]>
Commit | Line | Data |
---|---|---|
f38b84ea | 1 | /* |
7d23a63a | 2 | * Copyright (c) 2009, 2010 Nicira Networks. |
f38b84ea BP |
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 <config.h> | |
18 | ||
19 | #include "json.h" | |
20 | ||
21 | #include <assert.h> | |
22 | #include <ctype.h> | |
23 | #include <errno.h> | |
24 | #include <float.h> | |
25 | #include <limits.h> | |
26 | #include <math.h> | |
27 | #include <string.h> | |
28 | ||
29 | #include "dynamic-string.h" | |
30 | #include "hash.h" | |
31 | #include "shash.h" | |
32 | #include "unicode.h" | |
33 | #include "util.h" | |
34 | ||
35 | /* The type of a JSON token. */ | |
36 | enum json_token_type { | |
37 | T_EOF = 0, | |
38 | T_BEGIN_ARRAY = '[', | |
39 | T_END_ARRAY = ']', | |
40 | T_BEGIN_OBJECT = '{', | |
41 | T_END_OBJECT = '}', | |
42 | T_NAME_SEPARATOR = ':', | |
43 | T_VALUE_SEPARATOR = ',', | |
44 | T_FALSE = UCHAR_MAX + 1, | |
45 | T_NULL, | |
46 | T_TRUE, | |
47 | T_INTEGER, | |
48 | T_REAL, | |
49 | T_STRING | |
50 | }; | |
51 | ||
52 | /* A JSON token. | |
53 | * | |
54 | * RFC 4627 doesn't define a lexical structure for JSON but I believe this to | |
55 | * be compliant with the standard. | |
56 | */ | |
57 | struct json_token { | |
58 | enum json_token_type type; | |
59 | union { | |
60 | double real; | |
61 | long long int integer; | |
62 | const char *string; | |
63 | } u; | |
64 | }; | |
65 | ||
66 | enum json_lex_state { | |
67 | JSON_LEX_START, /* Not inside a token. */ | |
68 | JSON_LEX_NUMBER, /* Reading a number. */ | |
69 | JSON_LEX_KEYWORD, /* Reading a keyword. */ | |
70 | JSON_LEX_STRING, /* Reading a quoted string. */ | |
71 | JSON_LEX_ESCAPE /* In a quoted string just after a "\". */ | |
72 | }; | |
73 | ||
74 | enum json_parse_state { | |
75 | JSON_PARSE_START, /* Beginning of input. */ | |
76 | JSON_PARSE_END, /* End of input. */ | |
77 | ||
78 | /* Objects. */ | |
79 | JSON_PARSE_OBJECT_INIT, /* Expecting '}' or an object name. */ | |
80 | JSON_PARSE_OBJECT_NAME, /* Expecting an object name. */ | |
81 | JSON_PARSE_OBJECT_COLON, /* Expecting ':'. */ | |
82 | JSON_PARSE_OBJECT_VALUE, /* Expecting an object value. */ | |
83 | JSON_PARSE_OBJECT_NEXT, /* Expecting ',' or '}'. */ | |
84 | ||
85 | /* Arrays. */ | |
86 | JSON_PARSE_ARRAY_INIT, /* Expecting ']' or a value. */ | |
87 | JSON_PARSE_ARRAY_VALUE, /* Expecting a value. */ | |
88 | JSON_PARSE_ARRAY_NEXT /* Expecting ',' or ']'. */ | |
89 | }; | |
90 | ||
91 | struct json_parser_node { | |
92 | struct json *json; | |
93 | }; | |
94 | ||
95 | /* A JSON parser. */ | |
96 | struct json_parser { | |
97 | int flags; | |
98 | ||
99 | /* Lexical analysis. */ | |
100 | enum json_lex_state lex_state; | |
101 | struct ds buffer; /* Buffer for accumulating token text. */ | |
0bdf342a BP |
102 | int line_number; |
103 | int column_number; | |
104 | int byte_number; | |
f38b84ea BP |
105 | |
106 | /* Parsing. */ | |
107 | enum json_parse_state parse_state; | |
108 | #define JSON_MAX_HEIGHT 1000 | |
109 | struct json_parser_node *stack; | |
110 | size_t height, allocated_height; | |
111 | char *member_name; | |
112 | ||
113 | /* Parse status. */ | |
114 | bool done; | |
115 | char *error; /* Error message, if any, null if none yet. */ | |
116 | }; | |
117 | ||
118 | static struct json *json_create(enum json_type type); | |
119 | static void json_parser_input(struct json_parser *, struct json_token *); | |
120 | ||
121 | static void json_error(struct json_parser *p, const char *format, ...) | |
122 | PRINTF_FORMAT(2, 3); | |
123 | \f | |
124 | const char * | |
125 | json_type_to_string(enum json_type type) | |
126 | { | |
127 | switch (type) { | |
128 | case JSON_NULL: | |
129 | return "null"; | |
130 | ||
131 | case JSON_FALSE: | |
132 | return "false"; | |
133 | ||
134 | case JSON_TRUE: | |
135 | return "true"; | |
136 | ||
137 | case JSON_OBJECT: | |
138 | return "object"; | |
139 | ||
140 | case JSON_ARRAY: | |
141 | return "array"; | |
142 | ||
143 | case JSON_INTEGER: | |
144 | case JSON_REAL: | |
145 | return "number"; | |
146 | ||
147 | case JSON_STRING: | |
148 | return "string"; | |
149 | ||
150 | case JSON_N_TYPES: | |
151 | default: | |
152 | return "<invalid>"; | |
153 | } | |
154 | } | |
155 | \f | |
156 | /* Functions for manipulating struct json. */ | |
157 | ||
158 | struct json * | |
159 | json_null_create(void) | |
160 | { | |
161 | return json_create(JSON_NULL); | |
162 | } | |
163 | ||
164 | struct json * | |
165 | json_boolean_create(bool b) | |
166 | { | |
167 | return json_create(b ? JSON_TRUE : JSON_FALSE); | |
168 | } | |
169 | ||
170 | struct json * | |
171 | json_string_create_nocopy(char *s) | |
172 | { | |
173 | struct json *json = json_create(JSON_STRING); | |
174 | json->u.string = s; | |
175 | return json; | |
176 | } | |
177 | ||
178 | struct json * | |
179 | json_string_create(const char *s) | |
180 | { | |
181 | return json_string_create_nocopy(xstrdup(s)); | |
182 | } | |
183 | ||
184 | struct json * | |
185 | json_array_create_empty(void) | |
186 | { | |
187 | struct json *json = json_create(JSON_ARRAY); | |
188 | json->u.array.elems = NULL; | |
189 | json->u.array.n = 0; | |
190 | json->u.array.n_allocated = 0; | |
191 | return json; | |
192 | } | |
193 | ||
194 | void | |
195 | json_array_add(struct json *array_, struct json *element) | |
196 | { | |
197 | struct json_array *array = json_array(array_); | |
198 | if (array->n >= array->n_allocated) { | |
199 | array->elems = x2nrealloc(array->elems, &array->n_allocated, | |
200 | sizeof *array->elems); | |
201 | } | |
202 | array->elems[array->n++] = element; | |
203 | } | |
204 | ||
205 | void | |
206 | json_array_trim(struct json *array_) | |
207 | { | |
208 | struct json_array *array = json_array(array_); | |
209 | if (array->n < array->n_allocated){ | |
210 | array->n_allocated = array->n; | |
211 | array->elems = xrealloc(array->elems, array->n * sizeof *array->elems); | |
212 | } | |
213 | } | |
214 | ||
215 | struct json * | |
216 | json_array_create(struct json **elements, size_t n) | |
217 | { | |
218 | struct json *json = json_create(JSON_ARRAY); | |
219 | json->u.array.elems = elements; | |
220 | json->u.array.n = n; | |
221 | json->u.array.n_allocated = n; | |
222 | return json; | |
223 | } | |
224 | ||
f6f8c3ba BP |
225 | struct json * |
226 | json_array_create_1(struct json *elem0) | |
227 | { | |
228 | struct json **elems = xmalloc(sizeof *elems); | |
229 | elems[0] = elem0; | |
230 | return json_array_create(elems, 1); | |
231 | } | |
232 | ||
f38b84ea BP |
233 | struct json * |
234 | json_array_create_2(struct json *elem0, struct json *elem1) | |
235 | { | |
236 | struct json **elems = xmalloc(2 * sizeof *elems); | |
237 | elems[0] = elem0; | |
238 | elems[1] = elem1; | |
239 | return json_array_create(elems, 2); | |
240 | } | |
241 | ||
242 | struct json * | |
243 | json_array_create_3(struct json *elem0, struct json *elem1, struct json *elem2) | |
244 | { | |
245 | struct json **elems = xmalloc(3 * sizeof *elems); | |
246 | elems[0] = elem0; | |
247 | elems[1] = elem1; | |
248 | elems[2] = elem2; | |
249 | return json_array_create(elems, 3); | |
250 | } | |
251 | ||
252 | struct json * | |
253 | json_object_create(void) | |
254 | { | |
255 | struct json *json = json_create(JSON_OBJECT); | |
256 | json->u.object = xmalloc(sizeof *json->u.object); | |
257 | shash_init(json->u.object); | |
258 | return json; | |
259 | } | |
260 | ||
261 | struct json * | |
262 | json_integer_create(long long int integer) | |
263 | { | |
264 | struct json *json = json_create(JSON_INTEGER); | |
265 | json->u.integer = integer; | |
266 | return json; | |
267 | } | |
268 | ||
269 | struct json * | |
270 | json_real_create(double real) | |
271 | { | |
272 | struct json *json = json_create(JSON_REAL); | |
273 | json->u.real = real; | |
274 | return json; | |
275 | } | |
276 | ||
277 | void | |
278 | json_object_put(struct json *json, const char *name, struct json *value) | |
279 | { | |
280 | shash_add(json->u.object, name, value); | |
281 | } | |
282 | ||
283 | void | |
284 | json_object_put_string(struct json *json, const char *name, const char *value) | |
285 | { | |
286 | json_object_put(json, name, json_string_create(value)); | |
287 | } | |
288 | ||
289 | const char * | |
290 | json_string(const struct json *json) | |
291 | { | |
292 | assert(json->type == JSON_STRING); | |
293 | return json->u.string; | |
294 | } | |
295 | ||
296 | struct json_array * | |
297 | json_array(const struct json *json) | |
298 | { | |
299 | assert(json->type == JSON_ARRAY); | |
300 | return (struct json_array *) &json->u.array; | |
301 | } | |
302 | ||
303 | struct shash * | |
304 | json_object(const struct json *json) | |
305 | { | |
306 | assert(json->type == JSON_OBJECT); | |
307 | return (struct shash *) json->u.object; | |
308 | } | |
309 | ||
310 | bool | |
311 | json_boolean(const struct json *json) | |
312 | { | |
313 | assert(json->type == JSON_TRUE || json->type == JSON_FALSE); | |
314 | return json->type == JSON_TRUE; | |
315 | } | |
316 | ||
317 | double | |
318 | json_real(const struct json *json) | |
319 | { | |
320 | assert(json->type == JSON_REAL || json->type == JSON_INTEGER); | |
321 | return json->type == JSON_REAL ? json->u.real : json->u.integer; | |
322 | } | |
323 | ||
324 | int64_t | |
325 | json_integer(const struct json *json) | |
326 | { | |
327 | assert(json->type == JSON_INTEGER); | |
328 | return json->u.integer; | |
329 | } | |
330 | \f | |
331 | static void json_destroy_object(struct shash *object); | |
332 | static void json_destroy_array(struct json_array *array); | |
333 | ||
334 | /* Frees 'json' and everything it points to, recursively. */ | |
335 | void | |
336 | json_destroy(struct json *json) | |
337 | { | |
338 | if (json) { | |
339 | switch (json->type) { | |
340 | case JSON_OBJECT: | |
341 | json_destroy_object(json->u.object); | |
342 | break; | |
343 | ||
344 | case JSON_ARRAY: | |
345 | json_destroy_array(&json->u.array); | |
346 | break; | |
347 | ||
348 | case JSON_STRING: | |
349 | free(json->u.string); | |
350 | break; | |
351 | ||
352 | case JSON_NULL: | |
353 | case JSON_FALSE: | |
354 | case JSON_TRUE: | |
355 | case JSON_INTEGER: | |
356 | case JSON_REAL: | |
357 | break; | |
358 | ||
359 | case JSON_N_TYPES: | |
360 | NOT_REACHED(); | |
361 | } | |
362 | free(json); | |
363 | } | |
364 | } | |
365 | ||
366 | static void | |
367 | json_destroy_object(struct shash *object) | |
368 | { | |
369 | struct shash_node *node, *next; | |
370 | ||
371 | SHASH_FOR_EACH_SAFE (node, next, object) { | |
372 | struct json *value = node->data; | |
373 | ||
374 | json_destroy(value); | |
375 | shash_delete(object, node); | |
376 | } | |
377 | shash_destroy(object); | |
378 | free(object); | |
379 | } | |
380 | ||
381 | static void | |
382 | json_destroy_array(struct json_array *array) | |
383 | { | |
384 | size_t i; | |
385 | ||
386 | for (i = 0; i < array->n; i++) { | |
387 | json_destroy(array->elems[i]); | |
388 | } | |
389 | free(array->elems); | |
390 | } | |
391 | \f | |
392 | static struct json *json_clone_object(const struct shash *object); | |
393 | static struct json *json_clone_array(const struct json_array *array); | |
394 | ||
395 | /* Returns a deep copy of 'json'. */ | |
396 | struct json * | |
397 | json_clone(const struct json *json) | |
398 | { | |
399 | switch (json->type) { | |
400 | case JSON_OBJECT: | |
401 | return json_clone_object(json->u.object); | |
402 | ||
403 | case JSON_ARRAY: | |
404 | return json_clone_array(&json->u.array); | |
405 | ||
406 | case JSON_STRING: | |
407 | return json_string_create(json->u.string); | |
408 | ||
409 | case JSON_NULL: | |
410 | case JSON_FALSE: | |
411 | case JSON_TRUE: | |
412 | return json_create(json->type); | |
413 | ||
414 | case JSON_INTEGER: | |
415 | return json_integer_create(json->u.integer); | |
416 | ||
417 | case JSON_REAL: | |
418 | return json_real_create(json->u.real); | |
419 | ||
420 | case JSON_N_TYPES: | |
421 | default: | |
422 | NOT_REACHED(); | |
423 | } | |
424 | } | |
425 | ||
426 | static struct json * | |
427 | json_clone_object(const struct shash *object) | |
428 | { | |
429 | struct shash_node *node; | |
430 | struct json *json; | |
431 | ||
432 | json = json_object_create(); | |
433 | SHASH_FOR_EACH (node, object) { | |
434 | struct json *value = node->data; | |
435 | json_object_put(json, node->name, json_clone(value)); | |
436 | } | |
437 | return json; | |
438 | } | |
439 | ||
440 | static struct json * | |
441 | json_clone_array(const struct json_array *array) | |
442 | { | |
443 | struct json **elems; | |
444 | size_t i; | |
445 | ||
446 | elems = xmalloc(array->n * sizeof *elems); | |
447 | for (i = 0; i < array->n; i++) { | |
448 | elems[i] = json_clone(array->elems[i]); | |
449 | } | |
450 | return json_array_create(elems, array->n); | |
451 | } | |
452 | \f | |
453 | static size_t | |
454 | json_hash_object(const struct shash *object, size_t basis) | |
455 | { | |
456 | const struct shash_node **nodes; | |
457 | size_t n, i; | |
458 | ||
459 | nodes = shash_sort(object); | |
460 | n = shash_count(object); | |
461 | for (i = 0; i < n; i++) { | |
462 | const struct shash_node *node = nodes[i]; | |
463 | basis = hash_string(node->name, basis); | |
464 | basis = json_hash(node->data, basis); | |
465 | } | |
466 | return basis; | |
467 | } | |
468 | ||
469 | static size_t | |
470 | json_hash_array(const struct json_array *array, size_t basis) | |
471 | { | |
472 | size_t i; | |
473 | ||
474 | basis = hash_int(array->n, basis); | |
475 | for (i = 0; i < array->n; i++) { | |
476 | basis = json_hash(array->elems[i], basis); | |
477 | } | |
478 | return basis; | |
479 | } | |
480 | ||
481 | size_t | |
482 | json_hash(const struct json *json, size_t basis) | |
483 | { | |
484 | switch (json->type) { | |
485 | case JSON_OBJECT: | |
486 | return json_hash_object(json->u.object, basis); | |
487 | ||
488 | case JSON_ARRAY: | |
489 | return json_hash_array(&json->u.array, basis); | |
490 | ||
491 | case JSON_STRING: | |
492 | return hash_string(json->u.string, basis); | |
493 | ||
494 | case JSON_NULL: | |
495 | case JSON_FALSE: | |
496 | case JSON_TRUE: | |
497 | return hash_int(json->type << 8, basis); | |
498 | ||
499 | case JSON_INTEGER: | |
500 | return hash_int(json->u.integer, basis); | |
501 | ||
502 | case JSON_REAL: | |
503 | return hash_double(json->u.real, basis); | |
504 | ||
505 | case JSON_N_TYPES: | |
506 | default: | |
507 | NOT_REACHED(); | |
508 | } | |
509 | } | |
510 | ||
511 | static bool | |
512 | json_equal_object(const struct shash *a, const struct shash *b) | |
513 | { | |
6e57173f | 514 | struct shash_node *a_node; |
f38b84ea BP |
515 | |
516 | if (shash_count(a) != shash_count(b)) { | |
517 | return false; | |
518 | } | |
519 | ||
6e57173f BP |
520 | SHASH_FOR_EACH (a_node, a) { |
521 | struct shash_node *b_node = shash_find(b, a_node->name); | |
522 | if (!b_node || !json_equal(a_node->data, b_node->data)) { | |
f38b84ea BP |
523 | return false; |
524 | } | |
525 | } | |
526 | ||
527 | return true; | |
528 | } | |
529 | ||
530 | static bool | |
531 | json_equal_array(const struct json_array *a, const struct json_array *b) | |
532 | { | |
533 | size_t i; | |
534 | ||
535 | if (a->n != b->n) { | |
536 | return false; | |
537 | } | |
538 | ||
539 | for (i = 0; i < a->n; i++) { | |
540 | if (!json_equal(a->elems[i], b->elems[i])) { | |
541 | return false; | |
542 | } | |
543 | } | |
544 | ||
545 | return true; | |
546 | } | |
547 | ||
548 | bool | |
549 | json_equal(const struct json *a, const struct json *b) | |
550 | { | |
551 | if (a->type != b->type) { | |
552 | return false; | |
553 | } | |
554 | ||
555 | switch (a->type) { | |
556 | case JSON_OBJECT: | |
557 | return json_equal_object(a->u.object, b->u.object); | |
558 | ||
559 | case JSON_ARRAY: | |
560 | return json_equal_array(&a->u.array, &b->u.array); | |
561 | ||
562 | case JSON_STRING: | |
563 | return !strcmp(a->u.string, b->u.string); | |
564 | ||
565 | case JSON_NULL: | |
566 | case JSON_FALSE: | |
567 | case JSON_TRUE: | |
568 | return true; | |
569 | ||
570 | case JSON_INTEGER: | |
571 | return a->u.integer == b->u.integer; | |
572 | ||
573 | case JSON_REAL: | |
574 | return a->u.real == b->u.real; | |
575 | ||
576 | case JSON_N_TYPES: | |
577 | default: | |
578 | NOT_REACHED(); | |
579 | } | |
580 | } | |
581 | \f | |
582 | /* Lexical analysis. */ | |
583 | ||
584 | static void | |
585 | json_lex_keyword(struct json_parser *p) | |
586 | { | |
587 | struct json_token token; | |
588 | const char *s; | |
589 | ||
590 | s = ds_cstr(&p->buffer); | |
591 | if (!strcmp(s, "false")) { | |
592 | token.type = T_FALSE; | |
593 | } else if (!strcmp(s, "true")) { | |
594 | token.type = T_TRUE; | |
595 | } else if (!strcmp(s, "null")) { | |
596 | token.type = T_NULL; | |
597 | } else { | |
598 | json_error(p, "invalid keyword '%s'", s); | |
599 | return; | |
600 | } | |
601 | json_parser_input(p, &token); | |
602 | } | |
603 | ||
604 | static void | |
605 | json_lex_number(struct json_parser *p) | |
606 | { | |
607 | const char *cp = ds_cstr(&p->buffer); | |
608 | unsigned long long int significand = 0; | |
a105c27b | 609 | struct json_token token; |
f38b84ea BP |
610 | int sig_digits = 0; |
611 | bool imprecise = false; | |
612 | bool negative = false; | |
613 | int pow10 = 0; | |
614 | ||
615 | /* Leading minus sign. */ | |
616 | if (*cp == '-') { | |
617 | negative = true; | |
618 | cp++; | |
619 | } | |
620 | ||
621 | /* At least one integer digit, but 0 may not be used as a leading digit for | |
622 | * a longer number. */ | |
623 | significand = 0; | |
624 | sig_digits = 0; | |
625 | if (*cp == '0') { | |
626 | cp++; | |
627 | if (isdigit(*cp)) { | |
628 | json_error(p, "leading zeros not allowed"); | |
629 | return; | |
630 | } | |
631 | } else if (isdigit(*cp)) { | |
632 | do { | |
633 | if (significand <= ULLONG_MAX / 10) { | |
634 | significand = significand * 10 + (*cp - '0'); | |
635 | sig_digits++; | |
636 | } else { | |
637 | pow10++; | |
638 | if (*cp != '0') { | |
639 | imprecise = true; | |
640 | } | |
641 | } | |
642 | cp++; | |
643 | } while (isdigit(*cp)); | |
644 | } else { | |
645 | json_error(p, "'-' must be followed by digit"); | |
646 | return; | |
647 | } | |
648 | ||
649 | /* Optional fraction. */ | |
650 | if (*cp == '.') { | |
651 | cp++; | |
652 | if (!isdigit(*cp)) { | |
653 | json_error(p, "decimal point must be followed by digit"); | |
654 | return; | |
655 | } | |
656 | do { | |
657 | if (significand <= ULLONG_MAX / 10) { | |
658 | significand = significand * 10 + (*cp - '0'); | |
659 | sig_digits++; | |
660 | pow10--; | |
661 | } else if (*cp != '0') { | |
662 | imprecise = true; | |
663 | } | |
664 | cp++; | |
665 | } while (isdigit(*cp)); | |
666 | } | |
667 | ||
668 | /* Optional exponent. */ | |
669 | if (*cp == 'e' || *cp == 'E') { | |
670 | bool negative_exponent = false; | |
671 | int exponent; | |
672 | ||
673 | cp++; | |
674 | if (*cp == '+') { | |
675 | cp++; | |
676 | } else if (*cp == '-') { | |
677 | negative_exponent = true; | |
678 | cp++; | |
679 | } | |
680 | ||
681 | if (!isdigit(*cp)) { | |
682 | json_error(p, "exponent must contain at least one digit"); | |
683 | return; | |
684 | } | |
685 | ||
686 | exponent = 0; | |
687 | do { | |
688 | if (exponent >= INT_MAX / 10) { | |
689 | json_error(p, "exponent outside valid range"); | |
690 | return; | |
691 | } | |
692 | exponent = exponent * 10 + (*cp - '0'); | |
693 | cp++; | |
694 | } while (isdigit(*cp)); | |
695 | ||
696 | if (negative_exponent) { | |
697 | pow10 -= exponent; | |
698 | } else { | |
699 | pow10 += exponent; | |
700 | } | |
701 | } | |
702 | ||
703 | if (*cp != '\0') { | |
704 | json_error(p, "syntax error in number"); | |
705 | return; | |
706 | } | |
707 | ||
708 | /* Figure out number. | |
709 | * | |
710 | * We suppress negative zeros as a matter of policy. */ | |
711 | if (!significand) { | |
712 | struct json_token token; | |
713 | token.type = T_INTEGER; | |
714 | token.u.integer = 0; | |
715 | json_parser_input(p, &token); | |
716 | return; | |
717 | } | |
718 | ||
719 | if (!imprecise) { | |
720 | while (pow10 > 0 && significand < ULLONG_MAX / 10) { | |
721 | significand *= 10; | |
722 | sig_digits++; | |
723 | pow10--; | |
724 | } | |
725 | while (pow10 < 0 && significand % 10 == 0) { | |
726 | significand /= 10; | |
727 | sig_digits--; | |
728 | pow10++; | |
729 | } | |
730 | if (pow10 == 0 | |
731 | && significand <= (negative | |
732 | ? (unsigned long long int) LLONG_MAX + 1 | |
733 | : LLONG_MAX)) { | |
f38b84ea BP |
734 | token.type = T_INTEGER; |
735 | token.u.integer = negative ? -significand : significand; | |
736 | json_parser_input(p, &token); | |
737 | return; | |
738 | } | |
739 | } | |
740 | ||
a105c27b BP |
741 | token.type = T_REAL; |
742 | if (!str_to_double(ds_cstr(&p->buffer), &token.u.real)) { | |
743 | json_error(p, "number outside valid range"); | |
744 | return; | |
f38b84ea | 745 | } |
a105c27b BP |
746 | /* Suppress negative zero. */ |
747 | if (token.u.real == 0) { | |
748 | token.u.real = 0; | |
749 | } | |
750 | json_parser_input(p, &token); | |
f38b84ea BP |
751 | } |
752 | ||
7d23a63a BP |
753 | static const char * |
754 | json_lex_4hex(const char *cp, const char *end, int *valuep) | |
f38b84ea BP |
755 | { |
756 | int value, i; | |
757 | ||
7d23a63a BP |
758 | if (cp + 4 > end) { |
759 | return "quoted string ends within \\u escape"; | |
760 | } | |
761 | ||
f38b84ea BP |
762 | value = 0; |
763 | for (i = 0; i < 4; i++) { | |
764 | unsigned char c = *cp++; | |
765 | if (!isxdigit(c)) { | |
7d23a63a | 766 | return "malformed \\u escape"; |
f38b84ea BP |
767 | } |
768 | value = (value << 4) | hexit_value(c); | |
769 | } | |
770 | if (!value) { | |
7d23a63a | 771 | return "null bytes not supported in quoted strings"; |
f38b84ea BP |
772 | } |
773 | *valuep = value; | |
7d23a63a | 774 | return NULL; |
f38b84ea BP |
775 | } |
776 | ||
777 | static const char * | |
7d23a63a | 778 | json_lex_unicode(const char *cp, const char *end, struct ds *out) |
f38b84ea | 779 | { |
7d23a63a | 780 | const char *error; |
f38b84ea BP |
781 | int c0, c1; |
782 | ||
7d23a63a BP |
783 | error = json_lex_4hex(cp, end, &c0); |
784 | if (error) { | |
785 | ds_clear(out); | |
786 | ds_put_cstr(out, error); | |
f38b84ea BP |
787 | return NULL; |
788 | } | |
789 | cp += 4; | |
790 | if (!uc_is_leading_surrogate(c0)) { | |
7d23a63a | 791 | ds_put_utf8(out, c0); |
f38b84ea BP |
792 | return cp; |
793 | } | |
794 | ||
7d23a63a BP |
795 | if (cp + 2 > end || *cp++ != '\\' || *cp++ != 'u') { |
796 | ds_clear(out); | |
797 | ds_put_cstr(out, "malformed escaped surrogate pair"); | |
f38b84ea BP |
798 | return NULL; |
799 | } | |
800 | ||
7d23a63a BP |
801 | error = json_lex_4hex(cp, end, &c1); |
802 | if (error) { | |
803 | ds_clear(out); | |
804 | ds_put_cstr(out, error); | |
f38b84ea BP |
805 | return NULL; |
806 | } | |
807 | cp += 4; | |
808 | if (!uc_is_trailing_surrogate(c1)) { | |
7d23a63a BP |
809 | ds_clear(out); |
810 | ds_put_cstr(out, "second half of escaped surrogate pair is not " | |
811 | "trailing surrogate"); | |
f38b84ea BP |
812 | return NULL; |
813 | } | |
814 | ||
7d23a63a | 815 | ds_put_utf8(out, utf16_decode_surrogate_pair(c0, c1)); |
f38b84ea BP |
816 | return cp; |
817 | } | |
818 | ||
7d23a63a BP |
819 | bool |
820 | json_string_unescape(const char *in, size_t in_len, char **outp) | |
821 | { | |
822 | const char *end = in + in_len; | |
823 | bool ok = false; | |
824 | struct ds out; | |
825 | ||
826 | ds_init(&out); | |
827 | ds_reserve(&out, in_len); | |
828 | if (in_len > 0 && in[in_len - 1] == '\\') { | |
829 | ds_put_cstr(&out, "quoted string may not end with backslash"); | |
830 | goto exit; | |
831 | } | |
832 | while (in < end) { | |
833 | if (*in == '"') { | |
834 | ds_clear(&out); | |
835 | ds_put_cstr(&out, "quoted string may not include unescape \""); | |
836 | goto exit; | |
837 | } | |
838 | if (*in != '\\') { | |
839 | ds_put_char(&out, *in++); | |
f38b84ea BP |
840 | continue; |
841 | } | |
842 | ||
7d23a63a BP |
843 | in++; |
844 | switch (*in++) { | |
f38b84ea | 845 | case '"': case '\\': case '/': |
7d23a63a | 846 | ds_put_char(&out, in[-1]); |
f38b84ea BP |
847 | break; |
848 | ||
849 | case 'b': | |
7d23a63a | 850 | ds_put_char(&out, '\b'); |
f38b84ea BP |
851 | break; |
852 | ||
853 | case 'f': | |
7d23a63a | 854 | ds_put_char(&out, '\f'); |
f38b84ea BP |
855 | break; |
856 | ||
857 | case 'n': | |
7d23a63a | 858 | ds_put_char(&out, '\n'); |
f38b84ea BP |
859 | break; |
860 | ||
861 | case 'r': | |
7d23a63a | 862 | ds_put_char(&out, '\r'); |
f38b84ea BP |
863 | break; |
864 | ||
865 | case 't': | |
7d23a63a | 866 | ds_put_char(&out, '\t'); |
f38b84ea BP |
867 | break; |
868 | ||
869 | case 'u': | |
7d23a63a BP |
870 | in = json_lex_unicode(in, end, &out); |
871 | if (!in) { | |
f38b84ea BP |
872 | goto exit; |
873 | } | |
874 | break; | |
875 | ||
876 | default: | |
7d23a63a BP |
877 | ds_clear(&out); |
878 | ds_put_format(&out, "bad escape \\%c", in[-1]); | |
f38b84ea BP |
879 | goto exit; |
880 | } | |
881 | } | |
7d23a63a BP |
882 | ok = true; |
883 | ||
884 | exit: | |
885 | *outp = ds_cstr(&out); | |
886 | return ok; | |
887 | } | |
888 | ||
889 | static void | |
890 | json_parser_input_string(struct json_parser *p, const char *s) | |
891 | { | |
892 | struct json_token token; | |
f38b84ea BP |
893 | |
894 | token.type = T_STRING; | |
7d23a63a | 895 | token.u.string = s; |
f38b84ea | 896 | json_parser_input(p, &token); |
7d23a63a | 897 | } |
f38b84ea | 898 | |
7d23a63a BP |
899 | static void |
900 | json_lex_string(struct json_parser *p) | |
901 | { | |
902 | const char *raw = ds_cstr(&p->buffer); | |
903 | if (!strchr(raw, '\\')) { | |
904 | json_parser_input_string(p, raw); | |
905 | } else { | |
906 | char *cooked; | |
907 | ||
908 | if (json_string_unescape(raw, strlen(raw), &cooked)) { | |
909 | json_parser_input_string(p, cooked); | |
910 | } else { | |
911 | json_error(p, "%s", cooked); | |
912 | } | |
913 | ||
914 | free(cooked); | |
915 | } | |
f38b84ea BP |
916 | } |
917 | ||
918 | static bool | |
0bdf342a | 919 | json_lex_input(struct json_parser *p, unsigned char c) |
f38b84ea BP |
920 | { |
921 | struct json_token token; | |
922 | ||
0bdf342a BP |
923 | p->byte_number++; |
924 | if (c == '\n') { | |
925 | p->column_number = 0; | |
926 | p->line_number++; | |
927 | } else { | |
928 | p->column_number++; | |
929 | } | |
930 | ||
f38b84ea BP |
931 | switch (p->lex_state) { |
932 | case JSON_LEX_START: | |
933 | switch (c) { | |
934 | case ' ': case '\t': case '\n': case '\r': | |
935 | /* Nothing to do. */ | |
936 | return true; | |
937 | ||
938 | case 'a': case 'b': case 'c': case 'd': case 'e': | |
939 | case 'f': case 'g': case 'h': case 'i': case 'j': | |
940 | case 'k': case 'l': case 'm': case 'n': case 'o': | |
941 | case 'p': case 'q': case 'r': case 's': case 't': | |
942 | case 'u': case 'v': case 'w': case 'x': case 'y': | |
943 | case 'z': | |
944 | p->lex_state = JSON_LEX_KEYWORD; | |
945 | break; | |
946 | ||
947 | case '[': case '{': case ']': case '}': case ':': case ',': | |
948 | token.type = c; | |
949 | json_parser_input(p, &token); | |
950 | return true; | |
951 | ||
952 | case '-': | |
953 | case '0': case '1': case '2': case '3': case '4': | |
954 | case '5': case '6': case '7': case '8': case '9': | |
955 | p->lex_state = JSON_LEX_NUMBER; | |
956 | break; | |
957 | ||
958 | case '"': | |
959 | p->lex_state = JSON_LEX_STRING; | |
960 | return true; | |
961 | ||
962 | default: | |
963 | if (isprint(c)) { | |
964 | json_error(p, "invalid character '%c'", c); | |
965 | } else { | |
966 | json_error(p, "invalid character U+%04x", c); | |
967 | } | |
968 | return true; | |
969 | } | |
970 | break; | |
971 | ||
972 | case JSON_LEX_KEYWORD: | |
973 | if (!isalpha((unsigned char) c)) { | |
974 | json_lex_keyword(p); | |
975 | return false; | |
976 | } | |
977 | break; | |
978 | ||
979 | case JSON_LEX_NUMBER: | |
980 | if (!strchr(".0123456789eE-+", c)) { | |
981 | json_lex_number(p); | |
982 | return false; | |
983 | } | |
984 | break; | |
985 | ||
986 | case JSON_LEX_STRING: | |
987 | if (c == '\\') { | |
988 | p->lex_state = JSON_LEX_ESCAPE; | |
989 | } else if (c == '"') { | |
990 | json_lex_string(p); | |
991 | return true; | |
992 | } else if (c < 0x20) { | |
993 | json_error(p, "U+%04X must be escaped in quoted string", c); | |
994 | return true; | |
995 | } | |
996 | break; | |
997 | ||
998 | case JSON_LEX_ESCAPE: | |
999 | p->lex_state = JSON_LEX_STRING; | |
1000 | break; | |
1001 | ||
1002 | default: | |
1003 | abort(); | |
1004 | } | |
1005 | ds_put_char(&p->buffer, c); | |
1006 | return true; | |
1007 | } | |
1008 | \f | |
1009 | /* Parsing. */ | |
1010 | ||
1011 | /* Parses 'string' as a JSON object or array and returns a newly allocated | |
1012 | * 'struct json'. The caller must free the returned structure with | |
1013 | * json_destroy() when it is no longer needed. | |
1014 | * | |
1015 | * 'string' must be encoded in UTF-8. | |
1016 | * | |
1017 | * If 'string' is valid JSON, then the returned 'struct json' will be either an | |
1018 | * object (JSON_OBJECT) or an array (JSON_ARRAY). | |
1019 | * | |
1020 | * If 'string' is not valid JSON, then the returned 'struct json' will be a | |
1021 | * string (JSON_STRING) that describes the particular error encountered during | |
1022 | * parsing. (This is an acceptable means of error reporting because at its top | |
1023 | * level JSON must be either an object or an array; a bare string is not | |
1024 | * valid.) */ | |
1025 | struct json * | |
1026 | json_from_string(const char *string) | |
1027 | { | |
1028 | struct json_parser *p = json_parser_create(JSPF_TRAILER); | |
1029 | json_parser_feed(p, string, strlen(string)); | |
1030 | return json_parser_finish(p); | |
1031 | } | |
1032 | ||
1033 | /* Reads the file named 'file_name', parses its contents as a JSON object or | |
1034 | * array, and returns a newly allocated 'struct json'. The caller must free | |
1035 | * the returned structure with json_destroy() when it is no longer needed. | |
1036 | * | |
1037 | * The file must be encoded in UTF-8. | |
1038 | * | |
1039 | * See json_from_string() for return value semantics. | |
1040 | */ | |
1041 | struct json * | |
1042 | json_from_file(const char *file_name) | |
1043 | { | |
f38b84ea BP |
1044 | struct json *json; |
1045 | FILE *stream; | |
1046 | ||
f38b84ea BP |
1047 | stream = fopen(file_name, "r"); |
1048 | if (!stream) { | |
1049 | return json_string_create_nocopy( | |
1050 | xasprintf("error opening \"%s\": %s", file_name, strerror(errno))); | |
1051 | } | |
5562d6f5 BP |
1052 | json = json_from_stream(stream); |
1053 | fclose(stream); | |
1054 | ||
1055 | return json; | |
1056 | } | |
1057 | ||
1058 | /* Parses the contents of 'stream' as a JSON object or array, and returns a | |
1059 | * newly allocated 'struct json'. The caller must free the returned structure | |
1060 | * with json_destroy() when it is no longer needed. | |
1061 | * | |
1062 | * The file must be encoded in UTF-8. | |
1063 | * | |
1064 | * See json_from_string() for return value semantics. | |
1065 | */ | |
1066 | struct json * | |
1067 | json_from_stream(FILE *stream) | |
1068 | { | |
1069 | struct json_parser *p; | |
1070 | struct json *json; | |
f38b84ea | 1071 | |
f38b84ea BP |
1072 | p = json_parser_create(JSPF_TRAILER); |
1073 | for (;;) { | |
1074 | char buffer[BUFSIZ]; | |
1075 | size_t n; | |
1076 | ||
1077 | n = fread(buffer, 1, sizeof buffer, stream); | |
1078 | if (!n || json_parser_feed(p, buffer, n) != n) { | |
1079 | break; | |
1080 | } | |
1081 | } | |
1082 | json = json_parser_finish(p); | |
1083 | ||
f38b84ea BP |
1084 | if (ferror(stream)) { |
1085 | json_destroy(json); | |
1086 | json = json_string_create_nocopy( | |
5562d6f5 | 1087 | xasprintf("error reading JSON stream: %s", strerror(errno))); |
f38b84ea | 1088 | } |
f38b84ea BP |
1089 | |
1090 | return json; | |
1091 | } | |
1092 | ||
1093 | struct json_parser * | |
1094 | json_parser_create(int flags) | |
1095 | { | |
1096 | struct json_parser *p = xzalloc(sizeof *p); | |
1097 | p->flags = flags; | |
1098 | return p; | |
1099 | } | |
1100 | ||
1101 | size_t | |
1102 | json_parser_feed(struct json_parser *p, const char *input, size_t n) | |
1103 | { | |
1104 | size_t i; | |
1105 | for (i = 0; !p->done && i < n; ) { | |
1106 | if (json_lex_input(p, input[i])) { | |
1107 | i++; | |
1108 | } | |
1109 | } | |
1110 | return i; | |
1111 | } | |
1112 | ||
1113 | bool | |
1114 | json_parser_is_done(const struct json_parser *p) | |
1115 | { | |
1116 | return p->done; | |
1117 | } | |
1118 | ||
1119 | struct json * | |
1120 | json_parser_finish(struct json_parser *p) | |
1121 | { | |
1122 | struct json *json; | |
1123 | ||
1124 | switch (p->lex_state) { | |
1125 | case JSON_LEX_START: | |
1126 | break; | |
1127 | ||
1128 | case JSON_LEX_STRING: | |
1129 | case JSON_LEX_ESCAPE: | |
1130 | json_error(p, "unexpected end of input in quoted string"); | |
1131 | break; | |
1132 | ||
1133 | case JSON_LEX_NUMBER: | |
1134 | case JSON_LEX_KEYWORD: | |
1135 | json_lex_input(p, ' '); | |
1136 | break; | |
1137 | } | |
1138 | ||
1139 | if (p->parse_state == JSON_PARSE_START) { | |
1140 | json_error(p, "empty input stream"); | |
1141 | } else if (p->parse_state != JSON_PARSE_END) { | |
1142 | json_error(p, "unexpected end of input"); | |
1143 | } | |
1144 | ||
1145 | if (!p->error) { | |
1146 | assert(p->height == 1); | |
1147 | assert(p->stack[0].json != NULL); | |
1148 | json = p->stack[--p->height].json; | |
1149 | } else { | |
1150 | json = json_string_create_nocopy(p->error); | |
1151 | p->error = NULL; | |
1152 | } | |
1153 | ||
1154 | json_parser_abort(p); | |
1155 | ||
1156 | return json; | |
1157 | } | |
1158 | ||
1159 | void | |
1160 | json_parser_abort(struct json_parser *p) | |
1161 | { | |
1162 | if (p) { | |
1163 | ds_destroy(&p->buffer); | |
1164 | if (p->height) { | |
1165 | json_destroy(p->stack[0].json); | |
1166 | } | |
1167 | free(p->stack); | |
1168 | free(p->member_name); | |
1169 | free(p->error); | |
1170 | free(p); | |
1171 | } | |
1172 | } | |
1173 | ||
1174 | static struct json_parser_node * | |
1175 | json_parser_top(struct json_parser *p) | |
1176 | { | |
1177 | return &p->stack[p->height - 1]; | |
1178 | } | |
1179 | ||
1180 | static void | |
1181 | json_parser_put_value(struct json_parser *p, struct json *value) | |
1182 | { | |
1183 | struct json_parser_node *node = json_parser_top(p); | |
1184 | if (node->json->type == JSON_OBJECT) { | |
1185 | json_object_put(node->json, p->member_name, value); | |
1186 | free(p->member_name); | |
1187 | p->member_name = NULL; | |
1188 | } else if (node->json->type == JSON_ARRAY) { | |
1189 | json_array_add(node->json, value); | |
1190 | } else { | |
1191 | NOT_REACHED(); | |
1192 | } | |
1193 | } | |
1194 | ||
1195 | static struct json_parser_node * | |
1196 | json_parser_push(struct json_parser *p, | |
1197 | struct json *new_json, enum json_parse_state new_state) | |
1198 | { | |
1199 | if (p->height < JSON_MAX_HEIGHT) { | |
1200 | struct json_parser_node *node; | |
1201 | ||
1202 | if (p->height >= p->allocated_height) { | |
1203 | p->stack = x2nrealloc(p->stack, &p->allocated_height, | |
1204 | sizeof *p->stack); | |
1205 | } | |
1206 | ||
1207 | if (p->height > 0) { | |
1208 | json_parser_put_value(p, new_json); | |
1209 | } | |
1210 | ||
1211 | node = &p->stack[p->height++]; | |
1212 | node->json = new_json; | |
1213 | p->parse_state = new_state; | |
1214 | return node; | |
1215 | } else { | |
1216 | json_error(p, "input exceeds maximum nesting depth %d", | |
1217 | JSON_MAX_HEIGHT); | |
1218 | return NULL; | |
1219 | } | |
1220 | } | |
1221 | ||
1222 | static void | |
1223 | json_parser_push_object(struct json_parser *p) | |
1224 | { | |
1225 | json_parser_push(p, json_object_create(), JSON_PARSE_OBJECT_INIT); | |
1226 | } | |
1227 | ||
1228 | static void | |
1229 | json_parser_push_array(struct json_parser *p) | |
1230 | { | |
1231 | json_parser_push(p, json_array_create_empty(), JSON_PARSE_ARRAY_INIT); | |
1232 | } | |
1233 | ||
1234 | static void | |
1235 | json_parse_value(struct json_parser *p, struct json_token *token, | |
1236 | enum json_parse_state next_state) | |
1237 | { | |
1238 | struct json *value; | |
1239 | ||
1240 | switch (token->type) { | |
1241 | case T_FALSE: | |
1242 | value = json_boolean_create(false); | |
1243 | break; | |
1244 | ||
1245 | case T_NULL: | |
1246 | value = json_null_create(); | |
1247 | break; | |
1248 | ||
1249 | case T_TRUE: | |
1250 | value = json_boolean_create(true); | |
1251 | break; | |
1252 | ||
1253 | case '{': | |
1254 | json_parser_push_object(p); | |
1255 | return; | |
1256 | ||
1257 | case '[': | |
1258 | json_parser_push_array(p); | |
1259 | return; | |
1260 | ||
1261 | case T_INTEGER: | |
1262 | value = json_integer_create(token->u.integer); | |
1263 | break; | |
1264 | ||
1265 | case T_REAL: | |
1266 | value = json_real_create(token->u.real); | |
1267 | break; | |
1268 | ||
1269 | case T_STRING: | |
1270 | value = json_string_create(token->u.string); | |
1271 | break; | |
1272 | ||
1273 | case T_EOF: | |
1274 | case '}': | |
1275 | case ']': | |
1276 | case ':': | |
1277 | case ',': | |
1278 | default: | |
1279 | json_error(p, "syntax error expecting value"); | |
1280 | return; | |
1281 | } | |
1282 | ||
1283 | json_parser_put_value(p, value); | |
1284 | p->parse_state = next_state; | |
1285 | } | |
1286 | ||
1287 | static void | |
1288 | json_parser_pop(struct json_parser *p) | |
1289 | { | |
1290 | struct json_parser_node *node; | |
1291 | ||
1292 | /* Conserve memory. */ | |
1293 | node = json_parser_top(p); | |
1294 | if (node->json->type == JSON_ARRAY) { | |
1295 | json_array_trim(node->json); | |
1296 | } | |
1297 | ||
1298 | /* Pop off the top-of-stack. */ | |
1299 | if (p->height == 1) { | |
1300 | p->parse_state = JSON_PARSE_END; | |
1301 | if (!(p->flags & JSPF_TRAILER)) { | |
1302 | p->done = true; | |
1303 | } | |
1304 | } else { | |
1305 | p->height--; | |
1306 | node = json_parser_top(p); | |
1307 | if (node->json->type == JSON_ARRAY) { | |
1308 | p->parse_state = JSON_PARSE_ARRAY_NEXT; | |
1309 | } else if (node->json->type == JSON_OBJECT) { | |
1310 | p->parse_state = JSON_PARSE_OBJECT_NEXT; | |
1311 | } else { | |
1312 | NOT_REACHED(); | |
1313 | } | |
1314 | } | |
1315 | } | |
1316 | ||
1317 | static void | |
1318 | json_parser_input(struct json_parser *p, struct json_token *token) | |
1319 | { | |
1320 | switch (p->parse_state) { | |
1321 | case JSON_PARSE_START: | |
1322 | if (token->type == '{') { | |
1323 | json_parser_push_object(p); | |
1324 | } else if (token->type == '[') { | |
1325 | json_parser_push_array(p); | |
1326 | } else { | |
1327 | json_error(p, "syntax error at beginning of input"); | |
1328 | } | |
1329 | break; | |
1330 | ||
1331 | case JSON_PARSE_END: | |
1332 | json_error(p, "trailing garbage at end of input"); | |
1333 | break; | |
1334 | ||
1335 | case JSON_PARSE_OBJECT_INIT: | |
1336 | if (token->type == '}') { | |
1337 | json_parser_pop(p); | |
1338 | break; | |
1339 | } | |
1340 | /* Fall through. */ | |
1341 | case JSON_PARSE_OBJECT_NAME: | |
1342 | if (token->type == T_STRING) { | |
1343 | p->member_name = xstrdup(token->u.string); | |
1344 | p->parse_state = JSON_PARSE_OBJECT_COLON; | |
1345 | } else { | |
1346 | json_error(p, "syntax error parsing object expecting string"); | |
1347 | } | |
1348 | break; | |
1349 | ||
1350 | case JSON_PARSE_OBJECT_COLON: | |
1351 | if (token->type == ':') { | |
1352 | p->parse_state = JSON_PARSE_OBJECT_VALUE; | |
1353 | } else { | |
1354 | json_error(p, "syntax error parsing object expecting ':'"); | |
1355 | } | |
1356 | break; | |
1357 | ||
1358 | case JSON_PARSE_OBJECT_VALUE: | |
1359 | json_parse_value(p, token, JSON_PARSE_OBJECT_NEXT); | |
1360 | break; | |
1361 | ||
1362 | case JSON_PARSE_OBJECT_NEXT: | |
1363 | if (token->type == ',') { | |
1364 | p->parse_state = JSON_PARSE_OBJECT_NAME; | |
1365 | } else if (token->type == '}') { | |
1366 | json_parser_pop(p); | |
1367 | } else { | |
1368 | json_error(p, "syntax error expecting '}' or ','"); | |
1369 | } | |
1370 | break; | |
1371 | ||
1372 | case JSON_PARSE_ARRAY_INIT: | |
1373 | if (token->type == ']') { | |
1374 | json_parser_pop(p); | |
1375 | break; | |
1376 | } | |
1377 | /* Fall through. */ | |
1378 | case JSON_PARSE_ARRAY_VALUE: | |
1379 | json_parse_value(p, token, JSON_PARSE_ARRAY_NEXT); | |
1380 | break; | |
1381 | ||
1382 | case JSON_PARSE_ARRAY_NEXT: | |
1383 | if (token->type == ',') { | |
1384 | p->parse_state = JSON_PARSE_ARRAY_VALUE; | |
1385 | } else if (token->type == ']') { | |
1386 | json_parser_pop(p); | |
1387 | } else { | |
1388 | json_error(p, "syntax error expecting ']' or ','"); | |
1389 | } | |
1390 | break; | |
1391 | ||
1392 | default: | |
1393 | abort(); | |
1394 | } | |
1395 | ||
1396 | p->lex_state = JSON_LEX_START; | |
1397 | ds_clear(&p->buffer); | |
1398 | } | |
1399 | ||
1400 | static struct json * | |
1401 | json_create(enum json_type type) | |
1402 | { | |
1403 | struct json *json = xmalloc(sizeof *json); | |
1404 | json->type = type; | |
1405 | return json; | |
1406 | } | |
1407 | ||
1408 | static void | |
1409 | json_error(struct json_parser *p, const char *format, ...) | |
1410 | { | |
1411 | if (!p->error) { | |
0bdf342a | 1412 | struct ds msg; |
f38b84ea BP |
1413 | va_list args; |
1414 | ||
0bdf342a BP |
1415 | ds_init(&msg); |
1416 | ds_put_format(&msg, "line %d, column %d, byte %d: ", | |
1417 | p->line_number, p->column_number, p->byte_number); | |
f38b84ea | 1418 | va_start(args, format); |
0bdf342a | 1419 | ds_put_format_valist(&msg, format, args); |
f38b84ea BP |
1420 | va_end(args); |
1421 | ||
0bdf342a BP |
1422 | p->error = ds_steal_cstr(&msg); |
1423 | ||
f38b84ea BP |
1424 | p->done = true; |
1425 | } | |
1426 | } | |
1427 | \f | |
1428 | #define SPACES_PER_LEVEL 2 | |
1429 | ||
1430 | struct json_serializer { | |
36d802ae | 1431 | struct ds *ds; |
f38b84ea BP |
1432 | int depth; |
1433 | int flags; | |
1434 | }; | |
1435 | ||
36d802ae BP |
1436 | static void json_serialize(const struct json *, struct json_serializer *); |
1437 | static void json_serialize_object(const struct shash *object, | |
1438 | struct json_serializer *); | |
1439 | static void json_serialize_array(const struct json_array *, | |
1440 | struct json_serializer *); | |
1441 | static void json_serialize_string(const char *, struct ds *); | |
f38b84ea BP |
1442 | |
1443 | /* Converts 'json' to a string in JSON format, encoded in UTF-8, and returns | |
1444 | * that string. The caller is responsible for freeing the returned string, | |
1445 | * with free(), when it is no longer needed. | |
1446 | * | |
1447 | * If 'flags' contains JSSF_PRETTY, the output is pretty-printed with each | |
1448 | * nesting level introducing an additional indentation. Otherwise, the | |
1449 | * returned string does not contain any new-line characters. | |
1450 | * | |
1451 | * If 'flags' contains JSSF_SORT, members of objects in the output are sorted | |
1452 | * in bytewise lexicographic order for reproducibility. Otherwise, members of | |
1453 | * objects are output in an indeterminate order. | |
1454 | * | |
1455 | * The returned string is valid JSON only if 'json' represents an array or an | |
1456 | * object, since a bare literal does not satisfy the JSON grammar. */ | |
1457 | char * | |
1458 | json_to_string(const struct json *json, int flags) | |
36d802ae BP |
1459 | { |
1460 | struct ds ds; | |
1461 | ||
1462 | ds_init(&ds); | |
1463 | json_to_ds(json, flags, &ds); | |
1464 | return ds_steal_cstr(&ds); | |
1465 | } | |
1466 | ||
1467 | /* Same as json_to_string(), but the output is appended to 'ds'. */ | |
1468 | void | |
1469 | json_to_ds(const struct json *json, int flags, struct ds *ds) | |
f38b84ea BP |
1470 | { |
1471 | struct json_serializer s; | |
36d802ae BP |
1472 | |
1473 | s.ds = ds; | |
f38b84ea BP |
1474 | s.depth = 0; |
1475 | s.flags = flags; | |
36d802ae | 1476 | json_serialize(json, &s); |
f38b84ea BP |
1477 | } |
1478 | ||
1479 | static void | |
36d802ae | 1480 | json_serialize(const struct json *json, struct json_serializer *s) |
f38b84ea | 1481 | { |
36d802ae | 1482 | struct ds *ds = s->ds; |
f38b84ea BP |
1483 | |
1484 | switch (json->type) { | |
1485 | case JSON_NULL: | |
1486 | ds_put_cstr(ds, "null"); | |
1487 | break; | |
1488 | ||
1489 | case JSON_FALSE: | |
1490 | ds_put_cstr(ds, "false"); | |
1491 | break; | |
1492 | ||
1493 | case JSON_TRUE: | |
1494 | ds_put_cstr(ds, "true"); | |
1495 | break; | |
1496 | ||
1497 | case JSON_OBJECT: | |
36d802ae | 1498 | json_serialize_object(json->u.object, s); |
f38b84ea BP |
1499 | break; |
1500 | ||
1501 | case JSON_ARRAY: | |
36d802ae | 1502 | json_serialize_array(&json->u.array, s); |
f38b84ea BP |
1503 | break; |
1504 | ||
1505 | case JSON_INTEGER: | |
1506 | ds_put_format(ds, "%lld", json->u.integer); | |
1507 | break; | |
1508 | ||
1509 | case JSON_REAL: | |
1510 | ds_put_format(ds, "%.*g", DBL_DIG, json->u.real); | |
1511 | break; | |
1512 | ||
1513 | case JSON_STRING: | |
36d802ae | 1514 | json_serialize_string(json->u.string, ds); |
f38b84ea BP |
1515 | break; |
1516 | ||
1517 | case JSON_N_TYPES: | |
1518 | default: | |
1519 | NOT_REACHED(); | |
1520 | } | |
1521 | } | |
1522 | ||
1523 | static void | |
1524 | indent_line(struct json_serializer *s) | |
1525 | { | |
1526 | if (s->flags & JSSF_PRETTY) { | |
36d802ae BP |
1527 | ds_put_char(s->ds, '\n'); |
1528 | ds_put_char_multiple(s->ds, ' ', SPACES_PER_LEVEL * s->depth); | |
f38b84ea BP |
1529 | } |
1530 | } | |
1531 | ||
1532 | static void | |
36d802ae BP |
1533 | json_serialize_object_member(size_t i, const struct shash_node *node, |
1534 | struct json_serializer *s) | |
f38b84ea | 1535 | { |
36d802ae | 1536 | struct ds *ds = s->ds; |
f38b84ea BP |
1537 | |
1538 | if (i) { | |
1539 | ds_put_char(ds, ','); | |
1540 | indent_line(s); | |
1541 | } | |
1542 | ||
36d802ae | 1543 | json_serialize_string(node->name, ds); |
f38b84ea BP |
1544 | ds_put_char(ds, ':'); |
1545 | if (s->flags & JSSF_PRETTY) { | |
1546 | ds_put_char(ds, ' '); | |
1547 | } | |
36d802ae | 1548 | json_serialize(node->data, s); |
f38b84ea BP |
1549 | } |
1550 | ||
1551 | static void | |
36d802ae | 1552 | json_serialize_object(const struct shash *object, struct json_serializer *s) |
f38b84ea | 1553 | { |
36d802ae | 1554 | struct ds *ds = s->ds; |
f38b84ea BP |
1555 | |
1556 | ds_put_char(ds, '{'); | |
1557 | ||
1558 | s->depth++; | |
1559 | indent_line(s); | |
1560 | ||
1561 | if (s->flags & JSSF_SORT) { | |
1562 | const struct shash_node **nodes; | |
1563 | size_t n, i; | |
1564 | ||
1565 | nodes = shash_sort(object); | |
1566 | n = shash_count(object); | |
1567 | for (i = 0; i < n; i++) { | |
36d802ae | 1568 | json_serialize_object_member(i, nodes[i], s); |
f38b84ea BP |
1569 | } |
1570 | free(nodes); | |
1571 | } else { | |
1572 | struct shash_node *node; | |
1573 | size_t i; | |
1574 | ||
1575 | i = 0; | |
1576 | SHASH_FOR_EACH (node, object) { | |
36d802ae | 1577 | json_serialize_object_member(i++, node, s); |
f38b84ea BP |
1578 | } |
1579 | } | |
1580 | ||
1581 | ds_put_char(ds, '}'); | |
1582 | s->depth--; | |
1583 | } | |
1584 | ||
1585 | static void | |
36d802ae | 1586 | json_serialize_array(const struct json_array *array, struct json_serializer *s) |
f38b84ea | 1587 | { |
36d802ae | 1588 | struct ds *ds = s->ds; |
f38b84ea BP |
1589 | size_t i; |
1590 | ||
1591 | ds_put_char(ds, '['); | |
1592 | s->depth++; | |
1593 | ||
1594 | if (array->n > 0) { | |
1595 | indent_line(s); | |
1596 | ||
1597 | for (i = 0; i < array->n; i++) { | |
1598 | if (i) { | |
1599 | ds_put_char(ds, ','); | |
1600 | indent_line(s); | |
1601 | } | |
36d802ae | 1602 | json_serialize(array->elems[i], s); |
f38b84ea BP |
1603 | } |
1604 | } | |
1605 | ||
1606 | s->depth--; | |
1607 | ds_put_char(ds, ']'); | |
1608 | } | |
1609 | ||
1610 | static void | |
36d802ae | 1611 | json_serialize_string(const char *string, struct ds *ds) |
f38b84ea BP |
1612 | { |
1613 | uint8_t c; | |
1614 | ||
1615 | ds_put_char(ds, '"'); | |
1616 | while ((c = *string++) != '\0') { | |
1617 | switch (c) { | |
1618 | case '"': | |
1619 | ds_put_cstr(ds, "\\\""); | |
1620 | break; | |
1621 | ||
1622 | case '\\': | |
1623 | ds_put_cstr(ds, "\\\\"); | |
1624 | break; | |
1625 | ||
1626 | case '\b': | |
1627 | ds_put_cstr(ds, "\\b"); | |
1628 | break; | |
1629 | ||
1630 | case '\f': | |
1631 | ds_put_cstr(ds, "\\f"); | |
1632 | break; | |
1633 | ||
1634 | case '\n': | |
1635 | ds_put_cstr(ds, "\\n"); | |
1636 | break; | |
1637 | ||
1638 | case '\r': | |
1639 | ds_put_cstr(ds, "\\r"); | |
1640 | break; | |
1641 | ||
1642 | case '\t': | |
1643 | ds_put_cstr(ds, "\\t"); | |
1644 | break; | |
1645 | ||
1646 | default: | |
1647 | if (c >= 32) { | |
1648 | ds_put_char(ds, c); | |
1649 | } else { | |
1650 | ds_put_format(ds, "\\u%04x", c); | |
1651 | } | |
1652 | break; | |
1653 | } | |
1654 | } | |
1655 | ds_put_char(ds, '"'); | |
1656 | } |