The APL Orchard

Bubbler

12:03 AM

Math. Wins. Again.

Marshall

1:23 AM

Only loses to math+BQN.

jimfan

2:50 AM

@TessellatingHeckler So UTF-8 characters does give lower score because of multi-byte nature? I used to think only number of character matters...

TessellatingHeckler

@jimfan UTF-8 gives worse (higher number) score because many characters need multiple bytes each. Dyalog APL does not use ASCII character set, it uses a custom one, so it can represent the normal ⌊⌈⍣⍟⊂⊃∩⊇⊆... APL characters all in one byte each.

If you're just writing APL with numbers and text and standard things then you can count number of characters == number of bytes. If you go outside that, like those ≮ in the question, you need to switch to some other encoding like UTF-8 and count more bytes than characters

(does that make sense? I'm never sure with encoding things)

(otherwise, you could make a language with one symbol that solved the entire problem and count it as score 1, even though it took 40 bytes to encode the symbol - and that's the code)

Dyalog APL

<moon-child> nominally, there's apl classic, which uses one byte per symbol; and apl unicode, which uses utf8, so multiple bytes per symbol. However, there's github.com/abrudz/SBCS which makes an apl-specific encoding using one byte per symbol, which lets you count it that way for codegolf. However, this only works because there are fewer than 256 apl symbols

jimfan

3:11 AM

@TessellatingHeckler Okay. So far I only list number of characters (instead of byte) alongside with my answer. After all I feel char/byte count doesn't matter too much because there is always answer shorter than mine.

My personal view is byte count has nothing to do with preciseness of the language/answer itself. Penalising APL or other languages for its charset defeats the purpose somehow.

For example the C expression "4<=4" and APL expression '4≤4' essentially express the same meaning. APL loses because '≤' is composed of 3 bytes (0xE2 0x89 0xA40). Preciseness of language comes with cost. The cost however does not undermine preciseness and the beauty of it. And beauty prevails.

Dyalog APL

<moon-child> how about ~4>4 - tied :)

TessellatingHeckler

how can I detect from the first of one symbol like ( to the next )

all I can see online is stuff like if they're correctly balanced and nested, but that's not relevant here, it should work for "blah ) blah ( blah ( blah )" and just take the () pair

I have a long and convoluted boolean masking

⋄ {⍵/⍨~s∧⌽∨\⌽<(']'=⍵)∧s←∨\<\'['=⍵}⍣≡⊢'blah ] blah [ blah [ blah ]'

Dyalog APL

@TessellatingHeckler SYNTAX ERROR

TessellatingHeckler

no there isn't

I can start with {'[]'∘.=⍵} which looks like a decent start

Dyalog APL

<moon-child> @TessellatingHeckler yes there is. 'The function requires a left argument' on the first <

TessellatingHeckler

3:27 AM

@jimfan I think it's no fun when someone posts a golfing language answer and they say "in my language % means 'read a number from stdin and generate the nth Mersenne Prime so it's 1 char".

what

⋄ {⍵/⍨~s∧⌽∨\⌽<(']'=⍵)∧s←∨\<\'['=⍵}⍣≡⊢'blah ] blah [ blah [ blah ]'

Dyalog APL

@TessellatingHeckler SYNTAX ERROR

TessellatingHeckler

I ran that, it worked, then copy pasted it

but copy it back from here to Dyalog, it's broken

something is treating \ as an escape

that should be <\ scan

⋄{⍵/⍨~s∧⌽∨\⌽<\(']'=⍵)∧s←∨\<\'['=⍵}⍣≡⊢'blah ] blah [ blah [ blah ]'

Bubbler

@TessellatingHeckler Adding a 1-byte built-in for a challenge is a loophole.

The bot is somehow not catching up

Btw, there's ⍀ if you're worried about SE chat's weirdness.

⋄ +/2 3 4

Dyalog APL

@Bubbler 9

Bubbler

⋄ +/2 3 4

Dyalog APL

3:33 AM

@Bubbler 9

Bubbler

⋄ +/2 3 4

So it doesn't work with multiline code block

TessellatingHeckler

⋄{⍵/⍨~s∧⌽∨\⌽<\(']'=⍵)∧s←∨\<\'['=⍵}⍣≡⊢'blah ] blah [ blah [ blah ]'

I'm not posting a multiline code blockkkkkk

that was one line, no newline, four indent spaces

Dyalog APL

<moon-child> I'm not sure why it's not working. The bot is quite hackily written, and should probably be re-written

<klg> ⋄ -⍣=1

Execution timed out

TessellatingHeckler

bot aside, that does "first [" and "first ] after that" and a run of 1 1 1 between them

but it's so ugly

and it took so many hours to cobble together

Adám

4:44 AM

@TessellatingHeckler That is a multi-line block, in much the same way as 1 4⍴⎕A is a matrix, not a vector.

@TessellatingHeckler Cheating? ⋄ ∊'\[.*?]' '.'⎕S{~⍵.(Lengths⍴PatternNum)} 'blah ] blah [ blah [ blah ] blah'

Dyalog APL

@Adám 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0

Adám

5:03 AM

@TessellatingHeckler Proper APL solution: ⋄ {l∧⌽∨\⌽<\(']'=⍵)∧l←∨\'['=⍵} 'blah ] blah [ blah [ blah ] blah ] blah'

Dyalog APL

@Adám 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0

Adám

Oh, the Regex solution detects all such pairs, and you only wanted the first. ¯\_(⍨)_/¯ APL FTW.

Ah, here we go:

      Regex←≢↑0 1/⍨∘⊃'\[.*?]'⎕S 0 1
      ]runtime -c "Regex t" "APL t"

  Regex t → 3.5E¯6 |   0% ⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕
  APL t   → 1.4E¯6 | -60% ⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕
      t←⎕UCS 31+?1000⍴95
      ]runtime -c "Regex t" "APL t"

  Regex t → 2.5E¯5 |   0% ⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕
  APL t   → 1.5E¯6 | -94% ⎕⎕

Adám

6:30 AM

@TessellatingHeckler For large inputs, avoiding double-reversal improves performance:

      APL ←{l∧⌽∨\⌽<\(']'=⍵)∧l←∨\'['=⍵}
      APL2←{r∨l>∨\r←<\(']'=⍵)∧l←∨\'['=⍵}
      t←⎕UCS 31+?1e3⍴95
      ]runtime -c "APL t" "APL2 t"

  APL t  → 1.5E¯6 |   0% ⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕
  APL2 t → 1.8E¯6 | +16% ⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕
      t←⎕UCS 31+?1e4⍴95
      ]runtime -c "APL t" "APL2 t"

  APL t  → 3.8E¯6 |   0% ⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕
  APL2 t → 2.4E¯6 | -38% ⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕⎕
      t←⎕UCS 31+?1e5⍴95
      ]runtime -c "APL t" "APL2 t"

@TessellatingHeckler If you just want to extract the text, then simple indexing does the trick: ⋄ ((⊢↑⍨¯1+⍳∘']')⍳∘'['↓⊢) 'blah ] blah [ blah [ blah ] blah ] blah'

Dyalog APL

@Adám blah [ blah

Transcript for