Game Development

Einstein's well-tested "equivalence principle," which implies that crossing a black hole's event horizon should be an unremarkable event. A hypothetical astronaut passing across an event horizon would not even be aware of the transit.

Hawking proposes that black holes possess "apparent horizons" that only temporarily entrap matter and energy that can eventually reemerge as radiation. This outgoing radiation possesses all the original information about what fell into the black hole, although in radically different form.

Since the outgoing information is scrambled, Hawking writes, there's no practical way to reconstruct anything that fell in based on what comes out. The scrambling occurs because the apparent horizon is chaotic in nature, kind of like weather on Earth.

Pip

yep

Jon

i'm picturing hawking googling about black holes for the past 10 years

"I have an opinion on black holes now, you can read about it in my book"

Pip

lol

Kevin van der Velden

11:41 AM

Right? He is really well known for being a great physicist, you do know that?

Jon

i thought he was an alcoholic

Hjorthenify

If hawking radiation exists, then it will be a great time for physicist

Kevin van der Velden

@Jon it's probably pretty hard for someone who literally can't lift a bottle to be an alcoholic

Martin Sojka

@Jon The perfect one-way hash.

Hjorthenify

@KevinvanderVelden

Christoph

11:57 AM

so I finally found time to do some force-directed graphing

Kevin van der Velden

@Hjorthenify still hard to use if you can't put it on your own head =p

Christoph

uGFX on Teensy 3.1: draggable buttons and force directed widgets

►

Kevin van der Velden

Or fill it, or, well, anything

Hjorthenify

@KevinvanderVelden He has a helper.. Just put vodka og stronger in it and you're good to go :P

Jon

vid.me/i6o5

what they need is a giant mech to hold it upright

Kevin van der Velden

12:01 PM

Giant mechs are always the proper solution

@Hjorthenify we're not talking about him drinking alcohol, we're talking about being an alcoholic, it's really hard to drink without stopping if others can easily stop you by not assisting you

Hjorthenify

@KevinvanderVelden He's one of the worlds smartest scientists.. I'm sure he can figure something out :P

Kevin van der Velden

Like not being an alcoholist?

Hjorthenify

Exactly!

TreDubZedd

'Morning, All

user92578

Good Day!

Hjorthenify

12:07 PM

Hello!

user92578

I'll do one last run of the game and then I'm off to breaking it for a while again

Hjorthenify

your own game?

Jon

own your game.

user92578

yeah

Hjorthenify

I Soviet Russia game owns you

Jon

12:09 PM

SpaceX Falcon 9 Landing Follow Up

►

user92578

The BCR: The Big Collision Refactor

Hjorthenify

:P

I won't be coding this weekend..

user92578

y?

Hjorthenify

I'll be in Sweden without electricity :D

user92578

oh

user92578

12:10 PM

y?

Hjorthenify

Sometimes you just gotta go be one with nature

user92578

In Sweden?

Pip

@Jon hahaha nice

Hjorthenify

Yeah?

user92578

Why there? Don't you live like in Denmark?

Hjorthenify

12:12 PM

Yes, but Sweden has way more undisturbed nature

user92578

ah

Hjorthenify

I cant wait till I can play my game.. More than just ramming into the enemy's hitbox and watch it float away :P

Jon

this is cool

Penalty Kicks From FIFA 94 to 15

►

Chris

12:41 PM

one day we'll have crowds that don't suck

Jon

Yeah, i was looking at how the crowds doesn't seem to care about anything

lol

they are all looking at random spots and doing their own thing

Chris

3 models as animated billboards

Christoph

1:02 PM

so how about writing code for a "wild bunch of drunk steven hawkings" crowd? Shoud be fun and it's surely better than all the FIFA crowds until now summed up.

Jon

1:13 PM

done.

Jon

1:26 PM

heartstone plays good on android

SpartanDonut

Is that some cheap Hearthstone knock off? :P

Jon

shuddap

SpartanDonut

Don't worry. I won't tell anyone you can't afford the real Hearthstone.

Ali.S

5

SpartanDonut

Amazing

Ali.S

1:35 PM

lol

Jon

why even have a logout option

Martin Sojka

Well ... it's not like this is a new kind of error message. We had stuff like this for ages.

> suicide
You need to be alive to do that.

Jon

1:50 PM

> poop
You haven't eaten in 27 days.

Almo

2:01 PM

@Hjorthenify why is this image backward?

user92578

Aight I believe that the collision logic is in

user92578

Then onto changing all the references and class stuff etc

user92578

look at the errors flowing in

user92578

argh

user92578

Anyone around? I need some class structure help

TreDubZedd

2:15 PM

What's up?

SpartanDonut

Put properties and methods inside classes.

That should do it

user92578

So basically I've changed from using CBoundingBox everywhere to instead using CBounding everywhere, where CBounding can be instanced as either a circle or a box.

user92578

CBounding itself cannot do anything

TreDubZedd

Your own classes, or an API?

user92578

All the child classes of CBounding, CBoundingBox and CBoundingCircle need to implement collision methods for all the other types

user92578

2:17 PM

My own classes

user92578

But as I store all the collision items as CBounding pointers, I cannot call m_Bounding->Collide(other->m_Bounding) as they are both in the base type, not the actual collider type

user92578

That's my current problem

user92578

I'll put up an example pastebin to make it more clear

TreDubZedd

Make Collide virtual, and overload it (e.g. Collide(CBoundingBox* other) {...} and Collide(CBoundingCircle* other) {...} )

Christoph

The collision is not a property or capability of either bounding box, so how about a Collision class? The individual Bounding classes would have a geometry type and geometry parameters stored, and the collision has to make sense of them

user92578

2:20 PM

@TreDubZedd That's what I have

TreDubZedd

ah, ok; I understand now, I think

user92578

In the basics the problem is kinda described here: pastebin.com/Vq4NeCxs

user92578

This is starting to look like I'm going to need an enum for the type

Christoph

You need to determine the actual hitbox geometry at runtime. You're doing that implicitly by adding a layer of abstraction, which is not wrong but won't get you anywhere without further introspection. You can add a method that tells you about the size of the hitbox geometry in a certain direction, and then see if two hitboxes overlap

user92578

But the data that is required for a box and the data for a circle are different

user92578

2:33 PM

I don't think a uniform geometry solution will work here

user92578

Unless I just can't think this right (which maybe be possible), in which case could you maybe give me an concrete example of this in use?

Christoph

I'm sketching, hold on

There are two objects: a rectangle and a circle, representing two different hitbox geometries. The red line is the path between the two centers (if you can construct such a path, we're in luck).

TreDubZedd

According to my test, line 50 should know what to do...

Christoph

I've added two red circles for the rectangle: the outer is the largest circle around it. The black circle (other hitbox) intersects this circle, so it is at least close to our rectangle - but it doesn't touch the rectangle

user92578

@TreDubZedd It shouldn't. There's no method that takes in a CBounding*

user92578

2:39 PM

At least I believe so

user92578

Casting to a parent explicitly is a thing in C++, but not casting into a child class is not

TreDubZedd

pastebin.com/ycUPjEhu

user92578

So here's how I was thinking about doing it (but not 100% satisfied with the solution, but it's something that would definitely work): pastebin.com/TqJrHuaq

TreDubZedd

am I doing anything differently there?

user92578

Yeah

user92578

2:40 PM

You're taking in a Base* instead of a Sub*

TreDubZedd

yet line 49 prints out "Test: Sub Sub"

user92578

Virtual implementations do get passed with the base class

Christoph

the other red circle crosses the intersection of the center-to-center line and the rectangle. Your hitbox classes could at least provide those two distances for a given direction to another hitbox

user92578

@Christoph But I cannot simulate/test box vs box or box vs circle collisions with two circles instead of a box

Christoph

circles are actually easier than boxes because you only compare r1 + r2 vs distance

in the circle case, both red circles are the same

user92578

2:44 PM

@Christoph I know that

TreDubZedd

You could (probably) make it a two-step approach. other->PreCollide(this) with virtual void PreCollide(CBoundingBox* other) { other->Collide(this); }

user92578

@Christoph

user92578

See? The green circle does collide with the box

user92578

But it doesn't collide with the min circle of the box

user92578

2:47 PM

Hence no collision would be detected

Christoph

the smaller red circle is too small - it doesn't cross the intersection of the center-to-center line with the rectangle

in this situation it would be almost as large as the larger red circle

user92578

what is a "center-to-center" line?

TreDubZedd

The problem is that the parameter passed to Collide doesn't recognize what type it is...so flip it again, and call the virtual function on that object, forcing it to recognize its context

user92578

BTW Thanks for pointing out sketch.it for me :)

user92578

@TreDubZedd Doesn't work

TreDubZedd

2:50 PM

bummer

Christoph

the "center-to-center" line is the line between the centers of object a and object b. the red line in my sketch

user92578

pastebin.com/sNJhgVyq

user92578

@Christoph and how does the smaller circle "cross" that

user92578

Or "how does it not cross that line in my picture?"

user92578

2:52 PM

That's the line, right?

Christoph

yes, that's the line. the smaller red circle should cross the right edge of your rectangle

user92578

I'm sorry if I'm turning out like an total idiot and ass, but I'm just not getting how this works

user92578

@Christoph okay

user92578

So I'd need to dynamically re-position the smaller circle for every collision?

Christoph

yes. I've marked the intersection in my original sketch (blue circle)

Now if the black, large circle gets closer, it will eventually cross the point marked with a blue circle. Then you can be absolutely sure that the two boxes collide

user92578

2:56 PM

How about a case like this? Is the smaller one correctly positioned?

user92578

Christoph

no, smaller circle's center is at the center of the object ot belongs to - just the radius changes

user92578

okay

Christoph

can your rectangle rotate?

user92578

So like this then?

user92578

2:57 PM

user92578

@Christoph No.

user92578

Or do I need to figure out the closest point on the edge?

user92578

And set the radius to that?

Christoph

I'll do it step by step, starting with two objects:

figure out center-to-center line

find largest circle that contains the rectangle:

actually you can do that before figuring out the center-to-center line, because it is a property of the rectangle only

user92578

yeah

user92578

3:02 PM

i gotta go eating, but ill brb

user92578

thanks for your patience

Christoph

the black circle is outside this red circle, so no collision. I'm now moving the black circle closer to the rectangle

the black circle and the red one intersect, so it's a might-be-collision

Almo

@Jon @Chris Unity question for you guys

0

Q: What does the number in TextMesh.lineSpacing refer to?

I have a TextMesh with some text, and mutliple lines. The 4 examples have these settings: lineSpacing = 220, characterSize = 10 lineSpacing = 220, characterSize = 25 lineSpacing = 220, characterSize = 30 lineSpacing = 110, characterSize = 25 I don't understand what the lineSpacing number s...

unity

Christoph

now we figure out the smaller circle. it has the same center as the rectangle, and intersects the rectangle at the same point as the center-to-center line:

Jon

@Almo If I had to guess, it has somethign to do with the size of the textures AND the character size setting

because if you adjust the texture size that it generates from a font -- it completely changes everything

Christoph

3:09 PM

the black circle is outside this circle, so it's a "might-be-not-collision". It actually is no collision, the black rectangle and the black circle don't intersect. If the black circle moves closer, it will eventually intersect the smaller red circle:

Almo

@Jon you're probably right there.

Christoph

that's definitely a collision. If you want to get better results in that "might-be-collision" region, you need a different algorithm. From an implementation point of view, it's easy to calculate the radii for both red circles by implementing a virtual function for each geometry.

user92578

ah

user92578

ty for this

user4704

Still no new profiles for GDSE :(

user92578

3:18 PM

you're a mod. do somethiing!

user4704

It's all on SE.

user92578

"GIFF NEW PROFILES OR RIOT"

2

Christoph

the next step could be to implement a plane separation algorithm (see here: en.wikipedia.org/wiki/Hyperplane_separation_theorem). Again, any geometry you implement could provide a projection method for a given plane

gotta go now too

user92578

aight bye and thanks a lot @Christoph

TreDubZedd

@Tyyppi_77 en.wikipedia.org/wiki/Double_dispatch

user4704

3:31 PM

@Almo I would have guessed percentage-of-point-size initially

user4704

But your image seems to disprove that.

Almo

yeah, it's pretty weird.

Jon

its one of those unity internal secrets

user4704

I'm assuming it's some kind of similar typography-related unit.

user4704

The docs really do suck though.

Almo

3:32 PM

they do

user4704

I guess you could Reflector (or similar) the DLL defining TextMesh?

Almo

lineSpacing: the spacing between the lines.

user92578

@TreDubZedd Hmm so that kind of flipping

user4704

And maybe see what they do to the value before they pass it to whatever underlying API?

Jon

like i said, don't rule out texture size

it might do some funky calculation with that vlue

user4704

3:33 PM

@Tyyppi_77 Double-dispatch isn't really "flipping."

Almo

yeah

user92578

@JoshPetrie I would actually value your opinnion on this architectural issue I'm having

user92578

And okay, right.

user4704

Which is?

user92578

pastebin.com/Vq4NeCxs

user92578

3:34 PM

I've written a small example case there

user4704

Okay. So what is it you want to solve?

user92578

A way to call the correct Collide() method based on not the current type of the "bulletCircleOrWhatever" in this case, but the actual type of the object

user4704

Double dispatch is a typical method of doing that.

user4704

While it's a typical academic example, I'd rather avoid the issue outright and choose a different design for bounding objects and collision.

user4704

One that does not involve virtual methods on every collision check (and thus double-dispatch, which adds an extra level of indirection).

user92578

3:42 PM

While I'm reading about how I'd use double dispatch, I also thought about fixing it with a type enum and casting to a correct type.

user92578

But I guess this is not that good: pastebin.com/VwiUytrf

user92578

...than using either double dispatch or avoiding the problem?

user4704

Type-switching like that is less than ideal.

TreDubZedd

what is the ideal way to resolve collisions?

user92578

...with multiple collision geometry types

user92578

3:49 PM

That's pretty much the question behind all this

user4704

Avoid many collision geometry types.

user4704

Or at least avoid testing them against eachother.

user92578

I can't be the only one with a need for both boxes and circles

user92578

Stored in a uniform matter

user4704

Sure, but there's a couple things:

Jon

3:50 PM

twitch.tv/kenneynl

user4704

Double-dispatch for collision detection is predicated on the idea that you will have a many-to-many relationship between collision types, and that "many" is large or will become large over time.

user4704

It's preferable in that case to type-switching because it's more extensible.

user4704

But in practice you can get away with far fewer collision primitives.

user4704

In which case you can say that you understand the maintainability risks of type-switching and just do it.

user4704

Or opt for a system where you, for example, only do collision tests between same shapes, or do them that way but in a hierarchy of increases accuracy, et cetera.

user4704

3:52 PM

(depending on the needs of your game)

user92578

I understand the maintainability risks of type-switching, and I'm suggesting that someone will shout at me when I months from now come with some annoying maintaining problem with the type-switching system.

user4704

Yea, they will.

user4704

It's a fact of life.

user4704

Sometimes people "shout at me" for using bare pointers and I have to explain to them why std::shared_ptr is a shit default choice.

user4704

But that's a digression.

user4704

3:54 PM

Once you understand the concept between double-dispatch you can optimize a good portion of it away.

user4704

You can contrive, for example, a set of flags that represent the collision type used by a thing.

user4704

A bitwise or mathematical combination of those flags can create an index into a table referring to the appropriate collision resolution routine.

user4704

This reduces the cost back to a single indirection.

user4704

Further, I've found it useful to put collision tests into actual types.

user4704

Such that there is a "LineCircleCollision" class or similar.

user4704

3:57 PM

This can let you store certain aspects of some computations that you will repeat, thus avoiding repeated work. You can simply store the appropriate object in the appropriate slot in the "table" (typically just a switch, at this point)

user4704

You can then further optimize this by trying, if it doesn't otherwise impact the rest of your logic, to sort collision checks so objects with the same bitwise combination of collision primitves are all checked first

user92578

Suddenly:

user92578

@JoshPetrie Aight, I've got all that copied to a text file for later use :)

Pip

Such answer

0

A: I want to blit a standing man on pygame

Your problem here is that you are re-initializing the running_man variable inside of the while loop to zero every single frame. I suggest moving it to the outside of the loop so that it is only initialized once. Another problem that I see right off of the bat is that your event loop is inside o...