textoutputdev.cc

将pdf文档转换为高质量的html文档

	minBase = newMinBase;

	// look for words on the line below the current bottom edge of
	// the block
	newMaxBase = maxBase;
	for (baseIdx = pool->getBaseIdx(maxBase);
	     baseIdx <= pool->getBaseIdx(maxBase + lineSpace);
	     ++baseIdx) {
	  word0 = NULL;
	  word1 = pool->getPool(baseIdx);
	  while (word1) {
	    if (word1->base > maxBase &&
		word1->base <= maxBase + 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;
	      newMaxBase = word2->base;
	    } else {
	      word0 = word1;
	      word1 = word1->next;
	    }
	  }
	}
	maxBase = newMaxBase;

	// look for words that are on lines already in the block, and
	// that overlap the block horizontally
	for (baseIdx = pool->getBaseIdx(minBase - intraLineSpace);
	     baseIdx <= pool->getBaseIdx(maxBase + intraLineSpace);
	     ++baseIdx) {
	  word0 = NULL;
	  word1 = pool->getPool(baseIdx);
	  while (word1) {
	    if (word1->base >= minBase - intraLineSpace &&
		word1->base <= maxBase + intraLineSpace &&
		((rot == 0 || rot == 2)
		 ? (word1->xMin < blk->xMax + colSpace1 &&
		    word1->xMax > blk->xMin - colSpace1)
		 : (word1->yMin < blk->yMax + colSpace1 &&
		    word1->yMax > blk->yMin - colSpace1)) &&
		fabs(word1->fontSize - fontSize) <
		  maxBlockFontSizeDelta2 * 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;
	    } else {
	      word0 = word1;
	      word1 = word1->next;
	    }
	  }
	}

	// only check for outlying words (the next two chunks of code)
	// if we didn't find anything else
	if (found) {
	  continue;
	}

	// scan down the left side of the block, looking for words
	// that are near (but not overlapping) the block; if there are
	// three or fewer, add them to the block
	n = 0;
	for (baseIdx = pool->getBaseIdx(minBase - intraLineSpace);
	     baseIdx <= pool->getBaseIdx(maxBase + intraLineSpace);
	     ++baseIdx) {
	  word1 = pool->getPool(baseIdx);
	  while (word1) {
	    if (word1->base >= minBase - intraLineSpace &&
		word1->base <= maxBase + intraLineSpace &&
		((rot == 0 || rot == 2)
		 ? (word1->xMax <= blk->xMin &&
		    word1->xMax > blk->xMin - colSpace2)
		 : (word1->yMax <= blk->yMin &&
		    word1->yMax > blk->yMin - colSpace2)) &&
		fabs(word1->fontSize - fontSize) <
		  maxBlockFontSizeDelta3 * fontSize) {
	      ++n;
	      break;
	    }
	    word1 = word1->next;
	  }
	}
	if (n > 0 && n <= 3) {
	  for (baseIdx = pool->getBaseIdx(minBase - intraLineSpace);
	       baseIdx <= pool->getBaseIdx(maxBase + intraLineSpace);
	       ++baseIdx) {
	    word0 = NULL;
	    word1 = pool->getPool(baseIdx);
	    while (word1) {
	      if (word1->base >= minBase - intraLineSpace &&
		  word1->base <= maxBase + intraLineSpace &&
		  ((rot == 0 || rot == 2)
		   ? (word1->xMax <= blk->xMin &&
		      word1->xMax > blk->xMin - colSpace2)
		   : (word1->yMax <= blk->yMin &&
		      word1->yMax > blk->yMin - colSpace2)) &&
		  fabs(word1->fontSize - fontSize) <
		    maxBlockFontSizeDelta3 * fontSize) {
		word2 = word1;
		if (word0) {
		  word0->next = word1->next;
		} else {
		  pool->setPool(baseIdx, word1->next);
		}
		word1 = word1->next;
		word2->next = NULL;
		blk->addWord(word2);
		if (word2->base < minBase) {
		  minBase = word2->base;
		} else if (word2->base > maxBase) {
		  maxBase = word2->base;
		}
		found = gTrue;
		break;
	      } else {
		word0 = word1;
		word1 = word1->next;
	      }
	    }
	  }
	}

	// scan down the right side of the block, looking for words
	// that are near (but not overlapping) the block; if there are
	// three or fewer, add them to the block
	n = 0;
	for (baseIdx = pool->getBaseIdx(minBase - intraLineSpace);
	     baseIdx <= pool->getBaseIdx(maxBase + intraLineSpace);
	     ++baseIdx) {
	  word1 = pool->getPool(baseIdx);
	  while (word1) {
	    if (word1->base >= minBase - intraLineSpace &&
		word1->base <= maxBase + intraLineSpace &&
		((rot == 0 || rot == 2)
		 ? (word1->xMin >= blk->xMax &&
		    word1->xMin < blk->xMax + colSpace2)
		 : (word1->yMin >= blk->yMax &&
		    word1->yMin < blk->yMax + colSpace2)) &&
		fabs(word1->fontSize - fontSize) <
		  maxBlockFontSizeDelta3 * fontSize) {
	      ++n;
	      break;
	    }
	    word1 = word1->next;
	  }
	}
	if (n > 0 && n <= 3) {
	  for (baseIdx = pool->getBaseIdx(minBase - intraLineSpace);
	       baseIdx <= pool->getBaseIdx(maxBase + intraLineSpace);
	       ++baseIdx) {
	    word0 = NULL;
	    word1 = pool->getPool(baseIdx);
	    while (word1) {
	      if (word1->base >= minBase - intraLineSpace &&
		  word1->base <= maxBase + intraLineSpace &&
		  ((rot == 0 || rot == 2)
		   ? (word1->xMin >= blk->xMax &&
		      word1->xMin < blk->xMax + colSpace2)
		   : (word1->yMin >= blk->yMax &&
		      word1->yMin < blk->yMax + colSpace2)) &&
		  fabs(word1->fontSize - fontSize) <
		    maxBlockFontSizeDelta3 * fontSize) {
		word2 = word1;
		if (word0) {
		  word0->next = word1->next;
		} else {
		  pool->setPool(baseIdx, word1->next);
		}
		word1 = word1->next;
		word2->next = NULL;
		blk->addWord(word2);
		if (word2->base < minBase) {
		  minBase = word2->base;
		} else if (word2->base > maxBase) {
		  maxBase = word2->base;
		}
		found = gTrue;
		break;
	      } else {
		word0 = word1;
		word1 = word1->next;
	      }
	    }
	  }
	}

      } while (found);

      //~ need to compute the primary writing mode (horiz/vert) in
      //~ addition to primary rotation

      // coalesce the block, and add it to the list
      blk->coalesce(uMap);
      if (lastBlk) {
	lastBlk->next = blk;
      } else {
	blkList = blk;
      }
      lastBlk = blk;
      count[rot] += blk->charCount;
      if (primaryRot < 0 || count[rot] > count[primaryRot]) {
	primaryRot = rot;
      }
      ++nBlocks;
    }
  }

#if 0 // for debugging 
  printf("*** rotation ***\n");
  for (rot = 0; rot < 4; ++rot) {
    printf("  %d: %6d\n", rot, count[rot]);
  }
  printf("  primary rot = %d\n", primaryRot);
  printf("\n");
#endif

#if 0 // for debugging
  printf("*** blocks ***\n");
  for (blk = blkList; blk; blk = blk->next) {
    printf("block: rot=%d x=%.2f..%.2f y=%.2f..%.2f\n",
	   blk->rot, blk->xMin, blk->xMax, blk->yMin, blk->yMax);
    for (line = blk->lines; line; line = line->next) {
      printf("  line: x=%.2f..%.2f y=%.2f..%.2f base=%.2f\n",
	     line->xMin, line->xMax, line->yMin, line->yMax, line->base);
      for (word0 = line->words; word0; word0 = word0->next) {
	printf("    word: x=%.2f..%.2f y=%.2f..%.2f base=%.2f fontSize=%.2f space=%d: '",
	       word0->xMin, word0->xMax, word0->yMin, word0->yMax,
	       word0->base, word0->fontSize, word0->spaceAfter);
	for (i = 0; i < word0->len; ++i) {
	  fputc(word0->text[i] & 0xff, stdout);
	}
	printf("'\n");
      }
    }
  }
  printf("\n");
#endif

  // determine the primary direction
  lrCount = 0;
  for (blk = blkList; blk; blk = blk->next) {
    for (line = blk->lines; line; line = line->next) {
      for (word0 = line->words; word0; word0 = word0->next) {
	for (i = 0; i < word0->len; ++i) {
	  if (unicodeTypeL(word0->text[i])) {
	    ++lrCount;
	  } else if (unicodeTypeR(word0->text[i])) {
	    --lrCount;
	  }
	}
      }
    }
  }
  primaryLR = lrCount >= 0;

#if 0 // for debugging
  printf("*** direction ***\n");
  printf("lrCount = %d\n", lrCount);
  printf("primaryLR = %d\n", primaryLR);
#endif

  //----- column assignment

  // sort blocks into xy order for column assignment
  blocks = (TextBlock **)gmallocn(nBlocks, sizeof(TextBlock *));
  for (blk = blkList, i = 0; blk; blk = blk->next, ++i) {
    blocks[i] = blk;
  }
  qsort(blocks, nBlocks, sizeof(TextBlock *), &TextBlock::cmpXYPrimaryRot);

  // column assignment
  for (i = 0; i < nBlocks; ++i) {
    blk0 = blocks[i];
    col1 = 0;
    for (j = 0; j < i; ++j) {
      blk1 = blocks[j];
      col2 = 0; // make gcc happy
      switch (primaryRot) {
      case 0:
	if (blk0->xMin > blk1->xMax) {
	  col2 = blk1->col + blk1->nColumns + 3;
	} else if (blk1->xMax == blk1->xMin) {
	  col2 = blk1->col;
	} else {
	  col2 = blk1->col + (int)(((blk0->xMin - blk1->xMin) /
				    (blk1->xMax - blk1->xMin)) *
				   blk1->nColumns);
	}
	break;
      case 1:
	if (blk0->yMin > blk1->yMax) {
	  col2 = blk1->col + blk1->nColumns + 3;
	} else if (blk1->yMax == blk1->yMin) {
	  col2 = blk1->col;
	} else {
	  col2 = blk1->col + (int)(((blk0->yMin - blk1->yMin) /
				    (blk1->yMax - blk1->yMin)) *
				   blk1->nColumns);
	}
	break;
      case 2:
	if (blk0->xMax < blk1->xMin) {
	  col2 = blk1->col + blk1->nColumns + 3;
	} else if (blk1->xMin == blk1->xMax) {
	  col2 = blk1->col;
	} else {
	  col2 = blk1->col + (int)(((blk0->xMax - blk1->xMax) /
				    (blk1->xMin - blk1->xMax)) *
				   blk1->nColumns);
	}
	break;
      case 3:
	if (blk0->yMax < blk1->yMin) {
	  col2 = blk1->col + blk1->nColumns + 3;
	} else if (blk1->yMin == blk1->yMax) {
	  col2 = blk1->col;
	} else {
	  col2 = blk1->col + (int)(((blk0->yMax - blk1->yMax) /
				    (blk1->yMin - blk1->yMax)) *
				   blk1->nColumns);
	}
	break;
      }
      if (col2 > col1) {
	col1 = col2;
      }
    }
    blk0->col = col1;
    for (line = blk0->lines; line; line = line->next) {
      for (j = 0; j <= line->len; ++j) {
	line->col[j] += col1;
      }
    }
  }

#if 0 // for debugging
  printf("*** blocks, after column assignment ***\n");
  for (blk = blkList; blk; blk = blk->next) {
    printf("block: rot=%d x=%.2f..%.2f y=%.2f..%.2f col=%d nCols=%d\n",
	   blk->rot, blk->xMin, blk->xMax, blk->yMin, blk->yMax, blk->col,
	   blk->nColumns);
    for (line = blk->lines; line; line = line->next) {
      printf("  line:\n");
      for (word0 = line->words; word0; word0 = word0->next) {
	printf("    word: x=%.2f..%.2f y=%.2f..%.2f base=%.2f fontSize=%.2f space=%d: '",
	       word0->xMin, word0->xMax, word0->yMin, word0->yMax,
	       word0->base, word0->fontSize, word0->spaceAfter);
	for (i = 0; i < word0->len; ++i) {
	  fputc(word0->text[i] & 0xff, stdout);
	}
	printf("'\n");
      }
    }
  }
  printf("\n");
#endif

  //----- reading order sort

  // sort blocks into yx order (in preparation for reading order sort)
  qsort(blocks, nBlocks, sizeof(TextBlock *), &TextBlock::cmpYXPrimaryRot);

  // compute space on left and right sides of each block
  for (i = 0; i < nBlocks; ++i) {
    blk0 = blocks[i];
    for (j = 0; j < nBlocks; ++j) {
      blk1 = blocks[j];
      if (blk1 != blk0) {
	blk0->updatePriMinMax(blk1);
      }
    }
  }

#if 0 // for debugging
  printf("*** blocks, after yx sort ***\n");
  for (i = 0; i < nBlocks; ++i) {
    blk = blocks[i];
    printf("block: rot=%d x=%.2f..%.2f y=%.2f..%.2f space=%.2f..%.2f\n",
	   blk->rot, blk->xMin, blk->xMax, blk->yMin, blk->yMax,
	   blk->priMin, blk->priMax);
    for (line = blk->lines; line; line = line->next) {
      printf("  line:\n");
      for (word0 = line->words; word0; word0 = word0->next) {
	printf("    word: x=%.2f..%.2f y=%.2f..%.2f base=%.2f fontSize=%.2f space=%d: '",
	       word0->xMin, word0->xMax, word0->yMin, word0->yMax,
	       word0->base, word0->fontSize, word0->spaceAfter);
	for (j = 0; j < word0->len; ++j) {
	  fputc(word0->text[j] & 0xff, stdout);
	}
	printf("'\n");
      }
    }
  }
  printf("\n");
#endif

  // build the flows
  //~ this needs to be adjusted for writing mode (vertical text)
  //~ this also needs to account for right-to-left column ordering
  blkArray = (TextBlock **)gmallocn(nBlocks, sizeof(TextBlock *));
  memcpy(blkArray, blocks, nBlocks * sizeof(TextBlock *));
  flows = lastFlow = NULL;
  firstBlkIdx = 0;
  nBlocksLeft = nBlocks;
  while (nBlocksLeft > 0) {

    // find the upper-left-most block
    for (; !blkArray[firstBlkIdx]; ++firstBlkIdx) ;
    i = firstBlkIdx;
    blk = blkArray[i];
    for (j = firstBlkIdx + 1; j < nBlocks; ++j) {
      blk1 = blkArray[j];
      if (blk1) {
	if (blk && blk->secondaryDelta(blk1) > 0) {
	  break;
	}
	if (blk1->primaryCmp(blk) < 0) {
	  i = j;
	  blk = blk1;
	}
      }
    }
    blkArray[i] = NULL;
    --nBlocksLeft;
    blk->next = NULL;

    // create a new flow, starting with the upper-left-most block
    flow = new TextFlow(this, blk);
    if (lastFlow) {
      lastFlow->next = flow;
    } else {
      flows = flow;
    }
    lastFlow = flow;
    fontSize = blk->lines->words->fontSize;

    // push the upper-left-most block on the stack
    blk->stackNext = NULL;
    blkStack = blk;

    // find the other blocks in this flow
    while (blkStack) {

      // find the upper-left-most block under (but within
      // maxBlockSpacing of) the top block on the stack
      blkSpace = maxBlockSpacing * blkStack->lines->words->fontSize;
      blk = NULL;
      i = -1;
      for (j = firstBlkIdx; j < nBlocks; ++j) {
	blk1 = blkArray[j];
	if (blk1) {
	  if (blkStack->secondaryDelta(blk1) > blkSpace) {
	    break;
	  }
	  if (blk && blk->secondaryDelta(blk1) > 0) {
	    break;
	  }
	  if (blk1->isBelow(blkStack) &&
	      (!blk || blk1->primaryCmp(blk) < 0)) {
	    i = j;
	    blk = blk1;
	  }
	}
      }

      // if a suitable block was found, add it to the flow and push it
      // onto the stack
      if (blk && flow->blockFits(blk, blkStack)) {
	blkArray[i] = NULL;
	--nBlocksLeft;
	blk->next = NULL;
	flow->addBlock(blk);
	fontSize = blk->lines->words->fontSize;
	blk->stackNext = blkStack;
	blkStack = blk;

      // otherwise (if there is no block under the top block or the
      // block is not suitable), pop the stack
      } else {
	blkStack = blkStack->stackNext;
      }
    }
  }
  gfree(blkArray);

#if 0 // for debugging
  printf("*** flows ***\n");
  for (flow = flows; flow; flow = flow->next) {
    printf("flow: x=%.2f..%.2f y=%.2f..%.2f pri:%.2f..%.2f\n",
	   flow->xMin, flow->xMax, flow->yMin, flow->yMax,
	   flow->priMin, flow->priMax);
    for (blk = flow->blocks; blk; blk = blk->next) {
      printf("  block: rot=%d x=%.2f..%.2f y=%.2f..%.2f pri=%.2f..%.2f\n",
	     blk->rot, blk->xMin, blk->xMax, blk->yMin, blk->yMax,
	     blk->priMin, blk->priMax);
      for (line = blk->lines; line; line = line->next) {
	printf("    line:\n");
	for (word0 = line->words; word0; word0 = word0->next) {
	  printf("      word: x=%.2f..%.2f y=%.2f..%.2f base=%.2f fontSize=%.2f space=