SE Diplomacy

user61230

00:03

....whyyyy

user61230

00:36

There. Have a /data and /gamestate folder.

user61230

I'm going to enter the map into the map file... urgh.

user61230

(Thankfully, the test cases finally pass.)

user61230

03:07

There. Agh. Have a graph of a map, too.

user61230

Thank the gods, it passes its unit tests.

user61230

github.com/telthien/diplomaSE/blob/master/data/map.txt

Air

03:25

So, correct me if I'm wrong, but looking at your map file I don't see that it distinguishes explicitly between land and water territories?

user61230

The first token is L/W for land or water.

user61230

But no, it doesn't explicitly distinguish between the two.

Air

The first token is for the type of adjacency, not the type of node

user61230

Hmm, I probably could have picked a clearer naming convention.

user61230

L means that it's adjacent when considered as a land unit; W means it's adjacent when considered as a water unit.

user61230

03:27

In other words, to an army, L\N\WAL\LON,LVP,YOR indicates that Wales is adjacent to London, Liverpool, and Yorkshire.

Air

Got it.

user61230

To a fleet, W\N\WAL\IRI,ENG,LVP,LON indicates that Wales is adjacent to the Irish Sea, English Channel, Liverpool, and London.

user61230

My next commit will add a comment to the list explaining what's going on there

Air

And Y/N is supply.

user61230

Mhm.

Air

03:31

What interests me about this approach is that it keeps units in legal positions based on an assumption of starting from a legal position.

(I'm not criticizing here, just thinking out loud.)

user61230

(I understand, s'fine :] )

Air

So it's a little awkward if the task was, given an arbitrary type of unit and arbitrary location, determine whether it's a legal location

user61230

Hmm...

Air

i.e., is it land or water

but you should never have to solve that

user61230

Actually, I don't think so. In order for something to be a legal move placement, it must have adjacent territories.

user61230

03:33

So if you wanted to check if a fleet were in a legal position, you'd check to see if len(AdjacentWater)>0

Air

Hmm, and an army in the ocean?

I guess it would look up from the L entry

user61230

Yeah, the Irish Sea (for instance) shouldn't have any land connections.

user61230

If a unit is somewhere, it must have gotten there somehow and/or must have somewhere to go.

Air

That's pretty slick.

user61230

Thanks :]

Air

03:37

It makes sense for a game that's all about resolving orders, most of them dealing with strictly adjacent relationships. (friggin' convoys...)

user61230

(in the base rules, there are actual literal paradoxes that revolve around convoys)

user61230

(they added rules later to try to resolve them)

Air

Yeah, I read about a few of them

It's kind of frustrating because I want to have a logical, reality-based interpretation for things like two supported attacks bouncing against a convoying fleet

Something that says, uhhh, that's a lot of action going down there, maybe you're not going to be able to convoy through that

user61230

To be fair, that's what the resolution usually involves?

Air

Yeah, but the logical path to get there is far from a no-brainer

user61230

03:42

yes, that >_>

user61230

Hrrrrm. I'm trying to figure out how to implement orders, now...

Air

Probably want to first check that they're legal, before attempting to resolve anything

I have to admit that so far there are a few things about Go that make my Pythonista brain twitch

user61230

I'm curious, like what?

Air

Like why string literals in import statements?

like for ... := instead of for ... in

like I'm not sure I yet understand how to think about types in Go

> although Go has static types the language attempts to make types feel lighter weight than in typical OO languages

I can tell it's gonna take some hands-on time for me to grok what that's saying

user61230

yeah, I know what you mean

Air

03:53

I hope you don't write and refactor faster than I can catch up to the point of being able to contribute ;)

user61230

Go takes a lot of things that have pretty good definitions already and changes them all just a little bit so that you won't quite know what anything means anymore

user61230

like a pointer

Air

Uh oh

user61230

@Air don't worry, even this code took me longer than I'd like to admit .-.

Air

Reading the FAQ, I get the impression that Go is both interpreted and compiled, both statically and dynamically typed, it is every paradigm, it's modern and fast and the first two letters of Google and basically designed to be the forever hotness

all that, and a cute gopher, too

user61230

03:57

well, uh, I hate to do this, but uh, yes, the first two things are both true

Air

no!

user61230

no? :P

Air

good god

my head

It's a "no!" of disbelief

user61230

ehehe

Air

Too lazy to find the unicode interrobang

I imagine that the first meeting of the minds that led to the creation of Go involved heavy usage of the phrase, "...just crazy enough to work!"

user61230

03:59

that's... probably true

Air

Just tell me there's no global interpreter lock

user61230

the concurrent model in particular

user61230

oh god no, it's almost the opposite

user61230

Go gives you the goroutine, which basically acts as a job that gets assigned to a thread in the background without you needing to do anything

Air

That... sounds sexy as hell

user61230

04:00

you can basically say "go do these 500,000 items" and Go will schedule them and run them all on threads through automatic reallocation

user61230

it uses all threads you've allocated to it to execute those items

Air

Hmm, jumping abruptly back to dipSE stuff, I wonder if it would make sense to analyze the list of orders to identify sets of potentially interactive orders and resolve them in bunches

Then you could do it in parallel

(Not that performance is ever going to be a thing for this app)

user61230

(eh, Go gets C-level speeds a lot of the time, particularly for stuff like this)

user61230

user61230

hrrrrm

user61230

04:04

I see what you're saying?

Air

So like, if we look at our most recent set of orders

I think F Spa(nc)-MAO and A Por-Spa would have to be bunched

Since, if the former fails, the latter bounces

But something like A Mos-Liv is totally isolated

user61230

Yeah

Air

War, Gal, Bud are together

user61230

topological sorting

Air

I dunno, though. I could see it costing as much to identify these groups as you'd gain by resolving them in parallel.

Is this a DAG?

user61230

04:07

thank the gods, yes

user61230

not when drawn on a map, but the order of precedence for order resolution is

user61230

you can say that Gal depends on War, and Bud depends on War, so War resolves first

Air

Okay... I was gonna say, every move is potentially two-way

user61230

so, there's a bit of tomfoolery I'm thinking of doing along these lines

user61230

Instead of first running a topo sort, we run cycle detection

user61230

04:09

cut out any nodes that are in a cycle, and treat them as a single node

user61230

thankfully, in Diplomacy, it isn't possible to have two connected cycles

user61230

so that would only need to be done once

user61230

at that point, since all cycles have been removed, it's once again DAG

Air

What does a cycle look like in orders?

user61230

04:11

or something more complicated, like:

user61230

Air

This is fascinating, but we're quickly reaching the end of my graph theory/discrete maths knowledge

I remember my roommate talking about it when he was taking discrete in college, but at the time I was baking a cake and not really listening

In other news, I updated my desktop to Windows 10 yesterday since I'm home sick with nothing better to do

I have some ragrets

user61230

Hmmm?

Air

It is gross. I'm about ready to abandon Windows for good, after some 20 years.

jimsug

I quite like Windows 10

Although I'm running it on my Surface Pro 3 and my desktop runs Ubuntu.

Air

04:19

I upgraded from 7, which I was perfectly okay with

I'm on my laptop right now which is still running 7

jimsug

I think 7 > 10 is a smoother transition than 7 > 8/8.1, though.

Not that it has to be a good change, but I think it's a less jarring change.

user61230

I'm debating whether to update my Surface before college...

jimsug

I was very briefly on 8.1 on the Surface... it was good, too.

Air

I appreciate that the idea of what users want to see during an installation has evolved over the years

jimsug

I don't think I got to use it enough to be able to say whether it's better than 10, but it was good.

Air

04:21

But I have never preferred a pulsating screen that says "this is taking longer than expected, please don't turn the computer off" for 30+ minutes to the old rapidly-scrolling console of things happening or the slightly less old status bar

Also, "hey choose quick setup, we'll totally read your data, but it's easy, just do it"

Nope. Custom settings, off, off, off, off, off.

Guess what? We fucked with the start menu again. Now there are tiles. Tiles are so hot right now.

I freely admit this is more a "get off my lawn" reaction than a serious technical critique.

But, I was running 2k Pro as recently as 2012

It's possible I am not so into change

jimsug

04:39

maybe :P

user61230