Re: Support for 'where' primitive '⍸' ?

[Dr . Jürgen Sauermann]

Hi,

after investigating ⍸ I have added it to my TODO list.

Although its monadic form could be easily emulated with standard APL 
primitives,
the dyadic form seems to be useful (find the interval into which a value 
falls) and,
more importantly, not so easy to implement in APL.

Still wondering if monadic ⍸ B with non-boolean B has a real-life use case?

Best Regards,
Jürgen


On 11/27/20 9:34 PM, Dr. Jürgen Sauermann wrote:
> Hi Jay,
>
> thanks a lot for the pointers.
>
> Best Regards,
> Jürgen
>
>
> On 11/27/20 2:48 PM, Jay Foad wrote:
> > Dyalog's documentation for Where (monadic ⍸) is here:
> > http://help.dyalog.com/18.0/#Language/Primitive%20Functions/Where.htm
> > It was implemented long after Mastering Dyalog APL was written.
> >
> > Re: monadic ⍳ with vector arguments, I believe Dyalog got this from
> > the original NARS: see https://aplwiki.com/wiki/NARS: "Additionally,
> > Index Generator was extended to allow a vector argument".
> >
> > Jay.
> >
> > On Fri, 27 Nov 2020 at 13:42, Dr. Jürgen Sauermann
> > <mail@jürgen-sauermann.de> wrote:
> > > Hi Russ,
> > >
> > > welcome to APL. And welcome to bug-apl@gnu.org. The latter is a 
> > > moderated list, so please subscribe to
> > > it because otherwise each of your emails will require (manual) 
> > > approval by myself.
> > >
> > > Back to your question, ⍸ is definitely not a standard APL primitive. 
> > > My only source of information about Dyalog APL,
> > > the book with the title Mastering Dyalog APL - A Complete 
> > > Introduction to Dyalog APL, seems not to mention it
> > > either.
> > >
> > > I am somewhat reluctant when it comes to introducing non-standard 
> > > APL functions into GNU APL. In my
> > > opinion, which is certainly arguable, non-standard extensions of APL 
> > > may lure APL programmers into
> > > writing non-standard APL programs, which in turen undermines and 
> > > even contradicts the principles of
> > > free software.
> > >
> > > For a non-standard extension to make it into APL, the extension has 
> > > to smoothly integrate into APL. One
> > > of the rare examples where this was the case was monadic ⍳ with 
> > > vector arguments adopted from
> > > Dyalog APL. That extension was sort of filling a rather arbitrary 
> > > gap in standard APL, making APL
> > > more consistent than it was before.
> > >
> > > Some other extension in Dyalog did not make it into GNU APL because 
> > > IMHO the additional benefit that
> > > they provided was not worth the incompatibility that every 
> > > non-standard extension unavoidably creates.
> > >
> > > I cannot judge to which kind of extensions ⍸ belongs since I do not 
> > > know what it actually does. I will consider
> > > any opinions posted on bug-apl@gnu.org, though.
> > >
> > > Best Regards,
> > > Jürgen
> > >
> > >
> > >
> > > On 11/26/20 11:15 PM, Russtopia wrote:
> > >
> > > Hi, newbie to APL so my sincere apologies in advance if this is 
> > > incorrect, an FAQ, or by design (intentionally not implemented).
> > >
> > > Dyalog APL appears to support monadic ⍸ (underscore-iota) primitive 
> > > as a 'where' function, eg.
> > >
> > >      b ← 0 0 1 1 0
> > >      ⍸b
> > > 3 4
> > >
> > > ..whereas GNU APL considers the ⍸ as an invalid symbol (SYNTAX ERROR)
> > >
> > > Equivalent seems to be
> > >      b / ⍳⍴ b
> > > 3 4
> > >
> > > .. just wondering if this is a Dyalog non-standard and if GNU APL 
> > > plans to implement it.
> > >
> > > Thanks,
> > > -Russ