The Nineteenth Byte

Erik the Outgolfer

00:21

@BMO um... why doesn't this work?

New Main Posts

0

Q: A well linked challenge

An activity I sometimes do when I'm bored is to write a couple of characters in matching pairs. I then draw lines (over the tops never below) to connect these characters. For example I might write \$abcbac\$ and then I would draw the lines as: Or I might write \$abbcac\$ Once I've drawn t...

Ørjan Johansen

00:53

@TRITICIMAGVS Mwahahaha

Wait I'm stupid

feersum

@NewMainPosts Does this not just reduced to whether there's 2 of the same character in a row?

Ørjan Johansen

@feersum No.

abcbca, for example

(As implied by the hint. Although I still haven't completely convinced myself the hint is true.)

feersum

It's not too clear to me what the hint is supposed to mean.

Ørjan Johansen

Well like in my example, bcbc is such a substring. You can connect it internally, and then all the connections elsewhere (just between a and a in this case) can just pass over it.

So one direction of the hint is clear.

feersum

Yes, instead of "a loop can be drawn over it" it would make more sense if it said that no external loops need to intersect it.

Ørjan Johansen

01:05

I think you mean lines, not loops.

feersum

Indeed, I was using "loop" as a synonym for line.

Ørjan Johansen

but they're different things here

feersum

That's why I didn't understand the hint.

I instantly forgot about loops because they seemed irrelevant to the problem.

Ørjan Johansen

Oh the hint is true actually. What I've not entirely convinced myself about is that you can assume the substring with a loop must be contiguous.

Hm but if it's not, then a line must pass over between two of its parts. Which means that line isolates a contiguous substring between them.

feersum

Any two valid loops could be joined into one, I suppose.

Ørjan Johansen

01:15

No wait, a line must pass over unless everything's even on one side

Neil

01:25

@Deadcode ah, nice prime test check... with that, I can do it in 61 bytes (including decimal conversion and balancing groups): ^((1{14})1(1)+)(?<=(?<!^\4+(..+))\2?)(?<-3>\1)+$(?(3)1)

@ASCII-only is it me or is the astifier broken right now? I haven't been able to use -a at all today

Ørjan Johansen

OK I convinced myself. If the looped region has more than one contiguous substring, and B and D are the two leftmost components, with C the contiguous substring between them, then either C is even and so can be looped itself, or it as an outward connection separating B from D, so B can be looped alone.

So it is enough to look at loops around contiguous substrings.

ASCII-only

@Neil :/

well i haven't updated recently

so i guess it's just something i must have missed

@ØrjanJohansen :| i see what you did there

Deadcode

@Neil Cool, but how can \3 ever be unset in that? It's (x)+ and it's inside a {1} group, so it can never be unset, no? Or does the balancing group construction unset it? (I still have a very limited understanding of balancing groups)

ASCII-only

@Deadcode yeah the -3 pops the most recent \3 iirc?

Neil

01:46

right, each time I match the \1 I pop a \3, and I want there to be none left when I finish matching

.net allows multiple named capturing groups with the same name, which, combined with balancing groups, makes matching balanced constructs very easy

something like ($(?=.*(?$))|\[(?=.*(?]))|\{(?=.*(?\}))|<(?=.*(?>))|(?<-b>\k))‌+ (I may have misremembered the syntax)

each time you hit an open bracket you peek ahead for a close bracket so it's on the balancing group ready to be matched and popped when you run out of open brackets

Deadcode

02:07

@Neil Does that mean it could be an actually Accept-able answer to my SO question Is it possible to erase a capture group that has already matched, making it non-participating??

ASCII-only

02:28

@Deadcode yes

iirc balancing groups is .NET's answer to PCRE's recursion or something like that

might want to look into .NET regex, it's basically about as powerful as PCRE

Deadcode

Well, if you write it up nicely as an answer to that question I'll give it at least 150 bounty.

Deadcode

02:49

@Neil Are you going to edit your answer? Also, how is this not 60 bytes? (55 + 5 for decimal to unary)

@Neil And also, how does it have anything to do with my prime check? I don't see any resemblance.

ASCII-only

04:15

hmm

how bad of an idea would it be to have macros with different numbers of args based on their argument?

e.g. new takes 1 argument (type) -> new foo, which then becomes a function that takes as many arguments as the type's constructor does (well, because it is the constructor syntax)

TRITICIMAGVS

05:21

@ØrjanJohansen I think my hint was wrong or at least not worded very clearly. I've rewritten it to be right. (Also your thought process here is pretty much the exact one I went through :))

ASCII-only

06:52

wow

how tf are there so many progamming languages being designed

Esolanging Fruit

07:53

@ASCII-only Ah yes, r/ProgrammingLanguages.

The subreddit for discussion of random hobby projects that no one cares about, along with confused people from r/programming.

At least it doesn't get as many off-topic posts as r/rust.

flawr

09:18

@flawr We failed to celebrate TNB's 8th birthday D:

ASCII-only

09:49

@ØrjanJohansen coc?

ASCII-only

10:04

oh

ASCII-only

10:33

if anyone's still here: in an ideal language, what should compare_to return?

an enum value of either less, equal, or greater? if so, what to name the enum?

ais523

@Deadcode thanks for the information; this in turn implies that a backreference is always a substring of the string being matched, which looks like a fact that would be useful in writing optimized regex engines that handle backreferences (which is what I was interested in)

@ASCII-only the normal name seems to be something along the lines of "ordering"

ASCII-only

hmm. i was going to go for comparison_result - not entirely sure ordering describes it well/is a good word to describe it

also... would simple enums (just the values) or java-style enums be better

ais523

Java-style enums are broken, at least the way Java does it

ASCII-only

assume not the way java does it :P

Neil

@Deadcode I wrote up stackoverflow.com/a/54502001/182705 as nicely as I could.

ASCII-only

10:40

@ais523 i guess i mean... enums that contain more than just a list of names

Neil

@Deadcode it was 61 because I was using Retina 0.8.2 at the time and that takes 6 bytes for decimal to unary

ais523

do something like Rust or C#, where you can attach methods to arbitrary classes after the fact (but the methods are in the namespace of the person attaching them)

that way you can have all your enums just being enums without having to work out all the functionality they'd need in advance

ASCII-only

@ais523 in c# enums aren't classes :P

ais523

can you attach methods to them anyway? if not, that seems wrong

(rust lets you attach methods to primitives, but it's not monkeypatching because the methods are in your own namespace and so they won't interfere with anyone who's not trying to use them)

ASCII-only

@ais523 extension method in C#? yes

also well... actually in this language every method is just essentially a global function in disguise >_>

would there me a nice alternative to java's attaching values to the enum members? hmm

as in Day.THURSDAY could have "Thursday" as a member in Java (so day.getName() would return the string the value was constructed with)

Neil

10:45

@Deadcode it was the way you reinterpreted 14) equals n> as is prime and p times (p plus 14) equals n>

ASCII-only

well actually. what are the advantages/disadvantages of sum types and enums?

Deadcode

@Neil Thank you :)

I tested it and it works.

I think in most circumstances there would only be one to delete at a time.

Neil

yeah, well I wasn't sure with your crazy regexes ;-)

Deadcode

11:33

@ais523 You want to write a regex engine? Did you look at mine?

@Neil I'm actually curious about that – why did you stick with 0.8.2 for as long as you did?

Neil

@Deadcode I've got it bookmarked on tio, haven't got around to updating

I only use Retina 1 when I need a feature 0.8.2 doesn't have

Deadcode

The 1 byte savings for decimal-to-unary seems like reason enough :)

Unless you want the scores of new Retina programs to be directly comparable with those of older ones

Erik the Outgolfer

14:18

@BMO so... regarding the last ping, I discovered that rules don't work at all

Erik the Outgolfer

16:11

also... FR: make xA + 0A mean that there must be no more than x A, but there must be x A

MilkyWay90

17:00

can somebody teach me how .rule files work?

New Main Posts

0

Q: A Turtle Finds a Portal

A Turtle Finds a Portal The turtle wants to move along the grid to get to his food. He wants to know how many moves it will take for him to get there. As well since he is slow he has teleporters set up around his domain that he will utilize if it shortens his path. Or avoid them if it lengthen...

code-golf grid

New Sandboxed Posts

21:58

0

A: Sandbox for Proposed Challenges

How much of the infinite board can I reach? code-golf grid The board is an infinite 2 dimensional square grid, like a limitless chess board. A piece with value N (an N-mover) can move to any square that is a distance of exactly the square root of N from its current square (measured centre to ...

ASCII-only

22:10

@MilkyWay90 :| read golly docs pls

@EriktheOutgolfer excuse me...

@EriktheOutgolfer magic...

@EriktheOutgolfer no thanks. that would make most things trivial... plus that is not how chemistry works (well, even less like how chemistry works). it's an esoteric language, not a non-esoteric language like ><> is

Erik the Outgolfer

0A in the left part means that there must be no As for the rule to be applicable

ASCII-only

yes

it is one of the few ways to get conditionals

otherwise the language wouldn't even be TC

Erik the Outgolfer

however, 2A+0A currently doesn't have a meaning

so I'm suggesting that

ASCII-only

@EriktheOutgolfer it does...

it's 2A...

atoms are combined as an optimization

Erik the Outgolfer

oh huh didn't notice

ASCII-only

22:14

but yeah. this is what 0A does

2A + 0A is well... just 2A >_>

@EriktheOutgolfer btw i've half done a more optimized version of alchemist

in case you might have been looking to do very long deterministic divisions :P

Erik the Outgolfer

lol I've barely made a fibonacci as of yet

ASCII-only

haha

well, BMO's method of keeping a state atom makes it a whole lot easier. and easy to keep deterministic

state atom being an atom that's always there to indicate which procedure you're executing rn. like m = division -> n = comparison -> o = reset -> m (switching states being like m + 0endcondition -> n)

Erik the Outgolfer

22:32

@ASCII-only well... with state atoms, it quickly gets messy... :P

ASCII-only

@EriktheOutgolfer how

Erik the Outgolfer

0A, 0B, 0C, ...

ASCII-only

hmm?

Erik the Outgolfer

don't you have to include a lot of zeroes in a lot of rules?

ASCII-only

no like... at the end of a division it's just m+0x->n

@EriktheOutgolfer never

generally, at most twice per state. at most

if you have more... 99.99% chance you're doing something wrong. no offense

check the existing answers (note: i use bmo's method, jo king just does it without a structure. it's golfier, but you gotta make sure you know what you're doing since it's not deterministic)

Erik the Outgolfer

22:36

so you're saying something like detecting the end condition of a step and changing the state atom then?

also... if it's non-deterministic, then you have to be sure the result is unaffected

ASCII-only

@EriktheOutgolfer exactly

@EriktheOutgolfer yes

@EriktheOutgolfer i guess... think of it (0a vs a) like pattern matching

oh yeah, anyone here have an opinion on sum types vs enums?

Rogem

23:19

I'm kind of curious, would it be permitted to gzip my answers in C, and state the language as gzip | gcc? This seems to consistently shed about 1/6 from my golfs...

ASCII-only

@Rogem gzip isn't a language, you'd have to use bash to un-gzip it

but... most importantly, it's not readable. which is really bad

Rogem

gcc will happily compile from stdin, so technically piping gzip -c to it is an extended compiler?

And I wouldn't call most machine code nor golfing languages readable either.

ASCII-only

@Rogem oi

golfing languages are as readable as C

it's just a matter of learning function names...

trichoplax

@Rogem Depends on your motivation. If you just want to get a lower score by switching language, there are languages shorter still. If you want the extra challenge of running all your code through gzip to see which is shortest (the shortest C code won't always give the shortest gzipped output) then I expect you could argue it's valid (even if you still get the odd downvote)

Rogem

Yeah, get your point. But, we also allow unary languages, which almost definitely aren't readable.

trichoplax

23:25

Related meta discussion: codegolf.meta.stackexchange.com/questions/17185/…

(I would say that justifies posting gzipped code, but I don't know how it would be received)

ASCII-only

@trichoplax yep. especially given the fact that it's just compressed C code

i.e. zero-effort golfing

Rogem

Oi

trichoplax

It's only zero effort if you're the only one using it. As soon as you have a competitor in the same "language", you'll find they beat you if you only golf the C first then compress it

Rogem

Finding the exact right types that UB in just the right way to shave off a byte from your 300 byte cat golf is an art.

trichoplax

Being horribly difficult to work with doesn't guarantee upvotes though

Rogem

23:30

I know, alright

ASCII-only

holy crap charcoal has 100 stars

trichoplax

@Rogem That wasn't meant to mean you won't get upvotes (or that upvotes would necessarily be your motivation). Certainly wouldn't be the most awkward language used here, and I still can't guess in advance what will be popular...

Rogem

@trichoplax meant that there was just a little bit irony in the statement (in the context of C being difficult to golf).

trichoplax

Oh I see. My expectation is that golfing C would be much easier than finding the C code that gives the shortest gzip output

For example you might find that unrolling a loop and making the code longer results in more repetition and a shorter gzip output

Rogem

Yep. And variable reuse would be especially important. And possibly stdint.h, which right now goes completely unused...

Another thing I still need to try is packing the ASCII...

Because 8-bit bytes are overrated.

Quintec

23:53

Yeah, make 64 bits a byte instead

Neil

well, I came up with a regex restricted-source answer... that was tough!

Rogem

Well... gcc cries murder when there's a non-ASCII in the code, so I was just planning on packing 8 chars into 56 bits.

Transcript for