Has anyone written any APL benchmarks for the computer language shootout?

Would be interesting to see how Dyalog stacks up

@Adám - Is there a way to download the bookmarked conversations so that I can convert them into an eBook for my (off-line) tablet (for study on the train)?

@JeffZeitlin Wouldn't printing them to PDF work for that?

@JeffZeitlin Yes, for now, I'm having one every 2-3 weeks.

It should be fairly easy to scrape the html content.

I'm testing a couple of things for peeling the data, including simple cut-paste into a (Unicode-supporting) text editor.

Hrm. Print to PDF/Save as PDF (in Chrome) gives me the decoration on the right (lesson name, date, links, etc) on every page. Cut-Paste gives me doubled participant names and dates.

OK, 'nuff tearing out what little hair I have remaining over font and formatting issues...

@Adám - With the understanding that there is going to be an obvious preference for Dyalog, what do you think of NARS2000 as an APL environment?

@JeffZeitlin NARS2000 is mainly an experimental interpreter to play with language extensions. It regularly changes existing stuff without warning. And it has pretty much no facilities for communicating with the world outside. Neither does it care (much) about performance.

"Communication with the world outside" meaning interacting with the file system more than at the most simplistic level, and interacting with other than the file system?

I have a rank operator question. Anyone got a second?

a←2 7⍴⍳2 ⋄ b←2 1⍴5+⍳2

Can I use the rank operator to a÷b?

let me see

yes

a ÷⍤1 ⊢b

That was my thought @FrownyFrog but doesn't seem to work for me

@FrownyFrog 👍

)help

⎕←1 2 3

I get a length error.

how do you use the bot again?

I thought as you were. Looks to be down?

⍞←'ABC'

@JamesHeslip ABC

⎕←123

@JamesHeslip
123

ok cool

a←2 7⍴⍳2 ⋄ b←2 1⍴5+⍳2 ⋄ ⍞←a÷⍤1⊢b

Where do I find an APL IME to use with the bot?

tio.run/##SyzI0U2pTMzJT///P/…

⋄ a←2 7⍴⍳2 ⋄ b←2 1⍴5+⍳2 ⋄ ⍞←a÷⍤1⊢b

@RichardPark start with ⋄ if you don't want to start with ⎕←/⍞←

ah thank you

⋄ a←2 7⍴⍳2 ⋄ b←2 1⍴5+⍳2 ⋄ ⍞←a÷⍤1⊢b

ಠ_ಠ bot

I'm confused lol

¯(^//^)/¯

⋄ a←2 7⍴⍳2 ⋄ b←2 1⍴5+⍳2 ⋄ ⎕←a÷⍤1⊢b ⊣ '?'

@dzaima
0.1666666667 0.3333333333 0.1666666667 0.3333333333 0.1666666667 0.3333333333 0.1666666667
0.2857142857 0.1428571429 0.2857142857 0.1428571429 0.2857142857 0.1428571429 0.2857142857

ah, so no ⍞.

cool

anyway @JamesHeslip it works

I guess I'm just different... My APL (D16) is giving a length error

that's worth reporting

my version 15 also reports length error

brb

Thanks

Doesn't seem to work in Version 17.0.35960

but does in 17.1.36847

try ¯1 as well

it is a known issue apparently about rank implementation and the way scalar dyadics use singleton extension (1 element vector and scalar equivalence)

@FrownyFrog Rank ¯1?

yes

and (as you have seen) it's been fixed in 17.1

@FrownyFrog I'm not sure what ¯1 does. I'm not too comfortable with rank.

Still gives a length error in 16 with ¯1 though.

I guess it can't give a differetn result, it's just equivalent to 1

if it's important for your stuff for it to work in older versions then send a report to support@dyalog (in Windows: Help → Email Dyalog)

¯1 applies to rank 1-less than the rank of right argument

or whichever argument sorry

⋄ a←2 7⍴⍳2 ⋄ b←2⍴5+⍳2 ⋄ ⎕←a÷⍤1 0⊢b ⊣ 'doesn't require singleton scalars'

@dzaima
0.1666666667 0.3333333333 0.1666666667 0.3333333333 0.1666666667 0.3333333333 0.1666666667
0.2857142857 0.1428571429 0.2857142857 0.1428571429 0.2857142857 0.1428571429 0.2857142857

so in this case yes because a and b are both rank 2

@RichardPark Not super important. More the case that I tried to get something working using rank in an application for a client who is still running 16 and it just failed. I got confused then...

@JamesHeslip yep sorry about that

We should be moving them to 17.1 soonish.

@RichardPark appreciate you looking into this

it's annoying because you can do things with rank in a more consistent manner than square bracket indexing (which you can't use with user defined functions, for example)

square bracket axis-operator*

yeah move to 17.1 is best solution

@JamesHeslip de nada

@dzaima Interesting. This works, though given what I've just seen I might make more of a statement for the team to not use rank until our move to 17.1.

yep, what a shame

@JamesHeslip personally i'd just not rely on scalar extension in any scenario

@dzaima I think the issue is my understanding of rank.

@JamesHeslip eh, i barely understand ⍤ too :p

Adám did a session on it yesterday. I plan to read it over this weekend. Been so busy this week...

@JamesHeslip No you were right with your statement but there was a bug related to a fundamental annoyance in the Dyalog APL array model - I've seen a couple of things where J has made it work with a refreshed and more consistent model, but APL has "historical reasons" so we gotta live with it

You can bypass that bug by replacing ÷ by an equivalent dfn. iirc