Certainly this is correct and that why I asked I had misread.
for any given text, dynamic huffman coding always encoded with
exactly the same table and there is a one to one mapping between bit
length vector and huffman codes.
I think hcodes is also stable in that it generates the same huffman
co
I thought the bit lengths were encoded in the file, not recreated by
an algorithm in the decoder?
Thanks,
--
Raul
On Fri, Sep 12, 2014 at 1:27 AM, bill lam wrote:
> I think this matters a lot because output of our encoder will
> also be decoded by other zlib compliant decoders such as zlib.so
I think this matters a lot because output of our encoder will
also be decoded by other zlib compliant decoders such as zlib.so
itself. Or I had mis-read your question?
Пт, 12 сен 2014, Raul Miller написал(а):
> If we define
>
> bitlen=:4 :0
> b=. 0~:x
> b #inv #@>(b#x) hcodes b#y
> )
>
> Doe
If we define
bitlen=:4 :0
b=. 0~:x
b #inv #@>(b#x) hcodes b#y
)
Does it matter that these numbers are sometimes different from those
used in another implementation?
Thanks,
--
Raul
On Fri, Sep 12, 2014 at 12:59 AM, bill lam wrote:
> Consider an example
>
>1 0 0 2 1 hcode
Hi Raul,
This article seems containing an algorithm for generating
dynamic huffman coding in a human language.
An Explanation of the `Deflate' Algorithm
http://www.zlib.net/feldspar.html
--
regards,
GPG key 1024D/4434BAB3 2008-08-24
gpg --k
Consider an example
1 0 0 2 1 hcodes_jzlib_ i.5
+-+---+---+-+---+
|1 1 1|1 1 0 0|1 1 0 1|0|1 0|
+-+---+---+-+---+
#&> 1 0 0 2 1 hcodes_jzlib_ i.5
3 4 4 1 2
(0~:1 0 0 2 1)* #&> 1 0 0 2 1 hcodes_jzlib_ i.5
3 0 0 1 2
classic huf
That's what I thought at first, also.
But, let's look at the example at
http://www.jsoftware.com/pipermail/programming/2014-September/039299.html
and the bit widths given at
http://www.jsoftware.com/pipermail/programming/2014-September/039327.html
Here's how it looks to me:
bits -:#@>F hcode
http://www.jsoftware.com/jwiki/Essays/Collatz%20Conjecture
On Thu, Sep 11, 2014 at 11:16 PM, Neill Robson
wrote:
> I've been working on various Project Euler problems in J, and I seem to
> always get stuck whenever a problem requires excessive/lengthy recursion.
> In particular, I've been workin
I've been working on various Project Euler problems in J, and I seem to
always get stuck whenever a problem requires excessive/lengthy recursion.
In particular, I've been working on problem number 14, which asks what
starting value (under 1 million) produces the longest Collatz sequence.
Now, I st
Thanks. I prefer reading essays in wikipedia or rfc for
human language descriptions rather than guessing from
computer language codes (perhpas exclusing J).
rfc 1950 and 1951 for png and zlib for specification is self
contained except for the algorithm of getting the dynamic
huffman code (not f
One more link:
http://www.opensource.apple.com/source/zlib/zlib-43/zlib/contrib/puff/puff.c
* puff.c is a simple inflate written to be an unambiguous way to specify the
* deflate format. It is not written for speed but rather simplicity.
Not sure if it's any use
On Thu, Sep 11, 2014 at 9:46
Thanks for the background. I don't know if I'm helping any but figured
I'd share one more thing (for now). I found a library,
https://github.com/lvandeve/lodepng, that has a permissive license and
successfully decoded and encoded a png file.
As simple as:
lodepng_decode32_file(&image, &width, &he
Yes, it does indeed:
u=. +
((u^:((# - 1:))@:{.) -: u/) (i.1)
1
Also,
erase'u'
1
((u^:((# - 1:))@:{.) -: u/) (i.1)
1
I have to digest it a little bit more but so far everything seems to be
falling into place; at least, your interpretation is consistent.
Thanks
On Thu, Sep 11, 20
It seems everyone use huffman code in its own way, at least huffman coding
in jpeg is different from that in deflate.
The intent to is to replace bmp with png as the default image format in jal
addons for j803. This is already almost fully done. With dynamic huffman
coding, png images can achieve
The bit widths are calculated from the huffman tree
See
http://stackoverflow.com/questions/759707/efficient-way-of-storing-huffman-tree
http://www.siggraph.org/education/materials/HyperGraph/video/mpeg/mpegfaq/huffman_tutorial.html
The timing is interesting considering we were talking about tre
Remember that the definition of u/ was:
"u/y applies the dyad u between the items of y".
The items of i.1 are {.i.1
And:
({.i.1) -: +/i.1
1
Does that make sense? (And, yes, I intentionally mangled english
grammar in my third sentence of this message.)
Thanks,
--
Raul
On Thu, Sep 11, 201
Now, that is interesting. Thus, it seems, the unofficial documentation
about / is not quite correct (depending on what the meaning of "is" is.).
I have another question: if "u is applied (#y)-1 times." How come,
(-: u^:0)i.1
1
?
On Thu, Sep 11, 2014 at 6:36 PM, Raul Miller wrote:
> y a
y always remains unchanged. Note also that the result is distinct from y:
(-: +/) i.1
0
Thanks,
--
Raul
On Thu, Sep 11, 2014 at 3:13 PM, Jose Mario Quintana
wrote:
> My only guess is:
> "m/y inserts successive verbs from the gerund m between items of y"
>
> So, if there is no "between item
Certainly.
"u/y applies the dyad u between the items of y".
This means that u is applied (#y)-1 times.
And an error would be inconsistent with the concept expressed for the
case where 0=#y.
Thanks,
--
Raul
On Thu, Sep 11, 2014 at 1:50 PM, Dan Bron wrote:
> Pepe wrote:
>> My only question is
Here's the code I came up with, with Bill's help:
bl_count=:3 :0 NB. y is result of freqs
0,}.<:#/.~(,~ [: i. 1 + >./)y
)
start_vals=: +:@+/\.&.|.@}:@,~&0
find_codes=:3 :0 NB. y is result of freqs
b=. bl_count y
v=. start_vals b
n=. /:~ ~.y-.0
o=. ;({./.~ /:~ (n (([#2:) #: ])&.> (*b)#v+&.>
I wrote:
"
I do not think the dictionary is wrong but perhaps it is incomplete. As you
pointed out, the tally is 1 and the neutral element intance provision does
not apply; so, for example, in the sentence (*/0) or in general (u/A) where
A is an atom the result is the atom (A) (even if u is undefin
Even NuVuc refers back to the DoJ as the authoritative, normative reference.
If pointing out gaps in the Dictionary on the Forums isn't the way to get it
updated, what is?
Please excuse typos; sent from a phone.
> On Sep 11, 2014, at 4:45 PM, Henry Rich wrote:
>
> Crying about the Dictionar
bill, I'd be interested in a solution but I don't think I can
contribute any more on this. I played with
https://code.google.com/p/miniz/ and became even more convinced of the
complexity. It seems as though the compressor can decide whether to
include the dictionary code table or not -- likely base
Crying about the Dictionary isn't going to get things documented. Yeah,
it's documented in NuVoc.
Henry Rich
On 9/11/2014 3:13 PM, Jose Mario Quintana wrote:
My only guess is:
"m/y inserts successive verbs from the gerund m between items of y"
So, if there is no "between items of y" inserts
Agreed.
- Original Message ---
Subject: Re: [Jprogramming] Replace one item of a list
From: Jose Mario Quintana
Date: Thu, 11 Sep 2014 16:05:26 -0400
To: Programming forum
I copied the wrong text from the Dictionary; it should have been:
"
u/y applies the dyad u be
I copied the wrong text from the Dictionary; it should have been:
"
u/y applies the dyad u between the items of y
"
Incidentally, as Thomas proposed, if it cannot insert the verb, then
issuing an error could not have been a valid alternative as well?
On Thu, Sep 11, 2014 at 3:13 PM, Jose Mario Q
I don't know matlab, but is this what you are looking for
]A=:i.2 4 7
0 1 2 3 4 5 6
7 8 9 10 11 12 13
14 15 16 17 18 19 20
21 22 23 24 25 26 27
28 29 30 31 32 33 34
35 36 37 38 39 40 41
42 43 44 45 46 47 48
49 50 51 52 53 54 55
+/"2 A
42 46 50 54 58 62 66
154 158 162 1
My only guess is:
"m/y inserts successive verbs from the gerund m between items of y"
So, if there is no "between items of y" inserts nothing and y remains
unchanged; but, it seems to me that the Dictionary could be more assertive
in this instance.
On Thu, Sep 11, 2014 at 2:13 PM, Dan Bron wrot
Pepe wrote:
> My only question is: Does the Dictionary support this behavior?
> Raul responded:
> Yes, it does.
I replied:
> I am intrigued. Can you elaborate?
Thomas followed-up:
> I assumed that by not mentioning it, the implementation
> is free to do what it chooses. It could be anythi
On Thu, Sep 11, 2014 at 1:50 PM, Dan Bron wrote:
> Pepe wrote:
> > My only question is: Does the Dictionary support this behavior?
>
> Raul responded:
> > Yes, it does.
>
> I am intrigued. Can you elaborate?
>
I assumed that by not mentioning it, the implementation is free to do what
it chooses
What I would expect to happen with u/ Y where 1=#Y is: if the neutral
element (E) is known then E u Y otherwise a domain error, with the ability
to specify E through (!.). I think my mindset is a bit more pure where as
the implementation is a bit more practical, albeit, as Pepe said, arbitrary.
On
Pepe wrote:
> My only question is: Does the Dictionary support this behavior?
Raul responded:
> Yes, it does.
I am intrigued. Can you elaborate?
In particular, can you help me see, given only the text of the Dictionary,
that u/y where 1=#y must be y ? And not, for example, an error?
-D
This codes seemed invalid.
1 is a prefix of 11 which is a prefix of 111. Suppose there
is a bit pattern of 1 1 , it is ambiguous to mean
[68,'1'] [68,'1']
or [65,'11']
The huffman code in rfc is canonical meaning there is exactly one
possible huffman codes for a given bit length vector. This
Pepe wrote:
> My only question is: Does the Dictionary support
> [the behavior that y ? u/y when 1=#y ]?
No, I don't believe it does. At the very least, it's not supported by the
vocabulary entry for / . That said, there are other areas where such
reasoning applies (e.g. ;. ), so maybe it
On Thu, Sep 11, 2014 at 1:02 PM, Jose Mario Quintana
wrote:
> My only question is: Does the Dictionary support this behavior?
Yes, it does.
Thanks,
--
Raul
--
For information about J forums see http://www.jsoftware.com/forums.
The important thing to understand is that Rank is a count of dimensions.
So u"1 means that the arguments to u have (at most) 1 dimension, and u"2
means that the arguments to u have (at most) 2 dimensions. Any 'extra'
dimensions get treated independently.
See also: https://en.wikipedia.org/wiki/Ra
Joe,
both of the 2 rules in rfc involve bit lengths, so that your
output would probably also need to display the bit length of
each code for validation.
try something like "ABDD" you may expect bit length
of each symbol shuold be something like A <: C <: D <: B
This may or may not since d
I wrote:
"
I do not think the dictionary is wrong but perhaps it is incomplete. As you
pointed out, the tally is 1 and the neutral element intance provision does
not apply; so, for example, in the sentence (*/0) or in general (u/A) where
A is an atom the result is the atom (A) (even if u is undefin
David Lambert wrote:
> I wrote the rosetta code red-black tree entry.
I haven't read/reviewed the Red-Black tree implementation on RosettaCode,
but it did remind me of an argument I once made for supplying I. with an
inverse [1]:
> I.-space is useful, particularly when the domain (input & ou
I wrote the rosetta code red-black tree entry. As Devon guessed, I
followed Robert Sedgewick's left leaning red-black tree paper. As the
notes page indicates, merely giving an example of symbols relies on
implementation detail. I'll post such an entry if there's an
agreement. 1r3 of the cod
First of all, rank conjunction <> axis specification. Matlab
seems using axis specification although I know nothing about it.
+/ or +/"r always insert + between "items" but will not change the
order of axis. Items of a vector is scalar; Items of a matrix is
its row; item of i.2 4 7 is a matrix of
For practical reasons the items of a vector are listed horizontally but the
items of higher dimensional arrays are listed vertically. You must
remember this when you interpret displays.
To illustrate, in
i. 3 2 4
0 1 2 3
4 5 6 7
8 9 10 11
12 13 14 15
16 17 18 19
20 21 22 23
I do not know where to find the documentation easily. However, the
definitions lurk in the z locale.
inv
^:_1
names_z_ 3
...
copath
copathnl
copathnlx
coreset
...
winpathsep
xedit
names_z_ 1
define
each
every
fapplylines
inv
inverse
items
leaf
rows
rxapply
rxmerge
table
On 09/
Yep, that's it. And, being that there's only one argument to work with
(the scalar), there's insufficient information to actually invoke the
dyad, so it isn't.
-Dan
- Original Message ---
Subject: Re: [Jprogramming] Replace one item of a list
From: David Lambert
Date: Thu
Ignore the pako.js example output... It was just outputting the binary
representation of A-Z, not the huffman code
This is what I meant to send
For ABCD:
[65,'11'],
[66,'0'],
[67,'10'],
[68,'1'],
[256,'111']
It still doesn't seem to be sorting correctly lexographically, but I'm
not really in my
Perhaps the confusion is simply that the insert adverb implies dyadic
invocation of the verb.
On Wed, Sep 10, 2014 at 9:57 AM, Linda Alvord
> >wrote:
> >
> >>
> >>Here's the example that is puzzling me.
> >>
> >> *i.5
> >>0 1 1 1 1
> >>*/"0 i.5
> >>0 1 2 3 4
> >>
> >>
> >It looks like
I think the prefix coding looks OK, but the 2 rules does not:
I modified the code[1] to allow passing in a string and outputting the codes
C:\temp>deflate ABCDEFGHIJKLMONPQRSTUVWXYZ
code 65 : 0
code 66 : 6 0110
code 67 : 8 0
Kyle M. Rudden wrote:
> Is there is list of non-primitive standard structures
> somewhere I am missing?
>
> NuVoc and other vocabulary lists tend to have primitives only.
> Where would I find a discussion / explanation of "#inv" on the wiki?
>
> Useful non-primitive structures like "each"
This is strange since every author must had decode its own encoded
data as a smoke test.
Did you test if huffman code or bit lengths it produced was
correct or not, ie it is a prefix coding and it satisfy the 2
rules in rfc.
Чт, 11 сен 2014, Joe Bogner написал(а):
> unfortunately the dynamic codi
to get a column sum, all of these would work
+/ +/"_1 i.2 4 7
196 204 212 220 228 236 244
+/ +/"2 i.2 4 7
196 204 212 220 228 236 244
+/ +/ i.2 4 7
/ or +/ in this case will only "remove" (reduce along) one dimension at a time.
- Original Message -
From: George Dallas
To: pro
> In my experience so far, very useful non-primitive structures like "each"
you have to discover by reading all available materials and monitoring
these discussions.
See the standard library docs at
http://www.jsoftware.com/help/user/library.htm .
For example, this includes the expand verb.
I al
as nice as # inv is, you may find this adverb more flexible and intuitive:
insertitem =: 1 : '(u"_ {. ]) , [ , u"_ }. ]'
the adverb argument is a verb that tells which item position to insert. The
verb argument is dyadic, but can be a constant verb. Add @:] to it to just use
y as the argument
Let’s say we have a rank 3 array A like so:
]A=.i.2 4 7
0 1 2 3 4 5 6
7 8 9 10 11 12 13
14 15 16 17 18 19 20
21 22 23 24 25 26 27
28 29 30 31 32 33 34
35 36 37 38 39 40 41
42 43 44 45 46 47 48
49 50 51 52 53 54 55
And we would like to add the rows. The rank conjunction
Useful and interesting ... but where would I find a discussion / explanation of
"#inv" on the wiki?
NuVoc and other vocabulary lists tend to have primitives only.
In my experience so far, very useful non-primitive structures like "each" you
have to discover by reading all available materials and
> Ben Gorte wrote:
> > I use an adverb called ins that behaves like } , except that it inserts :)
> In the argument to the adverb, does 1 ins refer to the item at index 1 in
> the argument, or the empty space between item 1 and item 2, or the empty
> space between item 0 and item 1, or what?
The
My solutions usually are from a beginner's attach on the problem, but
sometimes this may be helpful.
]A=:3 3$0
0 0 0
0 0 0
0 0 0
]B=: 0 1 3 2 {"1 A(,"1) 1
0 0 1 0
0 0 1 0
0 0 1 0
]C=:0 1 3 2{"2 B(,"2) 1
0 0 1 0
0 0 1 0
1 1 1 1
0 0 1 0
]D=:0 1 3{"2 C
0 0 1 0
0 0 1 0
0 0
unfortunately the dynamic coding in the putty fork doesn't seem to work:
deflate -c deflate.c > out
deflate -d out
decoding error: incorrect data checksum
it works fine with static tables
C:\temp>echo ABCD > ABCD
C:\temp>deflate -c ABCD > out
C:\temp>deflate -d out
ABCD
I added some debuggi
Fractional axes specs are an abomination. Don't bother trying to model it.
,: and ,:"r are far superior.
On Thu, Sep 11, 2014 at 7:00 AM, Dan Bron wrote:
> Ben Gorte wrote:
> > I use an adverb called ins that behaves like } , except that it inserts
> :)
>
> In the argument to the adverb, does
Ben Gorte wrote:
> I use an adverb called ins that behaves like } , except that it inserts :)
In the argument to the adverb, does 1 ins refer to the item at index 1 in
the argument, or the empty space between item 1 and item 2, or the empty
space between item 0 and item 1, or what?
If I recall co
> I use a verb called ins that behaves like } , except that it inserts :)
Correction: adverb
Sorry about that,
Ben
--
For information about J forums see http://www.jsoftware.com/forums.htm
#^:_1 is discussed at
http://www.jsoftware.com/jwiki/Vocabulary/Inverses#Expand_.23.5E:_1
and other inverses are discussed on the same page.
Henry Rich
On 9/11/2014 7:04 AM, Sebastian wrote:
Hi Raul and Dan,
Many thanks for your great help! :)
I found # as copy in the j vocabulary now, but
If you find the inv-thing quirky and complex numbers too, you can use e.g. 1 2
1#M to duplicate a row of M. Then put something else at the copy using }
(amend), but you were probably going to do that anyhow.
I use a verb called ins that behaves like } , except that it inserts :)
888 (3) ins
# inv is the same as #^:_1 (using the power conjunction). A list of
default inverses can be found at
http://www.jsoftware.com/help/dictionary/d202n.htm
#^:_1 is listed under point 6.
Kind regards,
Jan-Pieter
2014-09-11 13:04 GMT+02:00 Sebastian :
> Hi Raul and Dan,
>
> Many thanks for your grea
Why do the words matter? It seems the hcodes are independent of the words.
A
abcdef
F
3 1 4 1 5 9
F hcodes A
┌─┬───┬───┬───┬───┬───┐
│1 0 1│1 0 0 0│0 0│1 0 0 1│0 1│1 1│
└─┴───┴───┴───┴───┴───┘
F hcodes (?6#256){a.
┌─┬───┬───┬───┬───┬──
Hi Raul and Dan,
Many thanks for your great help! :)
I found # as copy in the j vocabulary now, but I cannot find #inv. Where
I have to search?
Sebastian
-- Originalnachricht --
Von: "Raul Miller"
An: "Programming forum"
Gesendet: 11.09.2014 12:58:37
Betreff: Re: [Jprogramming] Ext
Here's an example of how I frequently use non-truth-valued #
9!:7]9 1 1#'+|-'
Thanks,
--
Raul
On Thu, Sep 11, 2014 at 6:49 AM, Dan Bron wrote:
> Note also that you can use plain, uninverted # to expand arguments:
>
>1 1j1 1 (#!.1) 3 3 $ 0
> 0 0 0
> 0 0 0
> 1 1 1
> 0 0 0
>1 1j1 1 (
-Original Message-
From: programming-boun...@forums.jsoftware.com
[mailto:programming-boun...@forums.jsoftware.com] On Behalf Of bill lam
Sent: Thursday, September 11, 2014 3:46 AM
To: Programming forum
Subject: Re: [Jprogramming] zlib huffman coding
the frequencies (guessing from bit l
Here's a corrected implementation of find_codes
bl_count=:3 :0
0,}.<:#/.~(,~ [: i. 1 + >./)y
)
start_vals=: +:@+/\.&.|.@}:@,~&0
find_codes=:3 :0
b=. bl_count y
v=. start_vals b
n=. /:~ ~.y-.0
o=. ;({./.~ /:~ (n (([#2:) #: ])&.> (*b)#v+&.>i.&.>b
c /: o
)
Argument is the same as for huffma
Note also that you can use plain, uninverted # to expand arguments:
1 1j1 1 (#!.1) 3 3 $ 0
0 0 0
0 0 0
1 1 1
0 0 0
1 1j1 1 (#!.1"1) 3 3 $ 0
0 0 1 0
0 0 1 0
0 0 1 0
Which approach you pick often depends on whether it's easier for you to express
your expansion vector with length N (complex
oops, I made lots of typo.
.. without prior knowing
On Sep 11, 2014 3:45 PM, "bill lam" wrote:
> the frequencies (guessing from bit lengths) should be something like 2 3 1
> 1
> (2 3 1 1) hcodes 'ABCD'
>
> the hard part is the inverse problem: how to get the huffman code with
> prior kno
Here are some perhaps relevant examples:
1 1 0 1 #inv (3 3$ 0)
0 0 0
0 0 0
0 0 0
0 0 0
1 1 0 1 #inv!.1 (3 3$ 0)
0 0 0
0 0 0
1 1 1
0 0 0
1 1 0 1 #inv!.1"1 (3 3$ 0)
0 0 1 0
0 0 1 0
0 0 1 0
1 1 0 1 #inv!.1"1 (1 1 0 1) #inv!.1 (3 3$ 0)
0 0 1 0
0 0 1 0
1 1 1 1
0 0 1 0
1 1 0 1 # 1 1 0 1 #
Hi,
I know the verbs , ,. ,: to add rows, columns and dimensions to
matrices. It is easy to add these to begin or the end of the
corrosponding dimension, but what is to do, if I want to add one
row/column in the middle of a matrix? Is the only way to slice the
matrix in two pieces and join th
Thanks for your help and the great discussion.
-- Originalnachricht --
Von: "Roger Hui"
An: "Programming forum"
Gesendet: 08.09.2014 16:21:05
Betreff: Re: [Jprogramming] Replace one item of a list
a=: i.5
a
0 1 2 3 4
a=: 100 (2}) a
a
0 1 100 3 4
On Mon, Sep 8, 2014 at 7
Thanks, Dan and Thomas, for asking. I do not know the answers to Thomas'
questions. Everything about ordinal fractions is contained in this text:
https://www.dropbox.com/s/o23cz6aao6gpanx/pluk.of?dl=0
It is an ordinal fraction database. The file is actually a turbo pascal
program, because everyth
the frequencies (guessing from bit lengths) should be something like 2 3 1 1
(2 3 1 1) hcodes 'ABCD'
the hard part is the inverse problem: how to get the huffman code with
prior knowing the bits for each symbol. Your pointer to the putty
fork looks like helpful. The comment is in lines 861 to
75 matches
Mail list logo
