Large clumps are not clumped properly

[vdbma]

Expected behaviour

One large clump (set of adjacent masked cells) should be gathered in one single clump

Actual behaviour:

Not the case see bellow

Working example:

(not so minimal)

from clump_tracker.finder import find_clumps
import numpy as np
from dataclasses import dataclass
import numpy as np
from functools import cached_property



@dataclass
class FakeVTK:
    xE: np.ndarray
    yE: np.ndarray
    zE:np.ndarray
    data: dict[np.ndarray]
    
    @cached_property
    def x(self):
        return np.convolve([0.5,0.5],self.xE,mode="valid")
        
    @cached_property
    def y(self):
        return np.convolve([0.5,0.5],self.yE,mode="valid")

    @cached_property
    def z(self):
        return np.convolve([0.5,0.5],self.zE,mode="valid")


def _data(Nx,Ny,Nz):
    return {"RHO":np.zeros((Nx,Ny,Nz)),
           "VX1":np.zeros((Nx,Ny,Nz)),
           "VX2":np.zeros((Nx,Ny,Nz))}



def oneBump(Nx=1024,Ny=1024,Nz=1):
    
    V = FakeVTK(np.linspace(-10,10,Nx+1),
               np.linspace(-10,10,Ny+1),
                np.linspace(-.5,.5,Nz+1),
               _data(Nx,Ny,Nz))

    xx,yy,zz = np.meshgrid(V.x,V.y,V.z,indexing="ij")

    V.data["RHO"] += np.exp(-(xx*xx+yy*yy))*2

    return V

def condition(data):
    return np.asarray(data["RHO"] > 1)


V = oneBump(2048,2048)


coords = [V.x,V.y,V.z]
dcoords = [np.ediff1d(_) for _ in [V.xE,V.yE,V.zE]]

max_distance = 1.1*np.ediff1d(V.xE)[0]
clumps = find_clumps(V.data,*coords,*dcoords,
                     float(max_distance),gamma=2,condition=condition)

print(f"{len(clumps) = } should be 1")

len(clumps) = 50 should be 1

Illustration:

import matplotlib.pyplot as plt

xxE,yyE = np.meshgrid(V.xE,V.yE,indexing="ij")
fig, ax = plt.subplots()
for cl in clumps:
    ax.plot(cl.x,cl.y,"r+",alpha=.5)
c = ax.pcolormesh(xxE,yyE,condition(V.data)[:,:,0],cmap="binary")
ax.set_aspect("equal")

fig.colorbar(c)

There should be only one single red cross in the middle of the black disc.

[vdbma]

See #25