Here’s my attempt at a constructive approach. In haste, as Liz wants
to get away from this WiFi spot! So no time for explanation... and I
haven’t had an opportunity to digest Erland’s improvement to my
improvement to Lippa Esu’s verb. Sorry for brevity.
constructpart =: 3 : 0
:
d =. y
m =. x
r =. ,. col =. i. d
nr =. d#1 NB. size of nub of each row
done=. 1
SIZE=: 0
for_i. >: |. }: i.d do.
nadd=. i * #r
col =. nadd$i.i
r =. i#r
nr =. i#nr
NB. nr =. nr + col -.@e."0 1 r NB. less space but slower than...
nr =. nr + *./"1 col ~: r
if. m<: done =. >: done do. NB. check row nub-sizes ok
ok =. nr <: m
ok =. ok * (nr < m) +. (nr = m) * i >: <./"1 r
nr =. ok#nr
r =. (ok#col),. ok#r
col =. ok#col
else.
r =. col,. r
end.
SIZE =: SIZE >. #col NB. for comparison with sizes in other
approaches
end.
~.i.~"1 r#~ m = #@~."1 r
)
Its performance is somewhat worse in time, comparable in space with my
earlier efforts.
I don't know about comparison with Erland's modification.
No time for more - "can we make a move", says Liz!
Mike
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
