#!/usr/bin/perl
#
# Reads standard input, writes to standard output.
#
# The multi-pass structure is necessary for efficiency. If I do the
# whole file in one read, some of the operations in the second pass
# are pathetically slow.
#
# to do:
#   At the very end, after eating all tags,  escape all < and >.
#	tabs => & ?
# 	áè
#


$debug = 1;

%greek = (
	a => alpha,
	b => beta,
	g => gamma, G => Gamma,
	d => delta, D => Delta,
	e => epsilon,
	h => eta,
	q => theta,
	f => phi,
	j => phi,
	i => iota,
	k => kappa,
	l => lambda,
	m => mu,
	n => nu,
	p => pi,
	r => rho,
	t => tau,
	y => psi, Y => Psi,
	u => upsilon, U => Upsilon,
	z => zeta,
	x => xi,
	c => chi,
	w => omega, W => Omega,
	S => Sigma
);

#==================================== first pass ============================================

undef $/; # Perl idiom so we read whole file all at once

open (TEMPOUT,">temp001.tex") or die("Cannot open temp001.tex for output");
while (<>) {

  tr/\r/\n/;

  # ---------------- First, deal with stuff that would keep me from recognizing patterns later:

  # Strip out all newlines that occur inside of brackets.
  # They only seem to occur at the deepest nesting level, so I only
  # code for that possibility.
  for ($j=1; $j<=6; $j++) {
    s/<([^<>]*)\n([^<>]*)>/<$1$2>/g;
  }
  
  # Font size changes, like <C+SIZE 70>, are likely to be wrong in LaTeX.
  # Sometimes I used them to adjust size of superscripts and subscripts by hand.
  # They prevent me from recognizing certain patterns later on, too.
  # They must die.
  s/<C\+SIZE\s[\d+.]*>//g;
  s/<GTABS[^<>]*>//g;
  
  # Deal with stuff like this: <I>f <FONT "Symbol"><P>l
  s/<I>([\w \t\d]+)<FONT "Symbol"><P>/<I>$1<P><FONT "Symbol">/g;
  
  # Escape characters that have special significance in latex. Do this before
  # I start generating my own latex.
  s/\%/\\\%/g;
  s/\#/\\\#/g;
  s/\&/\\\&/g;
  s/\$/\\\$/g;
  s/\^/\\\^/g;
  s/\~/\$\sim\$/g;
  s/_/\\_/g;

  # ---------------- Recognize whole tags:
  #Equations rendered as pictures, e.g. <\& 32 22 9 "PICT14">:
  s/<[^<>]*PICT[^<>]*>//g;
  
  #Index
  s/<\*INDEX\s\d+\s\d+\s\d+\s\d+\s\d+\s\("([^"]+)"\s""\s""\)>/\\index{$1}/g;
  s/<\*INDEX\s\d+\s\d+\s\d+\s\d+\s\d+\s\("([^"]+)"\s""\s""\)\s\("([^"]+)"\s""\s""\)>/\\index{$1!$2}/g;
  
  #Greek
  s/<FONT "Symbol">(<P>)?(.)\s*<FONT "[^"]+">/\$\\greek{$2}\$/g;
	s/\\greek{a}/\\alpha /g;
	s/\\greek{b}/\\beta /g;
	s/\\greek{c}/\\chi /g;
	s/\\greek{D}/\\Delta /g;
	s/\\greek{d}/\\delta /g;
	s/\\greek{e}/\\epsilon /g;
	s/\\greek{f}/\\phi /g;
	s/\\greek{G}/\\Gamma /g;
	s/\\greek{g}/\\gamma /g;
	s/\\greek{h}/\\eta /g;
	s/\\greek{i}/\\iota /g;
	s/\\greek{j}/\\phi /g;
	s/\\greek{k}/\\kappa /g;
	s/\\greek{l}/\\lambda /g;
	s/\\greek{m}/\\mu /g;
	s/\\greek{n}/\\nu /g;
	s/\\greek{p}/\\pi /g;
	s/\\greek{q}/\\theta /g;
	s/\\greek{S}/\\Sigma /g;
	s/\\greek{r}/\\rho /g;
	s/\\greek{U}/\\Upsilon /g;
	s/\\greek{t}/\\tau /g;
	s/\\greek{u}/\\upsilon /g;
	s/\\greek{W}/\\Omega /g;
	s/\\greek{w}/\\omega /g;
	s/\\greek{x}/\\xi /g;
	s/\\greek{Y}/\\Psi /g;
	s/\\greek{y}/\\psi /g;
	s/\\greek{z}/\\zeta /g;

  # symbols
  s/\\greek{¢}/'/g;
  s/\\greek{µ}/\\propto/g;
  s/\$'\$/'/g;
  s/°/\\degunit/g;

  # ----------------- Itty bitties: ------------------------------
  s/<n>/  /g;
  s/<m>/   /g;
  s/\xd5/'/g;
  s/\xd2/``/g;
  s/\xd3/''/g;
  s/\xa1/\\degc/g;
  s/¹/'/g;
  s/³/``/g;
  s/²/''/g;

  # Dashes
  s/\321/---/g;
  s/\320/-/g;

  #scientific notation
  s/10<\+>\s*(\+?)(\d+)\s*<P>/\$10powxxx\{$1$2\}\$ /g;
  s/10<\+>\s*(\-+|<\\\-+>)(\d+)\s*<P>/\$10powxxx\{\-$2\}\$ /g;
  s/powxxx\{(\S)\}/powxxx$1/g;
  s/powxxx/\136/g; # strange
  s/<\\\->/---/g;
  s/<\\\-+>/---/g;

  # superscripts, subscripts, variables in italics, ...
  #First, simplify math expressions to make them easier to recognize and deal with later.
  #This happens /before/ we officially decide whether they're math expressions.
  s/<\+>\s?(\w+)\s?<.>/\$^{$1}\$/g;
  s/<\->\s?(\w+)\s?<.>/\$_{$1}\$/g;
  #s/<P>[\(\)\[\]\/\+\-]<I>/$1/g;
  #Leading or trailing whitespace:
  s/<I>(\s*)([\w \t\+\-\/]+)<P>/$1<I>$2<P>/g;
  s/<I>([\w \t\+\-\/]+[\w\+\-\/])(\s*)<P>/<I>$1<P>$2/g;
  #Try to recognize cases where it's /not/ math:
    #multiple words, space between them, nothing but alpha and punctuation:
    s/<I>([a-zA-Z\.\,\;\'\s]+)([a-zA-Z\.\,\;\']+)(\s*)<P>/\\emph{$1$2}$3/g;
    		# ... doesn't work because of (()) 
    #at least three alphas in a row:
    s/<I>([a-zA-Z\']{3,})<P>/\\emph{$1}/g; 
  #Do it:
  s/<I>\s?(\w+)\s?<(I|P)><\+>(\d)<.>/\$$1^$3\$/g;
  s/<I>\s?(\w+)\s?<(I|P)><\+>(\w+)<.>/\$$1^{$3}\$/g;
  s/<I>\s?(\w+)\s?<(I|P)><\->(\w+)<.>/\$$1_{$3}\$/g;
  s/<I>\s?([^<>]+)\s?<P>/\$$1\$/g;
  
  # ----------------- paragraph breaks ------------------------------
  s/<\\<>[^<>]*>/\n\n/g;
  
  # ----------------- detect displayed equations ------------------------------
  #aligned equations
  s/\t([^=\n]*)=([^=\n]+)\n+/\t$1 = $2 \\\\\n/g;
  
  # --------------- math stuff

  
  # minus sign, not m-dash:
  s/\-\-\-(\d)/-$1/g; 
  s/\-\-\-\$/-\$/g; 
  s/([^\n]*)\-\-\-([^\n]*)\\\\\n/$1-$2\\\\\n/g;
  
  # My known habits:
  s/   ,/   \\qquad   ,/g;
  s/   \./   \\qquad   ./g;
  
  print TEMPOUT $_;
}
close (TEMPOUT);
#==================================== second pass ============================================
$/ = "\n"; # record delimiter
open (TEMPIN,"<temp001.tex") or die("Cannot open temp001.tex for input");
open (TEMPOUT,">temp002.tex") or die("Cannot open temp002.tex for output");
while (<TEMPIN>) {
  #-------------- work more on displayed equations
  

  # Eliminate dollar signs in displayed equations:
  s/(.*)\$(.*)\$(.*)\\\\\n/$1$2$3\\\\\n/g; 
  s/(.*)\$(.*)\$(.*)\\\\\n/$1$2$3\\\\\n/g; 
  s/(.*)\$(.*)\$(.*)\\\\\n/$1$2$3\\\\\n/g; 
  s/(.*)\$(.*)\$(.*)\\\\\n/$1$2$3\\\\\n/g; 
  

  print TEMPOUT $_;  
}
close (TEMPIN);
close (TEMPOUT);
#==================================== third pass ============================================
undef $/;
open (TEMPIN,"<temp002.tex") or die("Cannot open temp002.tex for input");
open (TEMPOUT,">temp003.tex") or die("Cannot open temp003.tex for output");
while (<TEMPIN>) {
  #------------------- sectioning
  #<@Chapter title:><\<>Chapter title>0   Introduction and Review
  s/<[^<>]+><\\<>Chapter title>(\d+)?(\s+)?([^<>]+)\n/\\chapter\{$3\}\n/g;
  s/<\@Chapter title:>//g;
  s/<[^<>]+><\\<>Section title>(\d+)?(\s+)?([^<>]+)\n/\\section\{$3\}\n/g;
  s/<\@Section title:>//g;
  s/<[^<>]+><\\<>Subsection title>(\d+)?(\s+)?([^<>]+)\n/\\subsection\{$3\}\n/g;
  s/<\@Subsection title:>//g;
  
  # --------------------------Eliminate nearly all tags not dealt with so far:
  s/<[^@\\<>][^<>]*>//g;
  s/<\@:>//g; 					# anonymous paragraph styles
  s/<\@[^<>=]+=(<\@[^<>]+>)*>\n+//g; 	# style definitions
  
  # --------------------------displayed equations
  s/(.*)\\\\\n/\\begin\{equation\*\}\n\t$1 \\\\\n\\end\{equation\*\}\n/g;
  s/\\end\{equation\*\}\s*\\begin\{equation\*\}//g;
  s/\\\\\n\n/\\\\\n/g;
  s/\n{2,}\\begin\{equation\*\}/\n\\begin\{equation\*\}/g;
  s/\\begin\{equation\*\}\n((.*\\\\\n){2,})\\end\{equation\*\}/\\begin\{align\*\}\n$1\\end\{align\*\}/g;
  s/\[([^\[\]]+)\]\s*\\\\/   \\qquad  \\text\{\[$1\]\}  \\\\/g;
  s/\\\\\n\\end\{/\n\\end\{/g;
  
  # ----------------- Clean up math ------------------------------
  # Beautify scientific notation, e.g. 0.5x$10^{-3}$
  s/(\-?)([\d\.]+)x\$10\^/\$$1$2\\times10^/g;

  s/_o/_\\zu\{o\}/g;

  s/\$[ \t\d\.]*\$//g;
  s/\$[ \t\d\.]*=[ \t\d\.]*\$/=/g;
  
  # dollar signs with nothing between them but a math symbol:
  s/\$\s?([\+\-\/\=\(\)\,\|])\s?\$/$1/g;
  
  
  # Simplify one-character subscripts and superscripts:
  s/\^\{(.)\}/^$1/g;
  s/_\{(.)\}/_$1/g;
  
  # operators:
  s/(cos|sin|ln|exp|tan)([^a-z])/\\$1 $2/g;
    
  # Inch-marks to quotes.
  # Do this after all tags are dealt with, because tags have inch-mark quotes sometimes.
  s/"([^"]*)"/``$1''/g;
  
  
  # ------------------Eliminate all remaining tags
  #<@Body text, no indent:>
  s/<[^<>]*>//g;
  
  s/\n\n\n/\n\n/g;
  print TEMPOUT $_;  
}
close (TEMPIN);
close (TEMPOUT);
#==================================== fourth pass ============================================
$/ = "\n";
open (TEMPIN,"<temp003.tex") or die("Cannot open temp003.tex for input");
open (TEMPOUT,">-") or die("Cannot open stdout for output");
$in_align = 0;
$in_eqn = 0;
while (<TEMPIN>) {
  if (m/\\begin\{align\*\}/) {$in_align = 1; $in_eqn = 1;}
  if (m/\\end\{align\*\}/) {$in_align = 0; $in_eqn = 0;}
  if (m/\\begin\{equation\*\}/) {$in_align = 0; $in_eqn = 1;}
  if (m/\\end\{equation\*\}/) {$in_align = 0; $in_eqn = 0;}
  if ($in_align) {
    s/=/&=/g;
  }
  
  s/ _/_/g;
  
  if (!$in_eqn) {
  # catch variables that weren't in italics
  s/([\s\d])([qwertyuiopsdfghjklzxcvbnmQWERTYUOPSDFGHJKLZXCVBNM])([\s\,])/$1\$$2\$$3/g;
  # Similar, but avoid misinterpreting abbreviations like "i.e.":
  s/([\s\d])([qwertyuiopsdfghjklzxcvbnmQWERTYUOPSDFGHJKLZXCVBNM])(\.\s)/$1\$$2\$$3/g;
  
  
  # shuffle dollar signs around
   s/\s([\da-zA-Z\+\-\_\^\/]+)\$([^\$]+)\$/ \$$1$2\$/g;
   s/\$([^\$]+)\$([\da-zA-Z\+\-\_\^\/]+)\s/\$$1$2\$ /g;
  s/_(.)_(.)/_\{$1$2\}/g;
  s/([a-zA-Z])\$_/\$$1_/g;
  #s/\$([^\$]*)\$([^ \$]*)\$([^\$]*)\$/\$$1$2$3\$/g;# eat 4 dollar signs -> 2
  	# ... wrong, because allows blanks, could combine unrelated expressions
  
  s/\$([^ .,\$]*)\$([^ .,\$]*)/\$$1$2\$/g;
  s/([^ .,\$]*)\$([^ .,\$]*)\$/\$$1$2\$/g;
  
  s/\$([^ \$]*)\$([^ \$]*)\$([^ \$]*)\$/\$$1$2$3\$/g; # eat 4 dollar signs -> 2
  
  s/ ([^ \$]*)=([^ \$]*) / \$$1=$2\$ /g;
  
  s/,\$/\$,/g;
  s/\.\$/\$\./g;
  
  s/\$\s*\$//g;
  s/\$=\$//g;
  }
  
  if ($in_eqn) {
    s/\$//g;
  }
  
  $count = 0;
  while (m/.{70,}/ && $count<50) {
    s/(.{50,60}) (.{20,})/$1\n$2/;
    ++$count;
  }
  
  print TEMPOUT $_;  
}
close (TEMPIN);
close (TEMPOUT);