I certainly looked at the difference in the resulting tacit forms. I did not
look at the performance difference until you prompted me to
NB. Modest values for to get numbers in the believable range.
1000 (6!:2) 'f 1000'
0.00125151
1000 (6!:2) 'upperRightOnes 1000'
0.00124929
NB. 10^2 times as many inputs, comparisons, and results.
100 (6!:2) 'f 10000'
0.239497
100 (6!:2) 'upperRightOnes 10000'
0.237435
NB. Might as well measure the space, too.
(7!:2) 'f 1000'
1066624
(7!:2) 'upperRightOnes 1000'
1058816
but don't see a lot of difference. Is that the kind of timing test you were
thinking of? There's also the time taken to read the code, where not having to
think about hooks might balance out the time needed to recognize that identical
arguments are being supplied to both sides of the </ . I was hoping that the
hook (or the fork) would make it obvious to the reader and the implementor that
i. y only needed to be evaluated once.
I am delighted to discover that someone has taken the time to look at the trees
for these two forms and (apparently) discovered the common sub-expression so
that i. y is only evaluated once in each statement. I'm happy when I can
write code the way I think of it and find that someone is making it efficient
behind my back.
... peter
Linda Alvord wrote:
Did you compare the tacit verbs of these two functions?
f=: 13 :'(i.y)</i.y'
f
i. </ i.
5!:4 <'f'
-- i.
--+- / --- <
L- i.
^
upperRightOnes=: 13 : '(</ ]) i. y'
upperRightOnes
[: (</ ]) i.
5!:4 <'upperRightOnes'
-- [:
│ -- / --- <
--+----+- ]
L- i.
In situations like this I guess timing would be the referee.
Linda
-----Original Message-----
From: programming-boun...@forums.jsoftware.com
[mailto:programming-boun...@forums.jsoftware.com] On Behalf Of Peter B.
Kessler
Sent: Tuesday, November 06, 2012 2:26 PM
To: programm...@jsoftware.com
Subject: Re: [Jprogramming] Arc consistency in J
Linda Alvord wrote:
This sequence has a few problems:
A=. ? (2$n)$2 NB. generate random matrix of [0,1]
A=. A *. (i.n) </ i.n NB. make it upper diagonal, zeros on diagonal
] A=. A + |: 2*A NB. make it symmetrix referencing transpose in C.
All the ideas you used are correct. Maybe something like this is simpler:
f=: 13 :'(i.y)</i.y'
(f n)+|:2*f n=:4
0 1 1 1
2 0 1 1
2 2 0 1
2 2 2 0
Linda
I have no idea where this fits in the scheme of sudoku solvers, but the
repeated use of "i. y" in your definition of "f" seems awkward, and the
repeated use of "f n" in your last statement seems awkward. An alternative
to duplicating an expression is to evaluate it once and use it as an
argument to a monadic fork or hook, since those implicitly reference their
right argument for both the left and right verbs. So you could write
NB. The hook references i. y twice.
upperRightOnes=: 13 : '(</ ]) i. y'
upperRightOnes 4
0 1 1 1
0 0 1 1
0 0 0 1
0 0 0 0
NB. The outer fork references upperRightOnes y twice.
upperRightOnesLowerLeftTwos=: 13 : '(] + ([: |: (2&*)))
upperRightOnes y'
upperRightOnesLowerLeftTwos 4
0 1 1 1
2 0 1 1
2 2 0 1
2 2 2 0
I am a total novice at J, so you can ignore my observations on programming
style.
... peter
P.S. I can't believe that I'm actually advocating the use of a monadic hook,
since they are among the things I find most confusing, because of their
asymmetry. If you have trouble reading the hook, you can rewrite it as a
monadic fork
upperRightOnesFork=: 13 : '([ </ ]) (i. y)'
upperRightOnesFork 4
0 1 1 1
0 0 1 1
0 0 0 1
0 0 0 0
which seems sort of redundant with the definition of hook[1]. The use of
"[" instead of "]" for the left verb of the fork is due to Raul Miller[2],
and serves to remind us to think about what this verb would mean if used as
a dyad.
[1] http://www.jsoftware.com/help/dictionary/dictf.htm
[2] http://jsoftware.com/pipermail/programming/2012-August/028995.html
----------------------------------------------------------------------
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
