Inclusive Inverse World Matrix?

[bluntman]

I am trying to write a function in a program I am developing which exports .ma files including skinning information. I am having problems with the bind pose settings, specifically the .pm (preBindMatrix) attribute of the skinCluster. Maya help simply says this is "The inclusive matrix inverse of the driving transform at the time of bind." Well I have tried providing the inverse of the world space matrix of the joint at the time of the bind as this parameter, and it doesn't match at all with what Maya thinks it should be. How should this inverse inclusive matrix be calculated? (I am NOT using the Maya SDK, only my own matrix functions, which I know to work correctly).
Example of the discrepancy between my output and mayas:
My output:
setAttr ".pm[0]" -type "matrix"
0.820887 0.494885 0.285016 0.000000
-0.570779 0.727444 0.380837 0.000000
-0.018862 -0.475305 0.879619 0.000000
-1.000246 -0.989063 -0.350760 1.000000;
Mayas output:
setAttr ".pm[0]" -type "matrix"
0.820888 -0.50960 -0.25777 0
0.570777 0.717203 0.39978 0
-0.01885 -0.475314 0.87961 0
-1.32233 0.430686 0.40951 1;

Any help would be very much appriciated, I am really stuck, and there seems to be no detailed information about this anywhere.
Thanks!

[joojaa]

And you are using mel because of??? But i doubt maya calculates the matrix worng. Since if it would youd have ALOT of problems allrready.

Anyway you could eb a bit off becaue maya has every peculiar notion of localness

[rob-bateman]

QUOTE(Joojaa @ 08/21/06, 04:52 AM) [snapback]245689[/snapback]
And you are using mel because of???

He's not using mel he's writing his own program (seperate from Maya) to write Maya Ascii files, which i assume is in C++....

It looks like the handedness of your Matrix functions is the other way around to Maya. You probably want to convert the handedness then write. It looks like that should solve it.

[bluntman]

Thanks for the replies! Yes I am writing my own program in C++ to export Maya ASCII files, the reason I posted in MEL section is a) because .ma files use MEL commands to construct the scene, and because this is where I think I am most likely to find people who can help with a problem of this nature.
Also I wasn't trying to imply that the Maya SDK would transform matricies incorrectly when i said " (I am NOT using the Maya SDK, only my own matrix functions, which I know to work correctly)." only that, even though I am not using the Maya SDK to do my matrix maths, I know that my own functions work correctly, so the problem is not there.
I am exporting other matricies (for instance for the transform nodes of meshes) and these match up exactly with the output Maya gives. My program does use different handedness from Maya, Z is up, Y is into the screen and X is to the right. I'm not sure of the method of converting a matrix from left to right handed or vice versa, any ideas?

Thanks!

/edit
Also any chance of a definition of Mayas strange notion of localness?

[joojaa]

Bateman yes i know that hes witing a ma file. I do it all the time

QUOTE(bluntman @ 08/22/06, 11:48 AM) [snapback]245808[/snapback]
a) because .ma files use MEL commands to construct the scene

Actualy its just a bit of a subset of mel And indeed the ma file contains stuff thats not valid in mel! Or atleast its not doable inside a maya session.

Still i vote thei sia api question, because api is about program development, regardless of the fact thet your wiriting MA file.

Yes, bateman is right your handedness looks opposite. Your calculating a typical architecture space while mayas using what most eingeeneers use. this means that your rotations are opposite amongst other things (sich as z being in opposite direction)

Wonky:

Maya uses the notion that pivot isnt necceserily at ttransfrorm this maens maya may show you a objects transofrm hadle at a differebnt place then it is. But this does not affect bones.

[bluntman]

Okay I don't think this is the case. I have to provide a matrix which defines the pivot position of the mesh in the bind pose and my matrix matches up perfectly with the one Maya expects. The only matrix that doesn't match is the bind pose matrix which is an inverse. So I think my problem is in the inverse operation.
I thought I understood the concept of an inverse matrix, but now I am not so sure. The inverse matrix of a transform is the matrix which, when applied, takes the transform back to the origin. This is correct isn't it?
If this is the case then the translation components of the inverse matrix should be simply the negated values of the transform matrix, which in the 2 example matricies I provided isn't the case. It seems like the matrix Maya actually requires isn't simply the inverse of the global transform, if I understand it correctly.

[joojaa]

matrix inversion has nothng to do with negating values. Inverse is defined as the matrix B that multipliesd by A produces identity.

A*B=I

anyway read http://mathworld.wolfram.com/MatrixInverse.html

i still wote ite handedness. wich sisint so easy to reverse! its a definition thing. its like me saying counterclokwise is thetway and your saying its in other because you look into opposite direction of me.

[bluntman]

I still don't see how it can be a handedness problem if all the other matricies I write to .MA file work correctly!! I have checked to make sure my inverse matrix is correct (i.e. A * B = I) and it definitely is the inverse! Does anyone have any other suggestions of an explanation of why all my matricies work except this one? Or can someone direct me to where I can find more information on how maya is calculating this "bindPreMatrix"?
I think the "Pre" part of the name might be a hint indicating that the matrix is used in a pre-mutiply context which iirc changes the way it should be constructed.
This is driving me nuts!!!!