skia2/include/codec/SkCodec.h

894 lines
34 KiB
C
Raw Normal View History

/*
* Copyright 2015 Google Inc.
*
* Use of this source code is governed by a BSD-style license that can be
* found in the LICENSE file.
*/
#ifndef SkCodec_DEFINED
#define SkCodec_DEFINED
#include "../private/SkTemplates.h"
#include "SkColor.h"
#include "SkEncodedImageFormat.h"
#include "SkEncodedInfo.h"
#include "SkImageInfo.h"
#include "SkSize.h"
#include "SkStream.h"
#include "SkTypes.h"
#include "SkYUVSizeInfo.h"
Add support for multiple frames in SkCodec Add an interface to decode frames beyond the first in SkCodec, and add an implementation for SkGifCodec. Add getFrameData to SkCodec. This method reads ahead in the stream to return a vector containing meta data about each frame in the image. This is not required in order to decode frames beyond the first, but it allows a client to learn extra information: - how long the frame should be displayed - whether a frame should be blended with a prior frame, allowing the client to provide the prior frame to speed up decoding Add a new fields to SkCodec::Options: - fFrameIndex - fHasPriorFrame The API is designed so that SkCodec never caches frames. If a client wants a frame beyond the first, they specify the frame in Options.fFrameIndex. If the client does not have the frame's required frame (the frame that this frame must be blended on top of) cached, they pass false for Options.fHasPriorFrame. Unless the frame is independent, the codec will then recursively decode all frames necessary to decode fFrameIndex. If the client has the required frame cached, they can put it in the dst they pass to the codec, and the codec will only draw fFrameIndex onto it. Replace SkGifCodec's scanline decoding support with progressive decoding, and update the tests accordingly. Implement new APIs in SkGifCodec. Instead of using gif_lib, use GIFImageReader, imported from Chromium (along with its copyright headers) with the following changes: - SkGifCodec is now the client - Replace blink types - Combine GIFColorMap::buildTable and ::getTable into a method that creates and returns an SkColorTable - Input comes from an SkStream, instead of a SegmentReader. Add SkStreamBuffer, which buffers the (potentially partial) stream in order to decode progressively. (FIXME: This requires copying data that previously was read directly from the SegmentReader. Does this hurt performance? If so, can we fix it?) - Remove UMA code - Instead of reporting screen width and height to the client, allow the client to query for it - Fail earlier if the first frame AND screen have size of zero - Compute required previous frame when adding a new one - Move GIFParseQuery from GIFImageDecoder to GIFImageReader - Allow parsing up to a specific frame (to skip parsing the rest of the stream if a client only wants the first frame) - Compute whether the first frame has alpha and supports index 8, to create the SkImageInfo. This happens before reporting that the size has been decoded. Add GIFImageDecoder::haveDecodedRow to SkGifCodec, imported from Chromium (along with its copyright header), with the following changes: - Add support for sampling - Use the swizzler - Keep track of the rows decoded - Do *not* keep track of whether we've seen alpha Remove SkCodec::kOutOfOrder_SkScanlineOrder, which was only used by GIF scanline decoding. Call onRewind even if there is no stream (SkGifCodec needs to clear its decoded state so it will decode from the beginning). Add a method to SkSwizzler to access the offset into the dst, taking subsetting into account. Add a GM that animates a GIF. Add tests for the new APIs. *** Behavior changes: * Previously, we reported that an image with a subset frame and no transparent index was opaque and used the background index (if present) to fill the background. This is necessary in order to support index 8, but it does not match viewers/browsers I have seen. Examples: - Chromium and Gimp render the background transparent - Firefox, Safari, Linux Image Viewer, Safari Preview clip to the frame (for a single frame image) This CL matches Chromium's behavior and renders the background transparent. This allows us to have consistent behavior across products and simplifies the code (relative to what we would have to do to continue the old behavior on Android). It also means that we will no longer support index 8 for some GIFs. * Stop checking for GIFSTAMP - all GIFs should be either 89a or 87a. This matches Chromium. I suspect that bugs would have been reported if valid GIFs started with "GIFVER" instead of "GIF89a" or "GIF87a" (but did not decode in Chromium). *** Future work not included in this CL: * Move some checks out of haveDecodedRow, since they are the same for the entire frame e.g. - intersecting the frameRect with the full image size - whether there is a color table * Change when we write transparent pixels - In some cases, Chromium deemed this unnecessary, but I suspect it is slower than the fallback case. There will continue to be cases where we should *not* write them, but for e.g. the first pass where we have already cleared to transparent (which we may also be able to skip) writing the transparent pixels will not make anything incorrect. * Report color type and alpha type per frame - Depending on alpha values, disposal methods, frame rects, etc, subsequent frames may have different properties than the first. * Skip copies of the encoded data - We copy the encoded data in case the stream is one that cannot be rewound, so we can parse and then decode (possibly not immediately). For some input streams, this is unnecessary. - I was concerned this cause a performance regression, but on average the new code is faster than the old for the images I tested [1]. - It may cause a performance regression for Chromium, though, where we can always move back in the stream, so this should be addressed. Design doc: https://docs.google.com/a/google.com/document/d/12Qhf9T92MWfdWujQwCIjhCO3sw6pTJB5pJBwDM1T7Kc/ [1] https://docs.google.com/a/google.com/spreadsheets/d/19V-t9BfbFw5eiwBTKA1qOBkZbchjlTC5EIz6HFy-6RI/ GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=2045293002 Review-Url: https://codereview.chromium.org/2045293002
2016-10-24 16:03:26 +00:00
#include <vector>
class SkColorSpace;
class SkColorSpaceXform;
class SkData;
class SkPngChunkReader;
class SkSampler;
namespace DM {
Make SkPngCodec decode progressively. This is a step towards using SkCodec in Chromium, where progressive decoding is necessary. Switch from using png_read_row (which expects all the data to be available) to png_process_data, which uses callbacks when rows are available. Create a new API for SkCodec, which supports progressive decoding and scanline decoding. Future changes will switch the other clients off of startScanlineDecode and get/skip-Scanlines to the new API. Remove SkCodec::kNone_ScanlineOrder, which was only used for interlaced PNG images. In the new API, interlaced PNG fits kTopDown. Also remove updateCurrScanline(), which was only used by the old implementation for interlaced PNG. DMSrcSink: - In CodecSrc::kScanline_Mode, use the new method for scanline decoding for the supported formats (just PNG and PNG-in-ICO for now). fuzz.cpp: - Remove reference to kNone_ScanlineOrder SkCodec: - Add new APIs: - startIncrementalDecode - incrementalDecode - Remove kNone_SkScanlineOrder and updateCurrScanline() - Set fDstInfo and fOptions in getPixels(). This may not be necessary for all implementations, but it simplifies things for SkPngCodec. SkPngCodec: - Implement new APIs - Switch from sk_read_fn/png_read_row etc to png_process_data - Expand AutoCleanPng's role to decode the header and create the SkPngCodec - Make the interlaced PNG decoder report how many lines were initialized during an incomplete decode SkIcoCodec: - Implement the new APIs; supported for PNG in ICO SkSampledCodec: - Call the new method for decoding scanlines, and fall back to the old method if the new version is unimplemented - Remove references to kNone_SkScanlineOrder tests/CodecPartial: - Add a test which decodes part of an image, then finishes the decode, and compares it to the straightforward method tests/CodecTest: - Add a test which decodes all scanlines using the new method - Repurpose the Codec_stripes test to decode using the new method in sections rather than all at once - In the method check(), add a parameter for whether the image supports the new method of scanline decoding, and be explicit about whether an image supports incomplete - Test incomplete PNG decodes. We should have been doing it anyway for non-interlaced (except for an image that is too small - one row), but the new method supports interlaced incomplete as well - Make test_invalid_parameters test the new method - Add a test to ensure that it's safe to fall back to scanline decoding without rewinding BUG=skia:4211 The new version was generally faster than the old version (but not significantly so). Some raw performance differences can be found at https://docs.google.com/a/google.com/spreadsheets/d/1Gis3aRCEa72qBNDRMgGDg3jD-pMgO-FXldlNF9ejo4o/ Design doc can be found at https://docs.google.com/a/google.com/document/d/11Mn8-ePDKwVEMCjs3nWwSjxcSpJ_Cu8DF57KNtUmgLM/ GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1997703003 Review-Url: https://codereview.chromium.org/1997703003
2016-09-16 15:20:38 +00:00
class CodecSrc;
class ColorCodecSrc;
}
class ColorCodecBench;
/**
* Abstraction layer directly on top of an image codec.
*/
class SK_API SkCodec : SkNoncopyable {
public:
/**
* Minimum number of bytes that must be buffered in SkStream input.
*
* An SkStream passed to NewFromStream must be able to use this many
* bytes to determine the image type. Then the same SkStream must be
* passed to the correct decoder to read from the beginning.
*
* This can be accomplished by implementing peek() to support peeking
* this many bytes, or by implementing rewind() to be able to rewind()
* after reading this many bytes.
*/
static size_t MinBufferedBytesNeeded();
/**
* If this stream represents an encoded image that we know how to decode,
* return an SkCodec that can decode it. Otherwise return NULL.
*
* As stated above, this call must be able to peek or read
* MinBufferedBytesNeeded to determine the correct format, and then start
* reading from the beginning. First it will attempt to peek, and it
* assumes that if less than MinBufferedBytesNeeded bytes (but more than
* zero) are returned, this is because the stream is shorter than this,
* so falling back to reading would not provide more data. If peek()
* returns zero bytes, this call will instead attempt to read(). This
* will require that the stream can be rewind()ed.
*
* If SkPngChunkReader is not NULL, take a ref and pass it to libpng if
* the image is a png.
*
* If the SkPngChunkReader is not NULL then:
* If the image is not a PNG, the SkPngChunkReader will be ignored.
* If the image is a PNG, the SkPngChunkReader will be reffed.
* If the PNG has unknown chunks, the SkPngChunkReader will be used
* to handle these chunks. SkPngChunkReader will be called to read
* any unknown chunk at any point during the creation of the codec
* or the decode. Note that if SkPngChunkReader fails to read a
* chunk, this could result in a failure to create the codec or a
* failure to decode the image.
* If the PNG does not contain unknown chunks, the SkPngChunkReader
* will not be used or modified.
*
* If NULL is returned, the stream is deleted immediately. Otherwise, the
* SkCodec takes ownership of it, and will delete it when done with it.
*/
static SkCodec* NewFromStream(SkStream*, SkPngChunkReader* = NULL);
/**
* If this data represents an encoded image that we know how to decode,
* return an SkCodec that can decode it. Otherwise return NULL.
*
* If the SkPngChunkReader is not NULL then:
* If the image is not a PNG, the SkPngChunkReader will be ignored.
* If the image is a PNG, the SkPngChunkReader will be reffed.
* If the PNG has unknown chunks, the SkPngChunkReader will be used
* to handle these chunks. SkPngChunkReader will be called to read
* any unknown chunk at any point during the creation of the codec
* or the decode. Note that if SkPngChunkReader fails to read a
* chunk, this could result in a failure to create the codec or a
* failure to decode the image.
* If the PNG does not contain unknown chunks, the SkPngChunkReader
* will not be used or modified.
*/
static SkCodec* NewFromData(sk_sp<SkData>, SkPngChunkReader* = NULL);
static SkCodec* NewFromData(SkData* data, SkPngChunkReader* reader) {
return NewFromData(sk_ref_sp(data), reader);
}
virtual ~SkCodec();
/**
* Return the ImageInfo associated with this codec.
*/
Merge SkCodec with SkScanlineDecoder Benefits: - This mimics other decoding APIs (including the ones SkCodec relies on, e.g. a png_struct, which can be used to decode an entire image or one line at a time). - It allows a client to ask us to do what we can do efficiently - i.e. start from encoded data and either decode the whole thing or scanlines. - It removes the duplicate methods which appeared in both SkCodec and SkScanlineDecoder (some of which, e.g. in SkJpegScanlineDecoder, just call fCodec->sameMethod()). - It simplifies moving more checks into the base class (e.g. the examples in skbug.com/4284). BUG=skia:4175 BUG=skia:4284 ===================================================================== SkScanlineDecoder.h/.cpp: Removed. SkCodec.h/.cpp: Add methods, enums, and variables which were previously in SkScanlineDecoder. Default fCurrScanline to -1, as a sentinel that start has not been called. General changes: Convert SkScanlineDecoders to SkCodecs. General changes in SkCodec subclasses: Merge SkScanlineDecoder implementation into SkCodec. Most (all?) owned an SkCodec, so they now call this-> instead of fCodec->. SkBmpCodec.h/.cpp: Replace the unused rowOrder method with an override for onGetScanlineOrder. Make getDstRow const, since it is called by onGetY, which is const. SkCodec_libpng.h/.cpp: Make SkPngCodec an abstract class, with two subclasses which handle scanline decoding separately (they share code for decoding the entire image). Reimplement onReallyHasAlpha so that it can return the most recent result (e.g. after a scanline decode which only decoded part of the image) or a better answer (e.g. if the whole image is known to be opaque). Compute fNumberPasses early, so we know which subclass to instantiate. Make SkPngInterlaceScanlineDecoder use the base class' fCurrScanline rather than a separate variable. CodexTest.cpp: Add tests for the state changes in SkCodec (need to call start before decoding scanlines; calling getPixels means that start will need to be called again before decoding more scanlines). Add a test which decodes in stripes, currently only used for an interlaced PNG. TODO: Add tests for onReallyHasAlpha. Review URL: https://codereview.chromium.org/1365313002
2015-09-30 15:57:13 +00:00
const SkImageInfo& getInfo() const { return fSrcInfo; }
const SkEncodedInfo& getEncodedInfo() const { return fEncodedInfo; }
enum Origin {
kTopLeft_Origin = 1, // Default
kTopRight_Origin = 2, // Reflected across y-axis
kBottomRight_Origin = 3, // Rotated 180
kBottomLeft_Origin = 4, // Reflected across x-axis
kLeftTop_Origin = 5, // Reflected across x-axis, Rotated 90 CCW
kRightTop_Origin = 6, // Rotated 90 CW
kRightBottom_Origin = 7, // Reflected across x-axis, Rotated 90 CW
kLeftBottom_Origin = 8, // Rotated 90 CCW
kDefault_Origin = kTopLeft_Origin,
kLast_Origin = kLeftBottom_Origin,
};
/**
* Returns the image orientation stored in the EXIF data.
* If there is no EXIF data, or if we cannot read the EXIF data, returns kTopLeft.
*/
Origin getOrigin() const { return fOrigin; }
/**
* Return a size that approximately supports the desired scale factor.
* The codec may not be able to scale efficiently to the exact scale
* factor requested, so return a size that approximates that scale.
* The returned value is the codec's suggestion for the closest valid
* scale that it can natively support
*/
SkISize getScaledDimensions(float desiredScale) const {
// Negative and zero scales are errors.
SkASSERT(desiredScale > 0.0f);
if (desiredScale <= 0.0f) {
return SkISize::Make(0, 0);
}
// Upscaling is not supported. Return the original size if the client
// requests an upscale.
if (desiredScale >= 1.0f) {
return this->getInfo().dimensions();
}
return this->onGetScaledDimensions(desiredScale);
}
/**
* Return (via desiredSubset) a subset which can decoded from this codec,
* or false if this codec cannot decode subsets or anything similar to
* desiredSubset.
*
* @param desiredSubset In/out parameter. As input, a desired subset of
* the original bounds (as specified by getInfo). If true is returned,
* desiredSubset may have been modified to a subset which is
* supported. Although a particular change may have been made to
* desiredSubset to create something supported, it is possible other
* changes could result in a valid subset.
* If false is returned, desiredSubset's value is undefined.
* @return true if this codec supports decoding desiredSubset (as
* returned, potentially modified)
*/
bool getValidSubset(SkIRect* desiredSubset) const {
return this->onGetValidSubset(desiredSubset);
}
/**
* Format of the encoded data.
*/
SkEncodedImageFormat getEncodedFormat() const { return this->onGetEncodedFormat(); }
/**
* Used to describe the result of a call to getPixels().
*
* Result is the union of possible results from subclasses.
*/
enum Result {
/**
* General return value for success.
*/
kSuccess,
/**
* The input is incomplete. A partial image was generated.
*/
kIncompleteInput,
/**
* The generator cannot convert to match the request, ignoring
* dimensions.
*/
kInvalidConversion,
/**
* The generator cannot scale to requested size.
*/
kInvalidScale,
/**
* Parameters (besides info) are invalid. e.g. NULL pixels, rowBytes
* too small, etc.
*/
kInvalidParameters,
/**
* The input did not contain a valid image.
*/
kInvalidInput,
/**
* Fulfilling this request requires rewinding the input, which is not
* supported for this input.
*/
kCouldNotRewind,
Merge SkCodec with SkScanlineDecoder Benefits: - This mimics other decoding APIs (including the ones SkCodec relies on, e.g. a png_struct, which can be used to decode an entire image or one line at a time). - It allows a client to ask us to do what we can do efficiently - i.e. start from encoded data and either decode the whole thing or scanlines. - It removes the duplicate methods which appeared in both SkCodec and SkScanlineDecoder (some of which, e.g. in SkJpegScanlineDecoder, just call fCodec->sameMethod()). - It simplifies moving more checks into the base class (e.g. the examples in skbug.com/4284). BUG=skia:4175 BUG=skia:4284 ===================================================================== SkScanlineDecoder.h/.cpp: Removed. SkCodec.h/.cpp: Add methods, enums, and variables which were previously in SkScanlineDecoder. Default fCurrScanline to -1, as a sentinel that start has not been called. General changes: Convert SkScanlineDecoders to SkCodecs. General changes in SkCodec subclasses: Merge SkScanlineDecoder implementation into SkCodec. Most (all?) owned an SkCodec, so they now call this-> instead of fCodec->. SkBmpCodec.h/.cpp: Replace the unused rowOrder method with an override for onGetScanlineOrder. Make getDstRow const, since it is called by onGetY, which is const. SkCodec_libpng.h/.cpp: Make SkPngCodec an abstract class, with two subclasses which handle scanline decoding separately (they share code for decoding the entire image). Reimplement onReallyHasAlpha so that it can return the most recent result (e.g. after a scanline decode which only decoded part of the image) or a better answer (e.g. if the whole image is known to be opaque). Compute fNumberPasses early, so we know which subclass to instantiate. Make SkPngInterlaceScanlineDecoder use the base class' fCurrScanline rather than a separate variable. CodexTest.cpp: Add tests for the state changes in SkCodec (need to call start before decoding scanlines; calling getPixels means that start will need to be called again before decoding more scanlines). Add a test which decodes in stripes, currently only used for an interlaced PNG. TODO: Add tests for onReallyHasAlpha. Review URL: https://codereview.chromium.org/1365313002
2015-09-30 15:57:13 +00:00
/**
* This method is not implemented by this codec.
* FIXME: Perhaps this should be kUnsupported?
*/
kUnimplemented,
};
/**
* Whether or not the memory passed to getPixels is zero initialized.
*/
enum ZeroInitialized {
/**
* The memory passed to getPixels is zero initialized. The SkCodec
* may take advantage of this by skipping writing zeroes.
*/
kYes_ZeroInitialized,
/**
* The memory passed to getPixels has not been initialized to zero,
* so the SkCodec must write all zeroes to memory.
*
* This is the default. It will be used if no Options struct is used.
*/
kNo_ZeroInitialized,
};
/**
* Additional options to pass to getPixels.
*/
struct Options {
Options()
: fZeroInitialized(kNo_ZeroInitialized)
Add support for multiple frames in SkCodec Add an interface to decode frames beyond the first in SkCodec, and add an implementation for SkGifCodec. Add getFrameData to SkCodec. This method reads ahead in the stream to return a vector containing meta data about each frame in the image. This is not required in order to decode frames beyond the first, but it allows a client to learn extra information: - how long the frame should be displayed - whether a frame should be blended with a prior frame, allowing the client to provide the prior frame to speed up decoding Add a new fields to SkCodec::Options: - fFrameIndex - fHasPriorFrame The API is designed so that SkCodec never caches frames. If a client wants a frame beyond the first, they specify the frame in Options.fFrameIndex. If the client does not have the frame's required frame (the frame that this frame must be blended on top of) cached, they pass false for Options.fHasPriorFrame. Unless the frame is independent, the codec will then recursively decode all frames necessary to decode fFrameIndex. If the client has the required frame cached, they can put it in the dst they pass to the codec, and the codec will only draw fFrameIndex onto it. Replace SkGifCodec's scanline decoding support with progressive decoding, and update the tests accordingly. Implement new APIs in SkGifCodec. Instead of using gif_lib, use GIFImageReader, imported from Chromium (along with its copyright headers) with the following changes: - SkGifCodec is now the client - Replace blink types - Combine GIFColorMap::buildTable and ::getTable into a method that creates and returns an SkColorTable - Input comes from an SkStream, instead of a SegmentReader. Add SkStreamBuffer, which buffers the (potentially partial) stream in order to decode progressively. (FIXME: This requires copying data that previously was read directly from the SegmentReader. Does this hurt performance? If so, can we fix it?) - Remove UMA code - Instead of reporting screen width and height to the client, allow the client to query for it - Fail earlier if the first frame AND screen have size of zero - Compute required previous frame when adding a new one - Move GIFParseQuery from GIFImageDecoder to GIFImageReader - Allow parsing up to a specific frame (to skip parsing the rest of the stream if a client only wants the first frame) - Compute whether the first frame has alpha and supports index 8, to create the SkImageInfo. This happens before reporting that the size has been decoded. Add GIFImageDecoder::haveDecodedRow to SkGifCodec, imported from Chromium (along with its copyright header), with the following changes: - Add support for sampling - Use the swizzler - Keep track of the rows decoded - Do *not* keep track of whether we've seen alpha Remove SkCodec::kOutOfOrder_SkScanlineOrder, which was only used by GIF scanline decoding. Call onRewind even if there is no stream (SkGifCodec needs to clear its decoded state so it will decode from the beginning). Add a method to SkSwizzler to access the offset into the dst, taking subsetting into account. Add a GM that animates a GIF. Add tests for the new APIs. *** Behavior changes: * Previously, we reported that an image with a subset frame and no transparent index was opaque and used the background index (if present) to fill the background. This is necessary in order to support index 8, but it does not match viewers/browsers I have seen. Examples: - Chromium and Gimp render the background transparent - Firefox, Safari, Linux Image Viewer, Safari Preview clip to the frame (for a single frame image) This CL matches Chromium's behavior and renders the background transparent. This allows us to have consistent behavior across products and simplifies the code (relative to what we would have to do to continue the old behavior on Android). It also means that we will no longer support index 8 for some GIFs. * Stop checking for GIFSTAMP - all GIFs should be either 89a or 87a. This matches Chromium. I suspect that bugs would have been reported if valid GIFs started with "GIFVER" instead of "GIF89a" or "GIF87a" (but did not decode in Chromium). *** Future work not included in this CL: * Move some checks out of haveDecodedRow, since they are the same for the entire frame e.g. - intersecting the frameRect with the full image size - whether there is a color table * Change when we write transparent pixels - In some cases, Chromium deemed this unnecessary, but I suspect it is slower than the fallback case. There will continue to be cases where we should *not* write them, but for e.g. the first pass where we have already cleared to transparent (which we may also be able to skip) writing the transparent pixels will not make anything incorrect. * Report color type and alpha type per frame - Depending on alpha values, disposal methods, frame rects, etc, subsequent frames may have different properties than the first. * Skip copies of the encoded data - We copy the encoded data in case the stream is one that cannot be rewound, so we can parse and then decode (possibly not immediately). For some input streams, this is unnecessary. - I was concerned this cause a performance regression, but on average the new code is faster than the old for the images I tested [1]. - It may cause a performance regression for Chromium, though, where we can always move back in the stream, so this should be addressed. Design doc: https://docs.google.com/a/google.com/document/d/12Qhf9T92MWfdWujQwCIjhCO3sw6pTJB5pJBwDM1T7Kc/ [1] https://docs.google.com/a/google.com/spreadsheets/d/19V-t9BfbFw5eiwBTKA1qOBkZbchjlTC5EIz6HFy-6RI/ GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=2045293002 Review-Url: https://codereview.chromium.org/2045293002
2016-10-24 16:03:26 +00:00
, fSubset(nullptr)
, fFrameIndex(0)
, fHasPriorFrame(false)
, fPremulBehavior(SkTransferFunctionBehavior::kRespect)
{}
ZeroInitialized fZeroInitialized;
/**
* If not NULL, represents a subset of the original image to decode.
* Must be within the bounds returned by getInfo().
* If the EncodedFormat is SkEncodedImageFormat::kWEBP (the only one which
* currently supports subsets), the top and left values must be even.
*
Make SkPngCodec decode progressively. This is a step towards using SkCodec in Chromium, where progressive decoding is necessary. Switch from using png_read_row (which expects all the data to be available) to png_process_data, which uses callbacks when rows are available. Create a new API for SkCodec, which supports progressive decoding and scanline decoding. Future changes will switch the other clients off of startScanlineDecode and get/skip-Scanlines to the new API. Remove SkCodec::kNone_ScanlineOrder, which was only used for interlaced PNG images. In the new API, interlaced PNG fits kTopDown. Also remove updateCurrScanline(), which was only used by the old implementation for interlaced PNG. DMSrcSink: - In CodecSrc::kScanline_Mode, use the new method for scanline decoding for the supported formats (just PNG and PNG-in-ICO for now). fuzz.cpp: - Remove reference to kNone_ScanlineOrder SkCodec: - Add new APIs: - startIncrementalDecode - incrementalDecode - Remove kNone_SkScanlineOrder and updateCurrScanline() - Set fDstInfo and fOptions in getPixels(). This may not be necessary for all implementations, but it simplifies things for SkPngCodec. SkPngCodec: - Implement new APIs - Switch from sk_read_fn/png_read_row etc to png_process_data - Expand AutoCleanPng's role to decode the header and create the SkPngCodec - Make the interlaced PNG decoder report how many lines were initialized during an incomplete decode SkIcoCodec: - Implement the new APIs; supported for PNG in ICO SkSampledCodec: - Call the new method for decoding scanlines, and fall back to the old method if the new version is unimplemented - Remove references to kNone_SkScanlineOrder tests/CodecPartial: - Add a test which decodes part of an image, then finishes the decode, and compares it to the straightforward method tests/CodecTest: - Add a test which decodes all scanlines using the new method - Repurpose the Codec_stripes test to decode using the new method in sections rather than all at once - In the method check(), add a parameter for whether the image supports the new method of scanline decoding, and be explicit about whether an image supports incomplete - Test incomplete PNG decodes. We should have been doing it anyway for non-interlaced (except for an image that is too small - one row), but the new method supports interlaced incomplete as well - Make test_invalid_parameters test the new method - Add a test to ensure that it's safe to fall back to scanline decoding without rewinding BUG=skia:4211 The new version was generally faster than the old version (but not significantly so). Some raw performance differences can be found at https://docs.google.com/a/google.com/spreadsheets/d/1Gis3aRCEa72qBNDRMgGDg3jD-pMgO-FXldlNF9ejo4o/ Design doc can be found at https://docs.google.com/a/google.com/document/d/11Mn8-ePDKwVEMCjs3nWwSjxcSpJ_Cu8DF57KNtUmgLM/ GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1997703003 Review-Url: https://codereview.chromium.org/1997703003
2016-09-16 15:20:38 +00:00
* In getPixels and incremental decode, we will attempt to decode the
* exact rectangular subset specified by fSubset.
*
* In a scanline decode, it does not make sense to specify a subset
* top or subset height, since the client already controls which rows
* to get and which rows to skip. During scanline decodes, we will
* require that the subset top be zero and the subset height be equal
* to the full height. We will, however, use the values of
* subset left and subset width to decode partial scanlines on calls
* to getScanlines().
*/
const SkIRect* fSubset;
Add support for multiple frames in SkCodec Add an interface to decode frames beyond the first in SkCodec, and add an implementation for SkGifCodec. Add getFrameData to SkCodec. This method reads ahead in the stream to return a vector containing meta data about each frame in the image. This is not required in order to decode frames beyond the first, but it allows a client to learn extra information: - how long the frame should be displayed - whether a frame should be blended with a prior frame, allowing the client to provide the prior frame to speed up decoding Add a new fields to SkCodec::Options: - fFrameIndex - fHasPriorFrame The API is designed so that SkCodec never caches frames. If a client wants a frame beyond the first, they specify the frame in Options.fFrameIndex. If the client does not have the frame's required frame (the frame that this frame must be blended on top of) cached, they pass false for Options.fHasPriorFrame. Unless the frame is independent, the codec will then recursively decode all frames necessary to decode fFrameIndex. If the client has the required frame cached, they can put it in the dst they pass to the codec, and the codec will only draw fFrameIndex onto it. Replace SkGifCodec's scanline decoding support with progressive decoding, and update the tests accordingly. Implement new APIs in SkGifCodec. Instead of using gif_lib, use GIFImageReader, imported from Chromium (along with its copyright headers) with the following changes: - SkGifCodec is now the client - Replace blink types - Combine GIFColorMap::buildTable and ::getTable into a method that creates and returns an SkColorTable - Input comes from an SkStream, instead of a SegmentReader. Add SkStreamBuffer, which buffers the (potentially partial) stream in order to decode progressively. (FIXME: This requires copying data that previously was read directly from the SegmentReader. Does this hurt performance? If so, can we fix it?) - Remove UMA code - Instead of reporting screen width and height to the client, allow the client to query for it - Fail earlier if the first frame AND screen have size of zero - Compute required previous frame when adding a new one - Move GIFParseQuery from GIFImageDecoder to GIFImageReader - Allow parsing up to a specific frame (to skip parsing the rest of the stream if a client only wants the first frame) - Compute whether the first frame has alpha and supports index 8, to create the SkImageInfo. This happens before reporting that the size has been decoded. Add GIFImageDecoder::haveDecodedRow to SkGifCodec, imported from Chromium (along with its copyright header), with the following changes: - Add support for sampling - Use the swizzler - Keep track of the rows decoded - Do *not* keep track of whether we've seen alpha Remove SkCodec::kOutOfOrder_SkScanlineOrder, which was only used by GIF scanline decoding. Call onRewind even if there is no stream (SkGifCodec needs to clear its decoded state so it will decode from the beginning). Add a method to SkSwizzler to access the offset into the dst, taking subsetting into account. Add a GM that animates a GIF. Add tests for the new APIs. *** Behavior changes: * Previously, we reported that an image with a subset frame and no transparent index was opaque and used the background index (if present) to fill the background. This is necessary in order to support index 8, but it does not match viewers/browsers I have seen. Examples: - Chromium and Gimp render the background transparent - Firefox, Safari, Linux Image Viewer, Safari Preview clip to the frame (for a single frame image) This CL matches Chromium's behavior and renders the background transparent. This allows us to have consistent behavior across products and simplifies the code (relative to what we would have to do to continue the old behavior on Android). It also means that we will no longer support index 8 for some GIFs. * Stop checking for GIFSTAMP - all GIFs should be either 89a or 87a. This matches Chromium. I suspect that bugs would have been reported if valid GIFs started with "GIFVER" instead of "GIF89a" or "GIF87a" (but did not decode in Chromium). *** Future work not included in this CL: * Move some checks out of haveDecodedRow, since they are the same for the entire frame e.g. - intersecting the frameRect with the full image size - whether there is a color table * Change when we write transparent pixels - In some cases, Chromium deemed this unnecessary, but I suspect it is slower than the fallback case. There will continue to be cases where we should *not* write them, but for e.g. the first pass where we have already cleared to transparent (which we may also be able to skip) writing the transparent pixels will not make anything incorrect. * Report color type and alpha type per frame - Depending on alpha values, disposal methods, frame rects, etc, subsequent frames may have different properties than the first. * Skip copies of the encoded data - We copy the encoded data in case the stream is one that cannot be rewound, so we can parse and then decode (possibly not immediately). For some input streams, this is unnecessary. - I was concerned this cause a performance regression, but on average the new code is faster than the old for the images I tested [1]. - It may cause a performance regression for Chromium, though, where we can always move back in the stream, so this should be addressed. Design doc: https://docs.google.com/a/google.com/document/d/12Qhf9T92MWfdWujQwCIjhCO3sw6pTJB5pJBwDM1T7Kc/ [1] https://docs.google.com/a/google.com/spreadsheets/d/19V-t9BfbFw5eiwBTKA1qOBkZbchjlTC5EIz6HFy-6RI/ GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=2045293002 Review-Url: https://codereview.chromium.org/2045293002
2016-10-24 16:03:26 +00:00
/**
* The frame to decode.
*
* Only meaningful for multi-frame images.
*/
size_t fFrameIndex;
Add support for multiple frames in SkCodec Add an interface to decode frames beyond the first in SkCodec, and add an implementation for SkGifCodec. Add getFrameData to SkCodec. This method reads ahead in the stream to return a vector containing meta data about each frame in the image. This is not required in order to decode frames beyond the first, but it allows a client to learn extra information: - how long the frame should be displayed - whether a frame should be blended with a prior frame, allowing the client to provide the prior frame to speed up decoding Add a new fields to SkCodec::Options: - fFrameIndex - fHasPriorFrame The API is designed so that SkCodec never caches frames. If a client wants a frame beyond the first, they specify the frame in Options.fFrameIndex. If the client does not have the frame's required frame (the frame that this frame must be blended on top of) cached, they pass false for Options.fHasPriorFrame. Unless the frame is independent, the codec will then recursively decode all frames necessary to decode fFrameIndex. If the client has the required frame cached, they can put it in the dst they pass to the codec, and the codec will only draw fFrameIndex onto it. Replace SkGifCodec's scanline decoding support with progressive decoding, and update the tests accordingly. Implement new APIs in SkGifCodec. Instead of using gif_lib, use GIFImageReader, imported from Chromium (along with its copyright headers) with the following changes: - SkGifCodec is now the client - Replace blink types - Combine GIFColorMap::buildTable and ::getTable into a method that creates and returns an SkColorTable - Input comes from an SkStream, instead of a SegmentReader. Add SkStreamBuffer, which buffers the (potentially partial) stream in order to decode progressively. (FIXME: This requires copying data that previously was read directly from the SegmentReader. Does this hurt performance? If so, can we fix it?) - Remove UMA code - Instead of reporting screen width and height to the client, allow the client to query for it - Fail earlier if the first frame AND screen have size of zero - Compute required previous frame when adding a new one - Move GIFParseQuery from GIFImageDecoder to GIFImageReader - Allow parsing up to a specific frame (to skip parsing the rest of the stream if a client only wants the first frame) - Compute whether the first frame has alpha and supports index 8, to create the SkImageInfo. This happens before reporting that the size has been decoded. Add GIFImageDecoder::haveDecodedRow to SkGifCodec, imported from Chromium (along with its copyright header), with the following changes: - Add support for sampling - Use the swizzler - Keep track of the rows decoded - Do *not* keep track of whether we've seen alpha Remove SkCodec::kOutOfOrder_SkScanlineOrder, which was only used by GIF scanline decoding. Call onRewind even if there is no stream (SkGifCodec needs to clear its decoded state so it will decode from the beginning). Add a method to SkSwizzler to access the offset into the dst, taking subsetting into account. Add a GM that animates a GIF. Add tests for the new APIs. *** Behavior changes: * Previously, we reported that an image with a subset frame and no transparent index was opaque and used the background index (if present) to fill the background. This is necessary in order to support index 8, but it does not match viewers/browsers I have seen. Examples: - Chromium and Gimp render the background transparent - Firefox, Safari, Linux Image Viewer, Safari Preview clip to the frame (for a single frame image) This CL matches Chromium's behavior and renders the background transparent. This allows us to have consistent behavior across products and simplifies the code (relative to what we would have to do to continue the old behavior on Android). It also means that we will no longer support index 8 for some GIFs. * Stop checking for GIFSTAMP - all GIFs should be either 89a or 87a. This matches Chromium. I suspect that bugs would have been reported if valid GIFs started with "GIFVER" instead of "GIF89a" or "GIF87a" (but did not decode in Chromium). *** Future work not included in this CL: * Move some checks out of haveDecodedRow, since they are the same for the entire frame e.g. - intersecting the frameRect with the full image size - whether there is a color table * Change when we write transparent pixels - In some cases, Chromium deemed this unnecessary, but I suspect it is slower than the fallback case. There will continue to be cases where we should *not* write them, but for e.g. the first pass where we have already cleared to transparent (which we may also be able to skip) writing the transparent pixels will not make anything incorrect. * Report color type and alpha type per frame - Depending on alpha values, disposal methods, frame rects, etc, subsequent frames may have different properties than the first. * Skip copies of the encoded data - We copy the encoded data in case the stream is one that cannot be rewound, so we can parse and then decode (possibly not immediately). For some input streams, this is unnecessary. - I was concerned this cause a performance regression, but on average the new code is faster than the old for the images I tested [1]. - It may cause a performance regression for Chromium, though, where we can always move back in the stream, so this should be addressed. Design doc: https://docs.google.com/a/google.com/document/d/12Qhf9T92MWfdWujQwCIjhCO3sw6pTJB5pJBwDM1T7Kc/ [1] https://docs.google.com/a/google.com/spreadsheets/d/19V-t9BfbFw5eiwBTKA1qOBkZbchjlTC5EIz6HFy-6RI/ GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=2045293002 Review-Url: https://codereview.chromium.org/2045293002
2016-10-24 16:03:26 +00:00
/**
* If true, the dst already contains the prior frame.
*
* Only meaningful for multi-frame images.
*
* If fFrameIndex needs to be blended with a prior frame (as reported by
* getFrameInfo[fFrameIndex].fRequiredFrame), the client can set this to
* either true or false:
*
* true means that the prior frame is already in the dst, and this
* codec only needs to decode fFrameIndex and blend it with the dst.
* Options.fZeroInitialized is ignored in this case.
*
* false means that the dst does not contain the prior frame, so this
* codec needs to first decode the prior frame (which in turn may need
* to decode its prior frame).
*/
bool fHasPriorFrame;
/**
* Indicates whether we should do a linear premultiply or a legacy premultiply.
*
* In the case where the dst SkColorSpace is nullptr, this flag is ignored and
* we will always do a legacy premultiply.
*/
SkTransferFunctionBehavior fPremulBehavior;
};
/**
* Decode into the given pixels, a block of memory of size at
* least (info.fHeight - 1) * rowBytes + (info.fWidth *
* bytesPerPixel)
*
* Repeated calls to this function should give the same results,
* allowing the PixelRef to be immutable.
*
* @param info A description of the format (config, size)
* expected by the caller. This can simply be identical
* to the info returned by getInfo().
*
* This contract also allows the caller to specify
* different output-configs, which the implementation can
* decide to support or not.
*
* A size that does not match getInfo() implies a request
* to scale. If the generator cannot perform this scale,
* it will return kInvalidScale.
*
* If the info contains a non-null SkColorSpace, the codec
* will perform the appropriate color space transformation.
* If the caller passes in the same color space that was
* reported by the codec, the color space transformation is
* a no-op.
*
* If info is kIndex8_SkColorType, then the caller must provide storage for up to 256
* SkPMColor values in ctable. On success the generator must copy N colors into that storage,
* (where N is the logical number of table entries) and set ctableCount to N.
*
* If info is not kIndex8_SkColorType, then the last two parameters may be NULL. If ctableCount
* is not null, it will be set to 0.
*
Merge SkCodec with SkScanlineDecoder Benefits: - This mimics other decoding APIs (including the ones SkCodec relies on, e.g. a png_struct, which can be used to decode an entire image or one line at a time). - It allows a client to ask us to do what we can do efficiently - i.e. start from encoded data and either decode the whole thing or scanlines. - It removes the duplicate methods which appeared in both SkCodec and SkScanlineDecoder (some of which, e.g. in SkJpegScanlineDecoder, just call fCodec->sameMethod()). - It simplifies moving more checks into the base class (e.g. the examples in skbug.com/4284). BUG=skia:4175 BUG=skia:4284 ===================================================================== SkScanlineDecoder.h/.cpp: Removed. SkCodec.h/.cpp: Add methods, enums, and variables which were previously in SkScanlineDecoder. Default fCurrScanline to -1, as a sentinel that start has not been called. General changes: Convert SkScanlineDecoders to SkCodecs. General changes in SkCodec subclasses: Merge SkScanlineDecoder implementation into SkCodec. Most (all?) owned an SkCodec, so they now call this-> instead of fCodec->. SkBmpCodec.h/.cpp: Replace the unused rowOrder method with an override for onGetScanlineOrder. Make getDstRow const, since it is called by onGetY, which is const. SkCodec_libpng.h/.cpp: Make SkPngCodec an abstract class, with two subclasses which handle scanline decoding separately (they share code for decoding the entire image). Reimplement onReallyHasAlpha so that it can return the most recent result (e.g. after a scanline decode which only decoded part of the image) or a better answer (e.g. if the whole image is known to be opaque). Compute fNumberPasses early, so we know which subclass to instantiate. Make SkPngInterlaceScanlineDecoder use the base class' fCurrScanline rather than a separate variable. CodexTest.cpp: Add tests for the state changes in SkCodec (need to call start before decoding scanlines; calling getPixels means that start will need to be called again before decoding more scanlines). Add a test which decodes in stripes, currently only used for an interlaced PNG. TODO: Add tests for onReallyHasAlpha. Review URL: https://codereview.chromium.org/1365313002
2015-09-30 15:57:13 +00:00
* If a scanline decode is in progress, scanline mode will end, requiring the client to call
* startScanlineDecode() in order to return to decoding scanlines.
*
* @return Result kSuccess, or another value explaining the type of failure.
*/
Result getPixels(const SkImageInfo& info, void* pixels, size_t rowBytes, const Options*,
SkPMColor ctable[], int* ctableCount);
/**
* Simplified version of getPixels() that asserts that info is NOT kIndex8_SkColorType and
* uses the default Options.
*/
Result getPixels(const SkImageInfo& info, void* pixels, size_t rowBytes);
/**
* If decoding to YUV is supported, this returns true. Otherwise, this
* returns false and does not modify any of the parameters.
*
* @param sizeInfo Output parameter indicating the sizes and required
* allocation widths of the Y, U, and V planes.
* @param colorSpace Output parameter. If non-NULL this is set to kJPEG,
* otherwise this is ignored.
*/
bool queryYUV8(SkYUVSizeInfo* sizeInfo, SkYUVColorSpace* colorSpace) const {
if (nullptr == sizeInfo) {
return false;
}
return this->onQueryYUV8(sizeInfo, colorSpace);
}
/**
* Returns kSuccess, or another value explaining the type of failure.
* This always attempts to perform a full decode. If the client only
* wants size, it should call queryYUV8().
*
* @param sizeInfo Needs to exactly match the values returned by the
* query, except the WidthBytes may be larger than the
* recommendation (but not smaller).
* @param planes Memory for each of the Y, U, and V planes.
*/
Result getYUV8Planes(const SkYUVSizeInfo& sizeInfo, void* planes[3]) {
if (nullptr == planes || nullptr == planes[0] || nullptr == planes[1] ||
nullptr == planes[2]) {
return kInvalidInput;
}
if (!this->rewindIfNeeded()) {
return kCouldNotRewind;
}
return this->onGetYUV8Planes(sizeInfo, planes);
}
Make SkPngCodec decode progressively. This is a step towards using SkCodec in Chromium, where progressive decoding is necessary. Switch from using png_read_row (which expects all the data to be available) to png_process_data, which uses callbacks when rows are available. Create a new API for SkCodec, which supports progressive decoding and scanline decoding. Future changes will switch the other clients off of startScanlineDecode and get/skip-Scanlines to the new API. Remove SkCodec::kNone_ScanlineOrder, which was only used for interlaced PNG images. In the new API, interlaced PNG fits kTopDown. Also remove updateCurrScanline(), which was only used by the old implementation for interlaced PNG. DMSrcSink: - In CodecSrc::kScanline_Mode, use the new method for scanline decoding for the supported formats (just PNG and PNG-in-ICO for now). fuzz.cpp: - Remove reference to kNone_ScanlineOrder SkCodec: - Add new APIs: - startIncrementalDecode - incrementalDecode - Remove kNone_SkScanlineOrder and updateCurrScanline() - Set fDstInfo and fOptions in getPixels(). This may not be necessary for all implementations, but it simplifies things for SkPngCodec. SkPngCodec: - Implement new APIs - Switch from sk_read_fn/png_read_row etc to png_process_data - Expand AutoCleanPng's role to decode the header and create the SkPngCodec - Make the interlaced PNG decoder report how many lines were initialized during an incomplete decode SkIcoCodec: - Implement the new APIs; supported for PNG in ICO SkSampledCodec: - Call the new method for decoding scanlines, and fall back to the old method if the new version is unimplemented - Remove references to kNone_SkScanlineOrder tests/CodecPartial: - Add a test which decodes part of an image, then finishes the decode, and compares it to the straightforward method tests/CodecTest: - Add a test which decodes all scanlines using the new method - Repurpose the Codec_stripes test to decode using the new method in sections rather than all at once - In the method check(), add a parameter for whether the image supports the new method of scanline decoding, and be explicit about whether an image supports incomplete - Test incomplete PNG decodes. We should have been doing it anyway for non-interlaced (except for an image that is too small - one row), but the new method supports interlaced incomplete as well - Make test_invalid_parameters test the new method - Add a test to ensure that it's safe to fall back to scanline decoding without rewinding BUG=skia:4211 The new version was generally faster than the old version (but not significantly so). Some raw performance differences can be found at https://docs.google.com/a/google.com/spreadsheets/d/1Gis3aRCEa72qBNDRMgGDg3jD-pMgO-FXldlNF9ejo4o/ Design doc can be found at https://docs.google.com/a/google.com/document/d/11Mn8-ePDKwVEMCjs3nWwSjxcSpJ_Cu8DF57KNtUmgLM/ GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1997703003 Review-Url: https://codereview.chromium.org/1997703003
2016-09-16 15:20:38 +00:00
/**
* Prepare for an incremental decode with the specified options.
*
* This may require a rewind.
*
* @param dstInfo Info of the destination. If the dimensions do not match
* those of getInfo, this implies a scale.
* @param dst Memory to write to. Needs to be large enough to hold the subset,
* if present, or the full image as described in dstInfo.
* @param options Contains decoding options, including if memory is zero
* initialized and whether to decode a subset.
* @param ctable A pointer to a color table. When dstInfo.colorType() is
* kIndex8, this should be non-NULL and have enough storage for 256
* colors. The color table will be populated after decoding the palette.
* @param ctableCount A pointer to the size of the color table. When
* dstInfo.colorType() is kIndex8, this should be non-NULL. It will
* be modified to the true size of the color table (<= 256) after
* decoding the palette.
* @return Enum representing success or reason for failure.
*/
Result startIncrementalDecode(const SkImageInfo& dstInfo, void* dst, size_t rowBytes,
const SkCodec::Options*, SkPMColor* ctable, int* ctableCount);
Result startIncrementalDecode(const SkImageInfo& dstInfo, void* dst, size_t rowBytes,
const SkCodec::Options* options) {
return this->startIncrementalDecode(dstInfo, dst, rowBytes, options, nullptr, nullptr);
}
Result startIncrementalDecode(const SkImageInfo& dstInfo, void* dst, size_t rowBytes) {
return this->startIncrementalDecode(dstInfo, dst, rowBytes, nullptr, nullptr, nullptr);
}
/**
* Start/continue the incremental decode.
*
* Not valid to call before calling startIncrementalDecode().
*
* After the first call, should only be called again if more data has been
* provided to the source SkStream.
*
* Unlike getPixels and getScanlines, this does not do any filling. This is
* left up to the caller, since they may be skipping lines or continuing the
* decode later. In the latter case, they may choose to initialize all lines
* first, or only initialize the remaining lines after the first call.
*
* @param rowsDecoded Optional output variable returning the total number of
* lines initialized. Only meaningful if this method returns kIncompleteInput.
* Otherwise the implementation may not set it.
* Note that some implementations may have initialized this many rows, but
* not necessarily finished those rows (e.g. interlaced PNG). This may be
* useful for determining what rows the client needs to initialize.
* @return kSuccess if all lines requested in startIncrementalDecode have
* been completely decoded. kIncompleteInput otherwise.
*/
Result incrementalDecode(int* rowsDecoded = nullptr) {
if (!fStartedIncrementalDecode) {
return kInvalidParameters;
}
return this->onIncrementalDecode(rowsDecoded);
}
Merge SkCodec with SkScanlineDecoder Benefits: - This mimics other decoding APIs (including the ones SkCodec relies on, e.g. a png_struct, which can be used to decode an entire image or one line at a time). - It allows a client to ask us to do what we can do efficiently - i.e. start from encoded data and either decode the whole thing or scanlines. - It removes the duplicate methods which appeared in both SkCodec and SkScanlineDecoder (some of which, e.g. in SkJpegScanlineDecoder, just call fCodec->sameMethod()). - It simplifies moving more checks into the base class (e.g. the examples in skbug.com/4284). BUG=skia:4175 BUG=skia:4284 ===================================================================== SkScanlineDecoder.h/.cpp: Removed. SkCodec.h/.cpp: Add methods, enums, and variables which were previously in SkScanlineDecoder. Default fCurrScanline to -1, as a sentinel that start has not been called. General changes: Convert SkScanlineDecoders to SkCodecs. General changes in SkCodec subclasses: Merge SkScanlineDecoder implementation into SkCodec. Most (all?) owned an SkCodec, so they now call this-> instead of fCodec->. SkBmpCodec.h/.cpp: Replace the unused rowOrder method with an override for onGetScanlineOrder. Make getDstRow const, since it is called by onGetY, which is const. SkCodec_libpng.h/.cpp: Make SkPngCodec an abstract class, with two subclasses which handle scanline decoding separately (they share code for decoding the entire image). Reimplement onReallyHasAlpha so that it can return the most recent result (e.g. after a scanline decode which only decoded part of the image) or a better answer (e.g. if the whole image is known to be opaque). Compute fNumberPasses early, so we know which subclass to instantiate. Make SkPngInterlaceScanlineDecoder use the base class' fCurrScanline rather than a separate variable. CodexTest.cpp: Add tests for the state changes in SkCodec (need to call start before decoding scanlines; calling getPixels means that start will need to be called again before decoding more scanlines). Add a test which decodes in stripes, currently only used for an interlaced PNG. TODO: Add tests for onReallyHasAlpha. Review URL: https://codereview.chromium.org/1365313002
2015-09-30 15:57:13 +00:00
/**
* The remaining functions revolve around decoding scanlines.
*/
/**
* Prepare for a scanline decode with the specified options.
*
* After this call, this class will be ready to decode the first scanline.
*
* This must be called in order to call getScanlines or skipScanlines.
*
* This may require rewinding the stream.
*
* Not all SkCodecs support this.
*
* @param dstInfo Info of the destination. If the dimensions do not match
* those of getInfo, this implies a scale.
* @param options Contains decoding options, including if memory is zero
* initialized.
* @param ctable A pointer to a color table. When dstInfo.colorType() is
* kIndex8, this should be non-NULL and have enough storage for 256
* colors. The color table will be populated after decoding the palette.
* @param ctableCount A pointer to the size of the color table. When
* dstInfo.colorType() is kIndex8, this should be non-NULL. It will
* be modified to the true size of the color table (<= 256) after
* decoding the palette.
* @return Enum representing success or reason for failure.
*/
Result startScanlineDecode(const SkImageInfo& dstInfo, const SkCodec::Options* options,
SkPMColor ctable[], int* ctableCount);
Merge SkCodec with SkScanlineDecoder Benefits: - This mimics other decoding APIs (including the ones SkCodec relies on, e.g. a png_struct, which can be used to decode an entire image or one line at a time). - It allows a client to ask us to do what we can do efficiently - i.e. start from encoded data and either decode the whole thing or scanlines. - It removes the duplicate methods which appeared in both SkCodec and SkScanlineDecoder (some of which, e.g. in SkJpegScanlineDecoder, just call fCodec->sameMethod()). - It simplifies moving more checks into the base class (e.g. the examples in skbug.com/4284). BUG=skia:4175 BUG=skia:4284 ===================================================================== SkScanlineDecoder.h/.cpp: Removed. SkCodec.h/.cpp: Add methods, enums, and variables which were previously in SkScanlineDecoder. Default fCurrScanline to -1, as a sentinel that start has not been called. General changes: Convert SkScanlineDecoders to SkCodecs. General changes in SkCodec subclasses: Merge SkScanlineDecoder implementation into SkCodec. Most (all?) owned an SkCodec, so they now call this-> instead of fCodec->. SkBmpCodec.h/.cpp: Replace the unused rowOrder method with an override for onGetScanlineOrder. Make getDstRow const, since it is called by onGetY, which is const. SkCodec_libpng.h/.cpp: Make SkPngCodec an abstract class, with two subclasses which handle scanline decoding separately (they share code for decoding the entire image). Reimplement onReallyHasAlpha so that it can return the most recent result (e.g. after a scanline decode which only decoded part of the image) or a better answer (e.g. if the whole image is known to be opaque). Compute fNumberPasses early, so we know which subclass to instantiate. Make SkPngInterlaceScanlineDecoder use the base class' fCurrScanline rather than a separate variable. CodexTest.cpp: Add tests for the state changes in SkCodec (need to call start before decoding scanlines; calling getPixels means that start will need to be called again before decoding more scanlines). Add a test which decodes in stripes, currently only used for an interlaced PNG. TODO: Add tests for onReallyHasAlpha. Review URL: https://codereview.chromium.org/1365313002
2015-09-30 15:57:13 +00:00
/**
* Simplified version of startScanlineDecode() that asserts that info is NOT
* kIndex8_SkColorType and uses the default Options.
*/
Result startScanlineDecode(const SkImageInfo& dstInfo);
/**
* Write the next countLines scanlines into dst.
*
* Not valid to call before calling startScanlineDecode().
*
* @param dst Must be non-null, and large enough to hold countLines
* scanlines of size rowBytes.
* @param countLines Number of lines to write.
* @param rowBytes Number of bytes per row. Must be large enough to hold
* a scanline based on the SkImageInfo used to create this object.
* @return the number of lines successfully decoded. If this value is
* less than countLines, this will fill the remaining lines with a
* default value.
Merge SkCodec with SkScanlineDecoder Benefits: - This mimics other decoding APIs (including the ones SkCodec relies on, e.g. a png_struct, which can be used to decode an entire image or one line at a time). - It allows a client to ask us to do what we can do efficiently - i.e. start from encoded data and either decode the whole thing or scanlines. - It removes the duplicate methods which appeared in both SkCodec and SkScanlineDecoder (some of which, e.g. in SkJpegScanlineDecoder, just call fCodec->sameMethod()). - It simplifies moving more checks into the base class (e.g. the examples in skbug.com/4284). BUG=skia:4175 BUG=skia:4284 ===================================================================== SkScanlineDecoder.h/.cpp: Removed. SkCodec.h/.cpp: Add methods, enums, and variables which were previously in SkScanlineDecoder. Default fCurrScanline to -1, as a sentinel that start has not been called. General changes: Convert SkScanlineDecoders to SkCodecs. General changes in SkCodec subclasses: Merge SkScanlineDecoder implementation into SkCodec. Most (all?) owned an SkCodec, so they now call this-> instead of fCodec->. SkBmpCodec.h/.cpp: Replace the unused rowOrder method with an override for onGetScanlineOrder. Make getDstRow const, since it is called by onGetY, which is const. SkCodec_libpng.h/.cpp: Make SkPngCodec an abstract class, with two subclasses which handle scanline decoding separately (they share code for decoding the entire image). Reimplement onReallyHasAlpha so that it can return the most recent result (e.g. after a scanline decode which only decoded part of the image) or a better answer (e.g. if the whole image is known to be opaque). Compute fNumberPasses early, so we know which subclass to instantiate. Make SkPngInterlaceScanlineDecoder use the base class' fCurrScanline rather than a separate variable. CodexTest.cpp: Add tests for the state changes in SkCodec (need to call start before decoding scanlines; calling getPixels means that start will need to be called again before decoding more scanlines). Add a test which decodes in stripes, currently only used for an interlaced PNG. TODO: Add tests for onReallyHasAlpha. Review URL: https://codereview.chromium.org/1365313002
2015-09-30 15:57:13 +00:00
*/
int getScanlines(void* dst, int countLines, size_t rowBytes);
Merge SkCodec with SkScanlineDecoder Benefits: - This mimics other decoding APIs (including the ones SkCodec relies on, e.g. a png_struct, which can be used to decode an entire image or one line at a time). - It allows a client to ask us to do what we can do efficiently - i.e. start from encoded data and either decode the whole thing or scanlines. - It removes the duplicate methods which appeared in both SkCodec and SkScanlineDecoder (some of which, e.g. in SkJpegScanlineDecoder, just call fCodec->sameMethod()). - It simplifies moving more checks into the base class (e.g. the examples in skbug.com/4284). BUG=skia:4175 BUG=skia:4284 ===================================================================== SkScanlineDecoder.h/.cpp: Removed. SkCodec.h/.cpp: Add methods, enums, and variables which were previously in SkScanlineDecoder. Default fCurrScanline to -1, as a sentinel that start has not been called. General changes: Convert SkScanlineDecoders to SkCodecs. General changes in SkCodec subclasses: Merge SkScanlineDecoder implementation into SkCodec. Most (all?) owned an SkCodec, so they now call this-> instead of fCodec->. SkBmpCodec.h/.cpp: Replace the unused rowOrder method with an override for onGetScanlineOrder. Make getDstRow const, since it is called by onGetY, which is const. SkCodec_libpng.h/.cpp: Make SkPngCodec an abstract class, with two subclasses which handle scanline decoding separately (they share code for decoding the entire image). Reimplement onReallyHasAlpha so that it can return the most recent result (e.g. after a scanline decode which only decoded part of the image) or a better answer (e.g. if the whole image is known to be opaque). Compute fNumberPasses early, so we know which subclass to instantiate. Make SkPngInterlaceScanlineDecoder use the base class' fCurrScanline rather than a separate variable. CodexTest.cpp: Add tests for the state changes in SkCodec (need to call start before decoding scanlines; calling getPixels means that start will need to be called again before decoding more scanlines). Add a test which decodes in stripes, currently only used for an interlaced PNG. TODO: Add tests for onReallyHasAlpha. Review URL: https://codereview.chromium.org/1365313002
2015-09-30 15:57:13 +00:00
/**
* Skip count scanlines.
*
* Not valid to call before calling startScanlineDecode().
*
* The default version just calls onGetScanlines and discards the dst.
* NOTE: If skipped lines are the only lines with alpha, this default
* will make reallyHasAlpha return true, when it could have returned
* false.
*
* @return true if the scanlines were successfully skipped
* false on failure, possible reasons for failure include:
* An incomplete input image stream.
* Calling this function before calling startScanlineDecode().
* If countLines is less than zero or so large that it moves
* the current scanline past the end of the image.
Merge SkCodec with SkScanlineDecoder Benefits: - This mimics other decoding APIs (including the ones SkCodec relies on, e.g. a png_struct, which can be used to decode an entire image or one line at a time). - It allows a client to ask us to do what we can do efficiently - i.e. start from encoded data and either decode the whole thing or scanlines. - It removes the duplicate methods which appeared in both SkCodec and SkScanlineDecoder (some of which, e.g. in SkJpegScanlineDecoder, just call fCodec->sameMethod()). - It simplifies moving more checks into the base class (e.g. the examples in skbug.com/4284). BUG=skia:4175 BUG=skia:4284 ===================================================================== SkScanlineDecoder.h/.cpp: Removed. SkCodec.h/.cpp: Add methods, enums, and variables which were previously in SkScanlineDecoder. Default fCurrScanline to -1, as a sentinel that start has not been called. General changes: Convert SkScanlineDecoders to SkCodecs. General changes in SkCodec subclasses: Merge SkScanlineDecoder implementation into SkCodec. Most (all?) owned an SkCodec, so they now call this-> instead of fCodec->. SkBmpCodec.h/.cpp: Replace the unused rowOrder method with an override for onGetScanlineOrder. Make getDstRow const, since it is called by onGetY, which is const. SkCodec_libpng.h/.cpp: Make SkPngCodec an abstract class, with two subclasses which handle scanline decoding separately (they share code for decoding the entire image). Reimplement onReallyHasAlpha so that it can return the most recent result (e.g. after a scanline decode which only decoded part of the image) or a better answer (e.g. if the whole image is known to be opaque). Compute fNumberPasses early, so we know which subclass to instantiate. Make SkPngInterlaceScanlineDecoder use the base class' fCurrScanline rather than a separate variable. CodexTest.cpp: Add tests for the state changes in SkCodec (need to call start before decoding scanlines; calling getPixels means that start will need to be called again before decoding more scanlines). Add a test which decodes in stripes, currently only used for an interlaced PNG. TODO: Add tests for onReallyHasAlpha. Review URL: https://codereview.chromium.org/1365313002
2015-09-30 15:57:13 +00:00
*/
bool skipScanlines(int countLines);
Merge SkCodec with SkScanlineDecoder Benefits: - This mimics other decoding APIs (including the ones SkCodec relies on, e.g. a png_struct, which can be used to decode an entire image or one line at a time). - It allows a client to ask us to do what we can do efficiently - i.e. start from encoded data and either decode the whole thing or scanlines. - It removes the duplicate methods which appeared in both SkCodec and SkScanlineDecoder (some of which, e.g. in SkJpegScanlineDecoder, just call fCodec->sameMethod()). - It simplifies moving more checks into the base class (e.g. the examples in skbug.com/4284). BUG=skia:4175 BUG=skia:4284 ===================================================================== SkScanlineDecoder.h/.cpp: Removed. SkCodec.h/.cpp: Add methods, enums, and variables which were previously in SkScanlineDecoder. Default fCurrScanline to -1, as a sentinel that start has not been called. General changes: Convert SkScanlineDecoders to SkCodecs. General changes in SkCodec subclasses: Merge SkScanlineDecoder implementation into SkCodec. Most (all?) owned an SkCodec, so they now call this-> instead of fCodec->. SkBmpCodec.h/.cpp: Replace the unused rowOrder method with an override for onGetScanlineOrder. Make getDstRow const, since it is called by onGetY, which is const. SkCodec_libpng.h/.cpp: Make SkPngCodec an abstract class, with two subclasses which handle scanline decoding separately (they share code for decoding the entire image). Reimplement onReallyHasAlpha so that it can return the most recent result (e.g. after a scanline decode which only decoded part of the image) or a better answer (e.g. if the whole image is known to be opaque). Compute fNumberPasses early, so we know which subclass to instantiate. Make SkPngInterlaceScanlineDecoder use the base class' fCurrScanline rather than a separate variable. CodexTest.cpp: Add tests for the state changes in SkCodec (need to call start before decoding scanlines; calling getPixels means that start will need to be called again before decoding more scanlines). Add a test which decodes in stripes, currently only used for an interlaced PNG. TODO: Add tests for onReallyHasAlpha. Review URL: https://codereview.chromium.org/1365313002
2015-09-30 15:57:13 +00:00
/**
* The order in which rows are output from the scanline decoder is not the
* same for all variations of all image types. This explains the possible
* output row orderings.
*/
enum SkScanlineOrder {
/*
* By far the most common, this indicates that the image can be decoded
* reliably using the scanline decoder, and that rows will be output in
* the logical order.
*/
kTopDown_SkScanlineOrder,
/*
* This indicates that the scanline decoder reliably outputs rows, but
* they will be returned in reverse order. If the scanline format is
* kBottomUp, the nextScanline() API can be used to determine the actual
* y-coordinate of the next output row, but the client is not forced
* to take advantage of this, given that it's not too tough to keep
* track independently.
*
* For full image decodes, it is safe to get all of the scanlines at
* once, since the decoder will handle inverting the rows as it
* decodes.
*
* For subset decodes and sampling, it is simplest to get and skip
* scanlines one at a time, using the nextScanline() API. It is
* possible to ask for larger chunks at a time, but this should be used
* with caution. As with full image decodes, the decoder will handle
* inverting the requested rows, but rows will still be delivered
* starting from the bottom of the image.
*
* Upside down bmps are an example.
*/
kBottomUp_SkScanlineOrder,
};
/**
* An enum representing the order in which scanlines will be returned by
* the scanline decoder.
*
* This is undefined before startScanlineDecode() is called.
Merge SkCodec with SkScanlineDecoder Benefits: - This mimics other decoding APIs (including the ones SkCodec relies on, e.g. a png_struct, which can be used to decode an entire image or one line at a time). - It allows a client to ask us to do what we can do efficiently - i.e. start from encoded data and either decode the whole thing or scanlines. - It removes the duplicate methods which appeared in both SkCodec and SkScanlineDecoder (some of which, e.g. in SkJpegScanlineDecoder, just call fCodec->sameMethod()). - It simplifies moving more checks into the base class (e.g. the examples in skbug.com/4284). BUG=skia:4175 BUG=skia:4284 ===================================================================== SkScanlineDecoder.h/.cpp: Removed. SkCodec.h/.cpp: Add methods, enums, and variables which were previously in SkScanlineDecoder. Default fCurrScanline to -1, as a sentinel that start has not been called. General changes: Convert SkScanlineDecoders to SkCodecs. General changes in SkCodec subclasses: Merge SkScanlineDecoder implementation into SkCodec. Most (all?) owned an SkCodec, so they now call this-> instead of fCodec->. SkBmpCodec.h/.cpp: Replace the unused rowOrder method with an override for onGetScanlineOrder. Make getDstRow const, since it is called by onGetY, which is const. SkCodec_libpng.h/.cpp: Make SkPngCodec an abstract class, with two subclasses which handle scanline decoding separately (they share code for decoding the entire image). Reimplement onReallyHasAlpha so that it can return the most recent result (e.g. after a scanline decode which only decoded part of the image) or a better answer (e.g. if the whole image is known to be opaque). Compute fNumberPasses early, so we know which subclass to instantiate. Make SkPngInterlaceScanlineDecoder use the base class' fCurrScanline rather than a separate variable. CodexTest.cpp: Add tests for the state changes in SkCodec (need to call start before decoding scanlines; calling getPixels means that start will need to be called again before decoding more scanlines). Add a test which decodes in stripes, currently only used for an interlaced PNG. TODO: Add tests for onReallyHasAlpha. Review URL: https://codereview.chromium.org/1365313002
2015-09-30 15:57:13 +00:00
*/
SkScanlineOrder getScanlineOrder() const { return this->onGetScanlineOrder(); }
/**
* Returns the y-coordinate of the next row to be returned by the scanline
* decoder.
*
* This will equal fCurrScanline, except in the case of strangely
Add support for multiple frames in SkCodec Add an interface to decode frames beyond the first in SkCodec, and add an implementation for SkGifCodec. Add getFrameData to SkCodec. This method reads ahead in the stream to return a vector containing meta data about each frame in the image. This is not required in order to decode frames beyond the first, but it allows a client to learn extra information: - how long the frame should be displayed - whether a frame should be blended with a prior frame, allowing the client to provide the prior frame to speed up decoding Add a new fields to SkCodec::Options: - fFrameIndex - fHasPriorFrame The API is designed so that SkCodec never caches frames. If a client wants a frame beyond the first, they specify the frame in Options.fFrameIndex. If the client does not have the frame's required frame (the frame that this frame must be blended on top of) cached, they pass false for Options.fHasPriorFrame. Unless the frame is independent, the codec will then recursively decode all frames necessary to decode fFrameIndex. If the client has the required frame cached, they can put it in the dst they pass to the codec, and the codec will only draw fFrameIndex onto it. Replace SkGifCodec's scanline decoding support with progressive decoding, and update the tests accordingly. Implement new APIs in SkGifCodec. Instead of using gif_lib, use GIFImageReader, imported from Chromium (along with its copyright headers) with the following changes: - SkGifCodec is now the client - Replace blink types - Combine GIFColorMap::buildTable and ::getTable into a method that creates and returns an SkColorTable - Input comes from an SkStream, instead of a SegmentReader. Add SkStreamBuffer, which buffers the (potentially partial) stream in order to decode progressively. (FIXME: This requires copying data that previously was read directly from the SegmentReader. Does this hurt performance? If so, can we fix it?) - Remove UMA code - Instead of reporting screen width and height to the client, allow the client to query for it - Fail earlier if the first frame AND screen have size of zero - Compute required previous frame when adding a new one - Move GIFParseQuery from GIFImageDecoder to GIFImageReader - Allow parsing up to a specific frame (to skip parsing the rest of the stream if a client only wants the first frame) - Compute whether the first frame has alpha and supports index 8, to create the SkImageInfo. This happens before reporting that the size has been decoded. Add GIFImageDecoder::haveDecodedRow to SkGifCodec, imported from Chromium (along with its copyright header), with the following changes: - Add support for sampling - Use the swizzler - Keep track of the rows decoded - Do *not* keep track of whether we've seen alpha Remove SkCodec::kOutOfOrder_SkScanlineOrder, which was only used by GIF scanline decoding. Call onRewind even if there is no stream (SkGifCodec needs to clear its decoded state so it will decode from the beginning). Add a method to SkSwizzler to access the offset into the dst, taking subsetting into account. Add a GM that animates a GIF. Add tests for the new APIs. *** Behavior changes: * Previously, we reported that an image with a subset frame and no transparent index was opaque and used the background index (if present) to fill the background. This is necessary in order to support index 8, but it does not match viewers/browsers I have seen. Examples: - Chromium and Gimp render the background transparent - Firefox, Safari, Linux Image Viewer, Safari Preview clip to the frame (for a single frame image) This CL matches Chromium's behavior and renders the background transparent. This allows us to have consistent behavior across products and simplifies the code (relative to what we would have to do to continue the old behavior on Android). It also means that we will no longer support index 8 for some GIFs. * Stop checking for GIFSTAMP - all GIFs should be either 89a or 87a. This matches Chromium. I suspect that bugs would have been reported if valid GIFs started with "GIFVER" instead of "GIF89a" or "GIF87a" (but did not decode in Chromium). *** Future work not included in this CL: * Move some checks out of haveDecodedRow, since they are the same for the entire frame e.g. - intersecting the frameRect with the full image size - whether there is a color table * Change when we write transparent pixels - In some cases, Chromium deemed this unnecessary, but I suspect it is slower than the fallback case. There will continue to be cases where we should *not* write them, but for e.g. the first pass where we have already cleared to transparent (which we may also be able to skip) writing the transparent pixels will not make anything incorrect. * Report color type and alpha type per frame - Depending on alpha values, disposal methods, frame rects, etc, subsequent frames may have different properties than the first. * Skip copies of the encoded data - We copy the encoded data in case the stream is one that cannot be rewound, so we can parse and then decode (possibly not immediately). For some input streams, this is unnecessary. - I was concerned this cause a performance regression, but on average the new code is faster than the old for the images I tested [1]. - It may cause a performance regression for Chromium, though, where we can always move back in the stream, so this should be addressed. Design doc: https://docs.google.com/a/google.com/document/d/12Qhf9T92MWfdWujQwCIjhCO3sw6pTJB5pJBwDM1T7Kc/ [1] https://docs.google.com/a/google.com/spreadsheets/d/19V-t9BfbFw5eiwBTKA1qOBkZbchjlTC5EIz6HFy-6RI/ GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=2045293002 Review-Url: https://codereview.chromium.org/2045293002
2016-10-24 16:03:26 +00:00
* encoded image types (bottom-up bmps).
Merge SkCodec with SkScanlineDecoder Benefits: - This mimics other decoding APIs (including the ones SkCodec relies on, e.g. a png_struct, which can be used to decode an entire image or one line at a time). - It allows a client to ask us to do what we can do efficiently - i.e. start from encoded data and either decode the whole thing or scanlines. - It removes the duplicate methods which appeared in both SkCodec and SkScanlineDecoder (some of which, e.g. in SkJpegScanlineDecoder, just call fCodec->sameMethod()). - It simplifies moving more checks into the base class (e.g. the examples in skbug.com/4284). BUG=skia:4175 BUG=skia:4284 ===================================================================== SkScanlineDecoder.h/.cpp: Removed. SkCodec.h/.cpp: Add methods, enums, and variables which were previously in SkScanlineDecoder. Default fCurrScanline to -1, as a sentinel that start has not been called. General changes: Convert SkScanlineDecoders to SkCodecs. General changes in SkCodec subclasses: Merge SkScanlineDecoder implementation into SkCodec. Most (all?) owned an SkCodec, so they now call this-> instead of fCodec->. SkBmpCodec.h/.cpp: Replace the unused rowOrder method with an override for onGetScanlineOrder. Make getDstRow const, since it is called by onGetY, which is const. SkCodec_libpng.h/.cpp: Make SkPngCodec an abstract class, with two subclasses which handle scanline decoding separately (they share code for decoding the entire image). Reimplement onReallyHasAlpha so that it can return the most recent result (e.g. after a scanline decode which only decoded part of the image) or a better answer (e.g. if the whole image is known to be opaque). Compute fNumberPasses early, so we know which subclass to instantiate. Make SkPngInterlaceScanlineDecoder use the base class' fCurrScanline rather than a separate variable. CodexTest.cpp: Add tests for the state changes in SkCodec (need to call start before decoding scanlines; calling getPixels means that start will need to be called again before decoding more scanlines). Add a test which decodes in stripes, currently only used for an interlaced PNG. TODO: Add tests for onReallyHasAlpha. Review URL: https://codereview.chromium.org/1365313002
2015-09-30 15:57:13 +00:00
*
* Results are undefined when not in scanline decoding mode.
*/
int nextScanline() const { return this->outputScanline(fCurrScanline); }
/**
* Returns the output y-coordinate of the row that corresponds to an input
* y-coordinate. The input y-coordinate represents where the scanline
* is located in the encoded data.
*
* This will equal inputScanline, except in the case of strangely
* encoded image types (bottom-up bmps, interlaced gifs).
*/
int outputScanline(int inputScanline) const;
Merge SkCodec with SkScanlineDecoder Benefits: - This mimics other decoding APIs (including the ones SkCodec relies on, e.g. a png_struct, which can be used to decode an entire image or one line at a time). - It allows a client to ask us to do what we can do efficiently - i.e. start from encoded data and either decode the whole thing or scanlines. - It removes the duplicate methods which appeared in both SkCodec and SkScanlineDecoder (some of which, e.g. in SkJpegScanlineDecoder, just call fCodec->sameMethod()). - It simplifies moving more checks into the base class (e.g. the examples in skbug.com/4284). BUG=skia:4175 BUG=skia:4284 ===================================================================== SkScanlineDecoder.h/.cpp: Removed. SkCodec.h/.cpp: Add methods, enums, and variables which were previously in SkScanlineDecoder. Default fCurrScanline to -1, as a sentinel that start has not been called. General changes: Convert SkScanlineDecoders to SkCodecs. General changes in SkCodec subclasses: Merge SkScanlineDecoder implementation into SkCodec. Most (all?) owned an SkCodec, so they now call this-> instead of fCodec->. SkBmpCodec.h/.cpp: Replace the unused rowOrder method with an override for onGetScanlineOrder. Make getDstRow const, since it is called by onGetY, which is const. SkCodec_libpng.h/.cpp: Make SkPngCodec an abstract class, with two subclasses which handle scanline decoding separately (they share code for decoding the entire image). Reimplement onReallyHasAlpha so that it can return the most recent result (e.g. after a scanline decode which only decoded part of the image) or a better answer (e.g. if the whole image is known to be opaque). Compute fNumberPasses early, so we know which subclass to instantiate. Make SkPngInterlaceScanlineDecoder use the base class' fCurrScanline rather than a separate variable. CodexTest.cpp: Add tests for the state changes in SkCodec (need to call start before decoding scanlines; calling getPixels means that start will need to be called again before decoding more scanlines). Add a test which decodes in stripes, currently only used for an interlaced PNG. TODO: Add tests for onReallyHasAlpha. Review URL: https://codereview.chromium.org/1365313002
2015-09-30 15:57:13 +00:00
Add support for multiple frames in SkCodec Add an interface to decode frames beyond the first in SkCodec, and add an implementation for SkGifCodec. Add getFrameData to SkCodec. This method reads ahead in the stream to return a vector containing meta data about each frame in the image. This is not required in order to decode frames beyond the first, but it allows a client to learn extra information: - how long the frame should be displayed - whether a frame should be blended with a prior frame, allowing the client to provide the prior frame to speed up decoding Add a new fields to SkCodec::Options: - fFrameIndex - fHasPriorFrame The API is designed so that SkCodec never caches frames. If a client wants a frame beyond the first, they specify the frame in Options.fFrameIndex. If the client does not have the frame's required frame (the frame that this frame must be blended on top of) cached, they pass false for Options.fHasPriorFrame. Unless the frame is independent, the codec will then recursively decode all frames necessary to decode fFrameIndex. If the client has the required frame cached, they can put it in the dst they pass to the codec, and the codec will only draw fFrameIndex onto it. Replace SkGifCodec's scanline decoding support with progressive decoding, and update the tests accordingly. Implement new APIs in SkGifCodec. Instead of using gif_lib, use GIFImageReader, imported from Chromium (along with its copyright headers) with the following changes: - SkGifCodec is now the client - Replace blink types - Combine GIFColorMap::buildTable and ::getTable into a method that creates and returns an SkColorTable - Input comes from an SkStream, instead of a SegmentReader. Add SkStreamBuffer, which buffers the (potentially partial) stream in order to decode progressively. (FIXME: This requires copying data that previously was read directly from the SegmentReader. Does this hurt performance? If so, can we fix it?) - Remove UMA code - Instead of reporting screen width and height to the client, allow the client to query for it - Fail earlier if the first frame AND screen have size of zero - Compute required previous frame when adding a new one - Move GIFParseQuery from GIFImageDecoder to GIFImageReader - Allow parsing up to a specific frame (to skip parsing the rest of the stream if a client only wants the first frame) - Compute whether the first frame has alpha and supports index 8, to create the SkImageInfo. This happens before reporting that the size has been decoded. Add GIFImageDecoder::haveDecodedRow to SkGifCodec, imported from Chromium (along with its copyright header), with the following changes: - Add support for sampling - Use the swizzler - Keep track of the rows decoded - Do *not* keep track of whether we've seen alpha Remove SkCodec::kOutOfOrder_SkScanlineOrder, which was only used by GIF scanline decoding. Call onRewind even if there is no stream (SkGifCodec needs to clear its decoded state so it will decode from the beginning). Add a method to SkSwizzler to access the offset into the dst, taking subsetting into account. Add a GM that animates a GIF. Add tests for the new APIs. *** Behavior changes: * Previously, we reported that an image with a subset frame and no transparent index was opaque and used the background index (if present) to fill the background. This is necessary in order to support index 8, but it does not match viewers/browsers I have seen. Examples: - Chromium and Gimp render the background transparent - Firefox, Safari, Linux Image Viewer, Safari Preview clip to the frame (for a single frame image) This CL matches Chromium's behavior and renders the background transparent. This allows us to have consistent behavior across products and simplifies the code (relative to what we would have to do to continue the old behavior on Android). It also means that we will no longer support index 8 for some GIFs. * Stop checking for GIFSTAMP - all GIFs should be either 89a or 87a. This matches Chromium. I suspect that bugs would have been reported if valid GIFs started with "GIFVER" instead of "GIF89a" or "GIF87a" (but did not decode in Chromium). *** Future work not included in this CL: * Move some checks out of haveDecodedRow, since they are the same for the entire frame e.g. - intersecting the frameRect with the full image size - whether there is a color table * Change when we write transparent pixels - In some cases, Chromium deemed this unnecessary, but I suspect it is slower than the fallback case. There will continue to be cases where we should *not* write them, but for e.g. the first pass where we have already cleared to transparent (which we may also be able to skip) writing the transparent pixels will not make anything incorrect. * Report color type and alpha type per frame - Depending on alpha values, disposal methods, frame rects, etc, subsequent frames may have different properties than the first. * Skip copies of the encoded data - We copy the encoded data in case the stream is one that cannot be rewound, so we can parse and then decode (possibly not immediately). For some input streams, this is unnecessary. - I was concerned this cause a performance regression, but on average the new code is faster than the old for the images I tested [1]. - It may cause a performance regression for Chromium, though, where we can always move back in the stream, so this should be addressed. Design doc: https://docs.google.com/a/google.com/document/d/12Qhf9T92MWfdWujQwCIjhCO3sw6pTJB5pJBwDM1T7Kc/ [1] https://docs.google.com/a/google.com/spreadsheets/d/19V-t9BfbFw5eiwBTKA1qOBkZbchjlTC5EIz6HFy-6RI/ GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=2045293002 Review-Url: https://codereview.chromium.org/2045293002
2016-10-24 16:03:26 +00:00
// The required frame for an independent frame is marked as
// kNone.
static constexpr size_t kNone = static_cast<size_t>(-1);
/**
* Information about individual frames in a multi-framed image.
*/
struct FrameInfo {
/**
* The frame that this frame needs to be blended with, or
* kNone.
*/
size_t fRequiredFrame;
/**
* Number of milliseconds to show this frame.
*/
size_t fDuration;
/**
* Whether the end marker for this frame is contained in the stream.
*
* Note: this does not guarantee that an attempt to decode will be complete.
* There could be an error in the stream.
*/
bool fFullyReceived;
Add support for multiple frames in SkCodec Add an interface to decode frames beyond the first in SkCodec, and add an implementation for SkGifCodec. Add getFrameData to SkCodec. This method reads ahead in the stream to return a vector containing meta data about each frame in the image. This is not required in order to decode frames beyond the first, but it allows a client to learn extra information: - how long the frame should be displayed - whether a frame should be blended with a prior frame, allowing the client to provide the prior frame to speed up decoding Add a new fields to SkCodec::Options: - fFrameIndex - fHasPriorFrame The API is designed so that SkCodec never caches frames. If a client wants a frame beyond the first, they specify the frame in Options.fFrameIndex. If the client does not have the frame's required frame (the frame that this frame must be blended on top of) cached, they pass false for Options.fHasPriorFrame. Unless the frame is independent, the codec will then recursively decode all frames necessary to decode fFrameIndex. If the client has the required frame cached, they can put it in the dst they pass to the codec, and the codec will only draw fFrameIndex onto it. Replace SkGifCodec's scanline decoding support with progressive decoding, and update the tests accordingly. Implement new APIs in SkGifCodec. Instead of using gif_lib, use GIFImageReader, imported from Chromium (along with its copyright headers) with the following changes: - SkGifCodec is now the client - Replace blink types - Combine GIFColorMap::buildTable and ::getTable into a method that creates and returns an SkColorTable - Input comes from an SkStream, instead of a SegmentReader. Add SkStreamBuffer, which buffers the (potentially partial) stream in order to decode progressively. (FIXME: This requires copying data that previously was read directly from the SegmentReader. Does this hurt performance? If so, can we fix it?) - Remove UMA code - Instead of reporting screen width and height to the client, allow the client to query for it - Fail earlier if the first frame AND screen have size of zero - Compute required previous frame when adding a new one - Move GIFParseQuery from GIFImageDecoder to GIFImageReader - Allow parsing up to a specific frame (to skip parsing the rest of the stream if a client only wants the first frame) - Compute whether the first frame has alpha and supports index 8, to create the SkImageInfo. This happens before reporting that the size has been decoded. Add GIFImageDecoder::haveDecodedRow to SkGifCodec, imported from Chromium (along with its copyright header), with the following changes: - Add support for sampling - Use the swizzler - Keep track of the rows decoded - Do *not* keep track of whether we've seen alpha Remove SkCodec::kOutOfOrder_SkScanlineOrder, which was only used by GIF scanline decoding. Call onRewind even if there is no stream (SkGifCodec needs to clear its decoded state so it will decode from the beginning). Add a method to SkSwizzler to access the offset into the dst, taking subsetting into account. Add a GM that animates a GIF. Add tests for the new APIs. *** Behavior changes: * Previously, we reported that an image with a subset frame and no transparent index was opaque and used the background index (if present) to fill the background. This is necessary in order to support index 8, but it does not match viewers/browsers I have seen. Examples: - Chromium and Gimp render the background transparent - Firefox, Safari, Linux Image Viewer, Safari Preview clip to the frame (for a single frame image) This CL matches Chromium's behavior and renders the background transparent. This allows us to have consistent behavior across products and simplifies the code (relative to what we would have to do to continue the old behavior on Android). It also means that we will no longer support index 8 for some GIFs. * Stop checking for GIFSTAMP - all GIFs should be either 89a or 87a. This matches Chromium. I suspect that bugs would have been reported if valid GIFs started with "GIFVER" instead of "GIF89a" or "GIF87a" (but did not decode in Chromium). *** Future work not included in this CL: * Move some checks out of haveDecodedRow, since they are the same for the entire frame e.g. - intersecting the frameRect with the full image size - whether there is a color table * Change when we write transparent pixels - In some cases, Chromium deemed this unnecessary, but I suspect it is slower than the fallback case. There will continue to be cases where we should *not* write them, but for e.g. the first pass where we have already cleared to transparent (which we may also be able to skip) writing the transparent pixels will not make anything incorrect. * Report color type and alpha type per frame - Depending on alpha values, disposal methods, frame rects, etc, subsequent frames may have different properties than the first. * Skip copies of the encoded data - We copy the encoded data in case the stream is one that cannot be rewound, so we can parse and then decode (possibly not immediately). For some input streams, this is unnecessary. - I was concerned this cause a performance regression, but on average the new code is faster than the old for the images I tested [1]. - It may cause a performance regression for Chromium, though, where we can always move back in the stream, so this should be addressed. Design doc: https://docs.google.com/a/google.com/document/d/12Qhf9T92MWfdWujQwCIjhCO3sw6pTJB5pJBwDM1T7Kc/ [1] https://docs.google.com/a/google.com/spreadsheets/d/19V-t9BfbFw5eiwBTKA1qOBkZbchjlTC5EIz6HFy-6RI/ GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=2045293002 Review-Url: https://codereview.chromium.org/2045293002
2016-10-24 16:03:26 +00:00
};
/**
* Return info about the frames in the image.
*
* May require reading through the stream to determine info about the
* frames (including the count).
Add support for multiple frames in SkCodec Add an interface to decode frames beyond the first in SkCodec, and add an implementation for SkGifCodec. Add getFrameData to SkCodec. This method reads ahead in the stream to return a vector containing meta data about each frame in the image. This is not required in order to decode frames beyond the first, but it allows a client to learn extra information: - how long the frame should be displayed - whether a frame should be blended with a prior frame, allowing the client to provide the prior frame to speed up decoding Add a new fields to SkCodec::Options: - fFrameIndex - fHasPriorFrame The API is designed so that SkCodec never caches frames. If a client wants a frame beyond the first, they specify the frame in Options.fFrameIndex. If the client does not have the frame's required frame (the frame that this frame must be blended on top of) cached, they pass false for Options.fHasPriorFrame. Unless the frame is independent, the codec will then recursively decode all frames necessary to decode fFrameIndex. If the client has the required frame cached, they can put it in the dst they pass to the codec, and the codec will only draw fFrameIndex onto it. Replace SkGifCodec's scanline decoding support with progressive decoding, and update the tests accordingly. Implement new APIs in SkGifCodec. Instead of using gif_lib, use GIFImageReader, imported from Chromium (along with its copyright headers) with the following changes: - SkGifCodec is now the client - Replace blink types - Combine GIFColorMap::buildTable and ::getTable into a method that creates and returns an SkColorTable - Input comes from an SkStream, instead of a SegmentReader. Add SkStreamBuffer, which buffers the (potentially partial) stream in order to decode progressively. (FIXME: This requires copying data that previously was read directly from the SegmentReader. Does this hurt performance? If so, can we fix it?) - Remove UMA code - Instead of reporting screen width and height to the client, allow the client to query for it - Fail earlier if the first frame AND screen have size of zero - Compute required previous frame when adding a new one - Move GIFParseQuery from GIFImageDecoder to GIFImageReader - Allow parsing up to a specific frame (to skip parsing the rest of the stream if a client only wants the first frame) - Compute whether the first frame has alpha and supports index 8, to create the SkImageInfo. This happens before reporting that the size has been decoded. Add GIFImageDecoder::haveDecodedRow to SkGifCodec, imported from Chromium (along with its copyright header), with the following changes: - Add support for sampling - Use the swizzler - Keep track of the rows decoded - Do *not* keep track of whether we've seen alpha Remove SkCodec::kOutOfOrder_SkScanlineOrder, which was only used by GIF scanline decoding. Call onRewind even if there is no stream (SkGifCodec needs to clear its decoded state so it will decode from the beginning). Add a method to SkSwizzler to access the offset into the dst, taking subsetting into account. Add a GM that animates a GIF. Add tests for the new APIs. *** Behavior changes: * Previously, we reported that an image with a subset frame and no transparent index was opaque and used the background index (if present) to fill the background. This is necessary in order to support index 8, but it does not match viewers/browsers I have seen. Examples: - Chromium and Gimp render the background transparent - Firefox, Safari, Linux Image Viewer, Safari Preview clip to the frame (for a single frame image) This CL matches Chromium's behavior and renders the background transparent. This allows us to have consistent behavior across products and simplifies the code (relative to what we would have to do to continue the old behavior on Android). It also means that we will no longer support index 8 for some GIFs. * Stop checking for GIFSTAMP - all GIFs should be either 89a or 87a. This matches Chromium. I suspect that bugs would have been reported if valid GIFs started with "GIFVER" instead of "GIF89a" or "GIF87a" (but did not decode in Chromium). *** Future work not included in this CL: * Move some checks out of haveDecodedRow, since they are the same for the entire frame e.g. - intersecting the frameRect with the full image size - whether there is a color table * Change when we write transparent pixels - In some cases, Chromium deemed this unnecessary, but I suspect it is slower than the fallback case. There will continue to be cases where we should *not* write them, but for e.g. the first pass where we have already cleared to transparent (which we may also be able to skip) writing the transparent pixels will not make anything incorrect. * Report color type and alpha type per frame - Depending on alpha values, disposal methods, frame rects, etc, subsequent frames may have different properties than the first. * Skip copies of the encoded data - We copy the encoded data in case the stream is one that cannot be rewound, so we can parse and then decode (possibly not immediately). For some input streams, this is unnecessary. - I was concerned this cause a performance regression, but on average the new code is faster than the old for the images I tested [1]. - It may cause a performance regression for Chromium, though, where we can always move back in the stream, so this should be addressed. Design doc: https://docs.google.com/a/google.com/document/d/12Qhf9T92MWfdWujQwCIjhCO3sw6pTJB5pJBwDM1T7Kc/ [1] https://docs.google.com/a/google.com/spreadsheets/d/19V-t9BfbFw5eiwBTKA1qOBkZbchjlTC5EIz6HFy-6RI/ GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=2045293002 Review-Url: https://codereview.chromium.org/2045293002
2016-10-24 16:03:26 +00:00
*
* As such, future decoding calls may require a rewind.
*
* For single-frame images, this will return an empty vector.
*/
std::vector<FrameInfo> getFrameInfo() {
return this->onGetFrameInfo();
}
static constexpr int kRepetitionCountInfinite = -1;
/**
* Return the number of times to repeat, if this image is animated.
*
* May require reading the stream to find the repetition count.
*
* As such, future decoding calls may require a rewind.
*
* For single-frame images, this will return 0.
*/
int getRepetitionCount() {
return this->onGetRepetitionCount();
}
protected:
/**
* Takes ownership of SkStream*
*/
SkCodec(int width,
int height,
const SkEncodedInfo&,
SkStream*,
sk_sp<SkColorSpace>,
Origin = kTopLeft_Origin);
/**
* Takes ownership of SkStream*
* Allows the subclass to set the recommended SkImageInfo
*/
SkCodec(const SkEncodedInfo&,
const SkImageInfo&,
SkStream*,
Origin = kTopLeft_Origin);
virtual SkISize onGetScaledDimensions(float /*desiredScale*/) const {
// By default, scaling is not supported.
return this->getInfo().dimensions();
}
// FIXME: What to do about subsets??
/**
* Subclasses should override if they support dimensions other than the
* srcInfo's.
*/
virtual bool onDimensionsSupported(const SkISize&) {
return false;
}
virtual SkEncodedImageFormat onGetEncodedFormat() const = 0;
/**
* @param rowsDecoded When the encoded image stream is incomplete, this function
* will return kIncompleteInput and rowsDecoded will be set to
* the number of scanlines that were successfully decoded.
* This will allow getPixels() to fill the uninitialized memory.
*/
virtual Result onGetPixels(const SkImageInfo& info,
void* pixels, size_t rowBytes, const Options&,
SkPMColor ctable[], int* ctableCount,
int* rowsDecoded) = 0;
virtual bool onQueryYUV8(SkYUVSizeInfo*, SkYUVColorSpace*) const {
return false;
}
virtual Result onGetYUV8Planes(const SkYUVSizeInfo&, void*[3] /*planes*/) {
return kUnimplemented;
}
virtual bool onGetValidSubset(SkIRect* /*desiredSubset*/) const {
// By default, subsets are not supported.
return false;
}
/**
* If the stream was previously read, attempt to rewind.
*
* If the stream needed to be rewound, call onRewind.
* @returns true if the codec is at the right position and can be used.
* false if there was a failure to rewind.
*
* This is called by getPixels() and start(). Subclasses may call if they
* need to rewind at another time.
*/
bool SK_WARN_UNUSED_RESULT rewindIfNeeded();
/**
* Called by rewindIfNeeded, if the stream needed to be rewound.
*
* Subclasses should do any set up needed after a rewind.
*/
virtual bool onRewind() {
return true;
}
/**
* On an incomplete input, getPixels() and getScanlines() will fill any uninitialized
* scanlines. This allows the subclass to indicate what value to fill with.
*
* @param dstInfo Describes the destination.
* @return The value with which to fill uninitialized pixels.
*
* Note that we can interpret the return value as a 64-bit Float16 color, a SkPMColor,
* a 16-bit 565 color, an 8-bit gray color, or an 8-bit index into a color table,
* depending on the color type.
*/
uint64_t getFillValue(const SkImageInfo& dstInfo) const {
return this->onGetFillValue(dstInfo);
}
/**
* Some subclasses will override this function, but this is a useful default for the color
* types that we support. Note that for color types that do not use the full 64-bits,
* we will simply take the low bits of the fill value.
*
* The defaults are:
* kRGBA_F16_SkColorType: Transparent or Black, depending on the src alpha type
* kN32_SkColorType: Transparent or Black, depending on the src alpha type
* kRGB_565_SkColorType: Black
* kGray_8_SkColorType: Black
* kIndex_8_SkColorType: First color in color table
*/
virtual uint64_t onGetFillValue(const SkImageInfo& dstInfo) const;
/**
* Get method for the input stream
*/
SkStream* stream() {
return fStream.get();
}
Merge SkCodec with SkScanlineDecoder Benefits: - This mimics other decoding APIs (including the ones SkCodec relies on, e.g. a png_struct, which can be used to decode an entire image or one line at a time). - It allows a client to ask us to do what we can do efficiently - i.e. start from encoded data and either decode the whole thing or scanlines. - It removes the duplicate methods which appeared in both SkCodec and SkScanlineDecoder (some of which, e.g. in SkJpegScanlineDecoder, just call fCodec->sameMethod()). - It simplifies moving more checks into the base class (e.g. the examples in skbug.com/4284). BUG=skia:4175 BUG=skia:4284 ===================================================================== SkScanlineDecoder.h/.cpp: Removed. SkCodec.h/.cpp: Add methods, enums, and variables which were previously in SkScanlineDecoder. Default fCurrScanline to -1, as a sentinel that start has not been called. General changes: Convert SkScanlineDecoders to SkCodecs. General changes in SkCodec subclasses: Merge SkScanlineDecoder implementation into SkCodec. Most (all?) owned an SkCodec, so they now call this-> instead of fCodec->. SkBmpCodec.h/.cpp: Replace the unused rowOrder method with an override for onGetScanlineOrder. Make getDstRow const, since it is called by onGetY, which is const. SkCodec_libpng.h/.cpp: Make SkPngCodec an abstract class, with two subclasses which handle scanline decoding separately (they share code for decoding the entire image). Reimplement onReallyHasAlpha so that it can return the most recent result (e.g. after a scanline decode which only decoded part of the image) or a better answer (e.g. if the whole image is known to be opaque). Compute fNumberPasses early, so we know which subclass to instantiate. Make SkPngInterlaceScanlineDecoder use the base class' fCurrScanline rather than a separate variable. CodexTest.cpp: Add tests for the state changes in SkCodec (need to call start before decoding scanlines; calling getPixels means that start will need to be called again before decoding more scanlines). Add a test which decodes in stripes, currently only used for an interlaced PNG. TODO: Add tests for onReallyHasAlpha. Review URL: https://codereview.chromium.org/1365313002
2015-09-30 15:57:13 +00:00
/**
* The remaining functions revolve around decoding scanlines.
*/
/**
* Most images types will be kTopDown and will not need to override this function.
*/
virtual SkScanlineOrder onGetScanlineOrder() const { return kTopDown_SkScanlineOrder; }
const SkImageInfo& dstInfo() const { return fDstInfo; }
const SkCodec::Options& options() const { return fOptions; }
/**
* Returns the number of scanlines that have been decoded so far.
* This is unaffected by the SkScanlineOrder.
*
* Returns -1 if we have not started a scanline decode.
*/
int currScanline() const { return fCurrScanline; }
virtual int onOutputScanline(int inputScanline) const;
bool initializeColorXform(const SkImageInfo& dstInfo,
SkTransferFunctionBehavior premulBehavior);
SkColorSpaceXform* colorXform() const { return fColorXform.get(); }
Add support for multiple frames in SkCodec Add an interface to decode frames beyond the first in SkCodec, and add an implementation for SkGifCodec. Add getFrameData to SkCodec. This method reads ahead in the stream to return a vector containing meta data about each frame in the image. This is not required in order to decode frames beyond the first, but it allows a client to learn extra information: - how long the frame should be displayed - whether a frame should be blended with a prior frame, allowing the client to provide the prior frame to speed up decoding Add a new fields to SkCodec::Options: - fFrameIndex - fHasPriorFrame The API is designed so that SkCodec never caches frames. If a client wants a frame beyond the first, they specify the frame in Options.fFrameIndex. If the client does not have the frame's required frame (the frame that this frame must be blended on top of) cached, they pass false for Options.fHasPriorFrame. Unless the frame is independent, the codec will then recursively decode all frames necessary to decode fFrameIndex. If the client has the required frame cached, they can put it in the dst they pass to the codec, and the codec will only draw fFrameIndex onto it. Replace SkGifCodec's scanline decoding support with progressive decoding, and update the tests accordingly. Implement new APIs in SkGifCodec. Instead of using gif_lib, use GIFImageReader, imported from Chromium (along with its copyright headers) with the following changes: - SkGifCodec is now the client - Replace blink types - Combine GIFColorMap::buildTable and ::getTable into a method that creates and returns an SkColorTable - Input comes from an SkStream, instead of a SegmentReader. Add SkStreamBuffer, which buffers the (potentially partial) stream in order to decode progressively. (FIXME: This requires copying data that previously was read directly from the SegmentReader. Does this hurt performance? If so, can we fix it?) - Remove UMA code - Instead of reporting screen width and height to the client, allow the client to query for it - Fail earlier if the first frame AND screen have size of zero - Compute required previous frame when adding a new one - Move GIFParseQuery from GIFImageDecoder to GIFImageReader - Allow parsing up to a specific frame (to skip parsing the rest of the stream if a client only wants the first frame) - Compute whether the first frame has alpha and supports index 8, to create the SkImageInfo. This happens before reporting that the size has been decoded. Add GIFImageDecoder::haveDecodedRow to SkGifCodec, imported from Chromium (along with its copyright header), with the following changes: - Add support for sampling - Use the swizzler - Keep track of the rows decoded - Do *not* keep track of whether we've seen alpha Remove SkCodec::kOutOfOrder_SkScanlineOrder, which was only used by GIF scanline decoding. Call onRewind even if there is no stream (SkGifCodec needs to clear its decoded state so it will decode from the beginning). Add a method to SkSwizzler to access the offset into the dst, taking subsetting into account. Add a GM that animates a GIF. Add tests for the new APIs. *** Behavior changes: * Previously, we reported that an image with a subset frame and no transparent index was opaque and used the background index (if present) to fill the background. This is necessary in order to support index 8, but it does not match viewers/browsers I have seen. Examples: - Chromium and Gimp render the background transparent - Firefox, Safari, Linux Image Viewer, Safari Preview clip to the frame (for a single frame image) This CL matches Chromium's behavior and renders the background transparent. This allows us to have consistent behavior across products and simplifies the code (relative to what we would have to do to continue the old behavior on Android). It also means that we will no longer support index 8 for some GIFs. * Stop checking for GIFSTAMP - all GIFs should be either 89a or 87a. This matches Chromium. I suspect that bugs would have been reported if valid GIFs started with "GIFVER" instead of "GIF89a" or "GIF87a" (but did not decode in Chromium). *** Future work not included in this CL: * Move some checks out of haveDecodedRow, since they are the same for the entire frame e.g. - intersecting the frameRect with the full image size - whether there is a color table * Change when we write transparent pixels - In some cases, Chromium deemed this unnecessary, but I suspect it is slower than the fallback case. There will continue to be cases where we should *not* write them, but for e.g. the first pass where we have already cleared to transparent (which we may also be able to skip) writing the transparent pixels will not make anything incorrect. * Report color type and alpha type per frame - Depending on alpha values, disposal methods, frame rects, etc, subsequent frames may have different properties than the first. * Skip copies of the encoded data - We copy the encoded data in case the stream is one that cannot be rewound, so we can parse and then decode (possibly not immediately). For some input streams, this is unnecessary. - I was concerned this cause a performance regression, but on average the new code is faster than the old for the images I tested [1]. - It may cause a performance regression for Chromium, though, where we can always move back in the stream, so this should be addressed. Design doc: https://docs.google.com/a/google.com/document/d/12Qhf9T92MWfdWujQwCIjhCO3sw6pTJB5pJBwDM1T7Kc/ [1] https://docs.google.com/a/google.com/spreadsheets/d/19V-t9BfbFw5eiwBTKA1qOBkZbchjlTC5EIz6HFy-6RI/ GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=2045293002 Review-Url: https://codereview.chromium.org/2045293002
2016-10-24 16:03:26 +00:00
virtual std::vector<FrameInfo> onGetFrameInfo() {
// empty vector - this is not animated.
return std::vector<FrameInfo>{};
Add support for multiple frames in SkCodec Add an interface to decode frames beyond the first in SkCodec, and add an implementation for SkGifCodec. Add getFrameData to SkCodec. This method reads ahead in the stream to return a vector containing meta data about each frame in the image. This is not required in order to decode frames beyond the first, but it allows a client to learn extra information: - how long the frame should be displayed - whether a frame should be blended with a prior frame, allowing the client to provide the prior frame to speed up decoding Add a new fields to SkCodec::Options: - fFrameIndex - fHasPriorFrame The API is designed so that SkCodec never caches frames. If a client wants a frame beyond the first, they specify the frame in Options.fFrameIndex. If the client does not have the frame's required frame (the frame that this frame must be blended on top of) cached, they pass false for Options.fHasPriorFrame. Unless the frame is independent, the codec will then recursively decode all frames necessary to decode fFrameIndex. If the client has the required frame cached, they can put it in the dst they pass to the codec, and the codec will only draw fFrameIndex onto it. Replace SkGifCodec's scanline decoding support with progressive decoding, and update the tests accordingly. Implement new APIs in SkGifCodec. Instead of using gif_lib, use GIFImageReader, imported from Chromium (along with its copyright headers) with the following changes: - SkGifCodec is now the client - Replace blink types - Combine GIFColorMap::buildTable and ::getTable into a method that creates and returns an SkColorTable - Input comes from an SkStream, instead of a SegmentReader. Add SkStreamBuffer, which buffers the (potentially partial) stream in order to decode progressively. (FIXME: This requires copying data that previously was read directly from the SegmentReader. Does this hurt performance? If so, can we fix it?) - Remove UMA code - Instead of reporting screen width and height to the client, allow the client to query for it - Fail earlier if the first frame AND screen have size of zero - Compute required previous frame when adding a new one - Move GIFParseQuery from GIFImageDecoder to GIFImageReader - Allow parsing up to a specific frame (to skip parsing the rest of the stream if a client only wants the first frame) - Compute whether the first frame has alpha and supports index 8, to create the SkImageInfo. This happens before reporting that the size has been decoded. Add GIFImageDecoder::haveDecodedRow to SkGifCodec, imported from Chromium (along with its copyright header), with the following changes: - Add support for sampling - Use the swizzler - Keep track of the rows decoded - Do *not* keep track of whether we've seen alpha Remove SkCodec::kOutOfOrder_SkScanlineOrder, which was only used by GIF scanline decoding. Call onRewind even if there is no stream (SkGifCodec needs to clear its decoded state so it will decode from the beginning). Add a method to SkSwizzler to access the offset into the dst, taking subsetting into account. Add a GM that animates a GIF. Add tests for the new APIs. *** Behavior changes: * Previously, we reported that an image with a subset frame and no transparent index was opaque and used the background index (if present) to fill the background. This is necessary in order to support index 8, but it does not match viewers/browsers I have seen. Examples: - Chromium and Gimp render the background transparent - Firefox, Safari, Linux Image Viewer, Safari Preview clip to the frame (for a single frame image) This CL matches Chromium's behavior and renders the background transparent. This allows us to have consistent behavior across products and simplifies the code (relative to what we would have to do to continue the old behavior on Android). It also means that we will no longer support index 8 for some GIFs. * Stop checking for GIFSTAMP - all GIFs should be either 89a or 87a. This matches Chromium. I suspect that bugs would have been reported if valid GIFs started with "GIFVER" instead of "GIF89a" or "GIF87a" (but did not decode in Chromium). *** Future work not included in this CL: * Move some checks out of haveDecodedRow, since they are the same for the entire frame e.g. - intersecting the frameRect with the full image size - whether there is a color table * Change when we write transparent pixels - In some cases, Chromium deemed this unnecessary, but I suspect it is slower than the fallback case. There will continue to be cases where we should *not* write them, but for e.g. the first pass where we have already cleared to transparent (which we may also be able to skip) writing the transparent pixels will not make anything incorrect. * Report color type and alpha type per frame - Depending on alpha values, disposal methods, frame rects, etc, subsequent frames may have different properties than the first. * Skip copies of the encoded data - We copy the encoded data in case the stream is one that cannot be rewound, so we can parse and then decode (possibly not immediately). For some input streams, this is unnecessary. - I was concerned this cause a performance regression, but on average the new code is faster than the old for the images I tested [1]. - It may cause a performance regression for Chromium, though, where we can always move back in the stream, so this should be addressed. Design doc: https://docs.google.com/a/google.com/document/d/12Qhf9T92MWfdWujQwCIjhCO3sw6pTJB5pJBwDM1T7Kc/ [1] https://docs.google.com/a/google.com/spreadsheets/d/19V-t9BfbFw5eiwBTKA1qOBkZbchjlTC5EIz6HFy-6RI/ GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=2045293002 Review-Url: https://codereview.chromium.org/2045293002
2016-10-24 16:03:26 +00:00
}
virtual int onGetRepetitionCount() {
return 0;
}
void setUnsupportedICC(bool SkDEBUGCODE(value)) { SkDEBUGCODE(fUnsupportedICC = value); }
private:
const SkEncodedInfo fEncodedInfo;
const SkImageInfo fSrcInfo;
std::unique_ptr<SkStream> fStream;
bool fNeedsRewind;
const Origin fOrigin;
SkImageInfo fDstInfo;
SkCodec::Options fOptions;
std::unique_ptr<SkColorSpaceXform> fColorXform;
Make SkPngCodec decode progressively. This is a step towards using SkCodec in Chromium, where progressive decoding is necessary. Switch from using png_read_row (which expects all the data to be available) to png_process_data, which uses callbacks when rows are available. Create a new API for SkCodec, which supports progressive decoding and scanline decoding. Future changes will switch the other clients off of startScanlineDecode and get/skip-Scanlines to the new API. Remove SkCodec::kNone_ScanlineOrder, which was only used for interlaced PNG images. In the new API, interlaced PNG fits kTopDown. Also remove updateCurrScanline(), which was only used by the old implementation for interlaced PNG. DMSrcSink: - In CodecSrc::kScanline_Mode, use the new method for scanline decoding for the supported formats (just PNG and PNG-in-ICO for now). fuzz.cpp: - Remove reference to kNone_ScanlineOrder SkCodec: - Add new APIs: - startIncrementalDecode - incrementalDecode - Remove kNone_SkScanlineOrder and updateCurrScanline() - Set fDstInfo and fOptions in getPixels(). This may not be necessary for all implementations, but it simplifies things for SkPngCodec. SkPngCodec: - Implement new APIs - Switch from sk_read_fn/png_read_row etc to png_process_data - Expand AutoCleanPng's role to decode the header and create the SkPngCodec - Make the interlaced PNG decoder report how many lines were initialized during an incomplete decode SkIcoCodec: - Implement the new APIs; supported for PNG in ICO SkSampledCodec: - Call the new method for decoding scanlines, and fall back to the old method if the new version is unimplemented - Remove references to kNone_SkScanlineOrder tests/CodecPartial: - Add a test which decodes part of an image, then finishes the decode, and compares it to the straightforward method tests/CodecTest: - Add a test which decodes all scanlines using the new method - Repurpose the Codec_stripes test to decode using the new method in sections rather than all at once - In the method check(), add a parameter for whether the image supports the new method of scanline decoding, and be explicit about whether an image supports incomplete - Test incomplete PNG decodes. We should have been doing it anyway for non-interlaced (except for an image that is too small - one row), but the new method supports interlaced incomplete as well - Make test_invalid_parameters test the new method - Add a test to ensure that it's safe to fall back to scanline decoding without rewinding BUG=skia:4211 The new version was generally faster than the old version (but not significantly so). Some raw performance differences can be found at https://docs.google.com/a/google.com/spreadsheets/d/1Gis3aRCEa72qBNDRMgGDg3jD-pMgO-FXldlNF9ejo4o/ Design doc can be found at https://docs.google.com/a/google.com/document/d/11Mn8-ePDKwVEMCjs3nWwSjxcSpJ_Cu8DF57KNtUmgLM/ GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1997703003 Review-Url: https://codereview.chromium.org/1997703003
2016-09-16 15:20:38 +00:00
// Only meaningful during scanline decodes.
int fCurrScanline;
Merge SkCodec with SkScanlineDecoder Benefits: - This mimics other decoding APIs (including the ones SkCodec relies on, e.g. a png_struct, which can be used to decode an entire image or one line at a time). - It allows a client to ask us to do what we can do efficiently - i.e. start from encoded data and either decode the whole thing or scanlines. - It removes the duplicate methods which appeared in both SkCodec and SkScanlineDecoder (some of which, e.g. in SkJpegScanlineDecoder, just call fCodec->sameMethod()). - It simplifies moving more checks into the base class (e.g. the examples in skbug.com/4284). BUG=skia:4175 BUG=skia:4284 ===================================================================== SkScanlineDecoder.h/.cpp: Removed. SkCodec.h/.cpp: Add methods, enums, and variables which were previously in SkScanlineDecoder. Default fCurrScanline to -1, as a sentinel that start has not been called. General changes: Convert SkScanlineDecoders to SkCodecs. General changes in SkCodec subclasses: Merge SkScanlineDecoder implementation into SkCodec. Most (all?) owned an SkCodec, so they now call this-> instead of fCodec->. SkBmpCodec.h/.cpp: Replace the unused rowOrder method with an override for onGetScanlineOrder. Make getDstRow const, since it is called by onGetY, which is const. SkCodec_libpng.h/.cpp: Make SkPngCodec an abstract class, with two subclasses which handle scanline decoding separately (they share code for decoding the entire image). Reimplement onReallyHasAlpha so that it can return the most recent result (e.g. after a scanline decode which only decoded part of the image) or a better answer (e.g. if the whole image is known to be opaque). Compute fNumberPasses early, so we know which subclass to instantiate. Make SkPngInterlaceScanlineDecoder use the base class' fCurrScanline rather than a separate variable. CodexTest.cpp: Add tests for the state changes in SkCodec (need to call start before decoding scanlines; calling getPixels means that start will need to be called again before decoding more scanlines). Add a test which decodes in stripes, currently only used for an interlaced PNG. TODO: Add tests for onReallyHasAlpha. Review URL: https://codereview.chromium.org/1365313002
2015-09-30 15:57:13 +00:00
bool fStartedIncrementalDecode;
#ifdef SK_DEBUG
bool fUnsupportedICC = false;
#endif
Make SkPngCodec decode progressively. This is a step towards using SkCodec in Chromium, where progressive decoding is necessary. Switch from using png_read_row (which expects all the data to be available) to png_process_data, which uses callbacks when rows are available. Create a new API for SkCodec, which supports progressive decoding and scanline decoding. Future changes will switch the other clients off of startScanlineDecode and get/skip-Scanlines to the new API. Remove SkCodec::kNone_ScanlineOrder, which was only used for interlaced PNG images. In the new API, interlaced PNG fits kTopDown. Also remove updateCurrScanline(), which was only used by the old implementation for interlaced PNG. DMSrcSink: - In CodecSrc::kScanline_Mode, use the new method for scanline decoding for the supported formats (just PNG and PNG-in-ICO for now). fuzz.cpp: - Remove reference to kNone_ScanlineOrder SkCodec: - Add new APIs: - startIncrementalDecode - incrementalDecode - Remove kNone_SkScanlineOrder and updateCurrScanline() - Set fDstInfo and fOptions in getPixels(). This may not be necessary for all implementations, but it simplifies things for SkPngCodec. SkPngCodec: - Implement new APIs - Switch from sk_read_fn/png_read_row etc to png_process_data - Expand AutoCleanPng's role to decode the header and create the SkPngCodec - Make the interlaced PNG decoder report how many lines were initialized during an incomplete decode SkIcoCodec: - Implement the new APIs; supported for PNG in ICO SkSampledCodec: - Call the new method for decoding scanlines, and fall back to the old method if the new version is unimplemented - Remove references to kNone_SkScanlineOrder tests/CodecPartial: - Add a test which decodes part of an image, then finishes the decode, and compares it to the straightforward method tests/CodecTest: - Add a test which decodes all scanlines using the new method - Repurpose the Codec_stripes test to decode using the new method in sections rather than all at once - In the method check(), add a parameter for whether the image supports the new method of scanline decoding, and be explicit about whether an image supports incomplete - Test incomplete PNG decodes. We should have been doing it anyway for non-interlaced (except for an image that is too small - one row), but the new method supports interlaced incomplete as well - Make test_invalid_parameters test the new method - Add a test to ensure that it's safe to fall back to scanline decoding without rewinding BUG=skia:4211 The new version was generally faster than the old version (but not significantly so). Some raw performance differences can be found at https://docs.google.com/a/google.com/spreadsheets/d/1Gis3aRCEa72qBNDRMgGDg3jD-pMgO-FXldlNF9ejo4o/ Design doc can be found at https://docs.google.com/a/google.com/document/d/11Mn8-ePDKwVEMCjs3nWwSjxcSpJ_Cu8DF57KNtUmgLM/ GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1997703003 Review-Url: https://codereview.chromium.org/1997703003
2016-09-16 15:20:38 +00:00
/**
* Return whether these dimensions are supported as a scale.
*
* The codec may choose to cache the information about scale and subset.
* Either way, the same information will be passed to onGetPixels/onStart
* on success.
*
* This must return true for a size returned from getScaledDimensions.
*/
bool dimensionsSupported(const SkISize& dim) {
return dim == fSrcInfo.dimensions() || this->onDimensionsSupported(dim);
}
Merge SkCodec with SkScanlineDecoder Benefits: - This mimics other decoding APIs (including the ones SkCodec relies on, e.g. a png_struct, which can be used to decode an entire image or one line at a time). - It allows a client to ask us to do what we can do efficiently - i.e. start from encoded data and either decode the whole thing or scanlines. - It removes the duplicate methods which appeared in both SkCodec and SkScanlineDecoder (some of which, e.g. in SkJpegScanlineDecoder, just call fCodec->sameMethod()). - It simplifies moving more checks into the base class (e.g. the examples in skbug.com/4284). BUG=skia:4175 BUG=skia:4284 ===================================================================== SkScanlineDecoder.h/.cpp: Removed. SkCodec.h/.cpp: Add methods, enums, and variables which were previously in SkScanlineDecoder. Default fCurrScanline to -1, as a sentinel that start has not been called. General changes: Convert SkScanlineDecoders to SkCodecs. General changes in SkCodec subclasses: Merge SkScanlineDecoder implementation into SkCodec. Most (all?) owned an SkCodec, so they now call this-> instead of fCodec->. SkBmpCodec.h/.cpp: Replace the unused rowOrder method with an override for onGetScanlineOrder. Make getDstRow const, since it is called by onGetY, which is const. SkCodec_libpng.h/.cpp: Make SkPngCodec an abstract class, with two subclasses which handle scanline decoding separately (they share code for decoding the entire image). Reimplement onReallyHasAlpha so that it can return the most recent result (e.g. after a scanline decode which only decoded part of the image) or a better answer (e.g. if the whole image is known to be opaque). Compute fNumberPasses early, so we know which subclass to instantiate. Make SkPngInterlaceScanlineDecoder use the base class' fCurrScanline rather than a separate variable. CodexTest.cpp: Add tests for the state changes in SkCodec (need to call start before decoding scanlines; calling getPixels means that start will need to be called again before decoding more scanlines). Add a test which decodes in stripes, currently only used for an interlaced PNG. TODO: Add tests for onReallyHasAlpha. Review URL: https://codereview.chromium.org/1365313002
2015-09-30 15:57:13 +00:00
// Methods for scanline decoding.
virtual SkCodec::Result onStartScanlineDecode(const SkImageInfo& /*dstInfo*/,
const SkCodec::Options& /*options*/, SkPMColor* /*ctable*/, int* /*ctableCount*/) {
Merge SkCodec with SkScanlineDecoder Benefits: - This mimics other decoding APIs (including the ones SkCodec relies on, e.g. a png_struct, which can be used to decode an entire image or one line at a time). - It allows a client to ask us to do what we can do efficiently - i.e. start from encoded data and either decode the whole thing or scanlines. - It removes the duplicate methods which appeared in both SkCodec and SkScanlineDecoder (some of which, e.g. in SkJpegScanlineDecoder, just call fCodec->sameMethod()). - It simplifies moving more checks into the base class (e.g. the examples in skbug.com/4284). BUG=skia:4175 BUG=skia:4284 ===================================================================== SkScanlineDecoder.h/.cpp: Removed. SkCodec.h/.cpp: Add methods, enums, and variables which were previously in SkScanlineDecoder. Default fCurrScanline to -1, as a sentinel that start has not been called. General changes: Convert SkScanlineDecoders to SkCodecs. General changes in SkCodec subclasses: Merge SkScanlineDecoder implementation into SkCodec. Most (all?) owned an SkCodec, so they now call this-> instead of fCodec->. SkBmpCodec.h/.cpp: Replace the unused rowOrder method with an override for onGetScanlineOrder. Make getDstRow const, since it is called by onGetY, which is const. SkCodec_libpng.h/.cpp: Make SkPngCodec an abstract class, with two subclasses which handle scanline decoding separately (they share code for decoding the entire image). Reimplement onReallyHasAlpha so that it can return the most recent result (e.g. after a scanline decode which only decoded part of the image) or a better answer (e.g. if the whole image is known to be opaque). Compute fNumberPasses early, so we know which subclass to instantiate. Make SkPngInterlaceScanlineDecoder use the base class' fCurrScanline rather than a separate variable. CodexTest.cpp: Add tests for the state changes in SkCodec (need to call start before decoding scanlines; calling getPixels means that start will need to be called again before decoding more scanlines). Add a test which decodes in stripes, currently only used for an interlaced PNG. TODO: Add tests for onReallyHasAlpha. Review URL: https://codereview.chromium.org/1365313002
2015-09-30 15:57:13 +00:00
return kUnimplemented;
}
Make SkPngCodec decode progressively. This is a step towards using SkCodec in Chromium, where progressive decoding is necessary. Switch from using png_read_row (which expects all the data to be available) to png_process_data, which uses callbacks when rows are available. Create a new API for SkCodec, which supports progressive decoding and scanline decoding. Future changes will switch the other clients off of startScanlineDecode and get/skip-Scanlines to the new API. Remove SkCodec::kNone_ScanlineOrder, which was only used for interlaced PNG images. In the new API, interlaced PNG fits kTopDown. Also remove updateCurrScanline(), which was only used by the old implementation for interlaced PNG. DMSrcSink: - In CodecSrc::kScanline_Mode, use the new method for scanline decoding for the supported formats (just PNG and PNG-in-ICO for now). fuzz.cpp: - Remove reference to kNone_ScanlineOrder SkCodec: - Add new APIs: - startIncrementalDecode - incrementalDecode - Remove kNone_SkScanlineOrder and updateCurrScanline() - Set fDstInfo and fOptions in getPixels(). This may not be necessary for all implementations, but it simplifies things for SkPngCodec. SkPngCodec: - Implement new APIs - Switch from sk_read_fn/png_read_row etc to png_process_data - Expand AutoCleanPng's role to decode the header and create the SkPngCodec - Make the interlaced PNG decoder report how many lines were initialized during an incomplete decode SkIcoCodec: - Implement the new APIs; supported for PNG in ICO SkSampledCodec: - Call the new method for decoding scanlines, and fall back to the old method if the new version is unimplemented - Remove references to kNone_SkScanlineOrder tests/CodecPartial: - Add a test which decodes part of an image, then finishes the decode, and compares it to the straightforward method tests/CodecTest: - Add a test which decodes all scanlines using the new method - Repurpose the Codec_stripes test to decode using the new method in sections rather than all at once - In the method check(), add a parameter for whether the image supports the new method of scanline decoding, and be explicit about whether an image supports incomplete - Test incomplete PNG decodes. We should have been doing it anyway for non-interlaced (except for an image that is too small - one row), but the new method supports interlaced incomplete as well - Make test_invalid_parameters test the new method - Add a test to ensure that it's safe to fall back to scanline decoding without rewinding BUG=skia:4211 The new version was generally faster than the old version (but not significantly so). Some raw performance differences can be found at https://docs.google.com/a/google.com/spreadsheets/d/1Gis3aRCEa72qBNDRMgGDg3jD-pMgO-FXldlNF9ejo4o/ Design doc can be found at https://docs.google.com/a/google.com/document/d/11Mn8-ePDKwVEMCjs3nWwSjxcSpJ_Cu8DF57KNtUmgLM/ GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1997703003 Review-Url: https://codereview.chromium.org/1997703003
2016-09-16 15:20:38 +00:00
virtual Result onStartIncrementalDecode(const SkImageInfo& /*dstInfo*/, void*, size_t,
const SkCodec::Options&, SkPMColor*, int*) {
return kUnimplemented;
}
virtual Result onIncrementalDecode(int*) {
return kUnimplemented;
}
virtual bool onSkipScanlines(int /*countLines*/) { return false; }
Merge SkCodec with SkScanlineDecoder Benefits: - This mimics other decoding APIs (including the ones SkCodec relies on, e.g. a png_struct, which can be used to decode an entire image or one line at a time). - It allows a client to ask us to do what we can do efficiently - i.e. start from encoded data and either decode the whole thing or scanlines. - It removes the duplicate methods which appeared in both SkCodec and SkScanlineDecoder (some of which, e.g. in SkJpegScanlineDecoder, just call fCodec->sameMethod()). - It simplifies moving more checks into the base class (e.g. the examples in skbug.com/4284). BUG=skia:4175 BUG=skia:4284 ===================================================================== SkScanlineDecoder.h/.cpp: Removed. SkCodec.h/.cpp: Add methods, enums, and variables which were previously in SkScanlineDecoder. Default fCurrScanline to -1, as a sentinel that start has not been called. General changes: Convert SkScanlineDecoders to SkCodecs. General changes in SkCodec subclasses: Merge SkScanlineDecoder implementation into SkCodec. Most (all?) owned an SkCodec, so they now call this-> instead of fCodec->. SkBmpCodec.h/.cpp: Replace the unused rowOrder method with an override for onGetScanlineOrder. Make getDstRow const, since it is called by onGetY, which is const. SkCodec_libpng.h/.cpp: Make SkPngCodec an abstract class, with two subclasses which handle scanline decoding separately (they share code for decoding the entire image). Reimplement onReallyHasAlpha so that it can return the most recent result (e.g. after a scanline decode which only decoded part of the image) or a better answer (e.g. if the whole image is known to be opaque). Compute fNumberPasses early, so we know which subclass to instantiate. Make SkPngInterlaceScanlineDecoder use the base class' fCurrScanline rather than a separate variable. CodexTest.cpp: Add tests for the state changes in SkCodec (need to call start before decoding scanlines; calling getPixels means that start will need to be called again before decoding more scanlines). Add a test which decodes in stripes, currently only used for an interlaced PNG. TODO: Add tests for onReallyHasAlpha. Review URL: https://codereview.chromium.org/1365313002
2015-09-30 15:57:13 +00:00
virtual int onGetScanlines(void* /*dst*/, int /*countLines*/, size_t /*rowBytes*/) { return 0; }
/**
* On an incomplete decode, getPixels() and getScanlines() will call this function
* to fill any uinitialized memory.
*
* @param dstInfo Contains the destination color type
* Contains the destination alpha type
* Contains the destination width
* The height stored in this info is unused
* @param dst Pointer to the start of destination pixel memory
* @param rowBytes Stride length in destination pixel memory
* @param zeroInit Indicates if memory is zero initialized
* @param linesRequested Number of lines that the client requested
* @param linesDecoded Number of lines that were successfully decoded
*/
void fillIncompleteImage(const SkImageInfo& dstInfo, void* dst, size_t rowBytes,
ZeroInitialized zeroInit, int linesRequested, int linesDecoded);
Merge SkCodec with SkScanlineDecoder Benefits: - This mimics other decoding APIs (including the ones SkCodec relies on, e.g. a png_struct, which can be used to decode an entire image or one line at a time). - It allows a client to ask us to do what we can do efficiently - i.e. start from encoded data and either decode the whole thing or scanlines. - It removes the duplicate methods which appeared in both SkCodec and SkScanlineDecoder (some of which, e.g. in SkJpegScanlineDecoder, just call fCodec->sameMethod()). - It simplifies moving more checks into the base class (e.g. the examples in skbug.com/4284). BUG=skia:4175 BUG=skia:4284 ===================================================================== SkScanlineDecoder.h/.cpp: Removed. SkCodec.h/.cpp: Add methods, enums, and variables which were previously in SkScanlineDecoder. Default fCurrScanline to -1, as a sentinel that start has not been called. General changes: Convert SkScanlineDecoders to SkCodecs. General changes in SkCodec subclasses: Merge SkScanlineDecoder implementation into SkCodec. Most (all?) owned an SkCodec, so they now call this-> instead of fCodec->. SkBmpCodec.h/.cpp: Replace the unused rowOrder method with an override for onGetScanlineOrder. Make getDstRow const, since it is called by onGetY, which is const. SkCodec_libpng.h/.cpp: Make SkPngCodec an abstract class, with two subclasses which handle scanline decoding separately (they share code for decoding the entire image). Reimplement onReallyHasAlpha so that it can return the most recent result (e.g. after a scanline decode which only decoded part of the image) or a better answer (e.g. if the whole image is known to be opaque). Compute fNumberPasses early, so we know which subclass to instantiate. Make SkPngInterlaceScanlineDecoder use the base class' fCurrScanline rather than a separate variable. CodexTest.cpp: Add tests for the state changes in SkCodec (need to call start before decoding scanlines; calling getPixels means that start will need to be called again before decoding more scanlines). Add a test which decodes in stripes, currently only used for an interlaced PNG. TODO: Add tests for onReallyHasAlpha. Review URL: https://codereview.chromium.org/1365313002
2015-09-30 15:57:13 +00:00
/**
* Return an object which will allow forcing scanline decodes to sample in X.
*
* May create a sampler, if one is not currently being used. Otherwise, does
* not affect ownership.
*
Add support for multiple frames in SkCodec Add an interface to decode frames beyond the first in SkCodec, and add an implementation for SkGifCodec. Add getFrameData to SkCodec. This method reads ahead in the stream to return a vector containing meta data about each frame in the image. This is not required in order to decode frames beyond the first, but it allows a client to learn extra information: - how long the frame should be displayed - whether a frame should be blended with a prior frame, allowing the client to provide the prior frame to speed up decoding Add a new fields to SkCodec::Options: - fFrameIndex - fHasPriorFrame The API is designed so that SkCodec never caches frames. If a client wants a frame beyond the first, they specify the frame in Options.fFrameIndex. If the client does not have the frame's required frame (the frame that this frame must be blended on top of) cached, they pass false for Options.fHasPriorFrame. Unless the frame is independent, the codec will then recursively decode all frames necessary to decode fFrameIndex. If the client has the required frame cached, they can put it in the dst they pass to the codec, and the codec will only draw fFrameIndex onto it. Replace SkGifCodec's scanline decoding support with progressive decoding, and update the tests accordingly. Implement new APIs in SkGifCodec. Instead of using gif_lib, use GIFImageReader, imported from Chromium (along with its copyright headers) with the following changes: - SkGifCodec is now the client - Replace blink types - Combine GIFColorMap::buildTable and ::getTable into a method that creates and returns an SkColorTable - Input comes from an SkStream, instead of a SegmentReader. Add SkStreamBuffer, which buffers the (potentially partial) stream in order to decode progressively. (FIXME: This requires copying data that previously was read directly from the SegmentReader. Does this hurt performance? If so, can we fix it?) - Remove UMA code - Instead of reporting screen width and height to the client, allow the client to query for it - Fail earlier if the first frame AND screen have size of zero - Compute required previous frame when adding a new one - Move GIFParseQuery from GIFImageDecoder to GIFImageReader - Allow parsing up to a specific frame (to skip parsing the rest of the stream if a client only wants the first frame) - Compute whether the first frame has alpha and supports index 8, to create the SkImageInfo. This happens before reporting that the size has been decoded. Add GIFImageDecoder::haveDecodedRow to SkGifCodec, imported from Chromium (along with its copyright header), with the following changes: - Add support for sampling - Use the swizzler - Keep track of the rows decoded - Do *not* keep track of whether we've seen alpha Remove SkCodec::kOutOfOrder_SkScanlineOrder, which was only used by GIF scanline decoding. Call onRewind even if there is no stream (SkGifCodec needs to clear its decoded state so it will decode from the beginning). Add a method to SkSwizzler to access the offset into the dst, taking subsetting into account. Add a GM that animates a GIF. Add tests for the new APIs. *** Behavior changes: * Previously, we reported that an image with a subset frame and no transparent index was opaque and used the background index (if present) to fill the background. This is necessary in order to support index 8, but it does not match viewers/browsers I have seen. Examples: - Chromium and Gimp render the background transparent - Firefox, Safari, Linux Image Viewer, Safari Preview clip to the frame (for a single frame image) This CL matches Chromium's behavior and renders the background transparent. This allows us to have consistent behavior across products and simplifies the code (relative to what we would have to do to continue the old behavior on Android). It also means that we will no longer support index 8 for some GIFs. * Stop checking for GIFSTAMP - all GIFs should be either 89a or 87a. This matches Chromium. I suspect that bugs would have been reported if valid GIFs started with "GIFVER" instead of "GIF89a" or "GIF87a" (but did not decode in Chromium). *** Future work not included in this CL: * Move some checks out of haveDecodedRow, since they are the same for the entire frame e.g. - intersecting the frameRect with the full image size - whether there is a color table * Change when we write transparent pixels - In some cases, Chromium deemed this unnecessary, but I suspect it is slower than the fallback case. There will continue to be cases where we should *not* write them, but for e.g. the first pass where we have already cleared to transparent (which we may also be able to skip) writing the transparent pixels will not make anything incorrect. * Report color type and alpha type per frame - Depending on alpha values, disposal methods, frame rects, etc, subsequent frames may have different properties than the first. * Skip copies of the encoded data - We copy the encoded data in case the stream is one that cannot be rewound, so we can parse and then decode (possibly not immediately). For some input streams, this is unnecessary. - I was concerned this cause a performance regression, but on average the new code is faster than the old for the images I tested [1]. - It may cause a performance regression for Chromium, though, where we can always move back in the stream, so this should be addressed. Design doc: https://docs.google.com/a/google.com/document/d/12Qhf9T92MWfdWujQwCIjhCO3sw6pTJB5pJBwDM1T7Kc/ [1] https://docs.google.com/a/google.com/spreadsheets/d/19V-t9BfbFw5eiwBTKA1qOBkZbchjlTC5EIz6HFy-6RI/ GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=2045293002 Review-Url: https://codereview.chromium.org/2045293002
2016-10-24 16:03:26 +00:00
* Only valid during scanline decoding or incremental decoding.
*/
virtual SkSampler* getSampler(bool /*createIfNecessary*/) { return nullptr; }
Make SkPngCodec decode progressively. This is a step towards using SkCodec in Chromium, where progressive decoding is necessary. Switch from using png_read_row (which expects all the data to be available) to png_process_data, which uses callbacks when rows are available. Create a new API for SkCodec, which supports progressive decoding and scanline decoding. Future changes will switch the other clients off of startScanlineDecode and get/skip-Scanlines to the new API. Remove SkCodec::kNone_ScanlineOrder, which was only used for interlaced PNG images. In the new API, interlaced PNG fits kTopDown. Also remove updateCurrScanline(), which was only used by the old implementation for interlaced PNG. DMSrcSink: - In CodecSrc::kScanline_Mode, use the new method for scanline decoding for the supported formats (just PNG and PNG-in-ICO for now). fuzz.cpp: - Remove reference to kNone_ScanlineOrder SkCodec: - Add new APIs: - startIncrementalDecode - incrementalDecode - Remove kNone_SkScanlineOrder and updateCurrScanline() - Set fDstInfo and fOptions in getPixels(). This may not be necessary for all implementations, but it simplifies things for SkPngCodec. SkPngCodec: - Implement new APIs - Switch from sk_read_fn/png_read_row etc to png_process_data - Expand AutoCleanPng's role to decode the header and create the SkPngCodec - Make the interlaced PNG decoder report how many lines were initialized during an incomplete decode SkIcoCodec: - Implement the new APIs; supported for PNG in ICO SkSampledCodec: - Call the new method for decoding scanlines, and fall back to the old method if the new version is unimplemented - Remove references to kNone_SkScanlineOrder tests/CodecPartial: - Add a test which decodes part of an image, then finishes the decode, and compares it to the straightforward method tests/CodecTest: - Add a test which decodes all scanlines using the new method - Repurpose the Codec_stripes test to decode using the new method in sections rather than all at once - In the method check(), add a parameter for whether the image supports the new method of scanline decoding, and be explicit about whether an image supports incomplete - Test incomplete PNG decodes. We should have been doing it anyway for non-interlaced (except for an image that is too small - one row), but the new method supports interlaced incomplete as well - Make test_invalid_parameters test the new method - Add a test to ensure that it's safe to fall back to scanline decoding without rewinding BUG=skia:4211 The new version was generally faster than the old version (but not significantly so). Some raw performance differences can be found at https://docs.google.com/a/google.com/spreadsheets/d/1Gis3aRCEa72qBNDRMgGDg3jD-pMgO-FXldlNF9ejo4o/ Design doc can be found at https://docs.google.com/a/google.com/document/d/11Mn8-ePDKwVEMCjs3nWwSjxcSpJ_Cu8DF57KNtUmgLM/ GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1997703003 Review-Url: https://codereview.chromium.org/1997703003
2016-09-16 15:20:38 +00:00
friend class DM::CodecSrc; // for fillIncompleteImage
friend class SkSampledCodec;
friend class SkIcoCodec;
friend struct Sniffer; // for fUnsupportedICC
friend class AutoCleanPng; // for setUnsupportedICC()
};
#endif // SkCodec_DEFINED