format off
#define GetMatrixMember(entry2D, entry3D, default) \
{ \
if (mMatrix3D) { \
return mMatrix3D->entry3D; \
} \
return mMatrix2D->entry2D; \
}
#define Get3DMatrixMember(entry3D, default) \
{ \
if (mMatrix3D) { \
return mMatrix3D->entry3D; \
} \
return default; \
}
double A() const GetMatrixMember(_11, _11, 1.0)
double B() const GetMatrixMember(_12, _12, 0)
double C() const GetMatrixMember(_21, _21, 0)
double D() const GetMatrixMember(_22, _22, 1.0)
double E() const GetMatrixMember(_31, _41, 0)
double F() const GetMatrixMember(_32, _42, 0)
double M11() const GetMatrixMember(_11, _11, 1.0)
double M12() const GetMatrixMember(_12, _12, 0)
double M13() const Get3DMatrixMember(_13, 0)
double M14() const Get3DMatrixMember(_14, 0)
double M21() const GetMatrixMember(_21, _21, 0)
double M22() const GetMatrixMember(_22, _22, 1.0)
double M23() const Get3DMatrixMember(_23, 0)
double M24() const Get3DMatrixMember(_24, 0)
double M31() const Get3DMatrixMember(_31, 0)
double M32() const Get3DMatrixMember(_32, 0)
double M33() const Get3DMatrixMember(_33, 1.0)
double M34() const Get3DMatrixMember(_34, 0)
double M41() const GetMatrixMember(_31, _41, 0)
double M42() const GetMatrixMember(_32, _42, 0)
double M43() const Get3DMatrixMember(_43, 0)
double M44() const Get3DMatrixMember(_44, 1.0)
#undef GetMatrixMember
#undef Get3DMatrixMember
Bug 1186265 - Partially update DOMPoint, DOMQuad, DOMRect, DOMMatrix. r=bz Some notes: this does not fully bring us to compliance to the current spec. Instead, these are the fixes that I needed to make in order to make css/geometry/interfaces.html pass with the DOMPoint changes in the previous patches. I don't fully understand why that patch caused the test to fail the way it did, but it ended up being easier to fix our code than understand why the harness was falling over. The DOMQuad::QuadBounds class was the source of some confusion for me. Now that DOMRectReadOnly is a concrete class with members, I wanted to avoid wasting them. However, the spec is unclear as to whether a DOMQuad's bound's should be live -- that is because DOMQuad exposes DOMPoint, we can set its points after retrieving a QuadBounds object. Our current code is live, setting the points changes the QuadBounds. Chromium's current behavior is to never update the QuadBounds object. I've left our behavior untouched in this patch (and waste 4 doubles per QuadBounds object), but I am intending to file a bug to understand what the intent of the spec is. I wonder if the author intended the points to be DOMPointReadOnly instead. If so, we could simplify the DOMRectReadOnly code and get rid of the virtual getters, which would be nice. I also wasn't thrilled to put the DOMMatrix setters on the DOMMatrixReadOnly class, but for brevity and simplicity of implementation, I've made them public. I briefly considered making the setters protected on the ReadOnly version of the class, but I'm not convinced that having to explicitly make them public on the derived class is worth the extra copies of the names. MozReview-Commit-ID: CjdW4Nbnc6A --HG-- extra : rebase_source : 44489693afebff571a415b487e29fa6153288421
2018-03-30 02:19:31 +03:00
// Defined here so we can construct DOMMatrixReadOnly objects.
#define Set2DMatrixMember(entry2D, entry3D) \
{ \
if (mMatrix3D) { \
mMatrix3D->entry3D = v; \
} else { \
mMatrix2D->entry2D = v; \
} \
}
#define Set3DMatrixMember(entry3D, default) \
{ \
if (mMatrix3D || (v != default)) { \
Ensure3DMatrix(); \
mMatrix3D->entry3D = v; \
} \
}
void SetA(double v) Set2DMatrixMember(_11, _11)
void SetB(double v) Set2DMatrixMember(_12, _12)
void SetC(double v) Set2DMatrixMember(_21, _21)
void SetD(double v) Set2DMatrixMember(_22, _22)
void SetE(double v) Set2DMatrixMember(_31, _41)
void SetF(double v) Set2DMatrixMember(_32, _42)
void SetM11(double v) Set2DMatrixMember(_11, _11)
void SetM12(double v) Set2DMatrixMember(_12, _12)
void SetM13(double v) Set3DMatrixMember(_13, 0)
void SetM14(double v) Set3DMatrixMember(_14, 0)
void SetM21(double v) Set2DMatrixMember(_21, _21)
void SetM22(double v) Set2DMatrixMember(_22, _22)
void SetM23(double v) Set3DMatrixMember(_23, 0)
void SetM24(double v) Set3DMatrixMember(_24, 0)
void SetM31(double v) Set3DMatrixMember(_31, 0)
void SetM32(double v) Set3DMatrixMember(_32, 0)
void SetM33(double v) Set3DMatrixMember(_33, 1.0)
void SetM34(double v) Set3DMatrixMember(_34, 0)
void SetM41(double v) Set2DMatrixMember(_31, _41)
void SetM42(double v) Set2DMatrixMember(_32, _42)
void SetM43(double v) Set3DMatrixMember(_43, 0)
void SetM44(double v) Set3DMatrixMember(_44, 1.0)
; // semi-colon here to get clang-format to align properly from here on
Bug 1186265 - Partially update DOMPoint, DOMQuad, DOMRect, DOMMatrix. r=bz Some notes: this does not fully bring us to compliance to the current spec. Instead, these are the fixes that I needed to make in order to make css/geometry/interfaces.html pass with the DOMPoint changes in the previous patches. I don't fully understand why that patch caused the test to fail the way it did, but it ended up being easier to fix our code than understand why the harness was falling over. The DOMQuad::QuadBounds class was the source of some confusion for me. Now that DOMRectReadOnly is a concrete class with members, I wanted to avoid wasting them. However, the spec is unclear as to whether a DOMQuad's bound's should be live -- that is because DOMQuad exposes DOMPoint, we can set its points after retrieving a QuadBounds object. Our current code is live, setting the points changes the QuadBounds. Chromium's current behavior is to never update the QuadBounds object. I've left our behavior untouched in this patch (and waste 4 doubles per QuadBounds object), but I am intending to file a bug to understand what the intent of the spec is. I wonder if the author intended the points to be DOMPointReadOnly instead. If so, we could simplify the DOMRectReadOnly code and get rid of the virtual getters, which would be nice. I also wasn't thrilled to put the DOMMatrix setters on the DOMMatrixReadOnly class, but for brevity and simplicity of implementation, I've made them public. I briefly considered making the setters protected on the ReadOnly version of the class, but I'm not convinced that having to explicitly make them public on the derived class is worth the extra copies of the names. MozReview-Commit-ID: CjdW4Nbnc6A --HG-- extra : rebase_source : 44489693afebff571a415b487e29fa6153288421
2018-03-30 02:19:31 +03:00
#undef Set2DMatrixMember
#undef Set3DMatrixMember
// clang-format on
Bug 1186265 - Partially update DOMPoint, DOMQuad, DOMRect, DOMMatrix. r=bz Some notes: this does not fully bring us to compliance to the current spec. Instead, these are the fixes that I needed to make in order to make css/geometry/interfaces.html pass with the DOMPoint changes in the previous patches. I don't fully understand why that patch caused the test to fail the way it did, but it ended up being easier to fix our code than understand why the harness was falling over. The DOMQuad::QuadBounds class was the source of some confusion for me. Now that DOMRectReadOnly is a concrete class with members, I wanted to avoid wasting them. However, the spec is unclear as to whether a DOMQuad's bound's should be live -- that is because DOMQuad exposes DOMPoint, we can set its points after retrieving a QuadBounds object. Our current code is live, setting the points changes the QuadBounds. Chromium's current behavior is to never update the QuadBounds object. I've left our behavior untouched in this patch (and waste 4 doubles per QuadBounds object), but I am intending to file a bug to understand what the intent of the spec is. I wonder if the author intended the points to be DOMPointReadOnly instead. If so, we could simplify the DOMRectReadOnly code and get rid of the virtual getters, which would be nice. I also wasn't thrilled to put the DOMMatrix setters on the DOMMatrixReadOnly class, but for brevity and simplicity of implementation, I've made them public. I briefly considered making the setters protected on the ReadOnly version of the class, but I'm not convinced that having to explicitly make them public on the derived class is worth the extra copies of the names. MozReview-Commit-ID: CjdW4Nbnc6A --HG-- extra : rebase_source : 44489693afebff571a415b487e29fa6153288421
2018-03-30 02:19:31 +03:00
already_AddRefed<DOMMatrix> Translate(double aTx, double aTy,
double aTz = 0) const;
already_AddRefed<DOMMatrix> Scale(double aScaleX,
const Optional<double>& aScaleY,
double aScaleZ, double aOriginX,
double aOriginY, double aOriginZ) const;
already_AddRefed<DOMMatrix> Scale3d(double aScale, double aOriginX = 0,
double aOriginY = 0,
double aOriginZ = 0) const;
already_AddRefed<DOMMatrix> ScaleNonUniform(double aScaleX,
double aScaleY) const;
already_AddRefed<DOMMatrix> Rotate(double aRotX,
const Optional<double>& aRotY,
const Optional<double>& aRotZ) const;
already_AddRefed<DOMMatrix> RotateFromVector(double aX, double aY) const;
already_AddRefed<DOMMatrix> RotateAxisAngle(double aX, double aY, double aZ,
double aAngle) const;
already_AddRefed<DOMMatrix> SkewX(double aSx) const;
already_AddRefed<DOMMatrix> SkewY(double aSy) const;
already_AddRefed<DOMMatrix> Multiply(const DOMMatrixInit& aOther,
ErrorResult& aRv) const;
already_AddRefed<DOMMatrix> FlipX() const;
already_AddRefed<DOMMatrix> FlipY() const;
already_AddRefed<DOMMatrix> Inverse() const;
bool Is2D() const;
bool IsIdentity() const;
already_AddRefed<DOMPoint> TransformPoint(const DOMPointInit& aPoint) const;
void ToFloat32Array(JSContext* aCx, JS::MutableHandle<JSObject*> aResult,
ErrorResult& aRv) const;
void ToFloat64Array(JSContext* aCx, JS::MutableHandle<JSObject*> aResult,
ErrorResult& aRv) const;
void Stringify(nsAString& aResult, ErrorResult& aRv);
bool WriteStructuredClone(JSContext* aCx,
JSStructuredCloneWriter* aWriter) const;
const gfx::MatrixDouble* GetInternal2D() const {
if (Is2D()) {
return mMatrix2D.get();
}
return nullptr;
}
protected:
nsCOMPtr<nsISupports> mParent;
UniquePtr<gfx::MatrixDouble> mMatrix2D;
UniquePtr<gfx::Matrix4x4Double> mMatrix3D;
virtual ~DOMMatrixReadOnly() = default;
/**
* Sets data from a fully validated and fixed-up matrix init,
* where all of its members are properly defined.
* The init dictionary's dimension must match the matrix one.
*/
void SetDataFromMatrix2DInit(const DOMMatrix2DInit& aMatrixInit);
void SetDataFromMatrixInit(const DOMMatrixInit& aMatrixInit);
DOMMatrixReadOnly* SetMatrixValue(const nsACString&, ErrorResult&);
Bug 1186265 - Partially update DOMPoint, DOMQuad, DOMRect, DOMMatrix. r=bz Some notes: this does not fully bring us to compliance to the current spec. Instead, these are the fixes that I needed to make in order to make css/geometry/interfaces.html pass with the DOMPoint changes in the previous patches. I don't fully understand why that patch caused the test to fail the way it did, but it ended up being easier to fix our code than understand why the harness was falling over. The DOMQuad::QuadBounds class was the source of some confusion for me. Now that DOMRectReadOnly is a concrete class with members, I wanted to avoid wasting them. However, the spec is unclear as to whether a DOMQuad's bound's should be live -- that is because DOMQuad exposes DOMPoint, we can set its points after retrieving a QuadBounds object. Our current code is live, setting the points changes the QuadBounds. Chromium's current behavior is to never update the QuadBounds object. I've left our behavior untouched in this patch (and waste 4 doubles per QuadBounds object), but I am intending to file a bug to understand what the intent of the spec is. I wonder if the author intended the points to be DOMPointReadOnly instead. If so, we could simplify the DOMRectReadOnly code and get rid of the virtual getters, which would be nice. I also wasn't thrilled to put the DOMMatrix setters on the DOMMatrixReadOnly class, but for brevity and simplicity of implementation, I've made them public. I briefly considered making the setters protected on the ReadOnly version of the class, but I'm not convinced that having to explicitly make them public on the derived class is worth the extra copies of the names. MozReview-Commit-ID: CjdW4Nbnc6A --HG-- extra : rebase_source : 44489693afebff571a415b487e29fa6153288421
2018-03-30 02:19:31 +03:00
void Ensure3DMatrix();
DOMMatrixReadOnly(nsISupports* aParent, bool is2D) : mParent(aParent) {
if (is2D) {
mMatrix2D = MakeUnique<gfx::MatrixDouble>();
} else {
mMatrix3D = MakeUnique<gfx::Matrix4x4Double>();
}
}
static bool ReadStructuredCloneElements(JSStructuredCloneReader* aReader,
DOMMatrixReadOnly* matrix);
private:
DOMMatrixReadOnly() = delete;
DOMMatrixReadOnly(const DOMMatrixReadOnly&) = delete;
DOMMatrixReadOnly& operator=(const DOMMatrixReadOnly&) = delete;
};
class DOMMatrix : public DOMMatrixReadOnly {
public:
explicit DOMMatrix(nsISupports* aParent) : DOMMatrixReadOnly(aParent) {}
DOMMatrix(nsISupports* aParent, const DOMMatrixReadOnly& other)
: DOMMatrixReadOnly(aParent, other) {}
DOMMatrix(nsISupports* aParent, const gfx::Matrix4x4& aMatrix)
: DOMMatrixReadOnly(aParent, aMatrix) {}
DOMMatrix(nsISupports* aParent, const gfx::Matrix& aMatrix)
: DOMMatrixReadOnly(aParent, aMatrix) {}
static already_AddRefed<DOMMatrix> FromMatrix(
nsISupports* aParent, const DOMMatrixInit& aMatrixInit, ErrorResult& aRv);
static already_AddRefed<DOMMatrix> FromMatrix(
const GlobalObject& aGlobal, const DOMMatrixInit& aMatrixInit,
ErrorResult& aRv);
static already_AddRefed<DOMMatrix> FromFloat32Array(
const GlobalObject& aGlobal, const Float32Array& aArray32,
ErrorResult& aRv);
static already_AddRefed<DOMMatrix> FromFloat64Array(
const GlobalObject& aGlobal, const Float64Array& aArray64,
ErrorResult& aRv);
static already_AddRefed<DOMMatrix> Constructor(
const GlobalObject& aGlobal,
const Optional<UTF8StringOrUnrestrictedDoubleSequenceOrDOMMatrixReadOnly>&
aArg,
ErrorResult& aRv);
static already_AddRefed<DOMMatrix> ReadStructuredClone(
JSContext* aCx, nsIGlobalObject* aGlobal,
JSStructuredCloneReader* aReader);
virtual JSObject* WrapObject(JSContext* aCx,
JS::Handle<JSObject*> aGivenProto) override;
DOMMatrix* MultiplySelf(const DOMMatrixInit& aOther, ErrorResult& aRv);
DOMMatrix* PreMultiplySelf(const DOMMatrixInit& aOther, ErrorResult& aRv);
DOMMatrix* TranslateSelf(double aTx, double aTy, double aTz = 0);
DOMMatrix* ScaleSelf(double aScaleX, const Optional<double>& aScaleY,
double aScaleZ, double aOriginX, double aOriginY,
double aOriginZ);
DOMMatrix* Scale3dSelf(double aScale, double aOriginX = 0,
double aOriginY = 0, double aOriginZ = 0);
DOMMatrix* RotateSelf(double aRotX, const Optional<double>& aRotY,
const Optional<double>& aRotZ);
DOMMatrix* RotateFromVectorSelf(double aX, double aY);
DOMMatrix* RotateAxisAngleSelf(double aX, double aY, double aZ,
double aAngle);
DOMMatrix* SkewXSelf(double aSx);
DOMMatrix* SkewYSelf(double aSy);
DOMMatrix* InvertSelf();
DOMMatrix* SetMatrixValue(const nsACString&, ErrorResult&);
virtual ~DOMMatrix() = default;
private:
DOMMatrix(nsISupports* aParent, bool is2D)
: DOMMatrixReadOnly(aParent, is2D) {}
};
} // namespace dom
} // namespace mozilla
#endif /*MOZILLA_DOM_DOMMATRIX_H_*/