วีดีโอ

pog

it's actually all a scam

real

pos.x / cell width, pos.y / cell height is the hash function here for getting the cell by the position. But many seem to criticize me for this terminology.

​@@rustwithoutrust You're only listing inputs without explaining the concept you titled this video.

Oh yes!

Yes, working on quadtrees right now.

@@rustwithoutrust Good that you know about it 😀👍

For the evolution project? Yes I do, but only if there's been a move.

@@rustwithoutrust how would you change algorithm if every single object moved each frame at random direction? collision only with the circles ofc

What do you mean by "collision only with the circles"?

@@rustwithoutrust you know, the moving circles that move towards the triangles. i forgot how you named them in the video

@@rustwithoutrust hi, i woke up and realized that my 1 AM queston made no sense. i'm sorry for that one

Thanks! Well, in the actual demonstration code I store buckets containing objects in a 2D array and index it by [y][x]. My hash function gives me floor(pos.x / cell width), floor(pos.y / cell height), and I index the correct bucket. Isn't it what we're talking about? Yes, I'm working on quadtrees right now, but I don't yet see how they could be useful for finding the closest neighbor, because the closest neighbor isn't always in the same cell. Subdivisions are just some recursion, so it isn't terrible.

​@@rustwithoutrust My previous comment must have gotten delted, maybe because I included a link, but here's the gist: floor(pos.x / cell width), floor(pos.y / cell height) is usually not the hash function for spatial hashing, but rather the INPUT to the hash function. You then take the result of that hash function, modulo it by your array length and that's your index into an array of buckets. You can input any X and Y and don't need to pre-allocate a huge grid of cells. Any X and Y will always map to some bucket in your array. It basically works like a hash map with (cell X,cell Y) as input. Regarding quadtrees, I was only wondering about performance for searching defined areas. It also seems possible to make well-performing implementations for finding the closest neighbor using a quadtree (I found some papers by searching Google), but I haven't tried implementing something like that myself yet.

I think this really depends on cell size and search area size. I’m sure there’s a ratio between the two that is optimal for this sort of search method.

The hash function is pos.x / cell width, pos.y / cell height.

@@rustwithoutrust That's quantizing or binning, not hashing.

Essentially the grid is a hashmap where the key is (x, y) and the value is what is inside the cell. What we hash here is the pos, in order to index the correct cell.

@@rustwithoutrust I understand what you are saying, it just isn't hashing.

Then hashmaps don't involve hashing by that logic, if we get each other.

1. Nobody's defined what's "isn't that much", for what computer, and for what context. 2. We're talking about the ratios here: 0.1/1.7 ≈ 0.06, so the speed of the version with spatial partitioning is just 6% of the straightforward approach.

real

he's a super advanced AI written in Rust

Well, the hash function is `pos.x / cell width, pos.y / cell height`.

@@rustwithoutrust So basically you store all objects in a hash table, and to look up an object in a given cell - you use the hash provided? And you use a hash table instead of an array of indices to save memory because there are many empty cells?

OK, thanks for the feedback.

yeah it didn’t explain much but it got me interested in the topic

The title of the video didn’t exactly say that the video will teach you everything about Spatial Hashing and how to use it. I prefer to look at it as an introduction video to Spatial Hashing. Now it got me interested and want to learn more about it.

Quadtrees, but they're kinda confusing if not used carefully.

@@rustwithoutrust Hm, interesting, i'll look in to those, might come in useful for my 2d game engine project

@@MrTurbo_ They're basically dynamic cells: if a cell has more than X objects in it, it gets recursively split into 4 quadrants.

Yes, it's beautifully simple and fast 👍

It's sort of O(n.m) because there seem to be a different number of items vs agents. Or to put it another way, it's O(n) "per agent", which is not the whole story. If the items are fixed in the world, you can also consider creating a static space partition (Voronoi diagram), but the non grid-like complexity might not be worth it, and removing an item would incur a recalculation cost.

Oh, I get the confusion. I mean that if we take one object out of the whole collection and go through the rest of the objects for this one object, it's O(n) or O(n-1).

@@rustwithoutrust Ohhh okay I didn't realize you were talking about just for one cell, I thought you meant calculating the entire world.

Yeah, that part could have been kinda confusing.

I think they send signals or something, but that's a good idea.

Is it anything different to computing a geo hash? E.g using Uber’s H3 library

fr

@@David-7815 Yooo! Thanks!

@@rustwithoutrust Github link is broken.

No, it's just that I don't want people to accidentally click a hyperlink.

I allow you.

@@rustwithoutrust thank you, king rust

Did I miss the "hashing" part? I thought that spatial hashing had to do with converting the location of an object into a special hash value that can allow you to easily grab nearest neighbors. Either way, I like where you are going with this, and would love to see an implementation of quad trees

​​@@pumpkinhead002Well, the hashing is basically (pos.x / cell width, pos.y / cell height), and I agree that's not really the best term for that, but people seem to call it that way.

​@@pumpkinhead002Quadtrees is actually exactly what I'm working on right now, so that thing will be big.

@@fabricehategekimana5350 Thanks bud! I'd also really appreciate your review on my latest evolution project series and maybe your opinion on what I could add.

Thanks bud!

nice

​@@cobbcodingikr

Thanks for your feedback! I'll make a video with in-depth analysis of why bodies behave the way they do and demystify certain kinds of behavior they pick. And yes, I'll talk about the implementation too!

@@DerSolinski Well, I agree with you, but I don't think that's necessary a negative. Well, something different here is that the bodies sometimes end up with different properties from run to run due to conditions. Conditions are pseudorandom, so a lot of rains, for example, allow them to have so much food that they can afford everything, but when several droughts in a row take place, they try to have a low vision radius, a low speed, and a minimum amount of skills, so they save up energy. But generally, yes, it's close from run to run.

anyways this is dope.

@@samifawcett4246 Hey, this is a pretty good concept that'll definitely spice it all up. I'm currently working on improving performance during rains, but when I'm done with that, I'll try to implement something like that, because the curve of development here is pretty deep.

Well, the closest is all the skills with lower speed and vision distance, I guess.

they're already kinda mentioned in one of the videos, but I agree

@@rustwithoutrust still deserves a dedicated vidya

C mentioned :pog:

ayo

Based

Glad to hear!

Right, that wasn't the cleanest way. Actually, thought about it 😅 Great you've noticed that! Edit: No, you actually can't do it, because you can't lock twice at the same time.

that wouldn’t work since it will cause data races, another thread could occupy it when you were going to grab a write.

@@kevz3351 your right. in this case RwLock does not help

​@@kevz3351You guys are getting me confused now 🤔 Lemme see who's right later for myself.

fr fr

fax

source?

@@rustwithoutrust it came to me in a dream

@@cobbcoding expected

Good video, a lot of stuff covered at once. It might be a good idea to make several videos going in-depth on each topic.

@@cobbcoding The next one will be about properly using `unsafe`.

@@rustwithoutrust safely using unsafe

@@rustwithoutrustuse that meme with the crab with the gun

@@noahbarger1 I will.

I will. I'll also make a video for those who are literally just starting to program and explain everything I remember I struggled with when I was a beginner. What do you think?

@@rustwithoutrust yes, good

Thanks 👍

@@rustwithoutrust needed more skibidi rizzin up livvy dunn in ohio no cap tho. Other than that it was good.

​@@cobbcodingfax

POG