123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525 |
- #!/usr/bin/perl
- # Author: Trizen
- # Date: 15 December 2022
- # Edit: 19 March 2024
- # https://github.com/trizen
- # Compress/decompress files using LZ77 compression + Arithmetic Coding (in fixed bits).
- # Reference:
- # Basic arithmetic coder in C++
- # https://github.com/billbird/arith32
- use 5.020;
- use strict;
- use warnings;
- use experimental qw(signatures);
- use Getopt::Std qw(getopts);
- use File::Basename qw(basename);
- use List::Util qw(max);
- use constant {
- PKGNAME => 'LZA',
- VERSION => '0.01',
- FORMAT => 'lza',
- CHUNK_SIZE => 1 << 16,
- };
- # Container signature
- use constant SIGNATURE => uc(FORMAT) . chr(1);
- # Arithmetic Coding settings
- use constant BITS => 32;
- use constant MAX => oct('0b' . ('1' x BITS));
- sub usage {
- my ($code) = @_;
- print <<"EOH";
- usage: $0 [options] [input file] [output file]
- options:
- -e : extract
- -i <filename> : input filename
- -o <filename> : output filename
- -r : rewrite output
- -v : version number
- -h : this message
- examples:
- $0 document.txt
- $0 document.txt archive.${\FORMAT}
- $0 archive.${\FORMAT} document.txt
- $0 -e -i archive.${\FORMAT} -o document.txt
- EOH
- exit($code // 0);
- }
- sub version {
- printf("%s %s\n", PKGNAME, VERSION);
- exit;
- }
- sub valid_archive {
- my ($fh) = @_;
- if (read($fh, (my $sig), length(SIGNATURE), 0) == length(SIGNATURE)) {
- $sig eq SIGNATURE || return;
- }
- return 1;
- }
- sub main {
- my %opt;
- getopts('ei:o:vhr', \%opt);
- $opt{h} && usage(0);
- $opt{v} && version();
- my ($input, $output) = @ARGV;
- $input //= $opt{i} // usage(2);
- $output //= $opt{o};
- my $ext = qr{\.${\FORMAT}\z}io;
- if ($opt{e} || $input =~ $ext) {
- if (not defined $output) {
- ($output = basename($input)) =~ s{$ext}{}
- || die "$0: no output file specified!\n";
- }
- if (not $opt{r} and -e $output) {
- print "'$output' already exists! -- Replace? [y/N] ";
- <STDIN> =~ /^y/i || exit 17;
- }
- decompress_file($input, $output)
- || die "$0: error: decompression failed!\n";
- }
- elsif ($input !~ $ext || (defined($output) && $output =~ $ext)) {
- $output //= basename($input) . '.' . FORMAT;
- compress_file($input, $output)
- || die "$0: error: compression failed!\n";
- }
- else {
- warn "$0: don't know what to do...\n";
- usage(1);
- }
- }
- sub read_bit ($fh, $bitstring) {
- if (($$bitstring // '') eq '') {
- $$bitstring = unpack('b*', getc($fh) // return undef);
- }
- chop($$bitstring);
- }
- sub read_bits ($fh, $bits_len) {
- my $data = '';
- read($fh, $data, $bits_len >> 3);
- $data = unpack('B*', $data);
- while (length($data) < $bits_len) {
- $data .= unpack('B*', getc($fh) // return undef);
- }
- if (length($data) > $bits_len) {
- $data = substr($data, 0, $bits_len);
- }
- return $data;
- }
- sub delta_encode ($integers, $double = 0) {
- my @deltas;
- my $prev = 0;
- unshift(@$integers, scalar(@$integers));
- while (@$integers) {
- my $curr = shift(@$integers);
- push @deltas, $curr - $prev;
- $prev = $curr;
- }
- my $bitstring = '';
- foreach my $d (@deltas) {
- if ($d == 0) {
- $bitstring .= '0';
- }
- elsif ($double) {
- my $t = sprintf('%b', abs($d) + 1);
- my $l = sprintf('%b', length($t));
- $bitstring .= '1' . (($d < 0) ? '0' : '1') . ('1' x (length($l) - 1)) . '0' . substr($l, 1) . substr($t, 1);
- }
- else {
- my $t = sprintf('%b', abs($d));
- $bitstring .= '1' . (($d < 0) ? '0' : '1') . ('1' x (length($t) - 1)) . '0' . substr($t, 1);
- }
- }
- pack('B*', $bitstring);
- }
- sub delta_decode ($fh, $double = 0) {
- my @deltas;
- my $buffer = '';
- my $len = 0;
- for (my $k = 0 ; $k <= $len ; ++$k) {
- my $bit = read_bit($fh, \$buffer);
- if ($bit eq '0') {
- push @deltas, 0;
- }
- elsif ($double) {
- my $bit = read_bit($fh, \$buffer);
- my $bl = 0;
- ++$bl while (read_bit($fh, \$buffer) eq '1');
- my $bl2 = oct('0b1' . join('', map { read_bit($fh, \$buffer) } 1 .. $bl));
- my $int = oct('0b1' . join('', map { read_bit($fh, \$buffer) } 1 .. ($bl2 - 1)));
- push @deltas, ($bit eq '1' ? 1 : -1) * ($int - 1);
- }
- else {
- my $bit = read_bit($fh, \$buffer);
- my $n = 0;
- ++$n while (read_bit($fh, \$buffer) eq '1');
- my $d = oct('0b1' . join('', map { read_bit($fh, \$buffer) } 1 .. $n));
- push @deltas, ($bit eq '1' ? $d : -$d);
- }
- if ($k == 0) {
- $len = pop(@deltas);
- }
- }
- my @acc;
- my $prev = $len;
- foreach my $d (@deltas) {
- $prev += $d;
- push @acc, $prev;
- }
- return \@acc;
- }
- sub lz77_compression ($str, $uncompressed, $indices, $lengths) {
- my $la = 0;
- my $prefix = '';
- my @chars = split(//, $str);
- my $end = $#chars;
- while ($la <= $end) {
- my $n = 1;
- my $p = 0;
- my $tmp;
- my $token = $chars[$la];
- while ( $n < 255
- and $la + $n <= $end
- and ($tmp = index($prefix, $token, $p)) >= 0) {
- $p = $tmp;
- $token .= $chars[$la + $n];
- ++$n;
- }
- --$n;
- push @$indices, $p;
- push @$lengths, $n;
- push @$uncompressed, $chars[$la + $n];
- $la += $n + 1;
- $prefix .= $token;
- }
- return;
- }
- sub lz77_decompression ($uncompressed, $indices, $lengths) {
- my $ret = '';
- my $chunk = '';
- foreach my $i (0 .. $#{$uncompressed}) {
- $chunk .= substr($chunk, $indices->[$i], $lengths->[$i]) . chr($uncompressed->[$i]);
- if (length($chunk) >= CHUNK_SIZE) {
- $ret .= $chunk;
- $chunk = '';
- }
- }
- if ($chunk ne '') {
- $ret .= $chunk;
- }
- $ret;
- }
- sub create_cfreq ($freq) {
- my @cf;
- my $T = 0;
- foreach my $i (sort { $a <=> $b } keys %$freq) {
- $freq->{$i} // next;
- $cf[$i] = $T;
- $T += $freq->{$i};
- $cf[$i + 1] = $T;
- }
- return (\@cf, $T);
- }
- sub ac_encode ($bytes_arr) {
- my $enc = '';
- my $EOF_SYMBOL = (max(@$bytes_arr) // 0) + 1;
- my @bytes = (@$bytes_arr, $EOF_SYMBOL);
- my %freq;
- ++$freq{$_} for @bytes;
- my ($cf, $T) = create_cfreq(\%freq);
- if ($T > MAX) {
- die "Too few bits: $T > ${\MAX}";
- }
- my $low = 0;
- my $high = MAX;
- my $uf_count = 0;
- foreach my $c (@bytes) {
- my $w = $high - $low + 1;
- $high = ($low + int(($w * $cf->[$c + 1]) / $T) - 1) & MAX;
- $low = ($low + int(($w * $cf->[$c]) / $T)) & MAX;
- if ($high > MAX) {
- die "high > MAX: $high > ${\MAX}";
- }
- if ($low >= $high) { die "$low >= $high" }
- while (1) {
- if (($high >> (BITS - 1)) == ($low >> (BITS - 1))) {
- my $bit = $high >> (BITS - 1);
- $enc .= $bit;
- if ($uf_count > 0) {
- $enc .= join('', 1 - $bit) x $uf_count;
- $uf_count = 0;
- }
- $low <<= 1;
- ($high <<= 1) |= 1;
- }
- elsif (((($low >> (BITS - 2)) & 0x1) == 1) && ((($high >> (BITS - 2)) & 0x1) == 0)) {
- ($high <<= 1) |= (1 << (BITS - 1));
- $high |= 1;
- ($low <<= 1) &= ((1 << (BITS - 1)) - 1);
- ++$uf_count;
- }
- else {
- last;
- }
- $low &= MAX;
- $high &= MAX;
- }
- }
- $enc .= '0';
- $enc .= '1';
- while (length($enc) % 8 != 0) {
- $enc .= '1';
- }
- return ($enc, \%freq);
- }
- sub ac_decode ($fh, $freq) {
- my ($cf, $T) = create_cfreq($freq);
- my @dec;
- my $low = 0;
- my $high = MAX;
- my $enc = oct('0b' . join '', map { getc($fh) // 1 } 1 .. BITS);
- my @table;
- foreach my $i (sort { $a <=> $b } keys %$freq) {
- foreach my $j ($cf->[$i] .. $cf->[$i + 1] - 1) {
- $table[$j] = $i;
- }
- }
- my $EOF_SYMBOL = max(keys %$freq) // 0;
- while (1) {
- my $w = $high - $low + 1;
- my $ss = int((($T * ($enc - $low + 1)) - 1) / $w);
- my $i = $table[$ss] // last;
- last if ($i == $EOF_SYMBOL);
- push @dec, $i;
- $high = ($low + int(($w * $cf->[$i + 1]) / $T) - 1) & MAX;
- $low = ($low + int(($w * $cf->[$i]) / $T)) & MAX;
- if ($high > MAX) {
- die "error";
- }
- if ($low >= $high) { die "$low >= $high" }
- while (1) {
- if (($high >> (BITS - 1)) == ($low >> (BITS - 1))) {
- ($high <<= 1) |= 1;
- $low <<= 1;
- ($enc <<= 1) |= (getc($fh) // 1);
- }
- elsif (((($low >> (BITS - 2)) & 0x1) == 1) && ((($high >> (BITS - 2)) & 0x1) == 0)) {
- ($high <<= 1) |= (1 << (BITS - 1));
- $high |= 1;
- ($low <<= 1) &= ((1 << (BITS - 1)) - 1);
- $enc = (($enc >> (BITS - 1)) << (BITS - 1)) | (($enc & ((1 << (BITS - 2)) - 1)) << 1) | (getc($fh) // 1);
- }
- else {
- last;
- }
- $low &= MAX;
- $high &= MAX;
- $enc &= MAX;
- }
- }
- return \@dec;
- }
- sub create_ac_entry ($bytes, $out_fh) {
- my ($enc, $freq) = ac_encode($bytes);
- my $max_symbol = max(keys %$freq) // 0;
- my @freqs;
- foreach my $k (0 .. $max_symbol) {
- push @freqs, $freq->{$k} // 0;
- }
- push @freqs, length($enc) >> 3;
- print $out_fh delta_encode(\@freqs);
- print $out_fh pack("B*", $enc);
- }
- sub decode_ac_entry ($fh) {
- my @freqs = @{delta_decode($fh)};
- my $bits_len = pop(@freqs);
- my %freq;
- foreach my $i (0 .. $#freqs) {
- if ($freqs[$i]) {
- $freq{$i} = $freqs[$i];
- }
- }
- say "Encoded length: $bits_len";
- my $bits = read_bits($fh, $bits_len << 3);
- if ($bits_len > 0) {
- open my $bits_fh, '<:raw', \$bits;
- return ac_decode($bits_fh, \%freq);
- }
- return [];
- }
- # Compress file
- sub compress_file ($input, $output) {
- open my $fh, '<:raw', $input
- or die "Can't open file <<$input>> for reading: $!";
- my $header = SIGNATURE;
- # Open the output file for writing
- open my $out_fh, '>:raw', $output
- or die "Can't open file <<$output>> for write: $!";
- # Print the header
- print $out_fh $header;
- my (@uncompressed, @indices, @lengths);
- # Compress data
- while (read($fh, (my $chunk), CHUNK_SIZE)) {
- lz77_compression($chunk, \@uncompressed, \@indices, \@lengths);
- }
- @indices = unpack('C*', pack('S*', @indices));
- @uncompressed = unpack('C*', join('', @uncompressed));
- create_ac_entry(\@uncompressed, $out_fh);
- create_ac_entry(\@indices, $out_fh);
- create_ac_entry(\@lengths, $out_fh);
- # Close the file
- close $out_fh;
- }
- # Decompress file
- sub decompress_file ($input, $output) {
- # Open and validate the input file
- open my $fh, '<:raw', $input
- or die "Can't open file <<$input>> for reading: $!";
- valid_archive($fh) || die "$0: file `$input' is not a \U${\FORMAT}\E v${\VERSION} archive!\n";
- # Open the output file
- open my $out_fh, '>:raw', $output
- or die "Can't open file <<$output>> for writing: $!";
- my $uncompressed = decode_ac_entry($fh);
- my @indices = unpack('S*', pack('C*', @{decode_ac_entry($fh)}));
- my $lengths = decode_ac_entry($fh);
- print $out_fh lz77_decompression($uncompressed, \@indices, $lengths);
- # Close the file
- close $fh;
- close $out_fh;
- }
- main();
- exit(0);
|