| 1 | use strict; | ||
| 2 | use warnings; | ||
| 3 | |||
| 4 | { no warnings 'once'; # pass info back to Makefile.PL | ||
| 5 | $PDL::Core::Dev::EXTRAS{$::PDLMOD}{OBJECT} .= join '', map " $::PDLBASE/$_\$(OBJ_EXT)", qw(ricecomp); | ||
| 6 | $PDL::Core::Dev::EXTRAS{$::PDLMOD}{INC} .= qq{ "-I$::PDLBASE"}; | ||
| 7 | } | ||
| 8 | |||
| 9 | pp_addpm({At=>'Top'},<<'EOD'); | ||
| 10 | |||
| 11 | =head1 NAME | ||
| 12 | |||
| 13 | PDL::Compression - compression utilities | ||
| 14 | |||
| 15 | =head1 DESCRIPTION | ||
| 16 | |||
| 17 | These routines generally accept some data as a PDL and compress it | ||
| 18 | into a smaller PDL. Algorithms typically work on a single dimension | ||
| 19 | and broadcast over other dimensions, producing a broadcasted table of | ||
| 20 | compressed values if more than one dimension is fed in. | ||
| 21 | |||
| 22 | The Rice algorithm, in particular, is designed to be identical to the | ||
| 23 | RICE_1 algorithm used in internal FITS-file compression (see PDL::IO::FITS). | ||
| 24 | |||
| 25 | =head1 SYNOPSIS | ||
| 26 | |||
| 27 | use PDL::Compression | ||
| 28 | |||
| 29 | ($y,$xsize) = $x->rice_compress(); | ||
| 30 | $c = $y->rice_expand($xsize); | ||
| 31 | |||
| 32 | =cut | ||
| 33 | 3 | ▶ | |
|
Subroutine
BEGIN: called CC=1 SCAR=0.0
| |||
| 34 | 3 | ▶ | use strict; |
|
Subroutine
BEGIN: called CC=1 SCAR=0.0
| |||
| 35 | use warnings; | ||
| 36 | |||
| 37 | EOD | ||
| 38 | |||
| 39 | pp_addpm({At=>'Bot'},<<'EOD'); | ||
| 40 | |||
| 41 | =head1 AUTHORS | ||
| 42 | |||
| 43 | Copyright (C) 2010 Craig DeForest. | ||
| 44 | All rights reserved. There is no warranty. You are allowed | ||
| 45 | to redistribute this software / documentation under certain | ||
| 46 | conditions. For details, see the file COPYING in the PDL | ||
| 47 | distribution. If this file is separated from the PDL distribution, | ||
| 48 | the copyright notice should be included in the file. | ||
| 49 | |||
| 50 | The Rice compression library is derived from the similar library in | ||
| 51 | the CFITSIO 3.24 release, and is licensed under yet more more lenient | ||
| 52 | terms than PDL itself; that notice is present in the file "ricecomp.c". | ||
| 53 | |||
| 54 | =head1 BUGS | ||
| 55 | |||
| 56 | =over 3 | ||
| 57 | |||
| 58 | =item * Currently headers are ignored. | ||
| 59 | |||
| 60 | =item * Currently there is only one compression algorithm. | ||
| 61 | |||
| 62 | =back | ||
| 63 | |||
| 64 | =head1 TODO | ||
| 65 | |||
| 66 | =over 3 | ||
| 67 | |||
| 68 | =item * Add object encapsulation | ||
| 69 | |||
| 70 | =item * Add test suite | ||
| 71 | |||
| 72 | =back | ||
| 73 | |||
| 74 | =cut | ||
| 75 | |||
| 76 | EOD | ||
| 77 | |||
| 78 | pp_addpm(<<'EOD'); | ||
| 79 | |||
| 80 | =head1 METHODS | ||
| 81 | |||
| 82 | =cut | ||
| 83 | |||
| 84 | EOD | ||
| 85 | |||
| 86 | pp_def( | ||
| 87 | "rice_compress", | ||
| 88 | HandleBad => 0, | ||
| 89 | Pars => 'in(n); [o]out(m=CALC(ceil($SIZE(n) * 1.01))); indx[o]len()', | ||
| 90 | OtherPars => "int blocksize", # in OtherPars to avoid autopromotion | ||
| 91 | GenericTypes =>['B','S','US','L'], | ||
| 92 | Doc => <<'EOD', | ||
| 93 | =for ref | ||
| 94 | |||
| 95 | Squishes an input PDL along the 0 dimension by Rice compression. In | ||
| 96 | scalar context, you get back only the compressed PDL; in list context, | ||
| 97 | you also get back ancillary information that is required to uncompress | ||
| 98 | the data with rice_uncompress. | ||
| 99 | |||
| 100 | Multidimensional data are broadcasted over - each row is compressed | ||
| 101 | separately, and the returned PDL is squished to the maximum compressed | ||
| 102 | size of any row. If any of the streams could not be compressed (the | ||
| 103 | algorithm produced longer output), the corresponding length is set to -1 | ||
| 104 | and the row is treated as if it had length 0. | ||
| 105 | |||
| 106 | Rice compression only works on integer data types -- if you have | ||
| 107 | floating point data you must first quantize them. | ||
| 108 | |||
| 109 | The underlying algorithm is identical to the Rice compressor used in | ||
| 110 | CFITSIO (and is used by PDL::IO::FITS to load and save compressed FITS | ||
| 111 | images). | ||
| 112 | |||
| 113 | The optional blocksize indicates how many samples are to be compressed | ||
| 114 | as a unit; it defaults to 32. | ||
| 115 | |||
| 116 | How it works: | ||
| 117 | |||
| 118 | Rice compression is a subset of Golomb compression, and works on data sets | ||
| 119 | where variation between adjacent samples is typically small compared to the | ||
| 120 | dynamic range of each sample. In this implementation (originally written | ||
| 121 | by Richard White and contributed to CFITSIO in 1999), the data are divided | ||
| 122 | into blocks of samples (by default 32 samples per block). Each block | ||
| 123 | has a running difference applied, and the difference is bit-folded to make | ||
| 124 | it positive definite. High order bits of the difference stream are discarded, | ||
| 125 | and replaced with a unary representation; low order bits are preserved. Unary | ||
| 126 | representation is very efficient for small numbers, but large jumps could | ||
| 127 | give rise to ludicrously large bins in a plain Golomb code; such large jumps | ||
| 128 | ("high entropy" samples) are simply recorded directly in the output stream. | ||
| 129 | |||
| 130 | Working on astronomical or solar image data, typical compression ratios of | ||
| 131 | 2-3 are achieved. | ||
| 132 | EOD | ||
| 133 | PMCode => <<'EOD', | ||
| 134 | sub PDL::rice_compress { | ||
| 135 | my $in = shift; | ||
| 136 | my $blocksize = shift || 32; | ||
| 137 | ## Reject floating-point inputs | ||
| 138 | if( $in->type != byte && | ||
| 139 | $in->type != short && | ||
| 140 | $in->type != ushort && | ||
| 141 | $in->type != long | ||
| 142 | ) { | ||
| 143 | die("rice_compress: input needs to have type byte, short, ushort, or long, not ".($in->type)."\n"); | ||
| 144 | } | ||
| 145 | PDL::_rice_compress_int( $in, my $out=PDL->null, my $len=PDL->null, $blocksize ); | ||
| 146 | my $l = $len->max->sclr; | ||
| 147 | $out = $out->slice("0:".($l-1))->sever; | ||
| 148 | return wantarray ? ($out, $in->dim(0), $blocksize, $len) : $out; | ||
| 149 | } | ||
| 150 | EOD | ||
| 151 | CHeader => qq{#include "ricecomp.h"\n}, | ||
| 152 | Code => <<'EOD', | ||
| 153 | char *err = "error message not updated"; | ||
| 154 | $len() = fits_rcomp$TBSUL(_byte,_short,_short,)(&err, (void *)$P(in), | ||
| 155 | $SIZE(n), | ||
| 156 | (void *)$P(out), | ||
| 157 | $SIZE(m) * sizeof($GENERIC(out)), | ||
| 158 | $COMP(blocksize) | ||
| 159 | ); | ||
| 160 | if (err) $CROAK("%s", err); | ||
| 161 | EOD | ||
| 162 | ); | ||
| 163 | |||
| 164 | pp_def( | ||
| 165 | "rice_expand", | ||
| 166 | HandleBad=>0, | ||
| 167 | Pars=>'in(n); indx len(); [o]out(m);', | ||
| 168 | OtherPars=>'IV dim0 => m; int blocksize', | ||
| 169 | OtherParsDefaults=>{ blocksize=>32 }, | ||
| 170 | GenericTypes=>['B','S','US','L'], | ||
| 171 | Doc=><<'EOD', | ||
| 172 | =for ref | ||
| 173 | |||
| 174 | Unsquishes a PDL that has been squished by rice_compress. | ||
| 175 | EOD | ||
| 176 | CHeader => qq{#include "ricecomp.h"\n}, | ||
| 177 | Code=><<'EOD', | ||
| 178 | char *err = fits_rdecomp$TBSUL(_byte,_short,_short,)( (void *)$P(in), | ||
| 179 | $len(), | ||
| 180 | (void *)$P(out), | ||
| 181 | $SIZE(m), | ||
| 182 | $COMP(blocksize) | ||
| 183 | ); | ||
| 184 | if (err) $CROAK("%s", err); | ||
| 185 | EOD | ||
| 186 | ); | ||
| 187 | |||
| 188 | pp_done(); | ||