Show More
@@ -1,1212 +1,1223 b'' | |||
|
1 | 1 | /* |
|
2 | 2 | Copyright (c) 2013-2015, Facebook, Inc. |
|
3 | 3 | All rights reserved. |
|
4 | 4 | |
|
5 | 5 | Redistribution and use in source and binary forms, with or without |
|
6 | 6 | modification, are permitted provided that the following conditions are met: |
|
7 | 7 | |
|
8 | 8 | * Redistributions of source code must retain the above copyright notice, |
|
9 | 9 | this list of conditions and the following disclaimer. |
|
10 | 10 | |
|
11 | 11 | * Redistributions in binary form must reproduce the above copyright notice, |
|
12 | 12 | this list of conditions and the following disclaimer in the documentation |
|
13 | 13 | and/or other materials provided with the distribution. |
|
14 | 14 | |
|
15 | 15 | * Neither the name Facebook nor the names of its contributors may be used to |
|
16 | 16 | endorse or promote products derived from this software without specific |
|
17 | 17 | prior written permission. |
|
18 | 18 | |
|
19 | 19 | THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" |
|
20 | 20 | AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
|
21 | 21 | IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE |
|
22 | 22 | DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE |
|
23 | 23 | FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
|
24 | 24 | DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR |
|
25 | 25 | SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER |
|
26 | 26 | CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, |
|
27 | 27 | OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE |
|
28 | 28 | OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
|
29 | 29 | */ |
|
30 | 30 | |
|
31 | 31 | #include <Python.h> |
|
32 | 32 | #include <bytesobject.h> |
|
33 | 33 | #ifdef _MSC_VER |
|
34 | 34 | #define inline __inline |
|
35 | 35 | #if _MSC_VER >= 1800 |
|
36 | 36 | #include <stdint.h> |
|
37 | 37 | #else |
|
38 | 38 | // The compiler associated with Python 2.7 on Windows doesn't ship |
|
39 | 39 | // with stdint.h, so define the small subset that we use here. |
|
40 | 40 | typedef __int8 int8_t; |
|
41 | 41 | typedef __int16 int16_t; |
|
42 | 42 | typedef __int32 int32_t; |
|
43 | 43 | typedef __int64 int64_t; |
|
44 | 44 | typedef unsigned __int8 uint8_t; |
|
45 | 45 | typedef unsigned __int16 uint16_t; |
|
46 | 46 | typedef unsigned __int32 uint32_t; |
|
47 | 47 | typedef unsigned __int64 uint64_t; |
|
48 | 48 | #define UINT32_MAX 4294967295U |
|
49 | 49 | #endif |
|
50 | 50 | #endif |
|
51 | 51 | |
|
52 | 52 | // clang-format off |
|
53 | 53 | /* Return the smallest size int that can store the value */ |
|
54 | 54 | #define INT_SIZE(x) (((x) == ((int8_t)x)) ? 1 : \ |
|
55 | 55 | ((x) == ((int16_t)x)) ? 2 : \ |
|
56 | 56 | ((x) == ((int32_t)x)) ? 4 : 8) |
|
57 | 57 | |
|
58 | 58 | #define BSER_ARRAY 0x00 |
|
59 | 59 | #define BSER_OBJECT 0x01 |
|
60 | 60 | #define BSER_BYTESTRING 0x02 |
|
61 | 61 | #define BSER_INT8 0x03 |
|
62 | 62 | #define BSER_INT16 0x04 |
|
63 | 63 | #define BSER_INT32 0x05 |
|
64 | 64 | #define BSER_INT64 0x06 |
|
65 | 65 | #define BSER_REAL 0x07 |
|
66 | 66 | #define BSER_TRUE 0x08 |
|
67 | 67 | #define BSER_FALSE 0x09 |
|
68 | 68 | #define BSER_NULL 0x0a |
|
69 | 69 | #define BSER_TEMPLATE 0x0b |
|
70 | 70 | #define BSER_SKIP 0x0c |
|
71 | 71 | #define BSER_UTF8STRING 0x0d |
|
72 | 72 | // clang-format on |
|
73 | 73 | |
|
74 | 74 | // An immutable object representation of BSER_OBJECT. |
|
75 | 75 | // Rather than build a hash table, key -> value are obtained |
|
76 | 76 | // by walking the list of keys to determine the offset into |
|
77 | 77 | // the values array. The assumption is that the number of |
|
78 | 78 | // array elements will be typically small (~6 for the top |
|
79 | 79 | // level query result and typically 3-5 for the file entries) |
|
80 | 80 | // so that the time overhead for this is small compared to |
|
81 | 81 | // using a proper hash table. Even with this simplistic |
|
82 | 82 | // approach, this is still faster for the mercurial use case |
|
83 | 83 | // as it helps to eliminate creating N other objects to |
|
84 | 84 | // represent the stat information in the hgwatchman extension |
|
85 | 85 | // clang-format off |
|
86 | 86 | typedef struct { |
|
87 | 87 | PyObject_HEAD |
|
88 | 88 | PyObject *keys; // tuple of field names |
|
89 | 89 | PyObject *values; // tuple of values |
|
90 | 90 | } bserObject; |
|
91 | 91 | // clang-format on |
|
92 | 92 | |
|
93 | 93 | static Py_ssize_t bserobj_tuple_length(PyObject* o) { |
|
94 | 94 | bserObject* obj = (bserObject*)o; |
|
95 | 95 | |
|
96 | 96 | return PySequence_Length(obj->keys); |
|
97 | 97 | } |
|
98 | 98 | |
|
99 | 99 | static PyObject* bserobj_tuple_item(PyObject* o, Py_ssize_t i) { |
|
100 | 100 | bserObject* obj = (bserObject*)o; |
|
101 | 101 | |
|
102 | 102 | return PySequence_GetItem(obj->values, i); |
|
103 | 103 | } |
|
104 | 104 | |
|
105 | 105 | // clang-format off |
|
106 | 106 | static PySequenceMethods bserobj_sq = { |
|
107 | 107 | bserobj_tuple_length, /* sq_length */ |
|
108 | 108 | 0, /* sq_concat */ |
|
109 | 109 | 0, /* sq_repeat */ |
|
110 | 110 | bserobj_tuple_item, /* sq_item */ |
|
111 | 111 | 0, /* sq_ass_item */ |
|
112 | 112 | 0, /* sq_contains */ |
|
113 | 113 | 0, /* sq_inplace_concat */ |
|
114 | 114 | 0 /* sq_inplace_repeat */ |
|
115 | 115 | }; |
|
116 | 116 | // clang-format on |
|
117 | 117 | |
|
118 | 118 | static void bserobj_dealloc(PyObject* o) { |
|
119 | 119 | bserObject* obj = (bserObject*)o; |
|
120 | 120 | |
|
121 | 121 | Py_CLEAR(obj->keys); |
|
122 | 122 | Py_CLEAR(obj->values); |
|
123 | 123 | PyObject_Del(o); |
|
124 | 124 | } |
|
125 | 125 | |
|
126 | 126 | static PyObject* bserobj_getattrro(PyObject* o, PyObject* name) { |
|
127 | 127 | bserObject* obj = (bserObject*)o; |
|
128 | 128 | Py_ssize_t i, n; |
|
129 | 129 | PyObject* name_bytes = NULL; |
|
130 | 130 | PyObject* ret = NULL; |
|
131 | const char* namestr; | |
|
131 | const char* namestr = NULL; | |
|
132 | 132 | |
|
133 | 133 | if (PyIndex_Check(name)) { |
|
134 | 134 | i = PyNumber_AsSsize_t(name, PyExc_IndexError); |
|
135 | 135 | if (i == -1 && PyErr_Occurred()) { |
|
136 | 136 | goto bail; |
|
137 | 137 | } |
|
138 | ret = PySequence_GetItem(obj->values, i); | |
|
139 | goto bail; | |
|
140 | } | |
|
141 | 138 | |
|
142 | // We can be passed in Unicode objects here -- we don't support anything other | |
|
143 | // than UTF-8 for keys. | |
|
144 | if (PyUnicode_Check(name)) { | |
|
145 | name_bytes = PyUnicode_AsUTF8String(name); | |
|
146 | if (name_bytes == NULL) { | |
|
139 | if (i == 8 && PySequence_Size(obj->values) < 9) { | |
|
140 | // Hack alert: Python 3 removed support for os.stat().st_mtime | |
|
141 | // being an integer.Instead, if you need an integer, you have to | |
|
142 | // use os.stat()[stat.ST_MTIME] instead. stat.ST_MTIME is 8, and | |
|
143 | // our stat tuples are shorter than that, so we can detect | |
|
144 | // requests for index 8 on tuples shorter than that and return | |
|
145 | // st_mtime instead. | |
|
146 | namestr = "st_mtime"; | |
|
147 | } else { | |
|
148 | ret = PySequence_GetItem(obj->values, i); | |
|
147 | 149 | goto bail; |
|
148 | 150 | } |
|
149 | namestr = PyBytes_AsString(name_bytes); | |
|
150 | 151 | } else { |
|
151 | namestr = PyBytes_AsString(name); | |
|
152 | // We can be passed in Unicode objects here -- we don't support anything other | |
|
153 | // than UTF-8 for keys. | |
|
154 | if (PyUnicode_Check(name)) { | |
|
155 | name_bytes = PyUnicode_AsUTF8String(name); | |
|
156 | if (name_bytes == NULL) { | |
|
157 | goto bail; | |
|
158 | } | |
|
159 | namestr = PyBytes_AsString(name_bytes); | |
|
160 | } else { | |
|
161 | namestr = PyBytes_AsString(name); | |
|
162 | } | |
|
152 | 163 | } |
|
153 | 164 | |
|
154 | 165 | if (namestr == NULL) { |
|
155 | 166 | goto bail; |
|
156 | 167 | } |
|
157 | 168 | // hack^Wfeature to allow mercurial to use "st_size" to reference "size" |
|
158 | 169 | if (!strncmp(namestr, "st_", 3)) { |
|
159 | 170 | namestr += 3; |
|
160 | 171 | } |
|
161 | 172 | |
|
162 | 173 | n = PyTuple_GET_SIZE(obj->keys); |
|
163 | 174 | for (i = 0; i < n; i++) { |
|
164 | 175 | const char* item_name = NULL; |
|
165 | 176 | PyObject* key = PyTuple_GET_ITEM(obj->keys, i); |
|
166 | 177 | |
|
167 | 178 | item_name = PyBytes_AsString(key); |
|
168 | 179 | if (!strcmp(item_name, namestr)) { |
|
169 | 180 | ret = PySequence_GetItem(obj->values, i); |
|
170 | 181 | goto bail; |
|
171 | 182 | } |
|
172 | 183 | } |
|
173 | 184 | |
|
174 | 185 | PyErr_Format( |
|
175 | 186 | PyExc_AttributeError, "bserobject has no attribute '%.400s'", namestr); |
|
176 | 187 | bail: |
|
177 | 188 | Py_XDECREF(name_bytes); |
|
178 | 189 | return ret; |
|
179 | 190 | } |
|
180 | 191 | |
|
181 | 192 | // clang-format off |
|
182 | 193 | static PyMappingMethods bserobj_map = { |
|
183 | 194 | bserobj_tuple_length, /* mp_length */ |
|
184 | 195 | bserobj_getattrro, /* mp_subscript */ |
|
185 | 196 | 0 /* mp_ass_subscript */ |
|
186 | 197 | }; |
|
187 | 198 | |
|
188 | 199 | PyTypeObject bserObjectType = { |
|
189 | 200 | PyVarObject_HEAD_INIT(NULL, 0) |
|
190 | 201 | "bserobj_tuple", /* tp_name */ |
|
191 | 202 | sizeof(bserObject), /* tp_basicsize */ |
|
192 | 203 | 0, /* tp_itemsize */ |
|
193 | 204 | bserobj_dealloc, /* tp_dealloc */ |
|
194 | 205 | 0, /* tp_print */ |
|
195 | 206 | 0, /* tp_getattr */ |
|
196 | 207 | 0, /* tp_setattr */ |
|
197 | 208 | 0, /* tp_compare */ |
|
198 | 209 | 0, /* tp_repr */ |
|
199 | 210 | 0, /* tp_as_number */ |
|
200 | 211 | &bserobj_sq, /* tp_as_sequence */ |
|
201 | 212 | &bserobj_map, /* tp_as_mapping */ |
|
202 | 213 | 0, /* tp_hash */ |
|
203 | 214 | 0, /* tp_call */ |
|
204 | 215 | 0, /* tp_str */ |
|
205 | 216 | bserobj_getattrro, /* tp_getattro */ |
|
206 | 217 | 0, /* tp_setattro */ |
|
207 | 218 | 0, /* tp_as_buffer */ |
|
208 | 219 | Py_TPFLAGS_DEFAULT, /* tp_flags */ |
|
209 | 220 | "bserobj tuple", /* tp_doc */ |
|
210 | 221 | 0, /* tp_traverse */ |
|
211 | 222 | 0, /* tp_clear */ |
|
212 | 223 | 0, /* tp_richcompare */ |
|
213 | 224 | 0, /* tp_weaklistoffset */ |
|
214 | 225 | 0, /* tp_iter */ |
|
215 | 226 | 0, /* tp_iternext */ |
|
216 | 227 | 0, /* tp_methods */ |
|
217 | 228 | 0, /* tp_members */ |
|
218 | 229 | 0, /* tp_getset */ |
|
219 | 230 | 0, /* tp_base */ |
|
220 | 231 | 0, /* tp_dict */ |
|
221 | 232 | 0, /* tp_descr_get */ |
|
222 | 233 | 0, /* tp_descr_set */ |
|
223 | 234 | 0, /* tp_dictoffset */ |
|
224 | 235 | 0, /* tp_init */ |
|
225 | 236 | 0, /* tp_alloc */ |
|
226 | 237 | 0, /* tp_new */ |
|
227 | 238 | }; |
|
228 | 239 | // clang-format on |
|
229 | 240 | |
|
230 | 241 | typedef struct loads_ctx { |
|
231 | 242 | int mutable; |
|
232 | 243 | const char* value_encoding; |
|
233 | 244 | const char* value_errors; |
|
234 | 245 | uint32_t bser_version; |
|
235 | 246 | uint32_t bser_capabilities; |
|
236 | 247 | } unser_ctx_t; |
|
237 | 248 | |
|
238 | 249 | static PyObject* |
|
239 | 250 | bser_loads_recursive(const char** ptr, const char* end, const unser_ctx_t* ctx); |
|
240 | 251 | |
|
241 | 252 | static const char bser_true = BSER_TRUE; |
|
242 | 253 | static const char bser_false = BSER_FALSE; |
|
243 | 254 | static const char bser_null = BSER_NULL; |
|
244 | 255 | static const char bser_bytestring_hdr = BSER_BYTESTRING; |
|
245 | 256 | static const char bser_array_hdr = BSER_ARRAY; |
|
246 | 257 | static const char bser_object_hdr = BSER_OBJECT; |
|
247 | 258 | |
|
248 | 259 | static inline uint32_t next_power_2(uint32_t n) { |
|
249 | 260 | n |= (n >> 16); |
|
250 | 261 | n |= (n >> 8); |
|
251 | 262 | n |= (n >> 4); |
|
252 | 263 | n |= (n >> 2); |
|
253 | 264 | n |= (n >> 1); |
|
254 | 265 | return n + 1; |
|
255 | 266 | } |
|
256 | 267 | |
|
257 | 268 | // A buffer we use for building up the serialized result |
|
258 | 269 | struct bser_buffer { |
|
259 | 270 | char* buf; |
|
260 | 271 | int wpos, allocd; |
|
261 | 272 | uint32_t bser_version; |
|
262 | 273 | uint32_t capabilities; |
|
263 | 274 | }; |
|
264 | 275 | typedef struct bser_buffer bser_t; |
|
265 | 276 | |
|
266 | 277 | static int bser_append(bser_t* bser, const char* data, uint32_t len) { |
|
267 | 278 | int newlen = next_power_2(bser->wpos + len); |
|
268 | 279 | if (newlen > bser->allocd) { |
|
269 | 280 | char* nbuf = realloc(bser->buf, newlen); |
|
270 | 281 | if (!nbuf) { |
|
271 | 282 | return 0; |
|
272 | 283 | } |
|
273 | 284 | |
|
274 | 285 | bser->buf = nbuf; |
|
275 | 286 | bser->allocd = newlen; |
|
276 | 287 | } |
|
277 | 288 | |
|
278 | 289 | memcpy(bser->buf + bser->wpos, data, len); |
|
279 | 290 | bser->wpos += len; |
|
280 | 291 | return 1; |
|
281 | 292 | } |
|
282 | 293 | |
|
283 | 294 | static int bser_init(bser_t* bser, uint32_t version, uint32_t capabilities) { |
|
284 | 295 | bser->allocd = 8192; |
|
285 | 296 | bser->wpos = 0; |
|
286 | 297 | bser->buf = malloc(bser->allocd); |
|
287 | 298 | bser->bser_version = version; |
|
288 | 299 | bser->capabilities = capabilities; |
|
289 | 300 | if (!bser->buf) { |
|
290 | 301 | return 0; |
|
291 | 302 | } |
|
292 | 303 | |
|
293 | 304 | // Leave room for the serialization header, which includes |
|
294 | 305 | // our overall length. To make things simpler, we'll use an |
|
295 | 306 | // int32 for the header |
|
296 | 307 | #define EMPTY_HEADER "\x00\x01\x05\x00\x00\x00\x00" |
|
297 | 308 | |
|
298 | 309 | // Version 2 also carries an integer indicating the capabilities. The |
|
299 | 310 | // capabilities integer comes before the PDU size. |
|
300 | 311 | #define EMPTY_HEADER_V2 "\x00\x02\x00\x00\x00\x00\x05\x00\x00\x00\x00" |
|
301 | 312 | if (version == 2) { |
|
302 | 313 | bser_append(bser, EMPTY_HEADER_V2, sizeof(EMPTY_HEADER_V2) - 1); |
|
303 | 314 | } else { |
|
304 | 315 | bser_append(bser, EMPTY_HEADER, sizeof(EMPTY_HEADER) - 1); |
|
305 | 316 | } |
|
306 | 317 | |
|
307 | 318 | return 1; |
|
308 | 319 | } |
|
309 | 320 | |
|
310 | 321 | static void bser_dtor(bser_t* bser) { |
|
311 | 322 | free(bser->buf); |
|
312 | 323 | bser->buf = NULL; |
|
313 | 324 | } |
|
314 | 325 | |
|
315 | 326 | static int bser_long(bser_t* bser, int64_t val) { |
|
316 | 327 | int8_t i8; |
|
317 | 328 | int16_t i16; |
|
318 | 329 | int32_t i32; |
|
319 | 330 | int64_t i64; |
|
320 | 331 | char sz; |
|
321 | 332 | int size = INT_SIZE(val); |
|
322 | 333 | char* iptr; |
|
323 | 334 | |
|
324 | 335 | switch (size) { |
|
325 | 336 | case 1: |
|
326 | 337 | sz = BSER_INT8; |
|
327 | 338 | i8 = (int8_t)val; |
|
328 | 339 | iptr = (char*)&i8; |
|
329 | 340 | break; |
|
330 | 341 | case 2: |
|
331 | 342 | sz = BSER_INT16; |
|
332 | 343 | i16 = (int16_t)val; |
|
333 | 344 | iptr = (char*)&i16; |
|
334 | 345 | break; |
|
335 | 346 | case 4: |
|
336 | 347 | sz = BSER_INT32; |
|
337 | 348 | i32 = (int32_t)val; |
|
338 | 349 | iptr = (char*)&i32; |
|
339 | 350 | break; |
|
340 | 351 | case 8: |
|
341 | 352 | sz = BSER_INT64; |
|
342 | 353 | i64 = (int64_t)val; |
|
343 | 354 | iptr = (char*)&i64; |
|
344 | 355 | break; |
|
345 | 356 | default: |
|
346 | 357 | PyErr_SetString(PyExc_RuntimeError, "Cannot represent this long value!?"); |
|
347 | 358 | return 0; |
|
348 | 359 | } |
|
349 | 360 | |
|
350 | 361 | if (!bser_append(bser, &sz, sizeof(sz))) { |
|
351 | 362 | return 0; |
|
352 | 363 | } |
|
353 | 364 | |
|
354 | 365 | return bser_append(bser, iptr, size); |
|
355 | 366 | } |
|
356 | 367 | |
|
357 | 368 | static int bser_bytestring(bser_t* bser, PyObject* sval) { |
|
358 | 369 | char* buf = NULL; |
|
359 | 370 | Py_ssize_t len; |
|
360 | 371 | int res; |
|
361 | 372 | PyObject* utf = NULL; |
|
362 | 373 | |
|
363 | 374 | if (PyUnicode_Check(sval)) { |
|
364 | 375 | utf = PyUnicode_AsEncodedString(sval, "utf-8", "ignore"); |
|
365 | 376 | sval = utf; |
|
366 | 377 | } |
|
367 | 378 | |
|
368 | 379 | res = PyBytes_AsStringAndSize(sval, &buf, &len); |
|
369 | 380 | if (res == -1) { |
|
370 | 381 | res = 0; |
|
371 | 382 | goto out; |
|
372 | 383 | } |
|
373 | 384 | |
|
374 | 385 | if (!bser_append(bser, &bser_bytestring_hdr, sizeof(bser_bytestring_hdr))) { |
|
375 | 386 | res = 0; |
|
376 | 387 | goto out; |
|
377 | 388 | } |
|
378 | 389 | |
|
379 | 390 | if (!bser_long(bser, len)) { |
|
380 | 391 | res = 0; |
|
381 | 392 | goto out; |
|
382 | 393 | } |
|
383 | 394 | |
|
384 | 395 | if (len > UINT32_MAX) { |
|
385 | 396 | PyErr_Format(PyExc_ValueError, "string too big"); |
|
386 | 397 | res = 0; |
|
387 | 398 | goto out; |
|
388 | 399 | } |
|
389 | 400 | |
|
390 | 401 | res = bser_append(bser, buf, (uint32_t)len); |
|
391 | 402 | |
|
392 | 403 | out: |
|
393 | 404 | if (utf) { |
|
394 | 405 | Py_DECREF(utf); |
|
395 | 406 | } |
|
396 | 407 | |
|
397 | 408 | return res; |
|
398 | 409 | } |
|
399 | 410 | |
|
400 | 411 | static int bser_recursive(bser_t* bser, PyObject* val) { |
|
401 | 412 | if (PyBool_Check(val)) { |
|
402 | 413 | if (val == Py_True) { |
|
403 | 414 | return bser_append(bser, &bser_true, sizeof(bser_true)); |
|
404 | 415 | } |
|
405 | 416 | return bser_append(bser, &bser_false, sizeof(bser_false)); |
|
406 | 417 | } |
|
407 | 418 | |
|
408 | 419 | if (val == Py_None) { |
|
409 | 420 | return bser_append(bser, &bser_null, sizeof(bser_null)); |
|
410 | 421 | } |
|
411 | 422 | |
|
412 | 423 | // Python 3 has one integer type. |
|
413 | 424 | #if PY_MAJOR_VERSION < 3 |
|
414 | 425 | if (PyInt_Check(val)) { |
|
415 | 426 | return bser_long(bser, PyInt_AS_LONG(val)); |
|
416 | 427 | } |
|
417 | 428 | #endif // PY_MAJOR_VERSION < 3 |
|
418 | 429 | |
|
419 | 430 | if (PyLong_Check(val)) { |
|
420 | 431 | return bser_long(bser, PyLong_AsLongLong(val)); |
|
421 | 432 | } |
|
422 | 433 | |
|
423 | 434 | if (PyBytes_Check(val) || PyUnicode_Check(val)) { |
|
424 | 435 | return bser_bytestring(bser, val); |
|
425 | 436 | } |
|
426 | 437 | |
|
427 | 438 | if (PyFloat_Check(val)) { |
|
428 | 439 | double dval = PyFloat_AS_DOUBLE(val); |
|
429 | 440 | char sz = BSER_REAL; |
|
430 | 441 | |
|
431 | 442 | if (!bser_append(bser, &sz, sizeof(sz))) { |
|
432 | 443 | return 0; |
|
433 | 444 | } |
|
434 | 445 | |
|
435 | 446 | return bser_append(bser, (char*)&dval, sizeof(dval)); |
|
436 | 447 | } |
|
437 | 448 | |
|
438 | 449 | if (PyList_Check(val)) { |
|
439 | 450 | Py_ssize_t i, len = PyList_GET_SIZE(val); |
|
440 | 451 | |
|
441 | 452 | if (!bser_append(bser, &bser_array_hdr, sizeof(bser_array_hdr))) { |
|
442 | 453 | return 0; |
|
443 | 454 | } |
|
444 | 455 | |
|
445 | 456 | if (!bser_long(bser, len)) { |
|
446 | 457 | return 0; |
|
447 | 458 | } |
|
448 | 459 | |
|
449 | 460 | for (i = 0; i < len; i++) { |
|
450 | 461 | PyObject* ele = PyList_GET_ITEM(val, i); |
|
451 | 462 | |
|
452 | 463 | if (!bser_recursive(bser, ele)) { |
|
453 | 464 | return 0; |
|
454 | 465 | } |
|
455 | 466 | } |
|
456 | 467 | |
|
457 | 468 | return 1; |
|
458 | 469 | } |
|
459 | 470 | |
|
460 | 471 | if (PyTuple_Check(val)) { |
|
461 | 472 | Py_ssize_t i, len = PyTuple_GET_SIZE(val); |
|
462 | 473 | |
|
463 | 474 | if (!bser_append(bser, &bser_array_hdr, sizeof(bser_array_hdr))) { |
|
464 | 475 | return 0; |
|
465 | 476 | } |
|
466 | 477 | |
|
467 | 478 | if (!bser_long(bser, len)) { |
|
468 | 479 | return 0; |
|
469 | 480 | } |
|
470 | 481 | |
|
471 | 482 | for (i = 0; i < len; i++) { |
|
472 | 483 | PyObject* ele = PyTuple_GET_ITEM(val, i); |
|
473 | 484 | |
|
474 | 485 | if (!bser_recursive(bser, ele)) { |
|
475 | 486 | return 0; |
|
476 | 487 | } |
|
477 | 488 | } |
|
478 | 489 | |
|
479 | 490 | return 1; |
|
480 | 491 | } |
|
481 | 492 | |
|
482 | 493 | if (PyMapping_Check(val)) { |
|
483 | 494 | Py_ssize_t len = PyMapping_Length(val); |
|
484 | 495 | Py_ssize_t pos = 0; |
|
485 | 496 | PyObject *key, *ele; |
|
486 | 497 | |
|
487 | 498 | if (!bser_append(bser, &bser_object_hdr, sizeof(bser_object_hdr))) { |
|
488 | 499 | return 0; |
|
489 | 500 | } |
|
490 | 501 | |
|
491 | 502 | if (!bser_long(bser, len)) { |
|
492 | 503 | return 0; |
|
493 | 504 | } |
|
494 | 505 | |
|
495 | 506 | while (PyDict_Next(val, &pos, &key, &ele)) { |
|
496 | 507 | if (!bser_bytestring(bser, key)) { |
|
497 | 508 | return 0; |
|
498 | 509 | } |
|
499 | 510 | if (!bser_recursive(bser, ele)) { |
|
500 | 511 | return 0; |
|
501 | 512 | } |
|
502 | 513 | } |
|
503 | 514 | |
|
504 | 515 | return 1; |
|
505 | 516 | } |
|
506 | 517 | |
|
507 | 518 | PyErr_SetString(PyExc_ValueError, "Unsupported value type"); |
|
508 | 519 | return 0; |
|
509 | 520 | } |
|
510 | 521 | |
|
511 | 522 | static PyObject* bser_dumps(PyObject* self, PyObject* args, PyObject* kw) { |
|
512 | 523 | PyObject *val = NULL, *res; |
|
513 | 524 | bser_t bser; |
|
514 | 525 | uint32_t len, bser_version = 1, bser_capabilities = 0; |
|
515 | 526 | |
|
516 | 527 | static char* kw_list[] = {"val", "version", "capabilities", NULL}; |
|
517 | 528 | |
|
518 | 529 | if (!PyArg_ParseTupleAndKeywords( |
|
519 | 530 | args, |
|
520 | 531 | kw, |
|
521 | 532 | "O|ii:dumps", |
|
522 | 533 | kw_list, |
|
523 | 534 | &val, |
|
524 | 535 | &bser_version, |
|
525 | 536 | &bser_capabilities)) { |
|
526 | 537 | return NULL; |
|
527 | 538 | } |
|
528 | 539 | |
|
529 | 540 | if (!bser_init(&bser, bser_version, bser_capabilities)) { |
|
530 | 541 | return PyErr_NoMemory(); |
|
531 | 542 | } |
|
532 | 543 | |
|
533 | 544 | if (!bser_recursive(&bser, val)) { |
|
534 | 545 | bser_dtor(&bser); |
|
535 | 546 | if (errno == ENOMEM) { |
|
536 | 547 | return PyErr_NoMemory(); |
|
537 | 548 | } |
|
538 | 549 | // otherwise, we've already set the error to something reasonable |
|
539 | 550 | return NULL; |
|
540 | 551 | } |
|
541 | 552 | |
|
542 | 553 | // Now fill in the overall length |
|
543 | 554 | if (bser_version == 1) { |
|
544 | 555 | len = bser.wpos - (sizeof(EMPTY_HEADER) - 1); |
|
545 | 556 | memcpy(bser.buf + 3, &len, sizeof(len)); |
|
546 | 557 | } else { |
|
547 | 558 | len = bser.wpos - (sizeof(EMPTY_HEADER_V2) - 1); |
|
548 | 559 | // The BSER capabilities block comes before the PDU length |
|
549 | 560 | memcpy(bser.buf + 2, &bser_capabilities, sizeof(bser_capabilities)); |
|
550 | 561 | memcpy(bser.buf + 7, &len, sizeof(len)); |
|
551 | 562 | } |
|
552 | 563 | |
|
553 | 564 | res = PyBytes_FromStringAndSize(bser.buf, bser.wpos); |
|
554 | 565 | bser_dtor(&bser); |
|
555 | 566 | |
|
556 | 567 | return res; |
|
557 | 568 | } |
|
558 | 569 | |
|
559 | 570 | int bunser_int(const char** ptr, const char* end, int64_t* val) { |
|
560 | 571 | int needed; |
|
561 | 572 | const char* buf = *ptr; |
|
562 | 573 | int8_t i8; |
|
563 | 574 | int16_t i16; |
|
564 | 575 | int32_t i32; |
|
565 | 576 | int64_t i64; |
|
566 | 577 | |
|
567 | 578 | switch (buf[0]) { |
|
568 | 579 | case BSER_INT8: |
|
569 | 580 | needed = 2; |
|
570 | 581 | break; |
|
571 | 582 | case BSER_INT16: |
|
572 | 583 | needed = 3; |
|
573 | 584 | break; |
|
574 | 585 | case BSER_INT32: |
|
575 | 586 | needed = 5; |
|
576 | 587 | break; |
|
577 | 588 | case BSER_INT64: |
|
578 | 589 | needed = 9; |
|
579 | 590 | break; |
|
580 | 591 | default: |
|
581 | 592 | PyErr_Format( |
|
582 | 593 | PyExc_ValueError, "invalid bser int encoding 0x%02x", buf[0]); |
|
583 | 594 | return 0; |
|
584 | 595 | } |
|
585 | 596 | if (end - buf < needed) { |
|
586 | 597 | PyErr_SetString(PyExc_ValueError, "input buffer to small for int encoding"); |
|
587 | 598 | return 0; |
|
588 | 599 | } |
|
589 | 600 | *ptr = buf + needed; |
|
590 | 601 | switch (buf[0]) { |
|
591 | 602 | case BSER_INT8: |
|
592 | 603 | memcpy(&i8, buf + 1, sizeof(i8)); |
|
593 | 604 | *val = i8; |
|
594 | 605 | return 1; |
|
595 | 606 | case BSER_INT16: |
|
596 | 607 | memcpy(&i16, buf + 1, sizeof(i16)); |
|
597 | 608 | *val = i16; |
|
598 | 609 | return 1; |
|
599 | 610 | case BSER_INT32: |
|
600 | 611 | memcpy(&i32, buf + 1, sizeof(i32)); |
|
601 | 612 | *val = i32; |
|
602 | 613 | return 1; |
|
603 | 614 | case BSER_INT64: |
|
604 | 615 | memcpy(&i64, buf + 1, sizeof(i64)); |
|
605 | 616 | *val = i64; |
|
606 | 617 | return 1; |
|
607 | 618 | default: |
|
608 | 619 | return 0; |
|
609 | 620 | } |
|
610 | 621 | } |
|
611 | 622 | |
|
612 | 623 | static int bunser_bytestring( |
|
613 | 624 | const char** ptr, |
|
614 | 625 | const char* end, |
|
615 | 626 | const char** start, |
|
616 | 627 | int64_t* len) { |
|
617 | 628 | const char* buf = *ptr; |
|
618 | 629 | |
|
619 | 630 | // skip string marker |
|
620 | 631 | buf++; |
|
621 | 632 | if (!bunser_int(&buf, end, len)) { |
|
622 | 633 | return 0; |
|
623 | 634 | } |
|
624 | 635 | |
|
625 | 636 | if (buf + *len > end) { |
|
626 | 637 | PyErr_Format(PyExc_ValueError, "invalid string length in bser data"); |
|
627 | 638 | return 0; |
|
628 | 639 | } |
|
629 | 640 | |
|
630 | 641 | *ptr = buf + *len; |
|
631 | 642 | *start = buf; |
|
632 | 643 | return 1; |
|
633 | 644 | } |
|
634 | 645 | |
|
635 | 646 | static PyObject* |
|
636 | 647 | bunser_array(const char** ptr, const char* end, const unser_ctx_t* ctx) { |
|
637 | 648 | const char* buf = *ptr; |
|
638 | 649 | int64_t nitems, i; |
|
639 | 650 | int mutable = ctx->mutable; |
|
640 | 651 | PyObject* res; |
|
641 | 652 | |
|
642 | 653 | // skip array header |
|
643 | 654 | buf++; |
|
644 | 655 | if (!bunser_int(&buf, end, &nitems)) { |
|
645 | 656 | return 0; |
|
646 | 657 | } |
|
647 | 658 | *ptr = buf; |
|
648 | 659 | |
|
649 | 660 | if (nitems > LONG_MAX) { |
|
650 | 661 | PyErr_Format(PyExc_ValueError, "too many items for python array"); |
|
651 | 662 | return NULL; |
|
652 | 663 | } |
|
653 | 664 | |
|
654 | 665 | if (mutable) { |
|
655 | 666 | res = PyList_New((Py_ssize_t)nitems); |
|
656 | 667 | } else { |
|
657 | 668 | res = PyTuple_New((Py_ssize_t)nitems); |
|
658 | 669 | } |
|
659 | 670 | |
|
660 | 671 | for (i = 0; i < nitems; i++) { |
|
661 | 672 | PyObject* ele = bser_loads_recursive(ptr, end, ctx); |
|
662 | 673 | |
|
663 | 674 | if (!ele) { |
|
664 | 675 | Py_DECREF(res); |
|
665 | 676 | return NULL; |
|
666 | 677 | } |
|
667 | 678 | |
|
668 | 679 | if (mutable) { |
|
669 | 680 | PyList_SET_ITEM(res, i, ele); |
|
670 | 681 | } else { |
|
671 | 682 | PyTuple_SET_ITEM(res, i, ele); |
|
672 | 683 | } |
|
673 | 684 | // DECREF(ele) not required as SET_ITEM steals the ref |
|
674 | 685 | } |
|
675 | 686 | |
|
676 | 687 | return res; |
|
677 | 688 | } |
|
678 | 689 | |
|
679 | 690 | static PyObject* |
|
680 | 691 | bunser_object(const char** ptr, const char* end, const unser_ctx_t* ctx) { |
|
681 | 692 | const char* buf = *ptr; |
|
682 | 693 | int64_t nitems, i; |
|
683 | 694 | int mutable = ctx->mutable; |
|
684 | 695 | PyObject* res; |
|
685 | 696 | bserObject* obj; |
|
686 | 697 | |
|
687 | 698 | // skip array header |
|
688 | 699 | buf++; |
|
689 | 700 | if (!bunser_int(&buf, end, &nitems)) { |
|
690 | 701 | return 0; |
|
691 | 702 | } |
|
692 | 703 | *ptr = buf; |
|
693 | 704 | |
|
694 | 705 | if (mutable) { |
|
695 | 706 | res = PyDict_New(); |
|
696 | 707 | } else { |
|
697 | 708 | obj = PyObject_New(bserObject, &bserObjectType); |
|
698 | 709 | obj->keys = PyTuple_New((Py_ssize_t)nitems); |
|
699 | 710 | obj->values = PyTuple_New((Py_ssize_t)nitems); |
|
700 | 711 | res = (PyObject*)obj; |
|
701 | 712 | } |
|
702 | 713 | |
|
703 | 714 | for (i = 0; i < nitems; i++) { |
|
704 | 715 | const char* keystr; |
|
705 | 716 | int64_t keylen; |
|
706 | 717 | PyObject* key; |
|
707 | 718 | PyObject* ele; |
|
708 | 719 | |
|
709 | 720 | if (!bunser_bytestring(ptr, end, &keystr, &keylen)) { |
|
710 | 721 | Py_DECREF(res); |
|
711 | 722 | return NULL; |
|
712 | 723 | } |
|
713 | 724 | |
|
714 | 725 | if (keylen > LONG_MAX) { |
|
715 | 726 | PyErr_Format(PyExc_ValueError, "string too big for python"); |
|
716 | 727 | Py_DECREF(res); |
|
717 | 728 | return NULL; |
|
718 | 729 | } |
|
719 | 730 | |
|
720 | 731 | if (mutable) { |
|
721 | 732 | // This will interpret the key as UTF-8. |
|
722 | 733 | key = PyUnicode_FromStringAndSize(keystr, (Py_ssize_t)keylen); |
|
723 | 734 | } else { |
|
724 | 735 | // For immutable objects we'll manage key lookups, so we can avoid going |
|
725 | 736 | // through the Unicode APIs. This avoids a potentially expensive and |
|
726 | 737 | // definitely unnecessary conversion to UTF-16 and back for Python 2. |
|
727 | 738 | // TODO: On Python 3 the Unicode APIs are smarter: we might be able to use |
|
728 | 739 | // Unicode keys there without an appreciable performance loss. |
|
729 | 740 | key = PyBytes_FromStringAndSize(keystr, (Py_ssize_t)keylen); |
|
730 | 741 | } |
|
731 | 742 | |
|
732 | 743 | if (!key) { |
|
733 | 744 | Py_DECREF(res); |
|
734 | 745 | return NULL; |
|
735 | 746 | } |
|
736 | 747 | |
|
737 | 748 | ele = bser_loads_recursive(ptr, end, ctx); |
|
738 | 749 | |
|
739 | 750 | if (!ele) { |
|
740 | 751 | Py_DECREF(key); |
|
741 | 752 | Py_DECREF(res); |
|
742 | 753 | return NULL; |
|
743 | 754 | } |
|
744 | 755 | |
|
745 | 756 | if (mutable) { |
|
746 | 757 | PyDict_SetItem(res, key, ele); |
|
747 | 758 | Py_DECREF(key); |
|
748 | 759 | Py_DECREF(ele); |
|
749 | 760 | } else { |
|
750 | 761 | /* PyTuple_SET_ITEM steals ele, key */ |
|
751 | 762 | PyTuple_SET_ITEM(obj->values, i, ele); |
|
752 | 763 | PyTuple_SET_ITEM(obj->keys, i, key); |
|
753 | 764 | } |
|
754 | 765 | } |
|
755 | 766 | |
|
756 | 767 | return res; |
|
757 | 768 | } |
|
758 | 769 | |
|
759 | 770 | static PyObject* |
|
760 | 771 | bunser_template(const char** ptr, const char* end, const unser_ctx_t* ctx) { |
|
761 | 772 | const char* buf = *ptr; |
|
762 | 773 | int64_t nitems, i; |
|
763 | 774 | int mutable = ctx->mutable; |
|
764 | 775 | PyObject* arrval; |
|
765 | 776 | PyObject* keys; |
|
766 | 777 | Py_ssize_t numkeys, keyidx; |
|
767 | 778 | unser_ctx_t keys_ctx = {0}; |
|
768 | 779 | if (mutable) { |
|
769 | 780 | keys_ctx.mutable = 1; |
|
770 | 781 | // Decode keys as UTF-8 in this case. |
|
771 | 782 | keys_ctx.value_encoding = "utf-8"; |
|
772 | 783 | keys_ctx.value_errors = "strict"; |
|
773 | 784 | } else { |
|
774 | 785 | // Treat keys as bytestrings in this case -- we'll do Unicode conversions at |
|
775 | 786 | // lookup time. |
|
776 | 787 | } |
|
777 | 788 | |
|
778 | 789 | if (buf[1] != BSER_ARRAY) { |
|
779 | 790 | PyErr_Format(PyExc_ValueError, "Expect ARRAY to follow TEMPLATE"); |
|
780 | 791 | return NULL; |
|
781 | 792 | } |
|
782 | 793 | |
|
783 | 794 | // skip header |
|
784 | 795 | buf++; |
|
785 | 796 | *ptr = buf; |
|
786 | 797 | |
|
787 | 798 | // Load template keys. |
|
788 | 799 | // For keys we don't want to do any decoding right now. |
|
789 | 800 | keys = bunser_array(ptr, end, &keys_ctx); |
|
790 | 801 | if (!keys) { |
|
791 | 802 | return NULL; |
|
792 | 803 | } |
|
793 | 804 | |
|
794 | 805 | numkeys = PySequence_Length(keys); |
|
795 | 806 | |
|
796 | 807 | // Load number of array elements |
|
797 | 808 | if (!bunser_int(ptr, end, &nitems)) { |
|
798 | 809 | Py_DECREF(keys); |
|
799 | 810 | return 0; |
|
800 | 811 | } |
|
801 | 812 | |
|
802 | 813 | if (nitems > LONG_MAX) { |
|
803 | 814 | PyErr_Format(PyExc_ValueError, "Too many items for python"); |
|
804 | 815 | Py_DECREF(keys); |
|
805 | 816 | return NULL; |
|
806 | 817 | } |
|
807 | 818 | |
|
808 | 819 | arrval = PyList_New((Py_ssize_t)nitems); |
|
809 | 820 | if (!arrval) { |
|
810 | 821 | Py_DECREF(keys); |
|
811 | 822 | return NULL; |
|
812 | 823 | } |
|
813 | 824 | |
|
814 | 825 | for (i = 0; i < nitems; i++) { |
|
815 | 826 | PyObject* dict = NULL; |
|
816 | 827 | bserObject* obj = NULL; |
|
817 | 828 | |
|
818 | 829 | if (mutable) { |
|
819 | 830 | dict = PyDict_New(); |
|
820 | 831 | } else { |
|
821 | 832 | obj = PyObject_New(bserObject, &bserObjectType); |
|
822 | 833 | if (obj) { |
|
823 | 834 | obj->keys = keys; |
|
824 | 835 | Py_INCREF(obj->keys); |
|
825 | 836 | obj->values = PyTuple_New(numkeys); |
|
826 | 837 | } |
|
827 | 838 | dict = (PyObject*)obj; |
|
828 | 839 | } |
|
829 | 840 | if (!dict) { |
|
830 | 841 | fail: |
|
831 | 842 | Py_DECREF(keys); |
|
832 | 843 | Py_DECREF(arrval); |
|
833 | 844 | return NULL; |
|
834 | 845 | } |
|
835 | 846 | |
|
836 | 847 | for (keyidx = 0; keyidx < numkeys; keyidx++) { |
|
837 | 848 | PyObject* key; |
|
838 | 849 | PyObject* ele; |
|
839 | 850 | |
|
840 | 851 | if (**ptr == BSER_SKIP) { |
|
841 | 852 | *ptr = *ptr + 1; |
|
842 | 853 | ele = Py_None; |
|
843 | 854 | Py_INCREF(ele); |
|
844 | 855 | } else { |
|
845 | 856 | ele = bser_loads_recursive(ptr, end, ctx); |
|
846 | 857 | } |
|
847 | 858 | |
|
848 | 859 | if (!ele) { |
|
849 | 860 | goto fail; |
|
850 | 861 | } |
|
851 | 862 | |
|
852 | 863 | if (mutable) { |
|
853 | 864 | key = PyList_GET_ITEM(keys, keyidx); |
|
854 | 865 | PyDict_SetItem(dict, key, ele); |
|
855 | 866 | Py_DECREF(ele); |
|
856 | 867 | } else { |
|
857 | 868 | PyTuple_SET_ITEM(obj->values, keyidx, ele); |
|
858 | 869 | // DECREF(ele) not required as SET_ITEM steals the ref |
|
859 | 870 | } |
|
860 | 871 | } |
|
861 | 872 | |
|
862 | 873 | PyList_SET_ITEM(arrval, i, dict); |
|
863 | 874 | // DECREF(obj) not required as SET_ITEM steals the ref |
|
864 | 875 | } |
|
865 | 876 | |
|
866 | 877 | Py_DECREF(keys); |
|
867 | 878 | |
|
868 | 879 | return arrval; |
|
869 | 880 | } |
|
870 | 881 | |
|
871 | 882 | static PyObject* bser_loads_recursive( |
|
872 | 883 | const char** ptr, |
|
873 | 884 | const char* end, |
|
874 | 885 | const unser_ctx_t* ctx) { |
|
875 | 886 | const char* buf = *ptr; |
|
876 | 887 | |
|
877 | 888 | switch (buf[0]) { |
|
878 | 889 | case BSER_INT8: |
|
879 | 890 | case BSER_INT16: |
|
880 | 891 | case BSER_INT32: |
|
881 | 892 | case BSER_INT64: { |
|
882 | 893 | int64_t ival; |
|
883 | 894 | if (!bunser_int(ptr, end, &ival)) { |
|
884 | 895 | return NULL; |
|
885 | 896 | } |
|
886 | 897 | // Python 3 has one integer type. |
|
887 | 898 | #if PY_MAJOR_VERSION >= 3 |
|
888 | 899 | return PyLong_FromLongLong(ival); |
|
889 | 900 | #else |
|
890 | 901 | if (ival < LONG_MIN || ival > LONG_MAX) { |
|
891 | 902 | return PyLong_FromLongLong(ival); |
|
892 | 903 | } |
|
893 | 904 | return PyInt_FromSsize_t(Py_SAFE_DOWNCAST(ival, int64_t, Py_ssize_t)); |
|
894 | 905 | #endif // PY_MAJOR_VERSION >= 3 |
|
895 | 906 | } |
|
896 | 907 | |
|
897 | 908 | case BSER_REAL: { |
|
898 | 909 | double dval; |
|
899 | 910 | memcpy(&dval, buf + 1, sizeof(dval)); |
|
900 | 911 | *ptr = buf + 1 + sizeof(double); |
|
901 | 912 | return PyFloat_FromDouble(dval); |
|
902 | 913 | } |
|
903 | 914 | |
|
904 | 915 | case BSER_TRUE: |
|
905 | 916 | *ptr = buf + 1; |
|
906 | 917 | Py_INCREF(Py_True); |
|
907 | 918 | return Py_True; |
|
908 | 919 | |
|
909 | 920 | case BSER_FALSE: |
|
910 | 921 | *ptr = buf + 1; |
|
911 | 922 | Py_INCREF(Py_False); |
|
912 | 923 | return Py_False; |
|
913 | 924 | |
|
914 | 925 | case BSER_NULL: |
|
915 | 926 | *ptr = buf + 1; |
|
916 | 927 | Py_INCREF(Py_None); |
|
917 | 928 | return Py_None; |
|
918 | 929 | |
|
919 | 930 | case BSER_BYTESTRING: { |
|
920 | 931 | const char* start; |
|
921 | 932 | int64_t len; |
|
922 | 933 | |
|
923 | 934 | if (!bunser_bytestring(ptr, end, &start, &len)) { |
|
924 | 935 | return NULL; |
|
925 | 936 | } |
|
926 | 937 | |
|
927 | 938 | if (len > LONG_MAX) { |
|
928 | 939 | PyErr_Format(PyExc_ValueError, "string too long for python"); |
|
929 | 940 | return NULL; |
|
930 | 941 | } |
|
931 | 942 | |
|
932 | 943 | if (ctx->value_encoding != NULL) { |
|
933 | 944 | return PyUnicode_Decode( |
|
934 | 945 | start, (long)len, ctx->value_encoding, ctx->value_errors); |
|
935 | 946 | } else { |
|
936 | 947 | return PyBytes_FromStringAndSize(start, (long)len); |
|
937 | 948 | } |
|
938 | 949 | } |
|
939 | 950 | |
|
940 | 951 | case BSER_UTF8STRING: { |
|
941 | 952 | const char* start; |
|
942 | 953 | int64_t len; |
|
943 | 954 | |
|
944 | 955 | if (!bunser_bytestring(ptr, end, &start, &len)) { |
|
945 | 956 | return NULL; |
|
946 | 957 | } |
|
947 | 958 | |
|
948 | 959 | if (len > LONG_MAX) { |
|
949 | 960 | PyErr_Format(PyExc_ValueError, "string too long for python"); |
|
950 | 961 | return NULL; |
|
951 | 962 | } |
|
952 | 963 | |
|
953 | 964 | return PyUnicode_Decode(start, (long)len, "utf-8", "strict"); |
|
954 | 965 | } |
|
955 | 966 | |
|
956 | 967 | case BSER_ARRAY: |
|
957 | 968 | return bunser_array(ptr, end, ctx); |
|
958 | 969 | |
|
959 | 970 | case BSER_OBJECT: |
|
960 | 971 | return bunser_object(ptr, end, ctx); |
|
961 | 972 | |
|
962 | 973 | case BSER_TEMPLATE: |
|
963 | 974 | return bunser_template(ptr, end, ctx); |
|
964 | 975 | |
|
965 | 976 | default: |
|
966 | 977 | PyErr_Format(PyExc_ValueError, "unhandled bser opcode 0x%02x", buf[0]); |
|
967 | 978 | } |
|
968 | 979 | |
|
969 | 980 | return NULL; |
|
970 | 981 | } |
|
971 | 982 | |
|
972 | 983 | static int _pdu_info_helper( |
|
973 | 984 | const char* data, |
|
974 | 985 | const char* end, |
|
975 | 986 | uint32_t* bser_version_out, |
|
976 | 987 | uint32_t* bser_capabilities_out, |
|
977 | 988 | int64_t* expected_len_out, |
|
978 | 989 | off_t* position_out) { |
|
979 | 990 | uint32_t bser_version; |
|
980 | 991 | uint32_t bser_capabilities = 0; |
|
981 | 992 | int64_t expected_len; |
|
982 | 993 | |
|
983 | 994 | const char* start; |
|
984 | 995 | start = data; |
|
985 | 996 | // Validate the header and length |
|
986 | 997 | if (memcmp(data, EMPTY_HEADER, 2) == 0) { |
|
987 | 998 | bser_version = 1; |
|
988 | 999 | } else if (memcmp(data, EMPTY_HEADER_V2, 2) == 0) { |
|
989 | 1000 | bser_version = 2; |
|
990 | 1001 | } else { |
|
991 | 1002 | PyErr_SetString(PyExc_ValueError, "invalid bser header"); |
|
992 | 1003 | return 0; |
|
993 | 1004 | } |
|
994 | 1005 | |
|
995 | 1006 | data += 2; |
|
996 | 1007 | |
|
997 | 1008 | if (bser_version == 2) { |
|
998 | 1009 | // Expect an integer telling us what capabilities are supported by the |
|
999 | 1010 | // remote server (currently unused). |
|
1000 | 1011 | if (!memcpy(&bser_capabilities, &data, sizeof(bser_capabilities))) { |
|
1001 | 1012 | return 0; |
|
1002 | 1013 | } |
|
1003 | 1014 | data += sizeof(bser_capabilities); |
|
1004 | 1015 | } |
|
1005 | 1016 | |
|
1006 | 1017 | // Expect an integer telling us how big the rest of the data |
|
1007 | 1018 | // should be |
|
1008 | 1019 | if (!bunser_int(&data, end, &expected_len)) { |
|
1009 | 1020 | return 0; |
|
1010 | 1021 | } |
|
1011 | 1022 | |
|
1012 | 1023 | *bser_version_out = bser_version; |
|
1013 | 1024 | *bser_capabilities_out = (uint32_t)bser_capabilities; |
|
1014 | 1025 | *expected_len_out = expected_len; |
|
1015 | 1026 | *position_out = (off_t)(data - start); |
|
1016 | 1027 | return 1; |
|
1017 | 1028 | } |
|
1018 | 1029 | |
|
1019 | 1030 | // This function parses the PDU header and provides info about the packet |
|
1020 | 1031 | // Returns false if unsuccessful |
|
1021 | 1032 | static int pdu_info_helper( |
|
1022 | 1033 | PyObject* self, |
|
1023 | 1034 | PyObject* args, |
|
1024 | 1035 | uint32_t* bser_version_out, |
|
1025 | 1036 | uint32_t* bser_capabilities_out, |
|
1026 | 1037 | int64_t* total_len_out) { |
|
1027 | 1038 | const char* start = NULL; |
|
1028 | 1039 | const char* data = NULL; |
|
1029 | 1040 | int datalen = 0; |
|
1030 | 1041 | const char* end; |
|
1031 | 1042 | int64_t expected_len; |
|
1032 | 1043 | off_t position; |
|
1033 | 1044 | |
|
1034 | 1045 | if (!PyArg_ParseTuple(args, "s#", &start, &datalen)) { |
|
1035 | 1046 | return 0; |
|
1036 | 1047 | } |
|
1037 | 1048 | data = start; |
|
1038 | 1049 | end = data + datalen; |
|
1039 | 1050 | |
|
1040 | 1051 | if (!_pdu_info_helper( |
|
1041 | 1052 | data, |
|
1042 | 1053 | end, |
|
1043 | 1054 | bser_version_out, |
|
1044 | 1055 | bser_capabilities_out, |
|
1045 | 1056 | &expected_len, |
|
1046 | 1057 | &position)) { |
|
1047 | 1058 | return 0; |
|
1048 | 1059 | } |
|
1049 | 1060 | *total_len_out = (int64_t)(expected_len + position); |
|
1050 | 1061 | return 1; |
|
1051 | 1062 | } |
|
1052 | 1063 | |
|
1053 | 1064 | // Expected use case is to read a packet from the socket and then call |
|
1054 | 1065 | // bser.pdu_info on the packet. It returns the BSER version, BSER capabilities, |
|
1055 | 1066 | // and the total length of the entire response that the peer is sending, |
|
1056 | 1067 | // including the bytes already received. This allows the client to compute the |
|
1057 | 1068 | // data size it needs to read before it can decode the data. |
|
1058 | 1069 | static PyObject* bser_pdu_info(PyObject* self, PyObject* args) { |
|
1059 | 1070 | uint32_t version, capabilities; |
|
1060 | 1071 | int64_t total_len; |
|
1061 | 1072 | if (!pdu_info_helper(self, args, &version, &capabilities, &total_len)) { |
|
1062 | 1073 | return NULL; |
|
1063 | 1074 | } |
|
1064 | 1075 | return Py_BuildValue("kkL", version, capabilities, total_len); |
|
1065 | 1076 | } |
|
1066 | 1077 | |
|
1067 | 1078 | static PyObject* bser_pdu_len(PyObject* self, PyObject* args) { |
|
1068 | 1079 | uint32_t version, capabilities; |
|
1069 | 1080 | int64_t total_len; |
|
1070 | 1081 | if (!pdu_info_helper(self, args, &version, &capabilities, &total_len)) { |
|
1071 | 1082 | return NULL; |
|
1072 | 1083 | } |
|
1073 | 1084 | return Py_BuildValue("L", total_len); |
|
1074 | 1085 | } |
|
1075 | 1086 | |
|
1076 | 1087 | static PyObject* bser_loads(PyObject* self, PyObject* args, PyObject* kw) { |
|
1077 | 1088 | const char* data = NULL; |
|
1078 | 1089 | int datalen = 0; |
|
1079 | 1090 | const char* start; |
|
1080 | 1091 | const char* end; |
|
1081 | 1092 | int64_t expected_len; |
|
1082 | 1093 | off_t position; |
|
1083 | 1094 | PyObject* mutable_obj = NULL; |
|
1084 | 1095 | const char* value_encoding = NULL; |
|
1085 | 1096 | const char* value_errors = NULL; |
|
1086 | 1097 | unser_ctx_t ctx = {1, 0}; |
|
1087 | 1098 | |
|
1088 | 1099 | static char* kw_list[] = { |
|
1089 | 1100 | "buf", "mutable", "value_encoding", "value_errors", NULL}; |
|
1090 | 1101 | |
|
1091 | 1102 | if (!PyArg_ParseTupleAndKeywords( |
|
1092 | 1103 | args, |
|
1093 | 1104 | kw, |
|
1094 | 1105 | "s#|Ozz:loads", |
|
1095 | 1106 | kw_list, |
|
1096 | 1107 | &start, |
|
1097 | 1108 | &datalen, |
|
1098 | 1109 | &mutable_obj, |
|
1099 | 1110 | &value_encoding, |
|
1100 | 1111 | &value_errors)) { |
|
1101 | 1112 | return NULL; |
|
1102 | 1113 | } |
|
1103 | 1114 | |
|
1104 | 1115 | if (mutable_obj) { |
|
1105 | 1116 | ctx.mutable = PyObject_IsTrue(mutable_obj) > 0 ? 1 : 0; |
|
1106 | 1117 | } |
|
1107 | 1118 | ctx.value_encoding = value_encoding; |
|
1108 | 1119 | if (value_encoding == NULL) { |
|
1109 | 1120 | ctx.value_errors = NULL; |
|
1110 | 1121 | } else if (value_errors == NULL) { |
|
1111 | 1122 | ctx.value_errors = "strict"; |
|
1112 | 1123 | } else { |
|
1113 | 1124 | ctx.value_errors = value_errors; |
|
1114 | 1125 | } |
|
1115 | 1126 | data = start; |
|
1116 | 1127 | end = data + datalen; |
|
1117 | 1128 | |
|
1118 | 1129 | if (!_pdu_info_helper( |
|
1119 | 1130 | data, |
|
1120 | 1131 | end, |
|
1121 | 1132 | &ctx.bser_version, |
|
1122 | 1133 | &ctx.bser_capabilities, |
|
1123 | 1134 | &expected_len, |
|
1124 | 1135 | &position)) { |
|
1125 | 1136 | return NULL; |
|
1126 | 1137 | } |
|
1127 | 1138 | |
|
1128 | 1139 | data = start + position; |
|
1129 | 1140 | // Verify |
|
1130 | 1141 | if (expected_len + data != end) { |
|
1131 | 1142 | PyErr_SetString(PyExc_ValueError, "bser data len != header len"); |
|
1132 | 1143 | return NULL; |
|
1133 | 1144 | } |
|
1134 | 1145 | |
|
1135 | 1146 | return bser_loads_recursive(&data, end, &ctx); |
|
1136 | 1147 | } |
|
1137 | 1148 | |
|
1138 | 1149 | static PyObject* bser_load(PyObject* self, PyObject* args, PyObject* kw) { |
|
1139 | 1150 | PyObject *load, *string; |
|
1140 | 1151 | PyObject* fp = NULL; |
|
1141 | 1152 | PyObject* mutable_obj = NULL; |
|
1142 | 1153 | const char* value_encoding = NULL; |
|
1143 | 1154 | const char* value_errors = NULL; |
|
1144 | 1155 | |
|
1145 | 1156 | static char* kw_list[] = { |
|
1146 | 1157 | "fp", "mutable", "value_encoding", "value_errors", NULL}; |
|
1147 | 1158 | |
|
1148 | 1159 | if (!PyArg_ParseTupleAndKeywords( |
|
1149 | 1160 | args, |
|
1150 | 1161 | kw, |
|
1151 | 1162 | "OOzz:load", |
|
1152 | 1163 | kw_list, |
|
1153 | 1164 | &fp, |
|
1154 | 1165 | &mutable_obj, |
|
1155 | 1166 | &value_encoding, |
|
1156 | 1167 | &value_errors)) { |
|
1157 | 1168 | return NULL; |
|
1158 | 1169 | } |
|
1159 | 1170 | |
|
1160 | 1171 | load = PyImport_ImportModule("pywatchman.load"); |
|
1161 | 1172 | if (load == NULL) { |
|
1162 | 1173 | return NULL; |
|
1163 | 1174 | } |
|
1164 | 1175 | string = PyObject_CallMethod( |
|
1165 | 1176 | load, "load", "OOzz", fp, mutable_obj, value_encoding, value_errors); |
|
1166 | 1177 | Py_DECREF(load); |
|
1167 | 1178 | return string; |
|
1168 | 1179 | } |
|
1169 | 1180 | |
|
1170 | 1181 | // clang-format off |
|
1171 | 1182 | static PyMethodDef bser_methods[] = { |
|
1172 | 1183 | {"loads", (PyCFunction)bser_loads, METH_VARARGS | METH_KEYWORDS, |
|
1173 | 1184 | "Deserialize string."}, |
|
1174 | 1185 | {"load", (PyCFunction)bser_load, METH_VARARGS | METH_KEYWORDS, |
|
1175 | 1186 | "Deserialize a file object"}, |
|
1176 | 1187 | {"pdu_info", (PyCFunction)bser_pdu_info, METH_VARARGS, |
|
1177 | 1188 | "Extract PDU information."}, |
|
1178 | 1189 | {"pdu_len", (PyCFunction)bser_pdu_len, METH_VARARGS, |
|
1179 | 1190 | "Extract total PDU length."}, |
|
1180 | 1191 | {"dumps", (PyCFunction)bser_dumps, METH_VARARGS | METH_KEYWORDS, |
|
1181 | 1192 | "Serialize string."}, |
|
1182 | 1193 | {NULL, NULL, 0, NULL} |
|
1183 | 1194 | }; |
|
1184 | 1195 | |
|
1185 | 1196 | #if PY_MAJOR_VERSION >= 3 |
|
1186 | 1197 | static struct PyModuleDef bser_module = { |
|
1187 | 1198 | PyModuleDef_HEAD_INIT, |
|
1188 | 1199 | "bser", |
|
1189 | 1200 | "Efficient encoding and decoding of BSER.", |
|
1190 | 1201 | -1, |
|
1191 | 1202 | bser_methods |
|
1192 | 1203 | }; |
|
1193 | 1204 | // clang-format on |
|
1194 | 1205 | |
|
1195 | 1206 | PyMODINIT_FUNC PyInit_bser(void) { |
|
1196 | 1207 | PyObject* mod; |
|
1197 | 1208 | |
|
1198 | 1209 | mod = PyModule_Create(&bser_module); |
|
1199 | 1210 | PyType_Ready(&bserObjectType); |
|
1200 | 1211 | |
|
1201 | 1212 | return mod; |
|
1202 | 1213 | } |
|
1203 | 1214 | #else |
|
1204 | 1215 | |
|
1205 | 1216 | PyMODINIT_FUNC initbser(void) { |
|
1206 | 1217 | (void)Py_InitModule("bser", bser_methods); |
|
1207 | 1218 | PyType_Ready(&bserObjectType); |
|
1208 | 1219 | } |
|
1209 | 1220 | #endif // PY_MAJOR_VERSION >= 3 |
|
1210 | 1221 | |
|
1211 | 1222 | /* vim:ts=2:sw=2:et: |
|
1212 | 1223 | */ |
General Comments 0
You need to be logged in to leave comments.
Login now