On Sun, 23 Jan 2022, Elijah Stone wrote:
%"_1 1 can be written as %&.(0&|:`a:)
oops, %"_1 _
t. t: t.. t::
Maybe redundant but still worthwhile expansion: u t. 1 2 3 should use axes 1, 2, and 3 for the monadic, dyadic left, and dyadic right cases, respectively, as rank. u t. (1 2;a:;3) should move axes 1 _and_ 2 to the front in the monadic case, do nothing for dyadic left, and move axis 3 to the front for dyadic right. u t. (1;a:) t.. (2;1) should move axis 1 of the left argument to the front, axis 2 to the back, and axis 1 of the right argument to the back.
Possibly, u t. (<1 0j2) should move axis 1 to the front and axis 2 to the back. I am not sure if this is actually a good idea; but it permits more direct expression of certain things that would not be so straightforward otherwise, but those things are of questionable utility. If that is added, |: should be extended similarly.
harder to teach
Note it shares the property with " that f"k"l is not the same as f"l, which is pointed out by e.g. 'Gordian Knot' as a point of confusion. The situation may be slightly worse, in that f"k"l may give huge results which are obviously wrong, whereas f t. k t. l may give a wrong result with about the right size.
Regardless, I don't think teachability is a strong reason to avoid including features. However I do understand that others disagree (e.g. one reason for wanting to exclude tacit conjunctions is that they are obscure.)
---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
