While you already have a good answer, let me experiment on this question a
bit more, just for practice.
Let's define the matrix, the function, and the column to be patched:
m =: _.".,;._2 ]0 : 0
0 6 3 2
9 1 7 3
5 2 9 5
4 0 9 0
)
f =: +:
c =: 2
Now we can use an amend solution like the ones given above
m c}~"1 0 f c {"1 m
0 6 6 2
9 1 14 3
5 2 18 5
4 0 18 0
We can also make this a function of m like this:
(c}~"1 0 [: f c&{"1) m
0 6 6 2
9 1 14 3
5 2 18 5
4 0 18 0
amendcol =: 2 : 'v.}~"1 0 [: u. v.&{"1'
(f amendcol 2) m
0 6 6 2
9 1 14 3
5 2 18 5
4 0 18 0
If f is not rank 1, this applies it to the whole column, not to individual
elements or the whole matrix at once:
(i.@:# amendcol 2) m
0 6 0 2
9 1 1 3
5 2 2 5
4 0 3 0
Now let's try an alternate solution, this time using agenda.
m [EMAIL PROTECTED]@.]"0"1 c = i. # m
0 6 6 2
9 1 14 3
5 2 18 5
4 0 18 0
This however applies f atomwise. If we want to apply it to the whole
column, we must do a transposition (or is there an easier way?).
|: (|:m) [EMAIL PROTECTED]@.]"1 0 c = i. # m
0 6 6 2
9 1 14 3
5 2 18 5
4 0 18 0
|: (|:m) [`(i.@:#)@[EMAIL PROTECTED]"1 0 c = i. # m
0 6 0 2
9 1 1 3
5 2 2 5
4 0 3 0
However, as agenda only chooses between using f or not, why don't we try to
use power '^:' instead?
|: (|:m) [EMAIL PROTECTED]:(0:`]`[)"1 0] c = i. # m
0 6 6 2
9 1 14 3
5 2 18 5
4 0 18 0
We need this stupid '@]' hack only because of the broken semantics of
'u^:(v1`v2)'. Imo, it should behave like this:
x u^:(v1`v2) y be equivalent to u^:(x v1 y) (x v2 y)
Alternately we could use a different method to circumvent this:
|: (|:m) f^:((1&{::)`(0&{::))@:;"1 0] c = i. # m
0 6 6 2
9 1 14 3
5 2 18 5
4 0 18 0
Then, we could just write 'f^:(]`[)' instead of '[EMAIL PROTECTED]:(0:`]`[)'.
ambrus
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
