QImage Document::RenderPage (int num, double xRes, double yRes)
{
auto page = WrapPage (pdf_load_page (MuDoc_, num), MuDoc_);
if (!page)
return QImage ();
#if MUPDF_VERSION < 0x0102
const auto& rect = pdf_bound_page (MuDoc_, page.get ());
#else
fz_rect rect;
pdf_bound_page (MuDoc_, page.get (), &rect);
#endif
auto px = fz_new_pixmap (MuCtx_, fz_device_bgr, xRes * (rect.x1 - rect.x0), yRes * (rect.y1 - rect.y0));
fz_clear_pixmap (MuCtx_, px);
auto dev = fz_new_draw_device (MuCtx_, px);
#if MUPDF_VERSION < 0x0102
pdf_run_page (MuDoc_, page.get (), dev, fz_scale (xRes, yRes), NULL);
#else
fz_matrix matrix;
pdf_run_page (MuDoc_, page.get (), dev, fz_scale (&matrix, xRes, yRes), NULL);
#endif
fz_free_device (dev);
const int pxWidth = fz_pixmap_width (MuCtx_, px);
const int pxHeight = fz_pixmap_height (MuCtx_, px);
auto samples = fz_pixmap_samples (MuCtx_, px);
QImage temp (samples, pxWidth, pxHeight, QImage::Format_ARGB32);
QImage img (QSize (pxWidth, pxHeight), QImage::Format_ARGB32);
for (int y = 0; y < pxHeight; ++y)
{
auto target = reinterpret_cast<QRgb*> (img.scanLine (y));
const auto source = reinterpret_cast<QRgb*> (temp.scanLine (y));
std::memcpy (target, source, sizeof (source [0]) * pxWidth);
}
fz_drop_pixmap (MuCtx_, px);
temp = QImage (QSize (pxWidth, pxHeight), QImage::Format_ARGB32);
QPainter p;
p.begin (&temp);
p.fillRect (QRect (QPoint (0, 0), temp.size ()), Qt::white);
p.drawImage (0, 0, img);
p.end ();
return temp;
}
static fz_display_list* _pdf_doc_get_list(
struct _pdf_doc *self, int pageno)
{
pdf_page *page;
if (self->disps[pageno])
return self->disps[pageno];
page = _pdf_doc_get_page(self, pageno);
if (NULL == self->disps[pageno]) {
fz_error err;
fz_device *mdev;
self->disps[pageno] = fz_new_display_list();
mdev = fz_new_list_device(self->disps[pageno]);
err = pdf_run_page(self->xref, page, mdev, fz_identity);
if (err) {
mume_error(("pdf_run_page(%d): %d\n", pageno, err));
}
fz_free_device(mdev);
}
return self->disps[pageno];
}
void benchrenderpage(fz_context *ctx, pdf_document *xref, pdf_page *page, int pagenum)
{
fz_device *dev;
fz_pixmap *pix;
fz_bbox bbox;
mstimer timer;
timerstart(&timer);
bbox = fz_round_rect(pdf_bound_page(xref, page));
pix = fz_new_pixmap_with_bbox(ctx, fz_device_rgb, bbox);
fz_clear_pixmap_with_value(ctx, pix, 0xFF);
dev = fz_new_draw_device(ctx, pix);
fz_try(ctx) {
pdf_run_page(xref, page, dev, fz_identity, NULL);
timerstop(&timer);
logbench("pagerender %3d: %.2f ms\n", pagenum, timeinms(&timer));
}
fz_catch(ctx) {
logbench("Error: pdf_run_page() failed\n");
}
fz_drop_pixmap(ctx, pix);
fz_free_device(dev);
}
static void pdfapp_loadpage_pdf(pdfapp_t *app)
{
pdf_page *page;
fz_error error;
fz_device *mdev;
error = pdf_load_page(&page, app->xref, app->pageno - 1);
if (error)
pdfapp_error(app, error);
app->page_bbox = page->mediabox;
app->page_rotate = page->rotate;
app->page_links = page->links;
page->links = NULL;
/* Create display list */
app->page_list = fz_new_display_list();
mdev = fz_new_list_device(app->page_list);
error = pdf_run_page(app->xref, page, mdev, fz_identity);
if (error)
{
error = fz_rethrow(error, "cannot draw page %d in '%s'", app->pageno, app->doctitle);
pdfapp_error(app, error);
}
fz_free_device(mdev);
pdf_free_page(page);
pdf_age_store(app->xref->store, 3);
}
QImage Pdf::page(int i)
{
pdf_page* page = pdf_load_page(xref, i);
if (page == 0) {
printf("cannot load page %d\n", i);
return QImage();
}
static const float resolution = 300.0;
const float zoom = resolution / 72.0;
fz_matrix ctm = fz_translate(0, -page->mediabox.y1);
ctm = fz_concat(ctm, fz_scale(zoom, -zoom));
ctm = fz_concat(ctm, fz_rotate(page->rotate));
fz_bbox bbox = fz_round_rect(fz_transform_rect(ctm, page->mediabox));
fz_pixmap* pix = fz_new_pixmap_with_rect(ctx, fz_device_gray, bbox);
fz_clear_pixmap_with_color(pix, 255);
fz_device* dev = fz_new_draw_device(ctx, cache, pix);
pdf_run_page(xref, page, dev, ctm);
fz_free_device(dev);
int w = pix->w;
int h = pix->h;
QImage image(w, h, QImage::Format_MonoLSB);
QVector<QRgb> ct(2);
ct[0] = qRgb(255, 255, 255);
ct[1] = qRgb(0, 0, 0);
image.setColorTable(ct);
uchar* s = pix->samples;
int stride = image.bytesPerLine();
int bytes = w >> 3;
for (int line = 0; line < h; ++line) {
uchar* d = image.bits() + stride * line;
for (int col = 0; col < bytes; ++col) {
uchar data = 0;
for (int i = 0; i < 8; ++i) {
uchar v = *s++;
s++;
data >>= 1;
if (v < 128)
data |= 0x80;
}
*d++ = data;
}
}
fz_drop_pixmap(ctx, pix);
pdf_free_page(ctx, page);
return image;
}
JNIEXPORT void JNICALL
Java_com_artifex_mupdf_MuPDFCore_gotoPageInternal(JNIEnv *env, jobject thiz, int page)
{
float zoom;
fz_matrix ctm;
fz_bbox bbox;
fz_device *dev = NULL;
pdf_page *currentPage = NULL;
fz_var(dev);
fz_var(currentPage);
/* In the event of an error, ensure we give a non-empty page */
pageWidth = 100;
pageHeight = 100;
LOGE("Goto page %d...", page);
fz_try(ctx)
{
if (currentPageList != NULL)
{
fz_free_display_list(ctx, currentPageList);
currentPageList = NULL;
}
pagenum = page;
currentPage = pdf_load_page(xref, pagenum);
zoom = resolution / 72;
currentMediabox = pdf_bound_page(xref, currentPage);
ctm = fz_scale(zoom, zoom);
bbox = fz_round_rect(fz_transform_rect(ctm, currentMediabox));
pageWidth = bbox.x1-bbox.x0;
pageHeight = bbox.y1-bbox.y0;
/* Render to list */
currentPageList = fz_new_display_list(ctx);
dev = fz_new_list_device(ctx, currentPageList);
pdf_run_page(xref, currentPage, dev, fz_identity, NULL);
}
fz_catch(ctx)
{
LOGE("cannot make displaylist from page %d", pagenum);
}
pdf_free_page(ctx, currentPage);
currentPage = NULL;
fz_free_device(dev);
dev = NULL;
}
void PDFDocument::Render(
Document::PixelWriter* pw, int page, float zoom, int rotation) {
assert((page >= 0) && (page < GetNumPages()));
std::unique_lock<std::mutex> lock(_render_mutex);
// 1. Init MuPDF structures.
const fz_matrix& m = Transform(zoom, rotation);
pdf_page* page_struct = GetPage(page);
const fz_irect& bbox = GetBoundingBox(page_struct, m);
fz_pixmap* pixmap = fz_new_pixmap_with_bbox(
_fz_context, fz_device_rgb(_fz_context), FZ_OBJ(bbox), nullptr, 1);
fz_device* dev = fz_new_draw_device(_fz_context, FZ_OBJ(fz_identity), pixmap);
// 2. Render page.
fz_clear_pixmap_with_value(_fz_context, pixmap, 0xff);
pdf_run_page(
_fz_context, _pdf_document, page_struct, dev, FZ_OBJ(m), nullptr);
// 3. Write pixmap to buffer. The page is vertically divided into n equal
// stripes, each copied to pw by one thread.
assert(fz_pixmap_components(_fz_context, pixmap) == 4);
uint8_t* buffer =
reinterpret_cast<uint8_t*>(fz_pixmap_samples(_fz_context, pixmap));
const int num_cols = fz_pixmap_width(_fz_context, pixmap);
const int num_rows = fz_pixmap_height(_fz_context, pixmap);
ExecuteInParallel([=](int num_threads, int i) {
const int num_rows_per_thread = num_rows / num_threads;
const int y_begin = i * num_rows_per_thread;
const int y_end =
(i == num_threads - 1) ? num_rows : (i + 1) * num_rows_per_thread;
uint8_t* p = buffer + y_begin * num_cols * 4;
for (int y = y_begin; y < y_end; ++y) {
for (int x = 0; x < num_cols; ++x) {
pw->Write(x, y, p[0], p[1], p[2]);
p += 4;
}
}
});
// 4. Clean up.
fz_close_device(_fz_context, dev);
fz_drop_device(_fz_context, dev);
fz_drop_pixmap(_fz_context, pixmap);
}
std::string PDFDocument::GetPageText(int page, int line_sep) {
// 1. Init MuPDF structures.
pdf_page* page_struct = GetPage(page);
fz_stext_sheet* text_sheet = fz_new_stext_sheet(_fz_context);
// 2. Render page.
#if MUPDF_VERSION >= 10010
fz_stext_options stext_options = { 0 };
// See #elif MUPDF_VERSION >= 10009 block below.
fz_stext_page* text_page = fz_new_stext_page_from_page(
_fz_context, &(page_struct->super), text_sheet, &stext_options);
#elif MUPDF_VERSION >= 10009
// The function below is a wrapper around fz_run_page that uses a fresh
// device. We can't use pdf_run_page to gather the text for us.
// These notes are also left in here in case MuPDF's API changes again.
fz_stext_page* text_page = fz_new_stext_page_from_page(
_fz_context, &(page_struct->super), text_sheet);
#else
fz_stext_page* text_page = fz_new_text_page(_fz_context);
fz_device* dev = fz_new_stext_device(_fz_context, text_sheet, text_page);
// I've no idea what fz_{begin,end}_page do, but without them pdf_run_page
// segfaults :-/
fz_begin_page(_fz_context, dev, &fz_infinite_rect, &fz_identity);
pdf_run_page(
_fz_context, _pdf_document, page_struct, dev, &fz_identity, nullptr);
fz_end_page(_fz_context, dev);
#endif
// 3. Build text.
std::string r;
for (fz_page_block* page_block = text_page->blocks;
page_block < text_page->blocks + text_page->len;
++page_block) {
assert(page_block != nullptr);
if (page_block->type != FZ_PAGE_BLOCK_TEXT) {
continue;
}
fz_stext_block* const text_block = page_block->u.text;
assert(text_block != nullptr);
for (fz_stext_line* text_line = text_block->lines;
text_line < text_block->lines + text_block->len;
++text_line) {
assert(text_line != nullptr);
for (fz_stext_span* text_span = text_line->first_span;
text_span != nullptr;
text_span = text_span->next) {
for (int i = 0; i < text_span->len; ++i) {
const int c = text_span->text[i].c;
// A single UTF-8 character cannot take more than 4 bytes, but let's
// go for 8.
char buffer[8];
const int num_bytes = fz_runetochar(buffer, c);
assert(num_bytes <= static_cast<int>(sizeof(buffer)));
buffer[num_bytes] = '\0';
r += buffer;
}
}
if (!isspace(r.back())) {
r += line_sep;
}
}
}
// 4. Clean up.
fz_drop_stext_page(_fz_context, text_page);
fz_drop_stext_sheet(_fz_context, text_sheet);
return r;
}
std::string PDFDocument::GetPageText(int page, int line_sep) {
// 1. Init MuPDF structures.
pdf_page* page_struct = GetPage(page);
#if MUPDF_VERSION < 10012
fz_stext_sheet* text_sheet = fz_new_stext_sheet(_fz_context);
#endif
// 2. Render page.
#if MUPDF_VERSION >= 10012
fz_stext_options stext_options = {0};
// See #elif MUPDF_VERSION >= 10009 block below.
fz_stext_page* text_page = fz_new_stext_page_from_page(
_fz_context, &(page_struct->super), &stext_options);
#elif MUPDF_VERSION >= 10010
fz_stext_options stext_options = {0};
// See #elif MUPDF_VERSION >= 10009 block below.
fz_stext_page* text_page = fz_new_stext_page_from_page(
_fz_context, &(page_struct->super), text_sheet, &stext_options);
#elif MUPDF_VERSION >= 10009
// The function below is a wrapper around fz_run_page that uses a fresh
// device. We can't use pdf_run_page to gather the text for us.
// These notes are also left in here in case MuPDF's API changes again.
fz_stext_page* text_page = fz_new_stext_page_from_page(
_fz_context, &(page_struct->super), text_sheet);
#else
fz_stext_page* text_page = fz_new_text_page(_fz_context);
fz_device* dev = fz_new_stext_device(_fz_context, text_sheet, text_page);
// I've no idea what fz_{begin,end}_page do, but without them pdf_run_page
// segfaults :-/
fz_begin_page(_fz_context, dev, &fz_infinite_rect, &fz_identity);
pdf_run_page(
_fz_context, _pdf_document, page_struct, dev, &fz_identity, nullptr);
fz_end_page(_fz_context, dev);
#endif
// 3. Build text.
std::string r;
#if MUPDF_VERSION >= 10012
for (fz_stext_block* text_block = text_page->first_block;
text_block != nullptr; text_block = text_block->next) {
if (text_block->type != FZ_STEXT_BLOCK_TEXT) {
continue;
}
for (fz_stext_line* text_line = text_block->u.t.first_line;
text_line != nullptr; text_line = text_line->next) {
for (fz_stext_char* text_char = text_line->first_char;
text_char != nullptr; text_char = text_char->next) {
{
const int c = text_char->c;
#else
for (fz_page_block* page_block = text_page->blocks;
page_block < text_page->blocks + text_page->len; ++page_block) {
assert(page_block != nullptr);
if (page_block->type != FZ_PAGE_BLOCK_TEXT) {
continue;
}
fz_stext_block* const text_block = page_block->u.text;
assert(text_block != nullptr);
for (fz_stext_line* text_line = text_block->lines;
text_line < text_block->lines + text_block->len; ++text_line) {
assert(text_line != nullptr);
for (fz_stext_span* text_span = text_line->first_span;
text_span != nullptr; text_span = text_span->next) {
for (int i = 0; i < text_span->len; ++i) {
const int c = text_span->text[i].c;
#endif
// A single UTF-8 character cannot take more than 4 bytes, but let's
// go for 8.
char buffer[8];
const int num_bytes = fz_runetochar(buffer, c);
assert(num_bytes <= static_cast<int>(sizeof(buffer)));
buffer[num_bytes] = '\0';
r += buffer;
}
}
if (!isspace(r.back())) {
r += line_sep;
}
}
}
// 4. Clean up.
fz_drop_stext_page(_fz_context, text_page);
#if MUPDF_VERSION < 10012
fz_drop_stext_sheet(_fz_context, text_sheet);
#endif
return r;
}
PDFDocument::PDFOutlineItem::~PDFOutlineItem() {}
PDFDocument::PDFOutlineItem::PDFOutlineItem(fz_outline* src) {
if (src == nullptr) {
_dest_page = -1;
} else {
_title = src->title;
#if MUPDF_VERSION >= 10010
_dest_page = src->page;
#else
_dest_page = src->dest.ld.gotor.page;
#endif
}
}
int PDFDocument::PDFOutlineItem::GetDestPage() const { return _dest_page; }
PDFDocument::PDFOutlineItem* PDFDocument::PDFOutlineItem::Build(
fz_context* ctx, fz_outline* src) {
PDFOutlineItem* root = nullptr;
std::vector<std::unique_ptr<OutlineItem>> items;
BuildRecursive(src, &items);
fz_drop_outline(ctx, src);
if (items.empty()) {
return nullptr;
} else if (items.size() == 1) {
root = dynamic_cast<PDFOutlineItem*>(items[0].release());
} else {
root = new PDFOutlineItem(nullptr);
root->_title = DEFAULT_ROOT_OUTLINE_ITEM_TITLE;
root->_children.swap(items);
}
return root;
}
void PDFDocument::PDFOutlineItem::BuildRecursive(
fz_outline* src,
std::vector<std::unique_ptr<Document::OutlineItem>>* output) {
assert(output != nullptr);
for (fz_outline* i = src; i != nullptr; i = i->next) {
PDFOutlineItem* item = new PDFOutlineItem(i);
if (i->down != nullptr) {
BuildRecursive(i->down, &(item->_children));
}
output->push_back(std::unique_ptr<Document::OutlineItem>(item));
}
}
PDFDocument::PDFPageCache::PDFPageCache(int cache_size, PDFDocument* parent)
: Cache<int, pdf_page*>(cache_size), _parent(parent) {}
PDFDocument::PDFPageCache::~PDFPageCache() { Clear(); }
pdf_page* PDFDocument::PDFPageCache::Load(const int& page) {
std::unique_lock<std::mutex> lock(_mutex);
return pdf_load_page(_parent->_fz_context, _parent->_pdf_document, page);
}
void PDFDocument::PDFPageCache::Discard(
const int& page, pdf_page* const& page_struct) {
std::unique_lock<std::mutex> lock(_mutex);
pdf_drop_page(_parent->_fz_context, _parent->_pdf_document, page_struct);
}
