You may very well be correct that the box representation is the right way
to go. It does provide a nice separation between array land and data with
arbitrary structure.
I'll have to take some time to digest the various approaches presented here.
Cheers,
Mike
On Mon, Sep 8, 2014 at 11:09 AM, 'P
Thanks for the links, they make for interesting reading.
I agree that the elegance of the solution is the hard part. Nevertheless
there have been previous attempts at ragged arrays that we can potentially
learn from: nested arrays in APL and K to some degree. I was hoping to
find out more about
One can imagine that it's possible for
] x =. (<@i."0) 1+i.3
+-+---+-+
|0|0 1|0 1 2|
+-+---+-+
to be laid out in memory as [0 0 1 0 1 2] with an additional vector
representing the starting point of each rank-1 array [0 1 3 6]. I would be
willing to wager a large sum that this is much
In my example, I show how the predecessor-index data structure can be used
to express both depth-first and breadth-first trees. I also give examples
of doing some useful operations on trees using this structure. I don't yet
see any advantage to using ordinal fractions since I've seen no examples
That's it! Thanks. Linda
-Original Message-
From: programming-boun...@forums.jsoftware.com
[mailto:programming-boun...@forums.jsoftware.com] On Behalf Of Raul Miller
Sent: Monday, September 08, 2014 9:03 PM
To: Programming forum
Subject: Re: [Jprogramming] Ragged Array Shapes are Trees
Given what I understand of your style, I think you are looking for
h=: 13 :'([: < [:$>)"0 y'
Thanks,
--
Raul
On Mon, Sep 8, 2014 at 7:51 PM, Linda Alvord wrote:
>
> How can you define h to use the definition of &, so that the result is the
> same as that for f and g :
>
> ]A=:((i.2
Point taken.
But neither you nor I (for example) have bothered to be interested
enough in Red Black trees to implement anything better. And for a
problem that size, it's really not right to put an hour into
criticizing it when an hour could have been used to show what you mean
using an actual impl
How can you define h to use the definition of &, so that the result is the
same as that for f and g :
]A=:((i.2 3);(,0)) ; (i. 2 3)
┌─┬─┐
│┌─┬─┐│0 1 2│
││0 1 2│0││3 4 5│
││3 4 5│ ││ │
│└─┴─┘│ │
└─┴─┘
f=: 13 :'$ each y'
f A
┌─┬───┐
│2│2 3│
Yes, but to Devon's point, having a direct transliteration of an algorithm
originally expressed in a (very) different language doesn't tell us anything
about J's power (or not) to express that algorithm.
Any programming language worth discussing is Turing complete. Which means
they're all *capa
It is, nevertheless, an implementation.
And anyone that feels like doing so can supply other alternates (or
replace it, if they feel theirs is good enough).
Thanks,
--
Raul
On Mon, Sep 8, 2014 at 5:28 PM, Devon McCormick wrote:
> This looks like a transliteration from another language - not
When I tested this on Windows, installing Pandoc put it on the path, so
that no change was needed to the shell call in makescript.
Can you experiment to find out why this is not working on your machine? You
should be able to call pandoc from the command prompt with the same
expression.
On Mon, Se
Chris,
After posting I looked at joxygen.ijs and saw that it used Pandoc. I
downloaded & installed it. The installer put it in: C:\Users\Tom
Arneson\AppData\Local\Pandoc
I modified the shell line in makescript to:
shell '"C:\Users\Tom Arneson\AppData\Local\Pandoc\pandoc" ',m,' -c
"joxygen.css"
DevonMcCormick's example
C: |__n0 | |_n00 | |_n01 | |__n1 |__n10 | |__n100 |__n11 | |__n110
| |__n111 | |__n112 |__n12
is expressed by ordinal fraction like this:
0 |__1 | |__11 | |__12 | |__2 |__21 | |__211 |__22
| |__221 | |__222 | |__223 |__23
This tree
This looks like a transliteration from another language - not J-like at
all. Once clue is that it's several times as long as many of the other
implementations.
On Mon, Sep 8, 2014 at 4:05 PM, Joe Bogner wrote:
> I found an implementation of red-black trees on rosettacode:
> http://rosettacode.
I think that is awesome, Thomas. For those who didn't load the code, its much
easier to use than it looks as everything is a verb, and the whole thing is
like a mini stack language, and each operation produces intuitive and visible
results.
- Original Message -
From: Thomas Costigli
It works for me. Perhaps you don't have Pandoc installed? The docs need to
make it clear that this is needed to build the output.
On Mon, Sep 8, 2014 at 12:23 PM, tom arneson wrote:
> I'm trying to use Joxygen and the example in the wiki doesn't work for me:
>
>
>
> The script ~addons/docs/joxyg
I found an implementation of red-black trees on rosettacode:
http://rosettacode.org/wiki/Pattern_matching#J ... It might be relevant or
may not be.
On Mon, Sep 8, 2014 at 3:45 PM, Devon McCormick wrote:
> A little research clarified what we see here: apparently it's part of the
> definition of
Array representations and "pointer" representations of trees both have
their uses. In the end it depends on what problem you are trying to solve.
As Devon demonstrated, for his purpose, array representation was not a
hindrance at all. But, as Dan mentions, trees are also used to implement
abstract
A little research clarified what we see here: apparently it's part of the
definition of a binary tree that the left node be smaller than its parent
but the right one is greater. Right away, I see a problem for the
predecessor-index representation of a tree that I'm advocating as it does
not distin
I'm trying to use Joxygen and the example in the wiki doesn't work for me:
The script ~addons/docs/joxygen/test.ijs builds docs from source in the
~addons/docs/joxygen/testscripts directory, with output to ~temp/joxygen. The
output should match the example output in ~addons/docs/joxygen/testd
in terms of copying one node to another,
((1 ; 0)&{:: <@[ each amendT 0 ]) 0 ;<(< i.5);i.4
┌───┬─┐
│┌─┐│┌───┬───┐│
││0 1 2 3 4│││┌─┐│0 1 2 3││
│└─┘│││0 1 2 3 4││ ││
│ ││└─┘│ ││
│ │└
At a minimum, I hope we can be a tad more elegant than this:
http://www-cs.ccny.cuny.edu/~peter/dstest/h_bl_tree.c .
On Mon, Sep 8, 2014 at 3:03 PM, Devon McCormick wrote:
> Do you have a reference to a good example of this? Looking at the
> "before" and "after" pictures on the right here -
>
Do you have a reference to a good example of this? Looking at the "before"
and "after" pictures on the right here -
http://en.wikipedia.org/wiki/Self-balancing_binary_search_tree - the
rebalancing seems arbitrary as it preserves some relations but changes
others.
On Mon, Sep 8, 2014 at 2:30 PM,
This is a very natural way of expressing trees in J.
And, this shows some of the pain of using trees - no one tends to like
any particular tree implementation. The grass is always greener on the
other side of the fence.
(And the "point" of trees often seems to be mediocrely efficient -
defining a
Raul wrote:
> Note that J already supports trees.
Devon wrote:
> I have J code that uses trees which I run daily and
> have been doing so for years.
Pascal wrote:
> I think trees are done at least ok, if not "right" already.
Challenge: express, in J, the logic of rebalancing a heap (say, a Fi
I think trees are done at least ok, if not "right" already.
I think Michal might be thinking of reinventing the box. I view trees as lists
of lists, where the right hand side of "of" might as well be anything including
subtrees.
So tree's are inherently polymorphic, and even if limited to only
This completes the performance TCO comparison (for dyadic ! aka binomial
coefficients); the following code includes undefined utilities but the
fixed code shown for the verbs has a proper linear representation
JVERSION
Installer: j602a_win.exe
Engine: j701/2012-12-06/12:20/x
Library: 6.02.023
I expand on this in a little more detail in the Wiki page I linked to, but
(in my view), there are two aspects to address w.r.t. trees in J;
efficiency is one of the two questions, but the less interesting one
("efficiency is a SMOP" [1]).
The other, more interesting question, is "convenience". Is
"The best is the enemy of the good"
I have J code that uses trees which I run daily and have been doing so for
years. The simple and efficient way I handle trees - using an idea from
APL days - is outlined here:
http://www.jsoftware.com/jwiki/DevonMcCormick/Trees .
The code I run daily - which p
Note that J already supports trees.
After all, that's what directories are, and J supports directories
(albeit using foreigns).
As for the efficiency of trees, well... you can see for yourself, from
first hand experience. Operating systems, after all, have many many
years of effort which have gon
I do think trees, if we did it right (that is, treated as first class
citizens in their own right, as opposed to a subtype or a different
perspective on "arrays"), would be a powerful extension to the language.
That would take some serious thought, though: for one thing, trees, by
their nature, en
Linda wrote:
"
The verb is amend and the main issue is that you must separate 100 and 2 or
they become the list 100 2
"
Actually, amend (}) is an adverb and the verb is (2}).
Roger wrote:
"
a=: 100 (2}) a
"
The form that Roger gave makes that apparent:
test=. 2}
type'test'
┌┐
│verb
That is not really an issue in the general case where the index and/or the
replacement value vary. As one would expect in the more general case.
a=.100
b=.2
a b} A=.i.5
0 1 100 3 4
On Mon, Sep 8, 2014 at 9:13 AM, Linda Alvord
wrote:
> The verb is amend and the main issue is that you mu
The verb is amend and the main issue is that you must separate 100 and 2 or
they become the list 100 2
100 (2)} A=.i.5
0 1 100 3 4
(100) 2} A=.i.5
0 1 100 3 4
Linda
-Original Message-
From: programming-boun...@forums.jsoftware.com
[mailto:programming-boun...@forum
What does this mean "shapes extended to support ragged arrays without boxing"?
K has ragged arrays without boxing, but K's concept of shape only
applies to the non-ragged prefix dimensions.
Meanwhile, it seems to me that the guarantee you suggest would be
computationally expensive.
(Not saying t
Thanks Pascal, that's a nice formulation for the 2-level deep case. It
breaks down of course for deeper nesting but could serve as a base.
([:|. $ ;~ $ each) ((i.2 3);(,0)) ; (i. 2 3)
+-+---+
|2|+-+---+|
| ||2|2 3||
| |+-+---+|
+-+---+
There's no specific problem I'm trying to solve
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 7:18 AM, Sebastian wrote:
> Hi,
>
> What is actually the best practice to replace one item of a list?
>
> The only way I know feels a bit complicated:
>
> a =: i. 5
> 0 1 2 3 4
>
>
>
> (2 {. a), 100, (3 }. a)
>
Hi,
What is actually the best practice to replace one item of a list?
The only way I know feels a bit complicated:
a =: i. 5
0 1 2 3 4
(2 {. a), 100, (3 }. a)
0 1 100 3 4
Best regards,
Sebastian
--
For information about
Thanks for your help!
Best regards
Sebastian
-- Originalnachricht --
Von: "Devon McCormick"
An: "J-programming forum"
Gesendet: 29.08.2014 22:43:44
Betreff: Re: [Jprogramming] Copy & Paste
Also, if you run jconsole, a C&P executes each line as if they were
entered
serially.
On Fri
Thanks for your tips!
Best regards
Sebastian
-- Originalnachricht --
Von: "Joe Bogner"
An: programm...@jsoftware.com
Gesendet: 29.08.2014 11:48:21
Betreff: Re: [Jprogramming] Copy & Paste
In addition to Linda's answer:
for quick evaluation of scripts on the clipboard, you can use F8
40 matches
Mail list logo
