1223
choas
(quokk.au)
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
this post was submitted on 06 Jun 2025
1223 points (99.0% liked)
Programmer Humor
23995 readers
1404 users here now
Welcome to Programmer Humor!
This is a place where you can post jokes, memes, humor, etc. related to programming!
For sharing awful code theres also Programming Horror.
Rules
- Keep content in english
- No advertisements
- Posts must be related to programming or programmer topics
founded 2 years ago
MODERATORS
Oof, this reminds me of a personal experience.
Me: Oh this grapple system is easy, we'll just push the player's vector towards the destination vector.
Game: Oh but there's a small object in the way that cannot be moved. This will make an immense amount of collision data per tick.
Me: Can't we just ignore-
Game:
As another mod maker/game tinkerer...
Genuinely, how did you fix this?
If I understand this right, the problem would be... ignoring certain collision meshes/hulls while in grapple movement mode... but then if you stop your grapple while basically inside or intersecting with those meshes/hulls, now insane nonsense happens, right?
...
Assuming this is an OoT hookshot style, just throw the player directly at the grapple end point thing, and not a more complex and realistic 'throwing and climbing a rope like a mountain climber' style grapple... the way I would try to address this would be:
Give the grapple movement mode some kind of shut down mechanism/recovery.
Like... oh the player is still trying to grapple toward point A... but they aren't moving at anywhere near the speed they would be if they were unobstructed, cancel the grapple mechanic.
Or: oh, the player is in grapple movement mode, but they collided with something, and they're nowhere near the grapple end point, stop the grapple mechanic and stop moving them.
For either (or both) of these, at the end now transition the player into some kind of specifically designed 'grapple mode has failed due to an obstruction' state, where the player now gets some amount of damage, a 'collided into object' animation, during which the player gets repositioned into a 'collison safe/no collision violation' nearest position, like a 'get unstuck' check in an mmo or something.
Or another way would be: before the player actually begins being moved by the grapple... do the vector trace from the player, to the end point, and around that vector, quickly draw a large box, rectangular cuboid, perhaps with endcaps of some kind... that just projects what the straight line movement of the player's collision hull would be... maybe make it a bit bigger than the player's collision hull just to be safe...
...just 'draw' that real quick, if it intersects with anything, no can do chief, grapple attempt fails, doesn't engage.
That would at least work for static world objects that don't move, you'd still also need one or both of the above methods to handle colliding with things that can move, npcs, other dynamic objects.
I don't know the "right" answer, but I set it so if you hit something, it plays out some checks similar to as you described:
If we collide with something but its only waist high, then we will have the player stop the grapple and attempt to vault over whatever it is.
If we collide with something and its more than waist high, then we wait for a very small delay and see if we made any progress towards our destination. If not, end the grapple because something is in the way.
Ignore all collision damage otherwise when grappling. Either we get stopped on the way and give up, or make it and then end the grapple.
... And last but most horrible of all:
All my games are janky though so I don't think this is some ideal setup.
Edit: Cleaned up the collision damage part as I thought I handled it differently.
Yep, those first 3 are either exactly or almost exactly what I ended up with when I toyed around with making something similar, haha.
Honestly, I think what you are describing as 'janky workarounds'... are actually how you do this right, they are 'efficiently implemented game mechanics'.
Maybe the code could be cleaned up and de-spaghettified a bit, but I've seen many other systems like this in many games and mods.
If it seems stupid, but it works... it isn't stupid.
The word for that is actually 'clever'.
... you'd be amazed how much enterprise level business software, for instance, relies on some weird ancient library or function that literally has a comment in the code that says "I do not know why this works, but it does, DO NOT CHANGE".
.....
But also: oh god WATER.
Fuck video game water rofl.
I feel your pain.
You could plausibly implement some physics to deal with it. If the player is moving into a surface, move them along the part of their grapple movement component that's perpendicular to that surface. This will allow them to slide along walls/floors/ceilings realistically. For the case where they need to move "through" a small object, you could treat their collision as a sphere and have it collide with the object; for small objects, this could let them pass by. Eg. for grappling sideways over a small rock on the ground, their point of collision would be mostly below them and a bit to the right, but they're being pulled mostly straight to the right, so they would move perpendicular to the point of contact and move up-right over the rock, then continue their grapple path. Depending on your game's physics system there are other solutions, but for a typical game engine, that should work well.
That just is running into the problem the original comment was trying to avoid in the first place:
You are constantly jamming into the surface and doing a whole bunch of collision checks to basically scrape the player across the surface...
...because you have to keep doing those checks in a loop untill you determine the obstacle is finally cleared, and then switch back to unrestricted or 'normal' grapple-movement.
You have to keep doing 3d vector collision mesh check calculations for the whole time the player is being 'scraped'... because you don't know when to switch 'perpendicular movement only' mode off, otherwise... so this is inefficient.
Assuming this is a 3D environment... there's no way you can just totally null out one dimension of the movement vector unless the player is perfectly perpendicular hitting a perfectly perpendicular surface.
If your level design is any degree of complex, with objects beyond basically perfect boxes that are all perfectly orientes to the world grid... and if the player is allowed to rotate... this doesn't work, your calcs still always involve 3 dimensions.
What you're saying might work in a 2D game... or I guess 2.5D, maybe?... but it wouldn't work in a 3D game.
...
Something possibly, sort of like what you've described, I think? but not really?... another idea that might work would be:
Upon detecting a collision, before the player has gotten to the grapple end point... the grapple movement basically complexifies with more nodes.
So you use a pathfinding algorithm to draw, instead of just a line between two points... now you have a point of origin where the player is, the end point, and a third point that is off to the side of the obstruction.
Now for that first segment, now the grapple pulls the player perpendicular to the obstruction surface, so it isn't constantly colliding and doing friction... and then when the player clears the obstruction, hits that midpoint, the movent vector changes.
This is basically what I described with doing the 'draw a giant skinny box' to check if a player can do an unobstructed grapple... but now more complicated as it involves 3D pathfinding...
This could possibly work, but it would take a good deal more work to optimize this, to make your entire world work with 3d path finding... normally, nav meshes are just done on more or less flat ground, up to some degree of incline... but now you also have to do this on literally all surfaces.
Again... this might work ... but it would take a lot of game dev work to implement, as you'd have to fully 3d navmesh every level... and this potentially would not handle complex surfaces well.
3D, aerial pathfinding in a very complex environment ... to my knowledge, still isn't really a thing many games have done very well, efficiently, with a general system. It usually just a bunch of manually placed aerial nav nodes, particular to the level itself... very intensive, manual work.
...
You have an odd definition of 'realistically'.
...
Whoah whoah whoah wow ok gotta stop you there.
Spheres tend to be the absolute worst objects to use in a collision mesh or hull, because they are comprised of far, far more tris or rects than a box.
This is a terrible idea.
There is a reason hitboxes... are called 'boxes'.
I think what you are trying to describe is a common concept in games where many objects that are basically... clutter, vegetation, extra fluff... they just do not interact with the player collision mesh/hull at all, for many parts of the engine/game.
Like a uh, a small pile of trash or rock that doesn't interact with the core player movement controller, but it might interact with an inverse kinematics system that slightly modifies the player's animation so that their foot rests on top of the rubble or rock.
But uh... doing a 'estimate everything's size by bounding it with a sphere and then negating movement collision if its small?'
This is not something you'd want to call when the grapple attempt is started, it'd be a massive stutter or slowdown, you'd have to index every object in the level... and you'd end up with like, if you have a pile or array of many small things, all together... well individually they are all small, so you can phase through a pile of many small things that is in totality actually large.
This is the kind of thing you just design your whole game and level and objects around from the ground up.
Again this 'solution' of yours (which just entirely abandons the concept of just not colliding with small objects, which you literally just described) just causes the problem the original comment was trying to avoid: having to do a whole bunch of collision calcs every time any obstacle is encountered.
... You speak as if you know what you are talking about, but you clearly do not.
Have you ever actually mocked up a 3 physics scenario in a game engine, or modded an existing game in a manner that is very reliant on or interactive with its physics engine?
Unnecessary "...", and no, you don't loop the check until the obstacle is passed any more than you would "loop" the player's ordinary movement. As normal, each tick you attempt to move the player forward some distance. If there is an obstacle in the way, they'll move less distance, which is fine-- this prevents them from rocketing up walls if they're slightly below a target grapple point beyond the wall, as in the below scenario.
What would be more efficient? Depending on how the game physics work, the player's collision mesh is probably a capsule, simple box, or sphere. It's really not that expensive to add this check; the player is presumably already doing collision checks using their mesh every tick for like, standing on the ground and touching walls.
When did I ever say that you would accomplish this effect by nulling out one component of their movement vector? That idea is a fabrication of your own delusions. It's pretty easy to do a mesh collision check, get the normal of the tri the player collided with, and use that to remove all the player's movement in that direction. This is probably already part of the engine's physics calculations anyways!
This could work, especially if the grappling hook is one of those ones where gravity stops affecting you (could be good for gameplay, that's valid). But to construct this path in a realistic manner, you would need to do similar calculations to what you're saying are inefficient, except all at once instead of spread over multiple frames. If you simplify the pathfinding checks to make the movement simpler, you could in most cases do the same thing with the player collision checks. Depends on how you implement it though I suppose. Too specific to cover all cases in a general discussion.
It is realistic that if I grapple into a surface I will move a shorter distance than if I was grappling freely, yes. This is true without friction etc. as well. Think of the extreme case: grappling directly downwards into the floor, in which case I would not move at all.
LMAO are you kidding me??
First of all you could do a check using a proper sphere rather than a mesh with tris. This can actually be faster than using a box-- eg. checking if two spheres (or a sphere and a point) collide is literally just a distance check compared to their combined radii. I bet even sphere-tri collision is easier than tri-tri, although my game engine knowledge doesn't extend far enough to say for sure in that case.
They're called that because boxes are common, not because they're the best.
This entire line of critique is invalid because I wasn't saying that at all. I'm saying that as a consequence of the collisions, they could pass around an obstacle; not that they could go through it. A rock under the player as they grapple sideways would push them upwards and slightly away due to the angle of the collision, and they could then continue moving sideways as before.
How on God's green Earth could you possibly, after I literally just described the precise mechanism by which the player would interact with small objects, still believe that I meant they should simply pass through them??? Maybe if you read the whole post instead of replying to each sentence individually you would've made that connection. Yes, I see the irony; I did read your whole post first though.
If you apply the grapple as a force it's literally the same collision calcs the player makes every single tick. If you can't due to engine/game/etc. limitations, it's still not that much extra collision calculation.
Try me. I am extensively aware of the way physics is typically handled in games. I will admit I don't often use game engines, because I usually try to make 2d games from scratch and implement my own simple physics. But yes, I'm aware of how 3d engines handle physics as well.
You're the one who asked to open a gate to the fifth dimension, you can't then get upset that you broke 3+1 dimensional physics