textoutputdev.cc

source code: Covert TXT to PDF

//========================================================================
//
// TextOutputDev.cc
//
// Copyright 1997-2002 Glyph & Cog, LLC
//
//========================================================================

#include <aconf.h>

#ifdef USE_GCC_PRAGMAS
#pragma implementation
#endif

#include <stdio.h>
#include <stdlib.h>
#include <stddef.h>
#include <math.h>
#include <ctype.h>
#include "gmem.h"
#include "GString.h"
#include "GList.h"
#include "config.h"
#include "Error.h"
#include "GlobalParams.h"
#include "UnicodeMap.h"
#include "GfxState.h"
#include "TextOutputDev.h"

#ifdef MACOS
// needed for setting type/creator of MacOS files
#include "ICSupport.h"
#endif

//------------------------------------------------------------------------
// parameters
//------------------------------------------------------------------------

// Minium and maximum inter-word spacing (as a fraction of the average
// character width).
#define wordMinSpaceWidth 0.3
#define wordMaxSpaceWidth 2.0

// Default min and max inter-word spacing (when the average character
// width is unknown).
#define wordDefMinSpaceWidth 0.2
#define wordDefMaxSpaceWidth 1.5

// Max difference in x,y coordinates (as a fraction of the font size)
// allowed for duplicated text (fake boldface, drop shadows) which is
// to be discarded.
#define dupMaxDeltaX 0.2
#define dupMaxDeltaY 0.2

// Min overlap (as a fraction of the font size) required for two
// lines to be considered vertically overlapping.
#define lineOverlapSlack 0.5

// Max difference in baseline y coordinates (as a fraction of the font
// size) allowed for words which are to be grouped into a line, not
// including sub/superscripts.
#define lineMaxBaselineDelta 0.1

// Max ratio of font sizes allowed for words which are to be grouped
// into a line, not including sub/superscripts.
#define lineMaxFontSizeRatio 1.4

// Min spacing (as a fraction of the font size) allowed between words
// which are to be grouped into a line.
#define lineMinDeltaX -0.5

// Minimum vertical overlap (as a fraction of the font size) required
// for superscript and subscript words.
#define lineMinSuperscriptOverlap 0.3
#define lineMinSubscriptOverlap   0.3

// Min/max ratio of font sizes allowed for sub/superscripts compared to
// the base text.
#define lineMinSubscriptFontSizeRatio   0.4
#define lineMaxSubscriptFontSizeRatio   1.01
#define lineMinSuperscriptFontSizeRatio 0.4
#define lineMaxSuperscriptFontSizeRatio 1.01

// Max horizontal spacing (as a fraction of the font size) allowed
// before sub/superscripts.
#define lineMaxSubscriptDeltaX   0.2
#define lineMaxSuperscriptDeltaX 0.2

// Maximum vertical spacing (as a fraction of the font size) allowed
// for lines which are to be grouped into a block.
#define blkMaxSpacing 2.0

// Max ratio of primary font sizes allowed for lines which are to be
// grouped into a block.
#define blkMaxFontSizeRatio 1.3

// Min overlap (as a fraction of the font size) required for two
// blocks to be considered vertically overlapping.
#define blkOverlapSlack 0.5

// Max vertical spacing (as a fraction of the font size) allowed
// between blocks which are 'adjacent' when sorted by reading order.
#define blkMaxSortSpacing 2.0

// Max vertical offset (as a fraction of the font size) of the top and
// bottom edges allowed for blocks which are to be grouped into a
// flow.
#define flowMaxDeltaY 1.0

//------------------------------------------------------------------------
// TextFontInfo
//------------------------------------------------------------------------

TextFontInfo::TextFontInfo(GfxState *state) {
  double *textMat;
  double t1, t2, avgWidth, w;
  int n, i;

  gfxFont = state->getFont();
  textMat = state->getTextMat();
  horizScaling = state->getHorizScaling();
  if ((t1 = fabs(textMat[0])) > 0.01 &&
      (t2 = fabs(textMat[3])) > 0.01) {
    horizScaling *= t1 / t2;
  }

  if (!gfxFont) {
    minSpaceWidth = horizScaling * wordDefMinSpaceWidth;
    maxSpaceWidth = horizScaling * wordDefMaxSpaceWidth;
  } else if (gfxFont->isCIDFont()) {
    //~ handle 16-bit fonts
    minSpaceWidth = horizScaling * wordDefMinSpaceWidth;
    maxSpaceWidth = horizScaling * wordDefMaxSpaceWidth;
  } else {
    avgWidth = 0;
    n = 0;
    for (i = 0; i < 256; ++i) {
      w = ((Gfx8BitFont *)gfxFont)->getWidth(i);
      if (w > 0) {
	avgWidth += w;
	++n;
      }
    }
    avgWidth /= n;
    minSpaceWidth = horizScaling * wordMinSpaceWidth * avgWidth;
    maxSpaceWidth = horizScaling * wordMaxSpaceWidth * avgWidth;
  }

}

TextFontInfo::~TextFontInfo() {
}

GBool TextFontInfo::matches(GfxState *state) {
  double *textMat;
  double t1, t2, h;

  textMat = state->getTextMat();
  h = state->getHorizScaling();
  if ((t1 = fabs(textMat[0])) > 0.01 &&
      (t2 = fabs(textMat[3])) > 0.01) {
    h *= t1 / t2;
  }
  return state->getFont() == gfxFont &&
         fabs(h - horizScaling) < 0.01;
}

//------------------------------------------------------------------------
// TextWord
//------------------------------------------------------------------------

TextWord::TextWord(GfxState *state, double x0, double y0,
		   TextFontInfo *fontA, double fontSizeA) {
  GfxFont *gfxFont;
  double x, y;

  font = fontA;
  fontSize = fontSizeA;
  state->transform(x0, y0, &x, &y);
  if ((gfxFont = font->gfxFont)) {
    yMin = y - gfxFont->getAscent() * fontSize;
    yMax = y - gfxFont->getDescent() * fontSize;
  } else {
    // this means that the PDF file draws text without a current font,
    // which should never happen
    yMin = y - 0.95 * fontSize;
    yMax = y + 0.35 * fontSize;
  }
  if (yMin == yMax) {
    // this is a sanity check for a case that shouldn't happen -- but
    // if it does happen, we want to avoid dividing by zero later
    yMin = y;
    yMax = y + 1;
  }
  yBase = y;
  text = NULL;
  xRight = NULL;
  len = size = 0;
  spaceAfter = gFalse;
  next = NULL;

}


TextWord::~TextWord() {
  gfree(text);
  gfree(xRight);
}

void TextWord::addChar(GfxState *state, double x, double y,
		       double dx, double dy, Unicode u) {
  if (len == size) {
    size += 16;
    text = (Unicode *)grealloc(text, size * sizeof(Unicode));
    xRight = (double *)grealloc(xRight, size * sizeof(double));
  }
  text[len] = u;
  if (len == 0) {
    xMin = x;
  }
  xMax = xRight[len] = x + dx;
  ++len;
}

// Returns true if <this> comes before <word2> in xy order.
GBool TextWord::xyBefore(TextWord *word2) {
  return xMin < word2->xMin ||
	 (xMin == word2->xMin && yMin < word2->yMin);
}

// Merge another word onto the end of this one.
void TextWord::merge(TextWord *word2) {
  int i;

  xMax = word2->xMax;
  if (word2->yMin < yMin) {
    yMin = word2->yMin;
  }
  if (word2->yMax > yMax) {
    yMax = word2->yMax;
  }
  if (len + word2->len > size) {
    size = len + word2->len;
    text = (Unicode *)grealloc(text, size * sizeof(Unicode));
    xRight = (double *)grealloc(xRight, size * sizeof(double));
  }
  for (i = 0; i < word2->len; ++i) {
    text[len + i] = word2->text[i];
    xRight[len + i] = word2->xRight[i];
  }
  len += word2->len;
}

//------------------------------------------------------------------------
// TextLine
//------------------------------------------------------------------------

TextLine::TextLine() {
  words = NULL;
  text = NULL;
  xRight = NULL;
  col = NULL;
  len = 0;
  hyphenated = gFalse;
  pageNext = NULL;
  next = NULL;
  flowNext = NULL;
}

TextLine::~TextLine() {
  TextWord *w1, *w2;

  for (w1 = words; w1; w1 = w2) {
    w2 = w1->next;
    delete w1;
  }
  gfree(text);
  gfree(xRight);
  gfree(col);
}

// Returns true if <this> comes before <line2> in yx order, allowing
// slack for vertically overlapping lines.
GBool TextLine::yxBefore(TextLine *line2) {
  double dy;

  dy = lineOverlapSlack * fontSize;

  // non-overlapping case
  if (line2->yMin > yMax - dy ||
      line2->yMax < yMin + dy) {
    return yMin < line2->yMin ||
           (yMin == line2->yMin && xMin < line2->xMin);
  }

  // overlapping case
  return xMin < line2->xMin;
}

// Merge another line's words onto the end of this line.
void TextLine::merge(TextLine *line2) {
  TextWord *word;
  int newLen, i;

  xMax = line2->xMax;
  if (line2->yMin < yMin) {
    yMin = line2->yMin;
  }
  if (line2->yMax > yMax) {
    yMax = line2->yMax;
  }
  xSpaceR = line2->xSpaceR;
  for (word = words; word->next; word = word->next) ;
  word->spaceAfter = gTrue;
  word->next = line2->words;
  line2->words = NULL;
  newLen = len + 1 + line2->len;
  text = (Unicode *)grealloc(text, newLen * sizeof(Unicode));
  xRight = (double *)grealloc(xRight, newLen * sizeof(double));
  text[len] = (Unicode)0x0020;
  xRight[len] = line2->xMin;
  for (i = 0; i < line2->len; ++i) {
    text[len + 1 + i] = line2->text[i];
    xRight[len + 1 + i] = line2->xRight[i];
  }
  len = newLen;
  convertedLen += line2->convertedLen;
  hyphenated = line2->hyphenated;
}

//------------------------------------------------------------------------
// TextBlock
//------------------------------------------------------------------------

TextBlock::TextBlock() {
  lines = NULL;
  next = NULL;
}

TextBlock::~TextBlock() {
  TextLine *l1, *l2;

  for (l1 = lines; l1; l1 = l2) {
    l2 = l1->next;
    delete l1;
  }
}

// Returns true if <this> comes before <blk2> in xy order, allowing
// slack for vertically overlapping blocks.
GBool TextBlock::yxBefore(TextBlock *blk2) {
  double dy;

  dy = blkOverlapSlack * lines->fontSize;

  // non-overlapping case
  if (blk2->yMin > yMax - dy ||
      blk2->yMax < yMin + dy) {
    return yMin < blk2->yMin ||
           (yMin == blk2->yMin && xMin < blk2->xMin);
  }

  // overlapping case
  return xMin < blk2->xMin;
}

// Merge another block's line onto the right of this one.
void TextBlock::mergeRight(TextBlock *blk2) {
  lines->merge(blk2->lines);
  xMax = lines->xMax;
  yMin = lines->yMin;
  yMax = lines->yMax;
  xSpaceR = lines->xSpaceR;
}

// Merge another block's lines onto the bottom of this block.
void TextBlock::mergeBelow(TextBlock *blk2) {
  TextLine *line;

  if (blk2->xMin < xMin) {
    xMin = blk2->xMin;
  }
  if (blk2->xMax > xMax) {
    xMax = blk2->xMax;
  }
  yMax = blk2->yMax;
  if (blk2->xSpaceL > xSpaceL) {
    xSpaceL = blk2->xSpaceL;
  }
  if (blk2->xSpaceR < xSpaceR) {
    xSpaceR = blk2->xSpaceR;
  }
  if (blk2->maxFontSize > maxFontSize) {
    maxFontSize = blk2->maxFontSize;
  }
  for (line = lines; line->next; line = line->next) ;
  line->next = line->flowNext = blk2->lines;
  blk2->lines = NULL;
}

//------------------------------------------------------------------------
// TextFlow
//------------------------------------------------------------------------

TextFlow::TextFlow() {
  blocks = NULL;
  next = NULL;
}

TextFlow::~TextFlow() {
  TextBlock *b1, *b2;

  for (b1 = blocks; b1; b1 = b2) {
    b2 = b1->next;
    delete b1;
  }
}


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

TextPage::TextPage(GBool rawOrderA) {
  rawOrder = rawOrderA;
  curWord = NULL;
  font = NULL;
  fontSize = 0;
  nest = 0;
  nTinyChars = 0;
  words = wordPtr = NULL;
  lines = NULL;
  flows = NULL;
  fonts = new GList();
}

TextPage::~TextPage() {
  clear();
  delete fonts;
}

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

  // adjust the font size
  gfxFont = state->getFont();
  fontSize = 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
      fontSize *= 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
      fontSize *= w / 0.5;
    } else if (anyCode >= 0 &&
	       (w = ((Gfx8BitFont *)gfxFont)->getWidth(anyCode)) > 0) {
      // better than nothing: 0.5 is a generic character width
      fontSize *= w / 0.5;
    }
    fm = gfxFont->getFontMatrix();
    if (fm[0] != 0) {
      fontSize *= fabs(fm[3] / fm[0]);
    }
  }
}

void TextPage::beginWord(GfxState *state, double x0, double y0) {
  // This check is needed because Type 3 characters can contain
  // text-drawing operations (when TextPage is being used via
  // XOutputDev rather than TextOutputDev).
  if (curWord) {
    ++nest;
    return;
  }