word wrap for Graphviz

wordWrapShared.sh

## set -x
## This is a Linux (and maybe MacOS?) shell script.
## It will need to be modified to run on Windows.
## It performs word-wrap (https://en.wikipedia.org/wiki/Line_wrap_and_word_wrap) on the labels of certain Graphviz nodes.
## - not html-like labels
## - not record or Mrecord shaped nodes
## - not edge labels
## - not graph or cluster labels
## - not xlabels
## - it DOES NOT recognize any meaning of TAB character (ASCII 09) or
## of HTML non-breaking space: ( )
## - it DOES try to honor \n \r \l embedded within labels
##
## to use:
## - set (new) maxTextWidth attribute for each node where word-wrap is desired
## maxTextWidth units are points (1/72 of inch)
## - invoke this script with one argument: the name of the Graphviz input file
## - output will be a modified version of the input file, with appropriate
## linebreaks (\n, \r, or \l) inserted in the node labels
##
##
## personal note: this is NOT the "right" way to do word-wrap
## it is a true kludge
## but it seems to work reasonably well
## steve roush
##
E=dot
E=twopi # we want a fast engine
gvpr -c '
  BEGIN{
  node_t thisNode, newNode;
  int i, cnt;
    string nl, rl, ll, Bell, DC1, DC2, DC3, aChar;
    string p1, p2, justChar;
    nl="\\" + "n";
    rl="\\" + "r";
    ll="\\" + "l";
    // Bell, DC1, DC2, and DC3 are unprintable ASCII characters
    //  this program assumes that they will not be part of any label
    //  so we use them to convey formatting purposes
    Bell=sprintf("%c", 7); // ascii Bell (decimal 7)
    DC1=sprintf("%c", 17); // ascii DC1  (decimal 17)
    DC2=sprintf("%c", 18); // ascii DC1  (decimal 18)
    DC3=sprintf("%c", 19); // ascii DC1  (decimal 18)
}
N{
  if (hasAttr($, "maxTextWidth") && $.maxTextWidth!="" && $.maxTextWidth!="0"){
  }else{
    //printf(2, "Note:: no word-wrap for node: %s.  It does not have maxTextWidth attribute.\n", $.name);
    break;
  }
  //HASATTR($, height)
  if (hasAttr($, "height") && $.height!="")
    $.__saveHeight=$.height;
  else
    $.__saveHeight="";
  if (hasAttr($, "width") && $.width!="")
    $.__saveWidth=$.width;
  else
    $.__saveWidth="";
  if (hasAttr($, "label") && $.label!="" && (ishtml($.label)!=1)){
    ////  need to count \ chars before [nrl]
    ////  if the count is odd, replace with \[nrl] Bell nl
    ////  else leave alone    
    cnt=0;
    i=0;
    while (i<length($.label)){
      aChar=substr($.label,i,1);
      if (strcmp(aChar,"\\")==0) {
        cnt++;
      }else{
        if (((cnt%2)==1) && aChar=="[rnl]"){
          p1=substr($.label,0,i+1);
          p2=substr($.label,i+1);
          $.label=p1 + Bell + aChar + nl + p2;
          i+=length(Bell + aChar + nl);
        }
      cnt=0;
      }
      i++;
    }
    //  ugh, we need to find any justification chars (\[rln])
    //  and then propagate that info backwards (toward the front of the text)
    justChar="n";
    for (i=length($.label)-2;i>0;i--){
      aChar=substr($.label,i,1);
      if (strcmp(aChar,Bell)==0) {
        justChar=substr($.label,i+1,1);
      }else if (aChar==" "){
          p1=substr($.label,0,i);
          p2=substr($.label,i+1);
          if (justChar=="n") $.label=p1 + DC1 + p2;
          if (justChar=="l") $.label=p1 + DC2 + p2;
          if (justChar=="r") $.label=p1 + DC3 + p2;
      }
    }
    $.label=gsub($.label,DC1,nl+" "+nl); //  wrap \? around each space
    $.label=gsub($.label,DC2,ll+" "+ll); //  wrap \? around each space
    $.label=gsub($.label,DC3,rl+" "+rl); //  wrap \? around each space
  }
  }
  /***********************************************************
  **** for time being, do not worry about edge labels
  **** or headlabels, taillabels, xlabels, or graph/cluster labels
  *************************************************************/
' $1 | 
$E -G"linelength=999999" -Txdot |
/bin/gvpr -c '
  BEGIN{
  int cnt, i, tmpCnt, just, oldJust, len;
  int maxLineWidth, lineWidth, wordWidth; // units = points
  string Xstr, tok, thisWord, nl, rl, ll, Bell;
  float Xpos, Ypos;
  obj_t thisObj;
  nl="\\" + "n";
  rl="\\" + "r";
  ll="\\" + "l";
  Bell=sprintf("%c", 7); // ascii Bell (decimal 7)
  /////////////////////////////////////////////////
  //  index is (temporarily) broken, so ...
  /////////////////////////////////////////////////
  int myIndex(string Main, string Pattern){
    int _I, _L, _P, _RC ;
    string _SUB;
    _RC=-1;
    _L=length(Main);
    _P=length(Pattern);
    for (_I=0; _I<(_L-_P+1);_I++){
      _SUB=substr(Main,_I,_P);
      if (_SUB==Pattern){
        _RC=_I;
 break;
      }
    }
    return(_RC);
  }
  /////////////////////////////////////////////////
  int adjustCnt(string rawTxt, int cnt){
    string tmpStr;
    int j;
    tmpStr=rawTxt+" "; // pad for substr
    for (j=0;j<cnt;j++){
      if (strcmp(substr(tmpStr,j,2),"\\\\")==0){
        cnt++;
      }
    }
    return cnt;
  }
 ///////////////////////////////////////////////////////////////////
 string getTok(){
    string tstr;
    Xstr=sub(Xstr,"^+( )",""); //  strip leading spaces - ksh pattern
    i=myIndex(Xstr, " ");
    tstr=substr(Xstr,0,i);
    Xstr=substr(Xstr,i);
    Xstr=sub(Xstr,"^+( )",""); //   strip leading spaces - ksh pattern
    return(tstr);
  }
  /////////////////////////////////////////////////////////////////
  void ldrawParse(){
    tok=getTok();
    just=0; //  
    while (tok!=""){
      switch(tok){
      case "F": // font
        tok=getTok(); // font size, discard
        //cnt=1+(int)getTok();  // font name char count (plus leading -)
        cnt=(int)getTok(); // font name char count (plus leading -)
        cnt++;
        tok=substr(Xstr,0,cnt); // font name, discard
        Xstr=substr(Xstr,cnt); // strip fontname from Xstr
        break;
      case "c": // pen color
      case "C": // fill color
        cnt=1+(int)getTok(); // color char count (plus leading -)
        tok=substr(Xstr,0,cnt); // font name, discard
        Xstr=substr(Xstr,cnt); // strip fontname from Xstr
        break;
      case "T": // text
        // by now, we have put each "word" on a different line
        // time to stitch the words into lines
        Xpos=(float)getTok(); // X pos
        Ypos=(float)getTok(); // Y pos
        oldJust=just;
        just=(int)getTok(); // justify
        wordWidth=(int)getTok(); // word width
        cnt=(int)getTok(); // word char count 
        ///////////////////////////////////////////////////////////
        // if string contains //
        // count will be short - length of "//" is 1 !!!!!!
        // we have already taken care of \G \N \E ...
        ///////////////////////////////////////////////////////////
        Xstr=substr(Xstr,1); // chop off leading "-"
        cnt=adjustCnt(Xstr,cnt);
        // now we have adjusted character/byte count
        thisWord=substr(Xstr,0,cnt); // word
        ////////////     time to wrap      ////////////////////
        tmpCnt=wordWidth;
        if (substr(thisWord,0,1)==Bell){
          lineWidth=0;
          thisObj.label+=("\\"+substr(thisWord,1,1));
        }else if ((lineWidth+tmpCnt)<=maxLineWidth){
          lineWidth+=tmpCnt;
          thisObj.label+=thisWord;
        }else{
          // first, strip trailing space at end of previous line
          // ksh pattern matching does not support ^ or $, 
          // so grind it out
          len=length(thisObj.label);
          if(substr(thisObj.label, len-1)==" ")
            thisObj.label=substr(thisObj.label,0,len-1);
          // do not start line with space
          if (thisWord!=" "){
            lineWidth=wordWidth;
            // add justification
            if (oldJust==0) thisObj.label+=(nl+thisWord);
            if (oldJust==-1) thisObj.label+=(ll+thisWord);
            if (oldJust==1) thisObj.label+=(rl+thisWord);
          }
        }
        Xstr=substr(Xstr,cnt); // strip word from Xstr 
        Xstr=sub(Xstr,"^+( )",""); //  strip leading spaces - ksh pattern
        break;
     case "S": // style
        cnt=1+(int)getTok(); // style char count (plus leading -)
        tok=substr(Xstr,0,cnt); // style, discard
        Xstr=substr(Xstr,cnt); // strip style from Xstr
        break;
     case "t": // font characteristics
        cnt=1+(int)getTok(); // style char count (plus leading -)
        tok=substr(Xstr,0,cnt); // style, discard
        Xstr=substr(Xstr,cnt); // strip style from Xstr
        break;
     case "E": //   filled ellipse
     case "e": // unfilled ellipse
        tok=substr(Xstr,0,cnt); // X, discard
        tok=substr(Xstr,0,cnt); // Y, discard
        tok=substr(Xstr,0,cnt); // width, discard
        tok=substr(Xstr,0,cnt); // height, discard
        break;
     case "P": //   filled polygon
     case "p": // unfilled polygon
     case "L": // polyline
     case "B": // unfilled B-spline
     case "b": //   filled polygon
        cnt=1+(int)getTok(); // point (X,Y) count 
        for (i=0;i<cnt;i++){ // delete multiple point (pos) values
          tok=substr(Xstr,0,cnt); // X, discard
          tok=substr(Xstr,0,cnt); // Y, discard
        }
        Xstr=substr(Xstr,cnt); // strip style from Xstr
        break;
     case "I": // external image
        break;
     default:
        break;
    }
    if (length(Xstr)>0){
      tok=getTok();
    }else{
      break;
    }
  }
 }
}
N{
  thisObj=$;
  lineWidth=0;
  maxLineWidth=0; // ??
  if (hasAttr($, "maxTextWidth") && $.maxTextWidth!="" && $.maxTextWidth!="0"){
     maxLineWidth=(int)(72. * maxTextWidth);
  }else{
    printf(2, "Note:: no word-wrap for node: %s.  It does not have maxTextWidth attribute.\n", $.name);
    break;
  }
  if (hasAttr($, "__saveHeight") && $.__saveHeight!="")
    $.height=$.__saveHeight;
  else
    $.height="";
  if (hasAttr($, "__saveWidth") && $.__saveWidth!="")
    $.width=$.__saveWidth;
  else
    $.width="";
 if (hasAttr($, "_ldraw_") && $._ldraw_!=""){
    $.label=""; // we will rebuild
    Xstr=$._ldraw_;
    ldrawParse(); // do heavy lifting
    $._ldraw_="";
  }
}'