> But how to get rid of nested if-s?
>
> If cond1 do
> Something1
> Else
> If cond2 do
> Something2
> Else
> Something3
> End
> End
ni =: verb define
if. 2 | y do.
<:y
else.
if. y>100 do.
-: y
else.
y
end.
end.
)
cond1 =: 2&|
Something1 =: <:
cond2 =: >&100
Something2 =: -:
Something3 =: ]
tni =: [EMAIL PROTECTED](cond1 (* >:)&:-. cond2)
ni 3
2
ni 110
55
ni 48
48
tni 3
2
tni 110
55
tni 48
48
> Is there any rule of thumb for this case or is the approach
> everytime depending on the specifics of the case?
Sure: change how you think ;) Scalar languages, like C, focus on details;
array languages, like J, focus on the "big picture".
So in C you make many small decisions, which manifest as nested conditions.
You don't do that so much in J.
To answer your question a different way (which will be more useful immediately
but may be detrimental in the long run): If I were
converting a C program to J, I might re-write all the nested conditionals as
select statements. Then I would phrase those
statements in J using agenda @. as I did above.
A couple of caveats:
* tacit code is driven entirely by its arguments. So, if your
predicates rely on (mutable) global data, you cannot write
a tacit equivalent.
* Sometimes conditionals are nested because the inner condition should not
be executed unless the outer condition is true (or
false). For example:
if. noun=nc'variable' do.
if. variable>100do.
varible+17
end.
end.
> -----Original Message-----
> From: [EMAIL PROTECTED]
> [mailto:[EMAIL PROTECTED] On Behalf Of Kairit Sirts
> Sent: Saturday, January 19, 2008 10:12 AM
> To: 'Programming forum'
> Subject: RE: [Jprogramming] loopless code
>
> That was very helpful, thanks!
>
> But how to get rid of nested if-s?
>
> If cond1 do
> Something1
> Else
> If cond2 do
> Something2
> Else
> Something3
> End
> End
>
> Is there any rule of thumb for this case or is the approach
> everytime depending on the specifics of the case?
>
> Kairit Sirts
>
> > -----Original Message-----
> > From: [EMAIL PROTECTED] [mailto:programming-
> > [EMAIL PROTECTED] On Behalf Of Devon McCormick
> > Sent: Friday, January 18, 2008 10:16 PM
> > To: Programming forum
> > Subject: Re: [Jprogramming] loopless code
> >
> > Kairit -
> >
> > as a rule, in J, you should be suspicious of any loop and ask the
> > question:
> > am I doing something here that is intrinsically serial?
> >
> > In the example above, it seems evident to me the answer is
> no. Then
> > again, I've been doing this sort of thing for longer than
> the lifetime
> > of the average beagle, so YMMV.
> >
> > To summarize what the first loop accomplishes: keep subtracting
> > successive elements of the row (from a random number) until
> we get to or below zero.
> > So, to turn this around and state more simply: where is the
> > cumulative sum of the row greater than some arbitrary
> value? That is,
> > index=. (roll>+/\row) i. 0
> > gives the same index calculated by the loop.
> >
> > The latter part of this code, we want to make the index one if it's
> > zero but leave it alone otherwise, or
> > index=. 1>.index NB. Whichever is greater, one or the existing
> > value (since we know index >:0)
> >
> > So, the two loops become the simple expression
> > 1>.0 i.~roll>+/\row
> > and "roll" is calculated from "row" but used only once so
> there's no
> > point assigning it a name, so this becomes
> > 1>.0i.~(?+/row)>+/\row
> >
> > Presumably you want to do this to all the rows. One way to
> do this is
> > to fashion the statement above into a function
> > rowWork=: 13 : '1>.0 i.~(?+/y)>+/\y'
> > then apply this to each row of y:
> > rowWork"1 y
> >
> > Hope you find this helpful.
> >
> > On 1/18/08, Kairit Sirts <[EMAIL PROTECTED]> wrote:
> > >
> > > >
> > > > In fact, I think it would be a fruitful discussion to
> compare and
> > > contrast
> > > > the cases where we use:
> > > >
> > > > * "
> > > > * /
> > > > * ^:iteration_calculator
> > > > * ^:condition^:_
> > > > * $:
> > > > * for. and while.
> > > >
> > > > I say that because I struggled with the third sentence of this
> > > message. I
> > > > couldn't quite articulate the type of problem which would cause
> > > > one to
> > > use
> > > > $: instead of ^: .
> > > >
> > > > -Dan
> > >
> > > I myself am far too incompetent to start this fruitful
> discussion,
> > > but actually I do have several peaces of code which are
> implemented
> > > with for or while loops. I don't like them and I'm quite
> sure there
> > > are nicer ways
> > to
> > > write them, but it has been to complicated to me so far.
> > >
> > > For example the following peace:
> > >
> > > y is a matrix of shape n * n
> > > x is a number between 0..n-1
> > >
> > > roll =. ?+/ row =. x { y
> > > index =. 0
> > > while. roll > 0 do.
> > > roll. =. roll. - index { row
> > > index =: >: index
> > > end.
> > > if. index = 0 do.
> > > index =. 1
> > > end.
> > >
> > > For the last if-clause I figured out something like this:
> > >
> > > check =: =&0^:(=&0)
> > >
> > > And 'check index' seems to work as expected, but I'm not
> quite sure
> > > if
> > it
> > > really is working correctly
> > >
> > > Kairit Sirts
> > >
> > >
> > >
> --------------------------------------------------------------------
> > > -- For information about J forums see
> > > http://www.jsoftware.com/forums.htm
> > >
> >
> >
> >
> > --
> > Devon McCormick, CFA
> > ^me^ at acm.
> > org is my
> > preferred e-mail
> >
> ----------------------------------------------------------------------
> > 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
