On 2/27/08, Roelof K Brouwer <[EMAIL PROTECTED]> wrote:
> repl_diag_by_minus_sum_of_row=: 3 : 't-I*+/"1 t=.(-.I=.=/~i.#y)*y'
> repl_diag_by_minus_sum_of_row=: 13 : 't-I*+/"1 t=.(-.I=.=/~i.#y)*y'
>
> applying them to i.5 5 gives two different results with the 2nd giving an
> unintended result
First off, I prefer to avoid names unless I can come up with
good names which are shorter than their expressions.
Second, I examined the expressions
3 : 't-I*+/"1 t=.(-.I=.=/~i.#y)*y' i. 5 5
13 : 't-I*+/"1 t=.(-.I=.=/~i.#y)*y' i. 5 5
and I think that the 13 : expression evaluator has a bug
which can be worked around by changing that second
expression to
13 : 't-I*+/"[EMAIL PROTECTED] t=.(-.I=.=/~i.#y)*y' i. 5 5
Here's my thinking. I tried:
3 : 'I*+/"1 t=.(-.I=.=/~i.#y)*y' i. 5 5
13 : 'I*+/"1 t=.(-.I=.=/~i.#y)*y' i. 5 5
these results were different
But when I tried
3 : '+/"1 t=.(-.I=.=/~i.#y)*y' i. 5 5
13 : '+/"1 t=.(-.I=.=/~i.#y)*y' i. 5 5
this pair of expressions gave me results that looked the same.
So I then looked at
3 : 'I[+/"1 t=.(-.I=.=/~i.#y)*y' i. 5 5
13 : 'I[+/"1 t=.(-.I=.=/~i.#y)*y' i. 5 5
and, again, my results looked similar.
So then I looked at
3 : 'I;+/"1 t=.(-.I=.=/~i.#y)*y' i. 5 5
13 : 'I;+/"1 t=.(-.I=.=/~i.#y)*y' i. 5 5
and my results looked different.
So then I looked at the linear form for
13 : 'I;+/"1 t=.(-.I=.=/~i.#y)*y'
which is
([: =/~ [: i. #) ([ ; +/"1) ([: -. [: =/~ [: i. #) * ]
And, here, I can see that the verb ([ ; +/"1)
treats +/"1 as a dyad but the original expression
used it as a monad.
I can force it to be treated as a monad by replacing
+/"1 with +/"[EMAIL PROTECTED]
FYI,
--
Raul
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
