github/CleanTxt

#! /usr/bin/perl -w

# Script to take the output of nroff -man and remove all the backspacing and
# the page footers and the screen commands etc so that it is more usefully
# readable online. In fact, in the latest nroff, intermediate footers don't
# seem to be generated any more.

$blankcount = 0;
$lastwascut = 0;
$firstheader = 1;

# Input on STDIN; output to STDOUT.

while (<STDIN>)
  {
  s/\x1b\[\d+m//g;   # Remove screen controls "ESC [ number m"
  s/.\x8//g;         # Remove "char, backspace"

  # Handle header lines. Retain only the first one we encounter, but remove
  # the blank line that follows. Any others (e.g. at end of document) and the
  # following blank line are dropped.

  if (/^PCRE(\w*)\(([13])\)\s+PCRE\1\(\2\)$/)
    {
    if ($firstheader)
      {
      $firstheader = 0;
      print;
      $lastprinted = $_;
      $lastwascut = 0;
      }
    $_=<STDIN>;       # Remove a blank that follows
    next;
    }

  # Count runs of empty lines

  if (/^\s*$/)
    {
    $blankcount++;
    $lastwascut = 0;
    next;
    }

  # If a chunk of lines has been cut out (page footer) and the next line
  # has a different indentation, put back one blank line.

  if ($lastwascut && $blankcount < 1 && defined($lastprinted))
    {
    ($a) = $lastprinted =~ /^(\s*)/;
    ($b) = $_ =~ /^(\s*)/;
    $blankcount++ if ($a ne $b);
    }

  # We get here only when we have a non-blank line in hand. If it was preceded
  # by 3 or more blank lines, read the next 3 lines and see if they are blank.
  # If so, remove all 7 lines, and remember that we have just done a cut.

  if ($blankcount >= 3)
    {
    for ($i = 0; $i < 3; $i++)
      {
      $next[$i] = <STDIN>;
      $next[$i] = "" if !defined $next[$i];
      $next[$i] =~ s/\x1b\[\d+m//g;   # Remove screen controls "ESC [ number m"
      $next[$i] =~ s/.\x8//g;         # Remove "char, backspace"
      }

    # Cut out chunks of the form <3 blanks><non-blank><3 blanks>

    if ($next[0] =~ /^\s*$/ &&
        $next[1] =~ /^\s*$/ &&
        $next[2] =~ /^\s*$/)
      {
      $blankcount -= 3;
      $lastwascut = 1;
      }

    # Otherwise output the saved blanks, the current, and the next three
    # lines. Remember the last printed line.

    else
      {
      for ($i = 0; $i < $blankcount; $i++) { print "\n"; }
      print;
      for ($i = 0; $i < 3; $i++)
        {
        $next[$i] =~ s/.\x8//g;
        print $next[$i];
        $lastprinted = $_;
        }
      $lastwascut = 0;
      $blankcount = 0;
      }
    }

  # This non-blank line is not preceded by 3 or more blank lines. Output
  # any blanks there are, and the line. Remember it. Force two blank lines
  # before headings.

  else
    {
    $blankcount = 2 if /^\S/ && !/^Last updated/ && !/^Copyright/ &&
      defined($lastprinted);
    for ($i = 0; $i < $blankcount; $i++) { print "\n"; }
    print;
    $lastprinted = $_;
    $lastwascut = 0;
    $blankcount = 0;
    }
  }

# End