Chanhyeok Park

``` manifolds = [] for i in range(50): pts = abs(np.random.normal(0, 0.1, 3)) length = np.random.normal(0, 90, 3) manifolds.append( Manifold.cube(pts[0], pts[1], pts[2]).rotate(length[0], length[1], length[2]) ) @profile(precision=5) def my_func(): c =...

Nope! It shows the same behavior I just functioned (instead of direct for loop) it to use the memory profiler which is one of the python libraries. I debuged and...

Thanks for the suggestion! I'll try to make them with C++ and check if it reproduces the result and the leaks. If it does, I'll have to find it out...

I think I found the problem! I think it's not the leaks but the wrong behavior of the library. ``` int main(int argc, char **argv) { random_device rd; default_random_engine gen{...

I checked that the mesh are consistent. (Have the same number of vertices in the all the iterations) And tried to copy the manifold with AsOriginal() when assigning to c...

> Unless c and manifolds[i] are identical, then you should expect c = c + manifolds[i] to increase the complexity of c. > > So if the number of vertices...

I found the problem! First, the memory increase is coming from the large vector includesMeshID. In the function IncrementMeshIDs(), includesMeshID array size keep increases meaning that it does not decreases...

Oh, I think map is the right chose. It's hard to track all the mesh numbers with vector design when meshes are randomly allocated and freed.

Oh I got distracted by some other project while working on it. Since I didn't make any meaningful changes yet maybe @pca006132 can work on it!