the k tree

« first day (23 days earlier) ← previous day next day → last day (1512 days later) »

chrispsn

00:47

how do you make a keyed table in k7?

if the "flipped dict" is the only way of making a table in k7, i probably don't mind it; the '+' at the front of +[a: 1 2 3; b: 4 5 6] is suggestive of "t" ie table

also how does \u work? what do the cols represent?

also i found a kdb IDE: timestored.com/qstudio

chrispsn

01:04

feels like \lf, \lc etc should be dicts from the word go instead of needing to convert per groups.google.com/forum/#!topic/shaktidb/kyxGs9n2Vmc

yiyus

01:26

@chrispsn ([a:1;b:4];[a:2;b:5];[a:3;b:6])

chrispsn

@yiyus ah, nice!

@yiyus there's no way to get the opening and closing parentheses on their own lines without the newlines stuffing things up, is there?

so not:

tt:([a:1;b:4]
    [a:2;b:5]
    [a:3;b:6])

but:

tt:(
    [a:1;b:4]
    [a:2;b:5]
    [a:3;b:6]
)

the trailing ')' is not ideal for diffs

yiyus

01:43

wait. is that second form working for you?

btw, if what you want is to enter data "row by row", but you don't want to repeat the key every time: ab!/:(1 4;2 5;3 6)

chrispsn

@yiyus no, but it kind of works in oK (albeit not a table):

t: ([a: 1; b: 2]
    [a: 3; b: 4]
)

#t    / 2

@yiyus nice. do the col headers need to be symbols? eg:

t: `a`b!/:(1 4;2 5;3 6)

yiyus

@chrispsn: yes, they are symbols. sorry, the backticks got lost in formatting

2 hours later…

chrispsn

03:31

Any suggestions on how to structure community docs?

I set up a page here: k7contrib.gitlab.io/docs

only the table one is there at the moment - ideally the docs would show on the same page rather than loading a new one (frames?)

can be updated here: gitlab.com/k7contrib/docs

wasn't sure if a wiki was a better way to go

but gitlab seems sensible given that's where shakti is

any thoughts?

It doesn’t really work for mobile

also wasn't sure what to call the group given shakti seems to be using k7 as the foundation but providing more familiar layers on top of it so you don't need to use k syntax (see eg groups.google.com/d/msg/shaktidb/etsXDI9J4o4/HmefTXyBAwAJ)

anyway - idea is that the docs are a clickable version of the k quickref

4 hours later…

ktye

08:06

The docs look nice. What I want to do, is one line per verb/type, like:
!n enum: !3 / 0 1 2
!l odometer: !2 3 / (0 0 0 1 1 1;0 1 2 0 1 2)
!d keys: ! [a:1;b:2] / `a`b
and so on... because the verbs are so heavily overloaded which is not clear from the overview

chrispsn

@ktye agree - maybe the docs can be anchor links on a main page instead of separate pages, which would reduce frictions for making lots of small examples like that

i'll tweak

2 hours later…

chrispsn

09:49

ok, updated k7contrib.gitlab.io/docs

2 hours later…

chrispsn

11:51

@ktye these are now added

ktye

12:25

in ok, over and scan does not work with dicts:
+/ [a:1;b:2] returns 3 in k4 and k7
+\ [a:1;b:2] returns 3 in k4 but [a:1;b:3] in k7
Again: ngn anticipated this!

ngn

@ktye k4 doesn't support [] syntax for dicts, use `a`b!1 2 instead

ktye

@ngn you are right. but it does not return an error either!

ngn

@ktye why an error? reduction works on the values of a dict. scan scans the values and preserves the keys

ktye

@ngn i mean, k4 does not return an error, if you type [a:1;b:2]

ngn

@ktye it assumes it's a function application +/ [a:1;b:2] ←→ (a:1) +/ b:2

ktye

12:38

@ngn and the brackets alone? They are applied to what? [a:5;b:6] is no error but assigns the variables a and b

ngn

@ktye blocks

ktye

@ngn ok. Blocks for k3 and k4, but for kona it's a syntax error.

chrispsn

@ngn it wasn’t clear to me from that conversation why blocks are useful - it helps reduce the need for newline hacks?

ngn

@chrispsn yeah, for example if you want to do if(a){b;c;}else{d;e;} in k4, you could write it as $[a;[b;c];[d;e]] instead of resorting to hacks like $[a;if[1;b;c];if[1;d;e]]

chrispsn

@ngn ah, got it - thanks

2 hours later…

Nerd Rage

14:32

@chrispsn `a key [a:4 5 6;b:3 4 5] but select statements seem completely broken as of now.

1 hour later…

dzaima

15:38

what

2 hours later…

ngn

17:08

@chrispsn i can't believe a.w. is posting on a public forum, lol

@dzaima is that k7? it could be that the a.b syntax requires a's values to be a generic list, like in k4

or it could be a bug... idk

dzaima

17:32

also i thought namespaces were supposed to be immutable

ngn

@dzaima oh, that is serious! i think you should let them know

and does it really print dicts like that?!

dzaima

@ngn [a:1] prints as-is, but more than 2 keys seem to print like that

ngn

like in q - bad

1 hour later…

ngn

19:06

how peach works

rcabaco

19:58

what's the k equivalent of the APL compression (1 0 1 0 1/2 3 5 7 11)?

2 3 5 7 11[&1 0 1 0 1]?

ngn

20:24

@rcabaco yes. 2 3 5 7 11@&1 0 1 0 1 is a byte shorter, if that matters

rcabaco

that's nicer. thanks

it's unfortunate (for me at least) that Dyalog APL has been going with OOP instead of dfns for the tools they make

ngn

@rcabaco what do you mean? dfns are there to stay, as far as i'm aware

rcabaco

i was looking at their tools (conga, HTMLRenderer, etc..) and it's all OOP

i should say, OOP a traditional control structures.

s/a/and

ngn

@rcabaco well, i prefer dfns too but there's not much we can do about other people's coding habits

rcabaco

20:40

true.

ngn

if you mention it to Adám, he might be able to convey it to the dyalog team

rcabaco

i just commented from a novice's standpoint.

ngn

i don't even consider control structures and oop to be apl

they don't fit into the core language

dzaima

@ngn sometimes i like to think of Dyalog being 2 parts - the calculation APL and OOP & control structures as a system to utilize APL

rcabaco

in the context i was writing, it makes it harder to learn APL from their code

dzaima

20:52

@rcabaco imo HTML is better off being represented as OOP as it, well, is. idk about conga though, but from the little i've scrolled trough the docs it seems reasonable (though i haven't looked at the sources)

rcabaco

@dzaima you might be right. Perhaps it's just my self-inflicting programmer years kicking in.

and there are plenty of resources out there to learn from.

dzaima

when i heard oop I expected a couple dozen files but no, and just 6 classes! the code may be using :if and other constructs a lot, but there's really no other non-hacky way to do ifs

rcabaco

True. But then you lose the choice of not using classes, subclasses, method overriding, etc.. once you use it. At least i did not find a way not to.

dzaima

@rcabaco i really hate the argument of "if X exists, you must abuse it" sure, you can use OOP for a lot of things, but imo it should be pretty easy to tell when oop is better and when dfns/pure APL is better

rcabaco

21:09

I agree with you. Again, i just commented on the context of learning APL by reading code. I am quite sure they know much better what is the right approach.

dzaima

@rcabaco my learning of apl didn't involve reading code, so sadly i can't relate

rcabaco

may i ask how you learned?

dzaima

@rcabaco my own experimentation, these lessons, and the thing i feel most pushed me towards thinking more APLy - writing my own implementation

rcabaco

thank you for the link to the lessons

dzaima

also answers on PPCG are sometimes well explained and may push the boundaries of ones knowledge of the built-ins, and the APL orchard is a great place to ask for help & ask questions

« first day (23 days earlier) ← previous day next day → last day (1512 days later) »

all rooms

Transcript for

Mar30

Mar '1931

Apr1

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