textoutputdev.cc

将pdf文档转换为高质量的html文档

    }
  }
}

TextWordList::~TextWordList() {
  delete words;
}

int TextWordList::getLength() {
  return words->getLength();
}

TextWord *TextWordList::get(int idx) {
  if (idx < 0 || idx >= words->getLength()) {
    return NULL;
  }
  return (TextWord *)words->get(idx);
}

#endif // TEXTOUT_WORD_LIST

//------------------------------------------------------------------------
// TextPage
//------------------------------------------------------------------------

TextPage::TextPage(GBool rawOrderA) {
  int rot;

  rawOrder = rawOrderA;
  curWord = NULL;
  charPos = 0;
  curFont = NULL;
  curFontSize = 0;
  nest = 0;
  nTinyChars = 0;
  lastCharOverlap = gFalse;
  if (!rawOrder) {
    for (rot = 0; rot < 4; ++rot) {
      pools[rot] = new TextPool();
    }
  }
  flows = NULL;
  blocks = NULL;
  rawWords = NULL;
  rawLastWord = NULL;
  fonts = new GList();
  lastFindXMin = lastFindYMin = 0;
  haveLastFind = gFalse;
}

TextPage::~TextPage() {
  int rot;

  clear();
  if (!rawOrder) {
    for (rot = 0; rot < 4; ++rot) {
      delete pools[rot];
    }
  }
  delete fonts;
}

void TextPage::startPage(GfxState *state) {
  clear();
  if (state) {
    pageWidth = state->getPageWidth();
    pageHeight = state->getPageHeight();
  } else {
    pageWidth = pageHeight = 0;
  }
}

void TextPage::endPage() {
  if (curWord) {
    endWord();
  }
}

void TextPage::clear() {
  int rot;
  TextFlow *flow;
  TextWord *word;

  if (curWord) {
    delete curWord;
    curWord = NULL;
  }
  if (rawOrder) {
    while (rawWords) {
      word = rawWords;
      rawWords = rawWords->next;
      delete word;
    }
  } else {
    for (rot = 0; rot < 4; ++rot) {
      delete pools[rot];
    }
    while (flows) {
      flow = flows;
      flows = flows->next;
      delete flow;
    }
    gfree(blocks);
  }
  deleteGList(fonts, TextFontInfo);

  curWord = NULL;
  charPos = 0;
  curFont = NULL;
  curFontSize = 0;
  nest = 0;
  nTinyChars = 0;
  if (!rawOrder) {
    for (rot = 0; rot < 4; ++rot) {
      pools[rot] = new TextPool();
    }
  }
  flows = NULL;
  blocks = NULL;
  rawWords = NULL;
  rawLastWord = NULL;
  fonts = new GList();
}

void TextPage::updateFont(GfxState *state) {
  GfxFont *gfxFont;
  double *fm;
  char *name;
  int code, mCode, letterCode, anyCode;
  double w;
  int i;

  // get the font info object
  curFont = NULL;
  for (i = 0; i < fonts->getLength(); ++i) {
    curFont = (TextFontInfo *)fonts->get(i);
    if (curFont->matches(state)) {
      break;
    }
    curFont = NULL;
  }
  if (!curFont) {
    curFont = new TextFontInfo(state);
    fonts->append(curFont);
  }

  // adjust the font size
  gfxFont = state->getFont();
  curFontSize = state->getTransformedFontSize();
  if (gfxFont && gfxFont->getType() == fontType3) {
    // This is a hack which makes it possible to deal with some Type 3
    // fonts.  The problem is that it's impossible to know what the
    // base coordinate system used in the font is without actually
    // rendering the font.  This code tries to guess by looking at the
    // width of the character 'm' (which breaks if the font is a
    // subset that doesn't contain 'm').
    mCode = letterCode = anyCode = -1;
    for (code = 0; code < 256; ++code) {
      name = ((Gfx8BitFont *)gfxFont)->getCharName(code);
      if (name && name[0] == 'm' && name[1] == '\0') {
	mCode = code;
      }
      if (letterCode < 0 && name && name[1] == '\0' &&
	  ((name[0] >= 'A' && name[0] <= 'Z') ||
	   (name[0] >= 'a' && name[0] <= 'z'))) {
	letterCode = code;
      }
      if (anyCode < 0 && name &&
	  ((Gfx8BitFont *)gfxFont)->getWidth(code) > 0) {
	anyCode = code;
      }
    }
    if (mCode >= 0 &&
	(w = ((Gfx8BitFont *)gfxFont)->getWidth(mCode)) > 0) {
      // 0.6 is a generic average 'm' width -- yes, this is a hack
      curFontSize *= w / 0.6;
    } else if (letterCode >= 0 &&
	       (w = ((Gfx8BitFont *)gfxFont)->getWidth(letterCode)) > 0) {
      // even more of a hack: 0.5 is a generic letter width
      curFontSize *= w / 0.5;
    } else if (anyCode >= 0 &&
	       (w = ((Gfx8BitFont *)gfxFont)->getWidth(anyCode)) > 0) {
      // better than nothing: 0.5 is a generic character width
      curFontSize *= w / 0.5;
    }
    fm = gfxFont->getFontMatrix();
    if (fm[0] != 0) {
      curFontSize *= fabs(fm[3] / fm[0]);
    }
  }
}

void TextPage::beginWord(GfxState *state, double x0, double y0) {
  double *fontm;
  double m[4], m2[4];
  int rot;

  // This check is needed because Type 3 characters can contain
  // text-drawing operations (when TextPage is being used via
  // {X,Win}SplashOutputDev rather than TextOutputDev).
  if (curWord) {
    ++nest;
    return;
  }

  // compute the rotation
  state->getFontTransMat(&m[0], &m[1], &m[2], &m[3]);
  if (state->getFont()->getType() == fontType3) {
    fontm = state->getFont()->getFontMatrix();
    m2[0] = fontm[0] * m[0] + fontm[1] * m[2];
    m2[1] = fontm[0] * m[1] + fontm[1] * m[3];
    m2[2] = fontm[2] * m[0] + fontm[3] * m[2];
    m2[3] = fontm[2] * m[1] + fontm[3] * m[3];
    m[0] = m2[0];
    m[1] = m2[1];
    m[2] = m2[2];
    m[3] = m2[3];
  }
  if (fabs(m[0] * m[3]) > fabs(m[1] * m[2])) {
    rot = (m[3] < 0) ? 0 : 2;
  } else {
    rot = (m[2] > 0) ? 1 : 3;
  }

  curWord = new TextWord(state, rot, x0, y0, charPos, curFont, curFontSize);
}

void TextPage::addChar(GfxState *state, double x, double y,
		       double dx, double dy,
		       CharCode c, int nBytes, Unicode *u, int uLen) {
  double x1, y1, w1, h1, dx2, dy2, base, sp, delta;
  GBool overlap;
  int i;

  // throw away chars that aren't inside the page bounds
  state->transform(x, y, &x1, &y1);
  if (x1 < 0 || x1 > pageWidth ||
      y1 < 0 || y1 > pageHeight) {
    charPos += nBytes;
    return;
  }

  // subtract char and word spacing from the dx,dy values
  sp = state->getCharSpace();
  if (c == (CharCode)0x20) {
    sp += state->getWordSpace();
  }
  state->textTransformDelta(sp * state->getHorizScaling(), 0, &dx2, &dy2);
  dx -= dx2;
  dy -= dy2;
  state->transformDelta(dx, dy, &w1, &h1);

  // check the tiny chars limit
  if (!globalParams->getTextKeepTinyChars() &&
      fabs(w1) < 3 && fabs(h1) < 3) {
    if (++nTinyChars > 50000) {
      charPos += nBytes;
      return;
    }
  }

  // break words at space character
  if (uLen == 1 && u[0] == (Unicode)0x20) {
    if (curWord) {
      ++curWord->charLen;
    }
    charPos += nBytes;
    endWord();
    return;
  }

  // start a new word if:
  // (1) this character doesn't fall in the right place relative to
  //     the end of the previous word (this places upper and lower
  //     constraints on the position deltas along both the primary
  //     and secondary axes), or
  // (2) this character overlaps the previous one (duplicated text), or
  // (3) the previous character was an overlap (we want each duplicated
  //     character to be in a word by itself at this stage)
  if (curWord && curWord->len > 0) {
    base = sp = delta = 0; // make gcc happy
    switch (curWord->rot) {
    case 0:
      base = y1;
      sp = x1 - curWord->xMax;
      delta = x1 - curWord->edge[curWord->len - 1];
      break;
    case 1:
      base = x1;
      sp = y1 - curWord->yMax;
      delta = y1 - curWord->edge[curWord->len - 1];
      break;
    case 2:
      base = y1;
      sp = curWord->xMin - x1;
      delta = curWord->edge[curWord->len - 1] - x1;
      break;
    case 3:
      base = x1;
      sp = curWord->yMin - y1;
      delta = curWord->edge[curWord->len - 1] - y1;
      break;
    }
    overlap = fabs(delta) < dupMaxPriDelta * curWord->fontSize &&
              fabs(base - curWord->base) < dupMaxSecDelta * curWord->fontSize;
    if (overlap || lastCharOverlap ||
	sp < -minDupBreakOverlap * curWord->fontSize ||
	sp > minWordBreakSpace * curWord->fontSize ||
	fabs(base - curWord->base) > 0.5) {
      endWord();
    }
    lastCharOverlap = overlap;
  } else {
    lastCharOverlap = gFalse;
  }

  if (uLen != 0) {
    // start a new word if needed
    if (!curWord) {
      beginWord(state, x, y);
    }

    // page rotation and/or transform matrices can cause text to be
    // drawn in reverse order -- in this case, swap the begin/end
    // coordinates and break text into individual chars
    if ((curWord->rot == 0 && w1 < 0) ||
	(curWord->rot == 1 && h1 < 0) ||
	(curWord->rot == 2 && w1 > 0) ||
	(curWord->rot == 3 && h1 > 0)) {
      endWord();
      beginWord(state, x + dx, y + dy);
      x1 += w1;
      y1 += h1;
      w1 = -w1;
      h1 = -h1;
    }

    // add the characters to the current word
    w1 /= uLen;
    h1 /= uLen;
    for (i = 0; i < uLen; ++i) {
      curWord->addChar(state, x1 + i*w1, y1 + i*h1, w1, h1, u[i]);
    }
  }
  if (curWord) {
    curWord->charLen += nBytes;
  }
  charPos += nBytes;
}

void TextPage::endWord() {
  // This check is needed because Type 3 characters can contain
  // text-drawing operations (when TextPage is being used via
  // {X,Win}SplashOutputDev rather than TextOutputDev).
  if (nest > 0) {
    --nest;
    return;
  }

  if (curWord) {
    addWord(curWord);
    curWord = NULL;
  }
}

void TextPage::addWord(TextWord *word) {
  // throw away zero-length words -- they don't have valid xMin/xMax
  // values, and they're useless anyway
  if (word->len == 0) {
    delete word;
    return;
  }

  if (rawOrder) {
    if (rawLastWord) {
      rawLastWord->next = word;
    } else {
      rawWords = word;
    }
    rawLastWord = word;
  } else {
    pools[word->rot]->addWord(word);
  }
}

void TextPage::coalesce(GBool physLayout) {
  UnicodeMap *uMap;
  TextPool *pool;
  TextWord *word0, *word1, *word2;
  TextLine *line;
  TextBlock *blkList, *blkStack, *blk, *lastBlk, *blk0, *blk1;
  TextBlock **blkArray;
  TextFlow *flow, *lastFlow;
  int rot, poolMinBaseIdx, baseIdx, startBaseIdx;
  double minBase, maxBase, newMinBase, newMaxBase;
  double fontSize, colSpace1, colSpace2, lineSpace, intraLineSpace, blkSpace;
  GBool found;
  int count[4];
  int lrCount;
  int firstBlkIdx, nBlocksLeft;
  int col1, col2;
  int i, j, n;

  if (rawOrder) {
    primaryRot = 0;
    primaryLR = gTrue;
    return;
  }

  uMap = globalParams->getTextEncoding();
  blkList = NULL;
  lastBlk = NULL;
  nBlocks = 0;
  primaryRot = -1;

#if 0 // for debugging
  printf("*** initial words ***\n");
  for (rot = 0; rot < 4; ++rot) {
    pool = pools[rot];
    for (baseIdx = pool->minBaseIdx; baseIdx <= pool->maxBaseIdx; ++baseIdx) {
      for (word0 = pool->getPool(baseIdx); word0; word0 = word0->next) {
	printf("    word: x=%.2f..%.2f y=%.2f..%.2f base=%.2f fontSize=%.2f rot=%d '",
	       word0->xMin, word0->xMax, word0->yMin, word0->yMax,
	       word0->base, word0->fontSize, rot*90);
	for (i = 0; i < word0->len; ++i) {
	  fputc(word0->text[i] & 0xff, stdout);
	}
	printf("'\n");
      }
    }
  }
  printf("\n");
#endif

  //----- assemble the blocks

  //~ add an outer loop for writing mode (vertical text)

  // build blocks for each rotation value
  for (rot = 0; rot < 4; ++rot) {
    pool = pools[rot];
    poolMinBaseIdx = pool->minBaseIdx;
    count[rot] = 0;

    // add blocks until no more words are left
    while (1) {

      // find the first non-empty line in the pool
      for (;
	   poolMinBaseIdx <= pool->maxBaseIdx &&
	     !pool->getPool(poolMinBaseIdx);
	   ++poolMinBaseIdx) ;
      if (poolMinBaseIdx > pool->maxBaseIdx) {
	break;
      }

      // look for the left-most word in the first four lines of the
      // pool -- this avoids starting with a superscript word
      startBaseIdx = poolMinBaseIdx;
      for (baseIdx = poolMinBaseIdx + 1;
	   baseIdx < poolMinBaseIdx + 4 && baseIdx <= pool->maxBaseIdx;
	   ++baseIdx) {
	if (!pool->getPool(baseIdx)) {
	  continue;
	}
	if (pool->getPool(baseIdx)->primaryCmp(pool->getPool(startBaseIdx))
	    < 0) {
	  startBaseIdx = baseIdx;
	}
      }

      // create a new block
      word0 = pool->getPool(startBaseIdx);
      pool->setPool(startBaseIdx, word0->next);
      word0->next = NULL;
      blk = new TextBlock(this, rot);
      blk->addWord(word0);

      fontSize = word0->fontSize;
      minBase = maxBase = word0->base;
      colSpace1 = minColSpacing1 * fontSize;
      colSpace2 = minColSpacing2 * fontSize;
      lineSpace = maxLineSpacingDelta * fontSize;
      intraLineSpace = maxIntraLineDelta * fontSize;

      // add words to the block
      do {
	found = gFalse;

	// look for words on the line above the current top edge of
	// the block
	newMinBase = minBase;
	for (baseIdx = pool->getBaseIdx(minBase);
	     baseIdx >= pool->getBaseIdx(minBase - lineSpace);
	     --baseIdx) {
	  word0 = NULL;
	  word1 = pool->getPool(baseIdx);
	  while (word1) {
	    if (word1->base < minBase &&
		word1->base >= minBase - lineSpace &&
		((rot == 0 || rot == 2)
		 ? (word1->xMin < blk->xMax && word1->xMax > blk->xMin)
		 : (word1->yMin < blk->yMax && word1->yMax > blk->yMin)) &&
		fabs(word1->fontSize - fontSize) <
		  maxBlockFontSizeDelta1 * fontSize) {
	      word2 = word1;
	      if (word0) {
		word0->next = word1->next;
	      } else {
		pool->setPool(baseIdx, word1->next);
	      }
	      word1 = word1->next;
	      word2->next = NULL;
	      blk->addWord(word2);
	      found = gTrue;
	      newMinBase = word2->base;
	    } else {
	      word0 = word1;
	      word1 = word1->next;
	    }
	  }
	}