gg2a works but can be inefficient in the b=. line
depending on how y is defined. gg2b is better
if y is not a boxed array.
gg2b=: 4 : 0
b=. |: x I.&> <"1}:y
(b +//. {:y) (<"1 ~.b)} 0$~>:#&>x
)
----- Original Message -----
From: Roger Hui <[EMAIL PROTECTED]>
Date: Monday, September 22, 2008 20:10
Subject: Re: [Jgeneral] A Sorting Problem
To: General forum <[email protected]>
> gg2a=: 4 : 0
> b=. |:x I.&> }:y
> (b +//. {:y) (<"1 ~.b)} (>:#&>x)$0
> )
>
> x=: /:~&.> 5 7 13 [EMAIL PROTECTED]&.> 1000
> y=: 4 1000 [EMAIL PROTECTED] 1000
>
> x (gg2a -: Gg2) y
> 1
>
> See http://www.jsoftware.com/jwiki/Essays/Key
> for some uses of "key".
>
>
>
> ----- Original Message -----
> From: "J. Patrick Harrington" <[EMAIL PROTECTED]>
> Date: Monday, September 22, 2008 19:44
> Subject: Re: [Jgeneral] A Sorting Problem
> To: General forum <[email protected]>
>
> >
> > Thanks, Raul! I've replaced the old Gg with Gg2:
> >
> > Gg2=: 4 : 0
> > 'f1 f2 f3'=. x
> > 'a1 a2 a3 a'=. y
> > b=. (f1 I. a1),.(f2 I. a2),.(f3 I. a3)
> > sums=. b +//. a
> > idx=. <"1 ~. b
> > 'n1 n2 n3'=. >: >#&.> f1;f2;f3
> > sums idx} S=. (n1,n2,n3)$0
> > )
> >
> > and now the sort adds negligible time to the program.
> > I need to learn about "key".
> > Patrick
> >
> > On Mon, 22 Sep 2008, Raul Miller wrote:
> > > On Mon, Sep 22, 2008 at 3:14 PM, J. Patrick Harrington
> > > <[EMAIL PROTECTED]> wrote:
> > >> So the problem looks like this: a -- intensity array,
> > size = np (nphotons >
> > >> 1e5)
> > >> a0, a1, a2 -- tags for position & angle, all size np
> > >> f0, f1, f2 -- boundaries of the cells into which I want
> > >> to sort the the photons.
> > >> Thus, b0=. f0 I. a0 will be the index which shows into
> > which f0 interval
> > >> each photon in "a" falls. The number
> > >> of boundaries might be ~40, so there will be 40^3=64000
> > >> cells to which the exiting photons might be assigned.
> > >
> > > I would be inclined to use J's key adverb to perform your sums.
> > > If I understand your description correctly, I think I would use:
> > > sums=: (|:b) +//. a
> > > bins=: ~.&.|: b
> > >
> > > I would also think about replacing b with its transpose since
> > > using it as a key suggests that would be more natural. Perhaps
> > > using an expression like:
> > > b=. (f0 I. a0),.(f1 I. a1),.(f2 I. a2)
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
