The Nineteenth Byte

Adám

12:00 AM

@Riker Largest temperature range work ever. Beats even astronauts.

DJMcMayhem

12:26 AM

Proof that vim is the greatest text editor ever invented

Zacharý

1:36 AM

@DJMcMayhem Wait ... what?!

Ordering a PIZZA in Vim?

Anonymous

So I'm thinking about writing a follow-up challenge to this one, but with mixed dice (for example, 1d12+1d8). Would that be too close to the original?

Zacharý

I personally don't think so.

Anonymous

Some of the same approaches can be used (like enumerating all possibilities and counting up the frequencies), but DFT-based approaches would need quite a bit of adapting.

Simply Beautiful Art

Oh, btw, I managed to golf it by checking if 0 wasn't in the array as well as some other small things. Woot woot! Here's the final product :D

HyperNeutrino

2:39 AM

@NieDzejkob No, it is not... That's really weird behavior. I'll have to look at why that's happening. Thanks for pointing that out!

HyperNeutrino

2:59 AM

CMP: Should I have Santa only be able to move 1 step per turn in Advent Challenge 2? Alternatively, should I make the elves gain extra steps too? It seems a bit easier than I originally planned, which might be a good thing, maybe not. I think I might want to just leave it because answers might already be being developed, but I'm not sure.

I don't know what I was thinking when I allowed Santa to take multiple steps because that takes the point out of the original challenge idea :( Oh well, I'm probably just going to simplify it down to "will Santa catch an elf", though I need to partially modify the backstory in that case :P

New Sandboxed Posts

3:57 AM

0

A: Sandbox for Proposed Challenges

Frequency Distribution of Mixed Dice Rolls A follow-up to this challenge Given a set of mixed dice, output the frequency distribution of rolling all of them and summing the rolled numbers on each die. For example, consider 1d12 + 1d8. The maximum and minimum rolls are 20 and 2, respectively, w...

Adám

5:00 AM

0

A: Community Promotion Ads - 2017

6

Anonymous

5:15 AM

@Adám I should add a bar for Actually

Steven H.

6:47 AM

@Adám That's an awesome thing that I never knew I needed.

Anonymous

7:31 AM

The only problem with that project is, the code is horribly unreadable, and so it's difficult to add more langs

Fatalize

8:07 AM

@Zgarb I think that's because ~+ will first generate a list of 1 element [X], and then for ~f it will backtrack on the values of X infinitely instead of trying new lengths of ~+ lists

On further testing, I actually have no idea

NieDzejkob

@DJMcMayhem I don't disagree with the statement, but that proof is weak - Emacs is ordering salads, without the help of a browser, and it's healthier.

Fatalize

I would have to look into it more with Prolog's debugger but I don't have time right now

NieDzejkob

(link)

Fatalize

@Adám that's pretty amazing

@Adám Can we contribute with other languages and if so can we get short instructions?

Adám

8:38 AM

@Mego That's easy due to commands.txt being almost completely regular. I'll add Actually when I have a few minutes.

user202729

I can't access Github...

Adám

@Mego The code doesn't need modification to to add more languages, only some plain-text character variables and possibly the CSS. I'll be happy to take requests.

user202729

(without a proxy, with a proxy the CSS are mis-rendered)

Adám

@user202729 Which language do you use? APL or Jelly or 05AB1E?

Fatalize

@Adám I don't understand how the code page is actually defined in the code. tcs seems to contain the symbols + the previous 2 are the shorcut, but e.g. were is the simple letter A defined?

Adám

8:41 AM

@Fatalize All I need is list of built-ins in as regular a format as possible. I'll be happy to take requests.

user202729

I can use a proxy and get over it. No problem. I also already have a keyboard.

Very useful. But isn't the ASCII characters extraneous?

(by ASCII character, I meant the printable ones that already can be typed)

Adám

@Fatalize The code page is not needed, and thus never defined. lbs is a list of strings. The first character is the language bar entry and the rest is the tooltip content (except for the backtick shortcuts and tab completions, which are filled in from bqk/bqv and tcs respectively).

Fatalize

yeah I got it while you were writing :)

user202729

But why don't you use full URL (https://abrudz.github.io/lb/jelly.js) instead of just /lb/jelly.js?

Adám

@user202729 No, if you hover over them (yeah, sorry, that needs a mouse/stylus) you get a tooltip with info about the symbol.

@user202729 Because the script is already in that dir.

user202729

8:47 AM

What is the "Jelly" link in

> Drag this link to your bookmarks bar: Jelly

at abrudz.github.io/lb/jelly ?

Adám

@Fatalize For compositions and/or tab completions I need a systematic format of that too.

@user202729 Just a bookmarklet which fetches the code, runs it and inserts the result in the current page.

user202729

Does it use full URL? (I suppose yes)

Adám

@user202729 yes

Anonymous

@Adám Thanks! To help out, I added a list of nice names for the various functions in Actually.

Anonymous

I also added the ability for Actually code to be compiled from those nice names, so Actually can be (marginally) more readable

Anonymous

8:59 AM

I don't know why I even bother trying to support Python 3.3 anymore

Adám

@Mego Where?

Anonymous

@Adám docs/nice_names.txt - each line contains 0 or more space-separated nice names for the command with the ordinal line_num

Anonymous

If there aren't any nice names, it's either not a command (like 0x00) or part of a literal (", ', :, [], 0123456789)

Anonymous

Though I need to fix numeric literal handling

Anonymous

Or maybe I won't

Anonymous

9:02 AM

:3.14+2.17i is plenty readable, right? :P

Sherlock9

I was just thinking about how Actually's commands could use names

Anonymous

@Sherlock9 Great minds and such :P

Adám

@Mego I can work with that. I'll remove all that are blank in commands.txt and then fill all the blanks in nice_names.txt with "SYNTAX" and then make the tooltip ?\nnice_name\ncommand

Sherlock9

I keep forgetting a[b] is E and it's documented as [a][b] when everything else is a[b:] and such

Good grief

Anonymous

I've been thinking about doing that for a while - I just finally sat down and did it

Anonymous

9:04 AM

@Sherlock9 Yeah I should probably make the docs more readable

Anonymous

@Adám That works for me

Sherlock9

Ah, were it not for my thesis proposal, I'd get back to work the tutorial again

Starting to think a Jelly-style wiki may not be a terrible idea

Anonymous

Possibly

Anonymous

But Actually's commands don't lend themselves to such a neat structure as Jelly's (niladic, monadic, dyadic)

Sherlock9

Unfortunately, suggesting such things generally leads to "and will you maintain it? :D"

Anonymous

9:06 AM

:P

Anonymous

If you get it started, I'll do my best to keep it updated

Anonymous

But bear in mind that I often forget to bump the version in setup.py before pushing :P

Sherlock9

No, but I think there are good categories for it: your original numeric, string, list, function but also stack and I/O

Anonymous

Yeah, the tests are more-or-less organized into categories like those

Sherlock9

Heck I'll type it up now

I'm procrastinating anyway

Might as well be productive

Anonymous

9:11 AM

The same thought process led to me finally adding nice names just now :P

Sherlock9

XD

Adám

9:26 AM

@Mego abrudz.github.io/lb/actually

2

Anonymous

@Adám Beautiful

Anonymous

My only complaint is that the keyboard obscures the TIO buttons

Adám

@Mego APL is great for rapid development. I just write some APL to convert your raw files to the rewuired JSON…

Anonymous

I use Actually for some rapid reformatting/translation stuff

Fatalize

@Adám pull requested

(I added a 2px padding to the bottom of the keyboard (it has 2px padding on the 3 other sides) so that the dot below some letters doesn't touch the border)

@Adám I think there's a problem with your Jelly bar: ÆæŒœ don't appear so people can't find the sequence for them (although they are pretty trivial)

Sherlock9

9:49 AM

@Mego should r -> range(len(a)) be put under Numeric or both List and String?

For that matter, should l -> len go under Numeric or both List and String?

Anonymous

I'd say list/str for both

Anonymous

I consider the categories to be based on the input(s), not the output(s)

Anonymous

Because of the concatenative nature, it makes more sense to categorize commands based on inputs (to support working forward) than to do the categorization based on outputs (which would support working backward).

Sherlock9

Well, I'm about halfway done with the Numeric list. If you have any quibbles with what I left out and left in (I put in a bunch of notes that say "vectorizes"), let me know

Anonymous

I'll look at it when I'm less tired

J c

10:13 AM

I'm trying to figure out how to "score" a combination of probabilities... is taking an average of them a silly thing to do?

Steven H.

What's the context?

J c

I'm trying to combine probabilities in a way that makes sense... if I have digram probabilities for the word "test" calculated as follows:
{"TE"=>0.00978, "ES"=>0.01319, "ST"=>0.01249}
Is there a good way to "score" that as likely to be an actual english word, as opposed to gibberish? And if so, how can I account for varying word lengths?

Steven H.

Like Markov chains?

J c

i'm not sure, those probabilities are just frequencies from a large corpus of english text, i'm not sure you can combine them in a chain or not

Steven H.

If you're trying to do Markov Chain likelihoods then averaging is a very silly thing to do

And those frequencies, as you described them, definitely define a Markov Chain

Score the probabilities as "likelihood of next character given first character", and do a product of all of them

i.e. your probabilities should be higher than what they currently are by a lot

Instead of measuring "ST" against all possible two-letter combinations, measure T against all the other letters that follow "S"

Sherlock9

10:19 AM

Never mind. Removed the "vectorizes" comments. Will just add it into the list page

Steven H.

That helps balance out sequences like "xy", which happen very infrequently overall but relatively frequently if "x" is given (which it is, thanks to the previous probability in the chain)

If any of the math is confusing we can probably talk about it somewhere else to not clutter up TNB

J c

awesome - i think that sounds exactly what i want, i'll see how far I get!

Steven H.

Keep in mind that this kind of Markov Chain isn't going to be perfect. The best ones are going to be ones that keep track of some sort of external state or lookbacks (like, say, three letter sequences, or four letter, or n-letter), but depending on how good you need the results to be your two-letter chaining should do fine

J c

makes sense - i'm trying to score the likelihood that the first portion of an email address is "real" or not by how much actual english content is in it, versus "mashing of the keyboard", i was going to start with 2 letter and go up to 3 depending on how things go

Fatalize

@Adám ᵗʰᵃⁿᵏˢ

Steven H.

10:31 AM

@Jc For that case specifically, make sure you ignore numbers or transliterate them to an un-L33T'd version. ab13m0n5t3r isn't the likeliest to appear in standard English content, but it's still understandable as "Able Monster" which signifies almost certainly real. People are generally more likely to have those non-alphabet characters in email addresses than they do elsewhere in their discussions

J c

good point - currently i'm stripping all but alpha characters, but i like your idea of translating leetspeak :)
btw, you mentioned that the probabilities should be higher and part of that is by doing a product of multiple ones - i think that makes them lower though (can chat somewhere else to avoid clutter if you prefer?)

Steven H.

Markov's House

For discussion relating to probabilities.

Markov's House

Transcript for