#!/usr/bin/env perl
#
# pkg-repgen: generates a binary repository for pkg-get
#
# requires prt-get
#
# html index generation code adapted from Jukka Heino's portspage
#
# usage: pkg-repgen [options] [directory [pkgname1..pkgnameN]]
#

use warnings;
use strict;
use Getopt::Long;
use Digest::file qw(digest_file_hex);

our $prtget = "/usr/bin/prt-get"; our $prtdir;
our $title = "CRUX Packages"; our $header; our $footer;
GetOptions("prtdir=s"=>\$prtdir, "title=s"=>\$title, "header=s"=>\$header, "footer=s"=>\$footer);

# Use compression-mode defined in pkgmk-conf
our $compress = "gz";
open CONFIG, "/etc/pkgmk.conf" or die "Could not open /etc/pkgmk.conf";
while (<CONFIG>) {
    $compress = $1 if m/^PKGMK_COMPRESSION_MODE=(.*)(\#|$)/;
}
close CONFIG;
$compress =~ s/["' ]//g;

$prtget .= " --no-std-config --config-set=\"prtdir $prtdir\"" if ($prtdir);

my @packages; my %isDup;

sub pkg_mtime {
	my $aName = $a; my $bName = $b;
	my $aTime; my $bTime;
	$aName =~ s/#.*//;
	$bName =~ s/#.*//;
	if ($aName lt $bName) { return -1; } 
	elsif ($aName gt $bName) { return 1; } 
	else {
		$aTime = (stat $a)[9];
		$bTime = (stat $b)[9];
	}
	if ($aTime le $bTime) { return -1; }
	else { return 1; }
}

my $pkgdir = shift @ARGV; my $quickMode=0;
if (($pkgdir) and (! -d "$pkgdir")) {
    print "usage: pkg-repgen [options] [directory [pkgname1..pkgnameN]]\n";
    exit 1;
}
my @dirlist = ("$pkgdir") ? glob("$pkgdir/*.pkg.tar.$compress") : glob("*.pkg.tar.$compress");
@dirlist = sort pkg_mtime @dirlist;
%isDup = map { $_ => 0 } @dirlist;

if (@ARGV) { # individual packages
    $quickMode=1;
    my @updates = sort @ARGV;
    while (my $name = shift @updates) {
        push @packages, grep { m/$name#.*\.pkg/ } @dirlist;
    }
} else { # the entire directory
    @packages = @dirlist;
}

# hashes to determine the package name ...
our %pname = map { $_ => (split /\//, $_)[-1] } @dirlist;
foreach my $p (@dirlist) { $pname{$p} =~ s/\#.*//; }

# ... or to look up the successor when merging old metadata files
my %followR; my %followH; my %followD; my @queue = @dirlist;
while (my $q = shift @queue) {
    ($#queue < 0) or ($pname{$q} ne $pname{$queue[0]}) or $isDup{$q} = 1;
}
 
# Populate some other hashes using a single run of prt-get
our %path; our %depends; our %descrip; our %flags;
my @validkeys = map { (split /\//, $_)[-1] } @dirlist;
map { s/\#.*// } @validkeys;
my %printme = map { $_ => 1 } @validkeys;
open (my $ppf, "$prtget printf '%n^%p^%e^%d^%E:%O:%R\n' |");
while (<$ppf>) {
    chomp;
    my ($name,$repo,$deps,$desc,$prepostread) = split /\^/;
    next if (! $printme{$name});
    $path{$name} = $repo . "/" . $name;
    $depends{$name} = $deps;
    $desc =~ s/\:/ /g;
    $descrip{$name} = $desc;
    $flags{$name} = $prepostread;
}
close ($ppf);

# Needed for alternating colors in the html index
my %parity = ( 0 => "even", 1 => "odd" );

# Generate the metadata files
($quickMode) ? pkg_single() : pkg_dir();

# Generate README and PKGINST
pkgreadscripts();

###################### individual packages ##########################
sub pkg_single {
  my ($oR, $oD, $oH, $nR, $nD, $nH, $oline, $oname);
  my $count = 0;      # needed for the html index
  my @dep_packages = @packages;
  my @idx_packages = @packages;

  my %firstrun = map { $_ => 0 } ("PKGREPO", "PKGDEPS", "index.html");
  open ($oR, "$pkgdir/PKGREPO")    or $firstrun{"PKGREPO"} = 1;
  open ($oD, "$pkgdir/PKGDEPS")    or $firstrun{"PKGDEPS"} = 1;
  open ($oH, "$pkgdir/index.html") or $firstrun{"index.html"} = 1;

  open ($nR, ">$pkgdir/PKGREPO.new");
  print "+ Updating specified entries in repository\n";
  RPKG: while (my $p =shift @packages) {
    my ($basename, $du, $md5, $ppr) = repodata($p);
    my $desc = (! $descrip{$pname{$p}}) ? "N.A." : $descrip{$pname{$p}};
    if ($firstrun{"PKGREPO"}==1) {
      printf $nR "%-s:%-s:%-s:%-s:%-s\n",$basename, $du, $md5, $desc, $ppr;
      next RPKG;
    }

    # Shift entries from the old repository until we find
    # a successor to the current package.
    while ( (! $followR{$pname{$p}}) and $oline = <$oR> ) {
        chomp($oline); $oname = $oline;
        $oname =~ s/\#.*//;
        print $nR "$oline\n" if ($oname lt $pname{$p});

        # before breaking out of the loop, append all the packages 
	# from the globbed queue that are lexographically earlier
        # than the current entry in the old repository.
	while ($pname{$p} le $oname) {
	  printf $nR "%-s:%-s:%-s:%-s:%-s\n", $basename, $du, $md5, $desc, $ppr;
	  next RPKG if (! $isDup{$p});
	  $p = shift @packages;
          ($basename, $du, $md5, $ppr) = repodata($p);
          $desc = (! $descrip{$pname{$p}}) ? "N.A." : $descrip{$pname{$p}};

          # save what got shifted from the repository if we're not going to
	  # print it now, but don't save packages that match the same glob.
          $followR{$pname{$p}} = "$oline\n" if ($pname{$p} lt $oname);
	}
    }

    # if the current package comes after everything in the old repository,
    # just append its metadata
    ($followR{$pname{$p}}) or printf $nR "%-s:%-s:%-s:%-s:%-s\n", $basename, $du, $md5, $desc, $ppr;

    next RPKG if (($isDup{$p}) or (! $followR{$pname{$p}}));

    # Arriving here means the current package is not a dup, and 
    # definitely has a successor in the old repository. But the 
    # next globbed package might be a more immediate successor. 
    # Decide which of the two possible successors comes first.
    # By defining a successor for the next package in the queue, 
    # we delay shifting entries off the old repo.
    if ((@packages) and ($pname{$packages[0]} le $followR{$pname{$p}})) {
           $followR{$pname{$packages[0]}} = $followR{$pname{$p}};
           next RPKG;
    } else {
	   print $nR $followR{$pname{$p}};
    }

  # Shift another package from the queue
  }

  # Likewise for the html index
  printheader(1);
  open ($nH, ">>$pkgdir/index.html.new");

  print "+ Updating specified entries in the html index\n";
  HPKG: while (my $p =shift @idx_packages) {
    my ($url, $pver, $desc, $date) = htmldata($p);

    if ($firstrun{"index.html"} == 1) {
        $count++;
        htmlrow($nH,$count,$pname{$p},$url,$pver,$desc,$date);
	next HPKG;
    }
    # Shift entries from the old html index until we find
    # a successor to the current package.
    while ( (! $followH{$pname{$p}}) and $oline=<$oH> ) {
        chomp($oline);
        # no need to copy the header, it should already be there
        next if ($oline !~ m/^<tr class="(odd|even)"/);
          
        $oname = $oline;
        $oname =~ s/.*a href="(.*)"/$1/; $oname =~ s/\%23.*//;
        if ($oname lt $pname{$p}) { $count++; print $nH "$oline\n"; }

        # before breaking out of the loop, append all the packages 
	# from the globbed queue that are lexographically earlier
        # than the current entry in the old html index.
        while ($pname{$p} le $oname) {
          $count++;
          htmlrow($nH,$count,$pname{$p},$url,$pver,$desc,$date);
	  next HPKG if (! $isDup{$p});
	  $p = shift @idx_packages;
          ($url, $pver, $desc, $date) = htmldata($p);

          # save what got shifted from the index if we're not going to print
	  # it now, but ignore packages that match the same glob.
          $followH{$pname{$p}} = "$oline\n" if ($pname{$p} lt $oname);
	}
    }

    # if the current package comes after everything in the old html index,
    # just append its metadata
    if (! $followH{$pname{$p}}) {
        $count++;
        htmlrow($nH,$count,$pname{$p},$url,$pver,$desc,$date);
    }

    next HPKG if (($isDup{$p}) or (! $followH{$pname{$p}}));

    # Arriving here means the current package is not a dup, and 
    # definitely has a successor in the old html index. But the 
    # next globbed package might be a more immediate successor.
    # Decide which of the two possible successors comes first. If the 
    # globbed package is the more immediate successor, use its name
    # as a key to retain the most recent entry from the html index.
    if ((@packages) and ($pname{$packages[0]} le $followH{$pname{$p}})) {
         $followH{$pname{$packages[0]}} = $followH{$pname{$p}};
         next HPKG;
    } else {
         $count++;
         $followH{$pname{$p}} =~ s/class="(even|odd)"/class="$parity{($count %2)}"/;
         print $nH $followH{$pname{$p}};
    }

  # Shift another package from the queue
  }

  # Likewise for the dependency map, but avoid creating duplicate entries
  open ($nD, ">$pkgdir/PKGDEPS.new");

  print "+ Updating specified entries in the depmap\n";
  DPKG: while (my $p =shift @dep_packages) {
    if ($firstrun{"PKGDEPS"}==1) {
       (! $depends{$pname{$p}}) or ($isDup{$p})
        or printf $nD "%-30s : %-s\n", $pname{$p}, $depends{$pname{$p}};
       next DPKG;
    }
    # Shift entries from the old depmap until we find a successor
    # to the current package
    while ( (! $followD{$pname{$p}}) and $oline = <$oD> ) {
        chomp($oline); $oname = $oline;
        $oname =~ s/\s*\:.*//;
        print $nD "$oline\n" if ($oname lt $pname{$p});
	while ($pname{$p} le $oname) {
	  if (! $isDup{$p}) {
            printf $nD "%-30s : %-s\n", $pname{$p}, $depends{$pname{$p}};
            next DPKG;
          } else {
	    $p = shift @dep_packages;
          }
          # save what got shifted from the depmap if we're not going to print
	  # it now, but ignore packages that match the same glob.
          $followD{$pname{$p}} = $oline if ($pname{$p} lt $oname);
	}
    }

    # if the current package comes after everything in the old depmap
    # and is not a dup, just append its metadata
    ($followD{$pname{$p}}) or ($isDup{$p}) or (! $depends{$pname{$p}})
    or printf $nD "%-30s : %-s\n", $pname{$p}, $depends{$pname{$p}};

    next DPKG if (($isDup{$p}) or (! $followD{$pname{$p}}));

    # Arriving here means the current package is not a dup, and 
    # definitely has a successor entry in the old depmap.
    # But the next globbed package might be a more immediate successor.
    # Decide which of the two possible successors comes first. If it's the
    # globbed package that comes next, save the old depmap entry.
    if ((@packages) and ($pname{$packages[0]} le $followD{$pname{$p}})) {
           $followD{$pname{$packages[0]}} = $followD{$pname{$p}};
	   next DPKG;
    } else {
           printf $nD $followD{$pname{$p}};
    }

  # Shift another package from the queue
  }

  # Done with all the packages that match command-line arguments.
  # Now append the tails of the old metadata files to their new counterparts.
  while ($firstrun{"index.html"}==0 and $oline = <$oH>) {
      $count++;
      $oline =~ s/class="(even|odd)"/class="$parity{($count % 2)}"/;
      print $nH $oline;
  }
  while ($firstrun{"PKGDEPS"}==0 and $oline = <$oD>) { print $nD $oline; }
  while ($firstrun{"PKGREPO"}==0 and $oline = <$oR>) { print $nR $oline; }

  close($nH);
  close($nD);
  close($nR);
  ($firstrun{"PKGREPO"}==1) or close($oR);
  ($firstrun{"PKGDEPS"}==1) or close($oD);
  ($firstrun{"index.html"}==1) or close($oH);

  foreach my $db (keys %firstrun) { rename("$pkgdir/$db.new", "$pkgdir/$db"); }
  printfooter($count) if ($firstrun{"index.html"} == 1);
}

######################## full repository ########################
sub pkg_dir {
    print "+ Generating dependencies\n";
    open (my $iD, ">$pkgdir/PKGDEPS");
    print "+ Generating repository\n";
    open (my $iR, ">$pkgdir/PKGREPO");
    printheader(0);
    my $count = 0;
    open (my $ih, ">>$pkgdir/index.html");
    foreach my $p (@packages) {
        my ($basename, $du, $md5, $ppr) = repodata($p);
	my ($url, $pver, $desc, $date) = htmldata($p);
        (! $depends{$pname{$p}}) or ($isDup{$p}) 
                or printf $iD "%-30s : %-s\n", $pname{$p}, $depends{$pname{$p}};
        printf $iR "%-s:%-s:%-s:%-s:%-s\n", $basename,$du,$md5,$desc,$ppr;
        $count++;
        htmlrow($ih,$count,$pname{$p},$url,$pver,$desc,$date);
    }
    close($ih);
    printfooter($count);
    close($iR);
    close($iD);
}

# consolidate all the README and install scripts for the available packages
sub pkgreadscripts {
    print "+ Generating README\n";
    open (my $fR, ">$pkgdir/PKGREAD");
    print $fR "# README files for repository. Do NOT remove this line.\n";

    print "+ Generating scripts\n";
    open (my $fS, ">$pkgdir/PKGINST");
    print $fS '#!/usr/bin/env bash
#
# PKGINST: pre- and post-install scripts for CRUX packages
#
run_script() {
  case "$1" in
';

    my %seen;
    foreach my $name (@dirlist) {
        $name =~ s/\#.*//; $name = (split /\//, $name)[-1];
	next if ($seen{$name});
        $seen{$name} = 1;
	next if (! $path{$name});
        if (-f "$path{$name}/README"){
            print $fR "##### PKGREADME: $name\n";
            open(my $readme, "$path{$name}/README");
            while (<$readme>){ print $fR $_; }
            close($readme);
        }
        foreach my $when ("pre", "post") {
            if (-f "$path{$name}/${when}-install"){
                print $fS "    $name.$when)\n";
                open(my $rs, "$path{$name}/${when}-install");
                while (<$rs>){
                    chomp;
                    (m/^\#(!.*sh|\s*EOF|\s*End)/) or print $fS "      $_\n";
                }
                close($rs);
                print $fS "    ;;\n";
            }
        }
    }
    print $fS "  esac\n}\n\n";
    print $fS '[ "$1" ] && [[ "$2" == @(pre|post) ]] && run_script "$1.$2"';
    print $fS "\n";
    close $fS;
    close $fR;
}

######################## html index subs ########################

sub printheader {
    my $isTemp = shift; my $ih;
    ($isTemp == 0) ? open ($ih, ">$pkgdir/index.html") : open ($ih, ">$pkgdir/index.html.new");
    print $ih <<EOH;
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.1//EN"
    "http://www.w3.org/TR/xhtml11/DTD/xhtml11.dtd">
<html xmlns="http://www.w3.org/1999/xhtml">
 <head>
EOH

    print $ih "  <title>$title</title>\n";

    print $ih <<EOH;
  <style type="text/css">
   body
   {
    font-family: Verdana, sans-serif;
    font-size: 85%;
    padding: 2em;
   }
   a
   {
    color: #67550d;
   }
   table
   {
    border: solid #e5dccf 1px;
    font-size: 85%;
   }
   td
   {
    padding: 6px;
   }
   tr.header
   {
    background-color: #e5dccf;
   }
   tr.odd
   {
    background-color: #f7f3ed;
   }
   tr.even
   {
    background-color: #fcf9f8;
   }
  </style>
  <meta http-equiv="Content-Type" content="text/html; charset=ISO-8859-1" />
 </head>
 <body>
EOH

    print $ih "  <h2>$title</h2>\n";
    if ($header) {
        open(FILE, $header) or die "Couldn't open header file";
        while (<FILE>) { print $ih "  " . $_; }
        close(FILE);
    }

    print $ih "  <table width=\"100%\" cellspacing=\"0\">\n";
    print $ih "   <tr class=\"header\"><td><b>Port</b></td><td><b>Version</b></td><td><b>Description</b></td>";
    print $ih "<td><b>Last modified</b></td>";
    print $ih "</tr>\n";
    close($ih);
}

sub htmlrow {
    my ($ih, $count, $name, $url, $version, $desc, $date) = @_;
    print $ih "<tr class=\"$parity{($count % 2)}\"><td>$name</td>";
    print $ih "<td><a href=\"$url\">$version</a></td>";
    print $ih "<td>$desc</td><td>$date</td></tr>\n";
}

sub printfooter {
    my $count = shift;
    open (my $ih, ">>$pkgdir/index.html");
    print $ih "  </table>\n";
    print $ih "  <p><b>$count packages</b></p>\n";
    if ($footer) {
        open(FILE, $footer) or die "Couldn't open footer file";
        while (<FILE>) { print $ih "  " . $_; }
        close(FILE);
    }
    print $ih "  <p><i>Generated by <a href=\"http://www.varlock.com\">pkg-repgen</a> on " . isotime() . ".</i></p>\n";
    print $ih <<EOH;
 </body>
</html>
EOH
    close($ih);
}

sub htmldata {
    my $p = shift;
    my ($pver, $url) = ($p, $p);
    $pver =~ s/.*\#//; $pver =~ s/\.pkg\.tar.*//;
    $url = (split /\//, $p)[-1]; $url =~ s/\#/\%23/;
    my $date = isotime( (stat($p))[9], 1);
    my $desc = (! $descrip{$pname{$p}}) ? "N.A." : $descrip{$pname{$p}};
    return $url, $pver, $desc, $date;
}

sub repodata {
    my $p = shift;
    my $basename = (split /\//, $p)[-1];
    my $du = (-s $p);
    my $md5 = digest_file_hex($p,"MD5");
    my $ppr = (! $flags{$pname{$p}}) ? "no:no:no" : $flags{$pname{$p}};
    return $basename, $du, $md5, $ppr;
}

sub isotime {
    my $time = (shift or time);
    my $accuracy = (shift or 2);
    my @t = gmtime ($time);
    my $year = $t[5] + 1900;
    my $month = sprintf("%02d", $t[4] + 1);
    my $day = sprintf("%02d", $t[3]);

    return "$year-$month-$day" if ($accuracy == 1);
    return "$year-$month-$day " . sprintf("%02d:%02d:%02d UTC", $t[2], $t[1], $t[0]);
}