Add BMesh and WM symmetrize operators

The symmetrize operation makes the input mesh elements symmetrical, but unlike mirroring it only copies in one direction. The edges and faces that cross the plane of symmetry are split as needed to enforce symmetry.

The symmetrize operator can be controlled with the "direction" property, which combines the choices of symmetry plane and positive-negative/negative-positive. The enum for this is BMO_SymmDirection.

Documentation: http://wiki.blender.org/index.php/User:Nicholasbishop/Symmetrize

[brechtvl, 2012-10-12 10:47:24]

This is really quite cool, I was expecting a simple vertex coordinates
symmetrize but it can actually mirror topology! Worked well in testing.

Some comments below.

https://codereview.appspot.com/6618059/diff/1/source/blender/bmesh/intern/bme...
File source/blender/bmesh/intern/bmesh_opdefines.c (right):

https://codereview.appspot.com/6618059/diff/1/source/blender/bmesh/intern/bme...
source/blender/bmesh/intern/bmesh_opdefines.c:1188: * Description is TODO
Guess this todo and the one a few lines lower is not difficult to fix?

https://codereview.appspot.com/6618059/diff/1/source/blender/bmesh/operators/...
File source/blender/bmesh/operators/bmo_symmetrize.c (right):

https://codereview.appspot.com/6618059/diff/1/source/blender/bmesh/operators/...
source/blender/bmesh/operators/bmo_symmetrize.c:38: #define SYMM_VERT_THRESHOLD
0.001f
I think these values are quite high, mirror for transform and particles uses
0.00002f (meshtools.c MOC_THRESH). For Sintel's head model I remember Campbell
set it to this low value because it didn't work with small detail. Not sure if
it should be user adjustable but probably it should be smaller?

https://codereview.appspot.com/6618059/diff/1/source/blender/bmesh/operators/...
source/blender/bmesh/operators/bmo_symmetrize.c:174: BLI_assert(r);
I have a feeling there is a corner case where r = 0 and lambda is used
uninitialized. I didn't try to recreate it, but it probably exists.

https://codereview.appspot.com/6618059/diff/1/source/blender/bmesh/operators/...
source/blender/bmesh/operators/bmo_symmetrize.c:196: if
(BLI_ghash_haskey(symm->vert_symm_map, e->v1))
This could call ghash_lookup without ghash_haskey first, will just return NULL.

https://codereview.appspot.com/6618059/diff/1/source/blender/editors/mesh/edi...
File source/blender/editors/mesh/editmesh_tools.c (right):

https://codereview.appspot.com/6618059/diff/1/source/blender/editors/mesh/edi...
source/blender/editors/mesh/editmesh_tools.c:5444: ot->prop =
RNA_def_enum(ot->srna, "direction", direction_items, 0,
It would be slightly more clear if 0 is replaced with BMO_SYMMETRIZE_NEGATIVE_X
here.

[nicholasbishop, 2012-10-14 01:43:03]

Update from review

[nicholasbishop, 2012-10-14 01:43:59]

> Guess this todo and the one a few lines lower is not difficult to fix?
Oops, forgot about this. Fixed.

> I think these values are quite high, mirror for transform and particles uses
> 0.00002f (meshtools.c MOC_THRESH). For Sintel's head model I remember Campbell
> set it to this low value because it didn't work with small detail. Not sure if
> it should be user adjustable but probably it should be smaller?
Changed to 0.00002f.

> I have a feeling there is a corner case where r = 0 and lambda is used
> uninitialized. I didn't try to recreate it, but it probably exists.
Added initialization to zero for lambda.

> This could call ghash_lookup without ghash_haskey first, will just return
NULL.
Removed the unnecessary BLI_ghash_haskey() calls.

> It would be slightly more clear if 0 is replaced with
BMO_SYMMETRIZE_NEGATIVE_X
> here.
Fixed.

I made one additional change: symmetrical edges with no faces attached were
getting deleted. Pulled out some code from symm_split_asymmetric_edges() into a
new symm_edge_crosses_axis() function, and in symm_kill_unused() check that the
edge is not symmetric.

Thanks for reviewing!

[nicholasbishop, 2012-10-14 03:02:10]

Fix for faces not being deleted when it has vertices on the axis but none on the
keep side

[brechtvl, 2012-10-15 09:52:43]

LGTM

Forgot one thing, this tool doesn't seem to be in any menu, but maybe that's
just missing from the patch.

[brechtvl, 2012-10-15 09:52:43]

LGTM

Forgot one thing, this tool doesn't seem to be in any menu, but maybe that's
just missing from the patch.

[nicholasbishop, 2012-10-15 13:58:26]

On 2012/10/15 09:52:43, brechtvl wrote:
Good point, what menu do you think it should go in?

> LGTM
> 
> Forgot one thing, this tool doesn't seem to be in any menu, but maybe that's
> just missing from the patch.

[brechtvl, 2012-10-15 14:44:59]

There doesn't seem to be a perfect for any of the submenus (maybe Mirror, but
not really). I'd put it in the top level Mesh menu and the W key specials menu.

[nicholasbishop, 2012-10-15 23:55:15]

Added the menu entries and committed as r51356.

Thanks for reviewing :)