R0J0hound's Forum Posts

  • The plugin is at the point where some design choices early on make it hard to change things without breaking stuff. Plus the event system's speed and limits make it hard to allow the amount of flexibility I'd like without being overly verbose and complex. Anyways, that's partially personal opinion. Construct makes some complex stuff simple and some simple things overly complex.

    As for your wishlist:

    *The view distance is just an artifact of how the fog is implemented. There are pros and cons of either. Rewriting the plugin to have alternative options for everything would be an option i suppose.

    *As I recall I try to fit the shadow frustum to just fit around the view frustum for maximum shadow detail. It's not perfect because some things off camera aren't being included due to some approximations on my part. There are alternate ways to do it, such as fitting the shadow frustum over the whole scene, but the tradeoff is giant scenes would have low res shadows. But I agree more work would need to be needed with shadows but it turned into just fiddling with settings to see what works well which was too time consuming.

    *blend modes would be nice.

    *opacity is already doable. You just have to turn on "transparent yes" in the object settings. The caveat is it's done per object. So it just draws objects back to front. Per polygon would be better, but more complex to do with how the renderer is implemented, also it still would have cases where the sorting would be off. Best would be an algo called "depth peeling" but it's slower.

    *normals are already in the plugin. As long as an obj file has normals in it you'll have normals, which is used for smooth shading. If you mean normal textures, then no, only one diffuse texture is used.

    *Collisions and raycasts are highly requested things. I have a good idea how to do them but I don't find I enjoy finding a way to make it usable in the limits of construct's event system.

    Overall, I may work on this more, but I'm more inclined to work on other things first if I find the time and motivation to code.

    -cheers

  • So, creating the mesh is something you’d probably do just once. Like:

    Start of layout

    — vertex (100,0,0)

    — vertex (100,100,0)

    — vertex (0,0,0)

    — save as mesh “triangle”

    Now it looks like you have a bunch of imagepoints that make up the polygon border of the image. You need to convert that to triangles. One simple way to do that is to do a triangle fan.

    Something like this. P is the imagepoint you want to start the fan from, best if you use a concave point. Ipx and ipy is sprite.imagepointX/y, and count is sprite.imagepointCount. I just didn’t want to type that all out. 

    Var p=1
Var i=0
Start of layout
— repeat count-2 times
— — set i to p
— — vertex(ipx(i), ipy(i), 0)
— — set i to (loopindex+p)%count+1
— — vertex(ipx(i), ipy(i), 0)
— — set i to (loopindex+p+1)%count+1
— — vertex(ipx(i), ipy(i), 0)
— save as mesh “triFan”

    Notice it matters what imagepoint you start with. See below. Also this won’t work for all shapes, but hopefully will. For any shape that doesn’t work with a triangle fan, you’d need full blown triangulation of polygons.

    Also I guess I didn’t specify uvs. For any imagepoint you can convert it to uvs with

    U=(Sprite.imagepointX(i)-sprite.bboxLeft)/sprite.width.

    V will be similar but with y, top and height.

  • There isn’t a distance limit so you just have to play with the numbers.

    Basically it’s a mix between the acceleration and spring stiffness. When the spring is stretched far enough it will counteract the acceleration.

    The damping makes it less springy. Set it to 0 to see it bounce all over. The damping slows it down.

    320 and 240 were just half the 640,480 window.

    I used the layout size to limit the position we use to the scroll area. It’s not needed if you had unbounded scrolling.

    The a and d variable are just the angle and distance from the camera to the mouse.

    There’s always some feature I forget about with my examples. You may still be about to use screen shake after you set the mouse position. Or we could do our own with some random values added to the scroll position.

  • CustomMovement is unwieldy, I never use it.

    Here's what I came up with:

    dropbox.com/s/jb23i3gx19ccktl/spring_camera.capx

    or radial instead

    dropbox.com/s/tt5p35r8r640mna/spring_camera_circle.capx

    Basically the camera is controlled by a damped spring which gives nice easing in and out. There are three variables you can tune:

    acceleration controls how fast it moves.

    stiffness controls how far you can look and how fast it returns to the center.

    damping gets rid of bouncing. higher values also make everything more sluggish. too low and there will be overshoot.

    Edit:

    Here the spring is applied to the whole camera instead of just the look around:

    dropbox.com/s/1lfz90zxj92bd9i/spring_camera_whole_camera.capx

    Kind of gives the impression of a real camera operator.

    Edit2:

    dropbox.com/s/42gs4g1tljsp35n/spring_camera_whole_camera2.capx

    Fix to ease down when scrolling to the edge of the layout.

  • Probably the solution is to not use lerp. Instead move the scroll around by applying an acceleration which would eliminate the jarring. The idea is you’d want an ease in and out instead of the jarring ease out that typical uses of lerp provides.

    I don’t have the time today to make a complete example but the basic logic would be:

    Mouse is in edge region? Accelerate camera to a max distance

    Else decelerate back to 0.

    Anyways there’s more to it than that. The acceleration provides the smooth motion and you can utilize a formula so you can gradually stop at a specific location.

  • That’s the only fancy thing you can do with expressions.

    From most complex

    TypeName(index).behavior.method(param1,param2)

    To most simple:

    TypeName.property

  • One way that’s pretty simple is to give your sprite an instance variable “target” and place the waypoints on the layout in the order you want to follow. You could alternately place the waypoints in any order and use an instance variable to set the order.

    Anyways the events would look like this: 

    Every tick
— sprite: move forward 100*dt pixels

Sprite: on collision with waypoint
— sprite: set target to waypoint.iid+1

Pick waypoint instance sprite.target
— sprite rotate 60*dt degrees toward (waypoint.x, waypoint.y)

    The 100 is the speed, and 60 is how fast it turns.

  • Try Construct 3

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

    Try Now Construct 3 users don't see these ads

  • Is the gltf file you’re using one of the ones included in examples for this plugin?

    I helped with the initial gltf loader part of the plugin. The plugin should be able to load any gltf file saved as embedded from blender, or any other program.

    With the gltf format it has three different variations:

    Gltf + bin

    Gltf embedded

    Glb (binary)

    The loader in this plugin only handles the embedded variation. Which seems like the format you’re using since it works with the Babylon test bed.

    Anyways if you’re able to share the gltf file or a c3p project with the file it may help to find the issue.

    From the error it looks like something isn’t loading correctly.

    Mikal

    It may be something as simple as the the exporter that made the gltf file not adding all the properties that the plugins loader expects. I thought it follows the spec but this may be a case where if that part isn’t there you’d just add it from the other info in the file.

  • int works the same as floor. They both round down, so your probability will be equal for all dice rolls with int(random(1,11)).

    Unless they changed the behavior of int to just truncate the float to an int. But you would only see that with negative numbers.

    int(0.5) = 0

    round(0.5) = 1

    floor(0.5) = 0

    ceil(0.5) = 1

  • Ultimately you can store the data any way you want. In this example I just stored it in a list of number,name;...

    Mainly because it was simple to parse.

    dropbox.com/s/ly7rfnnbg5c79m8/weighted_random.capx

  • If you put all those values in an array and just get one at random that would work.

    Array.at(int(random(Array.width)))

    For a more deluxe and compact version you could store the values and weights of each value. The sum of all the weights should be 1.

    Then to randomly pick a value based on the weights you’d:

    1. Generate a random value 0-1

    2. Loop over the array

    3. If the current weight<value? Then subtract the weight from the value and look at the next array position.

    4. If weight>=value then you’re done and the array position you’re on is the value to use.

    Anyways that’s the gist of it at least.

  • Is it all the odd numbers from 1 to 100?

    int(random(50))*2+1

    Or is it a different set of odd numbers?

  • Can’t open the project, but filtering can be done by looping over each entry in the array, checking its length and removing it if needed. The only catch is you’ll want to loop over it in reverse. 

    Array: for each x
Compare: len(array.at(array.width-1-array.curx) >3
— array: remove at array.width-1-array.curx

  • Skimming over it it doesn’t look like there’s anything preventing it from being ported over.

    The vertex shader portion doesn’t do anything, which is good since construct doesn’t let you modify that anyways.

    The fragment shader has a bunch of uniforms: floats and vec2s which can be two floats.

    The rest would depend on if GLSL is the same for webgl and OpenGL. So there may be a need for some fiddling.

  • That’s kinda what I meant with calculating. You’d effectively be just doing all the math that mouse.x/y does behind the scene. Which would work but I don’t think it provides much benefit.

    As for the lagging of the sprite behind the actual mouse cursor. I thought that was just because of latency between code and the drawn frames or 1/60th of a second.

    An alternative is the to set the mouse cursor from a sprite, which doesn’t lag behind but it’s on top of everything as it’s not truly a sprite then. That reason that doesn’t lag is the operating system is drawing it then which is more input event based instead of frame based.

    Anyways unless the lag you’re seeing is something different. In which case I’m all for tightening it up if possible.

    Edit:

    It just occurred to me that when you scroll some stuff doesn’t get recalculated till the end of the tick. For example viewportLeft(). It possibly could be the same for the mouse(layer).x expression. That could be lag you’re getting which is different than the other one I mentioned.

    You’ll have to test it but assuming that that setting the scroll position causes mouse.x to update till the next tick you can order your events like this.

    Basically before scrolling, save the scroll position to variables. Then after the scroll correct the position.

    Set sprite x to mouse.x

    Set oldScrollx to scrollx

    Set scroll position

    Set sprite x to self.x-oldScrollx+scrollx