wow, seems to work quite well already! blender 2.5 + ngons will rock. 
That was a great video, I can really see the benefits to using Bmesh and ripping editmesh out of Blender.
Weāll all be able to make thingimabobbies (how he says it) with no smoothing errors. But the correct smoothing on non-manifold meshes will have benefits if it can be supported in the BGE (where you need to watch polycount) as well like for Dragon wing membranes and wire fencing.
Just a bit of clarification hereā¦
Bmesh doesnāt make non-orientable surfaces orientable (this is where it differs from Radial edge/Partial entity which tries to treat non-manifolds as collections of orientable āshellsā). If you look at the example closely its actually a bit of an odd case, but if you did ārecalculate normalsā on that mesh it probably would break pretty quick (unless Joeedh did something really clever with that codeā¦)
Regardless, great to see mfoxdogg contributing his talents to a bmesh feature tour 
Cheers,
Briggs
Iāve been an artist in video games for 15 yearsā¦ never( ever ever) have I had a problem with ngons in any packageā¦ (blender not having them has caused me frustration!)
they help you keep a clean meshā¦ but whateverā¦
to answer your question: looking at the subdivide it can auto tesselateā¦ knife doesnāt seem to be in at the moment, but i guess thatād be an option tooā¦ the only other way to create ngons would be to dissolve edges or vertsā¦ but what would you expect?
as mfoxdog says, these preferences cannot be saved YETā¦ (same as keymaps, but I guess thatāll be in by release)
You can always triangulate the mesh anyway and iām sure itād be trivial to add a āselect all faces with more than 4 edgesā operatorā¦ if it doesnāt ship with one Iād writ one in python myself.
Playing with the latest Bmesh build, and its working VERY nice.
āYā is great, and this combined with Vertex+Edge select mode, dissolve & subdivide it can make the work flow so much nicer indeed.
Damn, I can never get used to all these improvementsā¦
Very awsome work!
Just a question: will ngons also work in the gameengine and is there a possible fps benefit from it?
benefit? probably not, would they work, well, I imagine everything will be rendered(rasterised?) on the video card as triangles either wayās so why wouldnāt it?
Looks pretty impressive if ngons are your thing. That said, count me in with Dim, Iād rather not have them arise spontaneously either. But Briggs once told me not to worry, so I wonāt .
One question though, why is connect the verts set to Y instead of F, which usually connects verts? Itās not as if thereās an overlap between the two.
I suspect the reasoning would be that there may be cases where you would want to create a new edge that does not bisect the ngon. (i.e. āFā) Whereas āYā does bisect the ngon.
Why these kinds of posts stubbornly keep popping up is beyond me. Aside from other Bmeshās technical advantages, n-gons are just extra aids in the modeling phase. They could effectively eliminate the current workarounds. They appear not because of the program, but because of your own actions. You donāt have to end up with n-gons. You usually re-connect verts when dissolving edges/verts as you go along for various reasons as redirecting edge flows or topology adjustments. I repeat: you donāt have to leave them as n-gons, you reconnect verts and leave them with quads or tris.
I donāt know why these posts pop up when virtually all 3d apps are n-gon-capable, including Wings3d. If you donāt like n-gons, use Blender 2.49b or earlier. Simple as that.
Anyway, seems Bmesh is beginning to take shape. Blender 2.5ās release might end up with Bmesh in place, after all.:yes:
EDIT: Am I missing something about n-gons or what? Iād appreciate any info where exactly these n-gon anxieties come from. Any workflow samples describing such?
Considering this is an option in just about every other 3d application including blender, I can confidently say yes.
I donāt know anything about Maya but I do know a thing or two about 3ds Max. Having worked with N-gons for some time let me just say that this will not be a problem. You do need to do some extra thinking to make quads when using the cut tool but for everything else the tools will almost never produce N-gons where you donāt want them.
Iād like to clarify the issue here. The problem isnāt that they canāt be connected - everybody understands that ngons are just a tool that can be turned into triangles at will. The thing is, I often start out with a cube and then over the whole modeling process keep subdiving edges until at one point Iāve got a 3000 polygon head. As youāll imagine, thatās quite a few subdivisions over the entire thing and one second every time to make tris or quads out of it quickly adds up. Heck, even if you donāt work in the thousands, turning that tesselation on every time is aggravating to say the least. All weāre asking for is the possibility of having the option of having tesselation on by default.
No need to argue about this, everyone can be made happy here I think
There will be no āauto-tesselationā as a global setting, since this would require a lot of decisions being made for users which may or may not suit their purposes. Users will have to understand the tools well enough to choose the options they want for each one. However with 2.5 there should be a system in place where you only have to do this once and then its saved in some kind of defaults for you.
Cheers,
Briggs
Iām sorry, I have a hard time imagining what you are describing here. Can you post a screen shot showing at least 4-6 stages of your workflow? (A short video migth be better, if possible.) Who knows, maybe weāre really missing something here. It might even be something that the devs could find essential and worth adding.
Iām afraid Iāve got no work-in-progress models left anywhere to show, but it isnāt too complicated or ground-breaking. Itās just that when modeling low-poly, you donāt want to add whole edge loops, so you select a couple of edges and subdivide them to create a mini-loop; which would always leave you with an ngon on each end. And itās usually good habit to have triangles in such areas, but auto-triangulation afterwards often has very different ideas from you as to what is a good place to place triangles.
If it works like Briggs says, though, Iām a happy Zwebbie and thereās no reason for any further discussion. And to be fair, I am a bit excited to actually see if ngons can change my workflow .
Two years using Blender, and 2.5 makes me feel like a dinosaur.
All be it, a very happy dinosaur. :yes:
Uhm,Bmesh seems 4 time slower in subdivision than the old data structure.
Without subdivision,itās also much slower,as with a medium mesh I canāt use mirror modifier with fast redraw(I mean acceptable fast).
WIth 2.49(or svn 2.5) I can use a mirror modifier and a subsurf modifier(2 level) on a 18845 quad mesh with a fast redraw,with Bmesh I canāt use subdivision and I canāt use mirror to match the same speedā¦
From what I understand bmesh should be faster in finding connectivity information(edge loops) but ,the base of modelling is moving points,itās too slow for that.
For now Itās too slow for my taste,I hope that the old edit mesh will continue to be used.
EDIT:
Maybe some partial redraw stuff could be used to improve performance.
Iād rather hope bmesh gets finally in and gets optimized - mirror is still probably going the way of conversion to old mesh and backā¦
Itās better FIRST optimizing and AFTER,when performance are acceptable,discarding the old edit mesh,the true is that for now performance are not good enough
Now there are some kind of models practically impossible to do(but really possible in 2.49)
hehe every coder goes by the same mantra, make it work first make it pretty later (including optimizations )
Thanks all for your wonderful comments, its those that keep me doing these demoes