Hi all! I was wondering which is the correct way of computing the maximum of a subfunction in a MixedElement space. Here I provide a minimal working example to illustrate my doubt:

```
from dolfinx import fem, mesh
import ufl
import numpy as np
from mpi4py import MPI
msh = mesh.create_rectangle(comm=MPI.COMM_WORLD,
points=((0.0, 0.0), (1.0, 1.0)), n=(2, 2),
cell_type=mesh.CellType.triangle,)
P1c = ufl.FiniteElement("Lagrange", msh.ufl_cell(), 1)
V = fem.FunctionSpace(msh, ufl.MixedElement([P1c, P1c]))
_, map1 = V.sub(0).collapse() # Mapping of the first subfunction
_, map2 = V.sub(1).collapse() # Mapping of the second subfunction
u = fem.Function(V)
np.random.seed(0)
u.sub(0).interpolate(lambda x: (np.random.rand(x.shape[1])))
u.sub(1).x.array[map2] = 2.0
u.x.scatter_forward()
u1, u2 = u.split()
print(u.x.array[map1].max())
print(u1.collapse().vector.max()[1])
```

I was expecting the last two lines to provide the same result but, actually, what one obtains is:

```
0.8917730007820798
0.8700121482468192
```

In fact, one can check that these two approaches provide different arrays of dofs:

```
print(u.x.array[map1])
print(u.sub(0).collapse().vector.array)
```

return

```
[0.891773 0.71518937 0.79172504 0.87001215 0.56804456 0.83261985
0.79915856 0.78052918 0.46147936]
[4.00604996e-311 7.01573217e-322 7.91725038e-001 8.70012148e-001
5.68044561e-001 8.32619846e-001 7.99158564e-001 7.80529176e-001
4.61479362e-001]
```

I do not understand why am I getting different outputs from the last two lines in my code. Moreover, I would like to know which approach should I use if I want to compute the maximum in parallel.

Thank you!