If we had a make manifold node, then do the boolon, then delete new genereated new faces from make manifold, some of these cases could still work in teori.
Yes, adding a Solidify Modifier to the operator is enough, the question is, is it worth it to automatically close non-solid shapes?
It is expected and documented that the Manifold Boolean will do nothing if any argument is not manifold. (One exception: argument 2 of a difference can be a simple plane.)
Manifold doesn’t really decimate, but it will “Simplify”, which means merging adjacent coplanar faces. Is that a useful function that isn’t available elsewhere?
There is no “make manifold” functionality in the Manifold library. I agree that it would be useful to make such a thing. It won’t be easy. One possible component of such code would be a hole filler.
1 Like
“It is expected and documented”
Perfect.
“Manifold doesn’t really decimate, but it will “Simplify”, which means merging adjacent coplanar faces. Is that a useful function that isn’t available elsewhere?”
Decimate/Planar/AngleLimit=0.1?
“One possible component of such code would be a hole filler.”
It’s worth considering.
Test A: I’ve found that Manifold preserves the operator’s material, which other methods don’t. Brilliant!
Test B: I have crossed the mesh, to see if it stops working when the normals are inverted, and at least it continues working in the areas that are not inverted, the other methods don’t. Better again.
Please attach the blend file you used to create that image to the original issue and I’ll look at it again. Maybe this is another, but related, problem.
It might be.
I’ve noticed this solver often creates faces that overlap or span between other faces like so ^
This could be a culprit perhaps? I’m not sure of course, I know nothing of the code.
But in the post just above using Suzzane you can see what I mean.
Nontheless I have attached two files in my original bug report:
1 Like
Pablo discovered in todays Blender Today live that Manifold is missing from edit mode’s Boolean operator (mesh.intersect_boolean)
Yeah, it was somewhat intentional on my part to leave it out of the edit mode boolean. I would have to convert everything from Mesh to BMesh, and then back again. Yeah, I can do it, but was a little bit waiting to see if users missed it. It is unclear to me how many people actually use edit mode boolean compared to modifier or geometry nodes versions of boolean. Also, since edit mode is more “single shot”, maybe in most cases the slowness of the Exact solver is acceptable in most cases for edit mode boolean.
But I do agree that from a “comprehensive and comprehensible product” point of view, it would make sense to add this solver to boolean edit mode too, and I may eventually do that (or someone else could…).
5 Likes
About a topic further up: the Manifold library has a PR that will fix the slowdown I had to introduce in order to prevent merging of coplanar faces that had original different faces in Blender. So once that is merged and when we get a new version of the manifold library into our dependencies, we should get faster again.
19 Likes
Hi Howard, first of all, thanks a lot for working on integrating this awesome new booleon to Blender! 
Super exciting stuff!
Has the new manifold libery update been added to Blender that fixes the slowdown yet? Or if not, will it happen before 4.5 Beta?
The have committed their PR that would allow me to fix the slowdown, but we have not yet rebuilt the blender dependency manifold library to incorporate that fix. Our platform maintainers have been waiting to batch up a bunch of rebuilds, and, I think, to see if any other fixes come in to the library. It would be nice if Manifold did an official release including that fix, and that may happen soon, but that is not mandatory for us to get the fix.
I am expecting that we should have this fix before 4.5 Beta. If not then, then definitely before the 4.5 release.
20 Likes
Thats fantastic to hear! Yay! 
: Thank you again Howard!
The new solver seems to succeed in every case I tested as long as the mesh is manifold.
Can’t compare the speed with other solvers beacause they just fail.
Great Job
4 Likes
Yes, very snappy. A bit faster than float but doesn’t fail as float. You can even animate it. Didn’t have an issue both in sculpt and GN.
15 Likes