Something is wrong here: $bits 286 (0=bits)-:0=F1 0 +/0=F1 9 +/0=bits 26
0 bits should mean that that symbol is not represented. But there are only 9 such symbols in F1. -- Raul On Wed, Sep 10, 2014 at 11:09 PM, bill lam <[email protected]> wrote: > I guess the description of the algorithm itself is correct, but > it does not mention how to get the bits of each symbol. The bit > lengths for classic and zlib huffman can be different. > > for testing, the bits for A from libz.so are > > ":&>_10<\bits > 10 10 10 10 10 10 10 10 10 10 > 10 10 10 10 10 10 10 10 10 10 > 10 10 10 10 10 10 10 10 10 10 > 10 10 10 10 10 10 10 10 10 10 > 10 10 10 10 10 10 10 10 10 10 > 10 10 10 10 10 10 10 10 10 10 > 10 10 10 10 10 10 10 10 10 10 > 10 10 10 10 10 10 10 10 10 10 > 10 10 10 10 10 10 10 10 10 10 > 10 10 10 10 10 10 10 10 10 10 > 10 10 10 10 10 10 10 10 10 10 > 10 10 10 10 10 10 10 10 10 10 > 10 10 10 10 10 10 10 10 10 10 > 10 10 10 10 10 10 10 10 10 10 > 10 10 10 11 11 10 10 10 10 11 > 11 10 10 10 10 10 10 10 10 10 > 10 10 10 10 10 10 10 10 10 10 > 10 10 10 10 10 10 10 10 10 10 > 10 10 10 10 10 10 10 10 10 10 > 10 10 10 10 10 10 10 10 10 10 > 10 10 10 10 10 10 10 10 10 10 > 10 10 10 10 10 10 10 10 10 10 > 10 10 10 10 10 10 10 10 10 10 > 10 10 10 10 10 10 10 10 10 10 > 10 10 10 10 10 10 10 10 10 10 > 10 10 10 10 10 10 10 0 0 0 > 0 0 0 0 0 0 0 0 0 0 > 0 0 0 0 10 2 0 0 0 0 > 0 0 0 0 0 1 > > this is the huffman dictionary stored in zlib stream and the actual > huffman codes are computed using the algorithm mentioned during decoding. > > Ср, 10 сен 2014, Raul Miller написал(а): >> I agree that there should not be any 2s. >> >> However, I think this means that there's something wrong with the >> description of the algorithm, or at least my understanding of it. >> >> I'll need some time to digest this. >> >> Thanks, >> >> -- >> Raul >> >> On Wed, Sep 10, 2014 at 9:39 PM, bill lam <[email protected]> wrote: >> > pardon me of forgetting telling another cavaet. >> > zlib huffman code is suboptimal so that bit length of code >> > for each symbol can be longer than that in classic huffman. >> > eg in my orignal data. >> > >> > 10{. ":@>F deflatecodes2 A >> > 1 1 0 0 0 0 0 0 0 0 >> > 1 1 1 1 1 1 1 1 1 1 0 >> > 1 1 1 1 1 1 1 1 1 1 1 >> > 1 1 1 1 1 1 1 1 1 2 0 >> > 1 1 1 1 1 1 1 1 1 2 1 >> > 1 1 1 1 1 1 1 1 2 1 0 >> > 1 1 0 0 0 0 0 0 0 1 >> > 1 1 0 0 0 0 0 0 1 0 >> > 1 1 0 0 0 0 0 0 1 1 >> > 1 1 0 0 0 0 0 1 0 0 >> > >> > there shouldn't be any 2's. It needs to increase bit length of >> > some code. >> > >> > Ср, 10 сен 2014, Raul Miller написал(а): >> >> Oops, sorry about that. >> >> >> >> Try it this way: >> >> >> >> deflatecodes2=:4 :0 >> >> L=. #@> x hcodes y >> >> U=. 0,~.L >> >> R=. ;<@(({. >./@(>#])&U #1:)@{. <@:+"1-@{.{."1 #:@i.@#)/.~L >> >> R/:;(</. i.@#)L >> >> ) >> >> >> >> ":@>F1 deflatecodes2 A1 >> >> 0 >> >> 1 1 0 >> >> 1 1 1 1 1 1 0 >> >> 1 1 1 1 1 1 1 >> >> 1 1 1 0 0 0 >> >> 1 1 1 0 0 1 >> >> 1 1 1 0 1 0 >> >> 1 1 1 0 1 1 >> >> 1 1 1 1 0 0 >> >> 1 1 1 1 0 1 >> >> 1 1 1 1 1 0 >> >> 1 0 >> >> >> >> Thanks, >> >> >> >> -- >> >> Raul >> >> >> >> >> >> On Wed, Sep 10, 2014 at 7:26 PM, bill lam <[email protected]> wrote: >> >> > for zlib, all huffman code of the same bit length are in sequential, >> >> > starting with a code an 0 bit in the end (even number), and no gaps in >> >> > the >> >> > block (consecutive). the symbols of these block represented are >> >> > sequential >> >> > but not consecutive (possibly gaps). So the zlib huffman code is >> >> > slightly >> >> > less efficient than the original huffman code but the advantage is >> >> > simpler >> >> > to store the table, just the bits used. section 3.2.6 gives an example >> >> > of >> >> > such table. >> >> > >> >> > deflatecodes can satisfy the 2 rules _but_ its result is invalid because >> >> > huffman code is a prefix coding. >> >> > >> >> > ,.F1 ( deflatecodes) A1 >> >> > ┌─────────────┐ >> >> > │0 │ >> >> > ├─────────────┤ >> >> > │1 1 0 │ >> >> > ├─────────────┤ >> >> > │1 1 1 0 0 1 0│ >> >> > ├─────────────┤ >> >> > │1 1 1 0 0 1 1│ >> >> > ├─────────────┤ >> >> > │1 1 1 0 0 0 │ >> >> > ├─────────────┤ >> >> > │1 1 1 0 0 1 │ >> >> > ├─────────────┤ >> >> > │1 1 1 0 1 0 │ >> >> > ├─────────────┤ >> >> > │1 1 1 0 1 1 │ >> >> > ├─────────────┤ >> >> > │1 1 1 1 0 0 │ >> >> > ├─────────────┤ >> >> > │1 1 1 1 0 1 │ >> >> > ├─────────────┤ >> >> > │1 1 1 1 1 0 │ >> >> > ├─────────────┤ >> >> > │1 0 │ >> >> > └─────────────┘ >> >> > 7 bit 1110010 is a prefix of 6 bit 111001 and this is illegal. >> >> > Instead 7 bit code block should start with >> >> > #.inv 2b11100 + 7 >> >> > 1 1 1 1 1 1 0 >> >> > because there are 6 symbols in the 6 bit code. (algorithm given in >> >> > section >> >> > 3.2.2). >> >> > >> >> > Any idea to fix deflatecodes so that it can give valid huffman codes? >> >> > Thanks. >> >> > On Sep 11, 2014 1:01 AM, "Raul Miller" <[email protected]> wrote: >> >> > >> >> >> I think the use of the term "consecutive" rather than "sequential" is >> >> >> telling. >> >> >> >> >> >> The described algorithm is: compute the huffman code lengths: >> >> >> #@>F1 hcodes A1 >> >> >> 1 3 7 7 6 6 6 6 6 6 6 2 >> >> >> >> >> >> Then assign ascending huffman codes first in length order and then >> >> >> within codes of the same length. >> >> >> >> >> >> Taken literally, that might be something like this: >> >> >> >> >> >> H=: 4 :0 >> >> >> L=.#@> x hcodes y >> >> >> U=.~.L >> >> >> ;<@(({.{.U e.~i.&.<:@{.)<@:+"1-@{.{."1 #:@i.@#)/.~L >> >> >> ) >> >> >> >> >> >> ":@>F1 H A1 >> >> >> 0 >> >> >> 1 1 0 >> >> >> 1 1 1 0 0 1 0 >> >> >> 1 1 1 0 0 1 1 >> >> >> 1 1 1 0 0 0 >> >> >> 1 1 1 0 0 1 >> >> >> 1 1 1 0 1 0 >> >> >> 1 1 1 0 1 1 >> >> >> 1 1 1 1 0 0 >> >> >> 1 1 1 1 0 1 >> >> >> 1 1 1 1 1 0 >> >> >> 1 0 >> >> >> >> >> >> But is this correct? Is it actually safe to leave the results like >> >> >> this - with all codes of the same length being consecutive to each >> >> >> other? >> >> >> >> >> >> F (hcodes -:&:(#@>) H) A >> >> >> 0 >> >> >> >> >> >> No. >> >> >> >> >> >> So... "consecutive" must refer only to the values used and not their >> >> >> order within the result. >> >> >> >> >> >> Perhaps something like this: >> >> >> >> >> >> deflatecodes=:4 :0 >> >> >> L=.#@> x hcodes y >> >> >> U=.~.L >> >> >> R=. ;<@(({.{.U e.~i.&.<:@{.)<@:+"1-@{.{."1 #:@i.@#)/.~L >> >> >> R/:;(</. i.@#)L >> >> >> ) >> >> >> >> >> >> F (hcodes -:&:(#@>) deflatecodes) A >> >> >> 1 >> >> >> >> >> >> There should be a better way of doing this, but this should at least >> >> >> get you started. >> >> >> >> >> >> Thanks, >> >> >> >> >> >> -- >> >> >> Raul >> >> >> >> >> >> >> >> >> On Wed, Sep 10, 2014 at 10:45 AM, bill lam <[email protected]> wrote: >> >> >> > For huffman coding used in zlib: >> >> >> > https://www.ietf.org/rfc/rfc1951.txt section 3.2.2. >> >> >> > >> >> >> > The Huffman codes used for each alphabet in the "deflate" >> >> >> > format have two additional rules: >> >> >> > >> >> >> > * All codes of a given bit length have lexicographically >> >> >> > consecutive values, in the same order as the symbols >> >> >> > they represent; >> >> >> > >> >> >> > * Shorter codes lexicographically precede longer codes. >> >> >> > I tried jwiki hcodes in >> >> >> > I try Roger's essay >> >> >> > http://www.jsoftware.com/jwiki/Essays/Huffman%20Coding >> >> >> > >> >> >> > hc=: 4 : 0 >> >> >> > if. 1=#x do. y >> >> >> > else. ((i{x),+/j{x) hc (i{y),<j{y [ i=. (i.#x) -. j=. 2{./:x end. >> >> >> > ) >> >> >> > >> >> >> > hcodes=: 4 : 0 >> >> >> > assert. x -:&$ y NB. weights and words have same shape >> >> >> > assert. (0<:x) *. 1=#$x NB. weights are non-negative >> >> >> > assert. 1 >: L.y NB. words are boxed not more than once >> >> >> > w=. ,&.> y NB. standardized words >> >> >> > assert. w -: ~.w NB. words are unique >> >> >> > t=. 0 {:: x hc w NB. minimal weight binary tree >> >> >> > ((< S: 0 t) i. w) { <@(1&=)@; S: 1 {:: t >> >> >> > ) >> >> >> > >> >> >> > but the coding produced is malformed for zlib. eg, >> >> >> > this is what I ran into trouble >> >> >> > >> >> >> > f1=: 1 256 17 1 1 9 1 >> >> >> > f2=: 2 1 0 1 255 0 1536 >> >> >> > F=: ,/(f1#f2) >> >> >> > A=: i.286 >> >> >> > >> >> >> > F hcodes A >> >> >> > >> >> >> > Or a shorter example >> >> >> > >> >> >> > A1=: i.12 >> >> >> > F1=: 2 1 0 0 0 0 0 0 0 0 0 1 >> >> >> > >> >> >> > F1 hcodes A1 >> >> >> > >> >> >> > Any idea? >> >> >> > >> >> >> > -- >> >> >> > regards, >> >> >> > ==================================================== >> >> >> > GPG key 1024D/4434BAB3 2008-08-24 >> >> >> > gpg --keyserver subkeys.pgp.net --recv-keys 4434BAB3 >> >> >> > gpg --keyserver subkeys.pgp.net --armor --export 4434BAB3 >> >> >> > ---------------------------------------------------------------------- >> >> >> > For information about J forums see >> >> >> > http://www.jsoftware.com/forums.htm >> >> >> ---------------------------------------------------------------------- >> >> >> For information about J forums see http://www.jsoftware.com/forums.htm >> >> >> >> >> > ---------------------------------------------------------------------- >> >> > For information about J forums see http://www.jsoftware.com/forums.htm >> >> ---------------------------------------------------------------------- >> >> For information about J forums see http://www.jsoftware.com/forums.htm >> > >> > -- >> > regards, >> > ==================================================== >> > GPG key 1024D/4434BAB3 2008-08-24 >> > gpg --keyserver subkeys.pgp.net --recv-keys 4434BAB3 >> > gpg --keyserver subkeys.pgp.net --armor --export 4434BAB3 >> > ---------------------------------------------------------------------- >> > For information about J forums see http://www.jsoftware.com/forums.htm >> ---------------------------------------------------------------------- >> For information about J forums see http://www.jsoftware.com/forums.htm > > -- > regards, > ==================================================== > GPG key 1024D/4434BAB3 2008-08-24 > gpg --keyserver subkeys.pgp.net --recv-keys 4434BAB3 > gpg --keyserver subkeys.pgp.net --armor --export 4434BAB3 > ---------------------------------------------------------------------- > For information about J forums see http://www.jsoftware.com/forums.htm ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
