Polyglot Development

@Qwerp-Derp Hello

3:22 AM

OK, somehow I think this polyglot is never going to work in REDGREEN :-D

imagine it's like Fission except there's no way to halt it and some of the instructions self-replicate

@ais523 Link? Google gets me this

catseye.tc/node/REDGREEN.html

in general, esolangs.org is a good way to find links to esolang specs

Oh yeah

I was hoping to add piet

but it is incompatible with ><>

btw how do turns work?

I know I can't go next

but I would ideally like to go as soon as possible

you can't go twice in a row, apart from that there's no restriction

two people can go back and forth indefinitely

I meant more informally it seems like there are probably 3 other people who would like to go

3:28 AM

so you can jump into 57 if you're quick after the posting of 56

nah, when it's this late, being able to post at all is fairly difficult

and we should probably not step on each other's toes

so we're all hoping someone else will post and reset the timer :-D

really I have so many ideas

I'm fairly low on them

busy trawling Esolang to try to find a language that works

Currently I am thinking of Brain-Flueue, Grass and Wise

but @0' said he wanted to do Grass

so I thought I would let him if he decided to go through with it

Wise will be really easy as far as I can tell

and grass wont be too hard

Brain-Flueue will be harder than Classic

3:33 AM

oh wow, I had an idea

and it's working out better than I ever expected

I could also do Crain-Flak

what number's the next language? 56?

@ais523 What's the idea?

yes

Rust

it has C-style comments, except they nest

meaning that all we need is an extra /* inside a C comment and we can split it away trivially

Wow that does sound like a good idea

3:38 AM

oh, except it doesn't understand linemarkers :-(

so the very first line of the program breaks it

that's probably unfixable

rip

I would imagine that the very first line is probably a deal breaker for a lot of languages

(in Rust, # can only legally be followed by [ if it appears at the start of the program)

yep

# is a great way to start because it's a comment in so many languages, and a linemarker in many others

but you can't handle every language at once

Have you tried python 1?

if it's at all possible to put something else there while keeping every language happy, it's probably a letter

also, no; I'd likely have better luck with perl 4 :-P

I wonder if there are any other languages with nesting /* comments + // comments, though, and that understand linemarkers

@ais523 Why's that? I don't know much about programming languages.

3:42 AM

oh, it's just that I know something about perl 4, but basically nothing about python 1

Ah

I have the opposite

Python is basically the only High level language I use

This challenge is really fun

I'm remised that I didn't get into it earlier

(or later for that matter, I should really be studying for finals)

oh right, it's easter

you should probably do that

big programming tasks are more fun in the time after exams IME, than in the time before

easter?

"easter holidays" is the name in the UK for the time after mainstream teaching has ended for the year but before the main exams of the year start

Oh ok

3:48 AM

and it's that time now, to within normal variation between university's schedules

I go to Uni in the states so I've never heard that term.

generally I find I have more work towards the end as professors try to cram in all the info they missed due to "snow days"

hmm, bc has # comments and spouts warnings on syntax it doesn't understand (it says error: in the warning message but it keeps running anyway)

I think that counts? seems a bit messy though, perhaps I'll leave that for when we're desperate

bc?

Currently Japt makes a bunch of errors, so maybe we are that desprate

there might be a way to get bc to work without errors, though, I'll try that

also it's one of those calculator programs that's had so many features added to it that it's Turing complete

Oh yeah

accidentally TC is my favorite thing

3:55 AM

it's not hard to make something TC by accident :-)

Mistah Figgins

4:28 AM

like little big planet

or powerpoint

I haven't looked at the powerpoint construction

but I suspect it's just a state machine, rather than a proof of Turing completeness

being able to run a Turing machine on a finite tape doesn't really prove anything, as it needs an infinite tape to get its power

4:47 AM

ooh, looks like I have a working entry already

this time in dc, which is fairly commonly used on PPCG

and is somewhere between an esolang and a practical language

A: Add a language to a polyglot

5:18 AM

0

56. dc, 1286 bytes #16 "(}23!@)(" 3//*v\D@;'[af2.qc]'#)"14";n4 #/*` PkPPX (22)S"[!(>7 7*,;68*,@;'1,@␉␉␉␉ P''53'S^'q #>␉ # >36!@␉ #`<` #<]+<[.>-]>[ #{ #z} # #=x<R+++++[D>+++++++EAL+++<-][pPLEASE,2<-#2DO,2SUB#1<-#52DO,2SUB#2<-#32DOREADOUT,2PLEASEGIVEUPFACiiipsddsdoh]>+.-- -. >][4O6O@ #x%+>+=ttt Z...

given that there was a language this easy still out there, I'd imagine that there are others

6:03 AM

the same trick I used for dc might work for Tcl (not sure, haven't looked into it too closely)

Versert looks like a reasonable possibility if we want another 2D language

11:53 AM

Ok I have Wise working in everything but nim and Incident

1:55 PM

I apparently missed a lot of chat. I've had my mind blown several times in catching up lol.

On the subject of dropping Cubix into evil: you have to make sure Alphuck doesn't parse it, as O in the input command.

Does anyone know how the haystack works rn?

nvr mind

Its now working in everything but Incident

@ais523 I think I looked at both bc and dc a while ago (in the late 20s, iirc, when I was still trying to use languages I'd heard of) but had no luck. Glad to see at least one of them has made it in now.

Somehow incident works?

I'll take it

Oh I see why

I accidentally did something smart

Ok that works

Also, in terms of easy languages to add, some of the interpreters I've found handle a bunch of random BF-derivatives that are just character swaps. The primary issue with those has been length (particularly once we began a more determined push to bring the VIP score down). If there were more with <4 character commands it'd be easier, but most that I found had too many commands that were longer than that.

Ok finally caught up on chat. Now to read these answer rundowns.

@WheatWizard one random warning about the main incident line: most of the cow code is being used as spacers. It can be shu…

A: Add a language to a polyglot

2:14 PM

0

57. Wise, 1300 bytes #16 "(}23!@)(" 3//*v\D@;'[af2.qc]'#)"14";n4 #/*` PkPPX (22)S"[!(>7 7*,;68*,@;'1,@␉␉␉␉ P''53'S^'q #>␉ # >36!@␉ #`<` #<]+<[.>-]>[ #{ #z} # #=x<R+++++[D>+++++++EAL+++<-][pPLEASE,2<-#2DO,2SUB#1<-#52DO,2SUB#2<-#32DOREADOUT,2PLEASEGIVEUPFACiiipsddsdoh]>+.-- -. >][4O6O@ #x%+>+=ttt...

3:12 PM

Has anyone tried starry?

I looked at starry for a bit. if i remember, whitespace is vital to stary, and we have Whitespace (the langeage) in the polyglot, which cares exclusively about whitespace

The Whitespace has run it's course by this line: #z}

Yeah starry cares a lot about whitespace.

My faulty memory is that starry would have to get past Whitespace's code, and that seemed too much of a hurdle.

I've done whitespace starry polyglots before

IIRC starry actally doesn't care what type of whitespace it encounters just how much there is

I'll look into it once the next answer is posted

whitespaces current solution is pretty ridged on whitespace in the top line. you might need to create a new whitespace solution to add starry.

I'd love to see it.

I can't believe 3 answers got posted today.

3:36 PM

@WheatWizard good write up with #55, you explained away some misconceptions I had with Japt and Thutu. Thanks

Ok, I really don't know how Japt works I just know it likes numbers in things.

I wouldn't take anything I say too seriously, its just what I think is happening not necessarily what is actually happening

well, yeah. but that's like everything for me.

I like how you frame the behavior it's simpler for me.

truth is irrelevant. :P

I barely know how most of these languages work anymore. I'm familiar enough with the usual suspects to pacify them with some trial and error, but that's about it.

I know nothing of 05AB1E, which is funny to me since I added that one. I also added Labyrinth and I'm mostly just surprised it's never been a real problem; no idea how it's getting along anymore lol.

3:51 PM

I like to think of it as specialization.

I'd bet you know how evil works in here better than anyone, just cuz you've worked it more.

That is probably the one I feel most comfortable with (ignoring all the simple BF variants). Thankfully now that just mostly means trying to vaguely remember what commands are dangerous in it.

I seem to get the Necromancer badge every time I answer the polyglot challenge nowadays

it seems like something's wrong with the name

Necromancer is one badge that is way easier to get on PPCG then on other sites

brainflak experts: what does [{}] mean?

negative pop

4:03 PM

and what happens if you pop an empty stack?

nothing

oh good

that probably hurts brainflak classic though

implicit zeros ftw

oh yeah big time

that will print 0

you can skip it with a loop if you really need to use it

{[{}]} does nothing

I'm trying to see if Tcl fits into the polyglot

it uses { … } as string literals

(basically because things like if are implemented as functions that take string literals as arguments, and eval them only if the condition holds; it's a weird language)

Ruby seems not to like [{…}] either, so I'll have to try something else

4:20 PM

my tester sense is telling me to look at lua again.

so far I haven't even managed to construct a legal Lua program that starts with #

# means "length of" in Lua, it's a unary operation, and the problem is that you can't start a Lua statement with an expression

lua accepts our line 1. I seem to remember finding something in the documentation a month back or so.

just not the line 2.

it's probably OK with linemarkers

hmm, I wonder if we could write the start of the program entirely in linemarkers?

lots of languages would like that, not just Lua

it's too early for talk this exciting.

I'll never get anything done.

Japt would throw a fit, I imagine, but we might be able to deal with that separately

4:46 PM

@Chance You were saying at one point that it would be pretty easy to add more shell scripting languages, and you linked a resource. Where was that, I can't seem to find it any longer

hyperpolyglot.org/unix-shells

I kinda liked the look of ksh. I was poking around yesterday.

But I never found an exploit.

actually, Lua doesn't use linemarkers

"The first line in the file is ignored if it starts with a #"

that means that we probably need a very long first line that does as much as we can

and to start the second line with a letter, most likely

lots of languages are OK to have assignments

7:06 PM

I there a lua statement that can start with #?

I gather that #"" returns the length of the string, but it doesn't like this as a standalone statement.

right, Lua statements aren't separated by semicolons, newlines, or anything really

(you can put semicolons or newlines there if you want to but they're optional)

this means that it has to be very restrictive with what can appear in a statement

and in particular, an expression can't appear at the start of a statement

# can only start an expression, so it isn't valid at the start of the program

(and in fact the Lua parser will error out as soon as it sees as # where a statement is expected)

the scripts don't like a=#"" either. :/

I think I could maybe work a trick like you did with dc, but all the #x block comment languages would still parse the lua line.

I think that can be made to work

Python would be the hardest choice

I believe q=[[ is the best way to start a Lua/scripting language polyglot

I've done that before

but then nim, julia, octave, ect. have problems.

(Perl doesn't allow arbitrary identifiers as variable names, but q is a keyword which happens to behave very nicely in this way; and it's valid syntax in nearly all the rest, except possibly the shells)

but yes, the nim/julia/octave set want you to start their block comments before you fall out of the # chain

8:11 PM

I don't think fish can naivelt be combined with the existing shell answers

fish uses an entirely different syntax for case/switch statements

fish is very syntactically different from zsh and bash

Yeah, fish is a weird one as far as shells go

It is a lot of fun though. I've actually decided to switch from bash to fish for my everyday purposes.

tcsh and fish might play nice though.

8:37 PM

I used to use it a fair bit. Never switched entirely too it because I had too much in my .bashrc that I didn't feel like reworking, but I had 'fish Fridays' where I made myself use it so I'd slowly build up the tweaks I most needed as I needed them.

Hm...looks like fish throws an error even if you try to exit at the beginning of the program

Seems to try and read the whole file beforehand, that might be a problem

Yeah I think fish is out altogether

I am currently looking at ksh, but it has the opposite problem where it behaves to similarly to the other shell languages

9:01 PM

Ok I'm going to stop working on this because Its not my turn anyway.

lol @ Fish Fridays.

So I found something interesting with japt while looking for a lua entry point. When I put [ '] at the end of line 1, the javascript transpolation (seen here: ethproductions.github.io/japt) the cuts off at after line 1 as well.

we might be able to stop japt from parsing the whole polyglot, which would free up some {} use in the shell space.

I'm guessing this might allow ksh.

I'll poke at this more tomorow.

Is that a crash/error of some sort?

I know nothing of Japt, but the sudden change feels that way. I also kept discovering languages today that silently crash but still exit with a 0 and no error text, which is making me question that more

9:47 PM

@Chance japt's still parsing; it just seems like it doesn't transpile anything from an unmatched [ to the end of the program

and '] places the closing ] inside a (very short) string literal

this doesn't appear to be a crash of any sort

it also isn't documented, though

I think as long as we make sure that the rest of the program contains more [ than ] from Japt's point of view, most things will be OK

actually there's definitely a parser bug here

given that ] after a $ count, and they shouldn't

10:08 PM

looks like this is the direction to go.

#16 "(}23!@)(" 3//*v\D@;'[af2.qc]'#)"14";n4
#/*1-40//[']

pyth and 05AB1E seem like the main challenges.

assuming we're willing to exploit a parser bug.

we may need to find a version of Japt that predates the challenge to see if the issue is still there

also that second line still looks hostile to Lua

or is there some other basis behind this?

oh yeah lua is done

I'm thinking this will open up {} for ksh

my sense is there is an exploit with to upper/ lower string functions

also the C/C++ section get it's put line cleaned up without the {} concern

oh, the idea is just to make braces easier to use in the rest of the code?

I think so.

it'll clean up the B-F section as well I think.

something I really like about this polyglot is how easy it is to tell what part of the code a language is for, in most cases

often it'll contain a literal number and you can look it up in the rundown

and if not, it's normally fairly distinctive

the bits that split code away from each other are the only confusing bits

10:18 PM

I wonder how easy it would actually be to invent a new language to add.

that, and all the random Incident/Prelude filler

oh, I assume A Pear Tree can be added trivially, it was invented for polyglots

but it was inspired by this question

at least it made it into the test driver. :P

yeah, there's only like two or three things it does better than Perl

but one of them happened to be exactly what I needed

$ apeartree.pl polyglot.poly
a partridge

looks like it's still working :-D

you should sneak it into an answer with #0.

I put "and a partridge in A Pear Tree" at the end of the rundown sometimes

last time was on Dec 25

not sure when next to do it, it seems worthwhile to save for a special occasion

it was a fun language to write, it was a thought experiment on "what sort of useful language would print a partridge in response to almost any input program"

and it turned out to be pretty useful on PPCG, often winning radiation-hardening challenges

strangely I don't think it's won any polyglot challenges yet, but those are rare and too many of them care about byte count (it's fairly verbose)

10:25 PM

I though about posting a polyglot challenge a while back, but decided to wait until this one ended.

we could do with more polyglot challenges, assuming they're sufficiently different from each other

I was thinking of a cops and robbers thread where cops basically declare a number of languages impossible to use in a polyglot with one another. and robbers try to prove them wrong.

kinda of a loose idea still

I like that, actually; I've had similar ideas myself

however, it's a bit risky to run as cops-and-robbers, because if the cops are competent there'll be nothing for the robbers to do

many people dislike that, and require that all answers should be robbable, which makes it hard for the challenge to work

on the other hand, requiring a proof wouldn't work either, and in cases where the proof can be done, it's uninteresting

for cops, most languages win. and any 2 can be used by robbers.

so I'm not sure if we could easily fit it into the existing challenge types

ah right, sets of more than 2 languages

10:30 PM

I think mostly greedy cops will get taken down by robbers.

I suggest we do it like this: a cop submission is a set of languages, and it can be cracked by either a) a polyglot in any two of those languages, or b) a cop submission with a superset of those languages that also isn't cracked (i.e. stealing other people's sets and adding languages to them is encouraged)

I think there should also be a requirement that each of the languages is capable of having programs written in them individually

as there are plenty of esolangs that have zero useful functionality

(and the program-to-be-defined should probably take input and produce output, to help cut down on that sort of loophole)

good call. yeah, I like all of that.

I'm thinking of things like Baby Language

I can only think of one PPCG challenge offhand where that's usable at all

(and it postdates the challenge, IIRC; or maybe not, I should check that)

oh wow, it doesn't

I am so going to answer this