3D in Pythonista?

zencuke

Additional note: What is hard in 3D graphics isn't the geometry, it is rendering, i.e. coloring the surfaces. Dooms biggest trick was to require all objects/walls to be perpendicular to the viewer. That simplified the rendering requirements substantially. Essentially every object in doom was a 2D shape with an image drawn on it. You scale to represent distance.

Webmaster4o

Are there any 3D libraries out there that could be made to work with numpy/PIL?

Moe

Take a look at SceneKit. While there is no python wrapper for it, you could use objc_util to accomplish that. It would probably make your life easier if you were to make a python wrapper first. To load that framework use

from objc_util import *

ObjcClass('NSBundle').bundleWithPath_('System/Library/Frameworks/SceneKit.framework').load()

omz

If you want to play with SceneKit, here's a starting point (requires beta #160023):

Note: You can use gestures (pan/pinch) to control the camera.

from objc_util import *
import ui
import math

load_framework('SceneKit')
SCNView, SCNScene, SCNBox, SCNText, SCNNode, SCNLight, SCNAction, UIFont = map(ObjCClass, ['SCNView', 'SCNScene', 'SCNBox', 'SCNText', 'SCNNode', 'SCNLight', 'SCNAction', 'UIFont'])

class SCNVector3 (Structure):
	_fields_ = [('x', c_float), ('y', c_float), ('z', c_float)]

@on_main_thread
def demo():
	main_view = ui.View()
	main_view_objc = ObjCInstance(main_view)
	scene_view = SCNView.alloc().initWithFrame_options_(((0, 0),(400, 400)), None).autorelease()
	scene_view.setAutoresizingMask_(18)
	scene_view.setAllowsCameraControl_(True)
	scene = SCNScene.scene()
	root_node = scene.rootNode()
	text_mesh = SCNText.textWithString_extrusionDepth_('Pythonista', 6.0)
	text_mesh.setFlatness_(0.2)
	text_mesh.setChamferRadius_(0.4)
	text_mesh.setFont_(UIFont.fontWithName_size_('HelveticaNeue-Bold', 18))
	bbox_min, bbox_max = SCNVector3(), SCNVector3()
	text_mesh.getBoundingBoxMin_max_(byref(bbox_min), byref(bbox_max), restype=None, argtypes=[POINTER(SCNVector3), POINTER(SCNVector3)])
	text_width = bbox_max.x - bbox_min.x
	text_node = SCNNode.nodeWithGeometry_(text_mesh)
	text_node.setCastsShadow_(True)
	text_container = SCNNode.node()
	text_container.addChildNode_(text_node)
	text_container.setPosition_((0, 40, 0))
	text_node.setPosition_((-text_width/2, 0, 0))
	box = SCNBox.boxWithWidth_height_length_chamferRadius_(100, 4, 100, 1)
	box_node = SCNNode.nodeWithGeometry_(box)
	root_node.addChildNode_(box_node)
	rotate_action = SCNAction.repeatActionForever_(SCNAction.rotateByX_y_z_duration_(0, math.pi*2, math.pi*2, 10))
	text_container.runAction_(rotate_action)
	root_node.addChildNode_(text_container)
	light_node = SCNNode.node()
	light_node.setPosition_((0, 100, 10))
	light_node.setRotation_((1, 0, 0, -math.pi/2))
	light = SCNLight.light()
	light.setType_('spot')
	light.setCastsShadow_(True)
	light.setColor_(UIColor.cyanColor().CGColor())
	light_node.setLight_(light)
	root_node.addChildNode_(light_node)
	scene_view.setScene_(scene)
	main_view_objc.addSubview_(scene_view)
	main_view.name = 'SceneKit Demo'
	main_view.present()

demo()

Utsira

Apologies for bumping a 2-year-old thread. When I try to run the above script in Python 3.5 I get an error at the map command saying:

no Objective-C class named 'b'SCNView'' found

Is this some 2.7 -> 3.5 issue? What is the extraneous 'b in the error message referring to?

thanks in advance.

ccc

The b'bytes string' is https://docs.python.org/3/reference/lexical_analysis.html#grammar-token-bytesprefix

SCNView is https://developer.apple.com/reference/scenekit/scnview

But I know nothing more.

omz

Try adding load_framework('SceneKit') before the line with the error.

abcabc

Thanks Omz . I tried also the basic code from following tutorial using your template and it is working fine.
https://code.tutsplus.com/tutorials/an-introduction-to-scenekit-fundamentals--cms-23847

from objc_util import *
import ui
import math

load_framework('SceneKit')

SCNView, SCNScene, SCNBox, SCNText, SCNNode, SCNLight, SCNCamera, SCNAction, UIFont = map(ObjCClass, ['SCNView', 'SCNScene', 'SCNBox', 'SCNText', 'SCNNode', 'SCNLight',  'SCNCamera', 'SCNAction', 'UIFont'])

class SCNVector3 (Structure):
    _fields_ = [('x', c_float), ('y', c_float), ('z', c_float)]
    
'''
https://code.tutsplus.com/tutorials/an-introduction-to-scenekit-fundamentals--cms-23847
'''
'''
override func viewDidLoad() {
    super.viewDidLoad()
     
    let sceneView = SCNView(frame: self.view.frame)
    self.view.addSubview(sceneView)
         
    let scene = SCNScene()
    sceneView.scene = scene
 
    let camera = SCNCamera()
    let cameraNode = SCNNode()
    cameraNode.camera = camera
    cameraNode.position = SCNVector3(x: 0.0, y: 0.0, z: 3.0)
 
    let light = SCNLight()
    light.type = SCNLightTypeOmni
    let lightNode = SCNNode()
    lightNode.light = light
    lightNode.position = SCNVector3(x: 1.5, y: 1.5, z: 1.5)
 
    let cubeGeometry = SCNBox(width: 1.0, height: 1.0, length: 1.0, chamferRadius: 0.0)
    let cubeNode = SCNNode(geometry: cubeGeometry)
 
    scene.rootNode.addChildNode(lightNode)
    scene.rootNode.addChildNode(cameraNode)
    scene.rootNode.addChildNode(cubeNode)
}
'''

@on_main_thread
def demo():
    main_view = ui.View()
    main_view_objc = ObjCInstance(main_view)
    scene_view = SCNView.alloc().initWithFrame_options_(((0, 0),(400, 400)), None).autorelease()
    scene_view.setAutoresizingMask_(18)
    scene_view.setAllowsCameraControl_(True)
    main_view_objc.addSubview_(scene_view)
    main_view.name = 'SceneKit Demo'
    
    scene = SCNScene.scene()
    scene_view.setScene_(scene)
    
    root_node = scene.rootNode()
    
    camera = SCNCamera.camera()
    camera_node = SCNNode.node()
    camera_node.setCamera(camera)
    camera_node.setPosition((0.0, 0.0, 3.0))

    light = SCNLight.light()
    light.setType_('omni')
    light_node = SCNNode.node()
    light_node.setLight_(light)
    light_node.setPosition((1.5, 1.5, 1.5))
    
    cube_geometry = SCNBox.boxWithWidth_height_length_chamferRadius_(1, 1, 1, 0)
    cube_node = SCNNode.nodeWithGeometry_(cube_geometry)
    
    root_node.addChildNode_(light_node)
    root_node.addChildNode_(camera_node)        
    root_node.addChildNode_(cube_node)
    
    main_view.present()

demo()

Full example here
https://gist.github.com/balachandrana/644840a223b636ecd84d14a595dad33b

Utsira

@omz thanks, it works now, that's awesome. I've used SceneKit quite a lot with Swift, but I'm still new to Python/ Pythonista. One question, how come you implement SCNVector3 as a Python class, rather than importing it with the other ObjC classes? Just trying to understand the implications of working with ObjC libraries in Pythonista. thanks.

dgelessus

SCNVector3 is a C struct, not an Objective-C class. Objective-C classes carry lots of information at runtime (their name, superclass, protocols, methods, etc.), and all Objective-C objects contain a pointer to their class. All of this information can be looked up via APIs provided by the Objective-C runtime library. That's what makes objc_util possible.

C structs, on the other hand, carry no runtime info at all. If you look at some struct data, you cannot see what its type is. For all you know, it might not even be a struct. It could also be an int, or an array of bytes. There is also no way to see what struct types there are, or what fields a certain struct type has.

So to be able to work with struct data, you first need to describe its type. In ctypes, you do that by creating a subclass of the ctypes.Structure class. Your subclass's _fields_ attribute lists the name and type of every field in the struct. Once you've done that, you can allocate instances of your struct in memory, and you can tell ctypes that a function takes or returns struct data of that type.

In this specific case, I don't think the declaration of SCNVector3 is actually needed. In the above code, you can see that all places where a SCNVector3 would be used in Swift, the Python/objc_util version simply uses a tuple. This is only possible because Objective-C is involved. All Objective-C methods have a type signature, which describes the types of their return value and all arguments, including structs. This allows objc_util to get some basic info about a method's struct arguments, which is enough to convert a Python tuple into correctly formated C struct data. This only works with Objective-C method arguments though, and not anywhere else. For plain C functions and variables, you still need to declare the struct type by hand.

Utsira

Thank you for your answer @dgelessus , really interesting.

bcraig_6973

@JadedTuna
I am new to this
Have you got a better Handel on how to do this
All I want to do is just using 3 Python learn to how create a 2d flat grid in 3d space
What are the step by steps to do this kind of thing ?

Brun0oO

Hi,

I got, from a swift project, some SceneKit assets (files with ‘.scn’ extension). I put them in a ###.scnassets directory and try to load them using the SCNScene.sceneNamed_ method.
This call fails, i got None as result.
Do you know if I need to make a kind of initialization for the SceneKit resources mechanism ?

Thanks for your help...