One thing you haven't talked about are the callbacks and C# events. Events do everything you require without the need to implement your own register/unregister functions (which gets really ugly when you have more than one callback in a class).
Chebyshev is pronounced "Tshebb - eee - sheff", or at least that's as close as I can render it for English pronunciation. If you want some sort of a mid-point between these norms, look at the p norms for 1 < p < 2 (for between Manhattan and Euclidean), or p > 2 (for between Euclidean and Chebyshev). No physical analogies for these, but useful (1 < p < 2 encourages orthogonal movement, but allows diagonal movement when it's a noticeable improvement, while p > 2 encourages diagonal movement, while still preventing the silly zig-zagging that you get on straight orthogonal lines with Chebyshev distance). That said, Chebyshev distance does make sense realistically: if you've got something moving on two perpendicular axes independently (think grabber arm moved by two motors, one in x direction, the other in y in one of those toy-grabby things), then Chebyshev distance is exactly the correct distance to use to estimate how long it will take to get between two points.
Hi! Mathf.Abs(... - ...) results for X and Y could go into variables making it no longer pure one liner, but much more readable and faster (due to the ABS and subtract being done once for each, instead of twice, which would be done in great most cases as in most cases you will compare non-neighbours therefore going through both parts of the function.) so it would look like: int absX = Mathf.Abs(this.X - tile.X); int absY = Mathf.Abs(this.Y - tile.Y); return (absX + absY) == 1 || (diagOkay && absX == 1 && absY == 1); PS. I know its gravedigging some old video, but because many other were speaking about this bit of code why not to kick it again :P Anyway, ignore me, and have a good day :P Cheers!
You could always skip the pronunciation issue and just call Chebyshev by it's other name, 'chessboard distance.' Named such after how a king moves in chess.
From my understanding, small functions like Mathf.abs() are typically inlined by the compiler, so function calls aren't really a thing here. I could be totally wrong though.
This is a very code-light episode. We're primarily going to be discussing different ways to calculate distance on a grid (Euclidean, Manhattan, Chebyshev) as well as responding to various comments and questions from TH-cam comments.
I know i´m waay behind with the videos, but hey, whatever. When implementing proper euclidean pathfinding, you could indeed cache the sqrt() thingy. Instead of telling the character to move from Tile A to Tile B, you could tell him the direction to move. By doing so, you could use the constant value of 1 for horizontal/vertical movement and a constant cached sqrt(2) for diagonal movement.
Quill - between my watching these tutorial videos at 1.5x speed (your voice works very well!) and trying to watch your Stellaris 'Sci-Fi Grand Strategy' series. You're basically taking up all my spare time right now. Which is annoying, because I'm missing out on some great TI6 games! Ahaha... Enjoying as always, thanks :)
I also was taught in Pascal at university. My GF (who is a few years younger than me) was absolutely horrified when that one came up in conversation one time.
If you are limiting characters to sitting in the middle of a tile then manhattan norm or the Chebyshev norm would be the best (Chebyshev would allow diagonal movement). This is because of the fact we are limited to tiles and they are cheaper to compute as the Chebyshev is just the maximum of x and y distances. If you allow characters to free from, then euclidean would be best because you can take the actual shortest path no matter where the character is.
A lot of performance things that are not actual issues, and probably wouldn't be. I would like to see you address more code structure things rather than performance micro-issues.
I'm with quill - I'd prefer verbosity in this case. But if you want to be a stickler (like these guys apparently do), you can save a mathf.abs() call by doing mathf.Abs(this.x+this.y - tile.x+tile.y) == 1
What's the measurement system again where every second diagonal counts as 2? It creates more realistic measurements while still allowing diagonal movement.
This might be silly but could it work out to basically blend smoothly between Manhattan- and Chebyshev distance? That should be equivalent to giving diagonals a distance of 1.5 which is fairly close to the euclidean distance. I'm not sure whether it's worth it calculation-wise but it seems reasonable? Edit: Ah, just, like, half a second later, you adressed this, essentially.
How hard would it be to make the job have percentage of completion. So that maybe one character completes it maybe 50% and then gets interupted, but when he gets back or another gets back, they only have to complete the last 50% percent?
Chebyshev:Ch - like match; e - like gAme (æ); b - like b :); y - is Close central unrounded vowel (if you look this article in wiki you will find audio file with pronunciation); sh - like SHarp; e - the same as previous 'e'; v - like violet [or like v :) ]Hope it will help you) Crap, youtube remake it in one single raw :((
"Manhattan distance is literally named after a city" - Actually, no, Manhattan is a borough of the New York City, named after the Manhattan Island on which it mostly lies. There are also a couple of cities across USA named Manhattan, but the Manhattan distance wasn't named after them.
+izikblu If I understand what you mean Abs will not work. Abs only gives you the magnitude of a vector if you are in 1 dimension. In order to get the magnitude of a vector in 2 or more dimensions you must do a square root.
hey with the amount of money you make why do you not make actual real projects you seem to understand code and the relative balances why not make a bigger project.
![Unity Base-Building Game Tutorial - Episode 21! [PATHFINDING - Part 1]](http://i.ytimg.com/vi/subTTBjGSC0/mqdefault.jpg)
![Unity Base-Building Game Tutorial - Episode 21! [PATHFINDING - Part 1]](/img/tr.png)
![Unity Base-Building Game Tutorial - Episode 22! [PATHFINDING - Part 2]](http://i.ytimg.com/vi/u5873yoCRvo/mqdefault.jpg)
![Unity Base-Building Game Tutorial - Episode 25! [NOT Cutting Corners]](http://i.ytimg.com/vi/UJ1nGARQtPw/mqdefault.jpg)
![Unity Base-Building Game Tutorial - Episode 17! [Job System Continued]](/img/n.gif)
One thing you haven't talked about are the callbacks and C# events. Events do everything you require without the need to implement your own register/unregister functions (which gets really ugly when you have more than one callback in a class).
Quill you should do again more Unity stuff.. it's so old and yet more advanced than half of yt tutorials
Chebyshev is pronounced "Tshebb - eee - sheff", or at least that's as close as I can render it for English pronunciation.
If you want some sort of a mid-point between these norms, look at the p norms for 1 < p < 2 (for between Manhattan and Euclidean), or p > 2 (for between Euclidean and Chebyshev). No physical analogies for these, but useful (1 < p < 2 encourages orthogonal movement, but allows diagonal movement when it's a noticeable improvement, while p > 2 encourages diagonal movement, while still preventing the silly zig-zagging that you get on straight orthogonal lines with Chebyshev distance).
That said, Chebyshev distance does make sense realistically: if you've got something moving on two perpendicular axes independently (think grabber arm moved by two motors, one in x direction, the other in y in one of those toy-grabby things), then Chebyshev distance is exactly the correct distance to use to estimate how long it will take to get between two points.
Hi!
Mathf.Abs(... - ...) results for X and Y could go into variables making it no longer pure one liner, but much more readable and faster (due to the ABS and subtract being done once for each, instead of twice, which would be done in great most cases as in most cases you will compare non-neighbours therefore going through both parts of the function.)
so it would look like:
int absX = Mathf.Abs(this.X - tile.X);
int absY = Mathf.Abs(this.Y - tile.Y);
return (absX + absY) == 1 || (diagOkay && absX == 1 && absY == 1);
PS. I know its gravedigging some old video, but because many other were speaking about this bit of code why not to kick it again :P
Anyway, ignore me, and have a good day :P
Cheers!
You could always skip the pronunciation issue and just call Chebyshev by it's other name, 'chessboard distance.' Named such after how a king moves in chess.
+Salvador A. Melo That's probably a REALLY good idea.
From my understanding, small functions like Mathf.abs() are typically inlined by the compiler, so function calls aren't really a thing here. I could be totally wrong though.
This is a very code-light episode. We're primarily going to be discussing different ways to calculate distance on a grid (Euclidean, Manhattan, Chebyshev) as well as responding to various comments and questions from TH-cam comments.
I know i´m waay behind with the videos, but hey, whatever.
When implementing proper euclidean pathfinding, you could indeed cache the sqrt() thingy. Instead of telling the character to move from Tile A to Tile B, you could tell him the direction to move. By doing so, you could use the constant value of 1 for horizontal/vertical movement and a constant cached sqrt(2) for diagonal movement.
Abs can be implemented with a XOR operation so it should be fast.
Quill - between my watching these tutorial videos at 1.5x speed (your voice works very well!) and trying to watch your Stellaris 'Sci-Fi Grand Strategy' series. You're basically taking up all my spare time right now. Which is annoying, because I'm missing out on some great TI6 games! Ahaha... Enjoying as always, thanks :)
3:01 "What the hell am I talking about? WELL! Let's play 5x5 Sudoku!"
So that one-liner already gives you your Manhattan-distance right away.
I laughed way too hard qwhen he said pop pop
I also was taught in Pascal at university. My GF (who is a few years younger than me) was absolutely horrified when that one came up in conversation one time.
This Boolean/Equation worked for me:
(Mathf.Abs((this.X - t.X) + (this.Y - t.Y)) == 1 ||
(diagnals && Mathf.Abs((this.X - t.X) * (this.Y - t.Y)) == 1))
Cheers!
If you are limiting characters to sitting in the middle of a tile then manhattan norm or the Chebyshev norm would be the best (Chebyshev would allow diagonal movement). This is because of the fact we are limited to tiles and they are cheaper to compute as the Chebyshev is just the maximum of x and y distances. If you allow characters to free from, then euclidean would be best because you can take the actual shortest path no matter where the character is.
Pop Pop! Community reference had me dying
Man your videos are amazing. Thanks so much
A lot of performance things that are not actual issues, and probably wouldn't be. I would like to see you address more code structure things rather than performance micro-issues.
I'm with quill - I'd prefer verbosity in this case. But if you want to be a stickler (like these guys apparently do), you can save a mathf.abs() call by doing mathf.Abs(this.x+this.y - tile.x+tile.y) == 1
Yup, you fooled me. I did not expect you to reference a programming language from the 1970s.
What's the measurement system again where every second diagonal counts as 2? It creates more realistic measurements while still allowing diagonal movement.
I just learned what a 1 unit diagonal is, I believe it is 1.4142
its just √2
This might be silly but could it work out to basically blend smoothly between Manhattan- and Chebyshev distance? That should be equivalent to giving diagonals a distance of 1.5 which is fairly close to the euclidean distance. I'm not sure whether it's worth it calculation-wise but it seems reasonable?
Edit: Ah, just, like, half a second later, you adressed this, essentially.
How hard would it be to make the job have percentage of completion. So that maybe one character completes it maybe 50% and then gets interupted, but when he gets back or another gets back, they only have to complete the last 50% percent?
I'm in college, and we do pascal. It sucks.
Chebyshev:Ch - like match; e - like gAme (æ); b - like b :); y - is Close central unrounded vowel (if you look this article in wiki you will find audio file with pronunciation); sh - like SHarp; e - the same as previous 'e'; v - like violet [or like v :) ]Hope it will help you) Crap, youtube remake it in one single raw :((
"Manhattan distance is literally named after a city" - Actually, no, Manhattan is a borough of the New York City, named after the Manhattan Island on which it mostly lies. There are also a couple of cities across USA named Manhattan, but the Manhattan distance wasn't named after them.
+quill18creates instead of square root/squaring, you can use Abs
+izikblu If I understand what you mean Abs will not work. Abs only gives you the magnitude of a vector if you are in 1 dimension. In order to get the magnitude of a vector in 2 or more dimensions you must do a square root.
later in the video he did what i was talking about (Abs(this.X-t.X) + Abs(this.Y -t.Y))
+izikblu that doesn't give you the same answer as sqrt((this.X-tile.X)^2 + (this.Y-tile.Y)^2) though. It doesn't give you the Euclidean distance.
ahh
it will look silly if it zig-zags... just go straight
hey with the amount of money you make why do you not make actual real projects you seem to understand code and the relative balances why not make a bigger project.
+1 for the Community reference. Pop Pop!