Something wrong with pop() [WEBGL constant is another issue]

[villares]

I'm sure we'll have 3D at some point :)

So let's put a small test example here:

from pyp5js import *

def setup():
    createCanvas(600, 600, WEBGL)
    
def draw():
  background(200)
  translate(-240, -100, 0)
  normalMaterial()
  push()
  rotateZ(frameCount * 0.01)
  rotateX(frameCount * 0.01)
  rotateY(frameCount * 0.01)
  box(50, 70, 100)
  pop()

start_p5(setup, draw)

I guess this is a bit early... but, who knows?

[villares]

Using _P5_INSTANCE.WEBGL I get this:
TypeError: _P5_INSTANCE.py_pop is not a function

[berinhard]

I got the same error here, but at least the 3D is working for now WEBGL.

[berinhard]

Updates to ongoing investigation:

The issue here is related to how Transcrypt converts the pop function from Python objects. With a more detailed version of the traceback, we can confirm that p5's pop was mapped to a new js function called py_pop:

Reading Transcrypt's source code, I was able to find the line where this mapping is defined. But I think it's a compile error, since our pop function is a regular one, defined by us, and pop functions only exist in Python via lists or dicts. Probably this will require a fix to Transcrypt.

[berinhard]

Transcrypt's __pragma__ seemed to deal with the problem. The following snippet in pyp5js/pytop5js.py worked for me:

def pop(*args):
    __pragma__ ('noalias', 'pop')
    _P5_INSTANCE.pop()
    __pragma__ ('alias', 'pop', 'py_pop')

[eduardostalinho]

Based on https://www.transcrypt.org/docs/html/special_facilities.html#id2, I think there are more instructions with this issue..

[berinhard]

@eduardostalinho I think this is true only for methods from P5.js that matches with any of those aliases definition. I spent today's morning taking a look at them and it seemed that case only occurs with p5.pop function.

Anyway, if we have the same error in the future with other functions, we'll already know how to fix them. So, I'm closing this issue as solved! Thanks for your PR =]

[kylepollina]

Hi all, I think this issue might have to open up again, or at least as a separate issue.

p5.Renderer objects also contain pop() and clear() methods that are renamed during the transcrypt stage. For example, a simple Python script like this:

from pyp5js import *

def setup():
    createCanvas(700, 700)

    gr = createGraphics(500, 500)
    gr.push()
    gr.pop()

will get compiled into this equivalent Javascript

export var setup = function () {
	if (arguments.length) {
		var __ilastarg0__ = arguments.length - 1;
		if (arguments [__ilastarg0__] && arguments [__ilastarg0__].hasOwnProperty ("__kwargtrans__")) {
			var __allkwargs0__ = arguments [__ilastarg0__--];
			for (var __attrib0__ in __allkwargs0__) {
			}
		}
	}
	else {
	}
	createCanvas (700, 700);
	var gr = createGraphics (500, 500);
	gr.push ();
	gr.py_pop ();
};

Which will result in the following error Uncaught TypeError: gr.py_pop is not a function. The same happens for the following:

gr = createGraphics(500, 500)
gr.clear()

gets transcrypted into

var gr = createGraphics (500, 500);
gr.push ();
gr.py_clear ();

Creates the error Uncaught TypeError: gr.py_clear is not a function.

I have tried this in the pyp5js.py file:

def createGraphics(*args):
    __pragma__('noalias', 'pop')
    graphics = _P5_INSTANCE.createGraphics(*args)
    __pragma__('alias', 'pop', 'py_pop')
    return graphics

but didn't fix anything. Do any of you have thoughts on how this might be avoided?

[kylepollina]

Oh you know what, nevermind. I now understand I can do this instead in my python script and it will work as expected:

from pyp5js import *

def setup():
    createCanvas(700, 700)

    gr = createGraphics(500, 500)
    gr.push()
    __pragma__('noalias', 'pop')
    gr.pop()
    __pragma__('alias', 'pop', 'py_pop')

The pragma just needs to be in my code instead of in the pyp5js library. I can open up a pull request with additional documentation on this quirk.

[berinhard]

Thanks for figuring this out @kylepollina! Unfortunately pyodide support (#124) is not ready yet. Once we have this done, these transcrypt specific fixes won't be necessary anymore =]