The APL Orchard

B. Wilson

12:48 AM

@Jeremygee "sometimes hard to follow threads on here" Seriously! I actually have no idea what comment this is replying to :P Glad to be of help.

B. Wilson

6:16 AM

Ah. No multiple indexed assignment?

Not really sure it could be defined in a natural way

Adám

7:04 AM

@B.Wilson Uh:

      a←⎕A
      b←⎕D
      (a[3])(b[2])←'x'
      a
ABxDEFGHIJKLMNOPQRSTUVWXYZ
      b
0x23456789

essie lovett

The syntax for the LHS is finicky, it would've taken me a few guesses

B. Wilson

@Adám Parens! I see.

Adám

I believe this isn't supported…

essie lovett

It would be nice to have the grammar documented

Adám

@B.Wilson Well, note that a[3] b[2] doesn't mean what you think; it means ⊂b — yet another reason to stop using brackets!

B. Wilson

7:13 AM

@Adám Because binding precedence makes it equivalent to (a[3] b)[2], right?

Adám

Yes.

If that wasn't the case, then 10 20 30[2] would be a rank error instead of giving 20, and it'd be incompatible with old (pre-nested arrays) APL.

Indeed, IBM broke this compatibility, making APL2 bind indexing stronger than stranding, so a[3] b[2] works more sensibly, but you need (10 20 30)[2]

B. Wilson

With the implicit each on modified assignment, for some reason I wa1s thinking that (a b)[1]← might set the 1+~⎕IOth element of both a and b.

Adám

That'd be an each on the bracket indexing itself, which surely can't work.

B. Wilson

The interpreter is Magic and designed by Harry Potter. It can do anything.

Adám

      a←⎕A
      b←⎕D
      (a b)⊣@1⍨←'x'
      a
xBCDEFGHIJKLMNOPQRSTUVWXYZ
      b
x123456789

↑ is only one character longer.

You can even write the shorter a b⊣@1⍨←'x' but this isn't recommended.

essie lovett

7:20 AM

Wow that's ugly :)

B. Wilson

Really? I think it's pretty direct.

Adám

If we'd had Structural Under, maybe we'd never added those strange indexed and selective assignments. (a b)←1⍢⊃¨a b

B. Wilson

I hate you, by the way. Always effortlessly producing nice little snippets like this :D

B. Wilson

7:39 AM

Nice. Prepending an empty vector lets us control the prototype of a potentially empty vector.

Adám

Relying on that seems dangerous. Better prepend a cell of the type you want, and then dropping it.

B. Wilson

Ah, okay. So that's an Implementation Detail™?

Adám

I'd say so, because 1. it varies between implementations, and 2. it varies between versions, and 3. it varies between array ranks (at least in Dyalog).

B. Wilson

Yikes. Thanks for warning!

Adám

@B.Wilson This paper may interest you.

B. Wilson

7:53 AM

Indeed, it does! Will read after work.

Silas Poulson

9:36 AM

@Adám do note that currently, that if you want to index with a variable as I initially read Adam's snippet, you need brakets so a b(⊣@y⍨)←'x'

Assignment is messy - in the nonbracketed case, the interpreter thinks it should be assigning to y so @ sees the symbol y not it's value

Adám

12:59 PM

Welcome to APL Quest 2018-5! Today's quest is What’s Your Sign? Revisited:

> Write an APL expression that, given a 2-element integer vector representing month and day, returns a character vector (string) of the corresponding Western zodiac sign.

Aries 	March 21–April 19 	Libra 	September 23–October 22
Taurus 	April 20–May 20 	Scorpio 	October 23–November 21
Gemini 	May 21–June 20 	Sagittarius 	November 22–December 21
Cancer 	June 21–July 22 	Capricorn 	December 22–January 19
Leo 	July 23–August 22 	Aquarius 	January 20–February 18
Virgo 	August 23–September 22 	Pisces 	February 19–March 20

Richard

⊃signs⌷⍨+/(⊂100 ⊥⍵)≥(100⊥⊢)¨dates

And the full version

Adám

btw, while the above table looks messy, you can copy and paste it into a character vector cv, and then do t←⎕CSV⍠'Separator'(⎕UCS 9)⊢cv 'S'

Richard

I should also have paid attention to the imput formatting perhaps

Adám

@Richard You can do something much more elegant.

Have a look at dyadic ⍸

Richard

ah, never really studied this dyadic way of the functon

reading it right now

ah nice :)

So it tells in which interval the values are

Adám

1:06 PM

Indeed.

Richard

But 100⊥ or something like that is still necesary. Or ⎕DT?

Adám

No, it handles TAO.

Richard

TAO?

Adám

apl.wiki/TAO

Basically, lexicographical ordering, so 1 20 precedes 2 1.

Richard

yes thanks, thats realy nice

Where does TAO apply? Not for (2 10)>(1 5), whcih will give (1 1) I hope

Adám

1:11 PM

Only ⍒⍋⍸

Oh, btw, we'll start the APL Quest 2 hours later beginning March 17. Note that this is between the dates when USA and EU change from winter to summer time, so watch out! We'll begin at 15:00 UTC

Richard

so in this case we dan do (2 3)⍸dates and use that as the index for the signs, or rotating them as you did previous time

Adám

Wrong order of arguments to ⍸ but yes.

Richard

ah yes, some other functions I also always switch the arguments...

Adám

With all this knowledge, and putting dates and signs into 1-letter variables, I have a 7-character solution.

Richard

:)

s⌷⍨d⍸⍵

Adám

1:18 PM

⊃⍨ but also, that fails when the date falls before the first cut-off.

The easiest is to rotate.

Richard

{⊃s⌽⍨d⍸⍵}

can't test it right now but something like this I think

Adám

Yes, that's perfect. I had:

(s d)←↓¯3 ¯2⌽2 12⍴⍉t[;1 3 2 4]
s←¯1↓¨s
d←(⍳⍤≢,⍤0⊢){2 2⊃' –'⎕vfi⍵}¨d
F←⊃d∘⍸⌽s⍨

Richard

t is test data?

Adám

25 mins ago, by Adám

btw, while the above table looks messy, you can copy and paste it into a character vector cv, and then do t←⎕CSV⍠'Separator'(⎕UCS 9)⊢cv 'S'

What is the most compact full definition? I have:

⊃((⍳⍤≢,⍤0⊢)19+1 0 2 1,2 4 2/2 4 3)∘⍸⌽'CapricornAquariusPiscesAriesTaurusGeminiCancerLeoVirgoLibraScorpioSagittarius'(∊⊂⊣)⎕A⍨

Richard

ok. Also intrigued by(⍳⍤≢,⍤0⊢)

Adám

1:31 PM

Oh, right, that's stupid of me.

(⍳⍤≢,⍪)

But either way, d{⍵,2 2⊃' –'⎕VFI⊃⍺}⍤0←⍳12 is more elegant. An example of putting significant amounts of code into the modifying function in modified assignment.

⊃((⍳⍤≢,⍪)19+1 0 2 1,2 4 2/2 4 3)∘⍸⌽'CapricornAquariusPiscesAriesTaurusGeminiCancerLeoVirgoLibraScorpioSagittarius'(∊⊂⊣)⎕A⍨

then.

And with 20.0, we could use write ⍳⍤≢⍛,⍪19+…

Richard

0←⍳12 I can not follow.

Adám

You know about A+←1, right?

Richard

asigning a function to A?

no

thats +∘1

Adám

That's incrementing A. Almost the same as A←A+1

Richard

ah , ok

Adám

1:37 PM

It exists in many programming languages. We call it modified assignment, but the world calls it

Augmented assignment

Augmented assignment (or compound assignment) is the name given to certain assignment operators in certain programming languages (especially those derived from C). An augmented assignment is generally used to replace a statement where an operator takes a variable as one of its arguments and then assigns the result back to the same variable. A simple example is x += 1 which is expanded to x = x + 1. Similar constructions are often available for various binary operators. In general, in languages offering this feature, most operators that can take a variable as one of their arguments and return a...

Dyalog allows any function, not just arithmetic primitives.

Here, the function used was {⍵,2 2⊃' –'⎕VFI⊃⍺}⍤0 so the expression is almost equivalent to d←d{⍵,2 2⊃' –'⎕VFI⊃⍺}⍤0⍳12

Richard

and ⍤ is rank in this case

Adám

Yes. Makes sense now?

@B.Wilson Did you want to post something, or shall we call it a day?

Richard

No not yet. A+←1 I get, but the 0 in 0←⍳12 I is nor a variable nor a function

Adám

No, the variable is d. 0 is just the right operand of ⍤ which is part of the modifying function.

B. Wilson

@Adám Sorry. Having clipboard problems.

Richard

1:49 PM

ok. I will play with it later on than. thanks for the food for thought :)

B. Wilson

Late as always. Here's my attempt:

 f←{a w←⍵
  n←'|Capricorn|Aquarius|Pisces|Aries|Taurus|Gemini'
  n,←'|Cancer|Leo|Virgo|Libra|Scorpio|Sagittarius'
  t←(12 2⍴1⌽2⊢⍤/1↓¨⊢⊂⍨'|'=⊢)n
  s←20 19 21 20 21 21 23 23 23 23 22 22
  ⊃t⌷⍨⎕←a(1+w≥a⌷s)}

Adám

@B.Wilson Do you want a mod to nuke the message to protect your PPI?

B. Wilson

@Adám Please!

Adám

@B.Wilson 1↓¨⊢⊂⍨'|'=⊢ is ⊢⊆⍨'|'≠⊢

@hyper-neutrino Can you nuke this to protect PPI?

B. Wilson

@Adám Ah! Partition, not Partitioned Enclose. Beautiful.

Not sure why the ⎕← snuck in there :P

Adám

1:53 PM

Debugging?

@B.Wilson No need for ⊢⍤ when you have a constant as left argument.

Actually pretty clever to choose day-of-month cut-off from the outset :-)

B. Wilson

Ah, yes. ⎕IO issues when pasting into test page.

Adám

But with that, I think we'll say, see you next week for 2018-6: What’s Your Angle?

B. Wilson

@Adám Lol. You can tell I just rabidly hacked this out :P Thanks.

Adám

Note that we'll begin at 15:00 UTC starting March 17!

hyper-neutrino

@Adám done

LdBeth

9:35 PM

does the dyalog IME for windows able to work without dyalog apl installation?

LdBeth

10:28 PM

nevermind, at least github.com/abrudz/Kbd works for me

Transcript for