mirror/kimageformats
1
0
Fork 0
mirror of https://invent.kde.org/frameworks/kimageformats.git synced 2025-06-03 17:08:08 -04:00
Code Issues Packages Projects Releases Wiki Activity
kimageformats/src/imageformats/xcf.cpp
Mirco Miranda fbf60f8bbb xcf: fix possible overflow 
Port to KF6 of MR !187
2023-09-05 11:32:32 +00:00

4307 lines
148 KiB
C++
Raw Blame History

/*
xcf.cpp: A Qt 5 plug-in for reading GIMP XCF image files
SPDX-FileCopyrightText: 2001 lignum Computing Inc. <allen@lignumcomputing.com>
SPDX-FileCopyrightText: 2004 Melchior FRANZ <mfranz@kde.org>
SPDX-License-Identifier: LGPL-2.1-or-later
*/
#include "util_p.h"
#include "xcf_p.h"
#include <QColorSpace>
#include <QDebug>
#include <QIODevice>
#include <QImage>
#include <QImageReader>
#include <QLoggingCategory>
#include <QPainter>
#include <QStack>
#include <QVector>
#include <QtEndian>
#if QT_VERSION < QT_VERSION_CHECK(6, 0, 0)
// To test the plugin with OSS FUZZ (in 2023) you need to make it compatible with Qt 5.
// The plugin has been made compatible with all Qt (5 and 6) versions supported by the KDE project.
#define XCF_QT5_SUPPORT
#endif
#ifndef XCF_QT5_SUPPORT
// Float images are not supported by Qt 5 and can be disabled in QT 6 to reduce memory usage.
// Unfortunately enabling/disabling this define results in slightly different images, so leave the default if possible.
#define USE_FLOAT_IMAGES // default uncommented
// Let's set a "reasonable" maximum size
#define MAX_IMAGE_WIDTH 300000
#define MAX_IMAGE_HEIGHT 300000
#else
// While it is possible to have images larger than 32767 pixels, QPainter seems unable to go beyond this threshold using Qt 5.
#define MAX_IMAGE_WIDTH 32767
#define MAX_IMAGE_HEIGHT 32767
#endif
#ifdef USE_FLOAT_IMAGES
#include <qrgbafloat.h>
#endif
#include <stdlib.h>
#include <string.h>
#include "gimp_p.h"
Q_DECLARE_LOGGING_CATEGORY(XCFPLUGIN)
Q_LOGGING_CATEGORY(XCFPLUGIN, "kf.imageformats.plugins.xcf", QtWarningMsg)
//#define DISABLE_TILE_PROFILE // default commented (comment to use the conversion as intended by Martin)
#define DISABLE_IMAGE_PROFILE // default uncommented (comment to use the conversion as intended by Martin)
#define DISABLE_TILE_PROFILE_CONV // default uncommented (comment to use the conversion as intended by Martin)
#define DISABLE_IMAGE_PROFILE_CONV // default uncommented (comment to use the conversion as intended by Martin)
const float INCHESPERMETER = (100.0f / 2.54f);
namespace
{
struct RandomTable {
// From glibc
static constexpr int rand_r(unsigned int *seed)
{
unsigned int next = *seed;
int result = 0;
next *= 1103515245;
next += 12345;
result = (unsigned int)(next / 65536) % 2048;
next *= 1103515245;
next += 12345;
result <<= 10;
result ^= (unsigned int)(next / 65536) % 1024;
next *= 1103515245;
next += 12345;
result <<= 10;
result ^= (unsigned int)(next / 65536) % 1024;
*seed = next;
return result;
}
constexpr RandomTable()
: values{}
{
unsigned int next = RANDOM_SEED;
for (int i = 0; i < RANDOM_TABLE_SIZE; i++) {
values[i] = rand_r(&next);
}
for (int i = 0; i < RANDOM_TABLE_SIZE; i++) {
int tmp{};
int swap = i + rand_r(&next) % (RANDOM_TABLE_SIZE - i);
tmp = values[i];
values[i] = values[swap];
values[swap] = tmp;
}
}
int values[RANDOM_TABLE_SIZE]{};
};
} // namespace {
/*!
* Each layer in an XCF file is stored as a matrix of
* 64-pixel by 64-pixel images. The GIMP has a sophisticated
* method of handling very large images as well as implementing
* parallel processing on a tile-by-tile basis. Here, though,
* we just read them in en-masse and store them in a matrix.
*/
typedef QVector<QVector<QImage>> Tiles;
class XCFImageFormat
{
Q_GADGET
public:
//! Properties which can be stored in an XCF file.
enum PropType {
PROP_END = 0,
PROP_COLORMAP = 1,
PROP_ACTIVE_LAYER = 2,
PROP_ACTIVE_CHANNEL = 3,
PROP_SELECTION = 4,
PROP_FLOATING_SELECTION = 5,
PROP_OPACITY = 6,
PROP_MODE = 7,
PROP_VISIBLE = 8,
PROP_LINKED = 9,
PROP_LOCK_ALPHA = 10,
PROP_APPLY_MASK = 11,
PROP_EDIT_MASK = 12,
PROP_SHOW_MASK = 13,
PROP_SHOW_MASKED = 14,
PROP_OFFSETS = 15,
PROP_COLOR = 16,
PROP_COMPRESSION = 17,
PROP_GUIDES = 18,
PROP_RESOLUTION = 19,
PROP_TATTOO = 20,
PROP_PARASITES = 21,
PROP_UNIT = 22,
PROP_PATHS = 23,
PROP_USER_UNIT = 24,
PROP_VECTORS = 25,
PROP_TEXT_LAYER_FLAGS = 26,
PROP_OLD_SAMPLE_POINTS = 27,
PROP_LOCK_CONTENT = 28,
PROP_GROUP_ITEM = 29,
PROP_ITEM_PATH = 30,
PROP_GROUP_ITEM_FLAGS = 31,
PROP_LOCK_POSITION = 32,
PROP_FLOAT_OPACITY = 33,
PROP_COLOR_TAG = 34,
PROP_COMPOSITE_MODE = 35,
PROP_COMPOSITE_SPACE = 36,
PROP_BLEND_SPACE = 37,
PROP_FLOAT_COLOR = 38,
PROP_SAMPLE_POINTS = 39,
MAX_SUPPORTED_PROPTYPE, // should always be at the end so its value is last + 1
};
Q_ENUM(PropType)
//! Compression type used in layer tiles.
enum XcfCompressionType : qint8 {
COMPRESS_INVALID = -1, /* our own */
COMPRESS_NONE = 0,
COMPRESS_RLE = 1,
COMPRESS_ZLIB = 2, /* unused */
COMPRESS_FRACTAL = 3, /* unused */
};
Q_ENUM(XcfCompressionType)
enum LayerModeType : quint32 {
GIMP_LAYER_MODE_NORMAL_LEGACY,
GIMP_LAYER_MODE_DISSOLVE,
GIMP_LAYER_MODE_BEHIND_LEGACY,
GIMP_LAYER_MODE_MULTIPLY_LEGACY,
GIMP_LAYER_MODE_SCREEN_LEGACY,
GIMP_LAYER_MODE_OVERLAY_LEGACY,
GIMP_LAYER_MODE_DIFFERENCE_LEGACY,
GIMP_LAYER_MODE_ADDITION_LEGACY,
GIMP_LAYER_MODE_SUBTRACT_LEGACY,
GIMP_LAYER_MODE_DARKEN_ONLY_LEGACY,
GIMP_LAYER_MODE_LIGHTEN_ONLY_LEGACY,
GIMP_LAYER_MODE_HSV_HUE_LEGACY,
GIMP_LAYER_MODE_HSV_SATURATION_LEGACY,
GIMP_LAYER_MODE_HSL_COLOR_LEGACY,
GIMP_LAYER_MODE_HSV_VALUE_LEGACY,
GIMP_LAYER_MODE_DIVIDE_LEGACY,
GIMP_LAYER_MODE_DODGE_LEGACY,
GIMP_LAYER_MODE_BURN_LEGACY,
GIMP_LAYER_MODE_HARDLIGHT_LEGACY,
GIMP_LAYER_MODE_SOFTLIGHT_LEGACY,
GIMP_LAYER_MODE_GRAIN_EXTRACT_LEGACY,
GIMP_LAYER_MODE_GRAIN_MERGE_LEGACY,
GIMP_LAYER_MODE_COLOR_ERASE_LEGACY,
GIMP_LAYER_MODE_OVERLAY,
GIMP_LAYER_MODE_LCH_HUE,
GIMP_LAYER_MODE_LCH_CHROMA,
GIMP_LAYER_MODE_LCH_COLOR,
GIMP_LAYER_MODE_LCH_LIGHTNESS,
GIMP_LAYER_MODE_NORMAL,
GIMP_LAYER_MODE_BEHIND,
GIMP_LAYER_MODE_MULTIPLY,
GIMP_LAYER_MODE_SCREEN,
GIMP_LAYER_MODE_DIFFERENCE,
GIMP_LAYER_MODE_ADDITION,
GIMP_LAYER_MODE_SUBTRACT,
GIMP_LAYER_MODE_DARKEN_ONLY,
GIMP_LAYER_MODE_LIGHTEN_ONLY,
GIMP_LAYER_MODE_HSV_HUE,
GIMP_LAYER_MODE_HSV_SATURATION,
GIMP_LAYER_MODE_HSL_COLOR,
GIMP_LAYER_MODE_HSV_VALUE,
GIMP_LAYER_MODE_DIVIDE,
GIMP_LAYER_MODE_DODGE,
GIMP_LAYER_MODE_BURN,
GIMP_LAYER_MODE_HARDLIGHT,
GIMP_LAYER_MODE_SOFTLIGHT,
GIMP_LAYER_MODE_GRAIN_EXTRACT,
GIMP_LAYER_MODE_GRAIN_MERGE,
GIMP_LAYER_MODE_VIVID_LIGHT,
GIMP_LAYER_MODE_PIN_LIGHT,
GIMP_LAYER_MODE_LINEAR_LIGHT,
GIMP_LAYER_MODE_HARD_MIX,
GIMP_LAYER_MODE_EXCLUSION,
GIMP_LAYER_MODE_LINEAR_BURN,
GIMP_LAYER_MODE_LUMA_DARKEN_ONLY,
GIMP_LAYER_MODE_LUMA_LIGHTEN_ONLY,
GIMP_LAYER_MODE_LUMINANCE,
GIMP_LAYER_MODE_COLOR_ERASE,
GIMP_LAYER_MODE_ERASE,
GIMP_LAYER_MODE_MERGE,
GIMP_LAYER_MODE_SPLIT,
GIMP_LAYER_MODE_PASS_THROUGH,
GIMP_LAYER_MODE_COUNT,
};
Q_ENUM(LayerModeType)
//! Type of individual layers in an XCF file.
enum GimpImageType : qint32 {
RGB_GIMAGE,
RGBA_GIMAGE,
GRAY_GIMAGE,
GRAYA_GIMAGE,
INDEXED_GIMAGE,
INDEXEDA_GIMAGE,
};
Q_ENUM(GimpImageType)
//! Type of individual layers in an XCF file.
enum GimpColorSpace : qint32 {
AutoColorSpace,
RgbLinearSpace,
RgbPerceptualSpace,
LabSpace,
};
Q_ENUM(GimpColorSpace);
//! Mode to use when compositing layer
enum GimpCompositeMode : qint32 {
CompositeAuto,
CompositeUnion,
CompositeClipBackdrop,
CompositeClipLayer,
CompositeIntersect,
};
Q_ENUM(GimpCompositeMode);
enum GimpPrecision : qint32 {
GIMP_PRECISION_U8_LINEAR = 100, /*< desc="8-bit linear integer" >*/
GIMP_PRECISION_U8_NON_LINEAR = 150, /*< desc="8-bit non-linear integer" >*/
GIMP_PRECISION_U8_PERCEPTUAL = 175, /*< desc="8-bit perceptual integer" >*/
GIMP_PRECISION_U16_LINEAR = 200, /*< desc="16-bit linear integer" >*/
GIMP_PRECISION_U16_NON_LINEAR = 250, /*< desc="16-bit non-linear integer" >*/
GIMP_PRECISION_U16_PERCEPTUAL = 275, /*< desc="16-bit perceptual integer" >*/
GIMP_PRECISION_U32_LINEAR = 300, /*< desc="32-bit linear integer" >*/
GIMP_PRECISION_U32_NON_LINEAR = 350, /*< desc="32-bit non-linear integer" >*/
GIMP_PRECISION_U32_PERCEPTUAL = 375, /*< desc="32-bit perceptual integer" >*/
GIMP_PRECISION_HALF_LINEAR = 500, /*< desc="16-bit linear floating point" >*/
GIMP_PRECISION_HALF_NON_LINEAR = 550, /*< desc="16-bit non-linear floating point" >*/
GIMP_PRECISION_HALF_PERCEPTUAL = 575, /*< desc="16-bit perceptual floating point" >*/
GIMP_PRECISION_FLOAT_LINEAR = 600, /*< desc="32-bit linear floating point" >*/
GIMP_PRECISION_FLOAT_NON_LINEAR = 650, /*< desc="32-bit non-linear floating point" >*/
GIMP_PRECISION_FLOAT_PERCEPTUAL = 675, /*< desc="32-bit perceptual floating point" >*/
GIMP_PRECISION_DOUBLE_LINEAR = 700, /*< desc="64-bit linear floating point" >*/
GIMP_PRECISION_DOUBLE_NON_LINEAR = 750, /*< desc="64-bit non-linear floating point" >*/
GIMP_PRECISION_DOUBLE_PERCEPTUAL = 775, /*< desc="64-bit perceptual floating point" >*/
};
Q_ENUM(GimpPrecision);
XCFImageFormat();
bool readXCF(QIODevice *device, QImage *image);
/*!
* Each GIMP image is composed of one or more layers. A layer can
* be one of any three basic types: RGB, grayscale or indexed. With an
* optional alpha channel, there are six possible types altogether.
*
* Note: there is only ever one instance of this structure. The
* layer info is discarded after it is merged into the final QImage.
*/
class Layer
{
public:
quint32 width; //!< Width of the layer
quint32 height; //!< Height of the layer
GimpImageType type; //!< Type of the layer (GimpImageType)
char *name; //!< Name of the layer
qint64 hierarchy_offset; //!< File position of Tile hierarchy
qint64 mask_offset; //!< File position of mask image
uint nrows; //!< Number of rows of tiles (y direction)
uint ncols; //!< Number of columns of tiles (x direction)
Tiles image_tiles; //!< The basic image
//! For Grayscale and Indexed images, the alpha channel is stored
//! separately (in this data structure, anyway).
Tiles alpha_tiles;
Tiles mask_tiles; //!< The layer mask (optional)
//! Additional information about a layer mask.
struct {
quint32 opacity;
float opacityFloat = 1.f;
quint32 visible;
quint32 show_masked;
uchar red, green, blue;
float redF, greenF, blueF; // Floats should override
quint32 tattoo;
} mask_channel;
XcfCompressionType compression = COMPRESS_INVALID; //!< tile compression method (CompressionType)
bool active; //!< Is this layer the active layer?
quint32 opacity = 255; //!< The opacity of the layer
float opacityFloat = 1.f; //!< The opacity of the layer, but floating point (both are set)
quint32 visible = 1; //!< Is the layer visible?
quint32 linked; //!< Is this layer linked (geometrically)
quint32 preserve_transparency; //!< Preserve alpha when drawing on layer?
quint32 apply_mask = 9; //!< Apply the layer mask? Use 9 as "uninitilized". Spec says "If the property does not appear for a layer which has a layer
//!< mask, it defaults to true (1).
// Robust readers should force this to false if the layer has no layer mask.
quint32 edit_mask; //!< Is the layer mask the being edited?
quint32 show_mask; //!< Show the layer mask rather than the image?
qint32 x_offset = 0; //!< x offset of the layer relative to the image
qint32 y_offset = 0; //!< y offset of the layer relative to the image
LayerModeType mode = GIMP_LAYER_MODE_NORMAL_LEGACY; //!< Combining mode of layer (LayerModeEffects)
quint32 tattoo; //!< (unique identifier?)
GimpColorSpace blendSpace = RgbLinearSpace; //!< What colorspace to use when blending
GimpColorSpace compositeSpace = RgbLinearSpace; //!< What colorspace to use when compositing
GimpCompositeMode compositeMode = CompositeUnion; //!< How to composite layer (union, clip, etc.)
//! As each tile is read from the file, it is buffered here.
#ifdef USE_FLOAT_IMAGES
uchar tile[quint64(TILE_WIDTH * TILE_HEIGHT * sizeof(QRgbaFloat32) * 1.5)];
#else
uchar tile[quint64(TILE_WIDTH * TILE_HEIGHT * sizeof(QRgba64) * 1.5)];
#endif
//! The data from tile buffer is copied to the Tile by this
//! method. Depending on the type of the tile (RGB, Grayscale,
//! Indexed) and use (image or mask), the bytes in the buffer are
//! copied in different ways.
bool (*assignBytes)(Layer &layer, uint i, uint j, const GimpPrecision &precision);
Layer(void)
: name(nullptr)
{
}
~Layer(void)
{
delete[] name;
}
Layer(const Layer &) = delete;
Layer &operator=(const Layer &) = delete;
QImage::Format qimageFormat(const GimpPrecision precision, uint num_colors = 0, bool legacyMode = false) const
{
int bpc = bytesPerChannel(precision);
#ifdef USE_FLOAT_IMAGES
bool float16 = !legacyMode && precision >= GIMP_PRECISION_HALF_LINEAR && precision <= GIMP_PRECISION_HALF_PERCEPTUAL;
bool float32 = !legacyMode && precision >= GIMP_PRECISION_FLOAT_LINEAR && precision <= GIMP_PRECISION_FLOAT_PERCEPTUAL;
#endif
if (legacyMode) {
bpc = std::min(bpc, 1);
}
switch (type) {
case RGB_GIMAGE:
if (opacity == OPAQUE_OPACITY) {
#ifdef USE_FLOAT_IMAGES
if (float16) {
return QImage::Format_RGBX16FPx4;
}
if (float32) {
return QImage::QImage::Format_RGBX32FPx4;
}
#endif
if (bpc == 1) {
return QImage::Format_RGBX8888;
} else if (bpc == 2 || bpc == 4) {
return QImage::Format_RGBX64;
} else {
qCDebug(XCFPLUGIN) << "Layer has invalid bpc" << bpc << precision;
return QImage::Format_Invalid;
}
}
Q_FALLTHROUGH();
case RGBA_GIMAGE:
#ifdef USE_FLOAT_IMAGES
if (float16) {
return QImage::Format_RGBA16FPx4;
}
if (float32) {
return QImage::QImage::Format_RGBA32FPx4;
}
#endif
if (bpc == 1) {
return QImage::Format_RGBA8888;
} else if (bpc == 2 || bpc == 4) {
return QImage::Format_RGBA64;
} else {
qCDebug(XCFPLUGIN) << "Layer has invalid bpc" << bpc;
return QImage::Format_Invalid;
}
break;
case GRAY_GIMAGE:
if (opacity == OPAQUE_OPACITY) {
return QImage::Format_Indexed8;
} // else, fall through to 32-bit representation
Q_FALLTHROUGH();
case GRAYA_GIMAGE:
return QImage::Format_RGBA8888;
break;
case INDEXED_GIMAGE:
// As noted in the table above, there are quite a few combinations
// which are possible with indexed images, depending on the
// presence of transparency (note: not translucency, which is not
// supported by The GIMP for indexed images) and the number of
// individual colors.
// Note: Qt treats a bitmap with a Black and White color palette
// as a mask, so only the "on" bits are drawn, regardless of the
// order color table entries. Otherwise (i.e., at least one of the
// color table entries is not black or white), it obeys the one-
// or two-color palette. Have to ask about this...
if (num_colors == 1 || num_colors == 2) {
return QImage::Format_MonoLSB;
} else {
return QImage::Format_Indexed8;
}
break;
case INDEXEDA_GIMAGE:
if (num_colors == 1) {
return QImage::Format_MonoLSB;
} else {
return QImage::Format_Indexed8;
}
}
qCWarning(XCFPLUGIN) << "Unhandled layer mode" << XCFImageFormat::LayerModeType(type);
return QImage::Format_Invalid;
}
};
/*!
* The in-memory representation of the XCF Image. It contains a few
* metadata items, but is mostly a container for the layer information.
*/
class XCFImage
{
public:
struct Header {
GimpPrecision precision = GIMP_PRECISION_U8_LINEAR; //!< Default precision (GimpPrecision)
quint32 width; //!< width of the XCF image
quint32 height; //!< height of the XCF image
qint32 type; //!< type of the XCF image (GimpImageBaseType)
} header;
XcfCompressionType compression = COMPRESS_RLE; //!< tile compression method (CompressionType)
float x_resolution = -1; //!< x resolution in dots per inch
float y_resolution = -1; //!< y resolution in dots per inch
qint32 tattoo; //!< (unique identifier?)
quint32 unit; //!< Units of The GIMP (inch, mm, pica, etc...)
qint32 num_colors = 0; //!< number of colors in an indexed image
QVector<QRgb> palette; //!< indexed image color palette
int num_layers; //!< number of layers
Layer layer; //!< most recently read layer
bool initialized; //!< Is the QImage initialized?
QImage image; //!< final QImage
QHash<QString,QByteArray> parasites; //!< parasites data
XCFImage(void)
: initialized(false)
{
}
QImage::Format qimageFormat() const
{
return layer.qimageFormat(header.precision, num_colors, true);
}
uint bytesPerChannel() const
{
return XCFImageFormat::bytesPerChannel(header.precision);
}
};
private:
static qint64 readOffsetPtr(QDataStream &stream)
{
if (stream.version() >= 11) {
qint64 ret;
stream >> ret;
return ret;
} else {
quint32 ret;
stream >> ret;
return ret;
}
}
//! In layer DISSOLVE mode, a random number is chosen to compare to a
//! pixel's alpha. If the alpha is greater than the random number, the
//! pixel is drawn. This table merely contains the random number seeds
//! for each ROW of an image. Therefore, the random numbers chosen
//! are consistent from run to run.
static int random_table[RANDOM_TABLE_SIZE];
static bool random_table_initialized;
static constexpr RandomTable randomTable{};
//! This table is used as a shared grayscale ramp to be set on grayscale
//! images. This is because Qt does not differentiate between indexed and
//! grayscale images.
static QVector<QRgb> grayTable;
//! This table provides the add_pixel saturation values (i.e. 250 + 250 = 255).
// static int add_lut[256][256]; - this is so lame waste of 256k of memory
static int add_lut(int, int);
//! The bottom-most layer is copied into the final QImage by this
//! routine.
typedef void (*PixelCopyOperation)(const Layer &layer, uint i, uint j, int k, int l, QImage &image, int m, int n);
//! Higher layers are merged into the final QImage by this routine.
typedef bool (*PixelMergeOperation)(const Layer &layer, uint i, uint j, int k, int l, QImage &image, int m, int n);
static bool modeAffectsSourceAlpha(const quint32 type);
bool loadImageProperties(QDataStream &xcf_io, XCFImage &image);
bool loadProperty(QDataStream &xcf_io, PropType &type, QByteArray &bytes, quint32 &rawType);
bool loadLayer(QDataStream &xcf_io, XCFImage &xcf_image);
bool loadLayerProperties(QDataStream &xcf_io, Layer &layer);
bool composeTiles(XCFImage &xcf_image);
void setGrayPalette(QImage &image);
void setPalette(XCFImage &xcf_image, QImage &image);
void setImageParasites(const XCFImage &xcf_image, QImage &image);
static bool assignImageBytes(Layer &layer, uint i, uint j, const GimpPrecision &precision);
bool loadHierarchy(QDataStream &xcf_io, Layer &layer, const GimpPrecision precision);
bool loadLevel(QDataStream &xcf_io, Layer &layer, qint32 bpp, const GimpPrecision precision);
static bool assignMaskBytes(Layer &layer, uint i, uint j, const GimpPrecision &precision);
bool loadMask(QDataStream &xcf_io, Layer &layer, const GimpPrecision precision);
bool loadChannelProperties(QDataStream &xcf_io, Layer &layer);
bool initializeImage(XCFImage &xcf_image);
bool loadTileRLE(QDataStream &xcf_io, uchar *tile, int size, int data_length, qint32 bpp, qint64 *bytesParsed);
static void copyLayerToImage(XCFImage &xcf_image);
static void copyRGBToRGB(const Layer &layer, uint i, uint j, int k, int l, QImage &image, int m, int n);
static void copyGrayToGray(const Layer &layer, uint i, uint j, int k, int l, QImage &image, int m, int n);
static void copyGrayToRGB(const Layer &layer, uint i, uint j, int k, int l, QImage &image, int m, int n);
static void copyGrayAToRGB(const Layer &layer, uint i, uint j, int k, int l, QImage &image, int m, int n);
static void copyIndexedToIndexed(const Layer &layer, uint i, uint j, int k, int l, QImage &image, int m, int n);
static void copyIndexedAToIndexed(const Layer &layer, uint i, uint j, int k, int l, QImage &image, int m, int n);
static void copyIndexedAToRGB(const Layer &layer, uint i, uint j, int k, int l, QImage &image, int m, int n);
static void mergeLayerIntoImage(XCFImage &xcf_image);
static bool mergeRGBToRGB(const Layer &layer, uint i, uint j, int k, int l, QImage &image, int m, int n);
static bool mergeGrayToGray(const Layer &layer, uint i, uint j, int k, int l, QImage &image, int m, int n);
static bool mergeGrayAToGray(const Layer &layer, uint i, uint j, int k, int l, QImage &image, int m, int n);
static bool mergeGrayToRGB(const Layer &layer, uint i, uint j, int k, int l, QImage &image, int m, int n);
static bool mergeGrayAToRGB(const Layer &layer, uint i, uint j, int k, int l, QImage &image, int m, int n);
static bool mergeIndexedToIndexed(const Layer &layer, uint i, uint j, int k, int l, QImage &image, int m, int n);
static bool mergeIndexedAToIndexed(const Layer &layer, uint i, uint j, int k, int l, QImage &image, int m, int n);
static bool mergeIndexedAToRGB(const Layer &layer, uint i, uint j, int k, int l, QImage &image, int m, int n);
static void initializeRandomTable();
static void dissolveRGBPixels(QImage &image, int x, int y);
static void dissolveAlphaPixels(QImage &image, int x, int y);
static uint bytesPerChannel(const GimpPrecision precision)
{
switch (precision) {
case GIMP_PRECISION_U8_LINEAR:
case GIMP_PRECISION_U8_NON_LINEAR:
case GIMP_PRECISION_U8_PERCEPTUAL:
return 1;
break;
case GIMP_PRECISION_U16_LINEAR:
case GIMP_PRECISION_U16_NON_LINEAR:
case GIMP_PRECISION_U16_PERCEPTUAL:
case GIMP_PRECISION_HALF_LINEAR:
case GIMP_PRECISION_HALF_NON_LINEAR:
case GIMP_PRECISION_HALF_PERCEPTUAL:
return 2;
break;
case GIMP_PRECISION_U32_LINEAR:
case GIMP_PRECISION_U32_NON_LINEAR:
case GIMP_PRECISION_U32_PERCEPTUAL:
case GIMP_PRECISION_FLOAT_LINEAR:
case GIMP_PRECISION_FLOAT_NON_LINEAR:
case GIMP_PRECISION_FLOAT_PERCEPTUAL:
return 4;
break;
case GIMP_PRECISION_DOUBLE_LINEAR:
case GIMP_PRECISION_DOUBLE_NON_LINEAR:
case GIMP_PRECISION_DOUBLE_PERCEPTUAL:
return 8;
break;
default:
qCDebug(XCFPLUGIN) << "Layer has invalid precision" << precision;
return 0;
}
}
public:
static bool readXCFHeader(QDataStream &ds, XCFImage::Header *header);
};
int XCFImageFormat::random_table[RANDOM_TABLE_SIZE];
bool XCFImageFormat::random_table_initialized;
constexpr RandomTable XCFImageFormat::randomTable;
QVector<QRgb> XCFImageFormat::grayTable;
bool XCFImageFormat::modeAffectsSourceAlpha(const quint32 type)
{
switch (type) {
case GIMP_LAYER_MODE_NORMAL_LEGACY:
case GIMP_LAYER_MODE_DISSOLVE:
case GIMP_LAYER_MODE_BEHIND_LEGACY:
return true;
case GIMP_LAYER_MODE_MULTIPLY_LEGACY:
case GIMP_LAYER_MODE_SCREEN_LEGACY:
case GIMP_LAYER_MODE_OVERLAY_LEGACY:
case GIMP_LAYER_MODE_DIFFERENCE_LEGACY:
case GIMP_LAYER_MODE_ADDITION_LEGACY:
case GIMP_LAYER_MODE_SUBTRACT_LEGACY:
case GIMP_LAYER_MODE_DARKEN_ONLY_LEGACY:
case GIMP_LAYER_MODE_LIGHTEN_ONLY_LEGACY:
case GIMP_LAYER_MODE_HSV_HUE_LEGACY:
case GIMP_LAYER_MODE_HSV_SATURATION_LEGACY:
case GIMP_LAYER_MODE_HSL_COLOR_LEGACY:
case GIMP_LAYER_MODE_HSV_VALUE_LEGACY:
case GIMP_LAYER_MODE_DIVIDE_LEGACY:
case GIMP_LAYER_MODE_DODGE_LEGACY:
case GIMP_LAYER_MODE_BURN_LEGACY:
case GIMP_LAYER_MODE_HARDLIGHT_LEGACY:
case GIMP_LAYER_MODE_SOFTLIGHT_LEGACY:
case GIMP_LAYER_MODE_GRAIN_EXTRACT_LEGACY:
case GIMP_LAYER_MODE_GRAIN_MERGE_LEGACY:
return false;
case GIMP_LAYER_MODE_COLOR_ERASE_LEGACY:
case GIMP_LAYER_MODE_OVERLAY:
case GIMP_LAYER_MODE_LCH_HUE:
case GIMP_LAYER_MODE_LCH_CHROMA:
case GIMP_LAYER_MODE_LCH_COLOR:
case GIMP_LAYER_MODE_LCH_LIGHTNESS:
return false;
case GIMP_LAYER_MODE_NORMAL:
return true;
case GIMP_LAYER_MODE_BEHIND:
case GIMP_LAYER_MODE_MULTIPLY:
case GIMP_LAYER_MODE_SCREEN:
case GIMP_LAYER_MODE_DIFFERENCE:
case GIMP_LAYER_MODE_ADDITION:
case GIMP_LAYER_MODE_SUBTRACT:
case GIMP_LAYER_MODE_DARKEN_ONLY:
case GIMP_LAYER_MODE_LIGHTEN_ONLY:
case GIMP_LAYER_MODE_HSV_HUE:
case GIMP_LAYER_MODE_HSV_SATURATION:
case GIMP_LAYER_MODE_HSL_COLOR:
case GIMP_LAYER_MODE_HSV_VALUE:
case GIMP_LAYER_MODE_DIVIDE:
case GIMP_LAYER_MODE_DODGE:
case GIMP_LAYER_MODE_BURN:
case GIMP_LAYER_MODE_HARDLIGHT:
case GIMP_LAYER_MODE_SOFTLIGHT:
case GIMP_LAYER_MODE_GRAIN_EXTRACT:
case GIMP_LAYER_MODE_GRAIN_MERGE:
case GIMP_LAYER_MODE_VIVID_LIGHT:
case GIMP_LAYER_MODE_PIN_LIGHT:
case GIMP_LAYER_MODE_LINEAR_LIGHT:
case GIMP_LAYER_MODE_HARD_MIX:
case GIMP_LAYER_MODE_EXCLUSION:
case GIMP_LAYER_MODE_LINEAR_BURN:
case GIMP_LAYER_MODE_LUMA_DARKEN_ONLY:
case GIMP_LAYER_MODE_LUMA_LIGHTEN_ONLY:
case GIMP_LAYER_MODE_LUMINANCE:
case GIMP_LAYER_MODE_COLOR_ERASE:
case GIMP_LAYER_MODE_ERASE:
case GIMP_LAYER_MODE_MERGE:
case GIMP_LAYER_MODE_SPLIT:
case GIMP_LAYER_MODE_PASS_THROUGH:
return false;
default:
qCWarning(XCFPLUGIN) << "Unhandled layer mode" << XCFImageFormat::LayerModeType(type);
return false;
}
}
//! Change a QRgb value's alpha only.
inline QRgb qRgba(const QRgb rgb, int a)
{
return ((a & 0xff) << 24 | (rgb & RGB_MASK));
}
/*!
* The constructor for the XCF image loader.
*/
XCFImageFormat::XCFImageFormat()
{
static_assert(sizeof(QRgb) == 4, "the code assumes sizeof(QRgb) == 4, if that's not your case, help us fix it :)");
}
/*!
* This initializes the tables used in the layer dissolving routines.
*/
void XCFImageFormat::initializeRandomTable()
{
// From GIMP "paint_funcs.c" v1.2
srand(RANDOM_SEED);
for (int i = 0; i < RANDOM_TABLE_SIZE; i++) {
random_table[i] = rand();
}
for (int i = 0; i < RANDOM_TABLE_SIZE; i++) {
int tmp;
int swap = i + rand() % (RANDOM_TABLE_SIZE - i);
tmp = random_table[i];
random_table[i] = random_table[swap];
random_table[swap] = tmp;
}
}
inline int XCFImageFormat::add_lut(int a, int b)
{
return qMin(a + b, 255);
}
bool XCFImageFormat::readXCFHeader(QDataStream &xcf_io, XCFImage::Header *header)
{
QByteArray tag(14, '\0');
if (xcf_io.readRawData(tag.data(), tag.size()) != tag.size()) {
qCDebug(XCFPLUGIN) << "XCF: read failure on header tag";
return false;
}
if (!tag.startsWith("gimp xcf") || !tag.endsWith('\0')) {
qCDebug(XCFPLUGIN) << "XCF: read called on non-XCF file";
return false;
}
// Remove null terminator
tag.chop(1);
if (tag.right(4) == "file") {
xcf_io.setVersion(0);
} else {
// Version 1 and onwards use the format "gimp xcf v###" instead of "gimp xcf file"
bool ok;
xcf_io.setVersion(tag.right(3).toInt(&ok));
if (!ok) {
qCDebug(XCFPLUGIN) << "Failed to parse version" << tag;
return false;
}
}
qCDebug(XCFPLUGIN) << "version" << xcf_io.version();
if (xcf_io.version() > 12) {
qCDebug(XCFPLUGIN) << "Unsupported version" << xcf_io.version();
return false;
}
xcf_io >> header->width >> header->height >> header->type;
if (xcf_io.version() >= 4) {
int precision;
xcf_io >> precision;
qCDebug(XCFPLUGIN) << "Precision" << GimpPrecision(precision);
if (xcf_io.version() < 7) {
switch (precision) {
case 0:
precision = GIMP_PRECISION_U8_NON_LINEAR;
break;
case 1:
precision = GIMP_PRECISION_U16_NON_LINEAR;
break;
case 2:
precision = GIMP_PRECISION_U32_LINEAR;
break;
case 3:
precision = GIMP_PRECISION_HALF_LINEAR;
break;
case 4:
precision = GIMP_PRECISION_FLOAT_LINEAR;
break;
default:
if (precision < GIMP_PRECISION_U8_LINEAR) {
qCWarning(XCFPLUGIN) << "Invalid precision read" << precision;
return false;
} else {
qCDebug(XCFPLUGIN) << "Unexpected precision" << precision << "in version" << xcf_io.version();
}
}
}
header->precision = GimpPrecision(precision);
}
qCDebug(XCFPLUGIN) << "tag:" << tag << " height: " << header->width << " width: " << header->height << " type: " << header->type;
if ((sizeof(void *) == 4 && qint64(header->width) * header->height > 16384 * 16384)) {
qCWarning(XCFPLUGIN) << "On 32-bits programs the maximum image size is limited to" << 16384 << "x" << 16384 << "px";
return false;
}
if (header->width > MAX_IMAGE_WIDTH || header->height > MAX_IMAGE_HEIGHT) {
qCWarning(XCFPLUGIN) << "The maximum image size is limited to" << MAX_IMAGE_WIDTH << "x" << MAX_IMAGE_HEIGHT << "px";
return false;
}
return true;
}
bool XCFImageFormat::readXCF(QIODevice *device, QImage *outImage)
{
XCFImage xcf_image;
QDataStream xcf_io(device);
if (!readXCFHeader(xcf_io, &xcf_image.header)) {
return false;
}
if (!loadImageProperties(xcf_io, xcf_image)) {
return false;
}
// The layers appear to be stored in top-to-bottom order. This is
// the reverse of how a merged image must be computed. So, the layer
// offsets are pushed onto a LIFO stack (thus, we don't have to load
// all the data of all layers before beginning to construct the
// merged image).
QStack<qint64> layer_offsets;
while (true) {
const qint64 layer_offset = readOffsetPtr(xcf_io);
if (layer_offset == 0) {
break;
}
if (layer_offset < 0) {
qCDebug(XCFPLUGIN) << "XCF: negative layer offset";
return false;
}
layer_offsets.push(layer_offset);
}
xcf_image.num_layers = layer_offsets.size();
if (layer_offsets.size() == 0) {
qCDebug(XCFPLUGIN) << "XCF: no layers!";
return false;
}
qCDebug(XCFPLUGIN) << xcf_image.num_layers << "layers";
// Load each layer and add it to the image
while (!layer_offsets.isEmpty()) {
qint64 layer_offset = layer_offsets.pop();
xcf_io.device()->seek(layer_offset);
if (!loadLayer(xcf_io, xcf_image)) {
return false;
}
}
if (!xcf_image.initialized) {
qCDebug(XCFPLUGIN) << "XCF: no visible layers!";
return false;
}
// The image was created: now I can set metadata and ICC color profile inside it.
setImageParasites(xcf_image, xcf_image.image);
*outImage = xcf_image.image;
return true;
}
/*!
* An XCF file can contain an arbitrary number of properties associated
* with the image (and layer and mask).
* \param xcf_io the data stream connected to the XCF image
* \param xcf_image XCF image data.
* \return true if there were no I/O errors.
*/
bool XCFImageFormat::loadImageProperties(QDataStream &xcf_io, XCFImage &xcf_image)
{
while (true) {
PropType type;
QByteArray bytes;
quint32 rawType;
if (!loadProperty(xcf_io, type, bytes, rawType)) {
qCDebug(XCFPLUGIN) << "XCF: error loading global image properties";
return false;
}
QDataStream property(bytes);
switch (type) {
case PROP_END:
return true;
case PROP_COMPRESSION:
property >> xcf_image.compression;
break;
case PROP_RESOLUTION:
property.setFloatingPointPrecision(QDataStream::SinglePrecision);
property >> xcf_image.x_resolution >> xcf_image.y_resolution;
break;
case PROP_TATTOO:
property >> xcf_image.tattoo;
break;
case PROP_PARASITES:
while (!property.atEnd()) {
char *tag;
quint32 size;
property.readBytes(tag, size);
quint32 flags;
QByteArray data;
property >> flags >> data;
// WARNING: you cannot add metadata to QImage here because it can be null.
// Adding a metadata to a QImage when it is null, does nothing (metas are lost).
if (tag) // store metadata for future use
xcf_image.parasites.insert(QString::fromUtf8(tag), data);
delete[] tag;
}
break;
case PROP_UNIT:
property >> xcf_image.unit;
break;
case PROP_PATHS: // This property is ignored.
break;
case PROP_USER_UNIT: // This property is ignored.
break;
case PROP_COLORMAP:
property >> xcf_image.num_colors;
if (xcf_image.num_colors < 0 || xcf_image.num_colors > 65535) {
return false;
}
xcf_image.palette = QVector<QRgb>();
xcf_image.palette.reserve(xcf_image.num_colors);
for (int i = 0; i < xcf_image.num_colors; i++) {
uchar r;
uchar g;
uchar b;
property >> r >> g >> b;
xcf_image.palette.push_back(qRgb(r, g, b));
}
break;
default:
qCDebug(XCFPLUGIN) << "XCF: unimplemented image property" << type << "(" << rawType << ")"
<< ", size " << bytes.size();
break;
}
}
}
/*!
* Read a single property from the image file. The property type is returned
* in type and the data is returned in bytes.
* \param xcf the image file data stream.
* \param type returns with the property type.
* \param bytes returns with the property data.
* \return true if there were no IO errors. */
bool XCFImageFormat::loadProperty(QDataStream &xcf_io, PropType &type, QByteArray &bytes, quint32 &rawType)
{
quint32 size;
xcf_io >> rawType;
if (rawType >= MAX_SUPPORTED_PROPTYPE) {
type = MAX_SUPPORTED_PROPTYPE;
// we don't support the property, but we still need to read from the device, assume it's like all the
// non custom properties that is data_length + data
xcf_io >> size;
xcf_io.skipRawData(size);
// return true because we don't really want to totally fail on an unsupported property since it may not be fatal
return true;
}
type = PropType(rawType);
char *data = nullptr;
// The colormap property size is not the correct number of bytes:
// The GIMP source xcf.c has size = 4 + ncolors, but it should be
// 4 + 3 * ncolors
if (type == PROP_COLORMAP) {
xcf_io >> size;
quint32 ncolors;
xcf_io >> ncolors;
size = 3 * ncolors + 4;
if (size > 65535 || size < 4) {
return false;
}
data = new char[size];
// since we already read "ncolors" from the stream, we put that data back
data[0] = 0;
data[1] = 0;
data[2] = ncolors >> 8;
data[3] = ncolors & 255;
// ... and read the remaining bytes from the stream
xcf_io.readRawData(data + 4, size - 4);
} else if (type == PROP_USER_UNIT) {
// The USER UNIT property size is not correct. I'm not sure why, though.
float factor;
qint32 digits;
xcf_io >> size >> factor >> digits;
for (int i = 0; i < 5; i++) {
char *unit_strings;
xcf_io >> unit_strings;
delete[] unit_strings;
if (xcf_io.device()->atEnd()) {
qCDebug(XCFPLUGIN) << "XCF: read failure on property " << type;
return false;
}
}
size = 0;
} else {
xcf_io >> size;
if (size > 256000) {
return false;
}
data = new char[size];
const quint32 dataRead = xcf_io.readRawData(data, size);
if (dataRead < size) {
qCDebug(XCFPLUGIN) << "XCF: loadProperty read less data than expected" << size << dataRead;
memset(&data[dataRead], 0, size - dataRead);
}
}
if (size != 0 && data) {
bytes = QByteArray(data, size);
}
delete[] data;
return true;
}
/*!
* Load a layer from the XCF file. The data stream must be positioned at
* the beginning of the layer data.
* \param xcf_io the image file data stream.
* \param xcf_image contains the layer and the color table
* (if the image is indexed).
* \return true if there were no I/O errors.
*/
bool XCFImageFormat::loadLayer(QDataStream &xcf_io, XCFImage &xcf_image)
{
Layer &layer(xcf_image.layer);
delete[] layer.name;
xcf_io >> layer.width >> layer.height >> layer.type >> layer.name;
// Don't want to keep passing this around, dumb XCF format
layer.compression = XcfCompressionType(xcf_image.compression);
if (!loadLayerProperties(xcf_io, layer)) {
return false;
}
qCDebug(XCFPLUGIN) << "layer: \"" << layer.name << "\", size: " << layer.width << " x " << layer.height << ", type: " << layer.type
<< ", mode: " << layer.mode << ", opacity: " << layer.opacity << ", visible: " << layer.visible << ", offset: " << layer.x_offset << ", "
<< layer.y_offset << ", compression" << layer.compression;
// Skip reading the rest of it if it is not visible. Typically, when
// you export an image from the The GIMP it flattens (or merges) only
// the visible layers into the output image.
if (layer.visible == 0) {
return true;
}
// If there are any more layers, merge them into the final QImage.
layer.hierarchy_offset = readOffsetPtr(xcf_io);
layer.mask_offset = readOffsetPtr(xcf_io);
if (layer.hierarchy_offset < 0) {
qCDebug(XCFPLUGIN) << "XCF: negative layer hierarchy_offset";
return false;
}
if (layer.mask_offset < 0) {
qCDebug(XCFPLUGIN) << "XCF: negative layer mask_offset";
return false;
}
// Allocate the individual tile QImages based on the size and type
// of this layer.
if (!composeTiles(xcf_image)) {
return false;
}
xcf_io.device()->seek(layer.hierarchy_offset);
// As tiles are loaded, they are copied into the layers tiles by
// this routine. (loadMask(), below, uses a slightly different
// version of assignBytes().)
layer.assignBytes = assignImageBytes;
if (!loadHierarchy(xcf_io, layer, xcf_image.header.precision)) {
return false;
}
if (layer.mask_offset != 0) {
// 9 means its not on the file. Spec says "If the property does not appear for a layer which has a layer mask, it defaults to true (1).
if (layer.apply_mask == 9) {
layer.apply_mask = 1;
}
xcf_io.device()->seek(layer.mask_offset);
if (!loadMask(xcf_io, layer, xcf_image.header.precision)) {
return false;
}
} else {
// Spec says "Robust readers should force this to false if the layer has no layer mask."
layer.apply_mask = 0;
}
// Now we should have enough information to initialize the final
// QImage. The first visible layer determines the attributes
// of the QImage.
if (!xcf_image.initialized) {
if (!initializeImage(xcf_image)) {
return false;
}
copyLayerToImage(xcf_image);
xcf_image.initialized = true;
} else {
const QColorSpace colorspaceBefore = xcf_image.image.colorSpace();
mergeLayerIntoImage(xcf_image);
if (xcf_image.image.colorSpace() != colorspaceBefore) {
qCDebug(XCFPLUGIN) << "Converting color space back to" << colorspaceBefore << "after layer composition";
xcf_image.image.convertToColorSpace(colorspaceBefore);
}
}
return true;
}
/*!
* An XCF file can contain an arbitrary number of properties associated
* with a layer.
* \param xcf_io the data stream connected to the XCF image.
* \param layer layer to collect the properties.
* \return true if there were no I/O errors.
*/
bool XCFImageFormat::loadLayerProperties(QDataStream &xcf_io, Layer &layer)
{
while (true) {
PropType type;
QByteArray bytes;
quint32 rawType;
if (!loadProperty(xcf_io, type, bytes, rawType)) {
qCDebug(XCFPLUGIN) << "XCF: error loading layer properties";
return false;
}
QDataStream property(bytes);
switch (type) {
case PROP_END:
return true;
case PROP_ACTIVE_LAYER:
layer.active = true;
break;
case PROP_OPACITY:
property >> layer.opacity;
layer.opacity = std::min(layer.opacity, 255u);
break;
case PROP_FLOAT_OPACITY:
// For some reason QDataStream isn't able to read the float (tried
// setting the endianness manually)
if (bytes.size() == 4) {
layer.opacityFloat = qFromBigEndian(*reinterpret_cast<float *>(bytes.data()));
} else {
qCDebug(XCFPLUGIN) << "XCF: Invalid data size for float:" << bytes.size();
}
break;
case PROP_VISIBLE:
property >> layer.visible;
break;
case PROP_LINKED:
property >> layer.linked;
break;
case PROP_LOCK_ALPHA:
property >> layer.preserve_transparency;
break;
case PROP_APPLY_MASK:
property >> layer.apply_mask;
break;
case PROP_EDIT_MASK:
property >> layer.edit_mask;
break;
case PROP_SHOW_MASK:
property >> layer.show_mask;
break;
case PROP_OFFSETS:
property >> layer.x_offset >> layer.y_offset;
break;
case PROP_MODE:
property >> layer.mode;
if (layer.mode >= GIMP_LAYER_MODE_COUNT) {
qCDebug(XCFPLUGIN) << "Found layer with unsupported mode" << LayerModeType(layer.mode) << "Defaulting to mode 0";
layer.mode = GIMP_LAYER_MODE_NORMAL_LEGACY;
}
break;
case PROP_TATTOO:
property >> layer.tattoo;
break;
case PROP_COMPOSITE_SPACE:
property >> layer.compositeSpace;
if (layer.compositeSpace < 0) {
layer.compositeSpace = GimpColorSpace(-layer.compositeSpace);
}
break;
case PROP_COMPOSITE_MODE:
property >> layer.compositeMode;
if (layer.compositeMode < 0) {
layer.compositeMode = XCFImageFormat::GimpCompositeMode(-layer.compositeMode);
}
break;
case PROP_BLEND_SPACE:
property >> layer.blendSpace;
if (layer.blendSpace) {
layer.blendSpace = GimpColorSpace(-layer.blendSpace);
}
break;
// Just for organization in the UI, doesn't influence rendering
case PROP_COLOR_TAG:
break;
// We don't support editing, so for now just ignore locking
case PROP_LOCK_CONTENT:
case PROP_LOCK_POSITION:
break;
default:
qCDebug(XCFPLUGIN) << "XCF: unimplemented layer property " << type << "(" << rawType << ")"
<< ", size " << bytes.size();
break;
}
}
}
/*!
* Compute the number of tiles in the current layer and allocate
* QImage structures for each of them.
* \param xcf_image contains the current layer.
*/
bool XCFImageFormat::composeTiles(XCFImage &xcf_image)
{
Layer &layer(xcf_image.layer);
layer.nrows = (layer.height + TILE_HEIGHT - 1) / TILE_HEIGHT;
layer.ncols = (layer.width + TILE_WIDTH - 1) / TILE_WIDTH;
qCDebug(XCFPLUGIN) << "IMAGE: height=" << xcf_image.header.height << ", width=" << xcf_image.header.width;
qCDebug(XCFPLUGIN) << "LAYER: height=" << layer.height << ", width=" << layer.width;
qCDebug(XCFPLUGIN) << "LAYER: rows=" << layer.nrows << ", columns=" << layer.ncols;
// NOTE: starting from GIMP 2.10, images can be very large. The 32K limit for width and height is obsolete
// and it was changed to 300000 (the same as Photoshop Big image). This plugin was able to open an RGB
// image of 108000x40000 pixels saved with GIMP 2.10
// SANITY CHECK: Catch corrupted XCF image file where the width or height
// of a tile is reported are bogus. See Bug# 234030.
if ((sizeof(void *) == 4 && qint64(layer.width) * layer.height > 16384 * 16384)) {
qCWarning(XCFPLUGIN) << "On 32-bits programs the maximum layer size is limited to" << 16384 << "x" << 16384 << "px";
return false;
}
if (layer.width > MAX_IMAGE_WIDTH || layer.height > MAX_IMAGE_HEIGHT) {
qCWarning(XCFPLUGIN) << "The maximum layer size is limited to" << MAX_IMAGE_WIDTH << "x" << MAX_IMAGE_HEIGHT << "px";
return false;
}
// NOTE: A layer is a named rectangular area of pixels which has a definite position with respect to the canvas.
// It may extend beyond the canvas or (more commonly) only cover some of it.
// SANITY CHECK: Avoid to load XCF with a layer grater than 10 times the final image
if (qint64(layer.width) * layer.height / 10 > qint64(xcf_image.header.width) * xcf_image.header.height) {
if (qint64(layer.width) * layer.height > 16384 * 16384) { // large layers only
qCWarning(XCFPLUGIN) << "Euristic sanity check: the image may be corrupted!";
return false;
}
}
#ifndef XCF_QT5_SUPPORT
// Qt 6 image allocation limit calculation: we have to check the limit here because the image is splitted in
// tiles of 64x64 pixels. The required memory to build the image is at least doubled because tiles are loaded
// and then the final image is created by copying the tiles inside it.
// NOTE: on Windows to open a 10GiB image the plugin uses 28GiB of RAM
qint64 channels = 1 + (layer.type == RGB_GIMAGE ? 2 : 0) + (layer.type == RGBA_GIMAGE ? 3 : 0);
if (qint64(layer.width) * qint64(layer.height) * channels * 2ll / 1024ll / 1024ll > QImageReader::allocationLimit()) {
qCDebug(XCFPLUGIN) << "Rejecting image as it exceeds the current allocation limit of" << QImageReader::allocationLimit() << "megabytes";
return false;
}
#endif
layer.image_tiles.resize(layer.nrows);
if (layer.type == GRAYA_GIMAGE || layer.type == INDEXEDA_GIMAGE) {
layer.alpha_tiles.resize(layer.nrows);
}
if (layer.mask_offset != 0) {
layer.mask_tiles.resize(layer.nrows);
}
for (uint j = 0; j < layer.nrows; j++) {
layer.image_tiles[j].resize(layer.ncols);
if (layer.type == GRAYA_GIMAGE || layer.type == INDEXEDA_GIMAGE) {
layer.alpha_tiles[j].resize(layer.ncols);
}
if (layer.mask_offset != 0) {
layer.mask_tiles[j].resize(layer.ncols);
}
}
const QImage::Format format = layer.qimageFormat(xcf_image.header.precision);
for (uint j = 0; j < layer.nrows; j++) {
for (uint i = 0; i < layer.ncols; i++) {
uint tile_width = (i + 1) * TILE_WIDTH <= layer.width ? TILE_WIDTH : layer.width - i * TILE_WIDTH;
uint tile_height = (j + 1) * TILE_HEIGHT <= layer.height ? TILE_HEIGHT : layer.height - j * TILE_HEIGHT;
// Try to create the most appropriate QImage (each GIMP layer
// type is treated slightly differently)
switch (layer.type) {
case RGB_GIMAGE:
case RGBA_GIMAGE:
layer.image_tiles[j][i] = QImage(tile_width, tile_height, format);
if (layer.image_tiles[j][i].isNull()) {
return false;
}
layer.image_tiles[j][i].setColorCount(0);
break;
case GRAY_GIMAGE:
layer.image_tiles[j][i] = QImage(tile_width, tile_height, QImage::Format_Indexed8);
if (layer.image_tiles[j][i].isNull()) {
return false;
}
layer.image_tiles[j][i].setColorCount(256);
setGrayPalette(layer.image_tiles[j][i]);
break;
case GRAYA_GIMAGE:
layer.image_tiles[j][i] = QImage(tile_width, tile_height, QImage::Format_Indexed8);
layer.image_tiles[j][i].setColorCount(256);
if (layer.image_tiles[j][i].isNull()) {
return false;
}
setGrayPalette(layer.image_tiles[j][i]);
layer.alpha_tiles[j][i] = QImage(tile_width, tile_height, QImage::Format_Indexed8);
if (layer.alpha_tiles[j][i].isNull()) {
return false;
}
layer.alpha_tiles[j][i].setColorCount(256);
setGrayPalette(layer.alpha_tiles[j][i]);
break;
case INDEXED_GIMAGE:
layer.image_tiles[j][i] = QImage(tile_width, tile_height, QImage::Format_Indexed8);
layer.image_tiles[j][i].setColorCount(xcf_image.num_colors);
if (layer.image_tiles[j][i].isNull()) {
return false;
}
setPalette(xcf_image, layer.image_tiles[j][i]);
break;
case INDEXEDA_GIMAGE:
layer.image_tiles[j][i] = QImage(tile_width, tile_height, QImage::Format_Indexed8);
if (layer.image_tiles[j][i].isNull()) {
return false;
}
layer.image_tiles[j][i].setColorCount(xcf_image.num_colors);
setPalette(xcf_image, layer.image_tiles[j][i]);
layer.alpha_tiles[j][i] = QImage(tile_width, tile_height, QImage::Format_Indexed8);
if (layer.alpha_tiles[j][i].isNull()) {
return false;
}
layer.alpha_tiles[j][i].setColorCount(256);
setGrayPalette(layer.alpha_tiles[j][i]);
}
if (layer.type != GRAYA_GIMAGE && layer.image_tiles[j][i].format() != format) {
qCWarning(XCFPLUGIN) << "Selected wrong tile format" << layer.image_tiles[j][i].format() << "expected" << format;
return false;
}
#ifndef DISABLE_TILE_PROFILE
switch (xcf_image.header.precision) {
case XCFImageFormat::GIMP_PRECISION_HALF_LINEAR:
case XCFImageFormat::GIMP_PRECISION_FLOAT_LINEAR:
case XCFImageFormat::GIMP_PRECISION_DOUBLE_LINEAR:
case XCFImageFormat::GIMP_PRECISION_U8_LINEAR:
case XCFImageFormat::GIMP_PRECISION_U16_LINEAR:
case XCFImageFormat::GIMP_PRECISION_U32_LINEAR:
layer.image_tiles[j][i].setColorSpace(QColorSpace::SRgbLinear);
break;
case XCFImageFormat::GIMP_PRECISION_HALF_NON_LINEAR:
case XCFImageFormat::GIMP_PRECISION_FLOAT_NON_LINEAR:
case XCFImageFormat::GIMP_PRECISION_DOUBLE_NON_LINEAR:
case XCFImageFormat::GIMP_PRECISION_U8_NON_LINEAR:
case XCFImageFormat::GIMP_PRECISION_U16_NON_LINEAR:
case XCFImageFormat::GIMP_PRECISION_U32_NON_LINEAR:
layer.image_tiles[j][i].setColorSpace(QColorSpace::SRgb);
break;
case XCFImageFormat::GIMP_PRECISION_HALF_PERCEPTUAL:
case XCFImageFormat::GIMP_PRECISION_FLOAT_PERCEPTUAL:
case XCFImageFormat::GIMP_PRECISION_DOUBLE_PERCEPTUAL:
case XCFImageFormat::GIMP_PRECISION_U8_PERCEPTUAL:
case XCFImageFormat::GIMP_PRECISION_U16_PERCEPTUAL:
case XCFImageFormat::GIMP_PRECISION_U32_PERCEPTUAL:
layer.image_tiles[j][i].setColorSpace(QColorSpace::SRgb);
break;
}
#endif
if (layer.mask_offset != 0) {
layer.mask_tiles[j][i] = QImage(tile_width, tile_height, QImage::Format_Indexed8);
layer.mask_tiles[j][i].setColorCount(256);
if (layer.mask_tiles[j][i].isNull()) {
return false;
}
setGrayPalette(layer.mask_tiles[j][i]);
}
}
}
return true;
}
/*!
* Apply a grayscale palette to the QImage. Note that Qt does not distinguish
* between grayscale and indexed images. A grayscale image is just
* an indexed image with a 256-color, grayscale palette.
* \param image image to set to a grayscale palette.
*/
void XCFImageFormat::setGrayPalette(QImage &image)
{
if (grayTable.isEmpty()) {
grayTable.resize(256);
for (int i = 0; i < 256; i++) {
grayTable[i] = qRgb(i, i, i);
}
}
image.setColorTable(grayTable);
}
/*!
* Copy the indexed palette from the XCF image into the QImage.
* \param xcf_image XCF image containing the palette read from the data stream.
* \param image image to apply the palette to.
*/
void XCFImageFormat::setPalette(XCFImage &xcf_image, QImage &image)
{
Q_ASSERT(xcf_image.num_colors == xcf_image.palette.size());
image.setColorTable(xcf_image.palette);
}
/*!
* Copy the parasites info to QImage.
* \param xcf_image XCF image containing the parasites read from the data stream.
* \param image image to apply the parasites data.
* \note Some comment taken from https://gitlab.gnome.org/GNOME/gimp/-/blob/master/devel-docs/parasites.txt
*/
void XCFImageFormat::setImageParasites(const XCFImage &xcf_image, QImage &image)
{
auto&& p = xcf_image.parasites;
auto keys = p.keys();
for (auto &&key : std::as_const(keys)) {
auto value = p.value(key);
if (value.isEmpty())
continue;
// "icc-profile" (IMAGE, PERSISTENT | UNDOABLE)
// This contains an ICC profile describing the color space the
// image was produced in. TIFF images stored in PhotoShop do
// oftentimes contain embedded profiles. An experimental color
// manager exists to use this parasite, and it will be used
// for interchange between TIFF and PNG (identical profiles)
if (key == QStringLiteral("icc-profile")) {
auto cs = QColorSpace::fromIccProfile(value);
if (cs.isValid())
image.setColorSpace(cs);
continue;
}
// "gimp-comment" (IMAGE, PERSISTENT)
// Standard GIF-style image comments. This parasite should be
// human-readable text in UTF-8 encoding. A trailing \0 might
// be included and is not part of the comment. Note that image
// comments may also be present in the "gimp-metadata" parasite.
if (key == QStringLiteral("gimp-comment")) {
value.replace('\0', QByteArray());
image.setText(QStringLiteral("Comment"), QString::fromUtf8(value));
continue;
}
// "gimp-image-metadata"
// Saved by GIMP 2.10.30 but it is not mentioned in the specification.
// It is an XML block with the properties set using GIMP.
if (key == QStringLiteral("gimp-image-metadata")) {
// NOTE: I arbitrary defined the metadata "XML:org.gimp.xml" because it seems
// a GIMP proprietary XML format (no xmlns defined)
value.replace('\0', QByteArray());
image.setText(QStringLiteral("XML:org.gimp.xml"), QString::fromUtf8(value));
continue;
}
#if 0 // Unable to generate it using latest GIMP version
// "gimp-metadata" (IMAGE, PERSISTENT)
// The metadata associated with the image, serialized as one XMP
// packet. This metadata includes the contents of any XMP, EXIF
// and IPTC blocks from the original image, as well as
// user-specified values such as image comment, copyright,
// license, etc.
if (key == QStringLiteral("gimp-metadata")) {
// NOTE: "XML:com.adobe.xmp" is the meta set by Qt reader when an
// XMP packet is found (e.g. when reading a PNG saved by Photoshop).
// I reused the same key because some programs could search for it.
value.replace('\0', QByteArray());
image.setText(QStringLiteral("XML:com.adobe.xmp"), QString::fromUtf8(value));
continue;
}
#endif
}
#ifdef DISABLE_IMAGE_PROFILE
// final colorspace checks
if (!image.colorSpace().isValid()) {
switch (xcf_image.header.precision) {
case XCFImageFormat::GIMP_PRECISION_HALF_LINEAR:
case XCFImageFormat::GIMP_PRECISION_FLOAT_LINEAR:
case XCFImageFormat::GIMP_PRECISION_DOUBLE_LINEAR:
case XCFImageFormat::GIMP_PRECISION_U8_LINEAR:
case XCFImageFormat::GIMP_PRECISION_U16_LINEAR:
case XCFImageFormat::GIMP_PRECISION_U32_LINEAR:
image.setColorSpace(QColorSpace::SRgbLinear);
break;
default:
image.setColorSpace(QColorSpace::SRgb);
break;
}
}
#endif
}
/*!
* Copy the bytes from the tile buffer into the image tile QImage, taking into
* account all the myriad different modes.
* \param layer layer containing the tile buffer and the image tile matrix.
* \param i column index of current tile.
* \param j row index of current tile.
*/
bool XCFImageFormat::assignImageBytes(Layer &layer, uint i, uint j, const GimpPrecision &precision)
{
QImage &image = layer.image_tiles[j][i];
const uchar *tile = layer.tile;
const int width = image.width();
const int height = image.height();
const int bytesPerLine = image.bytesPerLine();
uchar *bits = image.bits();
// Handle the special cases
if (layer.type == GRAYA_GIMAGE || layer.type == GRAY_GIMAGE || layer.type == INDEXEDA_GIMAGE) {
auto bpc = bytesPerChannel(precision);
for (int y = 0; y < height; y++) {
uchar *dataPtr = bits + y * bytesPerLine;
uchar *alphaPtr = nullptr;
if (!layer.alpha_tiles.isEmpty())
alphaPtr = layer.alpha_tiles[j][i].scanLine(y);
if (bpc == 4) {
#ifdef USE_FLOAT_IMAGES
if (precision < GimpPrecision::GIMP_PRECISION_HALF_LINEAR) {
for (int x = 0; x < width; x++) {
auto src = (const quint16 *)tile;
*dataPtr++ = qFromBigEndian<quint16>(src[0]) / 257;
if (alphaPtr) {
*alphaPtr++ = qFromBigEndian<quint16>(src[1]) / 257;
tile += sizeof(quint16) * 2;
} else {
tile += sizeof(quint16);
}
}
} else {
for (int x = 0; x < width; x++) {
auto src = (const float *)tile;
*dataPtr++ = qFromBigEndian<float>(src[0]) * 255;
if (alphaPtr) {
*alphaPtr++ = qFromBigEndian<float>(src[1]) * 255;
tile += sizeof(float) * 2;
} else {
tile += sizeof(float);
}
}
}
#else
for (int x = 0; x < width; x++) {
auto src = (const quint16 *)tile;
*dataPtr++ = qFromBigEndian<quint16>(src[0]) / 257;
if (alphaPtr) {
*alphaPtr++ = qFromBigEndian<quint16>(src[1]) / 257;
tile += sizeof(quint16) * 2;
} else {
tile += sizeof(quint16);
}
}
#endif
} else if (bpc == 2) {
#ifdef USE_FLOAT_IMAGES
if (precision < GimpPrecision::GIMP_PRECISION_HALF_LINEAR) {
for (int x = 0; x < width; x++) {
auto src = (const quint16 *)tile;
*dataPtr++ = qFromBigEndian<quint16>(src[0]) / 257;
if (alphaPtr)
*alphaPtr++ = qFromBigEndian<quint16>(src[1]) / 257;
tile += sizeof(QRgb);
}
} else {
for (int x = 0; x < width; x++) {
auto src = (const qfloat16 *)tile;
*dataPtr++ = qFromBigEndian<qfloat16>(src[0]) * 255;
if (alphaPtr)
*alphaPtr++ = qFromBigEndian<qfloat16>(src[1]) * 255;
tile += sizeof(QRgb);
}
}
#else
for (int x = 0; x < width; x++) {
auto src = (const quint16 *)tile;
*dataPtr++ = qFromBigEndian<quint16>(src[0]) / 257;
if (alphaPtr)
*alphaPtr++ = qFromBigEndian<quint16>(src[1]) / 257;
tile += sizeof(QRgb);
}
#endif
} else {
for (int x = 0; x < width; x++) {
if (tile[0] < image.colorCount())
*dataPtr++ = tile[0];
if (alphaPtr)
*alphaPtr++ = tile[1];
tile += sizeof(QRgb);
}
}
}
return true;
}
switch (image.format()) {
case QImage::Format_RGBX8888:
for (int y = 0; y < height; y++) {
uchar *dataPtr = image.scanLine(y);
for (int x = 0; x < width * 4; x += 4, tile += 4) {
dataPtr[x + 0] = tile[0];
dataPtr[x + 1] = tile[1];
dataPtr[x + 2] = tile[2];
dataPtr[x + 3] = 255;
}
}
break;
case QImage::Format_RGBA8888:
for (int y = 0; y < height; y++) {
const size_t bpl = width * 4;
memcpy(image.scanLine(y), tile + y * bpl, bpl);
}
break;
case QImage::Format_RGBX64:
for (int y = 0; y < height; y++) {
quint16 *dataPtr = (quint16 *)image.scanLine(y);
const size_t bpl = width * sizeof(QRgba64);
const quint16 *src = (const quint16 *)(tile + y * bpl);
for (int x = 0; x < width * 4; x += 4) {
dataPtr[x + 0] = qFromBigEndian(src[x + 0]);
dataPtr[x + 1] = qFromBigEndian(src[x + 1]);
dataPtr[x + 2] = qFromBigEndian(src[x + 2]);
dataPtr[x + 3] = 65535;
}
}
break;
#ifdef USE_FLOAT_IMAGES
case QImage::Format_RGBX16FPx4:
for (int y = 0; y < height; y++) {
qfloat16 *dataPtr = (qfloat16 *)image.scanLine(y);
const qfloat16 *src = (const qfloat16 *)(tile + y * width * sizeof(QRgbaFloat16));
for (int x = 0; x < width * 4; x += 4) {
dataPtr[x + 0] = qFromBigEndian(src[x + 0]);
dataPtr[x + 1] = qFromBigEndian(src[x + 1]);
dataPtr[x + 2] = qFromBigEndian(src[x + 2]);
dataPtr[x + 3] = qfloat16(1);
}
}
break;
case QImage::Format_RGBA16FPx4:
static_assert(sizeof(QRgbaFloat16) == sizeof(QRgba64), "Different sizes for float and int 16 bit pixels");
#endif
case QImage::Format_RGBA64:
for (int y = 0; y < height; y++) {
const size_t bpl = width * sizeof(QRgba64);
qFromBigEndian<qint16>(tile + y * bpl, width * 4, image.scanLine(y));
}
break;
#ifdef USE_FLOAT_IMAGES
case QImage::Format_RGBA32FPx4:
for (int y = 0; y < height; y++) {
const size_t bpl = width * sizeof(QRgbaFloat32);
qFromBigEndian<qint32>(tile + y * bpl, width * 4, image.scanLine(y));
}
break;
case QImage::Format_RGBX32FPx4:
for (int y = 0; y < height; y++) {
float *dataPtr = (float *)image.scanLine(y);
const float *src = (const float *)(tile + y * width * sizeof(QRgbaFloat32));
for (int x = 0; x < width * 4; x += 4) {
dataPtr[x + 0] = qFromBigEndian(src[x + 0]);
dataPtr[x + 1] = qFromBigEndian(src[x + 1]);
dataPtr[x + 2] = qFromBigEndian(src[x + 2]);
dataPtr[x + 3] = 1.f;
}
}
break;
#endif
case QImage::Format_Indexed8:
for (int y = 0; y < height; y++) {
uchar *dataPtr = bits + y * bytesPerLine;
for (int x = 0; x < width; x++) {
*dataPtr++ = tile[0];
tile += sizeof(QRgb);
}
}
break;
default:
qCWarning(XCFPLUGIN) << "Unhandled image format" << image.format() << "and/or layer type" << layer.type;
return false;
}
return true;
}
/*!
* The GIMP stores images in a "mipmap"-like hierarchy. As far as the QImage
* is concerned, however, only the top level (i.e., the full resolution image)
* is used.
* \param xcf_io the data stream connected to the XCF image.
* \param layer the layer to collect the image.
* \return true if there were no I/O errors.
*/
bool XCFImageFormat::loadHierarchy(QDataStream &xcf_io, Layer &layer, const GimpPrecision precision)
{
qint32 width;
qint32 height;
quint32 bpp;
xcf_io >> width >> height >> bpp;
const qint64 offset = readOffsetPtr(xcf_io);
qCDebug(XCFPLUGIN) << "width" << width << "height" << height << "bpp" << bpp << "offset" << offset;
if (offset < 0) {
qCDebug(XCFPLUGIN) << "XCF: negative hierarchy offset";
return false;
}
const bool isMask = layer.assignBytes == assignMaskBytes;
// make sure bpp is correct and complain if it is not
switch (layer.type) {
case RGB_GIMAGE:
if (bpp != 3 * bytesPerChannel(precision)) {
qCDebug(XCFPLUGIN) << "Found layer of type RGB but with bpp != 3" << bpp;
if (!isMask) {
return false;
}
}
break;
case RGBA_GIMAGE:
if (bpp != 4 * bytesPerChannel(precision)) {
qCDebug(XCFPLUGIN) << "Found layer of type RGBA but with bpp != 4, got" << bpp << "bpp";
if (!isMask) {
return false;
}
}
break;
case GRAY_GIMAGE:
if (bpp != 1 * bytesPerChannel(precision)) {
qCDebug(XCFPLUGIN) << "Found layer of type Gray but with bpp != 1" << bpp;
return false;
}
break;
case GRAYA_GIMAGE:
if (bpp != 2 * bytesPerChannel(precision)) {
qCDebug(XCFPLUGIN) << "Found layer of type Gray+Alpha but with bpp != 2" << bpp;
if (!isMask) {
return false;
}
}
break;
case INDEXED_GIMAGE:
if (bpp != 1 * bytesPerChannel(precision)) {
qCDebug(XCFPLUGIN) << "Found layer of type Indexed but with bpp != 1" << bpp;
return false;
}
break;
case INDEXEDA_GIMAGE:
if (bpp != 2 * bytesPerChannel(precision)) {
qCDebug(XCFPLUGIN) << "Found layer of type Indexed+Alpha but with bpp != 2" << bpp;
if (!isMask) {
return false;
}
}
break;
}
if (bpp > 4 * bytesPerChannel(precision)) {
qCDebug(XCFPLUGIN) << "bpp is" << bpp << "We don't support layers with bpp > 4";
return false;
}
// GIMP stores images in a "mipmap"-like format (multiple levels of
// increasingly lower resolution). Only the top level is used here,
// however.
quint32 junk;
do {
xcf_io >> junk;
if (xcf_io.device()->atEnd()) {
qCDebug(XCFPLUGIN) << "XCF: read failure on layer " << layer.name << " level offsets";
return false;
}
} while (junk != 0);
qint64 saved_pos = xcf_io.device()->pos();
xcf_io.device()->seek(offset);
if (!loadLevel(xcf_io, layer, bpp, precision)) {
return false;
}
xcf_io.device()->seek(saved_pos);
return true;
}
template<typename SourceFormat>
static bool convertFloatTo16Bit(uchar *output, quint64 outputSize, uchar *input)
{
SourceFormat *source = (SourceFormat *)(input);
for (quint64 offset = 0; offset < outputSize; offset++) {
((uint16_t *)output)[offset] = qToBigEndian(quint16(qBound(0., qFromBigEndian<SourceFormat>(source[offset]) * 65535. + 0.5, 65535.)));
}
return true;
}
/*!
* Load one level of the image hierarchy (but only the top level is ever used).
* \param xcf_io the data stream connected to the XCF image.
* \param layer the layer to collect the image.
* \param bpp the number of bytes in a pixel.
* \return true if there were no I/O errors.
* \sa loadTileRLE().
*/
bool XCFImageFormat::loadLevel(QDataStream &xcf_io, Layer &layer, qint32 bpp, const GimpPrecision precision)
{
qint32 width;
qint32 height;
xcf_io >> width >> height;
qint64 offset = readOffsetPtr(xcf_io);
if (offset < 0) {
qCDebug(XCFPLUGIN) << "XCF: negative level offset";
return false;
}
if (offset == 0) {
// offset 0 with rowsxcols != 0 is probably an error since it means we have tiles
// without data but just clear the bits for now instead of returning false
for (uint j = 0; j < layer.nrows; j++) {
for (uint i = 0; i < layer.ncols; i++) {
layer.image_tiles[j][i].fill(Qt::transparent);
if (layer.type == GRAYA_GIMAGE || layer.type == INDEXEDA_GIMAGE) {
layer.alpha_tiles[j][i].fill(Qt::transparent);
}
}
}
return true;
}
bool needConvert = true;
switch (precision) {
#ifdef USE_FLOAT_IMAGES
case GIMP_PRECISION_HALF_LINEAR:
case GIMP_PRECISION_HALF_NON_LINEAR:
case GIMP_PRECISION_HALF_PERCEPTUAL:
case GIMP_PRECISION_FLOAT_LINEAR:
case GIMP_PRECISION_FLOAT_NON_LINEAR:
case GIMP_PRECISION_FLOAT_PERCEPTUAL:
#endif
case GIMP_PRECISION_U8_LINEAR:
case GIMP_PRECISION_U8_NON_LINEAR:
case GIMP_PRECISION_U8_PERCEPTUAL:
case GIMP_PRECISION_U16_LINEAR:
case GIMP_PRECISION_U16_NON_LINEAR:
case GIMP_PRECISION_U16_PERCEPTUAL:
needConvert = false;
break;
default:
break;
}
const uint blockSize = TILE_WIDTH * TILE_HEIGHT * bpp * 1.5;
QVector<uchar> buffer;
if (needConvert) {
buffer.resize(blockSize * (bpp == 2 ? 2 : 1));
}
for (uint j = 0; j < layer.nrows; j++) {
for (uint i = 0; i < layer.ncols; i++) {
if (offset == 0) {
qCDebug(XCFPLUGIN) << "XCF: incorrect number of tiles in layer " << layer.name;
return false;
}
qint64 saved_pos = xcf_io.device()->pos();
qint64 offset2 = readOffsetPtr(xcf_io);
if (offset2 < 0) {
qCDebug(XCFPLUGIN) << "XCF: negative level offset";
return false;
}
// Evidently, RLE can occasionally expand a tile instead of compressing it!
if (offset2 == 0) {
offset2 = offset + blockSize;
}
xcf_io.device()->seek(offset);
qint64 bytesParsed = 0;
switch (layer.compression) {
case COMPRESS_NONE: {
if (xcf_io.version() > 11) {
qCDebug(XCFPLUGIN) << "Component reading not supported yet";
return false;
}
const int data_size = bpp * TILE_WIDTH * TILE_HEIGHT;
if (data_size > int(blockSize)) {
qCDebug(XCFPLUGIN) << "Tile data too big, we can only fit" << sizeof(layer.tile) << "but need" << data_size;
return false;
}
int dataRead = xcf_io.readRawData(reinterpret_cast<char *>(layer.tile), data_size);
if (dataRead < data_size) {
qCDebug(XCFPLUGIN) << "short read, expected" << data_size << "got" << dataRead;
return false;
}
bytesParsed = dataRead;
break;
}
case COMPRESS_RLE: {
int size = layer.image_tiles[j][i].width() * layer.image_tiles[j][i].height();
const uint data_size = size * bpp;
if (needConvert) {
if (data_size >= unsigned(buffer.size())) {
qCDebug(XCFPLUGIN) << "Tile data too big, we can only fit" << buffer.size() << "but need" << data_size;
return false;
}
} else {
if (data_size > sizeof(layer.tile)) {
qCDebug(XCFPLUGIN) << "Tile data too big, we can only fit" << sizeof(layer.tile) << "but need" << data_size;
return false;
}
if (blockSize > sizeof(layer.tile)) {
qCWarning(XCFPLUGIN) << "Too small tiles" << sizeof(layer.tile) << "this image requires" << blockSize << sizeof(QRgba64) << bpp;
return false;
}
}
if (!loadTileRLE(xcf_io, needConvert ? buffer.data() : layer.tile, size, offset2 - offset, bpp, &bytesParsed)) {
qCDebug(XCFPLUGIN) << "Failed to read RLE";
return false;
}
break;
}
default:
qCDebug(XCFPLUGIN) << "Unhandled compression" << layer.compression;
return false;
}
if (needConvert) {
if (bytesParsed > buffer.size()) {
qCDebug(XCFPLUGIN) << "Invalid number of bytes parsed" << bytesParsed << buffer.size();
return false;
}
switch (precision) {
case GIMP_PRECISION_U32_LINEAR:
case GIMP_PRECISION_U32_NON_LINEAR:
case GIMP_PRECISION_U32_PERCEPTUAL: {
quint32 *source = (quint32 *)(buffer.data());
for (quint64 offset = 0, len = buffer.size() / sizeof(quint32); offset < len; ++offset) {
((quint16 *)layer.tile)[offset] = qToBigEndian<quint16>(qFromBigEndian(source[offset]) / 65537);
}
break;
}
#ifndef USE_FLOAT_IMAGES
case GIMP_PRECISION_HALF_LINEAR:
case GIMP_PRECISION_HALF_NON_LINEAR:
case GIMP_PRECISION_HALF_PERCEPTUAL:
convertFloatTo16Bit<qfloat16>(layer.tile, buffer.size() / sizeof(qfloat16), buffer.data());
break;
case GIMP_PRECISION_FLOAT_LINEAR:
case GIMP_PRECISION_FLOAT_NON_LINEAR:
case GIMP_PRECISION_FLOAT_PERCEPTUAL:
convertFloatTo16Bit<float>(layer.tile, buffer.size() / sizeof(float), buffer.data());
break;
case GIMP_PRECISION_DOUBLE_LINEAR:
case GIMP_PRECISION_DOUBLE_NON_LINEAR:
case GIMP_PRECISION_DOUBLE_PERCEPTUAL:
convertFloatTo16Bit<double>(layer.tile, buffer.size() / sizeof(double), buffer.data());
break;
#else
case GIMP_PRECISION_DOUBLE_LINEAR:
case GIMP_PRECISION_DOUBLE_NON_LINEAR:
case GIMP_PRECISION_DOUBLE_PERCEPTUAL: {
double *source = (double *)(buffer.data());
for (quint64 offset = 0, len = buffer.size() / sizeof(double); offset < len; ++offset) {
((float *)layer.tile)[offset] = qToBigEndian<float>(float(qFromBigEndian(source[offset])));
}
break;
}
#endif
default:
qCWarning(XCFPLUGIN) << "Unsupported precision" << precision;
return false;
}
}
// The bytes in the layer tile are juggled differently depending on
// the target QImage. The caller has set layer.assignBytes to the
// appropriate routine.
if (!layer.assignBytes(layer, i, j, precision)) {
return false;
}
xcf_io.device()->seek(saved_pos);
offset = readOffsetPtr(xcf_io);
if (offset < 0) {
qCDebug(XCFPLUGIN) << "XCF: negative level offset";
return false;
}
}
}
return true;
}
/*!
* A layer can have a one channel image which is used as a mask.
* \param xcf_io the data stream connected to the XCF image.
* \param layer the layer to collect the mask image.
* \return true if there were no I/O errors.
*/
bool XCFImageFormat::loadMask(QDataStream &xcf_io, Layer &layer, const GimpPrecision precision)
{
qint32 width;
qint32 height;
char *name;
xcf_io >> width >> height >> name;
delete[] name;
if (!loadChannelProperties(xcf_io, layer)) {
return false;
}
const qint64 hierarchy_offset = readOffsetPtr(xcf_io);
if (hierarchy_offset < 0) {
qCDebug(XCFPLUGIN) << "XCF: negative mask hierarchy_offset";
return false;
}
xcf_io.device()->seek(hierarchy_offset);
layer.assignBytes = assignMaskBytes;
if (!loadHierarchy(xcf_io, layer, precision)) {
return false;
}
return true;
}
/*!
* This is the routine for which all the other code is simply
* infrastructure. Read the image bytes out of the file and
* store them in the tile buffer. This is passed a full 32-bit deep
* buffer, even if bpp is smaller. The caller can figure out what to
* do with the bytes.
*
* The tile is stored in "channels", i.e. the red component of all
* pixels, then the green component of all pixels, then blue then
* alpha, or, for indexed images, the color indices of all pixels then
* the alpha of all pixels.
*
* The data is compressed with "run length encoding". Some simple data
* integrity checks are made.
*
* \param xcf_io the data stream connected to the XCF image.
* \param tile the buffer to expand the RLE into.
* \param image_size number of bytes expected to be in the image tile.
* \param data_length number of bytes expected in the RLE.
* \param bpp number of bytes per pixel.
* \return true if there were no I/O errors and no obvious corruption of
* the RLE data.
*/
bool XCFImageFormat::loadTileRLE(QDataStream &xcf_io, uchar *tile, int image_size, int data_length, qint32 bpp, qint64 *bytesParsed)
{
uchar *data = tile;
uchar *xcfdata;
uchar *xcfodata;
uchar *xcfdatalimit;
int step = sizeof(QRgb);
switch (bpp) {
case 1:
case 2:
case 3:
case 4:
step = sizeof(QRgb);
break;
case 6:
case 8:
step = sizeof(QRgb) * 2;
break;
case 12:
case 16:
step = sizeof(QRgb) * 4;
break;
default:
qCDebug(XCFPLUGIN) << "XCF: unhandled bit depth" << bpp;
return false;
}
if (data_length < 0 || data_length > int(TILE_WIDTH * TILE_HEIGHT * step * 1.5)) {
qCDebug(XCFPLUGIN) << "XCF: invalid tile data length" << data_length;
return false;
}
xcfdata = xcfodata = new uchar[data_length];
const int dataRead = xcf_io.readRawData((char *)xcfdata, data_length);
if (dataRead <= 0) {
delete[] xcfodata;
qCDebug(XCFPLUGIN) << "XCF: read failure on tile" << dataRead;
return false;
}
if (dataRead < data_length) {
memset(&xcfdata[dataRead], 0, data_length - dataRead);
}
if (!xcf_io.device()->isOpen()) {
delete[] xcfodata;
qCDebug(XCFPLUGIN) << "XCF: read failure on tile";
return false;
}
xcfdatalimit = &xcfodata[data_length - 1];
for (int i = 0; i < bpp; ++i) {
data = tile + i;
int size = image_size;
while (size > 0) {
if (xcfdata > xcfdatalimit) {
goto bogus_rle;
}
uchar val = *xcfdata++;
uint length = val;
if (length >= 128) {
length = 255 - (length - 1);
if (length == 128) {
if (xcfdata >= xcfdatalimit) {
goto bogus_rle;
}
length = (*xcfdata << 8) + xcfdata[1];
xcfdata += 2;
}
size -= length;
if (size < 0) {
goto bogus_rle;
}
if (&xcfdata[length - 1] > xcfdatalimit) {
goto bogus_rle;
}
while (length-- > 0) {
*data = *xcfdata++;
data += step;
}
} else {
length += 1;
if (length == 128) {
if (xcfdata >= xcfdatalimit) {
goto bogus_rle;
}
length = (*xcfdata << 8) + xcfdata[1];
xcfdata += 2;
}
size -= length;
if (size < 0) {
goto bogus_rle;
}
if (xcfdata > xcfdatalimit) {
goto bogus_rle;
}
qintptr totalLength = qintptr(data - tile) + length * step;
if (totalLength >= image_size * step * 1.5) {
qCDebug(XCFPLUGIN) << "Ran out of space when trying to unpack image, over:" << totalLength - image_size << totalLength << image_size
<< length;
goto bogus_rle;
}
val = *xcfdata++;
while (length-- > 0) {
*data = val;
data += step;
}
}
}
}
*bytesParsed = qintptr(data - tile);
delete[] xcfodata;
return true;
bogus_rle:
qCDebug(XCFPLUGIN) << "The run length encoding could not be decoded properly";
delete[] xcfodata;
return false;
}
/*!
* An XCF file can contain an arbitrary number of properties associated
* with a channel. Note that this routine only reads mask channel properties.
* \param xcf_io the data stream connected to the XCF image.
* \param layer layer containing the mask channel to collect the properties.
* \return true if there were no I/O errors.
*/
bool XCFImageFormat::loadChannelProperties(QDataStream &xcf_io, Layer &layer)
{
while (true) {
PropType type;
QByteArray bytes;
quint32 rawType;
if (!loadProperty(xcf_io, type, bytes, rawType)) {
qCDebug(XCFPLUGIN) << "XCF: error loading channel properties";
return false;
}
QDataStream property(bytes);
switch (type) {
case PROP_END:
return true;
case PROP_OPACITY:
property >> layer.mask_channel.opacity;
layer.mask_channel.opacity = std::min(layer.mask_channel.opacity, 255u);
break;
case PROP_FLOAT_OPACITY:
// For some reason QDataStream isn't able to read the float (tried
// setting the endianness manually)
if (bytes.size() == 4) {
layer.mask_channel.opacityFloat = qFromBigEndian(*reinterpret_cast<float *>(bytes.data()));
} else {
qCDebug(XCFPLUGIN) << "XCF: Invalid data size for float:" << bytes.size();
}
break;
case PROP_VISIBLE:
property >> layer.mask_channel.visible;
break;
case PROP_SHOW_MASKED:
property >> layer.mask_channel.show_masked;
break;
case PROP_COLOR:
property >> layer.mask_channel.red >> layer.mask_channel.green >> layer.mask_channel.blue;
break;
case PROP_FLOAT_COLOR:
property >> layer.mask_channel.redF >> layer.mask_channel.greenF >> layer.mask_channel.blueF;
break;
case PROP_TATTOO:
property >> layer.mask_channel.tattoo;
break;
// Only used in edit mode
case PROP_LINKED:
break;
// Just for organization in the UI, doesn't influence rendering
case PROP_COLOR_TAG:
break;
// We don't support editing, so for now just ignore locking
case PROP_LOCK_CONTENT:
case PROP_LOCK_POSITION:
break;
default:
qCDebug(XCFPLUGIN) << "XCF: unimplemented channel property " << type << "(" << rawType << ")"
<< ", size " << bytes.size();
break;
}
}
}
/*!
* Copy the bytes from the tile buffer into the mask tile QImage.
* \param layer layer containing the tile buffer and the mask tile matrix.
* \param i column index of current tile.
* \param j row index of current tile.
*/
bool XCFImageFormat::assignMaskBytes(Layer &layer, uint i, uint j, const GimpPrecision &precision)
{
QImage &image = layer.mask_tiles[j][i];
if (image.depth() != 8) {
qCWarning(XCFPLUGIN) << "invalid bytes per pixel, we only do 8 bit masks" << image.depth();
return false;
}
uchar *tile = layer.tile;
const int width = image.width();
const int height = image.height();
const int bytesPerLine = image.bytesPerLine();
uchar *bits = image.bits();
auto bpc = bytesPerChannel(precision);
// mask management is a house of cards: the mask is always treated as 8 bit by the plugin
// (I don't want to twist the code) so it needs a conversion here.
// If previously converted the step is the type size, otherwise is the one set in loadTileRLE().
for (int y = 0; y < height; y++) {
uchar *dataPtr = bits + y * bytesPerLine;
#ifdef USE_FLOAT_IMAGES
if (bpc == 4) {
if (precision < GimpPrecision::GIMP_PRECISION_HALF_LINEAR) {
for (int x = 0; x < width; x++) {
*dataPtr++ = qFromBigEndian<quint16>(*(const quint16 *)tile) / 257;
tile += sizeof(quint16); // was converted to 16 bits in loadLevel()
}
} else {
for (int x = 0; x < width; x++) {
*dataPtr++ = qFromBigEndian<float>(*(const float *)tile) * 255;
tile += sizeof(QRgb); // yeah! see loadTileRLE()
}
}
} else if (bpc == 2) {
// when not converted, the step of a
if (precision < GimpPrecision::GIMP_PRECISION_HALF_LINEAR) {
for (int x = 0; x < width; x++) {
*dataPtr++ = qFromBigEndian<quint16>(*(const quint16 *)tile) / 257;
tile += sizeof(QRgb); // yeah! see loadTileRLE()
}
} else {
for (int x = 0; x < width; x++) {
*dataPtr++ = qFromBigEndian<qfloat16>(*(const qfloat16 *)tile) * 255;
tile += sizeof(QRgb); // yeah! see loadTileRLE()
}
}
}
#else
if (bpc == 2) {
for (int x = 0; x < width; x++) {
*dataPtr++ = qFromBigEndian<quint16>(*(const quint16 *)tile) / 257;
tile += sizeof(QRgb); // yeah! see loadTileRLE() / loadLevel()
}
} else if (bpc == 4) {
for (int x = 0; x < width; x++) {
*dataPtr++ = qFromBigEndian<quint16>(*(const quint16 *)tile) / 257;
tile += sizeof(quint16); // was converted to 16 bits in loadLevel()
}
}
#endif
else {
for (int x = 0; x < width; x++) {
*dataPtr++ = tile[0];
tile += sizeof(QRgb); // yeah! see loadTileRLE()
}
}
}
return true;
}
/*!
* Construct the QImage which will eventually be returned to the QImage
* loader.
*
* There are a couple of situations which require that the QImage is not
* exactly the same as The GIMP's representation. The full table is:
* \verbatim
* Grayscale opaque : 8 bpp indexed
* Grayscale translucent : 32 bpp + alpha
* Indexed opaque : 1 bpp if num_colors <= 2
* : 8 bpp indexed otherwise
* Indexed translucent : 8 bpp indexed + alpha if num_colors < 256
* : 32 bpp + alpha otherwise
* RGB opaque : 32 bpp
* RGBA translucent : 32 bpp + alpha
* \endverbatim
* Whether the image is translucent or not is determined by the bottom layer's
* alpha channel. However, even if the bottom layer lacks an alpha channel,
* it can still have an opacity < 1. In this case, the QImage is promoted
* to 32-bit. (Note this is different from the output from the GIMP image
* exporter, which seems to ignore this attribute.)
*
* Independently, higher layers can be translucent, but the background of
* the image will not show through if the bottom layer is opaque.
*
* For indexed images, translucency is an all or nothing effect.
* \param xcf_image contains image info and bottom-most layer.
*/
bool XCFImageFormat::initializeImage(XCFImage &xcf_image)
{
// (Aliases to make the code look a little better.)
Layer &layer(xcf_image.layer);
QImage &image(xcf_image.image);
switch (layer.type) {
case GRAY_GIMAGE:
if (layer.opacity == OPAQUE_OPACITY) {
image = imageAlloc(xcf_image.header.width, xcf_image.header.height, QImage::Format_Indexed8);
image.setColorCount(256);
if (image.isNull()) {
return false;
}
setGrayPalette(image);
image.fill(255);
break;
} // else, fall through to 32-bit representation
Q_FALLTHROUGH();
case GRAYA_GIMAGE:
case RGB_GIMAGE:
case RGBA_GIMAGE:
image = imageAlloc(xcf_image.header.width, xcf_image.header.height, xcf_image.qimageFormat());
if (image.isNull()) {
return false;
}
if (image.hasAlphaChannel()) {
image.fill(Qt::transparent);
} else {
image.fill(Qt::white);
}
break;
case INDEXED_GIMAGE:
// As noted in the table above, there are quite a few combinations
// which are possible with indexed images, depending on the
// presence of transparency (note: not translucency, which is not
// supported by The GIMP for indexed images) and the number of
// individual colors.
// Note: Qt treats a bitmap with a Black and White color palette
// as a mask, so only the "on" bits are drawn, regardless of the
// order color table entries. Otherwise (i.e., at least one of the
// color table entries is not black or white), it obeys the one-
// or two-color palette. Have to ask about this...
if (xcf_image.num_colors <= 2) {
image = imageAlloc(xcf_image.header.width, xcf_image.header.height, QImage::Format_MonoLSB);
image.setColorCount(xcf_image.num_colors);
if (image.isNull()) {
return false;
}
image.fill(0);
setPalette(xcf_image, image);
} else if (xcf_image.num_colors <= 256) {
image = imageAlloc(xcf_image.header.width, xcf_image.header.height, QImage::Format_Indexed8);
image.setColorCount(xcf_image.num_colors);
if (image.isNull()) {
return false;
}
image.fill(0);
setPalette(xcf_image, image);
}
break;
case INDEXEDA_GIMAGE:
if (xcf_image.num_colors == 1) {
// Plenty(!) of room to add a transparent color
xcf_image.num_colors++;
xcf_image.palette.resize(xcf_image.num_colors);
xcf_image.palette[1] = xcf_image.palette[0];
xcf_image.palette[0] = qRgba(255, 255, 255, 0);
image = imageAlloc(xcf_image.header.width, xcf_image.header.height, QImage::Format_MonoLSB);
image.setColorCount(xcf_image.num_colors);
if (image.isNull()) {
return false;
}
image.fill(0);
setPalette(xcf_image, image);
} else if (xcf_image.num_colors < 256) {
// Plenty of room to add a transparent color
xcf_image.num_colors++;
xcf_image.palette.resize(xcf_image.num_colors);
for (int c = xcf_image.num_colors - 1; c >= 1; c--) {
xcf_image.palette[c] = xcf_image.palette[c - 1];
}
xcf_image.palette[0] = qRgba(255, 255, 255, 0);
image = imageAlloc(xcf_image.header.width, xcf_image.header.height, QImage::Format_Indexed8);
image.setColorCount(xcf_image.num_colors);
if (image.isNull()) {
return false;
}
image.fill(0);
setPalette(xcf_image, image);
} else {
// No room for a transparent color, so this has to be promoted to
// true color. (There is no equivalent PNG representation output
// from The GIMP as of v1.2.)
image = imageAlloc(xcf_image.header.width, xcf_image.header.height, QImage::Format_ARGB32);
if (image.isNull()) {
return false;
}
image.fill(qRgba(255, 255, 255, 0));
}
break;
}
if (image.format() != xcf_image.qimageFormat()) {
qCWarning(XCFPLUGIN) << "Selected wrong format:" << image.format() << "expected" << layer.qimageFormat(xcf_image.header.precision);
return false;
}
// The final profile should be the one in the Parasite
// NOTE: if not set here, the colorSpace is aet in setImageParasites() (if no one defined in the parasites)
#ifndef DISABLE_IMAGE_PROFILE
switch (xcf_image.header.precision) {
case XCFImageFormat::GIMP_PRECISION_HALF_LINEAR:
case XCFImageFormat::GIMP_PRECISION_FLOAT_LINEAR:
case XCFImageFormat::GIMP_PRECISION_DOUBLE_LINEAR:
case XCFImageFormat::GIMP_PRECISION_U8_LINEAR:
case XCFImageFormat::GIMP_PRECISION_U16_LINEAR:
case XCFImageFormat::GIMP_PRECISION_U32_LINEAR:
image.setColorSpace(QColorSpace::SRgbLinear);
break;
case XCFImageFormat::GIMP_PRECISION_HALF_NON_LINEAR:
case XCFImageFormat::GIMP_PRECISION_FLOAT_NON_LINEAR:
case XCFImageFormat::GIMP_PRECISION_DOUBLE_NON_LINEAR:
case XCFImageFormat::GIMP_PRECISION_U8_NON_LINEAR:
case XCFImageFormat::GIMP_PRECISION_U16_NON_LINEAR:
case XCFImageFormat::GIMP_PRECISION_U32_NON_LINEAR:
image.setColorSpace(QColorSpace::SRgb);
break;
case XCFImageFormat::GIMP_PRECISION_HALF_PERCEPTUAL:
case XCFImageFormat::GIMP_PRECISION_FLOAT_PERCEPTUAL:
case XCFImageFormat::GIMP_PRECISION_DOUBLE_PERCEPTUAL:
case XCFImageFormat::GIMP_PRECISION_U8_PERCEPTUAL:
case XCFImageFormat::GIMP_PRECISION_U16_PERCEPTUAL:
case XCFImageFormat::GIMP_PRECISION_U32_PERCEPTUAL:
image.setColorSpace(QColorSpace::SRgb);
break;
}
#endif
if (xcf_image.x_resolution > 0 && xcf_image.y_resolution > 0) {
const float dpmx = xcf_image.x_resolution * INCHESPERMETER;
if (dpmx > float(std::numeric_limits<int>::max())) {
return false;
}
const float dpmy = xcf_image.y_resolution * INCHESPERMETER;
if (dpmy > float(std::numeric_limits<int>::max())) {
return false;
}
image.setDotsPerMeterX((int)dpmx);
image.setDotsPerMeterY((int)dpmy);
}
return true;
}
/*!
* Copy a layer into an image, taking account of the manifold modes. The
* contents of the image are replaced.
* \param xcf_image contains the layer and image to be replaced.
*/
void XCFImageFormat::copyLayerToImage(XCFImage &xcf_image)
{
Layer &layer(xcf_image.layer);
QImage &image(xcf_image.image);
PixelCopyOperation copy = nullptr;
switch (layer.type) {
case RGB_GIMAGE:
case RGBA_GIMAGE:
copy = copyRGBToRGB;
break;
case GRAY_GIMAGE:
if (layer.opacity == OPAQUE_OPACITY) {
copy = copyGrayToGray;
} else {
copy = copyGrayToRGB;
}
break;
case GRAYA_GIMAGE:
copy = copyGrayAToRGB;
break;
case INDEXED_GIMAGE:
copy = copyIndexedToIndexed;
break;
case INDEXEDA_GIMAGE:
if (xcf_image.image.depth() <= 8) {
copy = copyIndexedAToIndexed;
} else {
copy = copyIndexedAToRGB;
}
}
if (!copy) {
return;
}
// For each tile...
for (uint j = 0; j < layer.nrows; j++) {
uint y = j * TILE_HEIGHT;
for (uint i = 0; i < layer.ncols; i++) {
uint x = i * TILE_WIDTH;
// This seems the best place to apply the dissolve because it
// depends on the global position of each tile's
// pixels. Apparently it's the only mode which can apply to a
// single layer.
if (layer.mode == GIMP_LAYER_MODE_DISSOLVE) {
if (!random_table_initialized) {
initializeRandomTable();
random_table_initialized = true;
}
if (layer.type == RGBA_GIMAGE) {
dissolveRGBPixels(layer.image_tiles[j][i], x, y);
}
else if (layer.type == GRAYA_GIMAGE) {
dissolveAlphaPixels(layer.alpha_tiles[j][i], x, y);
}
}
// Shortcut for common case
if (copy == copyRGBToRGB && layer.apply_mask != 1) {
QPainter painter(&image);
painter.setOpacity(layer.opacity / 255.0);
painter.setCompositionMode(QPainter::CompositionMode_Source);
if (x + layer.x_offset < MAX_IMAGE_WIDTH &&
y + layer.y_offset < MAX_IMAGE_HEIGHT) {
painter.drawImage(x + layer.x_offset, y + layer.y_offset, layer.image_tiles[j][i]);
}
continue;
}
for (int l = 0; l < layer.image_tiles[j][i].height(); l++) {
for (int k = 0; k < layer.image_tiles[j][i].width(); k++) {
int m = x + k + layer.x_offset;
int n = y + l + layer.y_offset;
if (m < 0 || m >= image.width() || n < 0 || n >= image.height()) {
continue;
}
(*copy)(layer, i, j, k, l, image, m, n);
}
}
}
}
}
/*!
* Copy an RGB pixel from the layer to the RGB image. Straight-forward.
* The only thing this has to take account of is the opacity of the
* layer. Evidently, the GIMP exporter itself does not actually do this.
* \param layer source layer.
* \param i x tile index.
* \param j y tile index.
* \param k x pixel index of tile i,j.
* \param l y pixel index of tile i,j.
* \param image destination image.
* \param m x pixel of destination image.
* \param n y pixel of destination image.
*/
void XCFImageFormat::copyRGBToRGB(const Layer &layer, uint i, uint j, int k, int l, QImage &image, int m, int n)
{
if (image.depth() == 32) {
QRgb src = layer.image_tiles[j][i].pixel(k, l);
uchar src_a = layer.opacity;
if (layer.type == RGBA_GIMAGE) {
src_a = INT_MULT(src_a, qAlpha(src));
}
// Apply the mask (if any)
if (layer.apply_mask == 1 && layer.mask_tiles.size() > (int)j && layer.mask_tiles[j].size() > (int)i) {
src_a = INT_MULT(src_a, layer.mask_tiles[j][i].pixelIndex(k, l));
}
image.setPixel(m, n, qRgba(src, src_a));
} else if (image.depth() == 64) {
QRgba64 src = layer.image_tiles[j][i].pixelColor(k, l).rgba64();
quint16 src_a = layer.opacity;
if (layer.type == RGBA_GIMAGE) {
src_a = INT_MULT(src_a, qAlpha(src));
}
// Apply the mask (if any)
if (layer.apply_mask == 1 && layer.mask_tiles.size() > (int)j && layer.mask_tiles[j].size() > (int)i) {
src_a = INT_MULT(src_a, layer.mask_tiles[j][i].pixelIndex(k, l));
}
src.setAlpha(src_a);
image.setPixel(m, n, src);
}
}
/*!
* Copy a Gray pixel from the layer to the Gray image. Straight-forward.
* \param layer source layer.
* \param i x tile index.
* \param j y tile index.
* \param k x pixel index of tile i,j.
* \param l y pixel index of tile i,j.
* \param image destination image.
* \param m x pixel of destination image.
* \param n y pixel of destination image.
*/
void XCFImageFormat::copyGrayToGray(const Layer &layer, uint i, uint j, int k, int l, QImage &image, int m, int n)
{
int src = layer.image_tiles[j][i].pixelIndex(k, l);
image.setPixel(m, n, src);
}
/*!
* Copy a Gray pixel from the layer to an RGB image. Straight-forward.
* The only thing this has to take account of is the opacity of the
* layer. Evidently, the GIMP exporter itself does not actually do this.
* \param layer source layer.
* \param i x tile index.
* \param j y tile index.
* \param k x pixel index of tile i,j.
* \param l y pixel index of tile i,j.
* \param image destination image.
* \param m x pixel of destination image.
* \param n y pixel of destination image.
*/
void XCFImageFormat::copyGrayToRGB(const Layer &layer, uint i, uint j, int k, int l, QImage &image, int m, int n)
{
QRgb src = layer.image_tiles[j][i].pixel(k, l);
uchar src_a = layer.opacity;
image.setPixel(m, n, qRgba(src, src_a));
}
/*!
* Copy a GrayA pixel from the layer to an RGB image. Straight-forward.
* The only thing this has to take account of is the opacity of the
* layer. Evidently, the GIMP exporter itself does not actually do this.
* \param layer source layer.
* \param i x tile index.
* \param j y tile index.
* \param k x pixel index of tile i,j.
* \param l y pixel index of tile i,j.
* \param image destination image.
* \param m x pixel of destination image.
* \param n y pixel of destination image.
*/
void XCFImageFormat::copyGrayAToRGB(const Layer &layer, uint i, uint j, int k, int l, QImage &image, int m, int n)
{
QRgb src = layer.image_tiles[j][i].pixel(k, l);
uchar src_a = layer.alpha_tiles[j][i].pixelIndex(k, l);
src_a = INT_MULT(src_a, layer.opacity);
// Apply the mask (if any)
if (layer.apply_mask == 1 && layer.mask_tiles.size() > (int)j && layer.mask_tiles[j].size() > (int)i) {
src_a = INT_MULT(src_a, layer.mask_tiles[j][i].pixelIndex(k, l));
}
image.setPixel(m, n, qRgba(src, src_a));
}
/*!
* Copy an Indexed pixel from the layer to the Indexed image. Straight-forward.
* \param layer source layer.
* \param i x tile index.
* \param j y tile index.
* \param k x pixel index of tile i,j.
* \param l y pixel index of tile i,j.
* \param image destination image.
* \param m x pixel of destination image.
* \param n y pixel of destination image.
*/
void XCFImageFormat::copyIndexedToIndexed(const Layer &layer, uint i, uint j, int k, int l, QImage &image, int m, int n)
{
int src = layer.image_tiles[j][i].pixelIndex(k, l);
image.setPixel(m, n, src);
}
/*!
* Copy an IndexedA pixel from the layer to the Indexed image. Straight-forward.
* \param layer source layer.
* \param i x tile index.
* \param j y tile index.
* \param k x pixel index of tile i,j.
* \param l y pixel index of tile i,j.
* \param image destination image.
* \param m x pixel of destination image.
* \param n y pixel of destination image.
*/
void XCFImageFormat::copyIndexedAToIndexed(const Layer &layer, uint i, uint j, int k, int l, QImage &image, int m, int n)
{
uchar src = layer.image_tiles[j][i].pixelIndex(k, l);
uchar src_a = layer.alpha_tiles[j][i].pixelIndex(k, l);
src_a = INT_MULT(src_a, layer.opacity);
if (layer.apply_mask == 1 && layer.mask_tiles.size() > (int)j && layer.mask_tiles[j].size() > (int)i) {
src_a = INT_MULT(src_a, layer.mask_tiles[j][i].pixelIndex(k, l));
}
if (src_a > 127) {
src++;
} else {
src = 0;
}
image.setPixel(m, n, src);
}
/*!
* Copy an IndexedA pixel from the layer to an RGB image. Straight-forward.
* The only thing this has to take account of is the opacity of the
* layer. Evidently, the GIMP exporter itself does not actually do this.
* \param layer source layer.
* \param i x tile index.
* \param j y tile index.
* \param k x pixel index of tile i,j.
* \param l y pixel index of tile i,j.
* \param image destination image.
* \param m x pixel of destination image.
* \param n y pixel of destination image.
*/
void XCFImageFormat::copyIndexedAToRGB(const Layer &layer, uint i, uint j, int k, int l, QImage &image, int m, int n)
{
QRgb src = layer.image_tiles[j][i].pixel(k, l);
uchar src_a = layer.alpha_tiles[j][i].pixelIndex(k, l);
src_a = INT_MULT(src_a, layer.opacity);
// Apply the mask (if any)
if (layer.apply_mask == 1 && layer.mask_tiles.size() > (int)j && layer.mask_tiles[j].size() > (int)i) {
src_a = INT_MULT(src_a, layer.mask_tiles[j][i].pixelIndex(k, l));
}
// This is what appears in the GIMP window
if (src_a <= 127) {
src_a = 0;
} else {
src_a = OPAQUE_OPACITY;
}
image.setPixel(m, n, qRgba(src, src_a));
}
/*!
* Merge a layer into an image, taking account of the manifold modes.
* \param xcf_image contains the layer and image to merge.
*/
void XCFImageFormat::mergeLayerIntoImage(XCFImage &xcf_image)
{
Layer &layer(xcf_image.layer);
QImage &image(xcf_image.image);
PixelMergeOperation merge = nullptr;
if (!layer.opacity) {
return; // don't bother doing anything
}
if (layer.blendSpace == XCFImageFormat::AutoColorSpace) {
qCDebug(XCFPLUGIN) << "Auto blend space, defaulting to RgbLinearSpace (same as Gimp when writing this)";
layer.blendSpace = XCFImageFormat::RgbLinearSpace;
}
if (layer.blendSpace != XCFImageFormat::RgbLinearSpace) {
qCDebug(XCFPLUGIN) << "Unimplemented blend color space" << layer.blendSpace;
}
qCDebug(XCFPLUGIN) << "Blend color space" << layer.blendSpace;
if (layer.compositeSpace == XCFImageFormat::AutoColorSpace) {
qCDebug(XCFPLUGIN) << "Auto composite space, defaulting to RgbLinearSpace (same as Gimp when writing this)";
layer.compositeSpace = XCFImageFormat::RgbLinearSpace;
}
if (layer.compositeSpace != XCFImageFormat::RgbLinearSpace) {
qCDebug(XCFPLUGIN) << "Unimplemented composite color space" << layer.compositeSpace;
}
if (layer.compositeMode != XCFImageFormat::CompositeUnion) {
qCDebug(XCFPLUGIN) << "Unhandled composite mode" << layer.compositeMode;
}
switch (layer.type) {
case RGB_GIMAGE:
case RGBA_GIMAGE:
merge = mergeRGBToRGB;
break;
case GRAY_GIMAGE:
if (layer.opacity == OPAQUE_OPACITY) {
merge = mergeGrayToGray;
} else {
merge = mergeGrayToRGB;
}
break;
case GRAYA_GIMAGE:
if (xcf_image.image.depth() <= 8) {
merge = mergeGrayAToGray;
} else {
merge = mergeGrayAToRGB;
}
break;
case INDEXED_GIMAGE:
merge = mergeIndexedToIndexed;
break;
case INDEXEDA_GIMAGE:
if (xcf_image.image.depth() <= 8) {
merge = mergeIndexedAToIndexed;
} else {
merge = mergeIndexedAToRGB;
}
}
if (!merge) {
return;
}
if (merge == mergeRGBToRGB && layer.apply_mask != 1) {
int painterMode = -1;
switch (layer.mode) {
case GIMP_LAYER_MODE_NORMAL:
case GIMP_LAYER_MODE_NORMAL_LEGACY:
painterMode = QPainter::CompositionMode_SourceOver;
break;
case GIMP_LAYER_MODE_MULTIPLY:
case GIMP_LAYER_MODE_MULTIPLY_LEGACY:
painterMode = QPainter::CompositionMode_Multiply;
break;
case GIMP_LAYER_MODE_SCREEN:
case GIMP_LAYER_MODE_SCREEN_LEGACY:
painterMode = QPainter::CompositionMode_Screen;
break;
case GIMP_LAYER_MODE_OVERLAY:
case GIMP_LAYER_MODE_OVERLAY_LEGACY:
painterMode = QPainter::CompositionMode_Overlay;
break;
case GIMP_LAYER_MODE_DIFFERENCE:
case GIMP_LAYER_MODE_DIFFERENCE_LEGACY:
painterMode = QPainter::CompositionMode_Difference;
break;
case GIMP_LAYER_MODE_DARKEN_ONLY:
case GIMP_LAYER_MODE_DARKEN_ONLY_LEGACY:
painterMode = QPainter::CompositionMode_Darken;
break;
case GIMP_LAYER_MODE_LIGHTEN_ONLY:
case GIMP_LAYER_MODE_LIGHTEN_ONLY_LEGACY:
painterMode = QPainter::CompositionMode_Lighten;
break;
case GIMP_LAYER_MODE_DODGE:
case GIMP_LAYER_MODE_DODGE_LEGACY:
painterMode = QPainter::CompositionMode_ColorDodge;
break;
case GIMP_LAYER_MODE_BURN:
case GIMP_LAYER_MODE_BURN_LEGACY:
painterMode = QPainter::CompositionMode_ColorBurn;
break;
case GIMP_LAYER_MODE_HARDLIGHT:
case GIMP_LAYER_MODE_HARDLIGHT_LEGACY:
painterMode = QPainter::CompositionMode_HardLight;
break;
case GIMP_LAYER_MODE_SOFTLIGHT:
case GIMP_LAYER_MODE_SOFTLIGHT_LEGACY:
painterMode = QPainter::CompositionMode_SoftLight;
break;
case GIMP_LAYER_MODE_ADDITION:
case GIMP_LAYER_MODE_ADDITION_LEGACY:
painterMode = QPainter::CompositionMode_Plus;
break;
case GIMP_LAYER_MODE_EXCLUSION:
painterMode = QPainter::CompositionMode_Exclusion;
break;
// Not bothered to find what the QPainter equivalent is, or there is none
case GIMP_LAYER_MODE_GRAIN_EXTRACT:
case GIMP_LAYER_MODE_GRAIN_EXTRACT_LEGACY:
case GIMP_LAYER_MODE_GRAIN_MERGE:
case GIMP_LAYER_MODE_GRAIN_MERGE_LEGACY:
case GIMP_LAYER_MODE_COLOR_ERASE:
case GIMP_LAYER_MODE_COLOR_ERASE_LEGACY:
case GIMP_LAYER_MODE_LCH_HUE:
case GIMP_LAYER_MODE_LCH_CHROMA:
case GIMP_LAYER_MODE_LCH_COLOR:
case GIMP_LAYER_MODE_LCH_LIGHTNESS:
case GIMP_LAYER_MODE_BEHIND:
case GIMP_LAYER_MODE_BEHIND_LEGACY:
case GIMP_LAYER_MODE_SUBTRACT:
case GIMP_LAYER_MODE_SUBTRACT_LEGACY:
case GIMP_LAYER_MODE_HSV_HUE:
case GIMP_LAYER_MODE_HSV_SATURATION:
case GIMP_LAYER_MODE_HSL_COLOR:
case GIMP_LAYER_MODE_HSV_VALUE:
case GIMP_LAYER_MODE_DIVIDE:
case GIMP_LAYER_MODE_VIVID_LIGHT:
case GIMP_LAYER_MODE_PIN_LIGHT:
case GIMP_LAYER_MODE_LINEAR_LIGHT:
case GIMP_LAYER_MODE_HARD_MIX:
case GIMP_LAYER_MODE_LINEAR_BURN:
case GIMP_LAYER_MODE_LUMA_DARKEN_ONLY:
case GIMP_LAYER_MODE_LUMA_LIGHTEN_ONLY:
case GIMP_LAYER_MODE_LUMINANCE:
case GIMP_LAYER_MODE_ERASE:
case GIMP_LAYER_MODE_MERGE:
case GIMP_LAYER_MODE_SPLIT:
case GIMP_LAYER_MODE_PASS_THROUGH:
case GIMP_LAYER_MODE_HSV_HUE_LEGACY:
case GIMP_LAYER_MODE_HSV_SATURATION_LEGACY:
case GIMP_LAYER_MODE_HSL_COLOR_LEGACY:
case GIMP_LAYER_MODE_HSV_VALUE_LEGACY:
case GIMP_LAYER_MODE_DIVIDE_LEGACY:
qCDebug(XCFPLUGIN) << "No QPainter equivalent to" << layer.mode;
break;
// Special
case GIMP_LAYER_MODE_DISSOLVE:
case GIMP_LAYER_MODE_COUNT:
break;
}
if (painterMode != -1) {
QPainter painter(&image);
painter.setOpacity(layer.opacity / 255.0);
painter.setCompositionMode(QPainter::CompositionMode(painterMode));
qCDebug(XCFPLUGIN) << "Using QPainter for mode" << layer.mode;
for (uint j = 0; j < layer.nrows; j++) {
uint y = j * TILE_HEIGHT;
for (uint i = 0; i < layer.ncols; i++) {
uint x = i * TILE_WIDTH;
QImage &tile = layer.image_tiles[j][i];
if (x + layer.x_offset < MAX_IMAGE_WIDTH &&
y + layer.y_offset < MAX_IMAGE_HEIGHT) {
painter.drawImage(x + layer.x_offset, y + layer.y_offset, tile);
}
}
}
return;
}
}
#ifndef DISABLE_IMAGE_PROFILE_CONV // The final profile should be the one in the Parasite
if (layer.compositeSpace == XCFImageFormat::RgbPerceptualSpace && image.colorSpace() != QColorSpace::SRgb) {
qCDebug(XCFPLUGIN) << "Converting to composite color space" << layer.compositeSpace;
image.convertToColorSpace(QColorSpace::SRgb);
}
if (layer.compositeSpace == XCFImageFormat::RgbLinearSpace && image.colorSpace() != QColorSpace::SRgbLinear) {
qCDebug(XCFPLUGIN) << "Converting to composite color space" << layer.compositeSpace;
image.convertToColorSpace(QColorSpace::SRgbLinear);
}
#endif
for (uint j = 0; j < layer.nrows; j++) {
uint y = j * TILE_HEIGHT;
for (uint i = 0; i < layer.ncols; i++) {
uint x = i * TILE_WIDTH;
// This seems the best place to apply the dissolve because it
// depends on the global position of each tile's
// pixels. Apparently it's the only mode which can apply to a
// single layer.
if (layer.mode == GIMP_LAYER_MODE_DISSOLVE) {
if (!random_table_initialized) {
initializeRandomTable();
random_table_initialized = true;
}
if (layer.type == RGBA_GIMAGE) {
dissolveRGBPixels(layer.image_tiles[j][i], x, y);
}
else if (layer.type == GRAYA_GIMAGE) {
dissolveAlphaPixels(layer.alpha_tiles[j][i], x, y);
}
}
// Shortcut for common case
if (merge == mergeRGBToRGB && layer.apply_mask != 1 && layer.mode == GIMP_LAYER_MODE_NORMAL_LEGACY) {
QPainter painter(&image);
painter.setOpacity(layer.opacity / 255.0);
painter.setCompositionMode(QPainter::CompositionMode_SourceOver);
if (x + layer.x_offset < MAX_IMAGE_WIDTH &&
y + layer.y_offset < MAX_IMAGE_HEIGHT) {
painter.drawImage(x + layer.x_offset, y + layer.y_offset, layer.image_tiles[j][i]);
}
continue;
}
#ifndef DISABLE_TILE_PROFILE_CONV // not sure about that: left as old plugin
QImage &tile = layer.image_tiles[j][i];
if (layer.compositeSpace == XCFImageFormat::RgbPerceptualSpace && tile.colorSpace() != QColorSpace::SRgb) {
tile.convertToColorSpace(QColorSpace::SRgb);
}
if (layer.compositeSpace == XCFImageFormat::RgbLinearSpace && tile.colorSpace() != QColorSpace::SRgbLinear) {
tile.convertToColorSpace(QColorSpace::SRgbLinear);
}
#endif
for (int l = 0; l < layer.image_tiles[j][i].height(); l++) {
for (int k = 0; k < layer.image_tiles[j][i].width(); k++) {
int m = x + k + layer.x_offset;
int n = y + l + layer.y_offset;
if (m < 0 || m >= image.width() || n < 0 || n >= image.height()) {
continue;
}
if (!(*merge)(layer, i, j, k, l, image, m, n)) {
return;
}
}
}
}
}
}
/*!
* Merge an RGB pixel from the layer to the RGB image. Straight-forward.
* The only thing this has to take account of is the opacity of the
* layer. Evidently, the GIMP exporter itself does not actually do this.
* \param layer source layer.
* \param i x tile index.
* \param j y tile index.
* \param k x pixel index of tile i,j.
* \param l y pixel index of tile i,j.
* \param image destination image.
* \param m x pixel of destination image.
* \param n y pixel of destination image.
*/
bool XCFImageFormat::mergeRGBToRGB(const Layer &layer, uint i, uint j, int k, int l, QImage &image, int m, int n)
{
QRgb src = layer.image_tiles[j][i].pixel(k, l);
QRgb dst = image.pixel(m, n);
uchar src_r = qRed(src);
uchar src_g = qGreen(src);
uchar src_b = qBlue(src);
uchar src_a = qAlpha(src);
uchar dst_r = qRed(dst);
uchar dst_g = qGreen(dst);
uchar dst_b = qBlue(dst);
uchar dst_a = qAlpha(dst);
if (!src_a) {
return false; // nothing to merge
}
switch (layer.mode) {
case GIMP_LAYER_MODE_NORMAL:
case GIMP_LAYER_MODE_NORMAL_LEGACY:
break;
case GIMP_LAYER_MODE_MULTIPLY:
case GIMP_LAYER_MODE_MULTIPLY_LEGACY:
src_r = INT_MULT(src_r, dst_r);
src_g = INT_MULT(src_g, dst_g);
src_b = INT_MULT(src_b, dst_b);
src_a = qMin(src_a, dst_a);
break;
case GIMP_LAYER_MODE_DIVIDE:
case GIMP_LAYER_MODE_DIVIDE_LEGACY:
src_r = qMin((dst_r * 256) / (1 + src_r), 255);
src_g = qMin((dst_g * 256) / (1 + src_g), 255);
src_b = qMin((dst_b * 256) / (1 + src_b), 255);
src_a = qMin(src_a, dst_a);
break;
case GIMP_LAYER_MODE_SCREEN:
case GIMP_LAYER_MODE_SCREEN_LEGACY:
src_r = 255 - INT_MULT(255 - dst_r, 255 - src_r);
src_g = 255 - INT_MULT(255 - dst_g, 255 - src_g);
src_b = 255 - INT_MULT(255 - dst_b, 255 - src_b);
src_a = qMin(src_a, dst_a);
break;
case GIMP_LAYER_MODE_OVERLAY:
case GIMP_LAYER_MODE_OVERLAY_LEGACY:
src_r = INT_MULT(dst_r, dst_r + INT_MULT(2 * src_r, 255 - dst_r));
src_g = INT_MULT(dst_g, dst_g + INT_MULT(2 * src_g, 255 - dst_g));
src_b = INT_MULT(dst_b, dst_b + INT_MULT(2 * src_b, 255 - dst_b));
src_a = qMin(src_a, dst_a);
break;
case GIMP_LAYER_MODE_DIFFERENCE:
case GIMP_LAYER_MODE_DIFFERENCE_LEGACY:
src_r = dst_r > src_r ? dst_r - src_r : src_r - dst_r;
src_g = dst_g > src_g ? dst_g - src_g : src_g - dst_g;
src_b = dst_b > src_b ? dst_b - src_b : src_b - dst_b;
src_a = qMin(src_a, dst_a);
break;
case GIMP_LAYER_MODE_ADDITION:
case GIMP_LAYER_MODE_ADDITION_LEGACY:
src_r = add_lut(dst_r, src_r);
src_g = add_lut(dst_g, src_g);
src_b = add_lut(dst_b, src_b);
src_a = qMin(src_a, dst_a);
break;
case GIMP_LAYER_MODE_SUBTRACT:
case GIMP_LAYER_MODE_SUBTRACT_LEGACY:
src_r = dst_r > src_r ? dst_r - src_r : 0;
src_g = dst_g > src_g ? dst_g - src_g : 0;
src_b = dst_b > src_b ? dst_b - src_b : 0;
src_a = qMin(src_a, dst_a);
break;
case GIMP_LAYER_MODE_DARKEN_ONLY:
case GIMP_LAYER_MODE_DARKEN_ONLY_LEGACY:
src_r = dst_r < src_r ? dst_r : src_r;
src_g = dst_g < src_g ? dst_g : src_g;
src_b = dst_b < src_b ? dst_b : src_b;
src_a = qMin(src_a, dst_a);
break;
case GIMP_LAYER_MODE_LIGHTEN_ONLY:
case GIMP_LAYER_MODE_LIGHTEN_ONLY_LEGACY:
src_r = dst_r < src_r ? src_r : dst_r;
src_g = dst_g < src_g ? src_g : dst_g;
src_b = dst_b < src_b ? src_b : dst_b;
src_a = qMin(src_a, dst_a);
break;
case GIMP_LAYER_MODE_HSV_HUE:
case GIMP_LAYER_MODE_HSV_HUE_LEGACY: {
uchar new_r = dst_r;
uchar new_g = dst_g;
uchar new_b = dst_b;
RGBTOHSV(src_r, src_g, src_b);
RGBTOHSV(new_r, new_g, new_b);
new_r = src_r;
HSVTORGB(new_r, new_g, new_b);
src_r = new_r;
src_g = new_g;
src_b = new_b;
src_a = qMin(src_a, dst_a);
} break;
case GIMP_LAYER_MODE_HSV_SATURATION:
case GIMP_LAYER_MODE_HSV_SATURATION_LEGACY: {
uchar new_r = dst_r;
uchar new_g = dst_g;
uchar new_b = dst_b;
RGBTOHSV(src_r, src_g, src_b);
RGBTOHSV(new_r, new_g, new_b);
new_g = src_g;
HSVTORGB(new_r, new_g, new_b);
src_r = new_r;
src_g = new_g;
src_b = new_b;
src_a = qMin(src_a, dst_a);
} break;
case GIMP_LAYER_MODE_HSV_VALUE:
case GIMP_LAYER_MODE_HSV_VALUE_LEGACY: {
uchar new_r = dst_r;
uchar new_g = dst_g;
uchar new_b = dst_b;
RGBTOHSV(src_r, src_g, src_b);
RGBTOHSV(new_r, new_g, new_b);
new_b = src_b;
HSVTORGB(new_r, new_g, new_b);
src_r = new_r;
src_g = new_g;
src_b = new_b;
src_a = qMin(src_a, dst_a);
} break;
case GIMP_LAYER_MODE_HSL_COLOR:
case GIMP_LAYER_MODE_HSL_COLOR_LEGACY: {
uchar new_r = dst_r;
uchar new_g = dst_g;
uchar new_b = dst_b;
RGBTOHLS(src_r, src_g, src_b);
RGBTOHLS(new_r, new_g, new_b);
new_r = src_r;
new_b = src_b;
HLSTORGB(new_r, new_g, new_b);
src_r = new_r;
src_g = new_g;
src_b = new_b;
src_a = qMin(src_a, dst_a);
} break;
case GIMP_LAYER_MODE_DODGE:
case GIMP_LAYER_MODE_DODGE_LEGACY: {
uint tmp;
tmp = dst_r << 8;
tmp /= 256 - src_r;
src_r = (uchar)qMin(tmp, 255u);
tmp = dst_g << 8;
tmp /= 256 - src_g;
src_g = (uchar)qMin(tmp, 255u);
tmp = dst_b << 8;
tmp /= 256 - src_b;
src_b = (uchar)qMin(tmp, 255u);
src_a = qMin(src_a, dst_a);
} break;
case GIMP_LAYER_MODE_BURN:
case GIMP_LAYER_MODE_BURN_LEGACY: {
uint tmp;
tmp = (255 - dst_r) << 8;
tmp /= src_r + 1;
src_r = (uchar)qMin(tmp, 255u);
src_r = 255 - src_r;
tmp = (255 - dst_g) << 8;
tmp /= src_g + 1;
src_g = (uchar)qMin(tmp, 255u);
src_g = 255 - src_g;
tmp = (255 - dst_b) << 8;
tmp /= src_b + 1;
src_b = (uchar)qMin(tmp, 255u);
src_b = 255 - src_b;
src_a = qMin(src_a, dst_a);
} break;
case GIMP_LAYER_MODE_HARDLIGHT:
case GIMP_LAYER_MODE_HARDLIGHT_LEGACY: {
uint tmp;
if (src_r > 128) {
tmp = ((int)255 - dst_r) * ((int)255 - ((src_r - 128) << 1));
src_r = (uchar)qMin(255 - (tmp >> 8), 255u);
} else {
tmp = (int)dst_r * ((int)src_r << 1);
src_r = (uchar)qMin(tmp >> 8, 255u);
}
if (src_g > 128) {
tmp = ((int)255 - dst_g) * ((int)255 - ((src_g - 128) << 1));
src_g = (uchar)qMin(255 - (tmp >> 8), 255u);
} else {
tmp = (int)dst_g * ((int)src_g << 1);
src_g = (uchar)qMin(tmp >> 8, 255u);
}
if (src_b > 128) {
tmp = ((int)255 - dst_b) * ((int)255 - ((src_b - 128) << 1));
src_b = (uchar)qMin(255 - (tmp >> 8), 255u);
} else {
tmp = (int)dst_b * ((int)src_b << 1);
src_b = (uchar)qMin(tmp >> 8, 255u);
}
src_a = qMin(src_a, dst_a);
} break;
case GIMP_LAYER_MODE_SOFTLIGHT:
case GIMP_LAYER_MODE_SOFTLIGHT_LEGACY: {
uint tmpS;
uint tmpM;
tmpM = INT_MULT(dst_r, src_r);
tmpS = 255 - INT_MULT((255 - dst_r), (255 - src_r));
src_r = INT_MULT((255 - dst_r), tmpM) + INT_MULT(dst_r, tmpS);
tmpM = INT_MULT(dst_g, src_g);
tmpS = 255 - INT_MULT((255 - dst_g), (255 - src_g));
src_g = INT_MULT((255 - dst_g), tmpM) + INT_MULT(dst_g, tmpS);
tmpM = INT_MULT(dst_b, src_b);
tmpS = 255 - INT_MULT((255 - dst_b), (255 - src_b));
src_b = INT_MULT((255 - dst_b), tmpM) + INT_MULT(dst_b, tmpS);
src_a = qMin(src_a, dst_a);
} break;
case GIMP_LAYER_MODE_GRAIN_EXTRACT:
case GIMP_LAYER_MODE_GRAIN_EXTRACT_LEGACY: {
int tmp;
tmp = dst_r - src_r + 128;
tmp = qMin(tmp, 255);
tmp = qMax(tmp, 0);
src_r = (uchar)tmp;
tmp = dst_g - src_g + 128;
tmp = qMin(tmp, 255);
tmp = qMax(tmp, 0);
src_g = (uchar)tmp;
tmp = dst_b - src_b + 128;
tmp = qMin(tmp, 255);
tmp = qMax(tmp, 0);
src_b = (uchar)tmp;
src_a = qMin(src_a, dst_a);
} break;
case GIMP_LAYER_MODE_GRAIN_MERGE:
case GIMP_LAYER_MODE_GRAIN_MERGE_LEGACY: {
int tmp;
tmp = dst_r + src_r - 128;
tmp = qMin(tmp, 255);
tmp = qMax(tmp, 0);
src_r = (uchar)tmp;
tmp = dst_g + src_g - 128;
tmp = qMin(tmp, 255);
tmp = qMax(tmp, 0);
src_g = (uchar)tmp;
tmp = dst_b + src_b - 128;
tmp = qMin(tmp, 255);
tmp = qMax(tmp, 0);
src_b = (uchar)tmp;
src_a = qMin(src_a, dst_a);
} break;
case GIMP_LAYER_MODE_LINEAR_LIGHT: {
if (src_r <= 128) {
src_r = qBound(0, dst_r + 2 * src_r - 255, 255);
} else {
src_r = qBound(0, dst_r + 2 * (src_r - 128), 255);
}
if (src_g <= 128) {
src_g = qBound(0, dst_g + 2 * src_g - 255, 255);
} else {
src_g = qBound(0, dst_g + 2 * (src_g - 127), 255);
}
if (src_b <= 128) {
src_b = qBound(0, dst_b + 2 * src_b - 255, 255);
} else {
src_b = qBound(0, dst_b + 2 * (src_b - 127), 255);
}
} break;
case GIMP_LAYER_MODE_VIVID_LIGHT: {
// From http://www.simplefilter.de/en/basics/mixmods.html
float A[3];
A[0] = src_r / 255.;
A[1] = src_g / 255.;
A[2] = src_b / 255.;
float B[3];
B[0] = dst_r / 255.;
B[1] = dst_g / 255.;
B[2] = dst_b / 255.;
float C[3]{};
for (int i = 0; i < 3; i++) {
if (A[i] <= 0.5f) {
if (A[i] > 0.f) {
C[i] = 1.f - (1.f - B[i]) / (2.f * A[i]);
}
} else {
if (A[i] < 1.f) {
C[i] = B[i] / (2.f * (1.f - A[i]));
}
}
}
src_r = qBound(0.f, C[0] * 255.f, 255.f);
src_g = qBound(0.f, C[1] * 255.f, 255.f);
src_b = qBound(0.f, C[2] * 255.f, 255.f);
} break;
default:
qCWarning(XCFPLUGIN) << "Unhandled mode" << layer.mode;
return false;
}
src_a = INT_MULT(src_a, layer.opacity);
// Apply the mask (if any)
if (layer.apply_mask == 1 && layer.mask_tiles.size() > (int)j && layer.mask_tiles[j].size() > (int)i) {
src_a = INT_MULT(src_a, layer.mask_tiles[j][i].pixelIndex(k, l));
}
uchar new_r;
uchar new_g;
uchar new_b;
uchar new_a;
new_a = dst_a + INT_MULT(OPAQUE_OPACITY - dst_a, src_a);
const float src_ratio = new_a == 0 ? 1.0 : (float)src_a / new_a;
float dst_ratio = 1.0 - src_ratio;
new_r = (uchar)(src_ratio * src_r + dst_ratio * dst_r + EPSILON);
new_g = (uchar)(src_ratio * src_g + dst_ratio * dst_g + EPSILON);
new_b = (uchar)(src_ratio * src_b + dst_ratio * dst_b + EPSILON);
if (!modeAffectsSourceAlpha(layer.mode)) {
new_a = dst_a;
}
image.setPixel(m, n, qRgba(new_r, new_g, new_b, new_a));
return true;
}
/*!
* Merge a Gray pixel from the layer to the Gray image. Straight-forward.
* \param layer source layer.
* \param i x tile index.
* \param j y tile index.
* \param k x pixel index of tile i,j.
* \param l y pixel index of tile i,j.
* \param image destination image.
* \param m x pixel of destination image.
* \param n y pixel of destination image.
*/
bool XCFImageFormat::mergeGrayToGray(const Layer &layer, uint i, uint j, int k, int l, QImage &image, int m, int n)
{
int src = layer.image_tiles[j][i].pixelIndex(k, l);
image.setPixel(m, n, src);
return true;
}
/*!
* Merge a GrayA pixel from the layer to the Gray image. Straight-forward.
* \param layer source layer.
* \param i x tile index.
* \param j y tile index.
* \param k x pixel index of tile i,j.
* \param l y pixel index of tile i,j.
* \param image destination image.
* \param m x pixel of destination image.
* \param n y pixel of destination image.
*/
bool XCFImageFormat::mergeGrayAToGray(const Layer &layer, uint i, uint j, int k, int l, QImage &image, int m, int n)
{
int src = qGray(layer.image_tiles[j][i].pixel(k, l));
int dst = image.pixelIndex(m, n);
uchar src_a = layer.alpha_tiles[j][i].pixelIndex(k, l);
if (!src_a) {
return false; // nothing to merge
}
switch (layer.mode) {
case GIMP_LAYER_MODE_MULTIPLY:
case GIMP_LAYER_MODE_MULTIPLY_LEGACY: {
src = INT_MULT(src, dst);
} break;
case GIMP_LAYER_MODE_DIVIDE:
case GIMP_LAYER_MODE_DIVIDE_LEGACY: {
src = qMin((dst * 256) / (1 + src), 255);
} break;
case GIMP_LAYER_MODE_SCREEN:
case GIMP_LAYER_MODE_SCREEN_LEGACY: {
src = 255 - INT_MULT(255 - dst, 255 - src);
} break;
case GIMP_LAYER_MODE_OVERLAY:
case GIMP_LAYER_MODE_OVERLAY_LEGACY: {
src = INT_MULT(dst, dst + INT_MULT(2 * src, 255 - dst));
} break;
case GIMP_LAYER_MODE_DIFFERENCE:
case GIMP_LAYER_MODE_DIFFERENCE_LEGACY: {
src = dst > src ? dst - src : src - dst;
} break;
case GIMP_LAYER_MODE_ADDITION:
case GIMP_LAYER_MODE_ADDITION_LEGACY: {
src = add_lut(dst, src);
} break;
case GIMP_LAYER_MODE_SUBTRACT:
case GIMP_LAYER_MODE_SUBTRACT_LEGACY: {
src = dst > src ? dst - src : 0;
} break;
case GIMP_LAYER_MODE_DARKEN_ONLY:
case GIMP_LAYER_MODE_DARKEN_ONLY_LEGACY: {
src = dst < src ? dst : src;
} break;
case GIMP_LAYER_MODE_LIGHTEN_ONLY:
case GIMP_LAYER_MODE_LIGHTEN_ONLY_LEGACY: {
src = dst < src ? src : dst;
} break;
case GIMP_LAYER_MODE_DODGE:
case GIMP_LAYER_MODE_DODGE_LEGACY: {
uint tmp = dst << 8;
tmp /= 256 - src;
src = (uchar)qMin(tmp, 255u);
} break;
case GIMP_LAYER_MODE_BURN:
case GIMP_LAYER_MODE_BURN_LEGACY: {
uint tmp = (255 - dst) << 8;
tmp /= src + 1;
src = (uchar)qMin(tmp, 255u);
src = 255 - src;
} break;
case GIMP_LAYER_MODE_HARDLIGHT:
case GIMP_LAYER_MODE_HARDLIGHT_LEGACY: {
uint tmp;
if (src > 128) {
tmp = ((int)255 - dst) * ((int)255 - ((src - 128) << 1));
src = (uchar)qMin(255 - (tmp >> 8), 255u);
} else {
tmp = (int)dst * ((int)src << 1);
src = (uchar)qMin(tmp >> 8, 255u);
}
} break;
case GIMP_LAYER_MODE_SOFTLIGHT:
case GIMP_LAYER_MODE_SOFTLIGHT_LEGACY: {
uint tmpS;
uint tmpM;
tmpM = INT_MULT(dst, src);
tmpS = 255 - INT_MULT((255 - dst), (255 - src));
src = INT_MULT((255 - dst), tmpM) + INT_MULT(dst, tmpS);
} break;
case GIMP_LAYER_MODE_GRAIN_EXTRACT:
case GIMP_LAYER_MODE_GRAIN_EXTRACT_LEGACY: {
int tmp;
tmp = dst - src + 128;
tmp = qMin(tmp, 255);
tmp = qMax(tmp, 0);
src = (uchar)tmp;
} break;
case GIMP_LAYER_MODE_GRAIN_MERGE:
case GIMP_LAYER_MODE_GRAIN_MERGE_LEGACY: {
int tmp;
tmp = dst + src - 128;
tmp = qMin(tmp, 255);
tmp = qMax(tmp, 0);
src = (uchar)tmp;
} break;
default:
qCWarning(XCFPLUGIN) << "Unhandled mode" << layer.mode;
return false;
}
src_a = INT_MULT(src_a, layer.opacity);
// Apply the mask (if any)
if (layer.apply_mask == 1 && layer.mask_tiles.size() > (int)j && layer.mask_tiles[j].size() > (int)i) {
src_a = INT_MULT(src_a, layer.mask_tiles[j][i].pixelIndex(k, l));
}
uchar new_a = OPAQUE_OPACITY;
const float src_ratio = new_a == 0 ? 1.0 : (float)src_a / new_a;
float dst_ratio = 1.0 - src_ratio;
uchar new_g = (uchar)(src_ratio * src + dst_ratio * dst + EPSILON);
image.setPixel(m, n, new_g);
return true;
}
/*!
* Merge a Gray pixel from the layer to an RGB image. Straight-forward.
* The only thing this has to take account of is the opacity of the
* layer. Evidently, the GIMP exporter itself does not actually do this.
* \param layer source layer.
* \param i x tile index.
* \param j y tile index.
* \param k x pixel index of tile i,j.
* \param l y pixel index of tile i,j.
* \param image destination image.
* \param m x pixel of destination image.
* \param n y pixel of destination image.
*/
bool XCFImageFormat::mergeGrayToRGB(const Layer &layer, uint i, uint j, int k, int l, QImage &image, int m, int n)
{
QRgb src = layer.image_tiles[j][i].pixel(k, l);
uchar src_a = layer.opacity;
image.setPixel(m, n, qRgba(src, src_a));
return true;
}
/*!
* Merge a GrayA pixel from the layer to an RGB image. Straight-forward.
* The only thing this has to take account of is the opacity of the
* layer. Evidently, the GIMP exporter itself does not actually do this.
* \param layer source layer.
* \param i x tile index.
* \param j y tile index.
* \param k x pixel index of tile i,j.
* \param l y pixel index of tile i,j.
* \param image destination image.
* \param m x pixel of destination image.
* \param n y pixel of destination image.
*/
bool XCFImageFormat::mergeGrayAToRGB(const Layer &layer, uint i, uint j, int k, int l, QImage &image, int m, int n)
{
int src = qGray(layer.image_tiles[j][i].pixel(k, l));
int dst = qGray(image.pixel(m, n));
uchar src_a = layer.alpha_tiles[j][i].pixelIndex(k, l);
uchar dst_a = qAlpha(image.pixel(m, n));
if (!src_a) {
return false; // nothing to merge
}
switch (layer.mode) {
case GIMP_LAYER_MODE_MULTIPLY:
case GIMP_LAYER_MODE_MULTIPLY_LEGACY: {
src = INT_MULT(src, dst);
src_a = qMin(src_a, dst_a);
} break;
case GIMP_LAYER_MODE_DIVIDE:
case GIMP_LAYER_MODE_DIVIDE_LEGACY: {
src = qMin((dst * 256) / (1 + src), 255);
src_a = qMin(src_a, dst_a);
} break;
case GIMP_LAYER_MODE_SCREEN:
case GIMP_LAYER_MODE_SCREEN_LEGACY: {
src = 255 - INT_MULT(255 - dst, 255 - src);
src_a = qMin(src_a, dst_a);
} break;
case GIMP_LAYER_MODE_OVERLAY:
case GIMP_LAYER_MODE_OVERLAY_LEGACY: {
src = INT_MULT(dst, dst + INT_MULT(2 * src, 255 - dst));
src_a = qMin(src_a, dst_a);
} break;
case GIMP_LAYER_MODE_DIFFERENCE:
case GIMP_LAYER_MODE_DIFFERENCE_LEGACY: {
src = dst > src ? dst - src : src - dst;
src_a = qMin(src_a, dst_a);
} break;
case GIMP_LAYER_MODE_ADDITION:
case GIMP_LAYER_MODE_ADDITION_LEGACY: {
src = add_lut(dst, src);
src_a = qMin(src_a, dst_a);
} break;
case GIMP_LAYER_MODE_SUBTRACT:
case GIMP_LAYER_MODE_SUBTRACT_LEGACY: {
src = dst > src ? dst - src : 0;
src_a = qMin(src_a, dst_a);
} break;
case GIMP_LAYER_MODE_DARKEN_ONLY:
case GIMP_LAYER_MODE_DARKEN_ONLY_LEGACY: {
src = dst < src ? dst : src;
src_a = qMin(src_a, dst_a);
} break;
case GIMP_LAYER_MODE_LIGHTEN_ONLY:
case GIMP_LAYER_MODE_LIGHTEN_ONLY_LEGACY: {
src = dst < src ? src : dst;
src_a = qMin(src_a, dst_a);
} break;
case GIMP_LAYER_MODE_DODGE:
case GIMP_LAYER_MODE_DODGE_LEGACY: {
uint tmp = dst << 8;
tmp /= 256 - src;
src = (uchar)qMin(tmp, 255u);
src_a = qMin(src_a, dst_a);
} break;
case GIMP_LAYER_MODE_BURN:
case GIMP_LAYER_MODE_BURN_LEGACY: {
uint tmp = (255 - dst) << 8;
tmp /= src + 1;
src = (uchar)qMin(tmp, 255u);
src = 255 - src;
src_a = qMin(src_a, dst_a);
} break;
case GIMP_LAYER_MODE_HARDLIGHT:
case GIMP_LAYER_MODE_HARDLIGHT_LEGACY: {
uint tmp;
if (src > 128) {
tmp = ((int)255 - dst) * ((int)255 - ((src - 128) << 1));
src = (uchar)qMin(255 - (tmp >> 8), 255u);
} else {
tmp = (int)dst * ((int)src << 1);
src = (uchar)qMin(tmp >> 8, 255u);
}
src_a = qMin(src_a, dst_a);
} break;
case GIMP_LAYER_MODE_SOFTLIGHT:
case GIMP_LAYER_MODE_SOFTLIGHT_LEGACY: {
uint tmpS;
uint tmpM;
tmpM = INT_MULT(dst, src);
tmpS = 255 - INT_MULT((255 - dst), (255 - src));
src = INT_MULT((255 - dst), tmpM) + INT_MULT(dst, tmpS);
src_a = qMin(src_a, dst_a);
} break;
case GIMP_LAYER_MODE_GRAIN_EXTRACT:
case GIMP_LAYER_MODE_GRAIN_EXTRACT_LEGACY: {
int tmp;
tmp = dst - src + 128;
tmp = qMin(tmp, 255);
tmp = qMax(tmp, 0);
src = (uchar)tmp;
src_a = qMin(src_a, dst_a);
} break;
case GIMP_LAYER_MODE_GRAIN_MERGE:
case GIMP_LAYER_MODE_GRAIN_MERGE_LEGACY: {
int tmp;
tmp = dst + src - 128;
tmp = qMin(tmp, 255);
tmp = qMax(tmp, 0);
src = (uchar)tmp;
src_a = qMin(src_a, dst_a);
} break;
default:
qCWarning(XCFPLUGIN) << "Unhandled mode" << layer.mode;
return false;
}
src_a = INT_MULT(src_a, layer.opacity);
// Apply the mask (if any)
if (layer.apply_mask == 1 && layer.mask_tiles.size() > (int)j && layer.mask_tiles[j].size() > (int)i) {
src_a = INT_MULT(src_a, layer.mask_tiles[j][i].pixelIndex(k, l));
}
uchar new_a = dst_a + INT_MULT(OPAQUE_OPACITY - dst_a, src_a);
const float src_ratio = new_a == 0 ? 1.0 : (float)src_a / new_a;
float dst_ratio = 1.0 - src_ratio;
uchar new_g = (uchar)(src_ratio * src + dst_ratio * dst + EPSILON);
if (!modeAffectsSourceAlpha(layer.mode)) {
new_a = dst_a;
}
image.setPixel(m, n, qRgba(new_g, new_g, new_g, new_a));
return true;
}
/*!
* Merge an Indexed pixel from the layer to the Indexed image. Straight-forward.
* \param layer source layer.
* \param i x tile index.
* \param j y tile index.
* \param k x pixel index of tile i,j.
* \param l y pixel index of tile i,j.
* \param image destination image.
* \param m x pixel of destination image.
* \param n y pixel of destination image.
*/
bool XCFImageFormat::mergeIndexedToIndexed(const Layer &layer, uint i, uint j, int k, int l, QImage &image, int m, int n)
{
int src = layer.image_tiles[j][i].pixelIndex(k, l);
image.setPixel(m, n, src);
return true;
}
/*!
* Merge an IndexedA pixel from the layer to the Indexed image. Straight-forward.
* \param layer source layer.
* \param i x tile index.
* \param j y tile index.
* \param k x pixel index of tile i,j.
* \param l y pixel index of tile i,j.
* \param image destination image.
* \param m x pixel of destination image.
* \param n y pixel of destination image.
*/
bool XCFImageFormat::mergeIndexedAToIndexed(const Layer &layer, uint i, uint j, int k, int l, QImage &image, int m, int n)
{
uchar src = layer.image_tiles[j][i].pixelIndex(k, l);
uchar src_a = layer.alpha_tiles[j][i].pixelIndex(k, l);
src_a = INT_MULT(src_a, layer.opacity);
if (layer.apply_mask == 1 && layer.mask_tiles.size() > (int)j && layer.mask_tiles[j].size() > (int)i) {
src_a = INT_MULT(src_a, layer.mask_tiles[j][i].pixelIndex(k, l));
}
if (src_a > 127) {
src++;
image.setPixel(m, n, src);
}
return true;
}
/*!
* Merge an IndexedA pixel from the layer to an RGB image. Straight-forward.
* The only thing this has to take account of is the opacity of the
* layer. Evidently, the GIMP exporter itself does not actually do this.
* \param layer source layer.
* \param i x tile index.
* \param j y tile index.
* \param k x pixel index of tile i,j.
* \param l y pixel index of tile i,j.
* \param image destination image.
* \param m x pixel of destination image.
* \param n y pixel of destination image.
*/
bool XCFImageFormat::mergeIndexedAToRGB(const Layer &layer, uint i, uint j, int k, int l, QImage &image, int m, int n)
{
QRgb src = layer.image_tiles[j][i].pixel(k, l);
uchar src_a = layer.alpha_tiles[j][i].pixelIndex(k, l);
src_a = INT_MULT(src_a, layer.opacity);
// Apply the mask (if any)
if (layer.apply_mask == 1 && layer.mask_tiles.size() > (int)j && layer.mask_tiles[j].size() > (int)i) {
src_a = INT_MULT(src_a, layer.mask_tiles[j][i].pixelIndex(k, l));
}
// This is what appears in the GIMP window
if (src_a <= 127) {
src_a = 0;
} else {
src_a = OPAQUE_OPACITY;
}
image.setPixel(m, n, qRgba(src, src_a));
return true;
}
/*!
* Dissolving pixels: pick a random number between 0 and 255. If the pixel's
* alpha is less than that, make it transparent.
* \param image the image tile to dissolve.
* \param x the global x position of the tile.
* \param y the global y position of the tile.
*/
void XCFImageFormat::dissolveRGBPixels(QImage &image, int x, int y)
{
// The apparently spurious rand() calls are to wind the random
// numbers up to the same point for each tile.
for (int l = 0; l < image.height(); l++) {
unsigned int next = randomTable.values[(l + y) % RANDOM_TABLE_SIZE];
for (int k = 0; k < x; k++) {
RandomTable::rand_r(&next);
}
for (int k = 0; k < image.width(); k++) {
int rand_val = RandomTable::rand_r(&next) & 0xff;
QRgb pixel = image.pixel(k, l);
if (rand_val > qAlpha(pixel)) {
image.setPixel(k, l, qRgba(pixel, 0));
}
}
}
}
/*!
* Dissolving pixels: pick a random number between 0 and 255. If the pixel's
* alpha is less than that, make it transparent. This routine works for
* the GRAYA and INDEXEDA image types where the pixel alpha's are stored
* separately from the pixel themselves.
* \param image the alpha tile to dissolve.
* \param x the global x position of the tile.
* \param y the global y position of the tile.
*/
void XCFImageFormat::dissolveAlphaPixels(QImage &image, int x, int y)
{
// The apparently spurious rand() calls are to wind the random
// numbers up to the same point for each tile.
for (int l = 0; l < image.height(); l++) {
unsigned int next = randomTable.values[(l + y) % RANDOM_TABLE_SIZE];
for (int k = 0; k < x; k++) {
RandomTable::rand_r(&next);
}
for (int k = 0; k < image.width(); k++) {
int rand_val = RandomTable::rand_r(&next) & 0xff;
uchar alpha = image.pixelIndex(k, l);
if (rand_val > alpha) {
image.setPixel(k, l, 0);
}
}
}
}
///////////////////////////////////////////////////////////////////////////////
XCFHandler::XCFHandler()
{
}
bool XCFHandler::canRead() const
{
if (canRead(device())) {
setFormat("xcf");
return true;
}
return false;
}
bool XCFHandler::read(QImage *image)
{
XCFImageFormat xcfif;
return xcfif.readXCF(device(), image);
}
bool XCFHandler::write(const QImage &)
{
return false;
}
bool XCFHandler::supportsOption(ImageOption option) const
{
if (option == QImageIOHandler::Size)
return true;
return false;
}
QVariant XCFHandler::option(ImageOption option) const
{
QVariant v;
if (option == QImageIOHandler::Size) {
/*
* The image structure always starts at offset 0 in the XCF file.
* byte[9] "gimp xcf " File type identification
* byte[4] version XCF version
* "file": version 0
* "v001": version 1
* "v002": version 2
* "v003": version 3
* byte 0 Zero marks the end of the version tag.
* uint32 width Width of canvas
* uint32 height Height of canvas
*/
if (auto d = device()) {
// transactions works on both random and sequential devices
d->startTransaction();
auto ba9 = d->read(9); // "gimp xcf "
auto ba5 = d->read(4+1); // version + null terminator
auto ba = d->read(8); // width and height
d->rollbackTransaction();
if (ba9 == QByteArray("gimp xcf ") && ba5.size() == 5) {
QDataStream ds(ba);
quint32 width;
ds >> width;
quint32 height;
ds >> height;
if (ds.status() == QDataStream::Ok)
v = QVariant::fromValue(QSize(width, height));
}
}
}
return v;
}
bool XCFHandler::canRead(QIODevice *device)
{
if (!device) {
qCDebug(XCFPLUGIN) << "XCFHandler::canRead() called with no device";
return false;
}
if (device->isSequential()) {
return false;
}
const qint64 oldPos = device->pos();
if (!device->isSequential()) {
QDataStream ds(device);
XCFImageFormat::XCFImage::Header header;
bool failed = !XCFImageFormat::readXCFHeader(ds, &header);
ds.setDevice(nullptr);
device->seek(oldPos);
if (failed) {
return false;
}
switch (header.precision) {
case XCFImageFormat::GIMP_PRECISION_HALF_LINEAR:
case XCFImageFormat::GIMP_PRECISION_HALF_NON_LINEAR:
case XCFImageFormat::GIMP_PRECISION_HALF_PERCEPTUAL:
case XCFImageFormat::GIMP_PRECISION_FLOAT_LINEAR:
case XCFImageFormat::GIMP_PRECISION_FLOAT_NON_LINEAR:
case XCFImageFormat::GIMP_PRECISION_FLOAT_PERCEPTUAL:
case XCFImageFormat::GIMP_PRECISION_U8_LINEAR:
case XCFImageFormat::GIMP_PRECISION_U8_NON_LINEAR:
case XCFImageFormat::GIMP_PRECISION_U8_PERCEPTUAL:
case XCFImageFormat::GIMP_PRECISION_U16_LINEAR:
case XCFImageFormat::GIMP_PRECISION_U16_NON_LINEAR:
case XCFImageFormat::GIMP_PRECISION_U16_PERCEPTUAL:
case XCFImageFormat::GIMP_PRECISION_U32_LINEAR:
case XCFImageFormat::GIMP_PRECISION_U32_NON_LINEAR:
case XCFImageFormat::GIMP_PRECISION_U32_PERCEPTUAL:
break;
case XCFImageFormat::GIMP_PRECISION_DOUBLE_LINEAR:
case XCFImageFormat::GIMP_PRECISION_DOUBLE_NON_LINEAR:
case XCFImageFormat::GIMP_PRECISION_DOUBLE_PERCEPTUAL:
default:
qCDebug(XCFPLUGIN) << "unsupported precision" << header.precision;
return false;
}
return true;
}
char head[8];
qint64 readBytes = device->read(head, sizeof(head));
if (readBytes != sizeof(head)) {
if (device->isSequential()) {
while (readBytes > 0) {
device->ungetChar(head[readBytes-- - 1]);
}
} else {
device->seek(oldPos);
}
return false;
}
if (device->isSequential()) {
while (readBytes > 0) {
device->ungetChar(head[readBytes-- - 1]);
}
} else {
device->seek(oldPos);
}
return qstrncmp(head, "gimp xcf", 8) == 0;
}
QImageIOPlugin::Capabilities XCFPlugin::capabilities(QIODevice *device, const QByteArray &format) const
{
if (format == "xcf") {
return Capabilities(CanRead);
}
if (!format.isEmpty()) {
return {};
}
if (!device->isOpen()) {
return {};
}
Capabilities cap;
if (device->isReadable() && XCFHandler::canRead(device)) {
cap |= CanRead;
}
return cap;
}
QImageIOHandler *XCFPlugin::create(QIODevice *device, const QByteArray &format) const
{
QImageIOHandler *handler = new XCFHandler;
handler->setDevice(device);
handler->setFormat(format);
return handler;
}
// Just so I can get enum values printed
#include "xcf.moc"
#include "moc_xcf_p.cpp"
Reference in New Issue View Git Blame Copy Permalink