topic Re: Quantum effect - getParameterAtLength() in Fusion API and Scripts Forum

Quantum effect - getParameterAtLength()

MichaelT_123 — Tue, 10 Nov 2020 02:17:14 GMT

Hi TF360,

For a start, to create the base for the comprehension of the issue, please look at the picture below.

It represents the F360 API function:

(returnValue, parameter) = curveEvaluator3D_var.getParameterAtLength(fromParameter, length)

If LOGIC is the mother of all SCIENCES ( real, not imaginary ones), one would expect that by exceeding available size of a curve to the right, the function returned endParameter (as per getParameterExtents() values) . It is not the case. It returns startParameter.

In order to allow you to better sink into the gravity of the problem, let's imagine…

… a drunkard, in whatever direction he/she/it goes, always hit the same wall. Any physicist (even not completely sober) would describe this as the clear evidence of a quantum effect. Nobel prize… anyone?

Regards

MichaelT

Re: Quantum effect - getParameterAtLength()

kandennti — Tue, 10 Nov 2020 03:55:23 GMT

Hi @MichaelT_123 .

I made a sketch curve like the image and tested it.

The execution result looks like this.

startParameter:0.0 endParameter:1.0 startParameter + 1.0 -- returnValue:True -- parameter:0.03093685297119174 startParameter - 1.0 -- returnValue:True -- parameter:9.313225746154785e-10 endParameter + 1.0 -- returnValue:True -- parameter:0.0 endParameter - 1.0 -- returnValue:True -- parameter:0.9690631470288812 startParameter + :OK startParameter - :NG endParameter + :NG endParameter - :OK

All of the parameters were found to work correctly within the range.
Therefore, it seems necessary to check using the getParameterExtents method.

However, since "returnValue" is provided, I think it is a kinder design to return False if it is out of range.

Re: Quantum effect - getParameterAtLength()

MichaelT_123 — Tue, 10 Nov 2020 07:10:45 GMT

Hi Kandennti-San,

Thank you for your effort to check this out. Please find your code a little bit modified with the addition of putting in the real data from my experience of being hit by parametric quantum phenomena. I have not expected it, instead have spent two days scowering different section of my code (silly me).

Also are attached the file with a single sketch spline which was used for the testing.

#Fusion360API Python script import adsk.core, adsk.fusion, traceback _app = adsk.core.Application.cast(None) _ui = adsk.core.UserInterface.cast(None) def run(context): try: global _app, _ui _app = adsk.core.Application.get() _ui = _app.userInterface des :adsk.fusion.Design = _app.activeProduct root :adsk.fusion.Component = des.rootComponent skt :adsk.fusion.Sketch = root.sketches[0] crv :adsk.fusion.SketchCurve = skt.sketchCurves.sketchFittedSplines[0] geo :adsk.core.NurbsCurve3D= crv.geometry eva :adsk.core.CurveEvaluator3D = geo.evaluator # get Parameter range _, startParameter, endParameter = eva.getParameterExtents() print('startparameter:{}\nendparameter:{}\n'.format(startParameter, endParameter)) # -- getParameterAtLength test -- # startParameter + 1.0 length = +1.0 returnValue, parameter = eva.getParameterAtLength(startParameter, length) print('startParameter + 1.0 -- length:{0:<22} -- returnValue:{1:<4} -- parameter:{2:<22}'.format(length, returnValue, parameter)) # startParameter - 1.0 length = -1.0 returnValue, parameter = eva.getParameterAtLength(startParameter, length) print('startParameter - 1.0 -- length:{0:<22} -- returnValue:{1:<4} -- parameter:{2:<22} LeftFromStart gives Start~ OK!'.format(length, returnValue, parameter)) # endParameter + 1.0 length = +1.0 returnValue, parameter = eva.getParameterAtLength(endParameter, length) print('endParameter + 1.0 -- length:{0:<22} -- returnValue:{1:<4} -- parameter:{2:<22} RightFromEnd gives Start WRONG!'.format(length, returnValue, parameter)) # endParameter - 1.0 length = -1.0 returnValue, parameter = eva.getParameterAtLength(endParameter, length) print('endParameter - 1.0 -- length:{0:<22} -- returnValue:{1:<4} -- parameter:{2:<22}'.format(length, returnValue, parameter)) # fromParam by length @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ fromParam = 1.0 length = +0.023635877285719833 returnValue, parameter = eva.getParameterAtLength(fromParam, length) print('endParameter - 1.0 @@ length:{0:<22} @@ returnValue:{1:<4} @@ parameter:{2:<22} RightFromEnd gives Start WRONG!'.format(length, returnValue, parameter)) length = -0.023635877285719833 returnValue, parameter = eva.getParameterAtLength(fromParam, length) print('endParameter - 1.0 @@ length:{0:<22} @@ returnValue:{1:<4} @@ parameter:{2:<22}'.format(length, returnValue, parameter)) #@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ # -- getPointsAtParameters test -- print('') # startParameter + msg = 'startParameter + :' try: prms = [startParameter + (idx * 0.1) for idx in range(5)] returnValue, pnts = eva.getPointsAtParameters(prms) # [skt.sketchPoints.add(p) for p in pnts] msg += 'OK' except: msg += 'NG' print(msg) # startParameter - msg = 'startParameter - :' try: prms = [startParameter - (idx * 0.1) for idx in range(5)] returnValue, pnts = eva.getPointsAtParameters(prms) msg += 'OK' except: msg += 'NG' print(msg) # endParameter + msg = 'endParameter + :' try: prms = [endParameter + (idx * 0.1) for idx in range(5)] returnValue, pnts = eva.getPointsAtParameters(prms) msg += 'OK' except: msg += 'NG' print(msg) # endParameter - msg = 'endParameter - :' try: prms = [endParameter - (idx * 0.1) for idx in range(5)] returnValue, pnts = eva.getPointsAtParameters(prms) msg += 'OK' except: msg += 'NG' print(msg) except: if _ui: _ui.messageBox('Failed:\n{}'.format(traceback.format_exc())) ''' startParameter:0.0 endParameter:1.0 startParameter + 1.0 -- returnValue:True -- parameter:0.03093685297119174 startParameter - 1.0 -- returnValue:True -- parameter:9.313225746154785e-10 endParameter + 1.0 -- returnValue:True -- parameter:0.0 endParameter - 1.0 -- returnValue:True -- parameter:0.9690631470288812 startParameter + :OK startParameter - :NG endParameter + :NG endParameter - :OK ''' ''' #@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ startparameter:0.0 endparameter:1.0 startParameter + 1.0 -- length:1.0 -- returnValue:1 -- parameter:0.05831409622449987 startParameter - 1.0 -- length:-1.0 -- returnValue:1 -- parameter:9.313225746154785e-10 LeftFromStart gives Start~ OK! endParameter + 1.0 -- length:1.0 -- returnValue:1 -- parameter:0.0 RightFromEnd gives Start WRONG! endParameter - 1.0 -- length:-1.0 -- returnValue:1 -- parameter:0.9415661386942856 endParameter - 1.0 @@ length:0.023635877285719833 @@ returnValue:1 @@ parameter:0.0 RightFromEnd gives Start WRONG! endParameter - 1.0 @@ length:-0.023635877285719833 @@ returnValue:1 @@ parameter:0.9989981070240436 startParameter + :OK startParameter - :NG endParameter + :NG endParameter - :OK #@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ '''

Best Regards

MichaelT

Re: Quantum effect - getParameterAtLength()

kandennti — Tue, 10 Nov 2020 09:08:53 GMT

This is a little different story, but in the case of the surface, there is a SurfaceEvaluator object.

http://help.autodesk.com/view/fusion360/ENU/?guid=GUID-31dad9b2-ee1d-4216-8100-09ea44f9967a

The methods are also relatively similar to the CurveEvaluator3D Object.

@BrianEkins mentioned somewhere in the forums that the results will vary depending on how you get the SurfaceEvaluator.

・BRepFace.evaluator: finite. (taking into account the boundaries of the face)
・BRepFace.geometry.evaluator: infinite

In particular, the results of the SurfaceEvaluator.isParameterOnFace Method will be greatly affected.

https://help.autodesk.com/view/fusion360/ENU/?guid=GUID-e6682eb7-48f8-472d-9872-71da3c5ba8c8
(BRepFace.geometry.evaluator returns True even if the parameter is out of range.)

I assumed the same thing would happen with CurveEvaluator3D Object.

When I looked at the Help.

The red marks are provided by adsk.core and I expected it to be able to handle over-range parameters in the same way as CurveEvaluator3D.
Unfortunately, out-of-range parameters did not return the correct results.

I interpreted this as probably being a specification.

Re: Quantum effect - getParameterAtLength()

MichaelT_123 — Tue, 10 Nov 2020 11:00:59 GMT

Hi Kandennti-San,

What is/are Evaluator(s)?

Evaluators are a specific set of functions enveloping parametric geometrical object (2D, 3D, 4D, …)

Perhaps the simplest such object is a line:

Y = y*t

X = x*t

Where t ∈<t1,t2> ( t can be infinite too)

There are infinitesimally many different geometrical objects, and their analytics are varying too, so each one has its own SPECIFIC set of functions, F360 has named it "evaluator".

You have mentioned BrepFace object.

By analogy to a line if BRepFace the evaluator has parameters <u,v> constrained by respective ranges. If such constraints are not set BRepFace will become Surface (in some cases infinite one).

The core of evaluating functions is intact.

Regards

MichaelT

Re: Quantum effect - getParameterAtLength()

KrisKaplan — Tue, 10 Nov 2020 19:50:54 GMT

In this case, it is not related to topological bounded and unbounded evaluators.

This definitely just looks like a bug. Specifically with the NurbsCurve3D evaluator for temporary geometry and sketch geometry. There looks to be some mishandling of negative lengths (directions) in some edge cases, and in properly clamping the end values. (In some cases where the algorithm fails, it asserts (in debug builds) and returns 0.0, which just happens to be the start parameter in your case.)

I filed defect FUS-74761 for this issue.

If your purpose for calling getParameterAtLength is to get some sample points for the curve, you could try using the getStrokes API instead. The chordal deviation tolerance that it uses isn't the same as the arc length used by param at length, but it could be suitable for some uses.

BTW: The getPointsAtParameters tests in your script that write NG look to be as designed. It is an invalid argument to specify parameter values that are outside the curves declared parameter range, and the two NG tests walk outside the curve's parameter range.

Kris

Re: Quantum effect - getParameterAtLength()

MichaelT_123 — Wed, 11 Nov 2020 10:04:39 GMT

Hi Mr ChrisKaplan,

I am glad that the problem has also been identified on the F360 side. I hope it is in good hands/brains now and will be readily addressed. In the meantime, I use bandAid. It costs some μseconds, which are nothing in comparison to the time I have had spent to find this problematic treasure.

Regarding your suggestion of using getStrokes(), I use it from time to time. It is however very sensitive (in some cases) to tolerance parameter. It can generate a tremendous number of fit points.

Regards

MichaelT

P.S.

I have some other pickle, but one at the time...🤐

Re: Quantum effect - getParameterAtLength()

jenniferappl — Thu, 18 Apr 2024 15:12:50 GMT

That's a quirky but fascinating way to describe the issue with the F360 API function! It does seem paradoxical that exceeding the curve length would loop back to the startParameter rather than the expected endParameter. This behavior definitely brings a smile, imagining it as a quantum effect where no matter which direction the drunkard heads, they hit the same wall.
For a deeper understanding of this and potentially similar quirky behaviors in computational models, you might find some insightful resources at https://quantumai.co/. Sometimes, diving into the fundamental principles can shed light on these bewildering phenomena.