Definitely looks better in your example. 😄

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.

Reason I haven't merged it yet is because of backward compatibility. Do you think it'll break existing apps?

@Shchvova Yeah, maybe the // center mesh block wants to be in a if (fromVertices) condition. Even if nothing did break, it would avoid redoing the calculation.

Actually both of them need the center mesh code, I took it from the non-fromVertices and now it's being applied to both cases (pure Lua and buffered)

This is an initialization call, so centering the mesh will always be required.

Also the pure Lua version logic remains intact, the one that changed is the one using buffers so there shouldn't be any danger in this change.

@Shchvova I was looking over some of the code with @depilz and it seems like the vertex offset calculation is almost replicated in the tessellator's Update() method; I think the only difference is that the latter is done on the already-offseted vertices.

This doesn't affect the correctness, but seems like a legitimate optimization—meshes can have plenty of points, after all. Off-hand I didn't see if the situation would always occur or if the rect would want to be an optional input to Update(). (I could probably supply some some rough code in the next day or two.)

@Shchvova 20ee14a will prevent unnecessary updates when using regular tables to update a mesh.

Testing branch using lua tables to update meshes.

Testing branch using buffered memory data to update meshes.

They both look equally good but buffered is definitely a great improvement in performance .

@Shchvova this branch is ready to go. Can we merge it in?