I didn’t express that very well, sorry for that.
Rank helps when you want data parallelism.
Apply to /this/ data, after chopping it into chunks.
Not to different data computed from this one.
In merge sort, the data changes in each loop iteration
so you can’t use that. Usually, you resort to using ^:
or the fold family F: etc
I hope this is both more accutare and easier to understand.
Am 05.10.21 um 19:06 schrieb Adrien Mathieu:
Yes, indeed, that was the main issue I met when trying to solve this
puzzle, ie. when you solve a problem "using rank", it usually means your
operation could be parallelized (because the J specs do not specify the
order of operations), so there should be no time relation between each
appliance.
However, I had the impression that sometimes it is possible to translate
a time problem into a space problem, to use your words, for instance
using the \ and \. adverbs (if I had to translate this to Caml, I would
put it this way: if you could either solve you problem with recursion,
or with a big fold) — which, to me, counts as a "rank solution".
If what I say is a bit vague, it's because I'm not sure exactly to grasp
completely the limits between what can be solved "using rank", and what
can be solved "using recursion/loop". I don't even know if the two are
completely equivalent (ie. if you can get a solution with the same
complexity using both for every problem).
Adrien Mathieu
On 05/10/2021 19:00, Hauke Rehr wrote:
I don’t quite understand.
Rank always applies to data (space). Recursion to program flow (time).
Am 05.10.21 um 18:58 schrieb Adrien Mathieu:
Well, technically this answers the question, but it doesn't answer
/my/ question, since this is the functional-language approach of a
loop. Maybe I have to be more specific:
the answer you are giving is, to me, what you would code if you were
asked to translate, say, the Caml version of mergesort. So the loop
is transformed into recursion.
What I wanted to know if there is a way in which loop is translated
into rank (if you get what I mean).
Adrien Mathieu
On 05/10/2021 18:53, Gilles Kirouac wrote:
Does this page help?
https://code.jsoftware.com/wiki/NYCJUG/2012-05-08
See Merge-Sort Examples.
~ Gilles
Le 2021-10-05 à 12:03, Adrien Mathieu a écrit :
Hello,
I am a beginner, and I would like to know if there is a way to
write mergesort without using loops, or an ugly translation of
loops using the ^: conjunction.
In particular, I have the impression that the merge part is hard to
achieve.
Thanks in advance,
Adrien Mathieu
----------------------------------------------------------------------
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
--
----------------------
mail written using NEO
neo-layout.org
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
