datasketches 0.3.1 → 0.4.0

Sign up to get free protection for your applications and to get access to all the features.
Files changed (245) hide show
  1. checksums.yaml +4 -4
  2. data/CHANGELOG.md +10 -0
  3. data/NOTICE +1 -1
  4. data/README.md +0 -2
  5. data/ext/datasketches/cpc_wrapper.cpp +3 -3
  6. data/ext/datasketches/kll_wrapper.cpp +0 -10
  7. data/lib/datasketches/version.rb +1 -1
  8. data/lib/datasketches.rb +1 -1
  9. data/vendor/datasketches-cpp/CMakeLists.txt +23 -20
  10. data/vendor/datasketches-cpp/CODE_OF_CONDUCT.md +3 -0
  11. data/vendor/datasketches-cpp/CONTRIBUTING.md +50 -0
  12. data/vendor/datasketches-cpp/Doxyfile +2827 -0
  13. data/vendor/datasketches-cpp/LICENSE +0 -76
  14. data/vendor/datasketches-cpp/README.md +1 -3
  15. data/vendor/datasketches-cpp/common/CMakeLists.txt +12 -11
  16. data/vendor/datasketches-cpp/common/include/MurmurHash3.h +25 -27
  17. data/vendor/datasketches-cpp/common/include/common_defs.hpp +15 -10
  18. data/vendor/datasketches-cpp/common/include/count_zeros.hpp +11 -2
  19. data/vendor/datasketches-cpp/common/include/kolmogorov_smirnov.hpp +9 -6
  20. data/vendor/datasketches-cpp/common/include/memory_operations.hpp +5 -4
  21. data/vendor/datasketches-cpp/common/include/optional.hpp +148 -0
  22. data/vendor/datasketches-cpp/common/include/quantiles_sorted_view.hpp +95 -2
  23. data/vendor/datasketches-cpp/common/include/quantiles_sorted_view_impl.hpp +1 -1
  24. data/vendor/datasketches-cpp/common/include/serde.hpp +69 -20
  25. data/vendor/datasketches-cpp/common/test/CMakeLists.txt +2 -2
  26. data/vendor/datasketches-cpp/common/test/integration_test.cpp +6 -0
  27. data/vendor/datasketches-cpp/common/test/optional_test.cpp +85 -0
  28. data/vendor/datasketches-cpp/common/test/test_allocator.hpp +14 -14
  29. data/vendor/datasketches-cpp/{python/src/__init__.py → count/CMakeLists.txt} +25 -1
  30. data/vendor/datasketches-cpp/count/include/count_min.hpp +405 -0
  31. data/vendor/datasketches-cpp/count/include/count_min_impl.hpp +497 -0
  32. data/vendor/datasketches-cpp/{MANIFEST.in → count/test/CMakeLists.txt} +23 -20
  33. data/vendor/datasketches-cpp/count/test/count_min_allocation_test.cpp +155 -0
  34. data/vendor/datasketches-cpp/count/test/count_min_test.cpp +303 -0
  35. data/vendor/datasketches-cpp/cpc/include/cpc_common.hpp +14 -20
  36. data/vendor/datasketches-cpp/cpc/include/cpc_compressor.hpp +7 -4
  37. data/vendor/datasketches-cpp/cpc/include/cpc_compressor_impl.hpp +17 -17
  38. data/vendor/datasketches-cpp/cpc/include/cpc_confidence.hpp +3 -3
  39. data/vendor/datasketches-cpp/cpc/include/cpc_sketch.hpp +40 -40
  40. data/vendor/datasketches-cpp/cpc/include/cpc_sketch_impl.hpp +14 -11
  41. data/vendor/datasketches-cpp/cpc/include/cpc_union.hpp +35 -11
  42. data/vendor/datasketches-cpp/cpc/include/cpc_union_impl.hpp +8 -8
  43. data/vendor/datasketches-cpp/cpc/include/cpc_util.hpp +16 -8
  44. data/vendor/datasketches-cpp/cpc/include/u32_table.hpp +3 -2
  45. data/vendor/datasketches-cpp/cpc/include/u32_table_impl.hpp +5 -5
  46. data/vendor/datasketches-cpp/cpc/test/CMakeLists.txt +20 -7
  47. data/vendor/datasketches-cpp/cpc/test/cpc_sketch_deserialize_from_java_test.cpp +60 -0
  48. data/vendor/datasketches-cpp/cpc/test/cpc_sketch_serialize_for_java.cpp +38 -0
  49. data/vendor/datasketches-cpp/cpc/test/cpc_sketch_test.cpp +4 -29
  50. data/vendor/datasketches-cpp/cpc/test/cpc_union_test.cpp +4 -4
  51. data/vendor/datasketches-cpp/{tox.ini → density/CMakeLists.txt} +24 -8
  52. data/vendor/datasketches-cpp/density/include/density_sketch.hpp +256 -0
  53. data/vendor/datasketches-cpp/density/include/density_sketch_impl.hpp +543 -0
  54. data/vendor/datasketches-cpp/{python/datasketches/__init__.py → density/test/CMakeLists.txt} +15 -3
  55. data/vendor/datasketches-cpp/density/test/density_sketch_test.cpp +244 -0
  56. data/vendor/datasketches-cpp/fi/include/frequent_items_sketch.hpp +21 -9
  57. data/vendor/datasketches-cpp/fi/include/frequent_items_sketch_impl.hpp +6 -4
  58. data/vendor/datasketches-cpp/fi/include/reverse_purge_hash_map.hpp +9 -3
  59. data/vendor/datasketches-cpp/fi/test/CMakeLists.txt +14 -1
  60. data/vendor/datasketches-cpp/fi/test/frequent_items_sketch_deserialize_from_java_test.cpp +95 -0
  61. data/vendor/datasketches-cpp/fi/test/frequent_items_sketch_serialize_for_java.cpp +83 -0
  62. data/vendor/datasketches-cpp/fi/test/frequent_items_sketch_test.cpp +3 -42
  63. data/vendor/datasketches-cpp/hll/include/CouponList-internal.hpp +2 -2
  64. data/vendor/datasketches-cpp/hll/include/CouponList.hpp +3 -1
  65. data/vendor/datasketches-cpp/hll/include/Hll4Array-internal.hpp +19 -11
  66. data/vendor/datasketches-cpp/hll/include/Hll4Array.hpp +2 -5
  67. data/vendor/datasketches-cpp/hll/include/Hll6Array-internal.hpp +19 -7
  68. data/vendor/datasketches-cpp/hll/include/Hll6Array.hpp +1 -1
  69. data/vendor/datasketches-cpp/hll/include/Hll8Array-internal.hpp +98 -42
  70. data/vendor/datasketches-cpp/hll/include/Hll8Array.hpp +2 -0
  71. data/vendor/datasketches-cpp/hll/include/HllArray-internal.hpp +94 -61
  72. data/vendor/datasketches-cpp/hll/include/HllArray.hpp +20 -8
  73. data/vendor/datasketches-cpp/hll/include/HllSketch-internal.hpp +4 -4
  74. data/vendor/datasketches-cpp/hll/include/HllSketchImpl.hpp +3 -1
  75. data/vendor/datasketches-cpp/hll/include/HllSketchImplFactory.hpp +3 -21
  76. data/vendor/datasketches-cpp/hll/include/HllUnion-internal.hpp +8 -0
  77. data/vendor/datasketches-cpp/hll/include/HllUtil.hpp +14 -18
  78. data/vendor/datasketches-cpp/hll/include/coupon_iterator-internal.hpp +1 -1
  79. data/vendor/datasketches-cpp/hll/include/coupon_iterator.hpp +8 -2
  80. data/vendor/datasketches-cpp/hll/include/hll.hpp +79 -65
  81. data/vendor/datasketches-cpp/hll/test/CMakeLists.txt +14 -1
  82. data/vendor/datasketches-cpp/hll/test/HllUnionTest.cpp +7 -1
  83. data/vendor/datasketches-cpp/hll/test/ToFromByteArrayTest.cpp +0 -68
  84. data/vendor/datasketches-cpp/hll/test/hll_sketch_deserialize_from_java_test.cpp +69 -0
  85. data/vendor/datasketches-cpp/hll/test/hll_sketch_serialize_for_java.cpp +52 -0
  86. data/vendor/datasketches-cpp/kll/include/kll_helper.hpp +0 -1
  87. data/vendor/datasketches-cpp/kll/include/kll_helper_impl.hpp +2 -2
  88. data/vendor/datasketches-cpp/kll/include/kll_sketch.hpp +79 -53
  89. data/vendor/datasketches-cpp/kll/include/kll_sketch_impl.hpp +61 -132
  90. data/vendor/datasketches-cpp/kll/test/CMakeLists.txt +14 -1
  91. data/vendor/datasketches-cpp/kll/test/kll_sketch_deserialize_from_java_test.cpp +103 -0
  92. data/vendor/datasketches-cpp/kll/test/kll_sketch_serialize_for_java.cpp +62 -0
  93. data/vendor/datasketches-cpp/kll/test/kll_sketch_test.cpp +5 -40
  94. data/vendor/datasketches-cpp/quantiles/include/quantiles_sketch.hpp +76 -54
  95. data/vendor/datasketches-cpp/quantiles/include/quantiles_sketch_impl.hpp +66 -136
  96. data/vendor/datasketches-cpp/quantiles/test/CMakeLists.txt +14 -1
  97. data/vendor/datasketches-cpp/quantiles/test/quantiles_sketch_deserialize_from_java_test.cpp +84 -0
  98. data/vendor/datasketches-cpp/quantiles/test/quantiles_sketch_serialize_for_java.cpp +52 -0
  99. data/vendor/datasketches-cpp/quantiles/test/quantiles_sketch_test.cpp +15 -39
  100. data/vendor/datasketches-cpp/req/include/req_common.hpp +7 -3
  101. data/vendor/datasketches-cpp/req/include/req_compactor_impl.hpp +2 -4
  102. data/vendor/datasketches-cpp/req/include/req_sketch.hpp +105 -26
  103. data/vendor/datasketches-cpp/req/include/req_sketch_impl.hpp +50 -111
  104. data/vendor/datasketches-cpp/req/test/CMakeLists.txt +14 -1
  105. data/vendor/datasketches-cpp/req/test/req_sketch_deserialize_from_java_test.cpp +55 -0
  106. data/vendor/datasketches-cpp/{tuple/include/array_of_doubles_intersection_impl.hpp → req/test/req_sketch_serialize_for_java.cpp} +12 -7
  107. data/vendor/datasketches-cpp/req/test/req_sketch_test.cpp +3 -89
  108. data/vendor/datasketches-cpp/sampling/CMakeLists.txt +4 -0
  109. data/vendor/datasketches-cpp/sampling/include/ebpps_sample.hpp +210 -0
  110. data/vendor/datasketches-cpp/sampling/include/ebpps_sample_impl.hpp +535 -0
  111. data/vendor/datasketches-cpp/sampling/include/ebpps_sketch.hpp +281 -0
  112. data/vendor/datasketches-cpp/sampling/include/ebpps_sketch_impl.hpp +531 -0
  113. data/vendor/datasketches-cpp/sampling/include/var_opt_sketch.hpp +89 -32
  114. data/vendor/datasketches-cpp/sampling/include/var_opt_sketch_impl.hpp +33 -19
  115. data/vendor/datasketches-cpp/sampling/include/var_opt_union.hpp +13 -10
  116. data/vendor/datasketches-cpp/sampling/include/var_opt_union_impl.hpp +23 -19
  117. data/vendor/datasketches-cpp/sampling/test/CMakeLists.txt +55 -8
  118. data/vendor/datasketches-cpp/sampling/test/ebpps_allocation_test.cpp +96 -0
  119. data/vendor/datasketches-cpp/sampling/test/ebpps_sample_test.cpp +137 -0
  120. data/vendor/datasketches-cpp/sampling/test/ebpps_sketch_test.cpp +266 -0
  121. data/vendor/datasketches-cpp/sampling/test/var_opt_sketch_deserialize_from_java_test.cpp +81 -0
  122. data/vendor/datasketches-cpp/sampling/test/var_opt_sketch_serialize_for_java.cpp +54 -0
  123. data/vendor/datasketches-cpp/sampling/test/var_opt_sketch_test.cpp +33 -51
  124. data/vendor/datasketches-cpp/sampling/test/var_opt_union_deserialize_from_java_test.cpp +50 -0
  125. data/vendor/datasketches-cpp/sampling/test/var_opt_union_serialize_for_java.cpp +56 -0
  126. data/vendor/datasketches-cpp/sampling/test/var_opt_union_test.cpp +0 -20
  127. data/vendor/datasketches-cpp/theta/CMakeLists.txt +1 -0
  128. data/vendor/datasketches-cpp/theta/include/bit_packing.hpp +6279 -0
  129. data/vendor/datasketches-cpp/theta/include/bounds_on_ratios_in_sampled_sets.hpp +1 -0
  130. data/vendor/datasketches-cpp/theta/include/bounds_on_ratios_in_theta_sketched_sets.hpp +7 -6
  131. data/vendor/datasketches-cpp/theta/include/compact_theta_sketch_parser.hpp +14 -8
  132. data/vendor/datasketches-cpp/theta/include/compact_theta_sketch_parser_impl.hpp +60 -46
  133. data/vendor/datasketches-cpp/theta/include/theta_a_not_b.hpp +20 -5
  134. data/vendor/datasketches-cpp/theta/include/theta_constants.hpp +10 -4
  135. data/vendor/datasketches-cpp/theta/include/theta_helpers.hpp +3 -1
  136. data/vendor/datasketches-cpp/theta/include/theta_intersection.hpp +13 -5
  137. data/vendor/datasketches-cpp/theta/include/theta_intersection_base_impl.hpp +5 -5
  138. data/vendor/datasketches-cpp/theta/include/theta_intersection_impl.hpp +3 -3
  139. data/vendor/datasketches-cpp/theta/include/theta_jaccard_similarity.hpp +2 -1
  140. data/vendor/datasketches-cpp/theta/include/theta_jaccard_similarity_base.hpp +1 -0
  141. data/vendor/datasketches-cpp/theta/include/theta_set_difference_base_impl.hpp +1 -1
  142. data/vendor/datasketches-cpp/theta/include/theta_sketch.hpp +180 -33
  143. data/vendor/datasketches-cpp/theta/include/theta_sketch_impl.hpp +430 -130
  144. data/vendor/datasketches-cpp/theta/include/theta_union.hpp +17 -10
  145. data/vendor/datasketches-cpp/theta/include/theta_union_base_impl.hpp +10 -10
  146. data/vendor/datasketches-cpp/theta/include/theta_union_impl.hpp +3 -3
  147. data/vendor/datasketches-cpp/theta/include/theta_update_sketch_base.hpp +21 -6
  148. data/vendor/datasketches-cpp/theta/include/theta_update_sketch_base_impl.hpp +13 -3
  149. data/vendor/datasketches-cpp/theta/test/CMakeLists.txt +15 -1
  150. data/vendor/datasketches-cpp/theta/test/bit_packing_test.cpp +80 -0
  151. data/vendor/datasketches-cpp/theta/test/theta_sketch_deserialize_from_java_test.cpp +57 -0
  152. data/vendor/datasketches-cpp/theta/test/theta_sketch_serialize_for_java.cpp +61 -0
  153. data/vendor/datasketches-cpp/theta/test/theta_sketch_test.cpp +39 -188
  154. data/vendor/datasketches-cpp/theta/test/theta_union_test.cpp +25 -0
  155. data/vendor/datasketches-cpp/tuple/CMakeLists.txt +8 -7
  156. data/vendor/datasketches-cpp/tuple/include/array_of_doubles_sketch.hpp +19 -144
  157. data/vendor/datasketches-cpp/tuple/include/{array_of_doubles_a_not_b.hpp → array_tuple_a_not_b.hpp} +24 -16
  158. data/vendor/datasketches-cpp/tuple/include/{array_of_doubles_a_not_b_impl.hpp → array_tuple_a_not_b_impl.hpp} +4 -4
  159. data/vendor/datasketches-cpp/tuple/include/array_tuple_intersection.hpp +65 -0
  160. data/vendor/datasketches-cpp/tuple/include/array_tuple_intersection_impl.hpp +31 -0
  161. data/vendor/datasketches-cpp/tuple/include/array_tuple_sketch.hpp +237 -0
  162. data/vendor/datasketches-cpp/tuple/include/{array_of_doubles_sketch_impl.hpp → array_tuple_sketch_impl.hpp} +40 -41
  163. data/vendor/datasketches-cpp/tuple/include/array_tuple_union.hpp +81 -0
  164. data/vendor/datasketches-cpp/tuple/include/array_tuple_union_impl.hpp +43 -0
  165. data/vendor/datasketches-cpp/tuple/include/tuple_a_not_b.hpp +11 -2
  166. data/vendor/datasketches-cpp/tuple/include/tuple_intersection.hpp +17 -10
  167. data/vendor/datasketches-cpp/tuple/include/tuple_jaccard_similarity.hpp +2 -1
  168. data/vendor/datasketches-cpp/tuple/include/tuple_sketch.hpp +95 -32
  169. data/vendor/datasketches-cpp/tuple/include/tuple_sketch_impl.hpp +2 -1
  170. data/vendor/datasketches-cpp/tuple/include/tuple_union.hpp +19 -11
  171. data/vendor/datasketches-cpp/tuple/test/CMakeLists.txt +16 -1
  172. data/vendor/datasketches-cpp/tuple/test/aod_sketch_deserialize_from_java_test.cpp +76 -0
  173. data/vendor/datasketches-cpp/tuple/test/aod_sketch_serialize_for_java.cpp +62 -0
  174. data/vendor/datasketches-cpp/tuple/test/array_of_doubles_sketch_test.cpp +5 -129
  175. data/vendor/datasketches-cpp/tuple/test/engagement_test.cpp +85 -89
  176. data/vendor/datasketches-cpp/tuple/test/tuple_jaccard_similarity_test.cpp +3 -1
  177. data/vendor/datasketches-cpp/tuple/test/tuple_sketch_deserialize_from_java_test.cpp +47 -0
  178. data/vendor/datasketches-cpp/{python/src/datasketches.cpp → tuple/test/tuple_sketch_serialize_for_java.cpp} +16 -30
  179. data/vendor/datasketches-cpp/tuple/test/tuple_sketch_test.cpp +1 -1
  180. data/vendor/datasketches-cpp/version.cfg.in +1 -1
  181. metadata +61 -79
  182. data/vendor/datasketches-cpp/fi/test/items_sketch_string_from_java.sk +0 -0
  183. data/vendor/datasketches-cpp/fi/test/items_sketch_string_utf8_from_java.sk +0 -0
  184. data/vendor/datasketches-cpp/fi/test/longs_sketch_from_java.sk +0 -0
  185. data/vendor/datasketches-cpp/hll/test/array6_from_java.sk +0 -0
  186. data/vendor/datasketches-cpp/hll/test/compact_array4_from_java.sk +0 -0
  187. data/vendor/datasketches-cpp/hll/test/compact_set_from_java.sk +0 -0
  188. data/vendor/datasketches-cpp/hll/test/list_from_java.sk +0 -0
  189. data/vendor/datasketches-cpp/hll/test/updatable_array4_from_java.sk +0 -0
  190. data/vendor/datasketches-cpp/hll/test/updatable_set_from_java.sk +0 -0
  191. data/vendor/datasketches-cpp/kll/test/kll_sketch_from_java.sk +0 -0
  192. data/vendor/datasketches-cpp/pyproject.toml +0 -23
  193. data/vendor/datasketches-cpp/python/CMakeLists.txt +0 -81
  194. data/vendor/datasketches-cpp/python/README.md +0 -85
  195. data/vendor/datasketches-cpp/python/datasketches/PySerDe.py +0 -104
  196. data/vendor/datasketches-cpp/python/include/py_serde.hpp +0 -113
  197. data/vendor/datasketches-cpp/python/jupyter/CPCSketch.ipynb +0 -345
  198. data/vendor/datasketches-cpp/python/jupyter/FrequentItemsSketch.ipynb +0 -354
  199. data/vendor/datasketches-cpp/python/jupyter/HLLSketch.ipynb +0 -346
  200. data/vendor/datasketches-cpp/python/jupyter/KLLSketch.ipynb +0 -463
  201. data/vendor/datasketches-cpp/python/jupyter/ThetaSketchNotebook.ipynb +0 -403
  202. data/vendor/datasketches-cpp/python/pybind11Path.cmd +0 -21
  203. data/vendor/datasketches-cpp/python/src/cpc_wrapper.cpp +0 -90
  204. data/vendor/datasketches-cpp/python/src/fi_wrapper.cpp +0 -128
  205. data/vendor/datasketches-cpp/python/src/hll_wrapper.cpp +0 -134
  206. data/vendor/datasketches-cpp/python/src/kll_wrapper.cpp +0 -210
  207. data/vendor/datasketches-cpp/python/src/ks_wrapper.cpp +0 -68
  208. data/vendor/datasketches-cpp/python/src/py_serde.cpp +0 -111
  209. data/vendor/datasketches-cpp/python/src/quantiles_wrapper.cpp +0 -204
  210. data/vendor/datasketches-cpp/python/src/req_wrapper.cpp +0 -215
  211. data/vendor/datasketches-cpp/python/src/theta_wrapper.cpp +0 -172
  212. data/vendor/datasketches-cpp/python/src/vector_of_kll.cpp +0 -490
  213. data/vendor/datasketches-cpp/python/src/vo_wrapper.cpp +0 -173
  214. data/vendor/datasketches-cpp/python/tests/__init__.py +0 -16
  215. data/vendor/datasketches-cpp/python/tests/cpc_test.py +0 -64
  216. data/vendor/datasketches-cpp/python/tests/fi_test.py +0 -110
  217. data/vendor/datasketches-cpp/python/tests/hll_test.py +0 -130
  218. data/vendor/datasketches-cpp/python/tests/kll_test.py +0 -125
  219. data/vendor/datasketches-cpp/python/tests/quantiles_test.py +0 -126
  220. data/vendor/datasketches-cpp/python/tests/req_test.py +0 -126
  221. data/vendor/datasketches-cpp/python/tests/theta_test.py +0 -146
  222. data/vendor/datasketches-cpp/python/tests/vector_of_kll_test.py +0 -148
  223. data/vendor/datasketches-cpp/python/tests/vo_test.py +0 -125
  224. data/vendor/datasketches-cpp/req/test/req_float_empty_from_java.sk +0 -0
  225. data/vendor/datasketches-cpp/req/test/req_float_estimation_from_java.sk +0 -0
  226. data/vendor/datasketches-cpp/req/test/req_float_exact_from_java.sk +0 -0
  227. data/vendor/datasketches-cpp/req/test/req_float_raw_items_from_java.sk +0 -0
  228. data/vendor/datasketches-cpp/req/test/req_float_single_item_from_java.sk +0 -0
  229. data/vendor/datasketches-cpp/sampling/test/binaries_from_java.txt +0 -67
  230. data/vendor/datasketches-cpp/sampling/test/varopt_sketch_long_sampling.sk +0 -0
  231. data/vendor/datasketches-cpp/sampling/test/varopt_sketch_string_exact.sk +0 -0
  232. data/vendor/datasketches-cpp/sampling/test/varopt_union_double_sampling.sk +0 -0
  233. data/vendor/datasketches-cpp/setup.py +0 -110
  234. data/vendor/datasketches-cpp/theta/test/theta_compact_empty_from_java.sk +0 -0
  235. data/vendor/datasketches-cpp/theta/test/theta_compact_estimation_from_java.sk +0 -0
  236. data/vendor/datasketches-cpp/theta/test/theta_compact_exact_from_java.sk +0 -0
  237. data/vendor/datasketches-cpp/theta/test/theta_compact_single_item_from_java.sk +0 -0
  238. data/vendor/datasketches-cpp/tuple/include/array_of_doubles_intersection.hpp +0 -52
  239. data/vendor/datasketches-cpp/tuple/include/array_of_doubles_union.hpp +0 -81
  240. data/vendor/datasketches-cpp/tuple/include/array_of_doubles_union_impl.hpp +0 -43
  241. data/vendor/datasketches-cpp/tuple/test/aod_1_compact_empty_from_java.sk +0 -1
  242. data/vendor/datasketches-cpp/tuple/test/aod_1_compact_estimation_from_java.sk +0 -0
  243. data/vendor/datasketches-cpp/tuple/test/aod_1_compact_non_empty_no_entries_from_java.sk +0 -0
  244. data/vendor/datasketches-cpp/tuple/test/aod_2_compact_exact_from_java.sk +0 -0
  245. data/vendor/datasketches-cpp/tuple/test/aod_3_compact_empty_from_java.sk +0 -1
@@ -0,0 +1,543 @@
1
+ /*
2
+ * Licensed to the Apache Software Foundation (ASF) under one
3
+ * or more contributor license agreements. See the NOTICE file
4
+ * distributed with this work for additional information
5
+ * regarding copyright ownership. The ASF licenses this file
6
+ * to you under the Apache License, Version 2.0 (the
7
+ * "License"); you may not use this file except in compliance
8
+ * with the License. You may obtain a copy of the License at
9
+ *
10
+ * http://www.apache.org/licenses/LICENSE-2.0
11
+ *
12
+ * Unless required by applicable law or agreed to in writing,
13
+ * software distributed under the License is distributed on an
14
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
15
+ * KIND, either express or implied. See the License for the
16
+ * specific language governing permissions and limitations
17
+ * under the License.
18
+ */
19
+
20
+ #ifndef DENSITY_SKETCH_IMPL_HPP_
21
+ #define DENSITY_SKETCH_IMPL_HPP_
22
+
23
+ #include <algorithm>
24
+ #include <sstream>
25
+
26
+ #include "memory_operations.hpp"
27
+ #include "conditional_forward.hpp"
28
+
29
+ namespace datasketches {
30
+
31
+ template<typename T, typename K, typename A>
32
+ density_sketch<T, K, A>::density_sketch(uint16_t k, uint32_t dim, const K& kernel, const A& allocator):
33
+ kernel_(kernel),
34
+ k_(k),
35
+ dim_(dim),
36
+ num_retained_(0),
37
+ n_(0),
38
+ levels_(1, Level(allocator), allocator)
39
+ {
40
+ check_k(k);
41
+ }
42
+
43
+ template<typename T, typename K, typename A>
44
+ density_sketch<T, K, A>::density_sketch(uint16_t k, uint32_t dim, uint32_t num_retained, uint64_t n,
45
+ Levels&& levels, const K& kernel):
46
+ kernel_(kernel),
47
+ k_(k),
48
+ dim_(dim),
49
+ num_retained_(num_retained),
50
+ n_(n),
51
+ levels_(std::move(levels))
52
+ {
53
+ check_k(k);
54
+ }
55
+
56
+ template<typename T, typename K, typename A>
57
+ uint16_t density_sketch<T, K, A>::get_k() const {
58
+ return k_;
59
+ }
60
+
61
+ template<typename T, typename K, typename A>
62
+ uint32_t density_sketch<T, K, A>::get_dim() const {
63
+ return dim_;
64
+ }
65
+
66
+ template<typename T, typename K, typename A>
67
+ bool density_sketch<T, K, A>::is_empty() const {
68
+ return num_retained_ == 0;
69
+ }
70
+
71
+ template<typename T, typename K, typename A>
72
+ uint64_t density_sketch<T, K, A>::get_n() const {
73
+ return n_;
74
+ }
75
+
76
+ template<typename T, typename K, typename A>
77
+ uint32_t density_sketch<T, K, A>::get_num_retained() const {
78
+ return num_retained_;
79
+ }
80
+
81
+ template<typename T, typename K, typename A>
82
+ bool density_sketch<T, K, A>::is_estimation_mode() const {
83
+ return levels_.size() > 1;
84
+ }
85
+
86
+ template<typename T, typename K, typename A>
87
+ template<typename FwdVector>
88
+ void density_sketch<T, K, A>::update(FwdVector&& point) {
89
+ if (point.size() != dim_) throw std::invalid_argument("dimension mismatch");
90
+ while (num_retained_ >= k_ * levels_.size()) compact();
91
+ levels_[0].push_back(std::forward<FwdVector>(point));
92
+ ++num_retained_;
93
+ ++n_;
94
+ }
95
+
96
+ template<typename T, typename K, typename A>
97
+ template<typename FwdSketch>
98
+ void density_sketch<T, K, A>::merge(FwdSketch&& other) {
99
+ if (other.is_empty()) return;
100
+ if (other.dim_ != dim_) throw std::invalid_argument("dimension mismatch");
101
+ while (levels_.size() < other.levels_.size()) levels_.push_back(Level(levels_.get_allocator()));
102
+ for (unsigned height = 0; height < other.levels_.size(); ++height) {
103
+ std::copy(
104
+ forward_begin(conditional_forward<FwdSketch>(other.levels_[height])),
105
+ forward_end(conditional_forward<FwdSketch>(other.levels_[height])),
106
+ back_inserter(levels_[height])
107
+ );
108
+ }
109
+ num_retained_ += other.num_retained_;
110
+ n_ += other.n_;
111
+ while (num_retained_ >= k_ * levels_.size()) compact();
112
+ }
113
+
114
+ template<typename T, typename K, typename A>
115
+ T density_sketch<T, K, A>::get_estimate(const std::vector<T>& point) const {
116
+ if (is_empty()) throw std::runtime_error("operation is undefined for an empty sketch");
117
+ T density = 0;
118
+ for (unsigned height = 0; height < levels_.size(); ++height) {
119
+ for (const auto& p: levels_[height]) {
120
+ density += (1 << height) * kernel_(p, point) / n_;
121
+ }
122
+ }
123
+ return density;
124
+ }
125
+
126
+ template<typename T, typename K, typename A>
127
+ A density_sketch<T, K, A>::get_allocator() const {
128
+ return levels_.get_allocator();
129
+ }
130
+
131
+ template<typename T, typename K, typename A>
132
+ void density_sketch<T, K, A>::compact() {
133
+ for (unsigned height = 0; height < levels_.size(); ++height) {
134
+ if (levels_[height].size() >= k_) {
135
+ if (height + 1 >= levels_.size()) levels_.push_back(Level(levels_.get_allocator()));
136
+ compact_level(height);
137
+ break;
138
+ }
139
+ }
140
+ }
141
+
142
+ template<typename T, typename K, typename A>
143
+ void density_sketch<T, K, A>::compact_level(unsigned height) {
144
+ auto& level = levels_[height];
145
+ std::vector<bool> bits(level.size());
146
+ bits[0] = random_utils::random_bit();
147
+ std::random_shuffle(level.begin(), level.end());
148
+ for (unsigned i = 1; i < level.size(); ++i) {
149
+ T delta = 0;
150
+ for (unsigned j = 0; j < i; ++j) {
151
+ delta += (bits[j] ? 1 : -1) * kernel_(level[i], level[j]);
152
+ }
153
+ bits[i] = delta < 0;
154
+ }
155
+ for (unsigned i = 0; i < level.size(); ++i) {
156
+ if (bits[i]) {
157
+ levels_[height + 1].push_back(std::move(level[i]));
158
+ } else {
159
+ --num_retained_;
160
+ }
161
+ }
162
+ level.clear();
163
+ }
164
+
165
+ /* Serialized sketch layout:
166
+ * Int || Start Byte Addr:
167
+ * Addr:
168
+ * || 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 |
169
+ * 0 || Preamble_Ints | SerVer | FamID | Flags |------- K -------|---- unused -----|
170
+ *
171
+ * || 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 |
172
+ * 2 ||------------- Num Dimensions --------------|------ Num Retained Items ---------|
173
+ *
174
+ * || 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 |
175
+ * 4 ||---------------------------Items Seen Count (N)--------------------------------|
176
+ *
177
+ * Ints 2 and 3 are omitted when the sketch is empty, meaning Num Dimensions is stored at
178
+ * offset 8 in that case. Otherwise, Int 5 is the start of level data, consisting of the
179
+ * size of the level (as a uint32 value) followed by that number of points, with
180
+ * Num Dimensions per point.
181
+ */
182
+
183
+ template<typename T, typename K, typename A>
184
+ void density_sketch<T, K, A>::serialize(std::ostream& os) const {
185
+ const uint8_t preamble_ints = is_empty() ? PREAMBLE_INTS_SHORT : PREAMBLE_INTS_LONG;
186
+ write(os, preamble_ints);
187
+ const uint8_t ser_ver = SERIAL_VERSION;
188
+ write(os, ser_ver);
189
+ const uint8_t family = FAMILY_ID;
190
+ write(os, family);
191
+
192
+ // only empty is a valid flag
193
+ const uint8_t flags_byte = (is_empty() ? 1 << flags::IS_EMPTY : 0);
194
+ write(os, flags_byte);
195
+ write(os, k_);
196
+ const uint16_t unused = 0;
197
+ write(os, unused);
198
+ write(os, dim_);
199
+
200
+ if (is_empty())
201
+ return;
202
+
203
+ write(os, num_retained_);
204
+ write(os, n_);
205
+
206
+ // levels array -- uint32_t since a single level may be larger than k
207
+ size_t pt_size = sizeof(T) * dim_;
208
+ for (const Level& lvl : levels_) {
209
+ const uint32_t level_size = static_cast<uint32_t>(lvl.size());
210
+ write(os, level_size);
211
+ for (const Vector& pt : lvl) {
212
+ write(os, pt.data(), pt_size);
213
+ }
214
+ }
215
+ }
216
+
217
+ template<typename T, typename K, typename A>
218
+ auto density_sketch<T, K, A>::serialize(unsigned header_size_bytes) const -> vector_bytes {
219
+ const uint8_t preamble_ints = (is_empty() ? PREAMBLE_INTS_SHORT : PREAMBLE_INTS_LONG);
220
+
221
+ // pre-compute size
222
+ size_t size = header_size_bytes + preamble_ints * sizeof(uint32_t);
223
+ if (!is_empty())
224
+ for (const Level& lvl : levels_)
225
+ size += sizeof(uint32_t) + (lvl.size() * dim_ * sizeof(T));
226
+
227
+ vector_bytes bytes(size, 0, levels_.get_allocator());
228
+ uint8_t* ptr = bytes.data() + header_size_bytes;
229
+ const uint8_t* end_ptr = ptr + size;
230
+
231
+ ptr += copy_to_mem(preamble_ints, ptr);
232
+ const uint8_t ser_ver = SERIAL_VERSION;
233
+ ptr += copy_to_mem(ser_ver, ptr);
234
+ const uint8_t family = FAMILY_ID;
235
+ ptr += copy_to_mem(family, ptr);
236
+
237
+ // empty is the only valid flat
238
+ const uint8_t flags_byte = (is_empty() ? 1 << flags::IS_EMPTY : 0);
239
+ ptr += copy_to_mem(flags_byte, ptr);
240
+ ptr += copy_to_mem(k_, ptr);
241
+ ptr += sizeof(uint16_t); // 2 unused bytes
242
+ ptr += copy_to_mem(dim_, ptr);
243
+
244
+ if (is_empty())
245
+ return bytes;
246
+
247
+ ptr += copy_to_mem(num_retained_, ptr);
248
+ ptr += copy_to_mem(n_, ptr);
249
+
250
+ // levels array -- uint32_t since a single level may be larger than k
251
+ size_t pt_size = sizeof(T) * dim_;
252
+ for (const Level& lvl : levels_) {
253
+ ptr += copy_to_mem(static_cast<uint32_t>(lvl.size()), ptr);
254
+ for (const Vector& pt : lvl) {
255
+ ptr += copy_to_mem(pt.data(), ptr, pt_size);
256
+ }
257
+ }
258
+
259
+ if (ptr != end_ptr)
260
+ throw std::runtime_error("Actual output size does not equal expected output size");
261
+
262
+ return bytes;
263
+ }
264
+
265
+ template<typename T, typename K, typename A>
266
+ density_sketch<T, K, A> density_sketch<T, K, A>::deserialize(std::istream& is, const K& kernel, const A& allocator) {
267
+ const auto preamble_ints = read<uint8_t>(is);
268
+ const auto serial_version = read<uint8_t>(is);
269
+ const auto family_id = read<uint8_t>(is);
270
+ const auto flags_byte = read<uint8_t>(is);
271
+ const auto k = read<uint16_t>(is);
272
+ read<uint16_t>(is); // unused
273
+ const auto dim = read<uint32_t>(is);
274
+
275
+ check_k(k); // do we have constraints?
276
+ check_serial_version(serial_version); // a little redundant with the header check
277
+ check_family_id(family_id);
278
+ check_header_validity(preamble_ints, flags_byte, serial_version);
279
+
280
+ if (!is.good()) throw std::runtime_error("error reading from std::istream");
281
+ const bool is_empty = (flags_byte & (1 << flags::IS_EMPTY)) > 0;
282
+ if (is_empty) {
283
+ return density_sketch(k, dim, kernel, allocator);
284
+ }
285
+
286
+ const auto num_retained = read<uint32_t>(is);
287
+ const auto n = read<uint64_t>(is);
288
+
289
+ // levels arrays
290
+ size_t pt_size = sizeof(T) * dim;
291
+ Levels levels(allocator);
292
+ int64_t num_to_read = num_retained; // num_retrained is uint32_t so this allows error checking
293
+ while (num_to_read > 0) {
294
+ const auto level_size = read<uint32_t>(is);
295
+ Level lvl(allocator);
296
+ lvl.reserve(level_size);
297
+ for (uint32_t i = 0; i < level_size; ++i) {
298
+ Vector pt(dim, 0, allocator);
299
+ read(is, pt.data(), pt_size);
300
+ lvl.push_back(pt);
301
+ }
302
+ levels.push_back(lvl);
303
+ num_to_read -= lvl.size();
304
+ }
305
+
306
+ if (num_to_read != 0)
307
+ throw std::runtime_error("Error deserializing sketch: Incorrect number of items read");
308
+ if (!is.good()) throw std::runtime_error("error reading from std::istream");
309
+
310
+ return density_sketch(k, dim, num_retained, n, std::move(levels), kernel);
311
+ }
312
+
313
+ template<typename T, typename K, typename A>
314
+ density_sketch<T, K, A> density_sketch<T, K, A>::deserialize(const void* bytes, size_t size, const K& kernel, const A& allocator) {
315
+ ensure_minimum_memory(size, PREAMBLE_INTS_SHORT * sizeof(uint32_t));
316
+ const char* ptr = static_cast<const char*>(bytes);
317
+ const char* end_ptr = static_cast<const char*>(bytes) + size;
318
+ uint8_t preamble_ints;
319
+ ptr += copy_from_mem(ptr, preamble_ints);
320
+ uint8_t serial_version;
321
+ ptr += copy_from_mem(ptr, serial_version);
322
+ uint8_t family_id;
323
+ ptr += copy_from_mem(ptr, family_id);
324
+ uint8_t flags_byte;
325
+ ptr += copy_from_mem(ptr, flags_byte);
326
+ uint16_t k;
327
+ ptr += copy_from_mem(ptr, k);
328
+ uint16_t unused;
329
+ ptr += copy_from_mem(ptr, unused);
330
+ uint32_t dim;
331
+ ptr += copy_from_mem(ptr, dim);
332
+
333
+ check_k(k);
334
+ check_serial_version(serial_version); // a little redundant with the header check
335
+ check_family_id(family_id);
336
+ check_header_validity(preamble_ints, flags_byte, serial_version);
337
+
338
+ const bool is_empty = (flags_byte & (1 << flags::IS_EMPTY)) > 0;
339
+ if (is_empty) {
340
+ return density_sketch(k, dim, kernel, allocator);
341
+ }
342
+
343
+ ensure_minimum_memory(size, PREAMBLE_INTS_LONG * sizeof(uint32_t));
344
+ uint32_t num_retained;
345
+ ptr += copy_from_mem(ptr, num_retained);
346
+ uint64_t n;
347
+ ptr += copy_from_mem(ptr, n);
348
+
349
+ // Predicting the number of levels seems hard so determining the exact remaining
350
+ // size is also hard. But we need at least num_retained * dim * sizeof(T)
351
+ // bytes for the points so we can check that.
352
+ size_t pt_size = sizeof(T) * dim;
353
+ ensure_minimum_memory(end_ptr - ptr, num_retained * pt_size);
354
+
355
+ // levels arrays
356
+ Levels levels(allocator);
357
+ int64_t num_to_read = num_retained; // num_retained is uint32_t so this allows error checking
358
+ while (num_to_read > 0) {
359
+ uint32_t level_size;
360
+ ptr += copy_from_mem(ptr, level_size);
361
+ ensure_minimum_memory(end_ptr - ptr, level_size * pt_size);
362
+ Level lvl(allocator);
363
+ lvl.reserve(level_size);
364
+ for (uint32_t i = 0; i < level_size; ++i) {
365
+ Vector pt(dim, 0, allocator);
366
+ ptr += copy_from_mem(ptr, pt.data(), pt_size);
367
+ lvl.push_back(pt);
368
+ }
369
+ levels.push_back(lvl);
370
+ num_to_read -= lvl.size();
371
+ }
372
+
373
+ if (num_to_read != 0)
374
+ throw std::runtime_error("Error deserializing sketch: Incorrect number of items read");
375
+ if (ptr > end_ptr) throw std::runtime_error("Error deserializing sketch: Read beyond provided memory");
376
+
377
+ return density_sketch(k, dim, num_retained, n, std::move(levels), kernel);
378
+ }
379
+
380
+ template<typename T, typename K, typename A>
381
+ void density_sketch<T, K, A>::check_k(uint16_t k) {
382
+ if (k < 2)
383
+ throw std::invalid_argument("k must be > 1. Found: " + std::to_string(k));
384
+ }
385
+
386
+ template<typename T, typename K, typename A>
387
+ void density_sketch<T, K, A>::check_serial_version(uint8_t serial_version) {
388
+ if (serial_version == SERIAL_VERSION)
389
+ return;
390
+ else
391
+ throw std::invalid_argument("Possible corruption. Unrecognized serialization version: " + std::to_string(serial_version));
392
+ }
393
+
394
+ template<typename T, typename K, typename A>
395
+ void density_sketch<T, K, A>::check_family_id(uint8_t family_id) {
396
+ if (family_id == FAMILY_ID)
397
+ return;
398
+ else
399
+ throw std::invalid_argument("Possible corruption. Family id does not indicate density sketch: " + std::to_string(family_id));
400
+ }
401
+
402
+ template<typename T, typename K, typename A>
403
+ void density_sketch<T, K, A>::check_header_validity(uint8_t preamble_ints, uint8_t flags_byte, uint8_t serial_version) {
404
+ const bool empty = (flags_byte & (1 << flags::IS_EMPTY)) > 0;
405
+
406
+ if ((empty && preamble_ints == PREAMBLE_INTS_SHORT)
407
+ || (!empty && preamble_ints == PREAMBLE_INTS_LONG))
408
+ return;
409
+ else {
410
+ std::ostringstream os;
411
+ os << "Possible sketch corruption. Inconsistent state: "
412
+ << "preamble_ints = " << preamble_ints
413
+ << ", empty = " << (empty ? "true" : "false")
414
+ << ", serialization_version = " << serial_version;
415
+ throw std::invalid_argument(os.str());
416
+ }
417
+ }
418
+
419
+ template<typename T, typename K, typename A>
420
+ string<A> density_sketch<T, K, A>::to_string(bool print_levels, bool print_items) const {
421
+ // Using a temporary stream for implementation here does not comply with AllocatorAwareContainer requirements.
422
+ // The stream does not support passing an allocator instance, and alternatives are complicated.
423
+ std::ostringstream os;
424
+ os << "### Density sketch summary:" << std::endl;
425
+ os << " K : " << k_ << std::endl;
426
+ os << " Dim : " << dim_ << std::endl;
427
+ os << " Empty : " << (is_empty() ? "true" : "false") << std::endl;
428
+ os << " N : " << n_ << std::endl;
429
+ os << " Retained items : " << num_retained_ << std::endl;
430
+ os << " Estimation mode: " << (is_estimation_mode() ? "true" : "false") << std::endl;
431
+ os << " Levels : " << levels_.size() << std::endl;
432
+ os << "### End sketch summary" << std::endl;
433
+
434
+ if (print_levels) {
435
+ os << "### Density sketch levels:" << std::endl;
436
+ os << " height: size" << std::endl;
437
+ for (unsigned height = 0; height < levels_.size(); ++height) {
438
+ os << " " << height << ": "
439
+ << levels_[height].size() << std::endl;
440
+ }
441
+ os << "### End sketch levels" << std::endl;
442
+ }
443
+
444
+ if (print_items) {
445
+ os << "### Density sketch data:" << std::endl;
446
+ for (unsigned height = 0; height < levels_.size(); ++height) {
447
+ os << " level " << height << ": " << std::endl;
448
+ for (const auto& point: levels_[height]) {
449
+ os << " [";
450
+ bool first = true;
451
+ for (auto value: point) {
452
+ if (first) {
453
+ first = false;
454
+ } else {
455
+ os << ", ";
456
+ }
457
+ os << value;
458
+ }
459
+ os << "]" << std::endl;
460
+ }
461
+ }
462
+ os << "### End sketch data" << std::endl;
463
+ }
464
+ return string<A>(os.str().c_str(), levels_.get_allocator());
465
+ }
466
+
467
+ template<typename T, typename K, typename A>
468
+ auto density_sketch<T, K, A>::begin() const -> const_iterator {
469
+ return const_iterator(levels_.begin(), levels_.end());
470
+ }
471
+
472
+ template<typename T, typename K, typename A>
473
+ auto density_sketch<T, K, A>::end() const -> const_iterator {
474
+ return const_iterator(levels_.end(), levels_.end());
475
+ }
476
+
477
+ // iterator
478
+
479
+ template<typename T, typename K, typename A>
480
+ density_sketch<T, K, A>::const_iterator::const_iterator(LevelsIterator begin, LevelsIterator end):
481
+ levels_it_(begin),
482
+ levels_end_(end),
483
+ level_it_(),
484
+ height_(0)
485
+ {
486
+ // skip empty levels
487
+ while (levels_it_ != levels_end_) {
488
+ level_it_ = levels_it_->begin();
489
+ if (level_it_ != levels_it_->end()) break;
490
+ ++levels_it_;
491
+ ++height_;
492
+ }
493
+ }
494
+
495
+ template<typename T, typename K, typename A>
496
+ auto density_sketch<T, K, A>::const_iterator::operator++() -> const_iterator& {
497
+ ++level_it_;
498
+ if (level_it_ == levels_it_->end()) {
499
+ ++levels_it_;
500
+ ++height_;
501
+ // skip empty levels
502
+ while (levels_it_ != levels_end_) {
503
+ level_it_ = levels_it_->begin();
504
+ if (level_it_ != levels_it_->end()) break;
505
+ ++levels_it_;
506
+ ++height_;
507
+ }
508
+ }
509
+ return *this;
510
+ }
511
+
512
+ template<typename T, typename K, typename A>
513
+ auto density_sketch<T, K, A>::const_iterator::operator++(int) -> const_iterator& {
514
+ const_iterator tmp(*this);
515
+ operator++();
516
+ return tmp;
517
+ }
518
+
519
+ template<typename T, typename K, typename A>
520
+ bool density_sketch<T, K, A>::const_iterator::operator==(const const_iterator& other) const {
521
+ if (levels_it_ != other.levels_it_) return false;
522
+ if (levels_it_ == levels_end_) return true;
523
+ return level_it_ == other.level_it_;
524
+ }
525
+
526
+ template<typename T, typename K, typename A>
527
+ bool density_sketch<T, K, A>::const_iterator::operator!=(const const_iterator& other) const {
528
+ return !operator==(other);
529
+ }
530
+
531
+ template<typename T, typename K, typename A>
532
+ auto density_sketch<T, K, A>::const_iterator::operator*() const -> const value_type {
533
+ return value_type(*level_it_, 1ULL << height_);
534
+ }
535
+
536
+ template<typename T, typename K, typename A>
537
+ auto density_sketch<T, K, A>::const_iterator::operator->() const -> const return_value_holder<value_type> {
538
+ return **this;
539
+ }
540
+
541
+ } /* namespace datasketches */
542
+
543
+ #endif
@@ -15,8 +15,20 @@
15
15
  # specific language governing permissions and limitations
16
16
  # under the License.
17
17
 
18
- name = 'datasketches'
18
+ add_executable(density_test)
19
19
 
20
- from .PySerDe import *
20
+ target_link_libraries(density_test density common_test_lib)
21
21
 
22
- from _datasketches import *
22
+ set_target_properties(density_test PROPERTIES
23
+ CXX_STANDARD_REQUIRED YES
24
+ )
25
+
26
+ add_test(
27
+ NAME density_test
28
+ COMMAND density_test
29
+ )
30
+
31
+ target_sources(density_test
32
+ PRIVATE
33
+ density_sketch_test.cpp
34
+ )