pdf/pdfium/pdfium_page.cc

   1 // Copyright (c) 2010 The Chromium Authors. All rights reserved.
   2 // Use of this source code is governed by a BSD-style license that can be
   3 // found in the LICENSE file.
   4
   5 #include "pdf/pdfium/pdfium_page.h"
   6
   7 #include <math.h>
   8
   9 #include "base/logging.h"
  10 #include "base/strings/string_number_conversions.h"
  11 #include "base/strings/string_util.h"
  12 #include "base/strings/utf_string_conversions.h"
  13 #include "base/values.h"
  14 #include "pdf/pdfium/pdfium_engine.h"
  15
  16 // Used when doing hit detection.
  17 #define kTolerance 20.0
  18
  19 // Dictionary Value key names for returning the accessible page content as JSON.
  20 const char kPageWidth[] = "width";
  21 const char kPageHeight[] = "height";
  22 const char kPageTextBox[] = "textBox";
  23 const char kTextBoxLeft[] = "left";
  24 const char kTextBoxTop[]  = "top";
  25 const char kTextBoxWidth[] = "width";
  26 const char kTextBoxHeight[]  = "height";
  27 const char kTextBoxFontSize[] = "fontSize";
  28 const char kTextBoxNodes[] = "textNodes";
  29 const char kTextNodeType[] = "type";
  30 const char kTextNodeText[] = "text";
  31 const char kTextNodeURL[] = "url";
  32 const char kTextNodeTypeText[] = "text";
  33 const char kTextNodeTypeURL[] = "url";
  34 const char kDocLinkURLPrefix[] = "#page";
  35
  36 namespace chrome_pdf {
  37
  38 PDFiumPage::PDFiumPage(PDFiumEngine* engine,
  39                        int i,
  40                        const pp::Rect& r,
  41                        bool available)
  42     : engine_(engine),
  43       page_(NULL),
  44       text_page_(NULL),
  45       index_(i),
  46       rect_(r),
  47       calculated_links_(false),
  48       available_(available) {
  49 }
  50
  51 PDFiumPage::~PDFiumPage() {
  52 }
  53
  54 void PDFiumPage::Unload() {
  55   if (text_page_) {
  56     FPDFText_ClosePage(text_page_);
  57     text_page_ = NULL;
  58   }
  59
  60   if (page_) {
  61     if (engine_->form()) {
  62       FORM_OnBeforeClosePage(page_, engine_->form());
  63     }
  64     FPDF_ClosePage(page_);
  65     page_ = NULL;
  66   }
  67 }
  68
  69 FPDF_PAGE PDFiumPage::GetPage() {
  70   ScopedUnsupportedFeature scoped_unsupported_feature(engine_);
  71   if (!available_)
  72     return NULL;
  73   if (!page_) {
  74     page_ = FPDF_LoadPage(engine_->doc(), index_);
  75     if (page_ && engine_->form()) {
  76       FORM_OnAfterLoadPage(page_, engine_->form());
  77     }
  78   }
  79   return page_;
  80 }
  81
  82 FPDF_PAGE PDFiumPage::GetPrintPage() {
  83   ScopedUnsupportedFeature scoped_unsupported_feature(engine_);
  84   if (!available_)
  85     return NULL;
  86   if (!page_)
  87     page_ = FPDF_LoadPage(engine_->doc(), index_);
  88   return page_;
  89 }
  90
  91 void PDFiumPage::ClosePrintPage() {
  92   if (page_) {
  93     FPDF_ClosePage(page_);
  94     page_ = NULL;
  95   }
  96 }
  97
  98 FPDF_TEXTPAGE PDFiumPage::GetTextPage() {
  99   if (!available_)
 100     return NULL;
 101   if (!text_page_)
 102     text_page_ = FPDFText_LoadPage(GetPage());
 103   return text_page_;
 104 }
 105
 106 base::Value* PDFiumPage::GetAccessibleContentAsValue(int rotation) {
 107   base::DictionaryValue* node = new base::DictionaryValue();
 108
 109   if (!available_)
 110     return node;
 111
 112   double width = FPDF_GetPageWidth(GetPage());
 113   double height = FPDF_GetPageHeight(GetPage());
 114
 115   base::ListValue* text = new base::ListValue();
 116   int box_count = FPDFText_CountRects(GetTextPage(), 0, GetCharCount());
 117   for (int i = 0; i < box_count; i++) {
 118     double left, top, right, bottom;
 119     FPDFText_GetRect(GetTextPage(), i, &left, &top, &right, &bottom);
 120     text->Append(
 121         GetTextBoxAsValue(height, left, top, right, bottom, rotation));
 122   }
 123
 124   node->SetDouble(kPageWidth, width);
 125   node->SetDouble(kPageHeight, height);
 126   node->Set(kPageTextBox, text);  // Takes ownership of |text|
 127
 128   return node;
 129 }
 130
 131 base::Value* PDFiumPage::GetTextBoxAsValue(double page_height,
 132                                            double left, double top,
 133                                            double right, double bottom,
 134                                            int rotation) {
 135   base::string16 text_utf16;
 136   int char_count =
 137     FPDFText_GetBoundedText(GetTextPage(), left, top, right, bottom, NULL, 0);
 138   if (char_count > 0) {
 139     unsigned short* data = reinterpret_cast<unsigned short*>(
 140         WriteInto(&text_utf16, char_count + 1));
 141     FPDFText_GetBoundedText(GetTextPage(),
 142                             left, top, right, bottom,
 143                             data, char_count);
 144   }
 145   std::string text_utf8 = base::UTF16ToUTF8(text_utf16);
 146
 147   FPDF_LINK link = FPDFLink_GetLinkAtPoint(GetPage(), left, top);
 148   Area area;
 149   std::vector<LinkTarget> targets;
 150   if (link) {
 151     targets.push_back(LinkTarget());
 152     area = GetLinkTarget(link, &targets[0]);
 153   } else {
 154     pp::Rect rect(
 155         PageToScreen(pp::Point(), 1.0, left, top, right, bottom, rotation));
 156     GetLinks(rect, &targets);
 157     area = targets.size() == 0 ? TEXT_AREA : WEBLINK_AREA;
 158   }
 159
 160   int char_index = FPDFText_GetCharIndexAtPos(GetTextPage(), left, top,
 161                                               kTolerance, kTolerance);
 162   double font_size = FPDFText_GetFontSize(GetTextPage(), char_index);
 163
 164   base::DictionaryValue* node = new base::DictionaryValue();
 165   node->SetDouble(kTextBoxLeft, left);
 166   node->SetDouble(kTextBoxTop, page_height - top);
 167   node->SetDouble(kTextBoxWidth, right - left);
 168   node->SetDouble(kTextBoxHeight, top - bottom);
 169   node->SetDouble(kTextBoxFontSize, font_size);
 170
 171   base::ListValue* text_nodes = new base::ListValue();
 172
 173   if (area == DOCLINK_AREA) {
 174     std::string url = kDocLinkURLPrefix + base::IntToString(targets[0].page);
 175     text_nodes->Append(CreateURLNode(text_utf8, url));
 176   } else if (area == WEBLINK_AREA && link) {
 177     text_nodes->Append(CreateURLNode(text_utf8, targets[0].url));
 178   } else if (area == WEBLINK_AREA && !link) {
 179     size_t start = 0;
 180     for (size_t i = 0; i < targets.size(); ++i) {
 181       // Remove the extra NULL character at end.
 182       // Otherwise, find() will not return any matches.
 183       if (targets[i].url.size() > 0 &&
 184           targets[i].url[targets[i].url.size() - 1] == '\0') {
 185         targets[i].url.resize(targets[i].url.size() - 1);
 186       }
 187       // There should only ever be one NULL character
 188       DCHECK(targets[i].url[targets[i].url.size() - 1] != '\0');
 189
 190       // PDFium may change the case of generated links.
 191       std::string lowerCaseURL = base::StringToLowerASCII(targets[i].url);
 192       std::string lowerCaseText = base::StringToLowerASCII(text_utf8);
 193       size_t pos = lowerCaseText.find(lowerCaseURL, start);
 194       size_t length = targets[i].url.size();
 195       if (pos == std::string::npos) {
 196         // Check if the link is a "mailto:" URL
 197         if (lowerCaseURL.compare(0, 7, "mailto:") == 0) {
 198           pos = lowerCaseText.find(lowerCaseURL.substr(7), start);
 199           length -= 7;
 200         }
 201
 202         if (pos == std::string::npos) {
 203           // No match has been found.  This should never happen.
 204           continue;
 205         }
 206       }
 207
 208       std::string before_text = text_utf8.substr(start, pos - start);
 209       if (before_text.size() > 0)
 210         text_nodes->Append(CreateTextNode(before_text));
 211       std::string link_text = text_utf8.substr(pos, length);
 212       text_nodes->Append(CreateURLNode(link_text, targets[i].url));
 213
 214       start = pos + length;
 215     }
 216     std::string before_text = text_utf8.substr(start);
 217     if (before_text.size() > 0)
 218       text_nodes->Append(CreateTextNode(before_text));
 219   } else {
 220     text_nodes->Append(CreateTextNode(text_utf8));
 221   }
 222
 223   node->Set(kTextBoxNodes, text_nodes);  // Takes ownership of |text_nodes|.
 224   return node;
 225 }
 226
 227 base::Value* PDFiumPage::CreateTextNode(std::string text) {
 228   base::DictionaryValue* node = new base::DictionaryValue();
 229   node->SetString(kTextNodeType, kTextNodeTypeText);
 230   node->SetString(kTextNodeText, text);
 231   return node;
 232 }
 233
 234 base::Value* PDFiumPage::CreateURLNode(std::string text, std::string url) {
 235   base::DictionaryValue* node = new base::DictionaryValue();
 236   node->SetString(kTextNodeType, kTextNodeTypeURL);
 237   node->SetString(kTextNodeText, text);
 238   node->SetString(kTextNodeURL, url);
 239   return node;
 240 }
 241
 242 PDFiumPage::Area PDFiumPage::GetCharIndex(const pp::Point& point,
 243                                           int rotation,
 244                                           int* char_index,
 245                                           LinkTarget* target) {
 246   if (!available_)
 247     return NONSELECTABLE_AREA;
 248   pp::Point point2 = point - rect_.point();
 249   double new_x, new_y;
 250   FPDF_DeviceToPage(GetPage(), 0, 0, rect_.width(), rect_.height(),
 251         rotation, point2.x(), point2.y(), &new_x, &new_y);
 252
 253   int rv = FPDFText_GetCharIndexAtPos(
 254       GetTextPage(), new_x, new_y, kTolerance, kTolerance);
 255   *char_index = rv;
 256
 257   FPDF_LINK link = FPDFLink_GetLinkAtPoint(GetPage(), new_x, new_y);
 258   if (link) {
 259     // We don't handle all possible link types of the PDF. For example,
 260     // launch actions, cross-document links, etc.
 261     // In that case, GetLinkTarget() will return NONSELECTABLE_AREA
 262     // and we should proceed with area detection.
 263     PDFiumPage::Area area = GetLinkTarget(link, target);
 264     if (area != PDFiumPage::NONSELECTABLE_AREA)
 265       return area;
 266   }
 267
 268   if (rv < 0)
 269     return NONSELECTABLE_AREA;
 270
 271   return GetLink(*char_index, target) != -1 ? WEBLINK_AREA : TEXT_AREA;
 272 }
 273
 274 base::char16 PDFiumPage::GetCharAtIndex(int index) {
 275   if (!available_)
 276     return L'\0';
 277   return static_cast<base::char16>(FPDFText_GetUnicode(GetTextPage(), index));
 278 }
 279
 280 int PDFiumPage::GetCharCount() {
 281   if (!available_)
 282     return 0;
 283   return FPDFText_CountChars(GetTextPage());
 284 }
 285
 286 PDFiumPage::Area PDFiumPage::GetLinkTarget(
 287     FPDF_LINK link, PDFiumPage::LinkTarget* target) {
 288   FPDF_DEST dest = FPDFLink_GetDest(engine_->doc(), link);
 289   if (dest != NULL)
 290     return GetDestinationTarget(dest, target);
 291
 292   FPDF_ACTION action = FPDFLink_GetAction(link);
 293   if (action) {
 294     switch (FPDFAction_GetType(action)) {
 295       case PDFACTION_GOTO: {
 296           FPDF_DEST dest = FPDFAction_GetDest(engine_->doc(), action);
 297           if (dest)
 298             return GetDestinationTarget(dest, target);
 299           // TODO(gene): We don't fully support all types of the in-document
 300           // links. Need to implement that. There is a bug to track that:
 301           // http://code.google.com/p/chromium/issues/detail?id=55776
 302         } break;
 303       case PDFACTION_URI: {
 304           if (target) {
 305             size_t buffer_size =
 306                 FPDFAction_GetURIPath(engine_->doc(), action, NULL, 0);
 307             if (buffer_size > 1) {
 308               void* data = WriteInto(&target->url, buffer_size + 1);
 309               FPDFAction_GetURIPath(engine_->doc(), action, data, buffer_size);
 310             }
 311           }
 312           return WEBLINK_AREA;
 313         } break;
 314       // TODO(gene): We don't support PDFACTION_REMOTEGOTO and PDFACTION_LAUNCH
 315       // at the moment.
 316     }
 317   }
 318
 319   return NONSELECTABLE_AREA;
 320 }
 321
 322 PDFiumPage::Area PDFiumPage::GetDestinationTarget(
 323     FPDF_DEST destination, PDFiumPage::LinkTarget* target) {
 324   int page_index = FPDFDest_GetPageIndex(engine_->doc(), destination);
 325   if (target) {
 326     target->page = page_index;
 327   }
 328   return DOCLINK_AREA;
 329 }
 330
 331 int PDFiumPage::GetLink(int char_index, PDFiumPage::LinkTarget* target) {
 332   if (!available_)
 333     return -1;
 334
 335   CalculateLinks();
 336
 337   // Get the bounding box of the rect again, since it might have moved because
 338   // of the tolerance above.
 339   double left, right, bottom, top;
 340   FPDFText_GetCharBox(GetTextPage(), char_index, &left, &right, &bottom, &top);
 341
 342   pp::Point origin(
 343       PageToScreen(pp::Point(), 1.0, left, top, right, bottom, 0).point());
 344   for (size_t i = 0; i < links_.size(); ++i) {
 345     for (size_t j = 0; j < links_[i].rects.size(); ++j) {
 346       if (links_[i].rects[j].Contains(origin)) {
 347         if (target)
 348           target->url = links_[i].url;
 349         return i;
 350       }
 351     }
 352   }
 353   return -1;
 354 }
 355
 356 std::vector<int> PDFiumPage::GetLinks(pp::Rect text_area,
 357                                       std::vector<LinkTarget>* targets) {
 358   if (!available_)
 359     return std::vector<int>();
 360
 361   CalculateLinks();
 362
 363   std::vector<int> links;
 364
 365   for (size_t i = 0; i < links_.size(); ++i) {
 366     for (size_t j = 0; j < links_[i].rects.size(); ++j) {
 367       if (links_[i].rects[j].Intersects(text_area)) {
 368         if (targets) {
 369           LinkTarget target;
 370           target.url = links_[i].url;
 371           targets->push_back(target);
 372         }
 373         links.push_back(i);
 374       }
 375     }
 376   }
 377   return links;
 378 }
 379
 380 void PDFiumPage::CalculateLinks() {
 381   if (calculated_links_)
 382     return;
 383
 384   calculated_links_ = true;
 385   FPDF_PAGELINK links = FPDFLink_LoadWebLinks(GetTextPage());
 386   int count = FPDFLink_CountWebLinks(links);
 387   for (int i = 0; i < count; ++i) {
 388     base::string16 url;
 389     int url_length = FPDFLink_GetURL(links, i, NULL, 0);
 390     if (url_length > 1) {  // WriteInto needs at least 2 characters.
 391       unsigned short* data =
 392           reinterpret_cast<unsigned short*>(WriteInto(&url, url_length + 1));
 393       FPDFLink_GetURL(links, i, data, url_length);
 394     }
 395     Link link;
 396     link.url = base::UTF16ToUTF8(url);
 397
 398     // If the link cannot be converted to a pp::Var, then it is not possible to
 399     // pass it to JS. In this case, ignore the link like other PDF viewers.
 400     // See http://crbug.com/312882 for an example.
 401     pp::Var link_var(link.url);
 402     if (!link_var.is_string())
 403       continue;
 404
 405     // Make sure all the characters in the URL are valid per RFC 1738.
 406     // http://crbug.com/340326 has a sample bad PDF.
 407     // GURL does not work correctly, e.g. it just strips \t \r \n.
 408     bool is_invalid_url = false;
 409     for (size_t j = 0; j < link.url.length(); ++j) {
 410       // Control characters are not allowed.
 411       // 0x7F is also a control character.
 412       // 0x80 and above are not in US-ASCII.
 413       if (link.url[j] < ' ' || link.url[j] >= '\x7F') {
 414         is_invalid_url = true;
 415         break;
 416       }
 417     }
 418     if (is_invalid_url)
 419       continue;
 420
 421     int rect_count = FPDFLink_CountRects(links, i);
 422     for (int j = 0; j < rect_count; ++j) {
 423       double left, top, right, bottom;
 424       FPDFLink_GetRect(links, i, j, &left, &top, &right, &bottom);
 425       link.rects.push_back(
 426           PageToScreen(pp::Point(), 1.0, left, top, right, bottom, 0));
 427     }
 428     links_.push_back(link);
 429   }
 430   FPDFLink_CloseWebLinks(links);
 431 }
 432
 433 pp::Rect PDFiumPage::PageToScreen(const pp::Point& offset,
 434                                   double zoom,
 435                                   double left,
 436                                   double top,
 437                                   double right,
 438                                   double bottom,
 439                                   int rotation) {
 440   if (!available_)
 441     return pp::Rect();
 442
 443   int new_left, new_top, new_right, new_bottom;
 444   FPDF_PageToDevice(
 445       page_,
 446       static_cast<int>((rect_.x() - offset.x()) * zoom),
 447       static_cast<int>((rect_.y() - offset.y()) * zoom),
 448       static_cast<int>(ceil(rect_.width() * zoom)),
 449       static_cast<int>(ceil(rect_.height() * zoom)),
 450       rotation, left, top, &new_left, &new_top);
 451   FPDF_PageToDevice(
 452       page_,
 453       static_cast<int>((rect_.x() - offset.x()) * zoom),
 454       static_cast<int>((rect_.y() - offset.y()) * zoom),
 455       static_cast<int>(ceil(rect_.width() * zoom)),
 456       static_cast<int>(ceil(rect_.height() * zoom)),
 457       rotation, right, bottom, &new_right, &new_bottom);
 458
 459   // If the PDF is rotated, the horizontal/vertical coordinates could be
 460   // flipped.  See
 461   // http://www.netl.doe.gov/publications/proceedings/03/ubc/presentations/Goeckner-pres.pdf
 462   if (new_right < new_left)
 463     std::swap(new_right, new_left);
 464   if (new_bottom < new_top)
 465     std::swap(new_bottom, new_top);
 466
 467   return pp::Rect(
 468       new_left, new_top, new_right - new_left + 1, new_bottom - new_top + 1);
 469 }
 470
 471 PDFiumPage::Link::Link() {
 472 }
 473
 474 PDFiumPage::Link::~Link() {
 475 }
 476
 477 }  // namespace chrome_pdf