forked from rogersce/cnpy
-
Notifications
You must be signed in to change notification settings - Fork 0
/
cnpy.cpp
258 lines (205 loc) · 7.84 KB
/
cnpy.cpp
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
//Copyright (C) 2011 Carl Rogers
//Released under MIT License
//license available in LICENSE file, or at http://www.opensource.org/licenses/mit-license.php
#include"cnpy.h"
#include<complex>
#include<cstdlib>
#include<algorithm>
#include<cstring>
#include<iomanip>
#include<stdint.h>
#include<stdexcept>
#include <regex>
char cnpy::BigEndianTest() {
int x = 1;
return (((char *)&x)[0]) ? '<' : '>';
}
char cnpy::map_type(const std::type_info& t)
{
if(t == typeid(float) ) return 'f';
if(t == typeid(double) ) return 'f';
if(t == typeid(long double) ) return 'f';
if(t == typeid(int) ) return 'i';
if(t == typeid(char) ) return 'i';
if(t == typeid(short) ) return 'i';
if(t == typeid(long) ) return 'i';
if(t == typeid(long long) ) return 'i';
if(t == typeid(unsigned char) ) return 'u';
if(t == typeid(unsigned short) ) return 'u';
if(t == typeid(unsigned long) ) return 'u';
if(t == typeid(unsigned long long) ) return 'u';
if(t == typeid(unsigned int) ) return 'u';
if(t == typeid(bool) ) return 'b';
if(t == typeid(std::complex<float>) ) return 'c';
if(t == typeid(std::complex<double>) ) return 'c';
if(t == typeid(std::complex<long double>) ) return 'c';
else return '?';
}
template<> std::vector<char>& cnpy::operator+=(std::vector<char>& lhs, const std::string rhs) {
lhs.insert(lhs.end(),rhs.begin(),rhs.end());
return lhs;
}
template<> std::vector<char>& cnpy::operator+=(std::vector<char>& lhs, const char* rhs) {
//write in little endian
size_t len = strlen(rhs);
lhs.reserve(len);
for(size_t byte = 0; byte < len; byte++) {
lhs.push_back(rhs[byte]);
}
return lhs;
}
void cnpy::parse_npy_header(std::string &buffer, size_t &word_size,
std::vector<size_t> &shape, bool &fortran_order) {
assert(buffer.substr(0, 6) == "\x93NUMPY");
//uint8_t major_version = *reinterpret_cast<uint8_t *>(buffer.data() + 6);
//uint8_t minor_version = *reinterpret_cast<uint8_t *>(buffer.data() + 7);
uint16_t header_len =
(*reinterpret_cast<uint8_t *>(buffer.data() + 8)) |
((*reinterpret_cast<uint8_t *>(buffer.data() + 9)) << 8);
std::string header = buffer.substr(10, header_len);
size_t loc1, loc2;
//fortran order
loc1 = header.find("fortran_order")+16;
fortran_order = (header.substr(loc1,4) == "True" ? true : false);
//shape
loc1 = header.find("(");
loc2 = header.find(")");
std::regex num_regex("[0-9][0-9]*");
std::smatch sm;
shape.clear();
std::string str_shape = header.substr(loc1+1,loc2-loc1-1);
while(std::regex_search(str_shape, sm, num_regex)) {
shape.push_back(std::stoi(sm[0].str()));
str_shape = sm.suffix().str();
}
//endian, word size, data type
//byte order code | stands for not applicable.
//not sure when this applies except for byte array
loc1 = header.find("descr")+9;
bool littleEndian = (header[loc1] == '<' || header[loc1] == '|' ? true : false);
assert(littleEndian);
//char type = header[loc1+1];
//assert(type == map_type(T));
std::string str_ws = header.substr(loc1+2);
loc2 = str_ws.find("'");
word_size = atoi(str_ws.substr(0,loc2).c_str());
}
void cnpy::parse_npy_header(FILE* fp, size_t& word_size, std::vector<size_t>& shape, bool& fortran_order) {
char buffer[256];
size_t res = fread(buffer,sizeof(char),11,fp);
if(res != 11)
throw std::runtime_error("parse_npy_header: failed fread");
std::string header = fgets(buffer,256,fp);
assert(header[header.size()-1] == '\n');
size_t loc1, loc2;
//fortran order
loc1 = header.find("fortran_order");
if (loc1 == std::string::npos)
throw std::runtime_error("parse_npy_header: failed to find header keyword: 'fortran_order'");
loc1 += 16;
fortran_order = (header.substr(loc1,4) == "True" ? true : false);
//shape
loc1 = header.find("(");
loc2 = header.find(")");
if (loc1 == std::string::npos || loc2 == std::string::npos)
throw std::runtime_error("parse_npy_header: failed to find header keyword: '(' or ')'");
std::regex num_regex("[0-9][0-9]*");
std::smatch sm;
shape.clear();
std::string str_shape = header.substr(loc1+1,loc2-loc1-1);
while(std::regex_search(str_shape, sm, num_regex)) {
shape.push_back(std::stoi(sm[0].str()));
str_shape = sm.suffix().str();
}
//endian, word size, data type
//byte order code | stands for not applicable.
//not sure when this applies except for byte array
loc1 = header.find("descr");
if (loc1 == std::string::npos)
throw std::runtime_error("parse_npy_header: failed to find header keyword: 'descr'");
loc1 += 9;
bool littleEndian = (header[loc1] == '<' || header[loc1] == '|' ? true : false);
assert(littleEndian);
//char type = header[loc1+1];
//assert(type == map_type(T));
std::string str_ws = header.substr(loc1+2);
loc2 = str_ws.find("'");
word_size = atoi(str_ws.substr(0,loc2).c_str());
}
void cnpy::parse_zip_footer(FILE* fp, uint16_t& nrecs, size_t& global_header_size, size_t& global_header_offset)
{
std::vector<char> footer(22);
fseek(fp,-22,SEEK_END);
size_t res = fread(&footer[0],sizeof(char),22,fp);
if(res != 22)
throw std::runtime_error("parse_zip_footer: failed fread");
uint16_t disk_no, disk_start, nrecs_on_disk, comment_len;
disk_no = *(uint16_t*) &footer[4];
disk_start = *(uint16_t*) &footer[6];
nrecs_on_disk = *(uint16_t*) &footer[8];
nrecs = *(uint16_t*) &footer[10];
global_header_size = *(uint32_t*) &footer[12];
global_header_offset = *(uint32_t*) &footer[16];
comment_len = *(uint16_t*) &footer[20];
assert(disk_no == 0);
assert(disk_start == 0);
assert(nrecs_on_disk == nrecs);
assert(comment_len == 0);
}
cnpy::NpyArray load_the_npy_file(FILE* fp) {
std::vector<size_t> shape;
size_t word_size;
bool fortran_order;
cnpy::parse_npy_header(fp,word_size,shape,fortran_order);
cnpy::NpyArray arr(shape, word_size, fortran_order);
size_t nread = fread(arr.data<char>(),1,arr.num_bytes(),fp);
if(nread != arr.num_bytes())
throw std::runtime_error("load_the_npy_file: failed fread");
return arr;
}
cnpy::NpyArray load_the_npz_array(libzip::file &file, uint64_t size) {
std::string buffer = file.read(size);
std::vector<size_t> shape;
size_t word_size;
bool fortran_order;
cnpy::parse_npy_header(buffer, word_size,
shape, fortran_order);
cnpy::NpyArray array(shape, word_size, fortran_order);
uint64_t offset = size - array.num_bytes();
memcpy(array.data<unsigned char>(), buffer.data() + offset,
array.num_bytes());
return array;
}
cnpy::npz_t cnpy::npz_load(std::string fname) {
libzip::archive zip(fname, ZIP_RDONLY);
cnpy::npz_t arrays;
for (auto stat : zip) {
if (!(stat.valid & (ZIP_STAT_NAME | ZIP_STAT_SIZE))) {
// Skip files without name or size
continue;
}
std::string name(stat.name);
auto file = zip.open(name);
// erase the lagging .npy
std::string varname = name.substr(0, name.size() - 4);
arrays[varname] = load_the_npz_array(file, stat.size);
}
return arrays;
}
cnpy::NpyArray cnpy::npz_load(std::string fname, std::string varname) {
libzip::archive zip(fname, ZIP_RDONLY);
std::string varfilename = varname + ".npy";
libzip::stat stat = zip.stat(varfilename);
if (!(stat.valid & ZIP_STAT_SIZE)) {
throw std::runtime_error("npz_load: No size associated with " + varname);
}
auto file = zip.open(varfilename);
return load_the_npz_array(file, stat.size);
}
cnpy::NpyArray cnpy::npy_load(std::string fname) {
FILE* fp = fopen(fname.c_str(), "rb");
if(!fp) throw std::runtime_error("npy_load: Unable to open file "+fname);
NpyArray arr = load_the_npy_file(fp);
fclose(fp);
return arr;
}