R0J0hound's Forum Posts

  • You can rotate an object around any point with

    Rotate a degrees clockwise

    Set position to (x-cx)*cos(a)-(y-cy)*sin(a)+cx, (x-cx)*sin(a)+(y-cy)*cos(a)+cy

    Where x,y is the point, cx,cy is the center, and a is the degrees to rotate by.

    So one possible solution could be to keep track of an offset dx,dy from the object’s xy to rotate around. Initially you’d set that to wherever on the layout minus the sprites position. Then after that you’d update it with a helper function. 

    Number rx=0
Number ry=0

Function vrotate(x,y,a)
— set rx to x*cos(a)-y*sin(a)
— set ry to x*sin(a)+y*cos(a)

Function rotateSprite(u, a)
Pick sprite by uid u
— sprite: set position to self.x+self.dx, self.y+self.dy
— sprite: rotate a degrees clockwise
— call vrotate(sprite.dx, sprite.dy, a)
— sprite: set dx to rx
— sprite: set dy to ry
— sprite: set position to self.x-self.dx, self.y-self.dy

    Alternatively a different approach could be to have dx,dy be values from 0-1 to specify a point within the objects quad. Then to use that as a center you’d set the angle between two set position actions. It basically undoes the offset, rotates, then moves by updated offset. 

    Set position to Self.x-(Self.dx*self.width*cos(self.angle)-self.dy*self.height*sin(self.angle)), Self.y-(Self.dx*self.width*sin(self.angle)+self.dy*self.height*cos(self.angle))
Set angle to anything
Set position to Self.x+(Self.dx*self.width*cos(self.angle)-self.dy*self.height*sin(self.angle)), Self.y+(Self.dx*self.width*sin(self.angle)+self.dy*self.height*cos(self.angle))

    Either way you need to change angle via events. Can’t use with a behavior that changes the angle.

  • Here's one solution just that creates sprites covering horizontal and vertical runs of the same tile. Used an array to mark tiles that were already used.

    dropbox.com/scl/fi/l78tvqp8lz6oyo7fojcck/runs_of_tilemap_tiles.capx

  • The canvas is rotating around the top left corner. You can change that in the properties to be centered, but you’d also need to move it by half the width and height when positioning it.

  • Easiest would be to only use obj files with uvs. If I was still working on the plugin I would just add 0,0 for missing uvs and improve the obj loader to give more understandable errors.

    Currently the plugin is on the back burner indefinitely. It is in need of a rewrite and the event system makes things more complicated than they need to be. And finally time and interest isn’t really there at the moment.

  • I see, like I said I’ve no idea about that game. Guess you could make the player move through the doorway when they hit it, delete those doors and seal up that way. But that’s something to fiddle with another day.

    In general it’s possible to do anything. You just have to break the problem up into multiple simpler doable steps.

  • You should be able to treat the 2d array as a 1d one. If 0 means empty do: 

    on function “shiftArray”
Repeat 16*16-1 times
Array: at (loopindex%16, int(loopindex/16)) =0
— array: set at (loopindex%16, int(loopindex/16)) to Array.at((loopindex+1)%16, int((loopindex+1)/16))
— array: set at ((loopindex+1)%16, int((loopindex+1)/16)) to 0

    So the process would be to 1st set the array location to 0 that you’d want to delete and then run that to shift. It only shifts by one space so if you delete multiple then you have to run that the same number of times.

    I’m assuming the array starts fully shifted but if it isn’t there is a way to modify it to do that. Or a brute force way to do it is to just run the function 256 times.

  • Try Construct 3

    Develop games in your browser. Powerful, performant & highly capable.

    Try Now Construct 3 users don't see these ads

  • The top one has to do with the order things are drawn. In 2d it just uses zorder to draw back to front. In 3d there are some 3d layer settings that might help I suppose, like there is a setting to draw by zorder or by distance to screen. I could be remembering it wrong.

    The goal is to draw the 3d objects first, then draw the stuff with transparency far from the camera to close. That would work well if the billboard sprites face the camera, but when the quads are at many different angles there isn't any way to perfectly sort them in some cases with just changing draw order. Anyways just some ideas to fiddle with, or maybe what construct does automatically makes that less feasible. idk

  • Never played that game and probably never will. Guess its just a bunch of rooms connected by doors? And you don't see the new rooms till you enter the door?

    Easiest would be to just make all the rooms manually, and just reveal new rooms as you open doors.

    Tilemaps and sprites are just tools, I'd be more concerned at making something that works before making something that's super efficient.

    Random generation can get rather involved. One general idea with your rooms is to have each room fit within a grid with the doors on the edges. With the prefab rooms you should be able to hide that it is a grid. Then there's the matter of making all the rooms connect. One idea could be to utilize something like WFC (waveform function collapse) to let you generate the rooms on the fly. Basically start with a random room, then for each door place a room that has matching doors with the surrounding rooms, and if there are any sides with no rooms it can randomly have a door or not.

    Here's a test of the room linking idea where it generates more as it goes with a wfc like algo. I got bored before doing rooms over it.

    dropbox.com/scl/fi/z5uzt5vf20t0hiccp6keu/maze_expand.capx

    Another idea involves just defining each room, and saving it to some kind of text format to be able to load again on the fly. This literally just picks the next room at random at the doorway and removes the previous when leaving. Side effect is if you try backtracking the rooms will be different behind you, although I suppose this could be made persistent. Only supports a door per side of the room but with some tweaks you could support multiple doors per side.

    dropbox.com/scl/fi/6olz5ar4dk5ma7l2jtz81/crazyPathFromRooms.capx

    Anyways, all this is just to give ideas and was me trying some things out. You can probably find other similar examples. Maybe search for binding of issac examples. Those generate maps from room templates.

  • Do you mean something like this? 

    1 2 3
4 5 6
7 8 9

    Say you remove the 2. 

    1 0 3
4 5 6
7 8 9

    Then it shifts like this? 

    1 3 4
5 6 7
8 9 0

  • An event that isn’t a sub event of anything.

  • Maybe another way to look at it is the new objects are added to the instance list as the last step of a top level event. Functions are run in place so their new objects aren't added till the end of the caller top level event.

    It's the single most annoying quirk of the picking system, it's hard to explain, and hard to understand, but hard to not run into. With most of the events I write I do my best to work around having to deal with it.

    Anyways, I was merely pointing out it's possible to pick new objects sooner than the end of the tick if needed. It's just another tool, but if you have a different approach that's good too.

  • You can pick newly created instances sooner than just the end of a tick. The next top level event is enough.

  • I can never seem to open google drive links. It always asks me to login. So I have no input on your events.

    As far as the physics stepping mode:

    Variable will be frame rate independent but according to the box2d docs a fixed timestep is better since it makes things more consistent with the simulation. Realistically a variable timestep would only affect constraints such as joints and contacts. Can’t imagine it would be bad for simply moving objects and applying forces.

    Fixed will make the simulation more consistent but defaults to a timestep of 1/60. That means if you have a faster screen refresh rate the sim will run faster.

    A solution would be to set the timestep to match the screen refresh rate. But I forget if the physics behavior even lets you change the timestep. If it doesn’t then never mind what else I have to say. That’s kind of a blocker.

    Browsers don’t provide a way to get the screen’s refresh rate but you can guess it from the frame dt. A better guess would be to average multiple frame dts, maybe over one second. Ideally you’d want the game to be running with a low cpu/gpu load when sampling the frames so there is minimal frame skipping which would throw the average off. But it should be possible to statistically figure out that bigger dts refer to skipped frames but I digress.

  • Not really.

    If the level data is in a grid then presumably you could load it into construct as a tilemap, otherwise as sprites. Tilemap is problematic if you can’t load a different image at runtime in construct. So probably you’d have to create tilemaps and sprites with the relevant textures first then parse the level data from JSON.

    Layers? You’d need to someone match up construct features with LDTK features. Construct is limited with what you can create or modify at runtime. Guess you could create a construct layer for every layer but that would be up to you.

    Atlases? Construct does its own atlases and you can’t really replace them at runtime. That’s why I said first create objects with the relevant textures then parse the json. Alternately you could load the texture atlas into a sprite and use distort maps to select sub textures. TiledBackground is another option but replacing texture at runtime is per instance with that.

    You could use events or JavaScript. Js is only an advantage if you make your own plugin here to be better tailored for LDTK features.

    If someone hasn’t posted a converter before it probably hasn’t been done.

    Even Tiled doesn’t have a full loader. Construct only lets you load a small subset of tiled features that match with tilemap features.

    In general getting the info out of the LDTK json should be straightforward enough once you start doing it. It’s the visual aspect that would require some creativity to do in construct imo.

  • Loops are actually easy. Algorithms can be harder, especially when making them from scratch.

    So what are you trying to do exactly? Do you want to know all the tiles connected to a certain tile? You can do that with a flood fill algorithm. That could help narrow your search for how people solved that kind of thing before.

    I get lost though when you mention other things you’re trying to do and I’m not seeing a lot of connection. It probably is because I’ve never played the game you’re replicating.