Polyglot Development

Wheat Wizard

00:00

I see why prelude is a problem

Chance

do you see where the prelude answer is produced?

Wheat Wizard

Yeah

in the python

wait hold up

Chance

It's this bit: 46(8+9+9+9+9+=!)

Wheat Wizard

ok

Thank god prelude has runtime output

Chance

this kind of needs to exist as far to the right as possible. and your brainflak code make it no longer on the lingest line

Wheat Wizard

00:06

ok

Do programs need to terminate?

Chance

yeah. crashing is a no go. :P

Wheat Wizard

ok

Chance

the debug menu in tio is super helpful for prelude

it can't have vertically aligned parentheses

Wheat Wizard

why?

Chance

but the debug will tell you which columns are broken... or maybe that's just in the regular output

oh yeah. that's regular output

Wheat Wizard

00:14

Whats the deal with brackets lining up?

I got the single line to work in prelude

Chance

it like reads all lines simultaneously or somthing

not totally sure.

Wheat Wizard

ok so just add spaces until it doesn't error?

Chance

I just know that it can't have more than 1 parenthesis per column

basically

it's a fiddly one to debug

Wheat Wizard

ok done

Chance

somtimes adding a space causes more misalignments.

oh? cool

Wheat Wizard

00:16

just added a space and it all worked out

Chance

beautiful.

Wheat Wizard

I'm sure I'm going to have to come back to this

Chance

yeah, incident will be a pain i think.

I'll check back in tomorow.

Wheat Wizard

Ok

thansk

Currently broken:

thutu
cubix

Unknown status

Deadfish~
Incident
SNUSP
Reng

The fact that cubix is broken deeply concerns me

Qwerp-Derp

01:07

I challenge someone to add a Lisp-like language into the polyglot :P

ais523

@Qwerp-Derp it'd be fairly easy if we could find a lisp-alike with # comments

the problem is getting past the first few lines

@WheatWizard cubix is another language that breaks on basically every submission

Wheat Wizard

Oh ok

its a size thing right?

I have not programmed in cubix alot

ais523

we're using the "capsule" technique (previously used for Hexagony, but we managed to give that a fixed location in the end); it's a bit of code that you move around the polyglot to get it into the position you need

basically you put the code into a cubix debugger, run it a bit until the IP is in some relatively harmless location

Wheat Wizard

Ok might not be that bad then

ais523

then put the capsule there

Wheat Wizard

01:26

I think I'm getting thutu

Wheat Wizard

01:44

@ais523 I'm a little confused as to how the current Thutu code is intended to work.

ais523

almost all of it does nothing

Wheat Wizard

Ok

ais523

let me figure out the relevant bits

Wheat Wizard

thanks

I think I found most of it

ais523

OK, it's the /=1/24=x

Wheat Wizard

01:46

/=1/24=x<$+@+-@@@@=>+<@@@=>+<?#d>+.--./

Oh ok

ais523

everything else is parsed but has no useful effect (e.g. if statements whose conditions don't match, setting values in memory that are never used, and so on)

Wheat Wizard

Ok

ais523

apart from lines starting with #, those aren't even parsed

Wheat Wizard

That looks like it should work though right?

ais523

and the way it actually executes is: the =1 ("start of program") is replaced with 24=x ("output 24") and a string literal that just gets written to memory

Wheat Wizard

01:47

it replaces =1 with 24=x which should output 24

ais523

then on the second iteration, =9 ("EOF") is input from standard input, and is then interpreted as the command =9 ("end the program")

so yes, that line should work by itself

Wheat Wizard

hm

ais523

most likely the problem is that before the /, you have something that's interpreted by Thutu as a regular expression

and if it, say, contains unbalanced parens, that part of the program is going to crash

also, the regular expression actually needs to match

Wheat Wizard

Oh yeah you can't see the code I'm looking at

ais523

there's a | just before the / in the current polyglot because most regular expressions ending with | match anything

but changes to the code might potentially cause the regex to not parse, or otherwise not match

Wheat Wizard

01:49

It is actually errorring on line 20 which I have not touched.

weird

Oh wait I am reading the error message wrong thats why

ok I think I can fix this

ais523

it may be referring to the line number in the compiled code, not the original code

Wheat Wizard

yeah I think it was

Chance

@Qwerp-Derp #45 PicoLisp. Boom.

Wheat Wizard

What does it mean by unrecognized row modifier?

ais523

nearly always that the end of the line has an invalid form

there are only a few things that can legally appear at the end of a line in a Thutu program

by far the most common is /

Wheat Wizard

01:54

ok I put the prelude code there so thats got to be the problem

Yep moving the prelude changed the error

ais523

simply moving the / to the end of the line, or adding a new one, will typically fix row modifier errors

Wheat Wizard

thanks

ais523

(there are other row modifiers but they're much trickier to use in a way that parses correctly)

Wheat Wizard

I'll keep that in minf

Does Thutu require me to escape all of my {?

ais523

they have to be balanced and contain a decimal number, or two decimal numbers separated by a comma

although come to think of it, a \Q earlier in the line may well escape the rest of the line

which would be really convenient if it worked

Wheat Wizard

01:58

Ok well Brain-flak has already ensured they are balanced and japt forced them to contain numbers.

what is going on with \Q

ais523

it's a quotemeta operator, it puts backslashes before things automatically

but I'm not entirely sure how it works

Wheat Wizard

ok fair enough

can't hurt to try

ais523

(note: this isn't part of Thutu's spec, it's an implementation detail of how the Thutu compiler does does regex)

Wheat Wizard

ok \Q does something but it doesn't seem to really fix anything

ais523

well it'd only matter for Thutu, and I'm not sure whether or not it'd do what we want there

Wheat Wizard

02:01

#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]>+.-- -. >][
#x%+>+=ttt Z_*.
#D>xU/-<+++L
#R+.----\).>]|
#[#[(}2}20l0v0x1k\4O6O@1kMoOMoOMoOMoOMOO0l0ix0jor0h0h1d111x0eU0yx0y0moO1d0y0e0e00m1d0i0fx0g0n0n11MoOMoOMoOMoOMoOMoOMoOMoOMoOMoOMoOMoOMoOmOoMOo0moo0n0tx0t0moO0f0t0gOOM0g0f0h0j0j0i0001k10vx0v0l111111^_)0

Chance

\ would escape the { for both japt and thutu. Might be easier.

Wheat Wizard

escaping them doesn't seem to actually fix it

ais523

thutu also won't like the unbalanced square or round brackets (although in regex, enclosing characters in square brackets tends to escape them in a different way, with the exception being hyphens)

Chance

What's a round bracket? (?

Wheat Wizard

Enclosing everything in a square bracket would be a rather bad idea, Classic really doesn't like things in square brackets

ais523

02:11

@Chance yes; I just picked the first description that came into my head

"parenthesis" would be more common and unambiguous, I think

although "round bracket" isn't unheard of

Wheat Wizard

Ok part of the problem was that perl's regex doesn't like [] so I had to put something in it. But I am still geting the really strange Unescaped left brace in regex is deprecated

ais523

ooh, if it's deprecated, that means it still works

so you might not have to fix it at all

that's just a warning, not an error

Wheat Wizard

oh good

Now I just have to get rid of nested quantifiers

Oh I get what it means by nested quatifiers

Thutu is done

:)

... and I've broken prelude

Thanks for the help on Thutu

prelude is back up

Wheat Wizard

02:35

Oh shit

I'm over the byte count

Oh wait I can't do math

I'm not

Qwerp-Derp

Does anyone on here know any Factor?

Wheat Wizard

How does one test Reng?

This Reng interpreter doesn't seem to work on my machine

Ok I figured it out by not being dumb

Reng works

Now I can't get SNUSP to work

am I stupid or are these online interpreters really confusing?

@Qwerp-Derp Well since I'm the only one left here, no

Ok I'm done for now. 4 languages are left: SNUSP, Deadfish~, Incident, and Cubix

Chance

04:00

@Qwerp-Derp sorry, no.

Chance

17:16

craop modular SNSUP doesn't work in my answer

Wheat Wizard

Oh ok thats why I couldn't get the interpreter to work

I was testing it on your answer

Chance

oh, no. actually it's been broken for a while

*gulp

oh wait... maybe im just running it wrong

ok, yeah it was me. crap. I broke snsup.

Wheat Wizard

So just the latest answer is broken?

How do you test SNUSP?

Chance

this

quirkster.com/iano/snusp/snusp-js.html

gotta click show then run

Wheat Wizard

ok

I think I was just hitting run

Chance

17:22

it's the $ in bash.

$ is a starting point for modular snsup

Wheat Wizard

yeah

I vaguely remember that

Chance

17:35

ok. fixed.

Wheat Wizard

20:25

Ok all thats left is Cubix, Incident and Deadfish~. Does anyone know how to test Incident or Deadfish~? I tried the link for Deadfish~ and the interpreter works odd

Chance

I've tested Incident manually. do you know how it creates it's tokens?

Wheat Wizard

No I know absolutely nothing about Incident right now

Chance

by manually, I mean using tools, to determine how it will be interpreted, and verifying that it is not functionally different from prior answers.

Wheat Wizard

yeah

Chance

You've run the test driver I assume

There is a section at the bottom fir incident.

Wheat Wizard

20:28

Oh cool thanks

how do I verify that the tokens do the same thing?

I assume if the tokens are the same I don't have to do anything

Chance

It's listing the tokens that will be found by the incident interpreter, and the way it finds it's tokens is weird.

True, but they will almost always be different.

Wheat Wizard

Ok I got my tokens

Chance

to quote ais523, a token is anything that appears exactly 3 times in the input, and isn't a subset of another token, and doesn't overlap another token.

Wheat Wizard

ok

Chance

so, for example, if the letter j appears exactly, that is a token.

the tokens derive meaning from their relative positions

the ^_ token is the jump token

jump tokens have the pattern of appearing once, and then twice consecutively.

Wheat Wizard

20:32

Ok I see two of that

Chance

does that make sense?

there are 3 total. one in the middle

Wheat Wizard

oh yeah there is

Chance

cool.

Wheat Wizard

so when it appears 3 times it makes 3 tokens not 1

I thought to make 3 tokens you would appear 9 times

Chance

I suppose you could say that, but the concept of toekns is a little above me in the first place. I kinda just know the rules to make this work.

the other jump token is }2

Wheat Wizard

20:34

yeah I have some of those too

ais523

ooh, conversation about Incident?

ask me if there's anything you're unclear on

Chance

so, everything between the sets of }2 ... }2}2 and ^_ ... ^_^_ are not executed

ais523

(it took me a while to get my head around the language, and it's my own language)

Wheat Wizard

well for me that is everything right now

except for tokenization

@Chance ok cool

ais523

the basic idea is that Incident looks for substrings that appear exactly three times in the input, and ignores everything else

also, the actual characters that make up the substrings are irrelevant, only the patterns in which they appear in the program

Chance

20:36

The center token is key to making Incident work.

ais523

this also means that if two potential tokens overlap, neither of them counts (because otherwise you couldn't tell what order they're in, = you couldn't figure out what pattern they were in)

the syntax highlighter uses brown underlines for disqualified tokens (and cyan underlines at the point of the overlaps)

Wheat Wizard

Here are my tokens for anyone that wants a look

I split them by 10s

#60 has a space after the echo

Chance

To ensure the same token is in the center as in the last answer I have followed the following rule: The number of tokens between the ^_ tokens minus the number of tokens between the }2 tokens must equal 7.

ais523

right, the centre token in the program is the one that does I/O

Wheat Wizard

ok

ais523

20:39

so if a different program reaches the centre, the program's control flow will be the same but the output will be different

in order to get the output 33, you need the token currently spelled as 0mo as the centre token

Wheat Wizard

@Chance That is not the case

currently I have 19 as the difference

ais523

and if there are an even number of tokens (like there are here), the token immediately before the centre point is the one that matters, at least according to the docs

@Chance well there are three 0mo tokens, and it doesn't matter which of the three is in the centre

Wheat Wizard

I need 0mo to be in the centre?

Chance

@ais523 true. I've just always used the rule of 7. :)

that's right

Wheat Wizard

ok well currently my center is 0f

Chance

20:43

so, looking at your tokens you have (p

ais523

you can normally recentre it by adding or removing tokens

Chance

I'm guessing you tokenized this in the beginning of the bash script

Wheat Wizard

yeah

Chance

perhaps if you swap these characters, since I don't believe the order matters, you can detokenize (p

ais523

the golfiest way is normally to make one of the existing tokens overlap, or not overlap, but putting a space or x or something inside one of the copies a token, to remove all three copies, is normally easiest

Wheat Wizard

20:44

I'll change ((pp to (p(p that way (p will appear 4 times

Chance

you're getting it.

Wheat Wizard

0mo is now centered

ais523

yep, that's the sort of trick we use

btw, what operating system are you on?

Chance

wow really?! Why is it never that easy for me???

Wheat Wizard

I'm on OSX right now but I have a Arch partition

and access to debian and ubuntu

ais523

20:46

the official Incident interpreter and syntax highlighter have a dependency on libdivsufsort but otherwise run fine on most computers with a working C toolchain

libdivsufsort is in at least the Ubuntu repos, it wouldn't surprise me if it were in some of the other repos too

and having the syntax highlighter and an interpreter handy makes working on the polyglot easier IMO

Wheat Wizard

Ok libdovsufsort is available for OSX through brew, but my C compiler is ancient

ais523

pre-C99 ancient?

Incident's written in C11 but I'm not sure I used any features that aren't in C99

Wheat Wizard

post C99 pre C11

ais523

you might as well try the tarball then: nethack4.org/esolangs/calesyta-2016.tar.gz (you'll need to change "c11" to "c99" in the Makefile)

2

Wheat Wizard

wait what

oh I see

99 < 11

Ok it really doesn't like the use of u8

ais523

20:59

oh right, it wouldn't

let me see how fixable that is

it's just in the debug output

Wheat Wizard

I switch OS's so that I can compile with C11

ais523

also I think gcc might use UTF-8 by default? in which case you could remove all the u8 prefixes and it'd still work

Wheat Wizard

Ok what exactly am I building rn? I am going with the assumption it is the Incident interpreter.

ais523

the makefile builds two programs

incident is an Incident interpreter, incilite is an Incident syntax highlighter

Wheat Wizard

ok cool

ais523

21:03

(use the -V argument to incilite if you want to highlight it directly in your terminal, rather than output HTML)

Wheat Wizard

ok libdivsufsort is not available for debian so I will be moving back to OSX

ok removing u8 works!

Ok I've got my syntax highlighted now what does any of this mean?

ais523

OK, so the bright colours are tokens

Wheat Wizard

ok

ais523

the dark underlined text are rejected tokens (things that appear three times but are rejected due to an overlap)

Wheat Wizard

I'm going to have to change my terminal colors then

ais523

21:09

the bright underlined text is the centre token (all three copies are underlined, IIRC)

oh right, OS X

bright colours are also boldened

but that might not be different enough

Wheat Wizard

Ok done

ais523

you can just run your program through incident to see if it works; if not, the syntax highlighting can help you correct it

but any underlining, except on 0mo, in the Incident "section" of the polyglot normally implies that you've introduced a token that screws up the existing code

Wheat Wizard

How do I pass the program to incident?

ais523

save in a file, give the filename on the command line

Wheat Wizard

ok like everything is underlined

hey it prints 33!

ais523

21:12

in that case it's working :-)

most of the underlines are in parts of the program that don't matter

some of them we introduced intentionally

Wheat Wizard

I also think that there is something wrong with the way my terminal is rendering ansii escapes

ais523

the actual main body of the Incident code is just from the third }2 to the first ^_, anything outside that underlining doesn't really matter other than to recenter the token

which terminal are you using?

Wheat Wizard

I'm using a slightly modified version of iTerm2

ais523

I have some terminal test files handy that I created for NetHack, I can help you do a terminal test if you like

iTerm2 is pretty good nowadays though IIRC

Wheat Wizard

nah it prints 33 so I'm all good

ais523

21:15

:-)

when the program actually works, there's no real reason to debug it

Wheat Wizard

ok all that remains is Deadfish~ and cubix

and then maybe everything I've broken fixing those two

ais523

the deadfish family of languages is… interesting

they're incredibly weak, and incredibly easy to implement

Wheat Wizard

yeah

ais523

deadfish~'s interpreter is here: github.com/craigmbooth/deadfish

Wheat Wizard

I'm surpised you guys have not used more of them

ais523

21:17

you have to give it the program via a redirection

oh, most deadfish interpreters do at least some error checking

Wheat Wizard

I have that but I haven't been able to get it to work

ais523

and won't even get past the leading #

Wheat Wizard

what is redirection

ais523

python deadfish.py < polyglot.deadfish

Wheat Wizard

oh ok

I've done that before

ais523

21:18

right, just didn't know what it was called?

actually, one of the great things about deadfish~ is that it introduces control flow, but is still incapable of even something as simple as an infinite loop

Wheat Wizard

infinite loops are overrated imo

just makes it hard to debug

48

ais523

yay

Wheat Wizard

that leaves just Cubix

ais523

unless you hardly added anything to the program, the odds of Cubix working first time are very low

Chance

I'm reaching for the champagne bottle.

ais523

21:21

you see the \4O6O@ in the last couple of programs?

that's the Cubix capsule

Wheat Wizard

err I don't see it

ais523

the basic idea is that you run the program you have in the Cubix debugger (it's linked from all the last few answers)

single-stepping

#[#[(}2}20l0v0x1k\4O6O@

there

err, first you remove the capsule, then run the program

Wheat Wizard

wait I don't have a capsule right now

ais523

when the Cubix control flow reaches a point which can easily be modified (typically the Incident code, but the evil code is also a good place), you put the capsule there

ah, OK

you can create a new one then

it's basically just 4060@ in the current execution direction; if that direction happens to be up or down, you reflect it with a mirror \ or / to make it left or right instead

Wheat Wizard

ok

where on the cube does execution start?

ais523

21:26

the top left corner of the leftmost cube face

going right

Wheat Wizard

cool

ais523

however, it sometimes feels like half of ASCII is control flow commands in Cubix (it's actually much less than that, but it has a lot of control flow commands)

so the IP tends to fly off randomly into space

Wheat Wizard

ok

ais523

and using a debugger to figure out when it ends up somewhere useful is recommended

Wheat Wizard

debugger?

Chance

21:27

ethproductions.github.io/cubix

ais523

right, that one

run it in single-step mode

Wheat Wizard

ok it gets stuck in a loop very quickly from where it starts

I have a limited amount of space to put the 4060@

ais523

if there's no room you may have to add an extra character somewhere on the top face in order to change the alignment of the code

to make the IP more cooperative

Wheat Wizard

hold up there is a 4060@

it just doesn't appear when I do a search?

ais523

it's 4 letter-O 6 letter-O @

there are no digit zeroes

Wheat Wizard

21:30

oh

thats why

ais523

if you're using a font where the two look similar it can be confusing

I hadn't even realised that was the problem until I saw a zero in your comment, the two look quite different in this font

Wheat Wizard

yeah they look different for me, I can tell that 0s are not Os but I don't realize that Os are not 0s especially when they are next to other numbers

Ok fixed cubix broke prelude

ais523

prelude's gotten more and more problematic over time

because as the program gains more lines with parentheses on, there's more chance of them aligning vertically and screwing things up

Wheat Wizard

fixed

ais523

oh good, the thing with Prelude is that sometimes fixing it is trivial, and sometimes it's nightmarish

when I added Bash I had to completely redo the Prelude control flow in the left half of the program

Wheat Wizard

21:38

ok whirl is broken

what is whirl

ais523

whirl only looks at 0 and 1

and ignores all other characters

Wheat Wizard

oh yeah I broke that

ais523

I recommend using other digits unless you badly need a 0 or 1 for some reason

Wheat Wizard

yeah

ais523

and if you do need them, placing them late in the program is likely to make them a no-op

Wheat Wizard

21:39

I actually deleted a zero I thought I added earlier

but it must have been one of whirl's zeros

I'll avoid 1's and zeros

I'm surprised whirl worked at all, I've added a lot of 1s and 0s

Ok everything on TIO works

ais523

IIRC the existing Whirl code ends with some sort of stabliser to make future commands not have much of an effect?

Wheat Wizard

thats good

ais523

quite a few languages do that

Underload used to, then someone golfed it off, it's been breaking more often since then

Wheat Wizard

SNUSP is working

Reng works

Deadfish~ works

Incident does not work

ais523

run it through the highlighter to see what's wrong

most common issue is destablising the centre token, second most common issue is accidentally detokenising something inside the main body of the Incident code

Chance

21:45

whirl only executes commands when a double zero is encountered, which is not common.

Wheat Wizard

I added a bunch of new tokens apparently

ais523

ah, that's what it is

Wheat Wizard

Is there a good character I should use as a spacer?

like a character that most programs don't care about that I can use to move parens and break up tokens and such?

ais523

x is the standard

Wheat Wizard

ok

I ended up being able to move something

Unfortunately I need to go to class now hopefully I can finish this later tonight (EST).

Chance

21:52

I'm so excited to read the explanation

Wheat Wizard

22:02

Ok I got 0mo in the center but it still doesn't work

ais523

compare it to a syntax highlighting of a version that works

that might show you what you broke

Chance

22:14

perhaps you placed the cubix capsule inside a token? It wouldn't be the first time.

Wheat Wizard

23:57

@Chance No moving cubix added tokens

I think it was during the fixing of prelude

Transcript for

Polyglot Development