BUG in Sparse Spatial Data Structures

[Linus-Civil]

Describe the bug
The interface function of sparse data structure does not return the result as expected.

To Reproduce

import taichi as ti
ti.init(arch=ti.gpu)

x=ti.field(ti.f32)
block1 = ti.root.pointer(ti.i,3)
block2 = block1.bitmasked(ti.j,3)
block2.place(x)

x[1,1] = 1.2  # this should activate block1 [1] and block2 [4]

@ti.kernel
def check():
    print(ti.is_active(block1,[1])) # output: 1, this result is right
    print(ti.is_active(block2,[4])) # output: 0, not the expected result. Moreover, block2 cannot be indexed with 2 indices here.
    print(ti.rescale_index(block2,block1,[4])) #output: [4], not the expected result. it should print "[1]" here, according to the tutorial.
    for i,j in x:
        print(i,j) #output: 1 1, this result is right

check()
ti.deactivate_all_snodes() #RuntimeError: [D:/a/taichi/taichi/taichi/transforms/type_check.cpp:visit@152] [$28] S1 has 1 indices. Indexed with 2.

Additional comments
I explained all the existing problems in the comments.

[yuanming-hu]

Thank you for pointing this out! I confirm there's a bug here. Will take a close look later. My response to this will not be quick (since I need to look deep inside), but this issue provides valuable information, thanks!