[Bf-taskforce25] BMesh integration status

[joe]

I've done quite a lot of work at integrating bmesh into the 2.5
codebase.  I've replaced EditMesh with BMesh (though not completely as
yet, a lot of stuff still uses EditMesh), changed mesh DNA, modified
DerivedMesh and made a good start at getting all the modifiers working
with ngons, and a bunch of other stuff.

BMesh doesn't directly replace EditMesh.  Instead, I created a wrapper
structure around BMesh called BMEditMesh, that stores a bmesh, it's
triangle tesselation, and a bunch of stuff that didn't belong in the
BMesh structure itself.  I've ported a lot of tools to this new
infrastructure, though there's still a lot to be done.  I've made the
remaining EditMesh tools interface with BMesh, though this hasn't
worked out well; I hope to port over all existing EditMesh tools
before commencing serious user testing.

The changes I made to mesh DNA are pretty much as we discussed
earlier.  MFaces are now used to store the tesselation of the mesh,
while a new layer (MPolys and MLoops) are used to store real faces.
MPolys logically represent a face, while the face boundary itself is
defined by a series of MLoops (stored in one contiguous array for all
faces).

The new face structures are:

typedef struct MPoly {
    int loopstart;
    int mat_nr;
    int totloop;
    int flag;
} MPoly;

typedef struct MLoop {
//vert, edge indices
    int v, e;
} MLoop;

Mesh {
<all the stuff in Mesh>
    MPoly *mpoly;
    MLoop *mloop;
}

MPoly->loopstart points into mesh->mloop.  MPoly->totloop defines the
number of MLoops, and together the mpoly and the mloops define the
boundary of the face.

On the DerivedMesh side, I added simple face iterators to DerivedMesh.
 I did this for two reasons; I wanted to evaluate the possibility of
porting DerivedMesh to be entirely iterator-based, and it also seemed
like using iterators would be easier then matching the design of the
rest of DerivedMesh, given the new split nature of faces into
MPolys/MLoops.  I'm no longer convinced this was a good idea, and I
may rewrite it if I get the chance.

The Subsurf, array, and mirror modifiers are now ngon-aware.  Subsurf
still has some work left to be done (e.g. with CustomData
interpolation).  Both array and mirror now use bmesh; array is mostly
working, but has a few minor things done, while mirror is pretty much
finished.

At this point, I've started tackling the UV editor/unwrapper code.
UVs are now stored per-loop, and I've managed to port much of the uv
editor code to this new representation in the past few days.  My plan
is to finish this, port over vertex/weight paint, and then concentrate
for a while on stabilizing mesh editmode; rewrite old EditMesh tools
in BMesh, fix bugs, and in general get that stabilized.  Once I've
gotten this next bit with the UV and vertex/weight paint code done, I
figure I'll be finished with the major refactors, and can concentrate
on getting what I already have stabilized.

Joe