The APL Orchard - 2021-03-11

« first day (1506 days earlier) ← previous day next day → last day (1445 days later) »

rak1507

00:58

how can I do ']load HttpCommand' on tio

⎕SE.SALT.Load'HttpCommand' doesn't work for me

user

01:19

@rak1507 I thought putting it in the input would work, but it doesn't work either.

rak1507

tio.run/##SyzI0U2pTMzJT////1HfVOdI9eLEnBL1R90tqXmJSTmpwY4@If//…

it must not be installed or whatever

which is a right pain

user

Just take a screenshot of your code running.

rak1507

nah

Razetime

02:03

whats the equivalent of ⍎ in J?

Marshall

@Razetime ".

Razetime

cool

It's called "Do"

rak1507

03:01

damn, why does f⍢⊤ not work in extended

should be equal to 2⊥f⊤⊢

2 hours later…

nathan rogers

04:44

any k people in the house I have a selection question

first does K have pattern matching? I'd like to match rows in a table that match a given pattern (1;2;3) someF (any; 2; any), ideally I'd have liked if ((1;2;3);(3;3;3))^(:;2;:) worked, but it doesn't

second, what is the idiomatic B/A or B⌿mat in K? I have a basic selection, but its inelegant and I'm trying to learn

{x@&~4=x}1 2 3 4

Adám

05:22

@rak1507 SALT isn't enabled by default on TIO.

user41805

@rak1507 looks like UpTack doesn't have its obverse defined github.com/abrudz/dyalog-apl-extended/blob/master/UpTack.dyalog

rak1507

@Adám oh, I tried enabling salt, but I didn't know you had to pass the full path

user41805

@rak1507 if you interested in making a pr, my commit may be helpful github.com/abrudz/dyalog-apl-extended/commit/…

rak1507

thanks, I'll have a look later

Adám

@rak1507 You don't even need SALT.

ngn

06:23

@nathanrogers yes x@&y is idiomatic

@nathanrogers so you want to match the rows that have a 2 at column [1]? t:(1 2 3;3 3 3); t@&2=t[;1]

Adám

07:09

@ngn What do you think about making K off-topic here other than from the perspective of APL (comparison, ideas), and instead asking people to take K questions and discussions to your K Tree?

2

2 hours later…

Morten Kromberg

08:49

Regarding the problem of determining the path of the currently running code: Do y'all think this is the question that one generally wants to ask, rather than the path of the script or source file that was used to START the current process?

If you "build" a workspace or a bound executable from source files in order to have a more easily distributable package, you won't have loaded the script files from the environment where your package has been deployed. At that point, your application root directory would be wherever the workspace or bound executable was found.

I'm doing some work on the next release of Link at the moment, and I'm exprimenting with adding a ⎕SE.Link.LaunchDir API function, which will tell you from whence things were started.

1 hour later…

ngn

09:59

@Adám you're one of the room owners here - do whatever you like

xpqz

10:31

Morning/afternoon/evening, APLers.

ab5tract

hey there @xpqz

xpqz

11:09

Can anyone suggest a clever way of achieving the follwing "left join":

Given two vectors, the first a vector of char vecs, the second a vector of two-element vectors, group the second based on the first component, returning the second components per group, inserting the empty vectors for groups not present.

e.g

X←'one' 'two' 'three' 'four'
Y←('one' 1)('one' 2)('three' 1)('four' 2)('four' 3)
R←⎕←X f Y
┌───┬┬─┬───┐
│1 2││1│2 3│
└───┴┴─┴───┘
R≡(1 2)(,⍬)(,1)(2 3)
1

Adám

@xpqz (⊃¨{⊂1↓⊢/¨⍵}⌸⊢)(X,⍥⊂¨0),Y

xpqz

wow, nice

rak1507

11:50

also {/∘(1⊃¨⍵)¨↓⍺∘.≡⊃¨⍵}

Adám

12:00

Nice, but probably inefficient.

rak1507

yeah :(

Marshall

It's x⊸⊐⊸⊔˝⍉>y in BQN (group continues to be awesome)

dzaima

12:52

@Marshall and I continue complaining about it ignoring trailing empties

Marshall

@dzaima Pretty easy to fix though.

dzaima

+66% length to fix an edge-case that really shouldn't exist in the first place

Marshall

13:17

@dzaima But there isn't a way to never have that edge case in Group, is there? The safe solution is still much shorter than either Dyalog version or the slow (and golfed) (≡⌜˜⟜x<∘/⎉1⊢)˝⍉>y.

rak1507

something called 'group' should group things properly

dzaima

@Marshall with 2 arguments, I don't think there's a way. But a 3-arg group taking keys;values;output keys, would. Unfortunately, the character restrictions make finding a built-in for that nigh impossible

Marshall

@rak1507 All the things go to the correct group. The problem is that maybe a group you wanted that has no things won't be there.

rak1507

all the things might go to the correct group, but all the groups don't go to the correct things

Marshall

@rak1507 The correct absence of things.

rak1507

13:22

yes

Marshall

@dzaima I mean, if you're implementing a database you can always just define Group←{k‿v‿o←𝕩⋄v⊔˜(o⊐k)∾≠o}.

dzaima

@Marshall i'd argue that's a pretty good replacement of ⊔ in general

rak1507

@dzaima thoughts on extending ⍀ to do a traditional rtl scan in dzaima/APL?

dzaima

@rak1507 it's already taken up by expand, which is a function

rak1507

ah and you can't do two things like dyalog does?

dzaima

13:31

@rak1507 that's called schizophrenia, and that is bad

rak1507

rtl scans though :( new symbol for it?

dzaima

@rak1507 is it really that hard to ⌽ & ⍨?

rak1507

yes

Marshall

@dzaima I think a version of ⊔ that always requires a final count would be a lot easier to use. But you can just program that way anyway, and the current ⊔ is much more convenient in cases where every group is guaranteed to have entries (like monadic ⊐⊸⊔).

dzaima

(i'm much more likely to add a ltr reduce than a rtl scan :p)

Adám

13:33

@dzaima `f` doesn't have a dyadic meaning in dzaima/APL, does it?

Marshall

Current Group is also a lot easier to structurally invert.

dzaima

@Adám it doesn't, but i'd rather not abuse it for a reduce (also ``f\``)

(dammit did i really not test my code escaping :|)

Elias Mårtenson

13:52

Wow. Getting the parser to do the right thing for short-form function definitions was a nightmare. Turns out that support dyalog-style is actually easier than my earlier proposed style. Since I still don't want to support full-on Dyalog style (for the reasons outlined earlier) I have temporarily gone for this:

add ⇐ +

This is the same as Dyalog, but a double struck arrow instead of single-struck.

Adám

@Adám This was fun. Do I get 200 rep now?

Bubbler

@Adám I can give you 200 :D

Adám

@Bubbler That'd be nice. Then I have more to give to you…

rak1507

dfns.roman would shorten that a bit I think

Adám

> No libraries that do roman numeral conversion for you

rak1507

14:03

oh

do you think I can put ⎕FIX'file:///opt/mdyalog/17.1/64/unicode/Library/Conga/HttpUtils.dyalog' in the header or should that be in the actual code?

Adám

@rak1507 Either way, but if you put it in the header, then it needs to go in the header of your answer too: "Dyalog APL with HttpUtils"

rak1507

is there a way to make 'input' in tio not act like a repl?

dzaima

@rak1507 what do you want it to act like? just be completely ignored?

Adám

@rak1507 As a single text array?

dzaima

heh, {⍞}⍣⊢0 in code stops interpreting further code and exits, but the more obvious ⎕off is disallowed :D

Adám

14:19

@dzaima That's because you're actually inside a :Namespace definition.

dzaima

@Adám well yeah, but why disallow ⎕off if one can get around that anyway

Adám

@dzaima TIO's interpreter stops because input is redirected and so there's no chance the user will enter more. Not a normal stop.

rak1507

@dzaima regular stdin that I can read with ⎕/⍞

dzaima

@rak1507 that's already the case

Adám

@rak1507 Uh, it is.

rak1507

14:25

oh

I must be doing something wrong then

my bad

Adám

@rak1507 TIO link, please?

rak1507

I'll try and fix it myself

dzaima

@Adám oh, it cathes (and hides) the error that dyalog creates. why though

Adám

@dzaima So that the exit code can indicate the error.

dzaima

@Adám oh, regular Dyalog 17.1 throws up even if the input in REPL mode is cut short

Adám

14:29

@dzaima Yeah, the -script was only an unofficial hack at the time.

dzaima

wait, -script and --script are different things?

well, whatever anything does, imo the correct behavior is to quietly exit when the REPL hits EOF, but give an error if a ⎕/⍞ gets EOF

rak1507

'thing'≡+'thing', til

1 hour later…

Adám

15:59

Announcement: BAA webinar now at zoom.us/j/858532665 with Passcode ← ×/1920 12 17

ngn

16:29

@xpqz one way to do it k (while it's still allowed here): make a dict mapping X's items to empty lists and amend it with the keys and values from Y: .@[X!(#X)#,();;,;]/+Y

(amend with ,)

something similar in apl:

K V←↓⍉↑Y ⋄ R←⍬⍨¨X ⋄ R[X⍳K],←V

xpqz

16:49

@ngn Nice!

Razetime

17:32

@Adám dammit i found this too late

Dyalog windows is such an upgrade from mac

nathan rogers

18:11

@ngn the ok docs refer to m_l where m is a boolean function, will remove where 1 from l, and also m#l will select. Does NGN support these? I couldn't get either of them working correctly

ngn

@nathanrogers only m#l now but i'll add m_l too. in the meantime you can use {~..}# instead of {..}_

Adám

@Razetime You may want to log issues against RIDE for what you were missing there.

xpqz

18:44

Can anyone call a CMC?

rak1507

yes

xpqz

CMC: read a .ini file, and return it as a nested namespace (en.wikipedia.org/wiki/INI_file). Assume it has no commets.

rak1507

no backslash escapes either?

xpqz

Nope.

Just key=value and [section] and empty lines

rak1507

can value be a string like test="whatever"

in an example I found, it has this

server=192.0.2.62
port=143
file="payroll.dat"

how come the third one has quotes but the first two don't

xpqz

18:50

I think quotes are optional. All uses I've seen ditch the quotes

I wonder if Dyalog has one built in on some random i-bar

rak1507

hmm I think this cmc might have to wait as I just felt an incredibly sharp pain in my side, brb

ngn

@xpqz replace a=b with a←'b' using a regex and ns.⍎ it :)

xpqz

Will that handle sections?

nathan rogers

@rak1507 :O ono get better

xpqz

@rak1507 hope you ok buddy

Adám

18:54

@xpqz github.com/aplteam/IniFiles

xpqz

1000+ loc !

rak1507

well idk what that was about but back to the apling

xpqz

mine is 5

ngn

@xpqz how should they be handled?

rak1507

@xpqz no nested sections right?

xpqz

18:57

The ini file is basically a js object, or a py dic. Only one level, yes.

rak1507

ok

ngn

[s] could be replaced with ns←#.s←⎕ns⍬ or simply ignored

xpqz

I want to be able to say: ns.sectionname.keyname basically

gist.github.com/xpqz/8a0b794614a5fe667b01673ea0c13a27

rak1507

{root←⎕NS⍬ ⋄ root⊣⍎¨'(.*)=(.*)'⎕R'ns.\1←''\2'''⊢'\[(.*)\]'⎕R'ns←root.\1←⎕NS⍬'⊢⊃⎕NGET⍵1}

ngn

@rak1507 shouldn't it be root.⍎¨ instead of ⍎¨?

rak1507

19:07

oh, I forgot you can execute code inside namespaces like that, that code does work as it mutates root and then returns it, but your method will be simpler

ngn

ah, you have root.\1 so it's fine

rak1507

{r⊣(r←⎕NS⍬)⍎¨'(.*)=(.*)'⎕R'ns.\1←''\2'''⊢'\[(.*)\]'⎕R'ns←\1←⎕NS⍬'⊃⎕NGET⍵1} works

ngn

cool

@rak1507 ⊢⊃ → ⊃

Adám

@rak1507 ).⍎ → )⍎

rak1507

edited

ngn

19:11

it would be 1 byte shorter with a single ⎕r with multiple patterns

rak1507

I'm not trying to code golf

ngn

but whyyy.. sigh of disappointment

Adám

@xpqz Can we assume all values are inside sections?

rak1507

@ngn readability :D

Adám

Meh, this turned out longer:

{⎕JSON⍠'Dialect' 'JSON5'1⌽'}{',2⌽∊'^\[(\w+)]' '^(.*)=(.*)'⎕R'},\1:{' '\1:"\2",'⊃⎕NGET⍵}

@rak1507 Doesn't handle empty lines, nor keys named ns

rak1507

19:27

oh empty lines

Adám

That's just (…)~⊂''

It also populates the namespace with an extra ns name.

rak1507

true

does your json one work?

Adám

I believe so.

rak1507

nice

Adám

It also has the benefit of being safe (no ⍎) and auto-mangling names that are invalid in APL.

rak1507

19:30

definitely the smarter way to do rather than eval, which is pretty hacky

yeah

Adám

Mine also handles = in values. (Well, it can, at the cost of another few bytes.)

@rak1507 Test input:

[sec-tion]
ns=

[]
===

Test run:

      ⎕JSON⍠'Compact' 0{⎕JSON⍠'Dialect' 'JSON5'⊢1⌽'}{',2⌽∊'^\[(.*)]' '^(.*?)=(.*)'⎕R'},"\1":{' '"\1":"\2",'⊢⍵}t
{
  "": {
    "": "=="
  },
  "sec-tion": {
    "ns": ""
  }
}

rak1507

if someone has that as a config they deserve whatever nonsense they get

xpqz

:)

rak1507

what's up with the "": "=="

that doesn't seem right

Adám

@rak1507 Why not? It comes from === which means the empty string key has the value "=="

rak1507

19:35

oh right, I was interpreting it as the key "=" has the value "="

but I see now how that would not make very much sense

Adám

Not sure which it should be, as there's no real Ini file standard.

rak1507

having = as a potentially valid key character introduces ambiguity

Adám

How so?

rak1507

foo==bar

is foo equal to =bar or is foo= equal to bar

Adám

I think the spec is that the first = stops the key name.

Question is if the key can be empty or not.

rak1507

19:38

yeah, that makes sense to me

I think an empty key being allowed is fair enough but I don't think a user should necessarily expect something sensible from it

Adám

@rak1507 The Windows registry is often represented with Ini files, and there are plenty of empty keys there.

rak1507

of course it is, typical windows

Adám

However, regedit exports inifiles with keys in quotes. Hm.

Ah, they have lots of fancy features…

rak1507

yeah I saw backslash escaping in the wiki article

Adám

20:02

@rak1507 Mine fails on values that are not in a section.

rak1507

ah

Adám

Fixing it isn't so easy.

Adám

20:20

@rak1507 @xpqz OK, exactly 100 bytes (without the file read) for a fairly solid implementation:

{⎕JSON⍠'Dialect' 'JSON5'⊢2⌽'}}{"":{',∊'^;.*' '^\[(.*)]' '^(.*?)=(.*)'⎕R'' '},"\1":{' '"\1":"\2",'⊢⍵}

rak1507

nice

Adám

Things before the first section go into a special ⍙ namespace.

(This is necessary to allow a "global" key to have the same name as a section.)

Also allows comments in the form of lines that begin with ;

Ignores empty lines.

Handles all key names (even empty ones) but a 0-length section name will replace the "root" section (if any).

Also handles JSON style escapes.

Tacit version saves 4 bytes:

⎕JSON⍠'Dialect' 'JSON5'∘∊2⌽'}}{"":{','^;.*' '^\[(.*)]' '^(.*?)=(.*)'⎕R'' '},"\1":{' '"\1":"\2",'

xpqz

20:39

@Adám good job!

Adám

Almost APLcart material?

« first day (1506 days earlier) ← previous day next day → last day (1445 days later) »

Transcript for

Mar '2111

The APL Orchard

apl.chat ― Learn, teach, ask, code, golf, & discuss usage. See ...

apl array-manipulation j tips

join this room
about this room

00:00

06:00

12:00

18:00

all times are UTC

site design / logo © 2025 Stack Exchange Inc; legal

mobile