Hi,
I got spanked for asking this in a J forum. Totally understandable. I think
I have an answer; it's easier to do in modern APLs/J, but in the software
I'm using at work, it isn't. That's fine, it's good to know that other
people had the same question and fixed it, but that I'm using old software,
I don't have the resources.
Thanks,
WIll
On 6/13/07, Mike Day <[EMAIL PROTECTED]> wrote:
Dan
I think a single "item" is what he asked for;
your suggestion is preferable to mine as it's
more general, especially if he needs hyperplanes
other than the first. It also avoids transposes
which could be expensive in a real application.
A problem might arise - does APL*PLUS III allow
the rank-reduction in the final step, +/[axis] ,
for a non-nested character or general nested
array? I expect it does.
Dick Bowman observes that modern APLs offer
forms of selection other than eg [i;j;]. Dyalog
APL now allows
item {squid} [axis] A
APL*PLUS III may well have a similar mechanism.
Perhaps Will would comment on our various
contributions before the powers that be shut
down this non-J thread. In particular, does
he need the J-language solutions?
Mike
Dan Baronet wrote:
> Selecting an item on a specific axis can also be performed using
compress (/ in APL, # in
> J).
> dim <- (shape array)[Axis] NB. dim =. Axis{$array
> sel <- (-item-[]IO) rotate dim take 1 NB. sel =. (-item) |. dim{.1
> +/[Axis] sel/[Axis] array NB. sel #"((#dim)-Axis) array
>
> This only works for a single item tho.
>
> --- Mike Day <[EMAIL PROTECTED]> wrote:
>
>
>> Not J-programming but...
>>
>> Here's a simple "non-nested array" method, easily
>> worked up into an APL function, which might do what
>> you want. I think it generalises to a rank higher
>> (or lower) than 3, but you'd best check. I suggest
>> you look at the APL*PLUS III definition of transpose,
>> which should allow you to do more along the same
>> lines.
>>
>> (NB. I'm adding extra line-feeds to beat T/bird's
>> line-wrapping)
>>
>> Using Jim Weigang's transliteration of APL primitives.
>>
>> Adding J expressions as this is the J prog group!
>> J's transpose is a bit different from APL's....
>>
>> NB. using APL index origin 1
>>
>> A <- 2 3 4 {rho} {iota} ! 4
>>
>> NB. J: A =. i. 2 3 4
>>
>> r <- (1-n) {rot} {iota} {rho} {rho} A
>>
>> NB. J: r =. n |. i. $ $ A NB. J origin 0
>>
>> rtA <- r {transpose} A
>>
>> NB. J: rtA =. r |: A
>>
>> tAn <- (1{drop}{rho rtA) {rho} , rtA
>>
>> NB. J: tAn =. (, $~ ([EMAIL PROTECTED])) rtA
>>
>> NB. tAn is the required sub-array, but might need transposing...
>>
>> An <- ({gradeup} (-1) {drop} r) {transpose} tAn
>>
>> NB. J: An =. (/: }. r) |: tAn
>>
>> Mike
>>
>>
>> Will Slade wrote:
>>
>>> Hello,
>>>
>>> I'm crossposting this in the Usenet APL group.
>>>
>>> I'm learning APL*PLUS III for work, and I have a question for you
>>> experienced APLers.
>>>
>>> Lets say I have a matrix, A, with dimensions 2 3 4.
>>>
>>> A„2 3 4 ½ ¼!4
>>>
>>> I can return the first dimension of A like so: A[1;;]
>>>
>>> Is there a way to return the nth dimension of A without knowing what
the
>>> dimensions of A are? I'd like the flexibility to arbitrarily return
>>> the nth
>>> dimension, but I don't know how to do this in APL, and the error
messages
>>> are a little cryptic for someone with my inexperience.
>>>
>>> Thanks,
>>> Will
>>> ----------------------------------------------------------------------
>>> For information about J forums see http://www.jsoftware.com/forums.htm
>>>
>>>
>>>
>> ----------------------------------------------------------------------
>> For information about J forums see http://www.jsoftware.com/forums.htm
>>
>>
>
>
>
>
>
____________________________________________________________________________________
> Boardwalk for $500? In 2007? Ha! Play Monopoly Here and Now (it's
updated for today's economy) at Yahoo! Games.
> http://get.games.yahoo.com/proddesc?gamekey=monopolyherenow
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
>
>
>
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
