The APL Orchard

1:21 PM

⎕←'ab' 123 ⍬

@ngn
┌──┬───┬┐
│ab│123││
└──┴───┴┘

⎕←'ab' 123 (⊂⍬)

@ngn
┌──┬───┬──┐
│ab│123│┌┐│
│  │   ││││
│  │   │└┘│
└──┴───┴──┘

@RichardPark question about Merge: the spec says we should use ⊂⍬ for missing data but then the example 5↑ All ←⊃Merge/ deathrate education ... with ]box on shows an empty cell for education in china, which means it's really ⍬. which one should we use?

1 hour later…

Feeds

2:27 PM

twitter.com/dyalogapl/status/1138074027854090240

2:37 PM

why is there such big overhead for ,¨ :|

2:50 PM

@dzaima this is starting to happen so often that i'm beginning to think it isn't a coincidence..

ok this is very broken
https://tio.run/##7Zy7btswGIX3PMV5BPEqaZUGTVoEFIZG9zKmBWogyBhkbgEPHjx07OgxY4u@i1@gj@CSTFzEbVEHBiVL1lEI6mJadnhAkf/HY76Zv7@ZL3a77erTdnX3LG0OT5lemJZX29X9AoDbzdqmrooSHW/uk9yHfjn8Ht8oxknpq9fv9Xzxzu3Qx5Yr6hdbv@sPb9GXfiajfrH1u1mEp9q@U9pMPj9aYPm7H7h/6hJ8@ryZeB4a5/@L/KPiVqFKHyafHy2w5uijg6ffx@tQqzqHSiFyCA2RIYcxMAoCQkEJyBRGQguoDDqBstDQFsZCuGsaUkG70q6UO00gc0gYt5cQroiBcgcJdEr9Yuv36rFV9DL6gKB@0UePt/3pZyz1i67foi/90iR5an@rw@/xQDFOSusrgqiI/Klom7KuCJ/InAicRs2ZrKZmZEvnYksES3/loUEeKUiqdCJV4rMuLkkSsMKDJJVBan@sYCCNv6TdC4EJuUNY2AQyg9Y@qTTAJAUpEFCUke6NnjZl/lBYd5PEoyVjOA4cIz1KqNnYiJEl7hsXJRJC…

3:09 PM

@ngn I haven't tried this problem yet but I'll look now and see if there needs to be a correction made - from a quick noodle it looks like ]Box displays a bit different if it's a higher-rank array

@ngn ⎕←{a←⍵ ⋄ a[2;2]←⊂⍬ ⋄ a}3 3⍴⍳4

⎕←{a←⍵ ⋄ a[2;2]←⊂⍬ ⋄ a}3 3⍴⍳4

@RichardPark
┌─┬─┬─┐
│1│2│3│
├─┼─┼─┤
│4│ │2│
├─┼─┼─┤
│3│4│1│
└─┴─┴─┘

But then again...

⎕←3 3⍴1 2 (⊂⍬) 3 4 5 6

@RichardPark
┌─┬─┬──┐
│1│2│┌┐│
│ │ ││││
│ │ │└┘│
├─┼─┼──┤
│3│4│5 │
├─┼─┼──┤
│6│1│2 │
└─┴─┴──┘

@RichardPark it should be ((⊂2 2)⊃a)← instead of a[2;2]←

square bracket indexing returns an enclosed item, so square bracket modified assignment does the opposite - it discloses whatever you give it

pick (⊃) doesn't have that behaviour

⎕←{⍵[1]}'ab' 'cd'

@ngn
┌──┐
│ab│
└──┘

3:16 PM

⎕←{1⊃⍵}'ab' 'cd'

@ngn
ab

Then my suspicion is that it is a mistake based on a solution involving modified assignment

Can you email the question to contest2019@dyalog.com and it'll get sent around

@dzaima ok so it's a problem somewhere between stdout & the shell interface - teeing the output to a file shows everything correctly. it seems it might be something failing to wrap lines, and without rlwrap everything goes smoothly so..

@RichardPark sure

3:32 PM

I'll reply to the email as well

But it looks like it's just ⍬

⎕←1⋄1 2 3

@RichardPark
1

⎕←a←1⋄a←a+3⋄a

@RichardPark
1

Sorry just figuring out how this works for multiple lines

⋄ a←1 ⋄ a←a+3 ⋄ ⎕←a

3:34 PM

@dzaima
4

@RichardPark yeah, most likely

@ngn Oop not quite - the explanation I've been given makes the same mistake I did I think...

@dzaima bug in rlwrap? interesting

@dzaima is there an easy way to get the correctly linewrapped tee'd file? i'd like to try rlwrap sh and then cat thefile

@ngn i just made a file with java -jar ... | tee log and called rlwrap on that

notes: i'm using gnome-terminal, rlwrap 0.41, terminal width=80 (and there are lines longer than that), broken characters seem to always appear on a lines start, the specific output varies from run to run, the only difference in the output being a millisecond counter

3:55 PM

@dzaima what do you type to get that output?

@ngn the specific code?

@dzaima yeah

whatever i need to do to reproduce it

the following saved somewhere and `)ex path/to/file`. note, it may run for a 100s, it's definitely not the best for testing :p
https://tio.run/##TVPBattAEL3rK/YmiYRahrrQQHpYH5ZA9qLiLDoqjVt6cGyyJRRsX5piZNUqLaHQSyFNKQhTejGB0ktBnzI/kr7Zld1Klnb2zXtvxrurZ/n5ZW7v7@n9p5djWnxIgsnwYiQQiWn3kJa3VN0d8IyqnwgFvXsr9jvTiKrfVF4Bifeb2kuqO1r9oarm1PKGqo1Xz6n4jKGpxT8oCHLvuaTFxygQIjSZkTpNtdaIMmOM5rnOFO6@0XyZkIlaSeOJGvkUkQLRM1LlBtkSDV8ST5ZqNlQyY2IGVxCNUqolplJKzbdz1M6R7YyWSqO6Sh2xtcNPc49IsyvoaR@W6BclwyAOrFu@ECwl034YDNycijUtv1J5G5w@94Sz8fkwFFR@E1PYexVWB/GL8Ss3s27tNrx2i@sciRMLmNlCTHYCKr83NSQOpuoLnl/TAWdmEztvsCGbR71d…

it can be "sped up" by changing L20s 65 to something like 90

hmm, it seems that maximizing & then unmaximizing is a required step

@dzaima i tried that but i don't get any bad characters

huh

4:13 PM

i suspect something interesting might have happened in the middle of a utf8 multibyte sequence - either java decided to flush its buffers or rlwrap got a sigwinch

well, without a repro there's no way to know. anyway

happens for me on rlwrap 0.43 too, no idea what else could be at play, with spamming resizing i can reproduce 100% of the time

3 hours later…

7:21 PM

@RichardPark AllMoves: the "White to move" example is missing e3 and e4

"Black to move" is totally wrong - there is no file i on a chessboard (is that f?) and some of the rook's moves seem vertically reversed

4 hours later…

11:07 PM

@dzaima Inspired by this, should (A f) be A f ⍵ or A∘f ⍵?

@Adám there's no difference between the two, no? or are you talking about the dyadic case

⎕←1 2+¨3 4 ⋄ ⎕←1 2∘+¨3 4

@Adám
4 6
┌───┬───┐
│4 5│5 6│
└───┴───┘

it would still be parsed as 1 2∘(+¨)3 4 though, without changing the way operators work

@dzaima Ah, I see.

11:23 PM

still really want good chars for left/right each. i wonder how bad would it be if i just used a combining diaeresis to get an opposite of ⍩: <̈

sadly seems almost nothing monospaced renders it correctly :/

@dzaima A f LeftEach B is A f¨⊂B and A f RightEach B is A∘f¨B?

@Adám yeah those exist, but they're ugly, take multiple chars, aren't obvious & are hard to read in a longer line

@dzaima I wasn't arguing against, just making sure I understand. Besides, the first one is very awkward for trains: A (f g h)¨∘⊂ B