Tree build refactor 1

[xywei]

This is the first of several PRs to refactor the tree build. This PR adds the very basic classes and methods to enable generation of uniform box mesh for volumential (thus dropping its dependency on another custom fork of boxtree).

Generating a uniform mesh can be done as follows:

import numpy as np
import boxtree as bt
import modepy as mp

tree = bt.tree_build._make_tob_root(2, [[-1, -1], [1, 1]])

n_levels = 3
for lev in range(n_levels):
    tree = tree.uniformly_refined()

quad = mp.LegendreGaussQuadrature(1)
x, q = bt.tree_build.distribute_quadrature_rule(tree, quad)

print(x)
assert np.isclose(sum(q), 4)

• Needs ListOfListBuilder: tolerate None arguments pyopencl#600

[inducer]

@xywei There are some CI failures here that I believe might be on the end of the new code. Could you take a look?

_ test_uniform_tree_of_boxes_convergence[<context factory for <pyopencl.Device 'pthread-Intel(R) Xeon(R) Platinum 8272CL CPU @ 2.60GHz' on 'Portable Computing Language'>>-4-3] _
[gw1] linux -- Python 3.10.6 /home/runner/work/boxtree/boxtree/.miniforge3/envs/testing/bin/python3
Traceback (most recent call last):
  File "/home/runner/work/boxtree/boxtree/test/test_tree_of_boxes.py", line 156, in test_uniform_tree_of_boxes_convergence
    assert eoc_rec.order_estimate() >= 2*order + 2 - eps
AssertionError: assert 6.59328736537206 >= (((2 * 4) + 2) - 0.01)
 +  where 6.59328736537206 = <bound method EOCRecorder.order_estimate of <pytools.convergence.EOCRecorder object at 0x7f581a3e82e0>>()
 +    where <bound method EOCRecorder.order_estimate of <pytools.convergence.EOCRecorder object at 0x7f581a3e82e0>> = <pytools.convergence.EOCRecorder object at 0x7f581a3e82e0>.order_estimate
_ test_uniform_tree_of_boxes_convergence[<context factory for <pyopencl.Device 'pthread-Intel(R) Xeon(R) Platinum 8272CL CPU @ 2.60GHz' on 'Portable Computing Language'>>-4-2] _
[gw2] linux -- Python 3.10.6 /home/runner/work/boxtree/boxtree/.miniforge3/envs/testing/bin/python3
Traceback (most recent call last):
  File "/home/runner/work/boxtree/boxtree/test/test_tree_of_boxes.py", line 156, in test_uniform_tree_of_boxes_convergence
    assert eoc_rec.order_estimate() >= 2*order + 2 - eps
AssertionError: assert 1.9475325801058603 >= (((2 * 4) + 2) - 0.01)
 +  where 1.9475325801058603 = <bound method EOCRecorder.order_estimate of <pytools.convergence.EOCRecorder object at 0x7faee4da2a40>>()
 +    where <bound method EOCRecorder.order_estimate of <pytools.convergence.EOCRecorder object at 0x7faee4da2a40>> = <pytools.convergence.EOCRecorder object at 0x7faee4da2a40>.order_estimate
_ test_uniform_tree_of_boxes_convergence[<context factory for <pyopencl.Device 'pthread-Intel(R) Xeon(R) Platinum 8272CL CPU @ 2.60GHz' on 'Portable Computing Language'>>-3-2] _
[gw3] linux -- Python 3.10.6 /home/runner/work/boxtree/boxtree/.miniforge3/envs/testing/bin/python3
Traceback (most recent call last):
  File "/home/runner/work/boxtree/boxtree/test/test_tree_of_boxes.py", line 156, in test_uniform_tree_of_boxes_convergence
    assert eoc_rec.order_estimate() >= 2*order + 2 - eps
AssertionError: assert 7.9662860043822175 >= (((2 * 3) + 2) - 0.01)
 +  where 7.9662860043822175 = <bound method EOCRecorder.order_estimate of <pytools.convergence.EOCRecorder object at 0x7fc2d49ebaf0>>()
 +    where <bound method EOCRecorder.order_estimate of <pytools.convergence.EOCRecorder object at 0x7fc2d49ebaf0>> = <pytools.convergence.EOCRecorder object at 0x7fc2d49ebaf0>.order_estimate

[inducer]

Thanks for addressing the previous set of concerns. Looked again, found some more, including a potentially significant one about box number ordering.

[inducer]

Unsubscribing... @-mention or request review once it's ready for a look or needs attention.