Hi,
improvements on the rank operator are already on my TO list (see
README-4-compliance).
I guess ⍤[] probably allows rank along any axis, but I have not yet
investigated that.
/// Jürgen
On 03/31/2014 09:34 AM, Elias Mårtenson wrote:
Jay, thanks for reminding me of the rank operator. I had forgotten
about that one. :-)
Yes, in this case it would definitely do the right thing, but as you
mention it won't allow you to work on other axes.
Regards,
Elias
On 31 March 2014 15:29, Jay Foad <[email protected]
<mailto:[email protected]>> wrote:
On 31 March 2014 04:51, Elias Mårtenson <[email protected]
<mailto:[email protected]>> wrote:
> I keep wanting to be able to apply an axis argument to the ¨ (EACH)
> operator. I.e, to sort an array rows, I wanted to do:
>
> {⍵[⍋⍵]}¨[2] foo
>
>
> Instead, I had to do:
>
> ⊃ {⍵[⍋⍵]}¨ ⊂[2] foo
>
>
> Would it make sense to be able to specify an axis number to the
the EACH
> operator as an extension?
Hi Elias,
Have you considered using the Rank operator?
{⍵[⍋⍵]}⍤1(4 3⍴'FEEFIEFOEFUM')
EEF
EFI
EFO
FMU
There is a bit of tension in APL between these two different ways of
solving the same kinds of problem: primitives-with-axis, and the Rank
operator.
- Primitives-with-axis are a bit ad hoc. Every primitive interprets
the axis specification in a different way, and some primitives (like
Each) don't support it at all.
- By contrast, the Rank operator is completely uniform and applies to
all functions.
- But the Rank operator restricts you to dividing the argument up
along the leading axes. (Some would argue that this restriction is a
good thing.) E.g. you can use it to sort each row of a matrix, but not
to sort each column.
Jay.
