R0J0hound's Forum Posts

  • Had a look. Looks like that example just has a terrainheight function that gives the z height of any xy point on the terrain. They used a bunch of formulas to define the shape of the terrain. Cool but not necessarily the easiest to tweak. They could have sampled a greyscale image or used a noise function to define the shape as well.

    The idea could be used with wedges. Basically pick the wedges under an xy, and calculate the z with some math. The main issue is that method isn’t enough to deal with walls too. It’s mostly good for smooth terrain. So wall logic would be separate.

  • I’d have to look.

    I’d think raycasting in the z direction would give the z position on the slope without sliding. There are various ways to do that but projecting the xy position onto the equation for the plane of the top face of the slope is one idea. But youd have to do some checks to consider what direction you were moving into the wedge so you’d hit it from the side and such.

    Or actually we could take the push out direction in my example and if it’s mostly a change of z then do the z ray cast.

  • I’d say what you're doing is fine. You could also add semicolons to the front and back of the string to be able to have one replace work for all positions. Then you could remove the semicolons at the ends with mid() but usually I leave them there if I’m doing that sort of thing.

    Do bear in mind that if two or more tokens are the same then the replace will remove them all.

    A general solution could be to make your own tokenat that returns the index of a token. And then you can remove that with a left and right I imagine.

  • The method works by finding the closest point on the shape’s surface and pushing the object away from that. A side effect of that is wall sliding. Guess you don’t like how it’s slower moving up a slope because of that.

    There isn’t really a fix with this method. Maybe one of the other ideas floating around would help.

  • Instead of touch.x which grabs the coordinates from the bottom layer you can use touch.x("TouchInderface") to get it from the specific 2d layer.

    That should solve the touch coordinates issue I think.

  • Most of the ideas have limited times they can be used. Events aren’t flexible in all ways so I generally decide what I want to do with consideration of what you can do with events.

    Stuff like interfaces lose me. I mean I could look up what they do but they seem like just an abstraction like much of object oriented stuff that just adds complexity.

    The event system is basically a mix of multiple things. The picking system which is kind of like functional programming, and without that it’s structured programming like a limited c without pointers or structures. On the object oriented front we have plugins which act as objects and families which are more like genetics than inheritance. I guess behaviors and custom actions could considered a bit object oriented too.

    As a whole the event system doesn’t adhere to any specific paradigm or even map directly to JavaScript. It’s just a little of everything with many useful parts, arguably missing parts, and a few really awkward parts that require you to choose a different approach when doing some things.

    In the end we all have our preference on how we want to structure or abstract our code.

    Personally I want to use as few languages as possible so I use events and js only to add missing engine features when absolutely necessary. The engine as a whole has good coverage of features but like anything it has limits which require doing things a different way. I don’t make plugins or effects because I hate dealing with build systems, I don’t want to maintain such addons and deal with comparability across browsers and systems. The reason I use a program like construct is because I enjoy the fast edit->run cycle to iterate on stuff. My goal is to do stuff in the least amount of events possible while remaining readable so only add abstractions where it simplifies stuff and makes it readable.

    Cheers

  • Try Construct 3

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

    Try Now Construct 3 users don't see these ads

  • You could set an instance variable to self.y and use “pick highest variable”.

    Or You could add a “for each ordered by sprite.y descending” with a stop loop action.

  • Yeah, the pick2 is for the reverse case.

    The pick alls aren’t needed but I added them to show that you need all the instances to be picked for this to work. Actually, it would work if the same instances were picked in both families too.

    This is what a normal filtering condition like this:

    Sprite1: width = sprite2.width

    Expands to roughly under the hood: 

    // Sprite1 and Sprite2 are lists of the current picked instances. 
for(let i=0; i<Sprite1.length; i++) 
if(Sprite1[i].width == Sprite2[i%Sprite2.length].width)
 Pick Sprite1[i]

    A normal filtering condition is stuff like compare width, compare x, pick by comparison, … etc. Static conditions such as is overlapping or pick nth instance do their own thing.

    Anyways the relevant bit is what instance of other types are used in the condition. In practice we tend to just filter the other types down to one instance before doing a condition like that but it can be useful when the instance counts are the same.

    Actions reference other instances in the same way and has more utility. Most useful is “sprite1: set position to sprite2” if sprite1.count=sprite2.count then each sprite2 will have a sprite1 on it. And if sprite1.count=2*sprite2.count then each sprite2 will have two sprite1 on it.

    And yep that’s how else works with picking. You’d have to use a for each to avoid that.

  • Here's an example that uses mesh distort to make any polygon. Events 4-5 are the relevant ones. It takes the points variable as a comma separated list of xy coordinates as input. It sets the mesh size to 2 x count/2, then loops over the top row, and loops backwards over the bottom row to set the points.

    construct.net/en/forum/construct-3/how-do-i-8/manipulate-mesh-points-175978

  • Here’s a rough way to do lift. Basically you find the velocity perpendicular to the wing with a dot product and apply a force from that to do lift. It’s modeled after a lift formula found on Wikipedia. You’d tune it by adjusting the lift variable. 

    Var vrel=0
Var lift = 0.01

Every tick
— set vrel to sprite.physics.velocityX*cos(sprite.angle +90) + sprite.physics.velocityY*sin(sprite.angle +90)
— sprite: apply force -sign(vrel)*lift*vrel*vrel at angle sprite.angle+90

    Drag can be done similarly but with sprite.angle instead of sprite.angle+90, and it would have its own drag variable to tune instead of lift.

    It can probably be modeled better by scaling the lift by the width of the wing and scaling the drag by the wings height. But that would require tuning the scaling variables more. It probably could give pretty convincing motion with a wing a tail and a prop to give thrust.

    There are probably many other improvements to make the model better

  • There’s a system action now to set the collision cell size. I don’t recall what release it was added.

  • You can indirectly do it with a distort mesh since that distorts the collision poly too.

    Say you have 64 point polygon you want to use, you’d take a sprite with a square collision poly and apply a 32x2 distort and position the points to the poly. It won’t necessarily look pretty so you’d have it invisible and attach the visual on top of that.

    So apart from distorts changing or adding collision poly points nothing is setup for modifying collision polys outside the editor.

    Maybe with JavaScript but you’d have to bypass the scripting api to get into the unsupported and undocumented innards that has a has the possibility of changing per release and breaking anything you do with it.

  • Well since cellular automata usually works by looking at neighboring cells a 2d array seems like an easy choice. Guess you could use a dictionary but wouldn't you need to add an entry for each cell?

    Both arrays and dictionaries are rated O(1) for accessing data, which means with a key or index they can access the value with one lookup. So Big O notation doesn't help too much. Arrays are more memory efficient since it's just a list of values, whereas dictionaries store keys and values and there could be other dead space reserved depending how the dictionaries are implemented. But I imagine the speed difference is negligible, you'll hit other bottlenecks first.

    Personally I'd choose the one that looks cleaner.

    Arrat.at(x,y) vs Dictionary.get(x&","&y) but whatever you want to do.

    All those other data objects are just arrays and dictionaries.

    cvs and json are parse text data and then internally are just arrays, and a mix of arrays and dictionaries. Binary is just a 1d array of numbers.

  • Construct's picking system is good for filtering instances but not great when you want to compare two different instances.

    plazatin's idea could look like this: 

    var other=0

pick sprite instance 1
-- set other to sprite.variable

pick sprite instance 2
sprite: variable = other
-- do something

    If you want to compare more than one value from another instance then you could store the iid instead: 

    var other=0

pick sprite instance 1
-- set other to sprite.iid

pick sprite instance 2
sprite: variable = sprite(other).variable
-- do something

    Another common approach is to add a family with just that type, and name it otherSprite. The only caveat is you'd need to add the variables to the family to make them accessable from both sprite and otherSprite. 

    pick otherSprite instance 1
pick sprite instance 2
sprite: variable = otherSprite.variable
-- do something

  • Well all I can find is the SpeechSynthesisVoice JavaScript api has a list of voices you can use that varies per device. I don’t see a way to add other voices from JavaScript.

    You can find lots of sites that provide a service to use your own voice to read text but most aren’t free and all look like they generate the audio on their servers?

    One library sounded promising for making your own tts voice but the setup had around 10 steps and it wasn’t clear how you could use the voice.

    Overall seems like a complicated mess.