lzbw2_file_compression.pl 6.6 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249
  1. #!/usr/bin/perl
  2. # Author: Trizen
  3. # Date: 05 September 2023
  4. # Edit: 11 April 2024
  5. # https://github.com/trizen
  6. # Compress/decompress files using LZ77 compression + fixed-width integers encoding + Burrows-Wheeler Transform (BWT) + Huffman coding.
  7. # References:
  8. # Data Compression (Summer 2023) - Lecture 13 - BZip2
  9. # https://youtube.com/watch?v=cvoZbBZ3M2A
  10. use 5.036;
  11. use Getopt::Std qw(getopts);
  12. use File::Basename qw(basename);
  13. use List::Util qw(uniq);
  14. use Compression::Util qw(:all);
  15. use constant {
  16. PKGNAME => 'LZBW2',
  17. VERSION => '0.01',
  18. FORMAT => 'lzbw2',
  19. COMPRESSED_BYTE => chr(1),
  20. UNCOMPRESSED_BYTE => chr(0),
  21. CHUNK_SIZE => 1 << 16, # higher value = better compression
  22. RANDOM_DATA_THRESHOLD => 1, # in ratio
  23. };
  24. # Container signature
  25. use constant SIGNATURE => uc(FORMAT) . chr(1);
  26. sub usage {
  27. my ($code) = @_;
  28. print <<"EOH";
  29. usage: $0 [options] [input file] [output file]
  30. options:
  31. -e : extract
  32. -i <filename> : input filename
  33. -o <filename> : output filename
  34. -r : rewrite output
  35. -v : version number
  36. -h : this message
  37. examples:
  38. $0 document.txt
  39. $0 document.txt archive.${\FORMAT}
  40. $0 archive.${\FORMAT} document.txt
  41. $0 -e -i archive.${\FORMAT} -o document.txt
  42. EOH
  43. exit($code // 0);
  44. }
  45. sub version {
  46. printf("%s %s\n", PKGNAME, VERSION);
  47. exit;
  48. }
  49. sub valid_archive {
  50. my ($fh) = @_;
  51. if (read($fh, (my $sig), length(SIGNATURE), 0) == length(SIGNATURE)) {
  52. $sig eq SIGNATURE || return;
  53. }
  54. return 1;
  55. }
  56. sub main {
  57. my %opt;
  58. getopts('ei:o:vhr', \%opt);
  59. $opt{h} && usage(0);
  60. $opt{v} && version();
  61. my ($input, $output) = @ARGV;
  62. $input //= $opt{i} // usage(2);
  63. $output //= $opt{o};
  64. my $ext = qr{\.${\FORMAT}\z}io;
  65. if ($opt{e} || $input =~ $ext) {
  66. if (not defined $output) {
  67. ($output = basename($input)) =~ s{$ext}{}
  68. || die "$0: no output file specified!\n";
  69. }
  70. if (not $opt{r} and -e $output) {
  71. print "'$output' already exists! -- Replace? [y/N] ";
  72. <STDIN> =~ /^y/i || exit 17;
  73. }
  74. decompress_file($input, $output)
  75. || die "$0: error: decompression failed!\n";
  76. }
  77. elsif ($input !~ $ext || (defined($output) && $output =~ $ext)) {
  78. $output //= basename($input) . '.' . FORMAT;
  79. compress_file($input, $output)
  80. || die "$0: error: compression failed!\n";
  81. }
  82. else {
  83. warn "$0: don't know what to do...\n";
  84. usage(1);
  85. }
  86. }
  87. # Compress file
  88. sub compress_file ($input, $output) {
  89. open my $fh, '<:raw', $input
  90. or die "Can't open file <<$input>> for reading: $!";
  91. my $header = SIGNATURE;
  92. # Open the output file for writing
  93. open my $out_fh, '>:raw', $output
  94. or die "Can't open file <<$output>> for write: $!";
  95. # Print the header
  96. print $out_fh $header;
  97. my $lengths_str = '';
  98. my $uncompressed_str = '';
  99. my @sizes;
  100. my @distances_block;
  101. open my $uc_fh, '>:raw', \$uncompressed_str;
  102. open my $len_fh, '>:raw', \$lengths_str;
  103. my $create_bz2_block = sub {
  104. scalar(@sizes) > 0 or return;
  105. print $out_fh COMPRESSED_BYTE;
  106. print $out_fh delta_encode(\@sizes);
  107. print $out_fh bz2_compress($uncompressed_str);
  108. print $out_fh bz2_compress($lengths_str);
  109. my $ratio = uniq(@distances_block) / @distances_block * 100;
  110. if ($ratio < 10) {
  111. print $out_fh COMPRESSED_BYTE;
  112. print $out_fh bz2_compress_symbolic(\@distances_block);
  113. }
  114. else {
  115. print $out_fh UNCOMPRESSED_BYTE;
  116. print $out_fh obh_encode(\@distances_block);
  117. }
  118. @sizes = ();
  119. @distances_block = ();
  120. open $uc_fh, '>:raw', \$uncompressed_str;
  121. open $len_fh, '>:raw', \$lengths_str;
  122. };
  123. # Compress data
  124. while (read($fh, (my $chunk), CHUNK_SIZE)) {
  125. my ($literals, $distances, $lengths) = lz77_encode($chunk);
  126. my $est_ratio = length($chunk) / (4 * scalar(@$literals));
  127. say "Est. ratio: ", $est_ratio, " (", scalar(@$literals), " uncompressed bytes)";
  128. if ($est_ratio > RANDOM_DATA_THRESHOLD) {
  129. push @sizes, scalar(@$literals);
  130. print $uc_fh pack('C*', @$literals);
  131. print $len_fh pack('C*', @$lengths);
  132. push @distances_block, @$distances;
  133. }
  134. else {
  135. say "Random data detected...";
  136. $create_bz2_block->();
  137. print $out_fh UNCOMPRESSED_BYTE;
  138. print $out_fh create_huffman_entry(string2symbols($chunk));
  139. }
  140. if (length($uncompressed_str) >= CHUNK_SIZE) {
  141. $create_bz2_block->();
  142. }
  143. }
  144. $create_bz2_block->();
  145. close $out_fh;
  146. }
  147. # Decompress file
  148. sub decompress_file ($input, $output) {
  149. # Open and validate the input file
  150. open my $fh, '<:raw', $input
  151. or die "Can't open file <<$input>> for reading: $!";
  152. valid_archive($fh) || die "$0: file `$input' is not a \U${\FORMAT}\E v${\VERSION} archive!\n";
  153. # Open the output file
  154. open my $out_fh, '>:raw', $output
  155. or die "Can't open file <<$output>> for writing: $!";
  156. while (!eof($fh)) {
  157. my $compression_byte = getc($fh) // die "decompression error";
  158. if ($compression_byte eq UNCOMPRESSED_BYTE) {
  159. say "Decoding random data...";
  160. print $out_fh pack('C*', @{decode_huffman_entry($fh)});
  161. next;
  162. }
  163. elsif ($compression_byte ne COMPRESSED_BYTE) {
  164. die "decompression error";
  165. }
  166. my @sizes = @{delta_decode($fh)};
  167. my @uncompressed = unpack('C*', bz2_decompress($fh));
  168. my @lengths = unpack('C*', bz2_decompress($fh));
  169. my @distances = @{(getc($fh) eq COMPRESSED_BYTE) ? bz2_decompress_symbolic($fh) : obh_decode($fh)};
  170. while (@uncompressed) {
  171. my $size = shift(@sizes) // die "decompression error";
  172. my @uncompressed_chunk = splice(@uncompressed, 0, $size);
  173. my @lengths_chunk = splice(@lengths, 0, $size);
  174. my @distances_chunk = splice(@distances, 0, $size);
  175. scalar(@uncompressed_chunk) == $size or die "decompression error";
  176. scalar(@lengths_chunk) == $size or die "decompression error";
  177. scalar(@distances_chunk) == $size or die "decompression error";
  178. print $out_fh lz77_decode(\@uncompressed_chunk, \@distances_chunk, \@lengths_chunk);
  179. }
  180. }
  181. close $fh;
  182. close $out_fh;
  183. }
  184. main();
  185. exit(0);