the k tree

« first day (843 days earlier) ← previous day next day → last day (692 days later) »

rak1507

00:35

groups.google.com/g/shaktidb/c/8dDaO_ZDg1I/m/xxeE7mB-BwAJ this really annoyed me, 'we want to make it easy', no you don't, if you wanted to make it easy you'd have some actual documentation

JohnE

01:03

> You don't have permission to access this content

3 hours later…

Razetime

04:21

@rak1507 it would be great to have a central place for K related documentation and info

4 hours later…

chrispsn

08:16

@coltim agree, though I think odometer is useful, and list keys are different to dict keys in that they don't 'move' with the values

@ngn that is still the same :)

3 hours later…

ngn

11:19

@rak1507 i think getting the implementation right first is more important. implementation leads, documentation follows. for popular software like shakti, they don't even have to lift a finger - some fan will eventually write the docs. there already are a few competing attempts.

chrispsn

11:48

it looks like 'rank sensitivity' is already a concept in j - see 'internal rank' code.jsoftware.com/wiki/Vocabulary/IFamily

i really need to read the j wiki back to front

also TIL the word 'brick'

ngn

@chrispsn yeah, but j has multidimensional arrays like apl (making a distinction between matrix and vector of vectors), so there's never ambiguity

"brick" - nice :)

chrispsn

12:21

it would be interesting if v worked directly on the list, v. gave indices and v: gave a boolean mask

maybe that was the idea but the use of . and : doesn't seem to signify a specific meaning in J

maybe it wouldn't generalise well?

rak1507

@ngn so, rather than they want to make it easy, they want other people to make it easy for them?

ngn

@rak1507 i think they want to make the implementation easy to use. what you seem to want is documentation to make it easy to understand.

rak1507

12:36

Ease of use and documentation might as well be the same thing

ngn

maybe, but they don't have to be made by the same person at the same time

rak1507

True, but if they actually wanted to make it easy like they claim, they would be doing more to try and document things rather than relying solely on community provided material imo

Razetime

13:04

I'm having a problem with dictionaries(i think)

this program gives 'stk error on run

but I'm not sure why it's overflowing

ngn

@Razetime 2|x - did you mean 2!x?

Razetime

dammit

how do I return the result of a variable declaration?

ngn

@Razetime like, you have a:b and it's the last statement in { } and you want to return a as result?

Razetime

yes

or inside a $[]

ah, :

ngn

@Razetime you could use { .. a:b; a} or { .. :a:b} (:x is "return")

@Razetime i think in $[] it should just work

Razetime

13:08

so assigning inside

ngn

assignment evaluates to :: only when it's the last expression in { }

Razetime

assignment in a block always needs :: right

otherwise it won't update outside

ngn

@Razetime yes, a:b is local assignment (local to a lambda), a::b is global

more or less like apl but without intermediate scopes

a:b is like apl's a←b and a::b is like apl's a⊢←b

Razetime

a::b no matter what is always global

ngn

yes

@rak1507 ^ :)

Razetime

13:31

is there a way to set integer precision?

ngn

14:21

@Razetime what is integer precision?

Razetime

@ngn also, I'm getting 'dom even when there's an EOL at EOF

@ngn err i figured a way around it

it was a thing that needed bigint arithmetic

ngn

@Razetime do you have an example?

Razetime

ngn.bitbucket.io/…

wait, I'll upload the file

i'll try changing the encoding again

apparently i set it so some obscure unix encoding

1 hour later…

Razetime

15:55

i am having a problem here

d takes proper divisor sum

but the problem is x=s and ~x=s are still giving the same result?

@ngn ↑

somehow it works perfectly on the online interpreter??

ngn

16:12

@Razetime interesting. i get the 1 1 for the last expression, both online and on my laptop

coltim

@ngn I get 0 1 there as well o_0

I think x=s returns 0h there, but the ~ on that doesn't flip the value

Razetime

maybe wsl is drunk?

it's still giving the same result here btw

2 hours later…

ngn

17:59

@Razetime @coltim i'm afraid i don't have a windows machine to debug it

coltim

@ngn I'm on linux though =|

ngn

@coltim and you get 0 1 for the last expression?

coltim

yeh; and whenever it's variable=integerLiteral where the integerLiteral is <= 32,767

casting x=s to any of b i l fixes it

x~s also works

ngn

huh

coltim

variable=integerList works too (even for values in the same ranges)

ngn

18:11

@coltim so, a:999;~a=999 prints 1?

@coltim which compiler do you use?

coltim

@ngn that prints 0

@ngn I had to insert a override CC=clang-10 in my make file since the default is gcc or something

ngn

@coltim another option is to export CC=clang-10 in .bashrc and use make -e instead of make

i need a (preferably small) repro to test on my linux machine

can't afford to have bugs like that. or else the golfing kids will laugh at me :)

coltim

@ngn maybe I'm getting confused or something but a:999;~a=998 should return 1 but returns 0 for me

rak1507

ngn.bitbucket.io/k/#eJxLtLK0tOTSSLS1tLSwVqgD05pcEBoAYh8G2w==

weird

ngn

@rak1507 that's good - repro in wasm! thanks

coltim

18:21

@rak1507 huh, there ~999=998 returns 0 as well

(locally for me that returns 1)

rak1507

ngn.bitbucket.io/k/#eJyrM7Q1NDLnqgNRFgAXFAMY

something interesting going on there

coltim

@ngn you could try replacing the -march=native on your end

ngn

@coltim no change (at least on freebsd)

dzaima

mostly unrelated thought - now that integer types are transparent to the user, it could be beneficial to remove `b`h and maybe even `i

ngn

@dzaima right, the time has come

(after fixing the bug, of course)

dzaima

18:32

@ngn well, if you'd remove `h impl, there would be no more bug :)

ngn

19:26

@dzaima thanks! that worked

i spent most of the time chasing a bug in cat() - 1 2,3 was giving (1;2;3) instead of 1 2 3

gotta find a way to collect people's golfs and turn them into tests (like my own golfs) because who knows what else might be broken..

dzaima

@ngn :D

@ngn it's gonna be quite some manual work in any way you go about it. But it should be possible to query with SEDE and go from there

4 hours later…

coltim

23:38

@ngn what's worked for me is to attempt to golf things. I almost feel like a (relatively) straightforward "throw an arg of each type at each builtin, including all flavors of empty lists and nulls" would cover a bunch of the low hanging stuff

throw in a few inplace amends, and all the flavors of reduces/scans

one of those "combine X characters" as a fuzz test would be interesting

« first day (843 days earlier) ← previous day next day → last day (692 days later) »

all rooms

Transcript for

Jun27

Jun '2128

Jun29

app.element.io/#/room/#ngnk:matrix.org

join this room
about this room

00:00

06:00

12:00

18:00

all times are UTC