OmniSciDB  06b3bd477c
 All Classes Namespaces Files Functions Variables Typedefs Enumerations Enumerator Friends Macros Pages
geo_compression.cpp
Go to the documentation of this file.
1 /*
2  * Copyright 2020 OmniSci, Inc.
3  *
4  * Licensed under the Apache License, Version 2.0 (the "License");
5  * you may not use this file except in compliance with the License.
6  * You may obtain a copy of the License at
7  *
8  * http://www.apache.org/licenses/LICENSE-2.0
9  *
10  * Unless required by applicable law or agreed to in writing, software
11  * distributed under the License is distributed on an "AS IS" BASIS,
12  * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13  * See the License for the specific language governing permissions and
14  * limitations under the License.
15  */
16 
17 #include "Shared/geo_compression.h"
19 #include "Shared/geo_types.h"
20 
21 namespace geospatial {
22 
23 int32_t get_compression_scheme(const SQLTypeInfo& ti) {
24  if (ti.get_compression() == kENCODING_GEOINT && ti.get_comp_param() == 32) {
25  return COMPRESSION_GEOINT32;
26  }
27  if (ti.get_compression() != kENCODING_NONE) {
28  throw std::runtime_error("Invalid compression");
29  }
30  return COMPRESSION_NONE;
31 }
32 
33 uint64_t compress_coord(double coord, const SQLTypeInfo& ti, bool x) {
34  if (ti.get_compression() == kENCODING_GEOINT && ti.get_comp_param() == 32) {
37  }
38  return *reinterpret_cast<uint64_t*>(may_alias_ptr(&coord));
39 }
40 
41 uint64_t compress_null_point(const SQLTypeInfo& ti, bool x) {
42  if (ti.get_compression() == kENCODING_GEOINT && ti.get_comp_param() == 32) {
45  }
46  double n = x ? NULL_ARRAY_DOUBLE : NULL_DOUBLE;
47  auto u = *reinterpret_cast<uint64_t*>(may_alias_ptr(&n));
48  return u;
49 }
50 
51 // Compress non-NULL geo coords; and also NULL POINT coords (special case)
52 std::vector<uint8_t> compress_coords(std::vector<double>& coords, const SQLTypeInfo& ti) {
53  CHECK(!coords.empty()) << "Coord compression received no data";
54  bool is_null_point = false;
55  if (!ti.get_notnull()) {
56  is_null_point = (ti.get_type() == kPOINT && coords[0] == NULL_ARRAY_DOUBLE);
57  }
58  std::vector<uint8_t> compressed_coords;
59  bool x = true;
60  bool is_geoint32 =
61  (ti.get_compression() == kENCODING_GEOINT && ti.get_comp_param() == 32);
62  size_t coord_data_size = (is_geoint32) ? (ti.get_comp_param() / 8) : sizeof(double);
63  for (auto coord : coords) {
64  uint64_t coord_data;
65  if (is_null_point) {
66  coord_data = compress_null_point(ti, x);
67  } else {
68  if (ti.get_output_srid() == 4326) {
69  if (x) {
70  if (coord < -180.0 || coord > 180.0) {
71  throw std::runtime_error("WGS84 longitude " + std::to_string(coord) +
72  " is out of bounds");
73  }
74  } else {
75  if (coord < -90.0 || coord > 90.0) {
76  throw std::runtime_error("WGS84 latitude " + std::to_string(coord) +
77  " is out of bounds");
78  }
79  }
80  }
81  if (is_geoint32) {
82  coord_data = compress_coord(coord, ti, x);
83  } else {
84  auto coord_data_ptr = reinterpret_cast<uint64_t*>(&coord);
85  coord_data = *coord_data_ptr;
86  }
87  }
88  for (size_t i = 0; i < coord_data_size; i++) {
89  compressed_coords.push_back(coord_data & 0xFF);
90  coord_data >>= 8;
91  }
92  x = !x;
93  }
94  return compressed_coords;
95 }
96 
97 template <typename T>
98 void unpack_geo_vector(std::vector<T>& output, const int8_t* input_ptr, const size_t sz) {
99  if (sz == 0) {
100  return;
101  }
102  auto elems = reinterpret_cast<const T*>(input_ptr);
103  CHECK_EQ(size_t(0), sz % sizeof(T));
104  const size_t num_elems = sz / sizeof(T);
105  output.resize(num_elems);
106  for (size_t i = 0; i < num_elems; i++) {
107  output[i] = elems[i];
108  }
109 }
110 
111 template <>
112 void unpack_geo_vector<int32_t>(std::vector<int32_t>& output,
113  const int8_t* input_ptr,
114  const size_t sz) {
115  if (sz == 0) {
116  return;
117  }
118  auto elems = reinterpret_cast<const int32_t*>(input_ptr);
119  CHECK_EQ(size_t(0), sz % sizeof(int32_t));
120  const size_t num_elems = sz / sizeof(int32_t);
121  output.resize(num_elems);
122  for (size_t i = 0; i < num_elems; i++) {
123  output[i] = elems[i];
124  }
125 }
126 
127 template <typename T>
128 void decompress_geo_coords_geoint32(std::vector<T>& dec,
129  const int8_t* enc,
130  const size_t sz) {
131  if (sz == 0) {
132  return;
133  }
134  const auto compressed_coords = reinterpret_cast<const int32_t*>(enc);
135  const auto num_coords = sz / sizeof(int32_t);
136  dec.resize(num_coords);
137  for (size_t i = 0; i < num_coords; i += 2) {
138  dec[i] = Geo_namespace::decompress_longitude_coord_geoint32(compressed_coords[i]);
139  dec[i + 1] =
140  Geo_namespace::decompress_lattitude_coord_geoint32(compressed_coords[i + 1]);
141  }
142 }
143 
144 template <>
145 std::shared_ptr<std::vector<double>> decompress_coords<double, SQLTypeInfo>(
146  const SQLTypeInfo& geo_ti,
147  const int8_t* coords,
148  const size_t coords_sz) {
149  auto decompressed_coords_ptr = std::make_shared<std::vector<double>>();
150  if (geo_ti.get_compression() == kENCODING_GEOINT) {
151  if (geo_ti.get_comp_param() == 32) {
152  decompress_geo_coords_geoint32(*decompressed_coords_ptr, coords, coords_sz);
153  }
154  } else {
155  CHECK_EQ(geo_ti.get_compression(), kENCODING_NONE);
156  unpack_geo_vector(*decompressed_coords_ptr, coords, coords_sz);
157  }
158  return decompressed_coords_ptr;
159 }
160 
161 template <>
162 std::shared_ptr<std::vector<double>> decompress_coords<double, int32_t>(
163  const int32_t& ic,
164  const int8_t* coords,
165  const size_t coords_sz) {
166  auto decompressed_coords_ptr = std::make_shared<std::vector<double>>();
167  if (ic == COMPRESSION_GEOINT32) {
168  decompress_geo_coords_geoint32(*decompressed_coords_ptr, coords, coords_sz);
169  } else {
171  unpack_geo_vector(*decompressed_coords_ptr, coords, coords_sz);
172  }
173  return decompressed_coords_ptr;
174 }
175 
176 bool is_null_point(const SQLTypeInfo& geo_ti,
177  const int8_t* coords,
178  const size_t coords_sz) {
179  if (geo_ti.get_type() == kPOINT && !geo_ti.get_notnull()) {
180  if (geo_ti.get_compression() == kENCODING_GEOINT) {
181  if (geo_ti.get_comp_param() == 32) {
182  return Geo_namespace::is_null_point_longitude_geoint32(*((int32_t*)coords));
183  }
184  } else {
186  return *((double*)coords) == NULL_ARRAY_DOUBLE;
187  }
188  }
189  return false;
190 }
191 
192 } // namespace geospatial
#define CHECK_EQ(x, y)
Definition: Logger.h:205
#define NULL_DOUBLE
Definition: sqltypes.h:185
uint64_t compress_null_point(const SQLTypeInfo &ti, bool x)
#define NULL_ARRAY_DOUBLE
Definition: sqltypes.h:193
DEVICE constexpr uint64_t compress_null_point_lattitude_geoint32()
HOST DEVICE SQLTypes get_type() const
Definition: sqltypes.h:258
DEVICE bool is_null_point_longitude_geoint32(const int32_t compressed)
std::string to_string(char const *&&v)
CHECK(cgen_state)
DEVICE uint64_t compress_longitude_coord_geoint32(const double coord)
#define COMPRESSION_NONE
void unpack_geo_vector(std::vector< T > &output, const int8_t *input_ptr, const size_t sz)
DEVICE constexpr uint64_t compress_null_point_longitude_geoint32()
DEVICE uint64_t compress_lattitude_coord_geoint32(const double coord)
DEVICE double decompress_lattitude_coord_geoint32(const int32_t compressed)
HOST DEVICE EncodingType get_compression() const
Definition: sqltypes.h:266
uint64_t compress_coord(double coord, const SQLTypeInfo &ti, bool x)
std::vector< uint8_t > compress_coords(std::vector< double > &coords, const SQLTypeInfo &ti)
#define COMPRESSION_GEOINT32
HOST DEVICE int get_comp_param() const
Definition: sqltypes.h:267
void unpack_geo_vector< int32_t >(std::vector< int32_t > &output, const int8_t *input_ptr, const size_t sz)
std::shared_ptr< std::vector< double > > decompress_coords< double, SQLTypeInfo >(const SQLTypeInfo &geo_ti, const int8_t *coords, const size_t coords_sz)
bool is_null_point(const SQLTypeInfo &geo_ti, const int8_t *coords, const size_t coords_sz)
int32_t get_compression_scheme(const SQLTypeInfo &ti)
DEVICE double decompress_longitude_coord_geoint32(const int32_t compressed)
HOST DEVICE bool get_notnull() const
Definition: sqltypes.h:265
std::shared_ptr< std::vector< double > > decompress_coords< double, int32_t >(const int32_t &ic, const int8_t *coords, const size_t coords_sz)
HOST DEVICE int get_output_srid() const
Definition: sqltypes.h:264
void decompress_geo_coords_geoint32(std::vector< T > &dec, const int8_t *enc, const size_t sz)