The Nineteenth Byte

3:01 AM

@ChrisJester-Young personally I'm starting to gravitate towards a statically typed compiled programming language, but with full macros and type inferrence support

Chris Jester-Young

@orlp You mean like Nemerle?

orlp

never heard of it

Alex A.

Nemerle

Nemerle is a general-purpose high-level statically typed programming language designed for platforms using the Common Language Infrastructure (.NET/Mono). It offers functional, object-oriented and imperative features. It has a simple C#-like syntax and a powerful metaprogramming system. In June 2012, the core developers of Nemerle were hired by the Czech software development company JetBrains. The team is focusing on the development of Nitra; a framework for implementing new and existing programming languages. This framework will likely be used to create future versions of Nemerle. The language...

orlp

on top of that I have another idea that I don't really see around

Alex A.

Curry on pancakes? Because that sounds weird

orlp

3:02 AM

I like functional programming, but only at a function level

Chris Jester-Young

The more I hear your ideas, the more I think you'll love Nemerle.

orlp

my idea was to mark functions as 'pure'

Chris Jester-Young

GCC has extensions for doing that.

orlp

meaning that given the same input they'll always return the same output

any function using only other pure functions is also pure

Chris Jester-Young

Granted, GCC extensions are for C and C++ only, but still.

orlp

3:04 AM

C++11 could've done this with constexpr but they made it totally backwards

Alex A.

Fortran 90+ (maybe 95+) has pure function

orlp

@ChrisJester-Young maybe I should rather say impure

pure would be the default

Chris Jester-Young

I see, impure as the annotation.

feersum

I always wonder how you are supposed to debug in Haskell, with the default pure functions like that

If you can't put a random print statement in smoewhere

orlp

the point is, pure implies constexpr, allows all kinds of optimizations / advantages that functional programming languages enjoy, but without being 'locked in' so to speak

fundamentally I like some advantages that func. programming has, but I personally believe in the von Neumann model

PhiNotPi

3:07 AM

So, in my language, lambdas are going to be implemented as a beautiful mix of Polish and Reverse Polish notation, so that everything can be done stack-based.

feersum

Sounds very readable.

orlp

I've been thinking about making a Haskell backend for Pyth5

however I don't know enough about Monads and stuff to know if it's possible

Alex A.

@PhiNotPi It'll be just like the border between Poland and Reverse Poland.

feersum

In OCaml, you can put side effects wherever the hell you want.

Alex A.

OCaml is chaos.

PhiNotPi

3:10 AM

So, if the normal way to perform some operations is A|+B* (array A, reduced with addition and then multiplied by B), then the lambda form will be *Y|+X

Doorknob

@Calvin'sHobbies Is that gzipped or not? I'd be willing to bet the percentage is at most half of that when taking into account gzip.

PhiNotPi

The | reduce is a modifier that has to be pushed on the stack before the + is parsed and added onto the lambda, so that, in the process of creating lambdas, the interpreter will know what modifiers are applied to each operator ahead of time. The * is regular Polish notation.

An equivalent way to write the lambda would be *|Y+X

The | modifies the first thing it can, which isn't the Y and is instead the + like before.

|*Y+X would be invalid though.

orlp

@ChrisJester-Young #!fold-case is an abomination :(

Chris Jester-Young

@orlp It is, except that it's the default behaviour in R5RS so it's required to be supported for compatibility with programs written from the R5RS era.

(That is to say, R5RS has case-insensitive symbols, and R6RS and R7RS have case-sensitive symbols by default.)

Uh, that's not strictly correct. R5RS symbols are case-sensitive too, but the R5RS reader case-folds any bare datums it reads.

orlp

@ChrisJester-Young I also do not like the way #t and #f are notated

Chris Jester-Young

3:21 AM

Why not? Then t and f are just readtable entries.

orlp

I also really like implicit conversion of bools to 0/1

Chris Jester-Young

In the same way that vectors are #(...) (so you can have a ( readtable entry), bytevectors are #u8(...) (unsigned 8-bit), which also allows you to have a u readtable entry (and s if you want to support signed).

orlp

I don't know what readtable is unfortunately :(

Chris Jester-Young

A readtable is a table that the reader uses to handle anything following a #.

orlp

@ChrisJester-Young if I had to invent a language, I'd have types such as i64 (64-bit int), u8 (unsigned 8-bit int), and bool would act as if it's u1, except having a bit more constructors

(there would be no type called u1 though)

Chris Jester-Young

3:25 AM

Of course, it's only worth having 8, 16, 32, and 64 (and maybe 128).

orlp

my ideal type system

has i/u8, 16, 32, 64, Int (arbitrary width), Rational, f32/64, bool

not sure how I would handle complex

Doorknob

@orlp Automatic bool-to-int coercion is... confusing at best. I've never seen the need, except in golf.

It's better to keep the types completely separate IMO.

Ypnypn

What's the point of so many int types?

orlp

@Doorknob It's incredibly useful, I use it a lot

Doorknob

@orlp For what?

@Ypnypn Why allocate 64 bits when you only need 8?

Ypnypn

3:28 AM

@Doorknob Yeah, but how often do you use 16 or 32?

orlp

@Doorknob binary flags, [1, 2, 3] + [4, 5, 6] * condition, counter += condition, etc

Doorknob

@Ypnypn When they're convenient. (And then they're super nice to have.)

orlp

@Ypnypn when you know you only need 16 bits

there would also be some template type, not sure of the name/or syntax, that would have arbitrary bit width

e.g. i64 = I<64>

but you could use I<31> if you'd please

Doorknob

@orlp That's needlessly confusing. Why not just if (condition) arr += [4, 5, 6] or if (condition) counter++? Seems much less... obfuscated.

orlp

@Doorknob I find it not confusing at all

Doorknob

3:31 AM

I think the if versions are much more readable.

They make much more intuitive sense.

orlp

I disagree, especially not in bigger expressions

Chris Jester-Young

I totally get using 0 for false and 1 for true, especially for assembly program (think SETcc in x86). But, meh.

Alex A.

@orlp Use Julia. :D

Doorknob

@orlp In any case, what you're really doing with counter += condition is "increment counter if condition," so why not write it that way instead of hiding it behind a coercion?

feersum

Array indexing would be a more annoying situation to not to have bool/int conversion.

Doorknob

3:32 AM

And for bigger expressions, ternaries also exist.

feersum

arr[cond]

Ypnypn

TI-Basic is also fun with this: 1>0<1 is false

Chris Jester-Young

@Doorknob Again, in assembly, being able to use SETcc instead of Jcc is faster, but that's a microoptimisation.

Doorknob

@feersum In what situation would you use that? The only one I can think of is if arr is const, in which case you can just use a ternary

Chris Jester-Young

e.g., SETZ AL == set AL to 1 if zero flag is set, otherwise to 0. JZ label == jump to label if zero flag is set. Conditional jumps are expensive.

(Well, less expensive if branch prediction is correct, but.)

feersum

3:34 AM

I'm sure I've used it, but can't remember one right now

Alex A.

@Doorknob In R, that syntax performs array subsetting.

Doorknob

@AlexA. What's... that?

feersum

Something I do a lot more is if(x & 1) to test for an odd number

Alex A.

@Doorknob What's R or what's array subsetting?

Chris Jester-Young

@AlexA. Surely the latter. Surely everyone knows what R is?

Alex A.

3:35 AM

Idk, just making sure

Doorknob

@feersum That's firmly in the "unnecessarily unreadable" realm for me. Especially considering a compiler is likely to optimize % 2 == 1 into that anyway.

@AlexA. Of course I know what R is. :P

feersum

@Doorknob What are you doing on Code Golf with your great love for verbose code :P

Alex A.

@Doorknob In R, you can select a piece of an array based on some condition. So if you have x <- c(1, 2, 3, 4), you can get the elements greater than 2 using x[x > 2], which returns the array c(3, 4).

Doorknob

@AlexA. Oh. That's not indexing with a bool though.

@feersum "Verbose"? Have you heard of such things as "readable code"? :P

Alex A.

@Doorknob x > 2 creates a bool vector

Chris Jester-Young

3:37 AM

@AlexA. x.select {|x| x > 2}

Doorknob

@AlexA. That's still not arr[true] or arr[false]

Chris Jester-Young

@AlexA. Ah yes, I remember that.

feersum

I find x & 1 more readable because there are much less words to parse and it is still obvious.

Alex A.

@Doorknob arr[true] in R would return the whole vector and arr[false] in R would return an empty vector.

Doorknob

Yes, but that's not what we were talking about.

Chris Jester-Young

3:39 AM

Anyway, I'm going to maximise my Apple event video now, so, I won't be reading this chatroom. :-P

orlp

ok this is a bit arbitrary but not an unfair example

def statistic(l):
    cnt = 0
    for e in l:
        if e % 2 == 1:
            cnt += 1
        if e < 0:
            cnt += 1
    return cnt

def statistic(l):
    return sum((e % 2) + (e < 0) for e in l)

Alex A.

@Doorknob I'm just saying what R does with that syntax since you asked feersum what one would do with that syntax.

I rarely have any idea what folks are talking about here because it's mostly over my head, but I try to contribute where I can. :P

orlp

I like the second version a lot more

Alex A.

(usually ends up being off topic)

orlp

maybe I'm weird like that

Alex A.

3:40 AM

I like the second version as well

Doorknob

@orlp ? 1 : 0 makes it much more clear what the intention is ("add one if it's negative"). And it makes it much easier to change if you want to add, say, 2 if it's negative ((e < 0) * 2 is much less readable than (e < 0) ? 2 : 0). Or if you want to add 2 if it's negative, 4 otherwise.

@AlexA. No, I asked him what you would do with indexing an array with a bool. The syntax doesn't really matter.

Alex A.

@Doorknob Okay, I apologize for interrupting, I misunderstood.

orlp

def statistic(l):
    return sum((1 if e % 2 == 1 else 0) + (1 if e < 0 else 0) for e in l)

sorry, I still like my version better

feersum

haha

Doorknob

Well, that's Python's fault, for having horrible horrible ternary syntax.

Even so, I'd argue that's much more intuitive than your version.

orlp

3:44 AM

def statistic(l):
    return sum((e % 2 ? 1 : 0) + (e < 0 ? 1 : 0) for e in l)

I still like my version better

even assuming if we had this ternary syntax

feersum

@orlp You forgot to put e % 2 != 0 !

Doorknob

The first ternary is unnecessary. And if you removed it, I would definitely prefer the second version.

orlp

@feersum well there you go

Doorknob

It makes much more sense, which is well worth the what, extra 8 or so chars?

orlp

@Doorknob I disagree

if you get in the mindset of bools == ints, then it's 8 extra chars to parse

and occam's razor applies

(e < 0) in my mind is a 0 or a 1

a[e < 0] in my mind is the first or the second element

Doorknob

3:47 AM

But bools aren't ints, and that mindset breaks down because you can't take Math.sin(aBool) or what have you.

orlp

a[e < 0:] either is the entire array, or shaves off the first characters

Doorknob

(Well, you could, but it wouldn't make much sense.)

orlp

@Doorknob Math.sin(aBool) is as nonsensical as Math.sin(1)

Alex A.

@orlp I'm used to programming in SAS, which only has two data types (neither of which are bools), so this is often how I think as well.

Doorknob

Have to go to sleep now.

Alex A.

3:47 AM

Goodnight!

orlp

that's just because sin makes no sense on integers

not because bools make no sense as integers

orlp

4:08 AM

a = array (e.g. [1, 2, 3])
s = string
i = integer (e.g. 5)
b = bool (e.g. (e < 0))
f = variable containing boolean flag

Falsey:
    [] (empty array)
    "" (empty string)
    0 (zero integer)

First or second element: a[b]
Full array or everything but first: a[b:]
Empty array or full array: a * b
Same array or cartesian product of itself: a ** b    (not in Python, is in Pyth)
Same integer or added one: i + b
Zero or unchanged: i * b
One or unchanged: i ** b
Toggling flag: f = 1 - f
Or: f |= b
And: f &= b

some examples / techniques with integer booleans

Dennis

@AlexA. I've updated my polyglot draft:

1

A: Sandbox for Proposed Challenges

Overcoming cluster size code-challenge polyglot string math ascii-art Fed up with the reliability of Flash storage, you decided to store all your programs on a good old 1,024,000 byte floppy. However, after copying not even 2,000 programs, the disk was full. How's that even possible? Skilled in...

feersum

@Dennis So "read a string" really means to read the entirety of stdin?

Reto Koradi

Are any of you going to try aditsu's pizza challenge? Looks pretty tough for code golf. At least right off the bat, I can't think of an approach that would be somewhat efficient and simple enough to make a good golfing algorithm. Not asking for proposed strategies, I want to give it some thought myself! :)

Dennis

@feersum Exactly. I'll add that to the post.

feersum

In most languages 'read a string' suggests to read a space-delimited token or maybe 1 line

Mauris

4:22 AM

I would say it's almost always a line.

feersum

C, C++, and Java read space-delimited strings.

Dennis

OK, I'll change it to read all input.

Or is there a compelling reason to restrict the input to a single line?

Mauris

Reading all input sounds fine.

Reading a single line will be more complex in, say, Brainfuck, and easier in, say, Python

orlp

@RetoKoradi enumerate diagonally over all possible 2d pizza shapes

feersum

Happy Birthday <all input> is a little lacking in verisimilitude

Dennis

4:26 AM

Yeah, I changed the HB input to line.

orlp

then enumerate over all positions that shape fits for the first piece, second piece, etc, until everything is covered or all positions exhausted

at the end if no result found print no pizza for you

feersum

@Dennis any maximum input length?

Dennis

I capped it at 255.

Reto Koradi

@orlp Yeah, would have to be something like that. I wrote a pentomino (well, polyomino) solver before. Not terribly difficult, except for some efficiency considerations. But still, it's more complicated than what I would normally want to golf. And in that case, at least the shape of the pieces is given.

Another basic approach for the given problem would be to start enumerating ways of splitting the input, and check if all the pieces are the same. Or some kind of hybrid between the two.

Optimizer

5:09 AM

Good thing you're doing the Internet a service by removing malicious tabs. ;) — Alex A. ♦ 1 hour ago

^ sums it all up

> Optimizer is the best PPCG user ever — Alex A. ♦ 1 hour ago

2

^ too

Dennis

5:32 AM

LOL. A downvote for my post and an upvote for its comment. Somebody can't take a joke.

3

A: Goodness, it's covered in tabs!

CJam, 30 24 23 bytes q{_9=NA=Seasi*' ?@?:N}/ I usually refuse to post malicious code on the internet… Try it online in the CJam interpreter.

Beta Decay

6:01 AM

> What can I say? Tabs are the way forward... — Alex A. ♦ 2 days ago

2

↑ You guys were all asleep but Alex confessed to me his darkest secret

orlp

6:34 AM

> You really think someone would do that? Go on the internet and tell lies? — Beta Decay 1 day ago

Beta Decay

> No, everyone on the internet tells the truth — orlp 1 day ago

aditsu

7:57 AM

@Dennis traitor!!!

grc

is SE down?

nvm it's working again

aditsu

@Dennis (btw, it wasn't me)

flawr

11:23 AM

@MarcDefiant Where r u from?=)

trichoplax

@flawr I don't think chat pings will notify a person unless they have been in chat recently enough for their name to show up in the ping autocomplete

(or unless a mod superpings them)

flawr

Too bad=/

trichoplax

I sometimes work around this by adding a temporary comment to someone's question or answer if I really need to get their attention

Or if you had something to discuss I think you can invite someone to chat

(I know yours was only a passing comment but I mean generally)

flawr

Hm well I didn't want to abuse comment messages for other stuff. It is just that I think I might know MarcDefiant IRL=)

SE should implement a PM feature...

Doorknob

@MarcDefiant ^

Martin Büttner

11:32 AM

@flawr status-declined

2

Doorknob

(super-duper-ping)

jrenk

@flawr SE should become Facebook Kappa

flawr

11:49 AM

Myspace?

Mauris

"Word search" more like "esolang search".

As in me hunting through esolangs.org, not the grid

Timwi

Yes, I did that too, and decided it was too boring and stopped.

Martin Büttner

@Mauris oh you did Prelude

Mauris

12:07 PM

Yeah!

Timwi

Another question I have to all of you regarding Funciton. In all of the library functions I wrote for it, I always included (in a comment) a pseudocode representation of each function. Do you think I should keep that, or does it kind of defeat the point of writing it in actual Funciton? I’m kinda tempted to remove them all.

Martin Büttner

@Mauris which reminds me once again that I keep forgetting about sandboxing/posting the two prelude metagolf challenges

@Timwi I don't see anything wrong with including documentation

github.com/mbuettner/retina/blob/master/Examples/Rule110.rgx

github.com/mbuettner/retina/blob/master/Examples/2TagSystem.rgx

Timwi

Documentation is English text. What I’m talking about is pseudocode that’s equivalent to the relevant Funciton function. The syntax I use for that pseudocode is almost consistent and unambiguous enough that it could totally be a functional programming language in its own right...

jrenk

I will make an Esolang called "Boobies" only containing following characters ( . Y . )

Martin Büttner

@Timwi I don't think it makes a difference whether the documentation is English language or pseudocode.

(or a translation of the code to a more intuitive language)

Timwi

12:17 PM

Alright then, I’ll keep it

Thanks. Gotta go now, bye