| /* |
| * Copyright 2012 The Android Open Source Project |
| * |
| * Use of this source code is governed by a BSD-style license that can be |
| * found in the LICENSE file. |
| */ |
| |
| #include "include/core/SkImageFilter.h" |
| |
| #include "include/core/SkCanvas.h" |
| #include "include/core/SkRect.h" |
| #include "include/effects/SkComposeImageFilter.h" |
| #include "include/private/SkSafe32.h" |
| #include "src/core/SkFuzzLogging.h" |
| #include "src/core/SkImageFilterCache.h" |
| #include "src/core/SkImageFilter_Base.h" |
| #include "src/core/SkLocalMatrixImageFilter.h" |
| #include "src/core/SkMatrixImageFilter.h" |
| #include "src/core/SkReadBuffer.h" |
| #include "src/core/SkSpecialImage.h" |
| #include "src/core/SkSpecialSurface.h" |
| #include "src/core/SkValidationUtils.h" |
| #include "src/core/SkWriteBuffer.h" |
| #if SK_SUPPORT_GPU |
| #include "include/gpu/GrContext.h" |
| #include "include/private/GrRecordingContext.h" |
| #include "src/gpu/GrColorSpaceXform.h" |
| #include "src/gpu/GrContextPriv.h" |
| #include "src/gpu/GrFixedClip.h" |
| #include "src/gpu/GrRecordingContextPriv.h" |
| #include "src/gpu/GrRenderTargetContext.h" |
| #include "src/gpu/GrTextureProxy.h" |
| #include "src/gpu/SkGr.h" |
| #endif |
| #include <atomic> |
| |
| /////////////////////////////////////////////////////////////////////////////////////////////////// |
| // SkImageFilter - A number of the public APIs on SkImageFilter downcast to SkImageFilter_Base |
| // in order to perform their actual work. |
| /////////////////////////////////////////////////////////////////////////////////////////////////// |
| |
| /** |
| * Returns the number of inputs this filter will accept (some inputs can |
| * be NULL). |
| */ |
| int SkImageFilter::countInputs() const { return as_IFB(this)->fInputs.count(); } |
| |
| /** |
| * Returns the input filter at a given index, or NULL if no input is |
| * connected. The indices used are filter-specific. |
| */ |
| const SkImageFilter* SkImageFilter::getInput(int i) const { |
| SkASSERT(i < this->countInputs()); |
| return as_IFB(this)->fInputs[i].get(); |
| } |
| |
| bool SkImageFilter::isColorFilterNode(SkColorFilter** filterPtr) const { |
| return as_IFB(this)->onIsColorFilterNode(filterPtr); |
| } |
| |
| SkIRect SkImageFilter::filterBounds(const SkIRect& src, const SkMatrix& ctm, |
| MapDirection direction, const SkIRect* inputRect) const { |
| if (kReverse_MapDirection == direction) { |
| SkIRect bounds = as_IFB(this)->onFilterNodeBounds(src, ctm, direction, inputRect); |
| return as_IFB(this)->onFilterBounds(bounds, ctm, direction, &bounds); |
| } else { |
| SkASSERT(!inputRect); |
| SkIRect bounds = as_IFB(this)->onFilterBounds(src, ctm, direction, nullptr); |
| bounds = as_IFB(this)->onFilterNodeBounds(bounds, ctm, direction, nullptr); |
| SkIRect dst; |
| as_IFB(this)->getCropRect().applyTo( |
| bounds, ctm, as_IFB(this)->affectsTransparentBlack(), &dst); |
| return dst; |
| } |
| } |
| |
| SkRect SkImageFilter::computeFastBounds(const SkRect& src) const { |
| if (0 == this->countInputs()) { |
| return src; |
| } |
| SkRect combinedBounds = this->getInput(0) ? this->getInput(0)->computeFastBounds(src) : src; |
| for (int i = 1; i < this->countInputs(); i++) { |
| const SkImageFilter* input = this->getInput(i); |
| if (input) { |
| combinedBounds.join(input->computeFastBounds(src)); |
| } else { |
| combinedBounds.join(src); |
| } |
| } |
| return combinedBounds; |
| } |
| |
| bool SkImageFilter::canComputeFastBounds() const { |
| if (as_IFB(this)->affectsTransparentBlack()) { |
| return false; |
| } |
| for (int i = 0; i < this->countInputs(); i++) { |
| const SkImageFilter* input = this->getInput(i); |
| if (input && !input->canComputeFastBounds()) { |
| return false; |
| } |
| } |
| return true; |
| } |
| |
| bool SkImageFilter::asAColorFilter(SkColorFilter** filterPtr) const { |
| SkASSERT(nullptr != filterPtr); |
| if (!this->isColorFilterNode(filterPtr)) { |
| return false; |
| } |
| if (nullptr != this->getInput(0) || (*filterPtr)->affectsTransparentBlack()) { |
| (*filterPtr)->unref(); |
| return false; |
| } |
| return true; |
| } |
| |
| sk_sp<SkImageFilter> SkImageFilter::MakeMatrixFilter(const SkMatrix& matrix, |
| SkFilterQuality filterQuality, |
| sk_sp<SkImageFilter> input) { |
| return SkMatrixImageFilter::Make(matrix, filterQuality, std::move(input)); |
| } |
| |
| sk_sp<SkImageFilter> SkImageFilter::makeWithLocalMatrix(const SkMatrix& matrix) const { |
| return SkLocalMatrixImageFilter::Make(matrix, this->refMe()); |
| } |
| |
| /////////////////////////////////////////////////////////////////////////////////////////////////// |
| // SkImageFilter_Base |
| /////////////////////////////////////////////////////////////////////////////////////////////////// |
| |
| static int32_t next_image_filter_unique_id() { |
| static std::atomic<int32_t> nextID{1}; |
| |
| int32_t id; |
| do { |
| id = nextID++; |
| } while (id == 0); |
| return id; |
| } |
| |
| SkImageFilter_Base::SkImageFilter_Base(sk_sp<SkImageFilter> const* inputs, |
| int inputCount, const CropRect* cropRect) |
| : fUsesSrcInput(false) |
| , fUniqueID(next_image_filter_unique_id()) { |
| fCropRect = cropRect ? *cropRect : CropRect(SkRect(), 0x0); |
| |
| fInputs.reset(inputCount); |
| |
| for (int i = 0; i < inputCount; ++i) { |
| if (!inputs[i] || as_IFB(inputs[i])->fUsesSrcInput) { |
| fUsesSrcInput = true; |
| } |
| fInputs[i] = inputs[i]; |
| } |
| } |
| |
| SkImageFilter_Base::~SkImageFilter_Base() { |
| SkImageFilterCache::Get()->purgeByImageFilter(this); |
| } |
| |
| bool SkImageFilter_Base::Common::unflatten(SkReadBuffer& buffer, int expectedCount) { |
| const int count = buffer.readInt(); |
| if (!buffer.validate(count >= 0)) { |
| return false; |
| } |
| if (!buffer.validate(expectedCount < 0 || count == expectedCount)) { |
| return false; |
| } |
| |
| SkASSERT(fInputs.empty()); |
| for (int i = 0; i < count; i++) { |
| fInputs.push_back(buffer.readBool() ? buffer.readImageFilter() : nullptr); |
| if (!buffer.isValid()) { |
| return false; |
| } |
| } |
| SkRect rect; |
| buffer.readRect(&rect); |
| if (!buffer.isValid() || !buffer.validate(SkIsValidRect(rect))) { |
| return false; |
| } |
| |
| uint32_t flags = buffer.readUInt(); |
| fCropRect = CropRect(rect, flags); |
| return buffer.isValid(); |
| } |
| |
| void SkImageFilter_Base::flatten(SkWriteBuffer& buffer) const { |
| buffer.writeInt(fInputs.count()); |
| for (int i = 0; i < fInputs.count(); i++) { |
| const SkImageFilter* input = this->getInput(i); |
| buffer.writeBool(input != nullptr); |
| if (input != nullptr) { |
| buffer.writeFlattenable(input); |
| } |
| } |
| buffer.writeRect(fCropRect.rect()); |
| buffer.writeUInt(fCropRect.flags()); |
| } |
| |
| sk_sp<SkSpecialImage> SkImageFilter_Base::filterImage(SkSpecialImage* src, const Context& context, |
| SkIPoint* offset) const { |
| SkASSERT(src && offset); |
| if (!context.isValid()) { |
| return nullptr; |
| } |
| |
| uint32_t srcGenID = fUsesSrcInput ? src->uniqueID() : 0; |
| const SkIRect srcSubset = fUsesSrcInput ? src->subset() : SkIRect::MakeWH(0, 0); |
| SkImageFilterCacheKey key(fUniqueID, context.ctm(), context.clipBounds(), srcGenID, srcSubset); |
| if (context.cache()) { |
| sk_sp<SkSpecialImage> result = context.cache()->get(key, offset); |
| if (result) { |
| return result; |
| } |
| } |
| |
| sk_sp<SkSpecialImage> result(this->onFilterImage(src, context, offset)); |
| |
| #if SK_SUPPORT_GPU |
| if (src->isTextureBacked() && result && !result->isTextureBacked()) { |
| // Keep the result on the GPU - this is still required for some |
| // image filters that don't support GPU in all cases |
| auto context = src->getContext(); |
| result = result->makeTextureImage(context); |
| } |
| #endif |
| |
| if (result && context.cache()) { |
| context.cache()->set(key, result.get(), *offset, this); |
| } |
| |
| return result; |
| } |
| |
| bool SkImageFilter_Base::canHandleComplexCTM() const { |
| // CropRects need to apply in the source coordinate system, but are not aware of complex CTMs |
| // when performing clipping. For a simple fix, any filter with a crop rect set cannot support |
| // complex CTMs until that's updated. |
| if (this->cropRectIsSet() || !this->onCanHandleComplexCTM()) { |
| return false; |
| } |
| const int count = this->countInputs(); |
| for (int i = 0; i < count; ++i) { |
| const SkImageFilter_Base* input = as_IFB(this->getInput(i)); |
| if (input && !input->canHandleComplexCTM()) { |
| return false; |
| } |
| } |
| return true; |
| } |
| |
| void SkImageFilter::CropRect::applyTo(const SkIRect& imageBounds, const SkMatrix& ctm, |
| bool embiggen, SkIRect* cropped) const { |
| *cropped = imageBounds; |
| if (fFlags) { |
| SkRect devCropR; |
| ctm.mapRect(&devCropR, fRect); |
| SkIRect devICropR = devCropR.roundOut(); |
| |
| // Compute the left/top first, in case we need to modify the right/bottom for a missing edge |
| if (fFlags & kHasLeft_CropEdge) { |
| if (embiggen || devICropR.fLeft > cropped->fLeft) { |
| cropped->fLeft = devICropR.fLeft; |
| } |
| } else { |
| devICropR.fRight = Sk32_sat_add(cropped->fLeft, devICropR.width()); |
| } |
| if (fFlags & kHasTop_CropEdge) { |
| if (embiggen || devICropR.fTop > cropped->fTop) { |
| cropped->fTop = devICropR.fTop; |
| } |
| } else { |
| devICropR.fBottom = Sk32_sat_add(cropped->fTop, devICropR.height()); |
| } |
| if (fFlags & kHasWidth_CropEdge) { |
| if (embiggen || devICropR.fRight < cropped->fRight) { |
| cropped->fRight = devICropR.fRight; |
| } |
| } |
| if (fFlags & kHasHeight_CropEdge) { |
| if (embiggen || devICropR.fBottom < cropped->fBottom) { |
| cropped->fBottom = devICropR.fBottom; |
| } |
| } |
| } |
| } |
| |
| bool SkImageFilter_Base::applyCropRect(const Context& ctx, const SkIRect& srcBounds, |
| SkIRect* dstBounds) const { |
| SkIRect tmpDst = this->onFilterNodeBounds(srcBounds, ctx.ctm(), kForward_MapDirection, nullptr); |
| fCropRect.applyTo(tmpDst, ctx.ctm(), this->affectsTransparentBlack(), dstBounds); |
| // Intersect against the clip bounds, in case the crop rect has |
| // grown the bounds beyond the original clip. This can happen for |
| // example in tiling, where the clip is much smaller than the filtered |
| // primitive. If we didn't do this, we would be processing the filter |
| // at the full crop rect size in every tile. |
| return dstBounds->intersect(ctx.clipBounds()); |
| } |
| |
| // Return a larger (newWidth x newHeight) copy of 'src' with black padding |
| // around it. |
| static sk_sp<SkSpecialImage> pad_image(SkSpecialImage* src, |
| const SkImageFilter_Base::OutputProperties& outProps, |
| int newWidth, int newHeight, int offX, int offY) { |
| // We would like to operate in the source's color space (so that we return an "identical" |
| // image, other than the padding. To achieve that, we'd create new output properties: |
| // |
| // SkImageFilter::OutputProperties outProps(src->getColorSpace()); |
| // |
| // That fails in at least two ways. For formats that are texturable but not renderable (like |
| // F16 on some ES implementations), we can't create a surface to do the work. For sRGB, images |
| // may be tagged with an sRGB color space (which leads to an sRGB config in makeSurface). But |
| // the actual config of that sRGB image on a device with no sRGB support is non-sRGB. |
| // |
| // Rather than try to special case these situations, we execute the image padding in the |
| // destination color space. This should not affect the output of the DAG in (almost) any case, |
| // because the result of this call is going to be used as an input, where it would have been |
| // switched to the destination space anyway. The one exception would be a filter that expected |
| // to consume unclamped F16 data, but the padded version of the image is pre-clamped to 8888. |
| // We can revisit this logic if that ever becomes an actual problem. |
| sk_sp<SkSpecialSurface> surf(src->makeSurface(outProps, SkISize::Make(newWidth, newHeight))); |
| if (!surf) { |
| return nullptr; |
| } |
| |
| SkCanvas* canvas = surf->getCanvas(); |
| SkASSERT(canvas); |
| |
| canvas->clear(0x0); |
| |
| src->draw(canvas, offX, offY, nullptr); |
| |
| return surf->makeImageSnapshot(); |
| } |
| |
| sk_sp<SkSpecialImage> SkImageFilter_Base::applyCropRectAndPad(const Context& ctx, |
| SkSpecialImage* src, |
| SkIPoint* srcOffset, |
| SkIRect* bounds) const { |
| const SkIRect srcBounds = SkIRect::MakeXYWH(srcOffset->x(), srcOffset->y(), |
| src->width(), src->height()); |
| |
| if (!this->applyCropRect(ctx, srcBounds, bounds)) { |
| return nullptr; |
| } |
| |
| if (srcBounds.contains(*bounds)) { |
| return sk_sp<SkSpecialImage>(SkRef(src)); |
| } else { |
| sk_sp<SkSpecialImage> img(pad_image(src, ctx.outputProperties(), |
| bounds->width(), bounds->height(), |
| Sk32_sat_sub(srcOffset->x(), bounds->x()), |
| Sk32_sat_sub(srcOffset->y(), bounds->y()))); |
| *srcOffset = SkIPoint::Make(bounds->x(), bounds->y()); |
| return img; |
| } |
| } |
| |
| SkIRect SkImageFilter_Base::onFilterBounds(const SkIRect& src, const SkMatrix& ctm, |
| MapDirection dir, const SkIRect* inputRect) const { |
| if (this->countInputs() < 1) { |
| return src; |
| } |
| |
| SkIRect totalBounds; |
| for (int i = 0; i < this->countInputs(); ++i) { |
| const SkImageFilter* filter = this->getInput(i); |
| SkIRect rect = filter ? filter->filterBounds(src, ctm, dir, inputRect) : src; |
| if (0 == i) { |
| totalBounds = rect; |
| } else { |
| totalBounds.join(rect); |
| } |
| } |
| |
| return totalBounds; |
| } |
| |
| SkIRect SkImageFilter_Base::onFilterNodeBounds(const SkIRect& src, const SkMatrix&, |
| MapDirection, const SkIRect*) const { |
| return src; |
| } |
| |
| sk_sp<SkSpecialImage> SkImageFilter_Base::filterInput(int index, |
| SkSpecialImage* src, |
| const Context& ctx, |
| SkIPoint* offset) const { |
| const SkImageFilter* input = this->getInput(index); |
| if (!input) { |
| return sk_sp<SkSpecialImage>(SkRef(src)); |
| } |
| |
| sk_sp<SkSpecialImage> result(as_IFB(input)->filterImage(src, this->mapContext(ctx), offset)); |
| |
| SkASSERT(!result || src->isTextureBacked() == result->isTextureBacked()); |
| |
| return result; |
| } |
| |
| SkImageFilter_Base::Context SkImageFilter_Base::mapContext(const Context& ctx) const { |
| SkIRect clipBounds = this->onFilterNodeBounds(ctx.clipBounds(), ctx.ctm(), |
| MapDirection::kReverse_MapDirection, |
| &ctx.clipBounds()); |
| return Context(ctx.ctm(), clipBounds, ctx.cache(), ctx.outputProperties()); |
| } |
| |
| #if SK_SUPPORT_GPU |
| sk_sp<SkSpecialImage> SkImageFilter_Base::DrawWithFP(GrRecordingContext* context, |
| std::unique_ptr<GrFragmentProcessor> fp, |
| const SkIRect& bounds, |
| const OutputProperties& outputProperties, |
| GrProtected isProtected) { |
| GrPaint paint; |
| paint.addColorFragmentProcessor(std::move(fp)); |
| paint.setPorterDuffXPFactory(SkBlendMode::kSrc); |
| |
| sk_sp<SkColorSpace> colorSpace = sk_ref_sp(outputProperties.colorSpace()); |
| GrColorType colorType = SkColorTypeToGrColorType(outputProperties.colorType()); |
| sk_sp<GrRenderTargetContext> renderTargetContext( |
| context->priv().makeDeferredRenderTargetContext( |
| SkBackingFit::kApprox, |
| bounds.width(), |
| bounds.height(), |
| colorType, |
| std::move(colorSpace), |
| 1, |
| GrMipMapped::kNo, |
| kBottomLeft_GrSurfaceOrigin, |
| nullptr, |
| SkBudgeted::kYes, |
| isProtected)); |
| if (!renderTargetContext) { |
| return nullptr; |
| } |
| |
| SkIRect dstIRect = SkIRect::MakeWH(bounds.width(), bounds.height()); |
| SkRect srcRect = SkRect::Make(bounds); |
| SkRect dstRect = SkRect::MakeWH(srcRect.width(), srcRect.height()); |
| GrFixedClip clip(dstIRect); |
| renderTargetContext->fillRectToRect(clip, std::move(paint), GrAA::kNo, SkMatrix::I(), dstRect, |
| srcRect); |
| |
| return SkSpecialImage::MakeDeferredFromGpu( |
| context, dstIRect, kNeedNewImageUniqueID_SpecialImage, |
| renderTargetContext->asTextureProxyRef(), |
| renderTargetContext->colorSpaceInfo().refColorSpace()); |
| } |
| |
| sk_sp<SkSpecialImage> SkImageFilter_Base::ImageToColorSpace(SkSpecialImage* src, |
| const OutputProperties& outProps) { |
| // There are several conditions that determine if we actually need to convert the source to the |
| // destination's color space. Rather than duplicate that logic here, just try to make an xform |
| // object. If that produces something, then both are tagged, and the source is in a different |
| // gamut than the dest. There is some overhead to making the xform, but those are cached, and |
| // if we get one back, that means we're about to use it during the conversion anyway. |
| auto colorSpaceXform = GrColorSpaceXform::Make(src->getColorSpace(), src->alphaType(), |
| outProps.colorSpace(), kPremul_SkAlphaType); |
| |
| if (!colorSpaceXform) { |
| // No xform needed, just return the original image |
| return sk_ref_sp(src); |
| } |
| |
| sk_sp<SkSpecialSurface> surf(src->makeSurface(outProps, |
| SkISize::Make(src->width(), src->height()))); |
| if (!surf) { |
| return sk_ref_sp(src); |
| } |
| |
| SkCanvas* canvas = surf->getCanvas(); |
| SkASSERT(canvas); |
| SkPaint p; |
| p.setBlendMode(SkBlendMode::kSrc); |
| src->draw(canvas, 0, 0, &p); |
| return surf->makeImageSnapshot(); |
| } |
| #endif |
| |
| // In repeat mode, when we are going to sample off one edge of the srcBounds we require the |
| // opposite side be preserved. |
| SkIRect SkImageFilter_Base::DetermineRepeatedSrcBound(const SkIRect& srcBounds, |
| const SkIVector& filterOffset, |
| const SkISize& filterSize, |
| const SkIRect& originalSrcBounds) { |
| SkIRect tmp = srcBounds; |
| tmp.adjust(-filterOffset.fX, -filterOffset.fY, |
| filterSize.fWidth - filterOffset.fX, filterSize.fHeight - filterOffset.fY); |
| |
| if (tmp.fLeft < originalSrcBounds.fLeft || tmp.fRight > originalSrcBounds.fRight) { |
| tmp.fLeft = originalSrcBounds.fLeft; |
| tmp.fRight = originalSrcBounds.fRight; |
| } |
| if (tmp.fTop < originalSrcBounds.fTop || tmp.fBottom > originalSrcBounds.fBottom) { |
| tmp.fTop = originalSrcBounds.fTop; |
| tmp.fBottom = originalSrcBounds.fBottom; |
| } |
| |
| return tmp; |
| } |
| |
| void SkImageFilter_Base::PurgeCache() { |
| SkImageFilterCache::Get()->purge(); |
| } |
| |
| static sk_sp<SkImageFilter> apply_ctm_to_filter(sk_sp<SkImageFilter> input, const SkMatrix& ctm, |
| SkMatrix* remainder, bool asBackdrop) { |
| if (ctm.isScaleTranslate() || as_IFB(input)->canHandleComplexCTM()) { |
| // The filter supports the CTM, so leave it as-is and 'remainder' stores the whole CTM |
| *remainder = ctm; |
| return input; |
| } |
| |
| // We have a complex CTM and a filter that can't support them, so it needs to use the matrix |
| // transform filter that resamples the image contents. Decompose the simple portion of the ctm |
| // into 'remainder' |
| SkMatrix ctmToEmbed; |
| SkSize scale; |
| if (ctm.decomposeScale(&scale, &ctmToEmbed)) { |
| // decomposeScale splits ctm into scale * ctmToEmbed, so bake ctmToEmbed into DAG |
| // with a matrix filter and return scale as the remaining matrix for the real CTM. |
| remainder->setScale(scale.fWidth, scale.fHeight); |
| |
| // ctmToEmbed is passed to SkMatrixImageFilter, which performs its transforms as if it were |
| // a pre-transformation before applying the image-filter context's CTM. In this case, we |
| // need ctmToEmbed to be a post-transformation (i.e. after the scale matrix since |
| // decomposeScale produces ctm = ctmToEmbed * scale). Giving scale^-1 * ctmToEmbed * scale |
| // to the matrix filter achieves this effect. |
| // TODO (michaelludwig) - When the original root node of a filter can be drawn directly to a |
| // device using ctmToEmbed, this abuse of SkMatrixImageFilter can go away. |
| ctmToEmbed.preScale(scale.fWidth, scale.fHeight); |
| ctmToEmbed.postScale(1.f / scale.fWidth, 1.f / scale.fHeight); |
| } else { |
| // Unable to decompose |
| // FIXME Ideally we'd embed the entire CTM as part of the matrix image filter, but |
| // the device <-> src bounds calculations for filters are very brittle under perspective, |
| // and can easily run into precision issues (wrong bounds that clip), or performance issues |
| // (producing large source-space images where 80% of the image is compressed into a few |
| // device pixels). A longer term solution for perspective-space image filtering is needed |
| // see skbug.com/9074 |
| if (ctm.hasPerspective()) { |
| *remainder = ctm; |
| return input; |
| } |
| |
| ctmToEmbed = ctm; |
| remainder->setIdentity(); |
| } |
| |
| if (asBackdrop) { |
| // In the backdrop case we also have to transform the existing device-space buffer content |
| // into the source coordinate space prior to the filtering. Non-backdrop filter inputs are |
| // already in the source space because of how the layer is drawn by SkCanvas. |
| SkMatrix invEmbed; |
| if (ctmToEmbed.invert(&invEmbed)) { |
| input = SkComposeImageFilter::Make(std::move(input), |
| SkMatrixImageFilter::Make(invEmbed, kLow_SkFilterQuality, nullptr)); |
| } |
| } |
| return SkMatrixImageFilter::Make(ctmToEmbed, kLow_SkFilterQuality, input); |
| } |
| |
| sk_sp<SkImageFilter> SkImageFilter_Base::applyCTM(const SkMatrix& ctm, SkMatrix* remainder) const { |
| return apply_ctm_to_filter(this->refMe(), ctm, remainder, false); |
| } |
| |
| sk_sp<SkImageFilter> SkImageFilter_Base::applyCTMForBackdrop(const SkMatrix& ctm, |
| SkMatrix* remainder) const { |
| return apply_ctm_to_filter(this->refMe(), ctm, remainder, true); |
| } |