datasketches 0.3.1 → 0.4.0
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +4 -4
- data/CHANGELOG.md +10 -0
- data/NOTICE +1 -1
- data/README.md +0 -2
- data/ext/datasketches/cpc_wrapper.cpp +3 -3
- data/ext/datasketches/kll_wrapper.cpp +0 -10
- data/lib/datasketches/version.rb +1 -1
- data/lib/datasketches.rb +1 -1
- data/vendor/datasketches-cpp/CMakeLists.txt +23 -20
- data/vendor/datasketches-cpp/CODE_OF_CONDUCT.md +3 -0
- data/vendor/datasketches-cpp/CONTRIBUTING.md +50 -0
- data/vendor/datasketches-cpp/Doxyfile +2827 -0
- data/vendor/datasketches-cpp/LICENSE +0 -76
- data/vendor/datasketches-cpp/README.md +1 -3
- data/vendor/datasketches-cpp/common/CMakeLists.txt +12 -11
- data/vendor/datasketches-cpp/common/include/MurmurHash3.h +25 -27
- data/vendor/datasketches-cpp/common/include/common_defs.hpp +15 -10
- data/vendor/datasketches-cpp/common/include/count_zeros.hpp +11 -2
- data/vendor/datasketches-cpp/common/include/kolmogorov_smirnov.hpp +9 -6
- data/vendor/datasketches-cpp/common/include/memory_operations.hpp +5 -4
- data/vendor/datasketches-cpp/common/include/optional.hpp +148 -0
- data/vendor/datasketches-cpp/common/include/quantiles_sorted_view.hpp +95 -2
- data/vendor/datasketches-cpp/common/include/quantiles_sorted_view_impl.hpp +1 -1
- data/vendor/datasketches-cpp/common/include/serde.hpp +69 -20
- data/vendor/datasketches-cpp/common/test/CMakeLists.txt +2 -2
- data/vendor/datasketches-cpp/common/test/integration_test.cpp +6 -0
- data/vendor/datasketches-cpp/common/test/optional_test.cpp +85 -0
- data/vendor/datasketches-cpp/common/test/test_allocator.hpp +14 -14
- data/vendor/datasketches-cpp/{python/src/__init__.py → count/CMakeLists.txt} +25 -1
- data/vendor/datasketches-cpp/count/include/count_min.hpp +405 -0
- data/vendor/datasketches-cpp/count/include/count_min_impl.hpp +497 -0
- data/vendor/datasketches-cpp/{MANIFEST.in → count/test/CMakeLists.txt} +23 -20
- data/vendor/datasketches-cpp/count/test/count_min_allocation_test.cpp +155 -0
- data/vendor/datasketches-cpp/count/test/count_min_test.cpp +303 -0
- data/vendor/datasketches-cpp/cpc/include/cpc_common.hpp +14 -20
- data/vendor/datasketches-cpp/cpc/include/cpc_compressor.hpp +7 -4
- data/vendor/datasketches-cpp/cpc/include/cpc_compressor_impl.hpp +17 -17
- data/vendor/datasketches-cpp/cpc/include/cpc_confidence.hpp +3 -3
- data/vendor/datasketches-cpp/cpc/include/cpc_sketch.hpp +40 -40
- data/vendor/datasketches-cpp/cpc/include/cpc_sketch_impl.hpp +14 -11
- data/vendor/datasketches-cpp/cpc/include/cpc_union.hpp +35 -11
- data/vendor/datasketches-cpp/cpc/include/cpc_union_impl.hpp +8 -8
- data/vendor/datasketches-cpp/cpc/include/cpc_util.hpp +16 -8
- data/vendor/datasketches-cpp/cpc/include/u32_table.hpp +3 -2
- data/vendor/datasketches-cpp/cpc/include/u32_table_impl.hpp +5 -5
- data/vendor/datasketches-cpp/cpc/test/CMakeLists.txt +20 -7
- data/vendor/datasketches-cpp/cpc/test/cpc_sketch_deserialize_from_java_test.cpp +60 -0
- data/vendor/datasketches-cpp/cpc/test/cpc_sketch_serialize_for_java.cpp +38 -0
- data/vendor/datasketches-cpp/cpc/test/cpc_sketch_test.cpp +4 -29
- data/vendor/datasketches-cpp/cpc/test/cpc_union_test.cpp +4 -4
- data/vendor/datasketches-cpp/{tox.ini → density/CMakeLists.txt} +24 -8
- data/vendor/datasketches-cpp/density/include/density_sketch.hpp +256 -0
- data/vendor/datasketches-cpp/density/include/density_sketch_impl.hpp +543 -0
- data/vendor/datasketches-cpp/{python/datasketches/__init__.py → density/test/CMakeLists.txt} +15 -3
- data/vendor/datasketches-cpp/density/test/density_sketch_test.cpp +244 -0
- data/vendor/datasketches-cpp/fi/include/frequent_items_sketch.hpp +21 -9
- data/vendor/datasketches-cpp/fi/include/frequent_items_sketch_impl.hpp +6 -4
- data/vendor/datasketches-cpp/fi/include/reverse_purge_hash_map.hpp +9 -3
- data/vendor/datasketches-cpp/fi/test/CMakeLists.txt +14 -1
- data/vendor/datasketches-cpp/fi/test/frequent_items_sketch_deserialize_from_java_test.cpp +95 -0
- data/vendor/datasketches-cpp/fi/test/frequent_items_sketch_serialize_for_java.cpp +83 -0
- data/vendor/datasketches-cpp/fi/test/frequent_items_sketch_test.cpp +3 -42
- data/vendor/datasketches-cpp/hll/include/CouponList-internal.hpp +2 -2
- data/vendor/datasketches-cpp/hll/include/CouponList.hpp +3 -1
- data/vendor/datasketches-cpp/hll/include/Hll4Array-internal.hpp +19 -11
- data/vendor/datasketches-cpp/hll/include/Hll4Array.hpp +2 -5
- data/vendor/datasketches-cpp/hll/include/Hll6Array-internal.hpp +19 -7
- data/vendor/datasketches-cpp/hll/include/Hll6Array.hpp +1 -1
- data/vendor/datasketches-cpp/hll/include/Hll8Array-internal.hpp +98 -42
- data/vendor/datasketches-cpp/hll/include/Hll8Array.hpp +2 -0
- data/vendor/datasketches-cpp/hll/include/HllArray-internal.hpp +94 -61
- data/vendor/datasketches-cpp/hll/include/HllArray.hpp +20 -8
- data/vendor/datasketches-cpp/hll/include/HllSketch-internal.hpp +4 -4
- data/vendor/datasketches-cpp/hll/include/HllSketchImpl.hpp +3 -1
- data/vendor/datasketches-cpp/hll/include/HllSketchImplFactory.hpp +3 -21
- data/vendor/datasketches-cpp/hll/include/HllUnion-internal.hpp +8 -0
- data/vendor/datasketches-cpp/hll/include/HllUtil.hpp +14 -18
- data/vendor/datasketches-cpp/hll/include/coupon_iterator-internal.hpp +1 -1
- data/vendor/datasketches-cpp/hll/include/coupon_iterator.hpp +8 -2
- data/vendor/datasketches-cpp/hll/include/hll.hpp +79 -65
- data/vendor/datasketches-cpp/hll/test/CMakeLists.txt +14 -1
- data/vendor/datasketches-cpp/hll/test/HllUnionTest.cpp +7 -1
- data/vendor/datasketches-cpp/hll/test/ToFromByteArrayTest.cpp +0 -68
- data/vendor/datasketches-cpp/hll/test/hll_sketch_deserialize_from_java_test.cpp +69 -0
- data/vendor/datasketches-cpp/hll/test/hll_sketch_serialize_for_java.cpp +52 -0
- data/vendor/datasketches-cpp/kll/include/kll_helper.hpp +0 -1
- data/vendor/datasketches-cpp/kll/include/kll_helper_impl.hpp +2 -2
- data/vendor/datasketches-cpp/kll/include/kll_sketch.hpp +79 -53
- data/vendor/datasketches-cpp/kll/include/kll_sketch_impl.hpp +61 -132
- data/vendor/datasketches-cpp/kll/test/CMakeLists.txt +14 -1
- data/vendor/datasketches-cpp/kll/test/kll_sketch_deserialize_from_java_test.cpp +103 -0
- data/vendor/datasketches-cpp/kll/test/kll_sketch_serialize_for_java.cpp +62 -0
- data/vendor/datasketches-cpp/kll/test/kll_sketch_test.cpp +5 -40
- data/vendor/datasketches-cpp/quantiles/include/quantiles_sketch.hpp +76 -54
- data/vendor/datasketches-cpp/quantiles/include/quantiles_sketch_impl.hpp +66 -136
- data/vendor/datasketches-cpp/quantiles/test/CMakeLists.txt +14 -1
- data/vendor/datasketches-cpp/quantiles/test/quantiles_sketch_deserialize_from_java_test.cpp +84 -0
- data/vendor/datasketches-cpp/quantiles/test/quantiles_sketch_serialize_for_java.cpp +52 -0
- data/vendor/datasketches-cpp/quantiles/test/quantiles_sketch_test.cpp +15 -39
- data/vendor/datasketches-cpp/req/include/req_common.hpp +7 -3
- data/vendor/datasketches-cpp/req/include/req_compactor_impl.hpp +2 -4
- data/vendor/datasketches-cpp/req/include/req_sketch.hpp +105 -26
- data/vendor/datasketches-cpp/req/include/req_sketch_impl.hpp +50 -111
- data/vendor/datasketches-cpp/req/test/CMakeLists.txt +14 -1
- data/vendor/datasketches-cpp/req/test/req_sketch_deserialize_from_java_test.cpp +55 -0
- data/vendor/datasketches-cpp/{tuple/include/array_of_doubles_intersection_impl.hpp → req/test/req_sketch_serialize_for_java.cpp} +12 -7
- data/vendor/datasketches-cpp/req/test/req_sketch_test.cpp +3 -89
- data/vendor/datasketches-cpp/sampling/CMakeLists.txt +4 -0
- data/vendor/datasketches-cpp/sampling/include/ebpps_sample.hpp +210 -0
- data/vendor/datasketches-cpp/sampling/include/ebpps_sample_impl.hpp +535 -0
- data/vendor/datasketches-cpp/sampling/include/ebpps_sketch.hpp +281 -0
- data/vendor/datasketches-cpp/sampling/include/ebpps_sketch_impl.hpp +531 -0
- data/vendor/datasketches-cpp/sampling/include/var_opt_sketch.hpp +89 -32
- data/vendor/datasketches-cpp/sampling/include/var_opt_sketch_impl.hpp +33 -19
- data/vendor/datasketches-cpp/sampling/include/var_opt_union.hpp +13 -10
- data/vendor/datasketches-cpp/sampling/include/var_opt_union_impl.hpp +23 -19
- data/vendor/datasketches-cpp/sampling/test/CMakeLists.txt +55 -8
- data/vendor/datasketches-cpp/sampling/test/ebpps_allocation_test.cpp +96 -0
- data/vendor/datasketches-cpp/sampling/test/ebpps_sample_test.cpp +137 -0
- data/vendor/datasketches-cpp/sampling/test/ebpps_sketch_test.cpp +266 -0
- data/vendor/datasketches-cpp/sampling/test/var_opt_sketch_deserialize_from_java_test.cpp +81 -0
- data/vendor/datasketches-cpp/sampling/test/var_opt_sketch_serialize_for_java.cpp +54 -0
- data/vendor/datasketches-cpp/sampling/test/var_opt_sketch_test.cpp +33 -51
- data/vendor/datasketches-cpp/sampling/test/var_opt_union_deserialize_from_java_test.cpp +50 -0
- data/vendor/datasketches-cpp/sampling/test/var_opt_union_serialize_for_java.cpp +56 -0
- data/vendor/datasketches-cpp/sampling/test/var_opt_union_test.cpp +0 -20
- data/vendor/datasketches-cpp/theta/CMakeLists.txt +1 -0
- data/vendor/datasketches-cpp/theta/include/bit_packing.hpp +6279 -0
- data/vendor/datasketches-cpp/theta/include/bounds_on_ratios_in_sampled_sets.hpp +1 -0
- data/vendor/datasketches-cpp/theta/include/bounds_on_ratios_in_theta_sketched_sets.hpp +7 -6
- data/vendor/datasketches-cpp/theta/include/compact_theta_sketch_parser.hpp +14 -8
- data/vendor/datasketches-cpp/theta/include/compact_theta_sketch_parser_impl.hpp +60 -46
- data/vendor/datasketches-cpp/theta/include/theta_a_not_b.hpp +20 -5
- data/vendor/datasketches-cpp/theta/include/theta_constants.hpp +10 -4
- data/vendor/datasketches-cpp/theta/include/theta_helpers.hpp +3 -1
- data/vendor/datasketches-cpp/theta/include/theta_intersection.hpp +13 -5
- data/vendor/datasketches-cpp/theta/include/theta_intersection_base_impl.hpp +5 -5
- data/vendor/datasketches-cpp/theta/include/theta_intersection_impl.hpp +3 -3
- data/vendor/datasketches-cpp/theta/include/theta_jaccard_similarity.hpp +2 -1
- data/vendor/datasketches-cpp/theta/include/theta_jaccard_similarity_base.hpp +1 -0
- data/vendor/datasketches-cpp/theta/include/theta_set_difference_base_impl.hpp +1 -1
- data/vendor/datasketches-cpp/theta/include/theta_sketch.hpp +180 -33
- data/vendor/datasketches-cpp/theta/include/theta_sketch_impl.hpp +430 -130
- data/vendor/datasketches-cpp/theta/include/theta_union.hpp +17 -10
- data/vendor/datasketches-cpp/theta/include/theta_union_base_impl.hpp +10 -10
- data/vendor/datasketches-cpp/theta/include/theta_union_impl.hpp +3 -3
- data/vendor/datasketches-cpp/theta/include/theta_update_sketch_base.hpp +21 -6
- data/vendor/datasketches-cpp/theta/include/theta_update_sketch_base_impl.hpp +13 -3
- data/vendor/datasketches-cpp/theta/test/CMakeLists.txt +15 -1
- data/vendor/datasketches-cpp/theta/test/bit_packing_test.cpp +80 -0
- data/vendor/datasketches-cpp/theta/test/theta_sketch_deserialize_from_java_test.cpp +57 -0
- data/vendor/datasketches-cpp/theta/test/theta_sketch_serialize_for_java.cpp +61 -0
- data/vendor/datasketches-cpp/theta/test/theta_sketch_test.cpp +39 -188
- data/vendor/datasketches-cpp/theta/test/theta_union_test.cpp +25 -0
- data/vendor/datasketches-cpp/tuple/CMakeLists.txt +8 -7
- data/vendor/datasketches-cpp/tuple/include/array_of_doubles_sketch.hpp +19 -144
- data/vendor/datasketches-cpp/tuple/include/{array_of_doubles_a_not_b.hpp → array_tuple_a_not_b.hpp} +24 -16
- data/vendor/datasketches-cpp/tuple/include/{array_of_doubles_a_not_b_impl.hpp → array_tuple_a_not_b_impl.hpp} +4 -4
- data/vendor/datasketches-cpp/tuple/include/array_tuple_intersection.hpp +65 -0
- data/vendor/datasketches-cpp/tuple/include/array_tuple_intersection_impl.hpp +31 -0
- data/vendor/datasketches-cpp/tuple/include/array_tuple_sketch.hpp +237 -0
- data/vendor/datasketches-cpp/tuple/include/{array_of_doubles_sketch_impl.hpp → array_tuple_sketch_impl.hpp} +40 -41
- data/vendor/datasketches-cpp/tuple/include/array_tuple_union.hpp +81 -0
- data/vendor/datasketches-cpp/tuple/include/array_tuple_union_impl.hpp +43 -0
- data/vendor/datasketches-cpp/tuple/include/tuple_a_not_b.hpp +11 -2
- data/vendor/datasketches-cpp/tuple/include/tuple_intersection.hpp +17 -10
- data/vendor/datasketches-cpp/tuple/include/tuple_jaccard_similarity.hpp +2 -1
- data/vendor/datasketches-cpp/tuple/include/tuple_sketch.hpp +95 -32
- data/vendor/datasketches-cpp/tuple/include/tuple_sketch_impl.hpp +2 -1
- data/vendor/datasketches-cpp/tuple/include/tuple_union.hpp +19 -11
- data/vendor/datasketches-cpp/tuple/test/CMakeLists.txt +16 -1
- data/vendor/datasketches-cpp/tuple/test/aod_sketch_deserialize_from_java_test.cpp +76 -0
- data/vendor/datasketches-cpp/tuple/test/aod_sketch_serialize_for_java.cpp +62 -0
- data/vendor/datasketches-cpp/tuple/test/array_of_doubles_sketch_test.cpp +5 -129
- data/vendor/datasketches-cpp/tuple/test/engagement_test.cpp +85 -89
- data/vendor/datasketches-cpp/tuple/test/tuple_jaccard_similarity_test.cpp +3 -1
- data/vendor/datasketches-cpp/tuple/test/tuple_sketch_deserialize_from_java_test.cpp +47 -0
- data/vendor/datasketches-cpp/{python/src/datasketches.cpp → tuple/test/tuple_sketch_serialize_for_java.cpp} +16 -30
- data/vendor/datasketches-cpp/tuple/test/tuple_sketch_test.cpp +1 -1
- data/vendor/datasketches-cpp/version.cfg.in +1 -1
- metadata +61 -79
- data/vendor/datasketches-cpp/fi/test/items_sketch_string_from_java.sk +0 -0
- data/vendor/datasketches-cpp/fi/test/items_sketch_string_utf8_from_java.sk +0 -0
- data/vendor/datasketches-cpp/fi/test/longs_sketch_from_java.sk +0 -0
- data/vendor/datasketches-cpp/hll/test/array6_from_java.sk +0 -0
- data/vendor/datasketches-cpp/hll/test/compact_array4_from_java.sk +0 -0
- data/vendor/datasketches-cpp/hll/test/compact_set_from_java.sk +0 -0
- data/vendor/datasketches-cpp/hll/test/list_from_java.sk +0 -0
- data/vendor/datasketches-cpp/hll/test/updatable_array4_from_java.sk +0 -0
- data/vendor/datasketches-cpp/hll/test/updatable_set_from_java.sk +0 -0
- data/vendor/datasketches-cpp/kll/test/kll_sketch_from_java.sk +0 -0
- data/vendor/datasketches-cpp/pyproject.toml +0 -23
- data/vendor/datasketches-cpp/python/CMakeLists.txt +0 -81
- data/vendor/datasketches-cpp/python/README.md +0 -85
- data/vendor/datasketches-cpp/python/datasketches/PySerDe.py +0 -104
- data/vendor/datasketches-cpp/python/include/py_serde.hpp +0 -113
- data/vendor/datasketches-cpp/python/jupyter/CPCSketch.ipynb +0 -345
- data/vendor/datasketches-cpp/python/jupyter/FrequentItemsSketch.ipynb +0 -354
- data/vendor/datasketches-cpp/python/jupyter/HLLSketch.ipynb +0 -346
- data/vendor/datasketches-cpp/python/jupyter/KLLSketch.ipynb +0 -463
- data/vendor/datasketches-cpp/python/jupyter/ThetaSketchNotebook.ipynb +0 -403
- data/vendor/datasketches-cpp/python/pybind11Path.cmd +0 -21
- data/vendor/datasketches-cpp/python/src/cpc_wrapper.cpp +0 -90
- data/vendor/datasketches-cpp/python/src/fi_wrapper.cpp +0 -128
- data/vendor/datasketches-cpp/python/src/hll_wrapper.cpp +0 -134
- data/vendor/datasketches-cpp/python/src/kll_wrapper.cpp +0 -210
- data/vendor/datasketches-cpp/python/src/ks_wrapper.cpp +0 -68
- data/vendor/datasketches-cpp/python/src/py_serde.cpp +0 -111
- data/vendor/datasketches-cpp/python/src/quantiles_wrapper.cpp +0 -204
- data/vendor/datasketches-cpp/python/src/req_wrapper.cpp +0 -215
- data/vendor/datasketches-cpp/python/src/theta_wrapper.cpp +0 -172
- data/vendor/datasketches-cpp/python/src/vector_of_kll.cpp +0 -490
- data/vendor/datasketches-cpp/python/src/vo_wrapper.cpp +0 -173
- data/vendor/datasketches-cpp/python/tests/__init__.py +0 -16
- data/vendor/datasketches-cpp/python/tests/cpc_test.py +0 -64
- data/vendor/datasketches-cpp/python/tests/fi_test.py +0 -110
- data/vendor/datasketches-cpp/python/tests/hll_test.py +0 -130
- data/vendor/datasketches-cpp/python/tests/kll_test.py +0 -125
- data/vendor/datasketches-cpp/python/tests/quantiles_test.py +0 -126
- data/vendor/datasketches-cpp/python/tests/req_test.py +0 -126
- data/vendor/datasketches-cpp/python/tests/theta_test.py +0 -146
- data/vendor/datasketches-cpp/python/tests/vector_of_kll_test.py +0 -148
- data/vendor/datasketches-cpp/python/tests/vo_test.py +0 -125
- data/vendor/datasketches-cpp/req/test/req_float_empty_from_java.sk +0 -0
- data/vendor/datasketches-cpp/req/test/req_float_estimation_from_java.sk +0 -0
- data/vendor/datasketches-cpp/req/test/req_float_exact_from_java.sk +0 -0
- data/vendor/datasketches-cpp/req/test/req_float_raw_items_from_java.sk +0 -0
- data/vendor/datasketches-cpp/req/test/req_float_single_item_from_java.sk +0 -0
- data/vendor/datasketches-cpp/sampling/test/binaries_from_java.txt +0 -67
- data/vendor/datasketches-cpp/sampling/test/varopt_sketch_long_sampling.sk +0 -0
- data/vendor/datasketches-cpp/sampling/test/varopt_sketch_string_exact.sk +0 -0
- data/vendor/datasketches-cpp/sampling/test/varopt_union_double_sampling.sk +0 -0
- data/vendor/datasketches-cpp/setup.py +0 -110
- data/vendor/datasketches-cpp/theta/test/theta_compact_empty_from_java.sk +0 -0
- data/vendor/datasketches-cpp/theta/test/theta_compact_estimation_from_java.sk +0 -0
- data/vendor/datasketches-cpp/theta/test/theta_compact_exact_from_java.sk +0 -0
- data/vendor/datasketches-cpp/theta/test/theta_compact_single_item_from_java.sk +0 -0
- data/vendor/datasketches-cpp/tuple/include/array_of_doubles_intersection.hpp +0 -52
- data/vendor/datasketches-cpp/tuple/include/array_of_doubles_union.hpp +0 -81
- data/vendor/datasketches-cpp/tuple/include/array_of_doubles_union_impl.hpp +0 -43
- data/vendor/datasketches-cpp/tuple/test/aod_1_compact_empty_from_java.sk +0 -1
- data/vendor/datasketches-cpp/tuple/test/aod_1_compact_estimation_from_java.sk +0 -0
- data/vendor/datasketches-cpp/tuple/test/aod_1_compact_non_empty_no_entries_from_java.sk +0 -0
- data/vendor/datasketches-cpp/tuple/test/aod_2_compact_exact_from_java.sk +0 -0
- data/vendor/datasketches-cpp/tuple/test/aod_3_compact_empty_from_java.sk +0 -1
@@ -33,58 +33,58 @@
|
|
33
33
|
|
34
34
|
namespace datasketches {
|
35
35
|
|
36
|
-
|
37
|
-
* High performance C++ implementation of Compressed Probabilistic Counting (CPC) Sketch
|
38
|
-
*
|
39
|
-
* This is a very compact (in serialized form) distinct counting sketch.
|
40
|
-
* The theory is described in the following paper:
|
41
|
-
* https://arxiv.org/abs/1708.06839
|
42
|
-
*
|
43
|
-
* author Kevin Lang
|
44
|
-
* author Alexander Saydakov
|
45
|
-
*/
|
46
|
-
|
47
|
-
// forward-declarations
|
36
|
+
// forward declarations
|
48
37
|
template<typename A> class cpc_sketch_alloc;
|
49
38
|
template<typename A> class cpc_union_alloc;
|
50
39
|
|
51
|
-
|
40
|
+
/// CPC sketch alias with default allocator
|
52
41
|
using cpc_sketch = cpc_sketch_alloc<std::allocator<uint8_t>>;
|
53
42
|
|
54
|
-
|
55
|
-
|
56
|
-
|
57
|
-
|
58
|
-
|
59
|
-
|
43
|
+
/**
|
44
|
+
* Allocation and initialization of global decompression (decoding) tables.
|
45
|
+
* Call this before anything else if you want to control the initialization time.
|
46
|
+
* For instance, to have this happen outside of a transaction context.
|
47
|
+
* Otherwise initialization happens on the first use (serialization or deserialization).
|
48
|
+
* It is safe to call more than once assuming no race conditions.
|
49
|
+
* This is not thread safe! Neither is the rest of the library.
|
50
|
+
*/
|
60
51
|
template<typename A> void cpc_init();
|
61
52
|
|
53
|
+
/**
|
54
|
+
* High performance C++ implementation of Compressed Probabilistic Counting (CPC) Sketch
|
55
|
+
*
|
56
|
+
* This is a very compact (in serialized form) distinct counting sketch.
|
57
|
+
* The theory is described in the following paper:
|
58
|
+
* https://arxiv.org/abs/1708.06839
|
59
|
+
*
|
60
|
+
* @author Kevin Lang
|
61
|
+
* @author Alexander Saydakov
|
62
|
+
*/
|
62
63
|
template<typename A>
|
63
64
|
class cpc_sketch_alloc {
|
64
65
|
public:
|
66
|
+
using allocator_type = A;
|
67
|
+
using vector_bytes = std::vector<uint8_t, typename std::allocator_traits<A>::template rebind_alloc<uint8_t>>;
|
68
|
+
using vector_u64 = std::vector<uint64_t, typename std::allocator_traits<A>::template rebind_alloc<uint64_t>>;
|
69
|
+
|
65
70
|
/**
|
66
71
|
* Creates an instance of the sketch given the lg_k parameter and hash seed.
|
67
72
|
* @param lg_k base 2 logarithm of the number of bins in the sketch
|
68
73
|
* @param seed for hash function
|
74
|
+
* @param allocator instance of an allocator
|
69
75
|
*/
|
70
76
|
explicit cpc_sketch_alloc(uint8_t lg_k = cpc_constants::DEFAULT_LG_K, uint64_t seed = DEFAULT_SEED, const A& allocator = A());
|
71
77
|
|
72
|
-
|
78
|
+
/// @return allocator
|
73
79
|
A get_allocator() const;
|
74
80
|
|
75
|
-
|
76
|
-
* @return configured lg_k of this sketch
|
77
|
-
*/
|
81
|
+
/// @return configured lg_k of this sketch
|
78
82
|
uint8_t get_lg_k() const;
|
79
83
|
|
80
|
-
|
81
|
-
* @return true if this sketch represents an empty set
|
82
|
-
*/
|
84
|
+
/// @return true if this sketch represents an empty set
|
83
85
|
bool is_empty() const;
|
84
86
|
|
85
|
-
|
86
|
-
* @return estimate of the distinct count of the input stream
|
87
|
-
*/
|
87
|
+
/// @return estimate of the distinct count of the input stream
|
88
88
|
double get_estimate() const;
|
89
89
|
|
90
90
|
/**
|
@@ -189,13 +189,14 @@ public:
|
|
189
189
|
* Otherwise two sketches that should represent overlapping sets will be disjoint
|
190
190
|
* For instance, for signed 32-bit values call update(int32_t) method above,
|
191
191
|
* which does widening conversion to int64_t, if compatibility with Java is expected
|
192
|
-
* @param
|
193
|
-
* @param
|
192
|
+
* @param value pointer to the data
|
193
|
+
* @param size of the data in bytes
|
194
194
|
*/
|
195
195
|
void update(const void* value, size_t size);
|
196
196
|
|
197
197
|
/**
|
198
198
|
* Returns a human-readable summary of this sketch
|
199
|
+
* @return a human-readable summary of this sketch
|
199
200
|
*/
|
200
201
|
string<A> to_string() const;
|
201
202
|
|
@@ -205,16 +206,13 @@ public:
|
|
205
206
|
*/
|
206
207
|
void serialize(std::ostream& os) const;
|
207
208
|
|
208
|
-
// This is a convenience alias for users
|
209
|
-
// The type returned by the following serialize method
|
210
|
-
using vector_bytes = vector_u8<A>;
|
211
|
-
|
212
209
|
/**
|
213
210
|
* This method serializes the sketch as a vector of bytes.
|
214
211
|
* An optional header can be reserved in front of the sketch.
|
215
212
|
* It is an uninitialized space of a given size.
|
216
213
|
* This header is used in Datasketches PostgreSQL extension.
|
217
214
|
* @param header_size_bytes space to reserve in front of the sketch
|
215
|
+
* @return serialized sketch as a vector of bytes
|
218
216
|
*/
|
219
217
|
vector_bytes serialize(unsigned header_size_bytes = 0) const;
|
220
218
|
|
@@ -222,6 +220,7 @@ public:
|
|
222
220
|
* This method deserializes a sketch from a given stream.
|
223
221
|
* @param is input stream
|
224
222
|
* @param seed the seed for the hash function that was used to create the sketch
|
223
|
+
* @param allocator instance of an Allocator
|
225
224
|
* @return an instance of a sketch
|
226
225
|
*/
|
227
226
|
static cpc_sketch_alloc<A> deserialize(std::istream& is, uint64_t seed = DEFAULT_SEED, const A& allocator = A());
|
@@ -231,6 +230,7 @@ public:
|
|
231
230
|
* @param bytes pointer to the array of bytes
|
232
231
|
* @param size the size of the array
|
233
232
|
* @param seed the seed for the hash function that was used to create the sketch
|
233
|
+
* @param allocator instance of an Allocator
|
234
234
|
* @return an instance of the sketch
|
235
235
|
*/
|
236
236
|
static cpc_sketch_alloc<A> deserialize(const void* bytes, size_t size, uint64_t seed = DEFAULT_SEED, const A& allocator = A());
|
@@ -246,10 +246,10 @@ public:
|
|
246
246
|
*/
|
247
247
|
static size_t get_max_serialized_size_bytes(uint8_t lg_k);
|
248
248
|
|
249
|
-
|
249
|
+
/// @private for internal use
|
250
250
|
uint32_t get_num_coupons() const;
|
251
251
|
|
252
|
-
|
252
|
+
/// @private for debugging
|
253
253
|
// this should catch some forms of corruption during serialization-deserialization
|
254
254
|
bool validate() const;
|
255
255
|
|
@@ -276,7 +276,7 @@ private:
|
|
276
276
|
uint32_t num_coupons; // the number of coupons collected so far
|
277
277
|
|
278
278
|
u32_table<A> surprising_value_table;
|
279
|
-
|
279
|
+
vector_bytes sliding_window;
|
280
280
|
uint8_t window_offset; // derivable from num_coupons, but made explicit for speed
|
281
281
|
uint8_t first_interesting_column; // This is part of a speed optimization
|
282
282
|
|
@@ -285,7 +285,7 @@ private:
|
|
285
285
|
|
286
286
|
// for deserialization and cpc_union::get_result()
|
287
287
|
cpc_sketch_alloc(uint8_t lg_k, uint32_t num_coupons, uint8_t first_interesting_column, u32_table<A>&& table,
|
288
|
-
|
288
|
+
vector_bytes&& window, bool has_hip, double kxp, double hip_est_accum, uint64_t seed);
|
289
289
|
|
290
290
|
inline void row_col_update(uint32_t row_col);
|
291
291
|
inline void update_sparse(uint32_t row_col);
|
@@ -308,7 +308,7 @@ private:
|
|
308
308
|
static inline uint8_t determine_correct_offset(uint8_t lg_k, uint64_t c);
|
309
309
|
|
310
310
|
// this produces a full-size k-by-64 bit matrix
|
311
|
-
vector_u64
|
311
|
+
vector_u64 build_bit_matrix() const;
|
312
312
|
|
313
313
|
static uint8_t get_preamble_ints(uint32_t num_coupons, bool has_hip, bool has_table, bool has_window);
|
314
314
|
inline void write_hip(std::ostream& os) const;
|
@@ -315,7 +315,7 @@ void cpc_sketch_alloc<A>::move_window() {
|
|
315
315
|
const uint32_t k = 1 << lg_k;
|
316
316
|
|
317
317
|
// Construct the full-sized bit matrix that corresponds to the sketch
|
318
|
-
vector_u64
|
318
|
+
vector_u64 bit_matrix = build_bit_matrix();
|
319
319
|
|
320
320
|
// refresh the KXP register on every 8th window shift.
|
321
321
|
if ((new_offset & 0x7) == 0) refresh_kxp(bit_matrix.data());
|
@@ -359,7 +359,7 @@ void cpc_sketch_alloc<A>::refresh_kxp(const uint64_t* bit_matrix) {
|
|
359
359
|
|
360
360
|
// for improved numerical accuracy, we separately sum the bytes of the U64's
|
361
361
|
double byte_sums[8]; // allocating on the stack
|
362
|
-
std::fill(byte_sums,
|
362
|
+
std::fill(byte_sums, byte_sums + 8, 0);
|
363
363
|
|
364
364
|
for (size_t i = 0; i < k; i++) {
|
365
365
|
uint64_t word = bit_matrix[i];
|
@@ -458,7 +458,7 @@ void cpc_sketch_alloc<A>::serialize(std::ostream& os) const {
|
|
458
458
|
}
|
459
459
|
|
460
460
|
template<typename A>
|
461
|
-
|
461
|
+
auto cpc_sketch_alloc<A>::serialize(unsigned header_size_bytes) const -> vector_bytes {
|
462
462
|
compressed_state<A> compressed(sliding_window.get_allocator());
|
463
463
|
compressed.table_data_words = 0;
|
464
464
|
compressed.table_num_entries = 0;
|
@@ -469,7 +469,7 @@ vector_u8<A> cpc_sketch_alloc<A>::serialize(unsigned header_size_bytes) const {
|
|
469
469
|
const bool has_window = compressed.window_data.size() > 0;
|
470
470
|
const uint8_t preamble_ints = get_preamble_ints(num_coupons, has_hip, has_table, has_window);
|
471
471
|
const size_t size = header_size_bytes + (preamble_ints + compressed.table_data_words + compressed.window_data_words) * sizeof(uint32_t);
|
472
|
-
|
472
|
+
vector_bytes bytes(size, 0, sliding_window.get_allocator());
|
473
473
|
uint8_t* ptr = bytes.data() + header_size_bytes;
|
474
474
|
ptr += copy_to_mem(preamble_ints, ptr);
|
475
475
|
const uint8_t serial_version = SERIAL_VERSION;
|
@@ -712,15 +712,18 @@ static const size_t CPC_MAX_PREAMBLE_SIZE_BYTES = 40;
|
|
712
712
|
template<typename A>
|
713
713
|
size_t cpc_sketch_alloc<A>::get_max_serialized_size_bytes(uint8_t lg_k) {
|
714
714
|
check_lg_k(lg_k);
|
715
|
-
if (lg_k <= CPC_EMPIRICAL_SIZE_MAX_LGK)
|
715
|
+
if (lg_k <= CPC_EMPIRICAL_SIZE_MAX_LGK) {
|
716
|
+
return CPC_EMPIRICAL_MAX_SIZE_BYTES[lg_k - cpc_constants::MIN_LG_K] + CPC_MAX_PREAMBLE_SIZE_BYTES;
|
717
|
+
}
|
716
718
|
const uint32_t k = 1 << lg_k;
|
717
719
|
return (int) (CPC_EMPIRICAL_MAX_SIZE_FACTOR * k) + CPC_MAX_PREAMBLE_SIZE_BYTES;
|
718
720
|
}
|
719
721
|
|
720
722
|
template<typename A>
|
721
723
|
void cpc_sketch_alloc<A>::check_lg_k(uint8_t lg_k) {
|
722
|
-
if (lg_k <
|
723
|
-
throw std::invalid_argument("lg_k must be >= " + std::to_string(
|
724
|
+
if (lg_k < cpc_constants::MIN_LG_K || lg_k > cpc_constants::MAX_LG_K) {
|
725
|
+
throw std::invalid_argument("lg_k must be >= " + std::to_string(cpc_constants::MIN_LG_K) + " and <= "
|
726
|
+
+ std::to_string(cpc_constants::MAX_LG_K) + ": " + std::to_string(lg_k));
|
724
727
|
}
|
725
728
|
}
|
726
729
|
|
@@ -731,14 +734,14 @@ uint32_t cpc_sketch_alloc<A>::get_num_coupons() const {
|
|
731
734
|
|
732
735
|
template<typename A>
|
733
736
|
bool cpc_sketch_alloc<A>::validate() const {
|
734
|
-
vector_u64
|
737
|
+
vector_u64 bit_matrix = build_bit_matrix();
|
735
738
|
const uint64_t num_bits_set = count_bits_set_in_matrix(bit_matrix.data(), 1ULL << lg_k);
|
736
739
|
return num_bits_set == num_coupons;
|
737
740
|
}
|
738
741
|
|
739
742
|
template<typename A>
|
740
743
|
cpc_sketch_alloc<A>::cpc_sketch_alloc(uint8_t lg_k, uint32_t num_coupons, uint8_t first_interesting_column,
|
741
|
-
u32_table<A>&& table,
|
744
|
+
u32_table<A>&& table, vector_bytes&& window, bool has_hip, double kxp, double hip_est_accum, uint64_t seed):
|
742
745
|
lg_k(lg_k),
|
743
746
|
seed(seed),
|
744
747
|
was_merged(!has_hip),
|
@@ -800,14 +803,14 @@ uint8_t cpc_sketch_alloc<A>::determine_correct_offset(uint8_t lg_k, uint64_t c)
|
|
800
803
|
}
|
801
804
|
|
802
805
|
template<typename A>
|
803
|
-
|
806
|
+
auto cpc_sketch_alloc<A>::build_bit_matrix() const -> vector_u64 {
|
804
807
|
const uint32_t k = 1 << lg_k;
|
805
808
|
if (window_offset > 56) throw std::logic_error("offset > 56");
|
806
809
|
|
807
810
|
// Fill the matrix with default rows in which the "early zone" is filled with ones.
|
808
811
|
// This is essential for the routine's O(k) time cost (as opposed to O(C)).
|
809
812
|
const uint64_t default_row = (static_cast<uint64_t>(1) << window_offset) - 1;
|
810
|
-
vector_u64
|
813
|
+
vector_u64 matrix(k, default_row, sliding_window.get_allocator());
|
811
814
|
|
812
815
|
if (num_coupons == 0) return matrix;
|
813
816
|
|
@@ -27,31 +27,55 @@
|
|
27
27
|
|
28
28
|
namespace datasketches {
|
29
29
|
|
30
|
-
|
30
|
+
/// CPC union alias with default allocator
|
31
|
+
using cpc_union = cpc_union_alloc<std::allocator<uint8_t>>;
|
32
|
+
|
33
|
+
/**
|
31
34
|
* High performance C++ implementation of Compressed Probabilistic Counting (CPC) Union
|
32
35
|
*
|
33
36
|
* author Kevin Lang
|
34
37
|
* author Alexander Saydakov
|
35
38
|
*/
|
36
|
-
|
37
|
-
// alias with default allocator for convenience
|
38
|
-
using cpc_union = cpc_union_alloc<std::allocator<uint8_t>>;
|
39
|
-
|
40
39
|
template<typename A>
|
41
40
|
class cpc_union_alloc {
|
42
41
|
public:
|
42
|
+
using vector_bytes = std::vector<uint8_t, typename std::allocator_traits<A>::template rebind_alloc<uint8_t>>;
|
43
|
+
using vector_u64 = std::vector<uint64_t, typename std::allocator_traits<A>::template rebind_alloc<uint64_t>>;
|
44
|
+
|
43
45
|
/**
|
44
46
|
* Creates an instance of the union given the lg_k parameter and hash seed.
|
45
47
|
* @param lg_k base 2 logarithm of the number of bins in the sketch
|
46
48
|
* @param seed for hash function
|
49
|
+
* @param allocator instance of an allocator
|
47
50
|
*/
|
48
51
|
explicit cpc_union_alloc(uint8_t lg_k = cpc_constants::DEFAULT_LG_K, uint64_t seed = DEFAULT_SEED, const A& allocator = A());
|
49
52
|
|
53
|
+
/**
|
54
|
+
* Copy constructor
|
55
|
+
* @param other union to be copied
|
56
|
+
*/
|
50
57
|
cpc_union_alloc(const cpc_union_alloc<A>& other);
|
58
|
+
|
59
|
+
/**
|
60
|
+
* Move constructor
|
61
|
+
* @param other union to be moved
|
62
|
+
*/
|
51
63
|
cpc_union_alloc(cpc_union_alloc<A>&& other) noexcept;
|
64
|
+
|
52
65
|
~cpc_union_alloc();
|
53
66
|
|
67
|
+
/**
|
68
|
+
* Copy assignment
|
69
|
+
* @param other union to be copied
|
70
|
+
* @return reference to this union
|
71
|
+
*/
|
54
72
|
cpc_union_alloc<A>& operator=(const cpc_union_alloc<A>& other);
|
73
|
+
|
74
|
+
/**
|
75
|
+
* Move assignment
|
76
|
+
* @param other union to be moved
|
77
|
+
* @return reference to this union
|
78
|
+
*/
|
55
79
|
cpc_union_alloc<A>& operator=(cpc_union_alloc<A>&& other) noexcept;
|
56
80
|
|
57
81
|
/**
|
@@ -73,14 +97,14 @@ public:
|
|
73
97
|
cpc_sketch_alloc<A> get_result() const;
|
74
98
|
|
75
99
|
private:
|
76
|
-
|
77
|
-
|
78
|
-
|
100
|
+
using AllocU8 = typename std::allocator_traits<A>::template rebind_alloc<uint8_t>;
|
101
|
+
using AllocU64 = typename std::allocator_traits<A>::template rebind_alloc<uint64_t>;
|
102
|
+
using AllocCpc = typename std::allocator_traits<A>::template rebind_alloc<cpc_sketch_alloc<A>>;
|
79
103
|
|
80
104
|
uint8_t lg_k;
|
81
105
|
uint64_t seed;
|
82
106
|
cpc_sketch_alloc<A>* accumulator;
|
83
|
-
vector_u64
|
107
|
+
vector_u64 bit_matrix;
|
84
108
|
|
85
109
|
template<typename S> void internal_update(S&& sketch); // to support both rvalue and lvalue
|
86
110
|
|
@@ -90,8 +114,8 @@ private:
|
|
90
114
|
void switch_to_bit_matrix();
|
91
115
|
void walk_table_updating_sketch(const u32_table<A>& table);
|
92
116
|
void or_table_into_matrix(const u32_table<A>& table);
|
93
|
-
void or_window_into_matrix(const
|
94
|
-
void or_matrix_into_matrix(const vector_u64
|
117
|
+
void or_window_into_matrix(const vector_bytes& sliding_window, uint8_t offset, uint8_t src_lg_k);
|
118
|
+
void or_matrix_into_matrix(const vector_u64& src_matrix, uint8_t src_lg_k);
|
95
119
|
void reduce_k(uint8_t new_lg_k);
|
96
120
|
};
|
97
121
|
|
@@ -33,8 +33,8 @@ seed(seed),
|
|
33
33
|
accumulator(nullptr),
|
34
34
|
bit_matrix(allocator)
|
35
35
|
{
|
36
|
-
if (lg_k <
|
37
|
-
throw std::invalid_argument("lg_k must be >= " + std::to_string(
|
36
|
+
if (lg_k < cpc_constants::MIN_LG_K || lg_k > cpc_constants::MAX_LG_K) {
|
37
|
+
throw std::invalid_argument("lg_k must be >= " + std::to_string(cpc_constants::MIN_LG_K) + " and <= " + std::to_string(cpc_constants::MAX_LG_K) + ": " + std::to_string(lg_k));
|
38
38
|
}
|
39
39
|
accumulator = new (AllocCpc(allocator).allocate(1)) cpc_sketch_alloc<A>(lg_k, seed, allocator);
|
40
40
|
}
|
@@ -166,7 +166,7 @@ void cpc_union_alloc<A>::internal_update(S&& sketch) {
|
|
166
166
|
// SLIDING mode involves inverted logic, so we can't just walk the source sketch.
|
167
167
|
// Instead, we convert it to a bitMatrix that can be OR'ed into the destination.
|
168
168
|
if (cpc_sketch_alloc<A>::flavor::SLIDING != src_flavor) throw std::logic_error("wrong flavor"); // Case D
|
169
|
-
vector_u64
|
169
|
+
vector_u64 src_matrix = sketch.build_bit_matrix();
|
170
170
|
or_matrix_into_matrix(src_matrix, sketch.get_lg_k());
|
171
171
|
}
|
172
172
|
|
@@ -203,7 +203,7 @@ cpc_sketch_alloc<A> cpc_union_alloc<A>::get_result_from_bit_matrix() const {
|
|
203
203
|
|
204
204
|
const uint8_t offset = cpc_sketch_alloc<A>::determine_correct_offset(lg_k, num_coupons);
|
205
205
|
|
206
|
-
|
206
|
+
vector_bytes sliding_window(k, 0, bit_matrix.get_allocator());
|
207
207
|
// don't need to zero the window's memory
|
208
208
|
|
209
209
|
// dynamically growing caused snowplow effect
|
@@ -289,7 +289,7 @@ void cpc_union_alloc<A>::or_table_into_matrix(const u32_table<A>& table) {
|
|
289
289
|
}
|
290
290
|
|
291
291
|
template<typename A>
|
292
|
-
void cpc_union_alloc<A>::or_window_into_matrix(const
|
292
|
+
void cpc_union_alloc<A>::or_window_into_matrix(const vector_bytes& sliding_window, uint8_t offset, uint8_t src_lg_k) {
|
293
293
|
if (lg_k > src_lg_k) throw std::logic_error("dst LgK > src LgK");
|
294
294
|
const uint64_t dst_mask = (1 << lg_k) - 1; // downsamples when dst lgK < src LgK
|
295
295
|
const uint32_t src_k = 1 << src_lg_k;
|
@@ -299,7 +299,7 @@ void cpc_union_alloc<A>::or_window_into_matrix(const vector_u8<A>& sliding_windo
|
|
299
299
|
}
|
300
300
|
|
301
301
|
template<typename A>
|
302
|
-
void cpc_union_alloc<A>::or_matrix_into_matrix(const vector_u64
|
302
|
+
void cpc_union_alloc<A>::or_matrix_into_matrix(const vector_u64& src_matrix, uint8_t src_lg_k) {
|
303
303
|
if (lg_k > src_lg_k) throw std::logic_error("dst LgK > src LgK");
|
304
304
|
const uint64_t dst_mask = (1 << lg_k) - 1; // downsamples when dst lgK < src LgK
|
305
305
|
const uint32_t src_k = 1 << src_lg_k;
|
@@ -315,10 +315,10 @@ void cpc_union_alloc<A>::reduce_k(uint8_t new_lg_k) {
|
|
315
315
|
|
316
316
|
if (bit_matrix.size() > 0) { // downsample the unioner's bit matrix
|
317
317
|
if (accumulator != nullptr) throw std::logic_error("accumulator is not null");
|
318
|
-
vector_u64
|
318
|
+
vector_u64 old_matrix = std::move(bit_matrix);
|
319
319
|
const uint8_t old_lg_k = lg_k;
|
320
320
|
const uint32_t new_k = 1 << new_lg_k;
|
321
|
-
bit_matrix = vector_u64
|
321
|
+
bit_matrix = vector_u64(new_k, 0, old_matrix.get_allocator());
|
322
322
|
lg_k = new_lg_k;
|
323
323
|
or_matrix_into_matrix(old_matrix, old_lg_k);
|
324
324
|
return;
|
@@ -89,7 +89,13 @@ static inline uint32_t warren_count_bits_set_in_matrix(const uint64_t* array, ui
|
|
89
89
|
|
90
90
|
// This code is Figure 5-9 in "Hacker's Delight" by Henry S. Warren.
|
91
91
|
|
92
|
-
#define
|
92
|
+
#define DATASKETCHES_CSA(h, l, a, b, c) \
|
93
|
+
{ \
|
94
|
+
uint64_t u = a ^ b; \
|
95
|
+
uint64_t v = c; \
|
96
|
+
h = (a & b) | (u & v); \
|
97
|
+
l = u ^ v; \
|
98
|
+
}
|
93
99
|
|
94
100
|
static inline uint32_t count_bits_set_in_matrix(const uint64_t* a, uint32_t length) {
|
95
101
|
if ((length & 0x7) != 0) throw std::invalid_argument("the length of the array must be a multiple of 8");
|
@@ -98,15 +104,15 @@ static inline uint32_t count_bits_set_in_matrix(const uint64_t* a, uint32_t leng
|
|
98
104
|
fours = twos = ones = 0;
|
99
105
|
|
100
106
|
for (uint32_t i = 0; i <= length - 8; i += 8) {
|
101
|
-
|
102
|
-
|
103
|
-
|
107
|
+
DATASKETCHES_CSA(twos_a, ones, ones, a[i+0], a[i+1]);
|
108
|
+
DATASKETCHES_CSA(twos_b, ones, ones, a[i+2], a[i+3]);
|
109
|
+
DATASKETCHES_CSA(fours_a, twos, twos, twos_a, twos_b);
|
104
110
|
|
105
|
-
|
106
|
-
|
107
|
-
|
111
|
+
DATASKETCHES_CSA(twos_a, ones, ones, a[i+4], a[i+5]);
|
112
|
+
DATASKETCHES_CSA(twos_b, ones, ones, a[i+6], a[i+7]);
|
113
|
+
DATASKETCHES_CSA(fours_b, twos, twos, twos_a, twos_b);
|
108
114
|
|
109
|
-
|
115
|
+
DATASKETCHES_CSA(eights, fours, fours, fours_a, fours_b);
|
110
116
|
|
111
117
|
total += warren_bit_count(eights);
|
112
118
|
}
|
@@ -119,6 +125,8 @@ static inline uint32_t count_bits_set_in_matrix(const uint64_t* a, uint32_t leng
|
|
119
125
|
return total;
|
120
126
|
}
|
121
127
|
|
128
|
+
#undef DATASKETCHES_CSA
|
129
|
+
|
122
130
|
// Here are some timings made with quickTestMerge.c
|
123
131
|
// for the "5 5" case:
|
124
132
|
|
@@ -38,6 +38,7 @@ static const uint32_t U32_TABLE_DOWNSIZE_DENOM = 4LL;
|
|
38
38
|
template<typename A>
|
39
39
|
class u32_table {
|
40
40
|
public:
|
41
|
+
using vector_u32 = std::vector<uint32_t, typename std::allocator_traits<A>::template rebind_alloc<uint32_t>>;
|
41
42
|
|
42
43
|
u32_table(const A& allocator);
|
43
44
|
u32_table(uint8_t lg_size, uint8_t num_valid_bits, const A& allocator);
|
@@ -54,7 +55,7 @@ public:
|
|
54
55
|
|
55
56
|
static u32_table make_from_pairs(const uint32_t* pairs, uint32_t num_pairs, uint8_t lg_k, const A& allocator);
|
56
57
|
|
57
|
-
vector_u32
|
58
|
+
vector_u32 unwrapping_get_items() const;
|
58
59
|
|
59
60
|
static void merge(
|
60
61
|
const uint32_t* arr_a, size_t start_a, size_t length_a, // input
|
@@ -70,7 +71,7 @@ private:
|
|
70
71
|
uint8_t lg_size; // log2 of number of slots
|
71
72
|
uint8_t num_valid_bits;
|
72
73
|
uint32_t num_items;
|
73
|
-
vector_u32
|
74
|
+
vector_u32 slots;
|
74
75
|
|
75
76
|
inline uint32_t lookup(uint32_t item) const;
|
76
77
|
inline void must_insert(uint32_t item);
|
@@ -151,8 +151,8 @@ void u32_table<A>::rebuild(uint8_t new_lg_size) {
|
|
151
151
|
const uint32_t old_size = 1 << lg_size;
|
152
152
|
const uint32_t new_size = 1 << new_lg_size;
|
153
153
|
if (new_size <= num_items) throw std::logic_error("new_size <= num_items");
|
154
|
-
vector_u32
|
155
|
-
slots = vector_u32
|
154
|
+
vector_u32 old_slots = std::move(slots);
|
155
|
+
slots = vector_u32(new_size, UINT32_MAX, old_slots.get_allocator());
|
156
156
|
lg_size = new_lg_size;
|
157
157
|
for (uint32_t i = 0; i < old_size; i++) {
|
158
158
|
if (old_slots[i] != UINT32_MAX) {
|
@@ -168,10 +168,10 @@ void u32_table<A>::rebuild(uint8_t new_lg_size) {
|
|
168
168
|
// and even then the subsequent sort would fix things up.
|
169
169
|
// The result is nearly sorted, so make sure to use an efficient sort for that case
|
170
170
|
template<typename A>
|
171
|
-
|
172
|
-
if (num_items == 0) return vector_u32
|
171
|
+
auto u32_table<A>::unwrapping_get_items() const -> vector_u32 {
|
172
|
+
if (num_items == 0) return vector_u32(slots.get_allocator());
|
173
173
|
const uint32_t table_size = 1 << lg_size;
|
174
|
-
vector_u32
|
174
|
+
vector_u32 result(num_items, 0, slots.get_allocator());
|
175
175
|
size_t i = 0;
|
176
176
|
size_t l = 0;
|
177
177
|
size_t r = num_items - 1;
|
@@ -20,16 +20,15 @@ add_executable(cpc_test)
|
|
20
20
|
target_link_libraries(cpc_test cpc common_test_lib)
|
21
21
|
|
22
22
|
set_target_properties(cpc_test PROPERTIES
|
23
|
-
CXX_STANDARD 11
|
24
23
|
CXX_STANDARD_REQUIRED YES
|
25
24
|
)
|
26
25
|
|
27
|
-
|
28
|
-
|
29
|
-
|
30
|
-
|
31
|
-
|
32
|
-
|
26
|
+
file(TO_CMAKE_PATH "${CMAKE_CURRENT_SOURCE_DIR}" CPC_TEST_BINARY_PATH)
|
27
|
+
string(APPEND CPC_TEST_BINARY_PATH "/")
|
28
|
+
target_compile_definitions(cpc_test
|
29
|
+
PRIVATE
|
30
|
+
TEST_BINARY_INPUT_PATH="${CPC_TEST_BINARY_PATH}"
|
31
|
+
)
|
33
32
|
|
34
33
|
add_test(
|
35
34
|
NAME cpc_test
|
@@ -43,3 +42,17 @@ target_sources(cpc_test
|
|
43
42
|
compression_test.cpp
|
44
43
|
cpc_sketch_allocation_test.cpp
|
45
44
|
)
|
45
|
+
|
46
|
+
if (SERDE_COMPAT)
|
47
|
+
target_sources(cpc_test
|
48
|
+
PRIVATE
|
49
|
+
cpc_sketch_deserialize_from_java_test.cpp
|
50
|
+
)
|
51
|
+
endif()
|
52
|
+
|
53
|
+
if (GENERATE)
|
54
|
+
target_sources(cpc_test
|
55
|
+
PRIVATE
|
56
|
+
cpc_sketch_serialize_for_java.cpp
|
57
|
+
)
|
58
|
+
endif()
|
@@ -0,0 +1,60 @@
|
|
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
|
+
#include <catch2/catch.hpp>
|
21
|
+
#include <fstream>
|
22
|
+
#include <cpc_sketch.hpp>
|
23
|
+
|
24
|
+
namespace datasketches {
|
25
|
+
|
26
|
+
// assume the binary sketches for this test have been generated by datasketches-java code
|
27
|
+
// in the subdirectory called "java" in the root directory of this project
|
28
|
+
static std::string testBinaryInputPath = std::string(TEST_BINARY_INPUT_PATH) + "../../java/";
|
29
|
+
|
30
|
+
TEST_CASE("cpc sketch", "[serde_compat]") {
|
31
|
+
const unsigned n_arr[] = {0, 100, 200, 2000, 20000};
|
32
|
+
for (const unsigned n: n_arr) {
|
33
|
+
std::ifstream is;
|
34
|
+
is.exceptions(std::ios::failbit | std::ios::badbit);
|
35
|
+
is.open(testBinaryInputPath + "cpc_n" + std::to_string(n) + "_java.sk", std::ios::binary);
|
36
|
+
const auto sketch = cpc_sketch::deserialize(is);
|
37
|
+
REQUIRE(sketch.is_empty() == (n == 0));
|
38
|
+
REQUIRE(sketch.get_estimate() == Approx(n).margin(n * 0.02));
|
39
|
+
}
|
40
|
+
}
|
41
|
+
|
42
|
+
TEST_CASE("cpc sketch negative one", "[serde_compat]") {
|
43
|
+
std::ifstream is;
|
44
|
+
is.exceptions(std::ios::failbit | std::ios::badbit);
|
45
|
+
is.open(testBinaryInputPath + "cpc_negative_one_java.sk", std::ios::binary);
|
46
|
+
auto sketch = cpc_sketch::deserialize(is);
|
47
|
+
REQUIRE_FALSE(sketch.is_empty());
|
48
|
+
REQUIRE(sketch.get_estimate() == Approx(1).margin(0.01));
|
49
|
+
sketch.update((uint64_t) -1);
|
50
|
+
sketch.update((int64_t) -1);
|
51
|
+
sketch.update((uint32_t) -1);
|
52
|
+
sketch.update((int32_t) -1);
|
53
|
+
sketch.update((uint16_t) -1);
|
54
|
+
sketch.update((int16_t) -1);
|
55
|
+
sketch.update((uint8_t) -1);
|
56
|
+
sketch.update((int8_t) -1);
|
57
|
+
REQUIRE(sketch.get_estimate() == Approx(1).margin(0.01));
|
58
|
+
}
|
59
|
+
|
60
|
+
} /* namespace datasketches */
|
@@ -0,0 +1,38 @@
|
|
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
|
+
#include <catch2/catch.hpp>
|
21
|
+
#include <fstream>
|
22
|
+
#include <cpc_sketch.hpp>
|
23
|
+
|
24
|
+
namespace datasketches {
|
25
|
+
|
26
|
+
TEST_CASE("cpc sketch generate", "[serialize_for_java]") {
|
27
|
+
const unsigned n_arr[] = {0, 100, 200, 2000, 20000};
|
28
|
+
for (const unsigned n: n_arr) {
|
29
|
+
cpc_sketch sketch;
|
30
|
+
for (unsigned i = 1; i <= n; ++i) sketch.update(i);
|
31
|
+
REQUIRE(sketch.is_empty() == (n == 0));
|
32
|
+
REQUIRE(sketch.get_estimate() == Approx(n).margin(n * 0.02));
|
33
|
+
std::ofstream os("cpc_n" + std::to_string(n) + "_cpp.sk", std::ios::binary);
|
34
|
+
sketch.serialize(os);
|
35
|
+
}
|
36
|
+
}
|
37
|
+
|
38
|
+
} /* namespace datasketches */
|