Branch 23.06 merge 23.04 (#1021)

* Implement `geom_equals` and binary predicates that depend only on it. (#926) This PR implements binary predicates that depend only on equality, which is implemented here using columnar comparison in python. I'm playing with benchmarks of this feature now. On only Point geometries, we begin to outperform geopandas at 50k points, with 60x performance at 10m points. Authors: - H. Thomson Comer (https://github.com/thomcom) Approvers: - Michael Wang (https://github.com/isVoid) URL: #926 * Add python API `pairwise_point_polygon_distance` (#988) This PR closes #756 , add `pairwise_point_polygon_distance` for python. Depend on #984 #976 Authors: - Michael Wang (https://github.com/isVoid) Approvers: - H. Thomson Comer (https://github.com/thomcom) URL: #988 * Add `dependency-file-generator` as `pre-commit` hook (#1008) Similarly to these [cudf](rapidsai/cudf#12819) and [cuml](rapidsai/cuml#5246) PRs, this PR adds an entry to `.pre-commit-config.yaml` to run the [dependency-file-generator](https://github.com/rapidsai/dependency-file-generator). It also adds an argument to the `rapidsai/shared-action-workflows/.github/workflows/checks.yaml` shared workflow to disable the `dependency-file-generator` from running in that shared workflow. This avoids having the `dependency-file-generator` run in two places since pre-commit is run in CI [here](https://github.com/rapidsai/cuspatial/blob/branch-23.04/ci/check_style.sh#L23). Authors: - AJ Schmidt (https://github.com/ajschmidt8) Approvers: - Ray Douglass (https://github.com/raydouglass) URL: #1008 * Add ZipCode Counting Notebook (#919) This PR adds a notebook that demonstrate the use of quadtree PiP with a custom `QuadTree` structure and joins dataframe. We can use this PR to discuss the possibility of adding the custom structure to the codebase. This also updates the dependency list to include notebook environment in "all" targeted conda envs. Authors: - Michael Wang (https://github.com/isVoid) Approvers: - H. Thomson Comer (https://github.com/thomcom) - Mark Harris (https://github.com/harrism) - AJ Schmidt (https://github.com/ajschmidt8) URL: #919 * Header-only `quadtree_point_in_polygon` (#979) Closes #985 Also contains cleanup of docs for other spatial join functions, correct ordering of stream and MR parameters, and adds missing C++17 property from tests cmake configuration. Authors: - Mark Harris (https://github.com/harrism) Approvers: - Michael Wang (https://github.com/isVoid) - Paul Taylor (https://github.com/trxcllnt) URL: #979 * Reduce gtest times (#1018) Fixes #1017. Reduces C++ gtest total time (on my PC) from 47.9 seconds to 20.08 seconds. Several tests were running large datasets and combinations of size parameters that would be better to run as benchmarks rather than gtests. Reducing these by a factor of 10-100 saves a lot of development time and still exercises the code. In the case of `HausdorffTest/1.10000Spaces10Points (4850 ms)`, reducing it to 1000 spaces, 10 points reduced the time by nearly 100x, likely because it's $O(N^2)$. I modified any test that used close to 1s or more total time, since most column-API tests use under that, and most header-only tests use under 0.2s. | Test | Time Before (s) | Time After (s) | Speedup | |---|---|---|---| | DERIVE_TRAJECTORIES_TEST_EXP | 14.49 | 0.27 | 53.7x | | HAUSDORFF_TEST_EXP | 9.21 | 0.26 | 35.4x | | UTILITY_TEST | 1.86 | 0.30 | 6.2x | | POINT_BOUNDING_BOXES_TEST_EXP | 1.35 | 0.15 | 9x | | TRAJECTORY_DISTANCES_AND_SPEEDS_TEST_EXP | 0.80 | 0.13 | 6.2x | | TOTAL | 47.9 | 20.08 | 2.4x | Before: ``` (rapids) coder ➜ ~/cuspatial/cpp/build/release $ ninja test [0/1] Running tests... Test project /home/coder/cuspatial/cpp/build/release Start 1: SINUSOIDAL_PROJECTION_TEST 1/45 Test #1: SINUSOIDAL_PROJECTION_TEST ................. Passed 0.81 sec Start 2: HAVERSINE_TEST 2/45 Test #2: HAVERSINE_TEST ............................. Passed 0.77 sec Start 3: HAUSDORFF_TEST 3/45 Test #3: HAUSDORFF_TEST ............................. Passed 0.75 sec Start 4: JOIN_POINT_TO_LINESTRING_SMALL_TEST 4/45 Test #4: JOIN_POINT_TO_LINESTRING_SMALL_TEST ........ Passed 0.73 sec Start 5: JOIN_POINT_IN_POLYGON_TEST 5/45 Test #5: JOIN_POINT_IN_POLYGON_TEST ................. Passed 0.79 sec Start 6: POINT_IN_POLYGON_TEST 6/45 Test #6: POINT_IN_POLYGON_TEST ...................... Passed 0.80 sec Start 7: PAIRWISE_POINT_IN_POLYGON_TEST 7/45 Test #7: PAIRWISE_POINT_IN_POLYGON_TEST ............. Passed 0.76 sec Start 8: POINT_QUADTREE_TEST 8/45 Test #8: POINT_QUADTREE_TEST ........................ Passed 0.76 sec Start 9: LINESTRING_BOUNDING_BOXES_TEST 9/45 Test #9: LINESTRING_BOUNDING_BOXES_TEST ............. Passed 0.76 sec Start 10: POLYGON_BOUNDING_BOXES_TEST 10/45 Test #10: POLYGON_BOUNDING_BOXES_TEST ................ Passed 0.80 sec Start 11: POINT_DISTANCE_TEST 11/45 Test #11: POINT_DISTANCE_TEST ........................ Passed 0.79 sec Start 12: POINT_LINESTRING_DISTANCE_TEST 12/45 Test #12: POINT_LINESTRING_DISTANCE_TEST ............. Passed 0.78 sec Start 13: LINESTRING_DISTANCE_TEST 13/45 Test #13: LINESTRING_DISTANCE_TEST ................... Passed 0.78 sec Start 14: POINT_POLYGON_DISTANCE_TEST 14/45 Test #14: POINT_POLYGON_DISTANCE_TEST ................ Passed 0.76 sec Start 15: LINESTRING_INTERSECTION_TEST 15/45 Test #15: LINESTRING_INTERSECTION_TEST ............... Passed 0.83 sec Start 16: POINT_LINESTRING_NEAREST_POINT_TEST 16/45 Test #16: POINT_LINESTRING_NEAREST_POINT_TEST ........ Passed 0.77 sec Start 17: QUADTREE_POLYGON_FILTERING_TEST 17/45 Test #17: QUADTREE_POLYGON_FILTERING_TEST ............ Passed 0.79 sec Start 18: QUADTREE_LINESTRING_FILTERING_TEST 18/45 Test #18: QUADTREE_LINESTRING_FILTERING_TEST ......... Passed 0.76 sec Start 19: TRAJECTORY_DISTANCES_AND_SPEEDS_TEST 19/45 Test #19: TRAJECTORY_DISTANCES_AND_SPEEDS_TEST ....... Passed 0.79 sec Start 20: DERIVE_TRAJECTORIES_TEST 20/45 Test #20: DERIVE_TRAJECTORIES_TEST ................... Passed 0.76 sec Start 21: TRAJECTORY_BOUNDING_BOXES_TEST 21/45 Test #21: TRAJECTORY_BOUNDING_BOXES_TEST ............. Passed 0.75 sec Start 22: SPATIAL_WINDOW_POINT_TEST 22/45 Test #22: SPATIAL_WINDOW_POINT_TEST .................. Passed 0.75 sec Start 23: UTILITY_TEST 23/45 Test #23: UTILITY_TEST ............................... Passed 1.86 sec Start 24: HAVERSINE_TEST_EXP 24/45 Test #24: HAVERSINE_TEST_EXP ......................... Passed 0.14 sec Start 25: POINT_DISTANCE_TEST_EXP 25/45 Test #25: POINT_DISTANCE_TEST_EXP .................... Passed 0.11 sec Start 26: POINT_LINESTRING_DISTANCE_TEST_EXP 26/45 Test #26: POINT_LINESTRING_DISTANCE_TEST_EXP ......... Passed 0.11 sec Start 27: POINT_POLYGON_DISTANCE_TEST_EXP 27/45 Test #27: POINT_POLYGON_DISTANCE_TEST_EXP ............ Passed 0.13 sec Start 28: HAUSDORFF_TEST_EXP 28/45 Test #28: HAUSDORFF_TEST_EXP ......................... Passed 9.21 sec Start 29: LINESTRING_DISTANCE_TEST_EXP 29/45 Test #29: LINESTRING_DISTANCE_TEST_EXP ............... Passed 0.17 sec Start 30: LINESTRING_INTERSECTION_TEST_EXP 30/45 Test #30: LINESTRING_INTERSECTION_TEST_EXP ........... Passed 0.19 sec Start 31: POINT_LINESTRING_NEAREST_POINT_TEST_EXP 31/45 Test #31: POINT_LINESTRING_NEAREST_POINT_TEST_EXP .... Passed 0.12 sec Start 32: SINUSOIDAL_PROJECTION_TEST_EXP 32/45 Test #32: SINUSOIDAL_PROJECTION_TEST_EXP ............. Passed 0.12 sec Start 33: POINTS_IN_RANGE_TEST_EXP 33/45 Test #33: POINTS_IN_RANGE_TEST_EXP ................... Passed 0.11 sec Start 34: POINT_IN_POLYGON_TEST_EXP 34/45 Test #34: POINT_IN_POLYGON_TEST_EXP .................. Passed 0.12 sec Start 35: PAIRWISE_POINT_IN_POLYGON_TEST_EXP 35/45 Test #35: PAIRWISE_POINT_IN_POLYGON_TEST_EXP ......... Passed 0.11 sec Start 36: DERIVE_TRAJECTORIES_TEST_EXP 36/45 Test #36: DERIVE_TRAJECTORIES_TEST_EXP ............... Passed 14.49 sec Start 37: POINT_BOUNDING_BOXES_TEST_EXP 37/45 Test #37: POINT_BOUNDING_BOXES_TEST_EXP .............. Passed 1.35 sec Start 38: POLYGON_BOUNDING_BOXES_TEST_EXP 38/45 Test #38: POLYGON_BOUNDING_BOXES_TEST_EXP ............ Passed 0.11 sec Start 39: LINESTRING_BOUNDING_BOXES_TEST_EXP 39/45 Test #39: LINESTRING_BOUNDING_BOXES_TEST_EXP ......... Passed 0.11 sec Start 40: TRAJECTORY_DISTANCES_AND_SPEEDS_TEST_EXP 40/45 Test #40: TRAJECTORY_DISTANCES_AND_SPEEDS_TEST_EXP ... Passed 0.80 sec Start 41: POINT_QUADTREE_TEST_EXP 41/45 Test #41: POINT_QUADTREE_TEST_EXP .................... Passed 0.12 sec Start 42: OPERATOR_TEST_EXP 42/45 Test #42: OPERATOR_TEST_EXP .......................... Passed 0.14 sec Start 43: FIND_TEST_EXP 43/45 Test #43: FIND_TEST_EXP .............................. Passed 0.13 sec Start 44: JOIN_POINT_IN_POLYGON_SMALL_TEST_EXP 44/45 Test #44: JOIN_POINT_IN_POLYGON_SMALL_TEST_EXP ....... Passed 0.11 sec Start 45: JOIN_POINT_IN_POLYGON_LARGE_TEST_EXP 45/45 Test #45: JOIN_POINT_IN_POLYGON_LARGE_TEST_EXP ....... Passed 0.13 sec 100% tests passed, 0 tests failed out of 45 Total Test time (real) = 47.07 sec ``` After: ``` (rapids) coder ➜ ~/cuspatial/cpp/build/release $ ninja test [0/1] Running tests... Test project /home/coder/cuspatial/cpp/build/release Start 1: SINUSOIDAL_PROJECTION_TEST 1/45 Test #1: SINUSOIDAL_PROJECTION_TEST ................. Passed 0.78 sec Start 2: HAVERSINE_TEST 2/45 Test #2: HAVERSINE_TEST ............................. Passed 0.75 sec Start 3: HAUSDORFF_TEST 3/45 Test #3: HAUSDORFF_TEST ............................. Passed 0.74 sec Start 4: JOIN_POINT_TO_LINESTRING_SMALL_TEST 4/45 Test #4: JOIN_POINT_TO_LINESTRING_SMALL_TEST ........ Passed 0.77 sec Start 5: JOIN_POINT_IN_POLYGON_TEST 5/45 Test #5: JOIN_POINT_IN_POLYGON_TEST ................. Passed 0.76 sec Start 6: POINT_IN_POLYGON_TEST 6/45 Test #6: POINT_IN_POLYGON_TEST ...................... Passed 0.78 sec Start 7: PAIRWISE_POINT_IN_POLYGON_TEST 7/45 Test #7: PAIRWISE_POINT_IN_POLYGON_TEST ............. Passed 0.74 sec Start 8: POINT_QUADTREE_TEST 8/45 Test #8: POINT_QUADTREE_TEST ........................ Passed 0.75 sec Start 9: LINESTRING_BOUNDING_BOXES_TEST 9/45 Test #9: LINESTRING_BOUNDING_BOXES_TEST ............. Passed 0.75 sec Start 10: POLYGON_BOUNDING_BOXES_TEST 10/45 Test #10: POLYGON_BOUNDING_BOXES_TEST ................ Passed 0.73 sec Start 11: POINT_DISTANCE_TEST 11/45 Test #11: POINT_DISTANCE_TEST ........................ Passed 0.73 sec Start 12: POINT_LINESTRING_DISTANCE_TEST 12/45 Test #12: POINT_LINESTRING_DISTANCE_TEST ............. Passed 0.74 sec Start 13: LINESTRING_DISTANCE_TEST 13/45 Test #13: LINESTRING_DISTANCE_TEST ................... Passed 0.76 sec Start 14: POINT_POLYGON_DISTANCE_TEST 14/45 Test #14: POINT_POLYGON_DISTANCE_TEST ................ Passed 0.76 sec Start 15: LINESTRING_INTERSECTION_TEST 15/45 Test #15: LINESTRING_INTERSECTION_TEST ............... Passed 0.78 sec Start 16: POINT_LINESTRING_NEAREST_POINT_TEST 16/45 Test #16: POINT_LINESTRING_NEAREST_POINT_TEST ........ Passed 0.77 sec Start 17: QUADTREE_POLYGON_FILTERING_TEST 17/45 Test #17: QUADTREE_POLYGON_FILTERING_TEST ............ Passed 0.75 sec Start 18: QUADTREE_LINESTRING_FILTERING_TEST 18/45 Test #18: QUADTREE_LINESTRING_FILTERING_TEST ......... Passed 0.77 sec Start 19: TRAJECTORY_DISTANCES_AND_SPEEDS_TEST 19/45 Test #19: TRAJECTORY_DISTANCES_AND_SPEEDS_TEST ....... Passed 0.74 sec Start 20: DERIVE_TRAJECTORIES_TEST 20/45 Test #20: DERIVE_TRAJECTORIES_TEST ................... Passed 0.75 sec Start 21: TRAJECTORY_BOUNDING_BOXES_TEST 21/45 Test #21: TRAJECTORY_BOUNDING_BOXES_TEST ............. Passed 0.74 sec Start 22: SPATIAL_WINDOW_POINT_TEST 22/45 Test #22: SPATIAL_WINDOW_POINT_TEST .................. Passed 0.75 sec Start 23: UTILITY_TEST 23/45 Test #23: UTILITY_TEST ............................... Passed 0.30 sec Start 24: HAVERSINE_TEST_EXP 24/45 Test #24: HAVERSINE_TEST_EXP ......................... Passed 0.12 sec Start 25: POINT_DISTANCE_TEST_EXP 25/45 Test #25: POINT_DISTANCE_TEST_EXP .................... Passed 0.12 sec Start 26: POINT_LINESTRING_DISTANCE_TEST_EXP 26/45 Test #26: POINT_LINESTRING_DISTANCE_TEST_EXP ......... Passed 0.12 sec Start 27: POINT_POLYGON_DISTANCE_TEST_EXP 27/45 Test #27: POINT_POLYGON_DISTANCE_TEST_EXP ............ Passed 0.13 sec Start 28: HAUSDORFF_TEST_EXP 28/45 Test #28: HAUSDORFF_TEST_EXP ......................... Passed 0.26 sec Start 29: LINESTRING_DISTANCE_TEST_EXP 29/45 Test #29: LINESTRING_DISTANCE_TEST_EXP ............... Passed 0.14 sec Start 30: LINESTRING_INTERSECTION_TEST_EXP 30/45 Test #30: LINESTRING_INTERSECTION_TEST_EXP ........... Passed 0.19 sec Start 31: POINT_LINESTRING_NEAREST_POINT_TEST_EXP 31/45 Test #31: POINT_LINESTRING_NEAREST_POINT_TEST_EXP .... Passed 0.11 sec Start 32: SINUSOIDAL_PROJECTION_TEST_EXP 32/45 Test #32: SINUSOIDAL_PROJECTION_TEST_EXP ............. Passed 0.11 sec Start 33: POINTS_IN_RANGE_TEST_EXP 33/45 Test #33: POINTS_IN_RANGE_TEST_EXP ................... Passed 0.13 sec Start 34: POINT_IN_POLYGON_TEST_EXP 34/45 Test #34: POINT_IN_POLYGON_TEST_EXP .................. Passed 0.11 sec Start 35: PAIRWISE_POINT_IN_POLYGON_TEST_EXP 35/45 Test #35: PAIRWISE_POINT_IN_POLYGON_TEST_EXP ......... Passed 0.14 sec Start 36: DERIVE_TRAJECTORIES_TEST_EXP 36/45 Test #36: DERIVE_TRAJECTORIES_TEST_EXP ............... Passed 0.27 sec Start 37: POINT_BOUNDING_BOXES_TEST_EXP 37/45 Test #37: POINT_BOUNDING_BOXES_TEST_EXP .............. Passed 0.15 sec Start 38: POLYGON_BOUNDING_BOXES_TEST_EXP 38/45 Test #38: POLYGON_BOUNDING_BOXES_TEST_EXP ............ Passed 0.13 sec Start 39: LINESTRING_BOUNDING_BOXES_TEST_EXP 39/45 Test #39: LINESTRING_BOUNDING_BOXES_TEST_EXP ......... Passed 0.12 sec Start 40: TRAJECTORY_DISTANCES_AND_SPEEDS_TEST_EXP 40/45 Test #40: TRAJECTORY_DISTANCES_AND_SPEEDS_TEST_EXP ... Passed 0.13 sec Start 41: POINT_QUADTREE_TEST_EXP 41/45 Test #41: POINT_QUADTREE_TEST_EXP .................... Passed 0.14 sec Start 42: OPERATOR_TEST_EXP 42/45 Test #42: OPERATOR_TEST_EXP .......................... Passed 0.14 sec Start 43: FIND_TEST_EXP 43/45 Test #43: FIND_TEST_EXP .............................. Passed 0.15 sec Start 44: JOIN_POINT_IN_POLYGON_SMALL_TEST_EXP 44/45 Test #44: JOIN_POINT_IN_POLYGON_SMALL_TEST_EXP ....... Passed 0.12 sec Start 45: JOIN_POINT_IN_POLYGON_LARGE_TEST_EXP 45/45 Test #45: JOIN_POINT_IN_POLYGON_LARGE_TEST_EXP ....... Passed 0.13 sec 100% tests passed, 0 tests failed out of 45 Total Test time (real) = 20.08 sec ``` Authors: - Mark Harris (https://github.com/harrism) Approvers: - Michael Wang (https://github.com/isVoid) - Paul Taylor (https://github.com/trxcllnt) URL: #1018 * Re-add enabled_check_generated_files:false --------- Co-authored-by: H. Thomson Comer <thomcom@gmail.com> Co-authored-by: Michael Wang <isVoid@users.noreply.github.com> Co-authored-by: AJ Schmidt <ajschmidt8@users.noreply.github.com>
rapidsai · Mar 30, 2023 · a7466af · a7466af
1 parent 6b88aa8
commit a7466af
Show file tree

Hide file tree

Showing 53 changed files with 3,767 additions and 851 deletions.
diff --git a/.github/workflows/pr.yaml b/.github/workflows/pr.yaml
@@ -24,6 +24,8 @@ jobs:
   checks:
     secrets: inherit
     uses: rapidsai/shared-action-workflows/.github/workflows/checks.yaml@branch-23.06
+    with:
+      enable_check_generated_files: false
   conda-cpp-build:
     needs: checks
     secrets: inherit

diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml
@@ -42,6 +42,11 @@ repos:
               - id: codespell
                 args: ["--config pyproject.toml"]
                 additional_dependencies: ["tomli"]
+      - repo: https://github.com/rapidsai/dependency-file-generator
+        rev: v1.5.1
+        hooks:
+            - id: rapids-dependency-file-generator
+              args: ["--clean"]
 
 default_language_version:
       python: python3
diff --git a/conda/environments/all_cuda-118_arch-x86_64.yaml b/conda/environments/all_cuda-118_arch-x86_64.yaml
@@ -23,17 +23,20 @@ dependencies:
 - myst-parser
 - nbsphinx
 - ninja
+- notebook
 - numpydoc
 - nvcc_linux-64=11.8
 - pre-commit
 - pydata-sphinx-theme
+- pydeck
 - pytest
 - pytest-cov
 - pytest-xdist
 - python>=3.8,<3.11
 - rmm=23.06
 - scikit-build>=0.13.1
 - setuptools
+- shapely
 - sphinx<6
 - sysroot_linux-64==2.17
 name: all_cuda-118_arch-x86_64
diff --git a/cpp/benchmarks/floating_point_equality.cu b/cpp/benchmarks/floating_point_equality.cu
@@ -45,14 +45,14 @@ template <class FloatsIter>
 void generate_floats(FloatsIter begin, FloatsIter end)
 {
   using T       = typename std::iterator_traits<FloatsIter>::value_type;
-  auto engine_x = deterministic_engine(std::distance(begin, end));
+  auto engine_x = cuspatial::test::deterministic_engine(std::distance(begin, end));
 
   auto lo = std::numeric_limits<T>::min();
   auto hi = std::numeric_limits<T>::max();
 
-  auto x_dist = make_uniform_dist(lo, hi);
+  auto x_dist = cuspatial::test::make_uniform_dist(lo, hi);
 
-  auto x_gen = value_generator{lo, hi, engine_x, x_dist};
+  auto x_gen = cuspatial::test::value_generator{lo, hi, engine_x, x_dist};
 
   thrust::tabulate(
     rmm::exec_policy(), begin, end, [x_gen] __device__(size_t n) mutable { return x_gen(n); });

diff --git a/cpp/benchmarks/points_in_range.cu b/cpp/benchmarks/points_in_range.cu
@@ -15,6 +15,7 @@
  */
 
 #include <benchmarks/fixture/rmm_pool_raii.hpp>
+
 #include <cuspatial_test/random.cuh>
 
 #include <cuspatial/detail/iterator.hpp>
@@ -54,14 +55,14 @@ using namespace cuspatial;
 template <class PointsIter, typename T>
 void generate_points(PointsIter begin, PointsIter end, vec_2d<T> range_min, vec_2d<T> range_max)
 {
-  auto engine_x = deterministic_engine(std::distance(begin, end));
-  auto engine_y = deterministic_engine(2 * std::distance(begin, end));
+  auto engine_x = cuspatial::test::deterministic_engine(std::distance(begin, end));
+  auto engine_y = cuspatial::test::deterministic_engine(2 * std::distance(begin, end));
 
-  auto x_dist = make_uniform_dist(range_min.x, range_max.x);
-  auto y_dist = make_uniform_dist(range_min.y, range_max.y);
+  auto x_dist = cuspatial::test::make_uniform_dist(range_min.x, range_max.x);
+  auto y_dist = cuspatial::test::make_uniform_dist(range_min.y, range_max.y);
 
-  auto x_gen = value_generator{range_min.x, range_max.x, engine_x, x_dist};
-  auto y_gen = value_generator{range_min.y, range_max.y, engine_y, y_dist};
+  auto x_gen = cuspatial::test::value_generator{range_min.x, range_max.x, engine_x, x_dist};
+  auto y_gen = cuspatial::test::value_generator{range_min.y, range_max.y, engine_y, y_dist};
 
   thrust::tabulate(rmm::exec_policy(), begin, end, [x_gen, y_gen] __device__(size_t n) mutable {
     return vec_2d<T>{x_gen(n), y_gen(n)};

diff --git a/cpp/include/cuspatial/experimental/detail/algorithm/is_point_in_polygon.cuh b/cpp/include/cuspatial/experimental/detail/algorithm/is_point_in_polygon.cuh
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2022, NVIDIA CORPORATION.
+ * Copyright (c) 2022-2023, NVIDIA CORPORATION.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.

diff --git a/cpp/include/cuspatial/experimental/detail/join/get_quad_and_local_point_indices.cuh b/cpp/include/cuspatial/experimental/detail/join/get_quad_and_local_point_indices.cuh
@@ -23,8 +23,11 @@
 namespace cuspatial {
 namespace detail {
 
-inline __device__ std::pair<uint32_t, uint32_t> get_quad_and_local_point_indices(
-  uint32_t const global_index, uint32_t const* point_offsets, uint32_t const* point_offsets_end)
+template <typename IndexType, typename PointOffsetIterator>
+inline __device__ std::pair<IndexType, IndexType> get_quad_and_local_point_indices(
+  IndexType const global_index,
+  PointOffsetIterator point_offsets_begin,
+  PointOffsetIterator point_offsets_end)
 {
   // Calculate the position in "point_offsets" that `global_index` falls between.
   // This position is the index of the poly/quad pair for this `global_index`.
@@ -33,10 +36,10 @@ inline __device__ std::pair<uint32_t, uint32_t> get_quad_and_local_point_indices
   // quadrant. Adding this zero-based position to the quadrant's first point position in the
   // quadtree yields the "global" position in the `point_indices` map.
   auto const local_point_offset =
-    thrust::upper_bound(thrust::seq, point_offsets, point_offsets_end, global_index) - 1;
+    thrust::upper_bound(thrust::seq, point_offsets_begin, point_offsets_end, global_index) - 1;
   return std::make_pair(
     // quad_poly_index
-    thrust::distance(point_offsets, local_point_offset),
+    thrust::distance(point_offsets_begin, local_point_offset),
     // local_point_index
     global_index - *local_point_offset);
 }

diff --git a/cpp/include/cuspatial/experimental/detail/join/intersection.cuh b/cpp/include/cuspatial/experimental/detail/join/intersection.cuh
@@ -17,6 +17,8 @@
 #pragma once
 
 #include <cuspatial/detail/utility/z_order.cuh>
+#include <cuspatial/experimental/geometry/box.hpp>
+#include <cuspatial/experimental/point_quadtree.cuh>
 
 #include <rmm/cuda_stream_view.hpp>
 #include <rmm/device_uvector.hpp>
@@ -69,37 +71,32 @@ inline int32_t remove_non_quad_intersections(InputIterator input_begin,
 }
 
 template <class T,
-          class KeyIterator,
-          class LevelIterator,
-          class IsInternalIterator,
           class BoundingBoxIterator,
           class NodeIndicesIterator,
           class BBoxIndicesIterator,
           class NodePairsIterator,
           class LeafPairsIterator>
-inline std::pair<int32_t, int32_t> find_intersections(KeyIterator keys_first,
-                                                      LevelIterator levels_first,
-                                                      IsInternalIterator is_internal_node_first,
+inline std::pair<int32_t, int32_t> find_intersections(point_quadtree_ref quadtree,
                                                       BoundingBoxIterator bounding_box_first,
                                                       NodeIndicesIterator node_indices,
                                                       BBoxIndicesIterator bbox_indices,
                                                       NodePairsIterator node_pairs,
                                                       LeafPairsIterator leaf_pairs,
                                                       int32_t num_pairs,
-                                                      T x_min,
-                                                      T y_min,
+                                                      vec_2d<T> const& v_min,
                                                       T scale,
                                                       int8_t max_depth,
                                                       rmm::cuda_stream_view stream)
 {
-  auto nodes_first = thrust::make_zip_iterator(keys_first, levels_first, is_internal_node_first);
+  auto nodes_first = thrust::make_zip_iterator(
+    quadtree.key_begin(), quadtree.level_begin(), quadtree.internal_node_flag_begin());
 
   thrust::transform(
     rmm::exec_policy(stream),
     thrust::make_zip_iterator(node_indices, bbox_indices),
     thrust::make_zip_iterator(node_indices, bbox_indices) + num_pairs,
     node_pairs,
-    [x_min, y_min, scale, max_depth, nodes = nodes_first, bboxes = bounding_box_first] __device__(
+    [v_min, scale, max_depth, nodes = nodes_first, bboxes = bounding_box_first] __device__(
       auto const& node_and_bbox) {
       auto const& node_idx = thrust::get<0>(node_and_bbox);
       auto const& bbox_idx = thrust::get<1>(node_and_bbox);
@@ -109,24 +106,20 @@ inline std::pair<int32_t, int32_t> find_intersections(KeyIterator keys_first,
       uint8_t const& level            = thrust::get<1>(node);
       uint8_t const& is_internal_node = thrust::get<2>(node);
 
-      auto const& bbox       = bboxes[bbox_idx];
-      auto const& bbox_min   = bbox.v1;
-      auto const& bbox_max   = bbox.v2;
-      auto const& poly_x_min = bbox_min.x;
-      auto const& poly_y_min = bbox_min.y;
-      auto const& poly_x_max = bbox_max.x;
-      auto const& poly_y_max = bbox_max.y;
+      box<T> const bbox        = bboxes[bbox_idx];
+      vec_2d<T> const bbox_min = bbox.v1;
+      vec_2d<T> const bbox_max = bbox.v2;
 
       T const key_x       = utility::z_order_x(key);
       T const key_y       = utility::z_order_y(key);
       T const level_scale = scale * (1 << (max_depth - 1 - level));
-      T const node_x_min  = x_min + (key_x + 0) * level_scale;
-      T const node_y_min  = y_min + (key_y + 0) * level_scale;
-      T const node_x_max  = x_min + (key_x + 1) * level_scale;
-      T const node_y_max  = y_min + (key_y + 1) * level_scale;
+      T const node_x_min  = v_min.x + (key_x + 0) * level_scale;
+      T const node_y_min  = v_min.y + (key_y + 0) * level_scale;
+      T const node_x_max  = v_min.x + (key_x + 1) * level_scale;
+      T const node_y_max  = v_min.y + (key_y + 1) * level_scale;
 
-      if ((node_x_min > poly_x_max) || (node_x_max < poly_x_min) || (node_y_min > poly_y_max) ||
-          (node_y_max < poly_y_min)) {
+      if ((node_x_min > bbox_max.x) || (node_x_max < bbox_min.x) || (node_y_min > bbox_max.y) ||
+          (node_y_max < bbox_min.y)) {
         // if no overlap, return type = none_indicator
         return thrust::make_tuple(none_indicator, level, node_idx, bbox_idx);
       }

diff --git a/cpp/include/cuspatial/experimental/detail/point_quadtree.cuh b/cpp/include/cuspatial/experimental/detail/point_quadtree.cuh
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2022, NVIDIA CORPORATION.
+ * Copyright (c) 2022-2023, NVIDIA CORPORATION.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -241,8 +241,8 @@ std::pair<rmm::device_uvector<uint32_t>, point_quadtree> quadtree_on_points(
   T scale,
   int8_t max_depth,
   int32_t max_size,
-  rmm::mr::device_memory_resource* mr,
-  rmm::cuda_stream_view stream)
+  rmm::cuda_stream_view stream,
+  rmm::mr::device_memory_resource* mr)
 {
   auto num_points = thrust::distance(points_first, points_last);
   if (num_points <= 0) {

diff --git a/cpp/include/cuspatial/experimental/detail/quadtree_bbox_filtering.cuh b/cpp/include/cuspatial/experimental/detail/quadtree_bbox_filtering.cuh
@@ -29,38 +29,27 @@
 
 namespace cuspatial {
 
-template <class KeyIterator,
-          class LevelIterator,
-          class IsInternalIterator,
-          class BoundingBoxIterator,
-          class T>
+template <class BoundingBoxIterator, class T>
 std::pair<rmm::device_uvector<uint32_t>, rmm::device_uvector<uint32_t>>
-join_quadtree_and_bounding_boxes(KeyIterator keys_first,
-                                 KeyIterator keys_last,
-                                 LevelIterator levels_first,
-                                 IsInternalIterator is_internal_nodes_first,
-                                 KeyIterator lengths_first,
-                                 KeyIterator offsets_first,
+join_quadtree_and_bounding_boxes(point_quadtree_ref quadtree,
                                  BoundingBoxIterator bounding_boxes_first,
                                  BoundingBoxIterator bounding_boxes_last,
-                                 T x_min,
-                                 T y_min,
+                                 vec_2d<T> const& v_min,
                                  T scale,
                                  int8_t max_depth,
-                                 rmm::mr::device_memory_resource* mr,
-                                 rmm::cuda_stream_view stream)
+                                 rmm::cuda_stream_view stream,
+                                 rmm::mr::device_memory_resource* mr)
 {
   static_assert(is_same<T, cuspatial::iterator_vec_base_type<BoundingBoxIterator>>(),
                 "Iterator value_type mismatch");
 
-  auto const num_nodes = std::distance(keys_first, keys_last);
   auto const num_boxes = std::distance(bounding_boxes_first, bounding_boxes_last);
 
   // Count the number of top-level nodes to start.
   // This could be provided explicitly, but count_if should be fast enough.
   auto num_top_level_leaves = thrust::count_if(rmm::exec_policy(stream),
-                                               levels_first,
-                                               levels_first + num_nodes,
+                                               quadtree.level_begin(),
+                                               quadtree.level_end(),
                                                thrust::placeholders::_1 == 0);
 
   auto num_pairs = num_top_level_leaves * num_boxes;
@@ -96,9 +85,7 @@ join_quadtree_and_bounding_boxes(KeyIterator keys_first,
 
   // Find intersections for all the top level quadrants and bounding boxes
   std::tie(num_parents, num_leaves) =
-    detail::find_intersections(keys_first,
-                               levels_first,
-                               is_internal_nodes_first,
+    detail::find_intersections(quadtree,
                                bounding_boxes_first,
                                // The top-level node indices
                                detail::make_counting_transform_iterator(
@@ -111,8 +98,7 @@ join_quadtree_and_bounding_boxes(KeyIterator keys_first,
                                // found intersecting quadrant and bbox indices for output
                                make_output_values_iter(),
                                num_pairs,
-                               x_min,
-                               y_min,
+                               v_min,
                                scale,
                                max_depth,
                                stream);
@@ -141,8 +127,8 @@ join_quadtree_and_bounding_boxes(KeyIterator keys_first,
     // Walk one level down and fill the current level's vectors with
     // the next level's quadrant info and bbox indices.
     std::tie(num_pairs, cur_types, cur_levels, cur_node_idxs, cur_bbox_idxs) =
-      detail::descend_quadtree(lengths_first,
-                               offsets_first,
+      detail::descend_quadtree(quadtree.length_begin(),
+                               quadtree.offset_begin(),
                                num_parents,
                                cur_types,
                                cur_levels,
@@ -152,9 +138,7 @@ join_quadtree_and_bounding_boxes(KeyIterator keys_first,
 
     // Find intersections for the the next level's quadrants and bboxes
     std::tie(num_parents, num_leaves) =
-      detail::find_intersections(keys_first,
-                                 levels_first,
-                                 is_internal_nodes_first,
+      detail::find_intersections(quadtree,
                                  bounding_boxes_first,
                                  cur_node_idxs.begin(),
                                  cur_bbox_idxs.begin(),
@@ -163,8 +147,7 @@ join_quadtree_and_bounding_boxes(KeyIterator keys_first,
                                  // found intersecting quadrant and bbox indices for output
                                  make_output_values_iter(),
                                  num_pairs,
-                                 x_min,
-                                 y_min,
+                                 v_min,
                                  scale,
                                  max_depth,
                                  stream);
@@ -177,7 +160,8 @@ join_quadtree_and_bounding_boxes(KeyIterator keys_first,
     // Copy the relevant node offsets into a temporary vector so we don't modify the quadtree column
     rmm::device_uvector<uint32_t> tmp_node_offsets(num_results, stream);
 
-    auto const iter = thrust::make_permutation_iterator(offsets_first, out_node_idxs.begin());
+    auto const iter =
+      thrust::make_permutation_iterator(quadtree.offset_begin(), out_node_idxs.begin());
 
     thrust::copy(rmm::exec_policy(stream), iter, iter + num_results, tmp_node_offsets.begin());