'''https://lz4.github.io/lz4/%|%LZ4%|%''' is a byte-oriented lossless compression algorithm optimized for speed. Its low-level implementations provide extremely fast decompression compared to the competitors. ** Tcl implementation ** The following pure Tcl implementation of the LZ4 decoder has been tested in Tcl 8.5, Tcl 8.6 and [Jim Tcl], but currently only works in Jim Tcl if it was compiled ''without'' UTF-8 support (see https://github.com/msteveb/jimtcl/blob/master/README.utf-8#L99%|%README.utf-8%|% for the reason). To decompress data compressed with LZ4 use the command `::lz4::decode $data $verify`. If `$verify` is true and the command `::xxhash::xxhash32` is available the decoder will verify the header and the content checksum in LZ4 frames, which lets you detect data corruption. A module implementing `::xxhash::xxhash32` is available on the wiki page [xxHash]. It requires [Critcl] to be fast. An alternative is to set `$verify` to zero, checksum the data before compression with a checksum algorithm implemented in C (e.g., [zlib crc32] in Tcl 8.6 or one of the [Tcllibc%|%C-accelerated] checksum commands in [Tcllib]) and verify its integrity yourself afterwards. Download with [wiki-reaper]: `wiki-reaper 48789 0 > lz4-0.2.3.tm` *** Code *** ====== # A pure Tcl LZ4 decoder. # Copyright (c) 2017 dbohdan # License: MIT # This decoder implements version 1.5 of the LZ4 frame spec. It can verify # checksums if the command ::xxhash::xxhash32 is available. It is known to work # in Tcl 8.5, Tcl 8.6, and Jim Tcl 0.74-0.77. namespace eval ::lz4 { variable version 0.2.3 # The following variable will be true in Jim Tcl and false in Tcl 8.x. variable jim [expr {![catch {info version}]}] } proc ::lz4::decode-block {data ptr endPtr window} { variable jim set result {} set rangeSubcmd [expr {$jim ? {byterange} : {range}}] while 1 { if {![binary scan $data "@$ptr cu" token]} { error {data truncated} } incr ptr 1 set litLen [expr {($token >> 4) & 0x0F}] set matchLen [expr {$token & 0x0F}] if {$litLen == 15} { while 1 { if {![binary scan $data "@$ptr cu" byte]} { error {data truncated} } incr ptr 1 incr litLen $byte if {$byte < 255} break } } if {![binary scan $data "@$ptr a$litLen" literals]} { error {data truncated} } incr ptr $litLen append window $literals append result $literals # The last sequence is incomplete. if {$ptr < $endPtr} { if {![binary scan $data "@$ptr su" offset]} { error {data truncated} } incr ptr 2 if {$matchLen == 15} { while 1 { if {![binary scan $data "@$ptr cu" byte]} { error {data truncated} } incr ptr 1 incr matchLen $byte if {$byte < 255} break } } incr matchLen 4 incr offset -1 set endOffset [expr { $offset - $matchLen > 0 ? $offset - $matchLen : 0 }] set overlapLen [expr { $offset - $matchLen > 0 ? 0 : $matchLen - $offset }] set match [string $rangeSubcmd $window end-$offset end-$endOffset] set matchRepeated [string repeat $match [expr { ($overlapLen / ($offset - $endOffset + 1)) + 2 }]] set matchWithOverlap [string $rangeSubcmd $matchRepeated \ 0 $matchLen-1] append window $matchWithOverlap append result $matchWithOverlap } if {$ptr == $endPtr} break if {$ptr > $endPtr} { error {read beyond block end} } } return [list $ptr $window $result] } proc ::lz4::decode-frame {data ptr verify} { # Decode and validate the header. if {![binary scan $data "@$ptr i" magic]} { error {data truncated} } incr ptr 4 set fieldsStartPtr $ptr if {$magic == 0x184D2204} { # Normal frame. } elseif {(0x184D2A50 <= $magic) && ($magic <= 0x184D2A5F)} { # Skippable frame. if {![binary scan $data "@$ptr iu" frameSize]} { error {data truncated} } incr ptr 4 incr ptr $frameSize return [list $ptr {}] } else { error "unexpected magic number: $magic" } set flags {} if {![binary scan $data "@$ptr cu cu" flags blockDescr]} { error {data truncated} } incr ptr 2 set flagsReserved [expr {($flags & 0b00000011) == 0}] set hasContentChecksum [expr {($flags & 0b00000100) == 0b00000100}] set hasContentSize [expr {($flags & 0b00001000) == 0b00001000}] set hasBlockChecksums [expr {($flags & 0b00010000) == 0b00010000}] set blockIndep [expr {($flags & 0b00100000) == 0b00100000}] set version [expr {($flags & 0b11000000) == 0b01000000}] if {!$flagsReserved} { error {FLG reserved bits aren't zero} } if {!$version} { error {frame version isn't "01"} } set blockDescrReserved [expr {($blockDescr & 0b10001111) == 0}] set blockMaxSize [expr {$blockDescr >> 4}] if {!$blockDescrReserved} { error {BD reserved bits aren't zero} } if {$blockMaxSize < 4} { error "invalid block maximum size ($blockMaxSize < 4)" } if {$hasContentSize} { if {![binary scan $data "@$ptr wu" uncompressedSize]} { error {data truncated} } incr ptr 8 } if {![binary scan $data "@$ptr cu" headerChecksum]} { error {data truncated} } if {$verify} { if {![binary scan $data \ "@$fieldsStartPtr a[expr {$ptr - $fieldsStartPtr}]" \ header]} { error {can't scan header fields to verify checksum\ (this shouldn't happen)} } if {(([::xxhash::xxhash32 $header 0] >> 8) & 0xff) != $headerChecksum} { error {frame header doesn't match checksum} } } incr ptr 1 # Decode the blocks. set window {} while 1 { if {![binary scan $data "@$ptr iu" blockSize]} { error {data truncated} } incr ptr 4 set compressed [expr {!($blockSize >> 31)}] set blockSize [expr {$blockSize & 0x7fffffff}] ;# Zero the highest bit. if {$blockSize == 0} break if {$compressed} { lassign [decode-block $data \ $ptr \ [expr {$ptr + $blockSize}] $window] \ ptr \ window \ decodedBlock if {$blockIndep} { set window {} } else { set window [string range $window end-0xFFFF end] } } else { if {![binary scan $data "@$ptr a$blockSize" decodedBlock]} { error {data truncated} } incr ptr $blockSize } append result $decodedBlock } # Decode the checksum. if {$hasContentChecksum} { if {![binary scan $data "@$ptr iu" contentChecksum]} { error {data truncated} } incr ptr 4 if {$verify && ([::xxhash::xxhash32 $result 0] != $contentChecksum)} { error {decoded data doesn't match checksum} } } return [list $ptr $result] } proc ::lz4::decode {data verify} { if {$verify && ([info commands ::xxhash::xxhash32] eq {})} { error {asked to verify checksums but [::xxhash::xxhash32] is absent} } set ptr 0 set result {} set len [string length $data] while {$ptr < $len} { lassign [decode-frame $data $ptr $verify] ptr frame append result $frame } return $result } proc ::lz4::assert-equal {actual expected} { if {$actual ne $expected} { if {[string length $actual] > 200} { set actual [string range $actual 0 199]... } if {[string length $expected] > 200} { set expected [string range $expected 0 199]... } error "expected \"$expected\",\n\ but got \"$actual\"" } } proc ::lz4::file-test {path canHash} { if {![file exists $path]} { puts stderr "can't find file \"$path\" -- skipping test" return } # Can't use -ignorestderr because Jim Tcl doesn't support it. if {[catch {exec lz4 --version 2>@1}]} { puts stderr {can't run lz4 -- skipping test} return } set ch [open $path rb] set data [read $ch] close $ch set ch [open [list |lz4 -c -9 $path]] fconfigure $ch -translation binary set dataCompressed [read $ch] close $ch assert-equal [decode $dataCompressed 0] $data if {$canHash} { assert-equal [decode $dataCompressed 1] $data } } proc ::lz4::value-test {compressed original canHash} { assert-equal [decode $compressed 0] $original if {$canHash} { assert-equal [decode $compressed 1] $original } } proc ::lz4::test {} { set canHash 1 if {[info commands ::xxhash::xxhash32] eq {}} { set canHash 0 puts stderr [join { {warning: [::xxhash::xxhash32] is absent; } {tests with checksum verification will be skipped} } {}] } set hello {Hello, World!} set helloCompressed [join { \x04\x22 \x4d\x18 \x64\x40 \xa7\x0d \x00\x00 \x80\x48 \x65\x6c \x6c\x6f \x2c\x20 \x57\x6f \x72\x6c \x64\x21 \x00\x00 \x00\x00 \xe8\x43 \xd0\x9e } {}] set seq {} for {set i 0} {$i < 4} {incr i} { append seq [string repeat $i 64] } set seqCompressed [join { \x04\x22 \x4d\x18 \x64\x40 \xa7\x1a \x00\x00 \x00\x1f \x30\x01 \x00\x2c \x1f\x31 \x01\x00 \x2c\x1f \x32\x01 \x00\x2c \x1f\x33 \x01\x00 \x27\x50 \x33\x33 \x33\x33 \x33\x00 \x00\x00 \x00\x80 \xf5\x97 \x31 } {}] set abc "abcabcabc 123123123 abcabcabc 123123123 abcabcabc123123123\n" set abcCompressed [join { \x04 \x22 \x4d \x18 \x64 \x40 \xa7 \x1d \x00 \x00 \x00 \x32 \x61 \x62 \x63 \x03 \x00 \x42 \x20 \x31 \x32 \x33 \x03 \x00 \x1f \x20 \x14 \x00 \x0a \xa0 \x31 \x32 \x33 \x31 \x32 \x33 \x31 \x32 \x33 \x0a \x00 \x00 \x00 \x00 \xc3 \x67 \x9d \xbf } {}] set blah [join { {1blah2HELLOblah3blah4blah foo bar blah !!!! } {213218372132-------------------} } {}] set blahCompressed [join { \x04 \x22 \x4d \x18 \x64 \x40 \xa7 \x3c \x00 \x00 \x00 \xb0 \x31 \x62 \x6c \x61 \x68 \x32 \x48 \x45 \x4c \x4c \x4f \x0a \x00 \x10 \x33 \x05 \x00 \x10 \x34 \x05 \x00 \x91 \x20 \x66 \x6f \x6f \x20 \x62 \x61 \x72 \x20 \x0d \x00 \xd0 \x21 \x21 \x21 \x21 \x20 \x32 \x31 \x33 \x32 \x31 \x38 \x33 \x37 \x08 \x00 \x19 \x2d \x01 \x00 \x50 \x2d \x2d \x2d \x2d \x2d \x00 \x00 \x00 \x00 \xa4 \xeb \xf6 \xac } {}] puts stderr {running tests} puts stderr {--- hello} value-test $helloCompressed $hello 0 if {$canHash} { puts stderr {--- checksum error} catch {decode $helloCompressed 1} err assert-equal $err {decoded data doesn't match checksum} } puts stderr {--- seq} value-test $seqCompressed $seq $canHash puts stderr {--- abc} value-test $abcCompressed $abc $canHash puts stderr {--- blah} value-test $blahCompressed $blah $canHash puts stderr {--- empty} value-test {} {} $canHash puts stderr {--- truncated} catch {decode [string range $seqCompressed 0 20] 0} err assert-equal $err {data truncated} puts stderr {--- multiple frames} assert-equal [decode $helloCompressed$seqCompressed$helloCompressed 0] \ $hello$seq$hello puts stderr {--- passwd} file-test /etc/passwd $canHash puts stderr {--- sh (binary)} file-test /bin/sh $canHash } # If this is the main script... if {[info exists argv0] && ([file tail [info script]] eq [file tail $argv0])} { if {$argv eq {--test}} { ::lz4::test } elseif {[lindex $argv 0] ne {}} { set verify 1 set argv [lassign $argv a] if {$a eq {--ignore-checksums}} { set verify 0 set argv [lassign $argv a] } if {$a eq {-}} { set ch stdin fconfigure stdin -translation binary } else { set ch [open $a rb] } set data [read $ch] close $ch fconfigure stdout -translation binary puts -nonewline [::lz4::decode $data $verify] } else { puts "usage: [info script] (--test | \[--ignore-checksums\] filename)" puts "Use \"-\" as the filename to read from the standard input." } } ====== *** Performance comparison *** ======none > echo 'puts [info patchlevel]' | tclsh 8.6.5 > lz4 -k -9 ffmpeg.exe Compressed 40859136 bytes into 18846155 bytes ==> 46.12% > time lz4 -d -c ffmpeg.exe.lz4 > /dev/null 0.03user 0.03system 0:00.08elapsed 76%CPU > time tclsh lz4.tcl ffmpeg.exe.lz4 > /dev/null 18.59user 0.32system 0:19.05elapsed 99%CPU ====== ** See also ** * [brotli] * [Zstandard], a faster alternative to [zlib] by the same developer as LZ4 <>Compression | Jim Package | Package