Discussion between ASCII-only and Jo King

« first day (14 days earlier) ← previous day last day (15 days later) »

BMO

00:04

@ASCII-only It just checks whether there were multiple rules applicable, if so the computation is not determinitic

I did some work on Alchemist, prints the universe and "compresses" rule:

tio.run/…

Up to suggestions for output etc.

Also maybe debugging atoms?

@ASCII-only I don't understand what effectivelydeterministic is supposed to do?

Did you do some work on detecting multiple-rule cycles or the like?

8 hours later…

ASCII-only

08:10

@BMO whether the output will be effectively identical

@BMO i never finished making it work

@BMO no - i'm not entirely sure how to do that, you'd need to identify which rules are part of a cycle among other things, and i just think it would require a pretty big overhaul to make it work efficiently

@BMO i'd still like the bogus rule detection, times a rule is applicable before the set of possible rules changes, and exit if program no longer outputs that are in my version

@BMO this would be nice, yes

ASCII-only

09:00

@JoKing re: the breadth-first search, wdym by same state or whatever? (for caching) - the same rules applicable?

Jo King

@ASCII-only if the number of each atom is the same

+ the same output and input has been used

ASCII-only

:/ if that happens would it not be an infinite loop

also wow fast response lol

Jo King

what if two branches return to the same state?

you don't want to calculate it twice

ASCII-only

ah makes sense

Jo King

i guess an infinite loop is also deterministic

just... infinite

ASCII-only

09:03

true

i won't be surprised if this complexity is way too much (too many possible states) for all but the simplest programs and fully deterministic programs though :/

also hmmm BMO has updated to show better debugging and to collapse rules

Jo King

yay

ooh, and character IO

ASCII-only

:O really? nice

no maximum times applicable though

also peter taylor appears to have a faster one

but... afaict he selects a random rule then calculates how many times it can be applied

while ignoring the fact that there may be multiple applicable rules. there might be something i'm not seeing though

Jo King

yeah, that wouldn't work

_->5a+c;a+c->b+c;c->

ASCII-only

also clear and dump

@JoKing nah that's just what i think is happening, not 100% sure that's what he's actually doing - he's linked his interpreter on the alkanes answer

Jo King

yeah, they haven't updated the readme yet

so what, clear clears the universe? and dump prints debug?

ASCII-only

09:16

uhh brb checking

yeah looks like it

Jo King

i'm imagining it in the form %a clearing a

ASCII-only

also D: so much refactoring

nope it clears everything :/

i guess it's just a debug feature

Jo King

ASCII-only

no idea why clearing the universe is a debug feature, but... apparently it is

i guess it has a use in normal programs to halt or reset to just _, but no idea if that;s intended

also this keyboard is so ~~weird~~ small, always mistype enter when i reach for '

Jo King

what about ;?

ASCII-only

09:20

yes lol didn't even notice that

anyway... i guess time to send a PR adding the optimizations in

i'm just going to give up on the determinism check for now and figuring out better ways to optimize

since if we can do better optimization that's essentially better determinism checking as well, e.g. detecting things like a+b and a+c being able to be run at the same time

ok @BMO actually would you possibly accept a PR that adds the bogus rule, executing a rule multiple times if it's safe, and exiting the program if there's a non-outputting infinite loop that it can detect?

also ignore this, but... i tried to half-implement the new commits before i gave up (most importantly it has nice infinity not the hacky `-1` to detect times applicable):
https://tio.run/##7Vzrd9vGcv/Ov2IvohsCEUk9nN62iqlU1sNRq4cryrnpkVQZBJYiGhBgANCmEvtvT@c3s4uXSNlJz@nph8onILE7Mzs7Mzs7M7vM2M@nvwd@8fJlVydhd//4vR8Ppnnnt/5X6uzg4vXbg9fH6iTWy2gc69MkL/wk0HlPXT/O9egxSZPHWa31x0h/eOMXhc6SXH3V/9TpzNJwEWsFsspN50U0i37VPZUtkjdZ@tBTiySaPPbUkR4vHtzBwOupgyKdybfTZL4oiKpAXhGdnjr7YURf8TjSNMwsSqKcwT31Yaoz3elEs3maFeowTYosjQfnaeKHg1HhF5qeWRQUFuLIL/zB4dTPGg0naRz6NNdm4yIJiihNGo1nUd6kde4/jnWrZW4GVS597yk/JqZ7ahIl4d@j…

Jo King

the alkanes answer is cool

ASCII-only

O_o

i can't even understand it lol

Jo King

i suspect that the custom interpreter works because the answer is fully deterministic

oh, i'm not saying i understand it

ASCII-only

@JoKing i'm pretty sure it's because of the optimization he used

see this: github.com/pjt33/Alchemist.Py/blob/master/Alchemist.py#L82

active_rule = random.choice(applicable)
repetition_count = max_repetition_count(active_rule, state)

Also, hmm. I wonder if a graph of the previous/next states might be useful for figuring out how to optimize Alchemist

Jo King

09:37

just tried it with

_->5a+c
a+c->c
c->Out_a

always outputs either 0 or 5

rip

ASCII-only

well rip. so it optimizes a bit too much?

Jo King

yeah, raising an issue now

hmm, would you combine identical rules into one?

cos that would affect random rule choices

i guess you could make rules have weights lol

6 hours later…

BMO

15:56

@JoKing I'm not yet sure about these atoms, currently % just clears the whole universe and ? dumps the complete state to stderr.

@ASCII-only I don't think it's possible in general, maybe recognizing some often recurring patterns is possible though.

@ASCII-only Sure, if something gets faster/more comfortable and nothing is changed with the probabilities of executions, I'd like that :)

@JoKing Yeah, atm that's the only way to implement weights (which is a bit restrictive since you can only have fractions)

3 hours later…

BMO

18:39

@ASCII-only I tried to adapt it, but I seem to have broken it:

tio.run/##rTv7V9vIzr/…

Also why is there a new D3 flag? It doesn't seem to be needed :S

The thing is now that I changed the code, transform is not needed..

4 hours later…

ASCII-only

22:18

@BMO hmm?

@BMO it was going to be so there was a debug level for effectivelyDeterministic

@BMO yeah sorry that was one i was working on

@BMO chat.stackexchange.com/transcript/message/48821655#48821655

it should have all the optimizations i mentioned

BMO

@ASCII-only I'll see, things changed quite a bit.. Before it was just a quick and dirty thing, I did some refactoring :(

`_->10a+8b+s
s+a->s
s+b->s
s+0b->Out_a`

^ this is still not effectively deterministic, I don't understand what that means?

ASCII-only

basically, output isn't guaranteed to be same

BMO

Ah, maybe shouldn't be an Out-atom, sorry

But same with 0a+0b+s->Out_"x" instead :(

BMO

22:55

@ASCII-only: imgur.com/a/9qtvIOY

But this stuff is not feasible at all, keeping track of all the variables is too much :(

ASCII-only

yeah, it'd only show part

BMO

@BMO github.com/bforte/bforte.github.io

Somehow I get a 404, probably takes some time (never did github.io static sites before)

Did some typescript to graph the stuff but it wip

Ah it needs /index.html: bforte.github.io/index.html

ASCII-only

23:30

@BMO _->2a;a->b;0b->c breaks it

BMO

Yeah, most of it doesn't really work :S

One thing is zero-rules ^^

Also needs constant refresh, etc.

« first day (14 days earlier) ← previous day last day (15 days later) »

all rooms

Transcript for

Feb4

Feb '1914

Mar1

Discussion between ASCII-only and Jo …

Imported from a comment discussion on codegolf.stackexchange.c...

code-golf decision-problem number number-theory

join this room
about this room

00:00

06:00

12:00

18:00

all times are UTC