From UV to point on Mesh = impossible?

[nonquad]

Ah yes I see. Clever idea.

Of course you are assuming the code only has to solve for a single moving point. If you had many points to solve for, or if the query point was jumping around erratically, your solution wouldn't help much.

Of course, for the situations where those limitations aren't a problem (which is undboubtedly a lot of situations), that idea would rock. Either way it could potentially be used to augment a more robust system.

For more general uses, my gut feeling is that Joojaa's on target with the quadtree thing. And I still feel that the index array is a good alternative contender.

The index array would be a lot easier to impliment, but the quadtree might be more effective. I doubt we'll know for sure till one of us tries some actual coding.

Unfortunately that has its own problems, since you can't really build a quadtree in mel. A fair comparison would require API coding experience.

[then8man]

For anyone else who looks at this thread (as it's fairly old):

 Although the manual methods are probably useful for other software, (eg. Blender), in Maya there are already nodes that do this. I use the 'follicle' node. There is another one called 'pointOnSurfaceInfo' (nurbsSurfaces only) and another for muscle attachments ('cMuscleSurfAttach', which I haven't really tried - but apparently sticks to a nurbsSurface or a specific mesh face with adjustable UV.)

The follicle node has U and V parameters, and can take either a mesh or a nurbsSurface as input. (Note: It treats the UVs of a nurbsSurface as normalised (0.0 to 1.0) even if they aren't.) It also aligns itself to the normal and U tangent of a surface.

The easiest way to create one of these nodes is to create a hair via the menu Dynamics/Hair/Create hair, then delete the hair nodes. Otherwise the node can be created directly, using 'createNode', and manually connected up; my example is using Python (PyMel) as I took it from one of my scripts but the main commands can transfer to MEL easily enough.

import pymel.core as pm
 
patch = pm.PyNode("pCube1")
patchIsNurb = False
useSmoothedMesh = False
name = "TestFollicle1"
 
# Create a follicle and name the transform
fol = pm.createNode('follicle', ss=1)
folXform = pm.rename(fol.getParent(), name)
#fol = folXform.getShape()  # This line is not required in PyMel; fol updates itself when renamed.
 
# Set up the follicle
pm.connectAttr(fol.outTranslate, folXform.t)
pm.connectAttr(fol.outRotate, folXform.r)
pm.connectAttr(patch.worldMatrix[0], fol.inputWorldMatrix)
if patchIsNurb:
    pm.connectAttr(patch.local, fol.inputSurface)
elif useSmoothedMesh:
    pm.connectAttr(patch.outSmoothMesh, fol.inputMesh)
else:
    pm.connectAttr(patch.outMesh, fol.inputMesh)
 
# For testing on a default pCube (otherwise it misses the UVs)
fol.parameterU.set(0.5)
fol.parameterV.set(0.375)

[joojaa]

Pretty old == Ancient. Back in 2002 Maya didn't have many of the needed nodes. Also the API was a bit harder to get into back then. Follicles didnt exist untill about 6 years later. There are more up to date discussion on the subject.

But then maya GUI was lightning fast back then. Not like the slow morass it is today, i blame QT.