Reordering Mesh Vertices

[user_deleted]

Hi !
I'm sure you all are aware of problem, when vertex order changes in a polygonal mesh (ie. due to cutting and rejoining). I need to reorder those vertices back, using another, topologicaly identical, mesh as a reference.
I'm an API newbie, so maybe my way of thinking is wrong

I hoped that i could do it this way:

1. having two meshes (named: ordered and unordered) for each vertex of unordered i search for closest point in ordered mesh. I then store it's index and write it's coordinates into unordered mesh pointArray. This way my meshes have points in matching order.
2. next step would be iterating through all unordered's edges, and replacing vertexIndices to those stored in step 1.

does that make sense ?
I already tried replacing points, but i don't see a method to alter edge's indices for a start and end vertex. Is it possible, perhaps with MFnMesh ? If i can't modify those, I'd have to create new mesh with completely rebuilded components.

I'd welcome any tips, directions or general advices on how to approach this problem.

thanks !
Michal

[joojaa]

yes you can do this... But theres a suprising answer to all the cutting and splicing, I'm going to enlighten how you could have avoided the entire hustle. THis topic quite much reflect the difference between good problem solvers versus normal ones*.

For some reason people work in this manner, they do the base shape thats the whole mesh, then they consider this the target mesh (and rather literary orbit their thoughts about this unable to see the answer). Then they cut it into 2 to make symmetrical blendshapes. then model and combine back for each and every side (so they end up with the double amount of shapes), it could be more pieces too.

Then comes a curse the fact that the vertex order of their TARGET is different form their 1/2 half blends. And they allways want to proceed to fix the blendshapes (did the problem and solution dawn to you here). What they have failed to notice, is that the blends all have same order**, or a few permutations of orders, and never investigate the truth because they are so set on their target. To elaborate they have:

1 x order mesh and 2-120 y order meshes

Now they always want to make the vertex order of all the meshes match the x order mesh, which sends them running in panic never to think about it again, or go i must use API like you did. Well what about cutting up the TARGET mesh and doing it by exact same recombination like all the others? Well it turns out this will make your target y order.

---

Now when it comes to the API solution, be sure to recognize that you only need to work on the smaller set of differences! Don't work on the bigger. Secondly its easier to reconstruct each mesh by reading positional of the reworked indices than to rewrite them to a copy of the original.

However the closest point solution will not work... You will need some other way of finding this out. Such as embedding a blind data before splitting up, a user set scripted array etc etc.

Tough nothing states you CANNOT write the edge info all you need is to create a long int array and connect it to edge array of your node (its called edge, look it up in your node reference) Note after that you need to redo the faces.

---

• Trick here is not to give the solution any definitive starting constraints, that may or may not be appropriate. Now here i don't consider anything other than a working blendshape the target and not constrained on the imaginary TARGET notion. Understanding the real question (which is not hwat most people ask for) is the solution here.
  ** The trick here is always to do the recombination efforts in similar topographic manner, se Maya does not do this by random, like some of you may think, rather in order. So while you can work as relaxed as you wish, you need to FOCUS on doing the combines and merges in one order, and you'll end up with the same order ALL the times. Note it's usually easier to work ON one halve thats the both sides of your final target, reason being that now each sides vertex order is the same, so they are orderingly interchangeable, now all you must do is remember the order in which to combine (left hand left body right body head), NOTE the left hand notion in combine is spatial so in mirroring don't do it right hand just because your named it left, but rather left as in what you se on screen. Here follows a demonstration of session where i did this manually (yes i know it could be shortened by SO MUCH more by actually developing the script, but it's just a transcrypt animation for your benefit ,very quickly done too):

CODE
file -f -new;
window -e- vis 0 $gCommandWindow;
polyCube -w 1 -h 1 -d 1 -sx 2 -sy 2 -sz 2 -ax 0 1 0 -tx 1 -ch 1;
viewFit;
annotate -tx "Think this is head!" -p 0 0.8 0;
setAttr "annotationShape1.displayArrow" 0;
refresh; pause -sec 1;
select -r pCube1.f[0]
          pCube1.f[2]
          pCube1.f[4]
          pCube1.f[6]
          pCube1.f[8]
          pCube1.f[10]
          pCube1.f[12]
          pCube1.f[14]
          pCube1.f[20:23];
refresh; pause -sec 1;
delete;
setAttr -type "string" annotationShape1.text "Think this is head! (now half a head)";
refresh; pause -sec 1;
select -r pCube1;
duplicate -rr;  
move -r 2 0 0;
viewFit -all;
refresh; pause -sec 1;
duplicate -rr;
move -r 2 0 0;
viewFit -all;
refresh; pause -sec 1;
duplicate -rr;
move -r 2 0 0;
viewFit -all;

refresh; pause -sec 1;

select -r persp;
for ($i=0;$i<4;$i++){
move -r (0.758996/4) (-1.121061/4) (0.0817991/4);
refresh;
}

select -r pCube2.vtx[5] pCube2.vtx[7] pCube2.vtx[9];
refresh; pause -sec 1;
for ($i=0;$i<4;$i++){
move -r 0.05 0.05 0;
refresh; pause -sec 0;
}
pause -sec 1;
select -r pCube3.vtx[3] pCube3.vtx[11] pCube3.vtx[16];
refresh; pause -sec 1;
for ($i=0;$i<4;$i++){
move -r -0.067 0 0;
refresh; pause -sec 0;
}

pause -sec 1;
select -r pCube4.vtx[1];
refresh; pause -sec 1;
for ($i=0;$i<4;$i++){
move -r 0.056 -0.04 0.08;
refresh; pause -sec 0;
}

setAttr -type "string" annotationShape1.text "Lets reasemble ME!";
refresh; pause -sec 1;
setAttr -type "string" annotationShape1.text "Lets reasemble ME.";
refresh; pause -sec 1;
setAttr -type "string" annotationShape1.text "Lets reasemble ME!";
refresh; pause -sec 1;

select -r pCube1;
refresh; pause -sec 1;
duplicate -rr;
duplicate -rr;
duplicate -rr;
duplicate -rr; scale -1 1 1;
duplicate -rr;
duplicate -rr;
duplicate -rr;

refresh; pause -sec 1;
select -r pCube1;
refresh;
select -add pCube11;
refresh; pause -sec 1;
constructionHistory -toggle false; // we dont need the extra shapes
CombinePolygons
refresh; pause -sec 1;
select -r polySurface1.e[8]
          polySurface1.e[10]
          polySurface1.e[12]
          polySurface1.e[14]
          polySurface1.e[16]
          polySurface1.e[18]
          polySurface1.e[20]
          polySurface1.e[22]
          polySurface1.e[36]
          polySurface1.e[38]
          polySurface1.e[40]
          polySurface1.e[42]
          polySurface1.e[44]
          polySurface1.e[46]
          polySurface1.e[48]
          polySurface1.e[50];
refresh;
performPolySewEdge 0;
refresh; pause -sec 1;
select -r pCube2;
setAttr "pCube2.translateX" 0;
duplicate -rr; scale -1 1 1;
select -r pCube5 pCube11;
CombinePolygons;
select -r polySurface2.e[8]
          polySurface2.e[10]
          polySurface2.e[12]
          polySurface2.e[14]
          polySurface2.e[16]
          polySurface2.e[18]
          polySurface2.e[20]
          polySurface2.e[22]
          polySurface2.e[36]
          polySurface2.e[38]
          polySurface2.e[40]
          polySurface2.e[42]
          polySurface2.e[44]
          polySurface2.e[46]
          polySurface2.e[48]
          polySurface2.e[50];
performPolySewEdge 0;
setAttr "polySurface2.translateX" -2;
refresh;
select -r pCube2 pCube8;
CombinePolygons;
select -r polySurface3.e[8]
          polySurface3.e[10]
          polySurface3.e[12]
          polySurface3.e[14]
          polySurface3.e[16]
          polySurface3.e[18]
          polySurface3.e[20]
          polySurface3.e[22]
          polySurface3.e[36]
          polySurface3.e[38]
          polySurface3.e[40]
          polySurface3.e[42]
          polySurface3.e[44]
          polySurface3.e[46]
          polySurface3.e[48]
          polySurface3.e[50];
performPolySewEdge 0;
setAttr "polySurface3.translateX" 2;
select -r pCube3;
setAttr "pCube3.translateX" 0;
duplicate -rr; scale -1 1 1;
select -r pCube6  pCube11;
CombinePolygons;
select -r polySurface4.e[8]
          polySurface4.e[10]
          polySurface4.e[12]
          polySurface4.e[14]
          polySurface4.e[16]
          polySurface4.e[18]
          polySurface4.e[20]
          polySurface4.e[22]
          polySurface4.e[36]
          polySurface4.e[38]
          polySurface4.e[40]
          polySurface4.e[42]
          polySurface4.e[44]
          polySurface4.e[46]
          polySurface4.e[48]
          polySurface4.e[50];
performPolySewEdge 0;
setAttr "polySurface4.translateX" -4;
select -r pCube3 pCube9;
CombinePolygons;
select -r polySurface5.e[8]
          polySurface5.e[10]
          polySurface5.e[12]
          polySurface5.e[14]
          polySurface5.e[16]
          polySurface5.e[18]
          polySurface5.e[20]
          polySurface5.e[22]
          polySurface5.e[36]
          polySurface5.e[38]
          polySurface5.e[40]
          polySurface5.e[42]
          polySurface5.e[44]
          polySurface5.e[46]
          polySurface5.e[48]
          polySurface5.e[50];
performPolySewEdge 0;
setAttr "polySurface5.translateX" 4;
select -r pCube4;
setAttr "pCube4.translateX" 0;
duplicate -rr; scale -1 1 1;
select -r pCube7 pCube11;
CombinePolygons;
select -r polySurface6.e[8]
          polySurface6.e[10]
          polySurface6.e[12]
          polySurface6.e[14]
          polySurface6.e[16]
          polySurface6.e[18]
          polySurface6.e[20]
          polySurface6.e[22]
          polySurface6.e[36]
          polySurface6.e[38]
          polySurface6.e[40]
          polySurface6.e[42]
          polySurface6.e[44]
          polySurface6.e[46]
          polySurface6.e[48]
          polySurface6.e[50];
performPolySewEdge 0;
setAttr "polySurface6.translateX" -6;
select -r pCube4 pCube10;
CombinePolygons;
select -r polySurface7.e[8]
          polySurface7.e[10]
          polySurface7.e[12]
          polySurface7.e[14]
          polySurface7.e[16]
          polySurface7.e[18]
          polySurface7.e[20]
          polySurface7.e[22]
          polySurface7.e[36]
          polySurface7.e[38]
          polySurface7.e[40]
          polySurface7.e[42]
          polySurface7.e[44]
          polySurface7.e[46]
          polySurface7.e[48]
          polySurface7.e[50];
performPolySewEdge 0;
setAttr "polySurface7.translateX" 6;
viewFit -all;
refresh; pause -sec 1;

setAttr -type "string" annotationShape1.text "Now some center symmetries!";
refresh; pause -sec 1;
setAttr -type "string" annotationShape1.text "Now some center symmetries.";
refresh; pause -sec 1;
setAttr -type "string" annotationShape1.text "Now some center symmetries!";
refresh; pause -sec 1;
select -r polySurface1;
duplicate -rr;
move -r 0 2 0;
refresh; pause -sec 1;
duplicate -rr;
move -r 0 2 0;
refresh; pause -sec 1;
select -r polySurface8.vtx[4] polySurface8.vtx[6] polySurface8.vtx[8];
move -r -ls -wd 0 0.414941 0;
refresh;
select -r polySurface9.vtx[2];
move -r -ls -wd 0 0 1.729639;

setAttr -type "string" annotationShape1.text "DONE! blend me to chek indexes";

Paste and execute! don't be supprised each and every CUBE has the exact same order even tough i've cut it about and recombined.

[user_deleted]

Hi !

Thanks alot for an extensive answer !
However I should have provided more details justyfying my intentions. Here's brief description of my mini pipeline:
I have two separate files: a skinned character (bones, muscles syflex'es and so on ) and a shading file with the same geometry and assigned shaders. The benefit here is, that at the same time two artist can work on skinning and UV mapping (different scenes). Later the animation is being transfered using deformationCaching plugin, that works based on vertexorder. The danger is that UV layouting is often performed in external app, and reimporting geometry (OBJ) chages that vertOrder. This could be solved with wrapDeforming shadingGeometry with geo imported from skinning file, but it's neither effective nor elegant. Tonight i'll invesitgate API little bit more

As to a blendShapes problem... I'd like to share a trick that i used to create corrective deformation blendShapes for both sides. Nothing spectacular,but someone might find it usefull.
To create a mirror copy of a blendShape target i create animated mesh which morphs from baseMesh to a target, mirror this by scaling -1 in one axis, and use as a wrap deformer on a copy of base mesh. This way i don't have to cut anything, and i can do meshes that are not perfectly symmetrical

Summing things up... I could get away without reordering vertices, but our work comfort will be much higher with this tool. It could save my ass from time to time :>. Another solution could be some clever way of copying UVs between unordered meshes...

thanks again, Michal

EDIT:
i just tried the mel you pasted. WOW this is cool, really !

[joojaa]

Yes i know of the trick its just pushing the problem solving one step ahead of whats neccesery! BUt working on the halves actualy makes for much better workflow after youve automated the assembler, you now have a halve to full builders and a full that will allways match, so you never NEED to have wholes for all shapes you can allways generate them. thus making changes work faster. (also did you know you can just inject the changed shape form another file to a blend using mel easily)

Swaping the uves is actualy easier, faster and safer because its about a 2 value per vertexface index array + aanother for the vertex uv only + a ndicator if i remeber this right of he bat (in the barebones solution) and thus requires you to do WAY less work. for quad mesh its 4*2+2+1of data to be copied about. For full mesh reconstruction its 3+3+1+4*2+2+1+8+4 operations in addition to findig target.

To be totaly honest you can do this quite easily in mel (and quite fast too, sorry no maya at hand)

But in api all you need to do is following pesudocode

do for all verts
find corresponding verex
get its uv indices (both per vertex starus and the per vertexface one)
copy it over to target meshes uv index
loop

While your at it make it so you can do this on a mesh from the file directly becasuse it sounds like that would be evan handier for your pipe,

If you havent by now you should read the mesh nodes node reference... Supprisingly the aswer of what you need to do is THERE. Sure api has hooks for this but its usualy easir to just write the damn data like described in the node ref. (because thetway you can get the aswer in one go!)

PS. Well i needed to find a way to communicate with very *&

[user_deleted]

OK, tonight i tried API some more, and skimmed through mesh node docs. I decided to try acting on edge plug directly. Here's code fragment, not working:

CODE
        MPlug plugEdges= dnFnMessed.findPlug("edge", stat);
    for(unsigned i= 0; i<plugEdges.numElements(); ++i) {
         MPlug plugEdge= plugEdges.elementByLogicalIndex(i, &stat);//doesn't work with physicalIndex,why

                 MPlug se= plugEdge.child(unsigned(0), &stat);//start of edge
         MPlug ee= plugEdge.child(unsigned(1), &stat);//end of edge

         int s, e;
         stat= se.getValue(s);//(kFailure): Unexpected Internal Failure
         stat= ee.getValue(e);//(kFailure): Unexpected Internal Failure        
    }

why do i get those unexpected internals ? I can't read those attrs in MEL either, what's the syntax for getAttr in this case?

-michal

[joojaa]

why do you insist in doing 10 times the work... just work on uvs. And no i said the uv thing can be done in mel i never said you could do the edge array stuff in mel, mel cant acces long3 format stuff but can acces float stuff. Besides the edge array must come in at creation time. maya does not update it if not.

you can get all uves in one go (wich is 10-100 times faster than getting them one by one)

getAttr pCubeShape1.uv[0:getAttr -s pCubeShape1.uv-1];

I mean you have allmost no benefit of this anyway all you need is to fetch UVS and if you need blend targets (yeah i know zbrush) then just move over the sorted position array to the blend node directly or to a copy of the original topo mesh wih is allready in the right order (this si usefull because you can directly transfeer the data to YOUR allready exsisting rig without doing EXTRA history steps). I mean its infinitely simpler form your point of view. Thisway dont need to DO the reorder anything in the edge array and form the users point of view it the same thing.

However if you NEED to reorder aeverything because you must do exactly like your badly desiggned parameter array says do the whole mesh form scratch using. OR you have some other design parameter yoru not telling me. The only reason youd do a full fledged reorder plug is if you wanted some arbitrary nonexistent order, in wich case its justified.

THEN use

MObject MFnMesh:: create ( int numVertices , int numPolygons , const MPointArray &vertexArray, const MIntArray &polygonCounts, const MIntArray &polygonConnects, MObject parentOrOwner, MStatus * ReturnStatus )

but as i asaid justt trasfeering the uvs is a REALY good alternative since its the only thing you will have hard time trasfering rest is piece of cake. AND you allready DO have a right edge order at your disposal, all you need is to redo the data therein

So what your doing here is from your description

CODE
[data]->[data]->[blendshape]->[targetdata]
[data]->[data]-> .                    [targetdata]
[data]->[data]->[transfer UV]->[targetdata]
[data]->[data]-> .               [targetdata]

while im tellig you to do:

CODE
[data]->[data]->[blendshape]->targetdata
[data]->[data]->targetdata

[mike-b--3]

Hey Joojaa,

Is there any chance you could update the script so it works in at least maya 2015?  I'm trying to wrap my head around these concepts, so I'm not really able to fix the script myself because I'm not even sure what its doing haha.  I'm trying to figure out how to match vertex numbes between 2 similar meshes, will this help me out?

Thanks!