Line data Source code
1 : /*-------------------------------------------------------------------------
2 : *
3 : * jsonapi.c
4 : * JSON parser and lexer interfaces
5 : *
6 : * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group
7 : * Portions Copyright (c) 1994, Regents of the University of California
8 : *
9 : * IDENTIFICATION
10 : * src/common/jsonapi.c
11 : *
12 : *-------------------------------------------------------------------------
13 : */
14 : #ifndef FRONTEND
15 : #include "postgres.h"
16 : #else
17 : #include "postgres_fe.h"
18 : #endif
19 :
20 : #include "common/jsonapi.h"
21 : #include "mb/pg_wchar.h"
22 : #include "port/pg_lfind.h"
23 :
24 : #ifdef JSONAPI_USE_PQEXPBUFFER
25 : #include "pqexpbuffer.h"
26 : #else
27 : #include "lib/stringinfo.h"
28 : #include "miscadmin.h"
29 : #endif
30 :
31 : /*
32 : * By default, we will use palloc/pfree along with StringInfo. In libpq,
33 : * use malloc and PQExpBuffer, and return JSON_OUT_OF_MEMORY on out-of-memory.
34 : */
35 : #ifdef JSONAPI_USE_PQEXPBUFFER
36 :
37 : #define STRDUP(s) strdup(s)
38 : #define ALLOC(size) malloc(size)
39 : #define ALLOC0(size) calloc(1, size)
40 : #define REALLOC realloc
41 : #define FREE(s) free(s)
42 :
43 : #define jsonapi_appendStringInfo appendPQExpBuffer
44 : #define jsonapi_appendBinaryStringInfo appendBinaryPQExpBuffer
45 : #define jsonapi_appendStringInfoChar appendPQExpBufferChar
46 : /* XXX should we add a macro version to PQExpBuffer? */
47 : #define jsonapi_appendStringInfoCharMacro appendPQExpBufferChar
48 : #define jsonapi_makeStringInfo createPQExpBuffer
49 : #define jsonapi_initStringInfo initPQExpBuffer
50 : #define jsonapi_resetStringInfo resetPQExpBuffer
51 : #define jsonapi_termStringInfo termPQExpBuffer
52 : #define jsonapi_destroyStringInfo destroyPQExpBuffer
53 :
54 : #else /* !JSONAPI_USE_PQEXPBUFFER */
55 :
56 : #define STRDUP(s) pstrdup(s)
57 : #define ALLOC(size) palloc(size)
58 : #define ALLOC0(size) palloc0(size)
59 : #define REALLOC repalloc
60 :
61 : #ifdef FRONTEND
62 : #define FREE pfree
63 : #else
64 : /*
65 : * Backend pfree() doesn't handle NULL pointers like the frontend's does; smooth
66 : * that over to reduce mental gymnastics. Avoid multiple evaluation of the macro
67 : * argument to avoid future hair-pulling.
68 : */
69 : #define FREE(s) do { \
70 : void *__v = (s); \
71 : if (__v) \
72 : pfree(__v); \
73 : } while (0)
74 : #endif
75 :
76 : #define jsonapi_appendStringInfo appendStringInfo
77 : #define jsonapi_appendBinaryStringInfo appendBinaryStringInfo
78 : #define jsonapi_appendStringInfoChar appendStringInfoChar
79 : #define jsonapi_appendStringInfoCharMacro appendStringInfoCharMacro
80 : #define jsonapi_makeStringInfo makeStringInfo
81 : #define jsonapi_initStringInfo initStringInfo
82 : #define jsonapi_resetStringInfo resetStringInfo
83 : #define jsonapi_termStringInfo(s) pfree((s)->data)
84 : #define jsonapi_destroyStringInfo destroyStringInfo
85 :
86 : #endif /* JSONAPI_USE_PQEXPBUFFER */
87 :
88 : /*
89 : * The context of the parser is maintained by the recursive descent
90 : * mechanism, but is passed explicitly to the error reporting routine
91 : * for better diagnostics.
92 : */
93 : typedef enum /* contexts of JSON parser */
94 : {
95 : JSON_PARSE_VALUE, /* expecting a value */
96 : JSON_PARSE_STRING, /* expecting a string (for a field name) */
97 : JSON_PARSE_ARRAY_START, /* saw '[', expecting value or ']' */
98 : JSON_PARSE_ARRAY_NEXT, /* saw array element, expecting ',' or ']' */
99 : JSON_PARSE_OBJECT_START, /* saw '{', expecting label or '}' */
100 : JSON_PARSE_OBJECT_LABEL, /* saw object label, expecting ':' */
101 : JSON_PARSE_OBJECT_NEXT, /* saw object value, expecting ',' or '}' */
102 : JSON_PARSE_OBJECT_COMMA, /* saw object ',', expecting next label */
103 : JSON_PARSE_END, /* saw the end of a document, expect nothing */
104 : } JsonParseContext;
105 :
106 : /*
107 : * Setup for table-driven parser.
108 : * These enums need to be separate from the JsonTokenType and from each other
109 : * so we can have all of them on the prediction stack, which consists of
110 : * tokens, non-terminals, and semantic action markers.
111 : */
112 :
113 : enum JsonNonTerminal
114 : {
115 : JSON_NT_JSON = 32,
116 : JSON_NT_ARRAY_ELEMENTS,
117 : JSON_NT_MORE_ARRAY_ELEMENTS,
118 : JSON_NT_KEY_PAIRS,
119 : JSON_NT_MORE_KEY_PAIRS,
120 : };
121 :
122 : enum JsonParserSem
123 : {
124 : JSON_SEM_OSTART = 64,
125 : JSON_SEM_OEND,
126 : JSON_SEM_ASTART,
127 : JSON_SEM_AEND,
128 : JSON_SEM_OFIELD_INIT,
129 : JSON_SEM_OFIELD_START,
130 : JSON_SEM_OFIELD_END,
131 : JSON_SEM_AELEM_START,
132 : JSON_SEM_AELEM_END,
133 : JSON_SEM_SCALAR_INIT,
134 : JSON_SEM_SCALAR_CALL,
135 : };
136 :
137 : /*
138 : * struct containing the 3 stacks used in non-recursive parsing,
139 : * and the token and value for scalars that need to be preserved
140 : * across calls.
141 : *
142 : * typedef appears in jsonapi.h
143 : */
144 : struct JsonParserStack
145 : {
146 : int stack_size;
147 : char *prediction;
148 : size_t pred_index;
149 : /* these two are indexed by lex_level */
150 : char **fnames;
151 : bool *fnull;
152 : JsonTokenType scalar_tok;
153 : char *scalar_val;
154 : };
155 :
156 : /*
157 : * struct containing state used when there is a possible partial token at the
158 : * end of a json chunk when we are doing incremental parsing.
159 : *
160 : * typedef appears in jsonapi.h
161 : */
162 : struct JsonIncrementalState
163 : {
164 : bool started;
165 : bool is_last_chunk;
166 : bool partial_completed;
167 : jsonapi_StrValType partial_token;
168 : };
169 :
170 : /*
171 : * constants and macros used in the nonrecursive parser
172 : */
173 : #define JSON_NUM_TERMINALS 13
174 : #define JSON_NUM_NONTERMINALS 5
175 : #define JSON_NT_OFFSET JSON_NT_JSON
176 : /* for indexing the table */
177 : #define OFS(NT) (NT) - JSON_NT_OFFSET
178 : /* classify items we get off the stack */
179 : #define IS_SEM(x) ((x) & 0x40)
180 : #define IS_NT(x) ((x) & 0x20)
181 :
182 : /*
183 : * These productions are stored in reverse order right to left so that when
184 : * they are pushed on the stack what we expect next is at the top of the stack.
185 : */
186 : static char JSON_PROD_EPSILON[] = {0}; /* epsilon - an empty production */
187 :
188 : /* JSON -> string */
189 : static char JSON_PROD_SCALAR_STRING[] = {JSON_SEM_SCALAR_CALL, JSON_TOKEN_STRING, JSON_SEM_SCALAR_INIT, 0};
190 :
191 : /* JSON -> number */
192 : static char JSON_PROD_SCALAR_NUMBER[] = {JSON_SEM_SCALAR_CALL, JSON_TOKEN_NUMBER, JSON_SEM_SCALAR_INIT, 0};
193 :
194 : /* JSON -> 'true' */
195 : static char JSON_PROD_SCALAR_TRUE[] = {JSON_SEM_SCALAR_CALL, JSON_TOKEN_TRUE, JSON_SEM_SCALAR_INIT, 0};
196 :
197 : /* JSON -> 'false' */
198 : static char JSON_PROD_SCALAR_FALSE[] = {JSON_SEM_SCALAR_CALL, JSON_TOKEN_FALSE, JSON_SEM_SCALAR_INIT, 0};
199 :
200 : /* JSON -> 'null' */
201 : static char JSON_PROD_SCALAR_NULL[] = {JSON_SEM_SCALAR_CALL, JSON_TOKEN_NULL, JSON_SEM_SCALAR_INIT, 0};
202 :
203 : /* JSON -> '{' KEY_PAIRS '}' */
204 : static char JSON_PROD_OBJECT[] = {JSON_SEM_OEND, JSON_TOKEN_OBJECT_END, JSON_NT_KEY_PAIRS, JSON_TOKEN_OBJECT_START, JSON_SEM_OSTART, 0};
205 :
206 : /* JSON -> '[' ARRAY_ELEMENTS ']' */
207 : static char JSON_PROD_ARRAY[] = {JSON_SEM_AEND, JSON_TOKEN_ARRAY_END, JSON_NT_ARRAY_ELEMENTS, JSON_TOKEN_ARRAY_START, JSON_SEM_ASTART, 0};
208 :
209 : /* ARRAY_ELEMENTS -> JSON MORE_ARRAY_ELEMENTS */
210 : static char JSON_PROD_ARRAY_ELEMENTS[] = {JSON_NT_MORE_ARRAY_ELEMENTS, JSON_SEM_AELEM_END, JSON_NT_JSON, JSON_SEM_AELEM_START, 0};
211 :
212 : /* MORE_ARRAY_ELEMENTS -> ',' JSON MORE_ARRAY_ELEMENTS */
213 : static char JSON_PROD_MORE_ARRAY_ELEMENTS[] = {JSON_NT_MORE_ARRAY_ELEMENTS, JSON_SEM_AELEM_END, JSON_NT_JSON, JSON_SEM_AELEM_START, JSON_TOKEN_COMMA, 0};
214 :
215 : /* KEY_PAIRS -> string ':' JSON MORE_KEY_PAIRS */
216 : static char JSON_PROD_KEY_PAIRS[] = {JSON_NT_MORE_KEY_PAIRS, JSON_SEM_OFIELD_END, JSON_NT_JSON, JSON_SEM_OFIELD_START, JSON_TOKEN_COLON, JSON_TOKEN_STRING, JSON_SEM_OFIELD_INIT, 0};
217 :
218 : /* MORE_KEY_PAIRS -> ',' string ':' JSON MORE_KEY_PAIRS */
219 : static char JSON_PROD_MORE_KEY_PAIRS[] = {JSON_NT_MORE_KEY_PAIRS, JSON_SEM_OFIELD_END, JSON_NT_JSON, JSON_SEM_OFIELD_START, JSON_TOKEN_COLON, JSON_TOKEN_STRING, JSON_SEM_OFIELD_INIT, JSON_TOKEN_COMMA, 0};
220 :
221 : /*
222 : * Note: there are also epsilon productions for ARRAY_ELEMENTS,
223 : * MORE_ARRAY_ELEMENTS, KEY_PAIRS and MORE_KEY_PAIRS
224 : * They are all the same as none require any semantic actions.
225 : */
226 :
227 : /*
228 : * Table connecting the productions with their director sets of
229 : * terminal symbols.
230 : * Any combination not specified here represents an error.
231 : */
232 :
233 : typedef struct
234 : {
235 : size_t len;
236 : char *prod;
237 : } td_entry;
238 :
239 : #define TD_ENTRY(PROD) { sizeof(PROD) - 1, (PROD) }
240 :
241 : static td_entry td_parser_table[JSON_NUM_NONTERMINALS][JSON_NUM_TERMINALS] =
242 : {
243 : /* JSON */
244 : [OFS(JSON_NT_JSON)][JSON_TOKEN_STRING] = TD_ENTRY(JSON_PROD_SCALAR_STRING),
245 : [OFS(JSON_NT_JSON)][JSON_TOKEN_NUMBER] = TD_ENTRY(JSON_PROD_SCALAR_NUMBER),
246 : [OFS(JSON_NT_JSON)][JSON_TOKEN_TRUE] = TD_ENTRY(JSON_PROD_SCALAR_TRUE),
247 : [OFS(JSON_NT_JSON)][JSON_TOKEN_FALSE] = TD_ENTRY(JSON_PROD_SCALAR_FALSE),
248 : [OFS(JSON_NT_JSON)][JSON_TOKEN_NULL] = TD_ENTRY(JSON_PROD_SCALAR_NULL),
249 : [OFS(JSON_NT_JSON)][JSON_TOKEN_ARRAY_START] = TD_ENTRY(JSON_PROD_ARRAY),
250 : [OFS(JSON_NT_JSON)][JSON_TOKEN_OBJECT_START] = TD_ENTRY(JSON_PROD_OBJECT),
251 : /* ARRAY_ELEMENTS */
252 : [OFS(JSON_NT_ARRAY_ELEMENTS)][JSON_TOKEN_ARRAY_START] = TD_ENTRY(JSON_PROD_ARRAY_ELEMENTS),
253 : [OFS(JSON_NT_ARRAY_ELEMENTS)][JSON_TOKEN_OBJECT_START] = TD_ENTRY(JSON_PROD_ARRAY_ELEMENTS),
254 : [OFS(JSON_NT_ARRAY_ELEMENTS)][JSON_TOKEN_STRING] = TD_ENTRY(JSON_PROD_ARRAY_ELEMENTS),
255 : [OFS(JSON_NT_ARRAY_ELEMENTS)][JSON_TOKEN_NUMBER] = TD_ENTRY(JSON_PROD_ARRAY_ELEMENTS),
256 : [OFS(JSON_NT_ARRAY_ELEMENTS)][JSON_TOKEN_TRUE] = TD_ENTRY(JSON_PROD_ARRAY_ELEMENTS),
257 : [OFS(JSON_NT_ARRAY_ELEMENTS)][JSON_TOKEN_FALSE] = TD_ENTRY(JSON_PROD_ARRAY_ELEMENTS),
258 : [OFS(JSON_NT_ARRAY_ELEMENTS)][JSON_TOKEN_NULL] = TD_ENTRY(JSON_PROD_ARRAY_ELEMENTS),
259 : [OFS(JSON_NT_ARRAY_ELEMENTS)][JSON_TOKEN_ARRAY_END] = TD_ENTRY(JSON_PROD_EPSILON),
260 : /* MORE_ARRAY_ELEMENTS */
261 : [OFS(JSON_NT_MORE_ARRAY_ELEMENTS)][JSON_TOKEN_COMMA] = TD_ENTRY(JSON_PROD_MORE_ARRAY_ELEMENTS),
262 : [OFS(JSON_NT_MORE_ARRAY_ELEMENTS)][JSON_TOKEN_ARRAY_END] = TD_ENTRY(JSON_PROD_EPSILON),
263 : /* KEY_PAIRS */
264 : [OFS(JSON_NT_KEY_PAIRS)][JSON_TOKEN_STRING] = TD_ENTRY(JSON_PROD_KEY_PAIRS),
265 : [OFS(JSON_NT_KEY_PAIRS)][JSON_TOKEN_OBJECT_END] = TD_ENTRY(JSON_PROD_EPSILON),
266 : /* MORE_KEY_PAIRS */
267 : [OFS(JSON_NT_MORE_KEY_PAIRS)][JSON_TOKEN_COMMA] = TD_ENTRY(JSON_PROD_MORE_KEY_PAIRS),
268 : [OFS(JSON_NT_MORE_KEY_PAIRS)][JSON_TOKEN_OBJECT_END] = TD_ENTRY(JSON_PROD_EPSILON),
269 : };
270 :
271 : /* the GOAL production. Not stored in the table, but will be the initial contents of the prediction stack */
272 : static char JSON_PROD_GOAL[] = {JSON_TOKEN_END, JSON_NT_JSON, 0};
273 :
274 : static inline JsonParseErrorType json_lex_string(JsonLexContext *lex);
275 : static inline JsonParseErrorType json_lex_number(JsonLexContext *lex, const char *s,
276 : bool *num_err, size_t *total_len);
277 : static inline JsonParseErrorType parse_scalar(JsonLexContext *lex, const JsonSemAction *sem);
278 : static JsonParseErrorType parse_object_field(JsonLexContext *lex, const JsonSemAction *sem);
279 : static JsonParseErrorType parse_object(JsonLexContext *lex, const JsonSemAction *sem);
280 : static JsonParseErrorType parse_array_element(JsonLexContext *lex, const JsonSemAction *sem);
281 : static JsonParseErrorType parse_array(JsonLexContext *lex, const JsonSemAction *sem);
282 : static JsonParseErrorType report_parse_error(JsonParseContext ctx, JsonLexContext *lex);
283 : static bool allocate_incremental_state(JsonLexContext *lex);
284 : static inline void set_fname(JsonLexContext *lex, char *fname);
285 :
286 : /* the null action object used for pure validation */
287 : const JsonSemAction nullSemAction =
288 : {
289 : NULL, NULL, NULL, NULL, NULL,
290 : NULL, NULL, NULL, NULL, NULL
291 : };
292 :
293 : /* sentinels used for out-of-memory conditions */
294 : static JsonLexContext failed_oom;
295 : static JsonIncrementalState failed_inc_oom;
296 :
297 : /* Parser support routines */
298 :
299 : /*
300 : * lex_peek
301 : *
302 : * what is the current look_ahead token?
303 : */
304 : static inline JsonTokenType
305 7811924 : lex_peek(JsonLexContext *lex)
306 : {
307 7811924 : return lex->token_type;
308 : }
309 :
310 : /*
311 : * lex_expect
312 : *
313 : * move the lexer to the next token if the current look_ahead token matches
314 : * the parameter token. Otherwise, report an error.
315 : */
316 : static inline JsonParseErrorType
317 315042 : lex_expect(JsonParseContext ctx, JsonLexContext *lex, JsonTokenType token)
318 : {
319 315042 : if (lex_peek(lex) == token)
320 314970 : return json_lex(lex);
321 : else
322 72 : return report_parse_error(ctx, lex);
323 : }
324 :
325 : /* chars to consider as part of an alphanumeric token */
326 : #define JSON_ALPHANUMERIC_CHAR(c) \
327 : (((c) >= 'a' && (c) <= 'z') || \
328 : ((c) >= 'A' && (c) <= 'Z') || \
329 : ((c) >= '0' && (c) <= '9') || \
330 : (c) == '_' || \
331 : IS_HIGHBIT_SET(c))
332 :
333 : /*
334 : * Utility function to check if a string is a valid JSON number.
335 : *
336 : * str is of length len, and need not be null-terminated.
337 : */
338 : bool
339 23 : IsValidJsonNumber(const char *str, size_t len)
340 : {
341 : bool numeric_error;
342 : size_t total_len;
343 23 : JsonLexContext dummy_lex = {0};
344 :
345 23 : if (len <= 0)
346 0 : return false;
347 :
348 : /*
349 : * json_lex_number expects a leading '-' to have been eaten already.
350 : *
351 : * having to cast away the constness of str is ugly, but there's not much
352 : * easy alternative.
353 : */
354 23 : if (*str == '-')
355 : {
356 2 : dummy_lex.input = str + 1;
357 2 : dummy_lex.input_length = len - 1;
358 : }
359 : else
360 : {
361 21 : dummy_lex.input = str;
362 21 : dummy_lex.input_length = len;
363 : }
364 :
365 23 : dummy_lex.token_start = dummy_lex.input;
366 :
367 23 : json_lex_number(&dummy_lex, dummy_lex.input, &numeric_error, &total_len);
368 :
369 23 : return (!numeric_error) && (total_len == dummy_lex.input_length);
370 : }
371 :
372 : /*
373 : * makeJsonLexContextCstringLen
374 : * Initialize the given JsonLexContext object, or create one
375 : *
376 : * If a valid 'lex' pointer is given, it is initialized. This can
377 : * be used for stack-allocated structs, saving overhead. If NULL is
378 : * given, a new struct is allocated.
379 : *
380 : * If need_escapes is true, ->strval stores the unescaped lexemes.
381 : * Unescaping is expensive, so only request it when necessary.
382 : *
383 : * If need_escapes is true or lex was given as NULL, then caller is
384 : * responsible for freeing the returned struct, either by calling
385 : * freeJsonLexContext() or (in backend environment) via memory context
386 : * cleanup.
387 : *
388 : * In shlib code, any out-of-memory failures will be deferred to time
389 : * of use; this function is guaranteed to return a valid JsonLexContext.
390 : */
391 : JsonLexContext *
392 27480 : makeJsonLexContextCstringLen(JsonLexContext *lex, const char *json,
393 : size_t len, int encoding, bool need_escapes)
394 : {
395 27480 : if (lex == NULL)
396 : {
397 4475 : lex = ALLOC0(sizeof(JsonLexContext));
398 4475 : if (!lex)
399 0 : return &failed_oom;
400 4475 : lex->flags |= JSONLEX_FREE_STRUCT;
401 : }
402 : else
403 23005 : memset(lex, 0, sizeof(JsonLexContext));
404 :
405 27480 : lex->errormsg = NULL;
406 27480 : lex->input = lex->token_terminator = lex->line_start = json;
407 27480 : lex->line_number = 1;
408 27480 : lex->input_length = len;
409 27480 : lex->input_encoding = encoding;
410 27480 : lex->need_escapes = need_escapes;
411 27480 : if (need_escapes)
412 : {
413 : /*
414 : * This call can fail in shlib code. We defer error handling to time
415 : * of use (json_lex_string()) since we might not need to parse any
416 : * strings anyway.
417 : */
418 21694 : lex->strval = jsonapi_makeStringInfo();
419 21694 : lex->flags |= JSONLEX_FREE_STRVAL;
420 : }
421 :
422 27480 : return lex;
423 : }
424 :
425 : /*
426 : * Allocates the internal bookkeeping structures for incremental parsing. This
427 : * can only fail in-band with shlib code.
428 : */
429 : #define JS_STACK_CHUNK_SIZE 64
430 : #define JS_MAX_PROD_LEN 10 /* more than we need */
431 : #define JSON_TD_MAX_STACK 6400 /* hard coded for now - this is a REALLY high
432 : * number */
433 : static bool
434 2083 : allocate_incremental_state(JsonLexContext *lex)
435 : {
436 : void *pstack,
437 : *prediction,
438 : *fnames,
439 : *fnull;
440 :
441 2083 : lex->inc_state = ALLOC0(sizeof(JsonIncrementalState));
442 2083 : pstack = ALLOC0(sizeof(JsonParserStack));
443 2083 : prediction = ALLOC(JS_STACK_CHUNK_SIZE * JS_MAX_PROD_LEN);
444 2083 : fnames = ALLOC(JS_STACK_CHUNK_SIZE * sizeof(char *));
445 2083 : fnull = ALLOC(JS_STACK_CHUNK_SIZE * sizeof(bool));
446 :
447 : #ifdef JSONAPI_USE_PQEXPBUFFER
448 980 : if (!lex->inc_state
449 980 : || !pstack
450 980 : || !prediction
451 980 : || !fnames
452 980 : || !fnull)
453 : {
454 0 : FREE(lex->inc_state);
455 0 : FREE(pstack);
456 0 : FREE(prediction);
457 0 : FREE(fnames);
458 0 : FREE(fnull);
459 :
460 0 : lex->inc_state = &failed_inc_oom;
461 0 : return false;
462 : }
463 : #endif
464 :
465 2083 : jsonapi_initStringInfo(&(lex->inc_state->partial_token));
466 2083 : lex->pstack = pstack;
467 2083 : lex->pstack->stack_size = JS_STACK_CHUNK_SIZE;
468 2083 : lex->pstack->prediction = prediction;
469 2083 : lex->pstack->fnames = fnames;
470 2083 : lex->pstack->fnull = fnull;
471 :
472 : /*
473 : * fnames between 0 and lex_level must always be defined so that
474 : * freeJsonLexContext() can handle them safely. inc/dec_lex_level() handle
475 : * the rest.
476 : */
477 : Assert(lex->lex_level == 0);
478 2083 : lex->pstack->fnames[0] = NULL;
479 :
480 2083 : lex->incremental = true;
481 2083 : return true;
482 : }
483 :
484 :
485 : /*
486 : * makeJsonLexContextIncremental
487 : *
488 : * Similar to above but set up for use in incremental parsing. That means we
489 : * need explicit stacks for predictions, field names and null indicators, but
490 : * we don't need the input, that will be handed in bit by bit to the
491 : * parse routine. We also need an accumulator for partial tokens in case
492 : * the boundary between chunks happens to fall in the middle of a token.
493 : *
494 : * In shlib code, any out-of-memory failures will be deferred to time of use;
495 : * this function is guaranteed to return a valid JsonLexContext.
496 : */
497 : JsonLexContext *
498 2083 : makeJsonLexContextIncremental(JsonLexContext *lex, int encoding,
499 : bool need_escapes)
500 : {
501 2083 : if (lex == NULL)
502 : {
503 1 : lex = ALLOC0(sizeof(JsonLexContext));
504 1 : if (!lex)
505 0 : return &failed_oom;
506 :
507 1 : lex->flags |= JSONLEX_FREE_STRUCT;
508 : }
509 : else
510 2082 : memset(lex, 0, sizeof(JsonLexContext));
511 :
512 2083 : lex->line_number = 1;
513 2083 : lex->input_encoding = encoding;
514 :
515 2083 : if (!allocate_incremental_state(lex))
516 : {
517 0 : if (lex->flags & JSONLEX_FREE_STRUCT)
518 : {
519 0 : FREE(lex);
520 0 : return &failed_oom;
521 : }
522 :
523 : /* lex->inc_state tracks the OOM failure; we can return here. */
524 0 : return lex;
525 : }
526 :
527 2083 : lex->need_escapes = need_escapes;
528 2083 : if (need_escapes)
529 : {
530 : /*
531 : * This call can fail in shlib code. We defer error handling to time
532 : * of use (json_lex_string()) since we might not need to parse any
533 : * strings anyway.
534 : */
535 126 : lex->strval = jsonapi_makeStringInfo();
536 126 : lex->flags |= JSONLEX_FREE_STRVAL;
537 : }
538 :
539 2083 : return lex;
540 : }
541 :
542 : void
543 1960 : setJsonLexContextOwnsTokens(JsonLexContext *lex, bool owned_by_context)
544 : {
545 1960 : if (lex->incremental && lex->inc_state->started)
546 : {
547 : /*
548 : * Switching this flag after parsing has already started is a
549 : * programming error.
550 : */
551 : Assert(false);
552 0 : return;
553 : }
554 :
555 1960 : if (owned_by_context)
556 980 : lex->flags |= JSONLEX_CTX_OWNS_TOKENS;
557 : else
558 980 : lex->flags &= ~JSONLEX_CTX_OWNS_TOKENS;
559 : }
560 :
561 : static inline bool
562 2598809 : inc_lex_level(JsonLexContext *lex)
563 : {
564 2598809 : if (lex->incremental && (lex->lex_level + 1) >= lex->pstack->stack_size)
565 : {
566 : size_t new_stack_size;
567 : char *new_prediction;
568 : char **new_fnames;
569 : bool *new_fnull;
570 :
571 38400 : new_stack_size = lex->pstack->stack_size + JS_STACK_CHUNK_SIZE;
572 :
573 38400 : new_prediction = REALLOC(lex->pstack->prediction,
574 : new_stack_size * JS_MAX_PROD_LEN);
575 : #ifdef JSONAPI_USE_PQEXPBUFFER
576 19200 : if (!new_prediction)
577 0 : return false;
578 : #endif
579 38400 : lex->pstack->prediction = new_prediction;
580 :
581 38400 : new_fnames = REALLOC(lex->pstack->fnames,
582 : new_stack_size * sizeof(char *));
583 : #ifdef JSONAPI_USE_PQEXPBUFFER
584 19200 : if (!new_fnames)
585 0 : return false;
586 : #endif
587 38400 : lex->pstack->fnames = new_fnames;
588 :
589 38400 : new_fnull = REALLOC(lex->pstack->fnull, new_stack_size * sizeof(bool));
590 : #ifdef JSONAPI_USE_PQEXPBUFFER
591 19200 : if (!new_fnull)
592 0 : return false;
593 : #endif
594 38400 : lex->pstack->fnull = new_fnull;
595 :
596 38400 : lex->pstack->stack_size = new_stack_size;
597 : }
598 :
599 2598809 : lex->lex_level += 1;
600 :
601 2598809 : if (lex->incremental)
602 : {
603 : /*
604 : * Ensure freeJsonLexContext() remains safe even if no fname is
605 : * assigned at this level.
606 : */
607 2598809 : lex->pstack->fnames[lex->lex_level] = NULL;
608 : }
609 :
610 2598809 : return true;
611 : }
612 :
613 : static inline void
614 959908 : dec_lex_level(JsonLexContext *lex)
615 : {
616 959908 : set_fname(lex, NULL); /* free the current level's fname, if needed */
617 959908 : lex->lex_level -= 1;
618 959908 : }
619 :
620 : static inline void
621 7498971 : push_prediction(JsonParserStack *pstack, td_entry entry)
622 : {
623 7498971 : memcpy(pstack->prediction + pstack->pred_index, entry.prod, entry.len);
624 7498971 : pstack->pred_index += entry.len;
625 7498971 : }
626 :
627 : static inline char
628 24221137 : pop_prediction(JsonParserStack *pstack)
629 : {
630 : Assert(pstack->pred_index > 0);
631 24221137 : return pstack->prediction[--pstack->pred_index];
632 : }
633 :
634 : static inline char
635 76 : next_prediction(JsonParserStack *pstack)
636 : {
637 : Assert(pstack->pred_index > 0);
638 76 : return pstack->prediction[pstack->pred_index - 1];
639 : }
640 :
641 : static inline bool
642 24523440 : have_prediction(JsonParserStack *pstack)
643 : {
644 24523440 : return pstack->pred_index > 0;
645 : }
646 :
647 : static inline void
648 1635398 : set_fname(JsonLexContext *lex, char *fname)
649 : {
650 1635398 : if (lex->flags & JSONLEX_CTX_OWNS_TOKENS)
651 : {
652 : /*
653 : * Don't leak prior fnames. If one hasn't been assigned yet,
654 : * inc_lex_level ensured that it's NULL (and therefore safe to free).
655 : */
656 436862 : FREE(lex->pstack->fnames[lex->lex_level]);
657 : }
658 :
659 1635398 : lex->pstack->fnames[lex->lex_level] = fname;
660 1635398 : }
661 :
662 : static inline char *
663 635734 : get_fname(JsonLexContext *lex)
664 : {
665 635734 : return lex->pstack->fnames[lex->lex_level];
666 : }
667 :
668 : static inline void
669 3267587 : set_fnull(JsonLexContext *lex, bool fnull)
670 : {
671 3267587 : lex->pstack->fnull[lex->lex_level] = fnull;
672 3267587 : }
673 :
674 : static inline bool
675 744 : get_fnull(JsonLexContext *lex)
676 : {
677 744 : return lex->pstack->fnull[lex->lex_level];
678 : }
679 :
680 : /*
681 : * Free memory in a JsonLexContext.
682 : *
683 : * There's no need for this if a *lex pointer was given when the object was
684 : * made, need_escapes was false, and json_errdetail() was not called; or if (in
685 : * backend environment) a memory context delete/reset is imminent.
686 : */
687 : void
688 6815 : freeJsonLexContext(JsonLexContext *lex)
689 : {
690 : static const JsonLexContext empty = {0};
691 :
692 6815 : if (!lex || lex == &failed_oom)
693 0 : return;
694 :
695 6815 : if (lex->flags & JSONLEX_FREE_STRVAL)
696 4600 : jsonapi_destroyStringInfo(lex->strval);
697 :
698 6815 : if (lex->errormsg)
699 953 : jsonapi_destroyStringInfo(lex->errormsg);
700 :
701 6815 : if (lex->incremental)
702 : {
703 2080 : jsonapi_termStringInfo(&lex->inc_state->partial_token);
704 2080 : FREE(lex->inc_state);
705 2080 : FREE(lex->pstack->prediction);
706 :
707 2080 : if (lex->flags & JSONLEX_CTX_OWNS_TOKENS)
708 : {
709 : int i;
710 :
711 : /* Clean up any tokens that were left behind. */
712 821410 : for (i = 0; i <= lex->lex_level; i++)
713 820430 : FREE(lex->pstack->fnames[i]);
714 : }
715 :
716 2080 : FREE(lex->pstack->fnames);
717 2080 : FREE(lex->pstack->fnull);
718 2080 : FREE(lex->pstack->scalar_val);
719 2080 : FREE(lex->pstack);
720 : }
721 :
722 6815 : if (lex->flags & JSONLEX_FREE_STRUCT)
723 4070 : FREE(lex);
724 : else
725 2745 : *lex = empty;
726 : }
727 :
728 : /*
729 : * pg_parse_json
730 : *
731 : * Publicly visible entry point for the JSON parser.
732 : *
733 : * lex is a lexing context, set up for the json to be processed by calling
734 : * makeJsonLexContext(). sem is a structure of function pointers to semantic
735 : * action routines to be called at appropriate spots during parsing, and a
736 : * pointer to a state object to be passed to those routines.
737 : *
738 : * If FORCE_JSON_PSTACK is defined then the routine will call the non-recursive
739 : * JSON parser. This is a useful way to validate that it's doing the right
740 : * thing at least for non-incremental cases. If this is on we expect to see
741 : * regression diffs relating to error messages about stack depth, but no
742 : * other differences.
743 : */
744 : JsonParseErrorType
745 26990 : pg_parse_json(JsonLexContext *lex, const JsonSemAction *sem)
746 : {
747 : #ifdef FORCE_JSON_PSTACK
748 : /*
749 : * We don't need partial token processing, there is only one chunk. But we
750 : * still need to init the partial token string so that freeJsonLexContext
751 : * works, so perform the full incremental initialization.
752 : */
753 : if (!allocate_incremental_state(lex))
754 : return JSON_OUT_OF_MEMORY;
755 :
756 : return pg_parse_json_incremental(lex, sem, lex->input, lex->input_length, true);
757 :
758 : #else
759 :
760 : JsonTokenType tok;
761 : JsonParseErrorType result;
762 :
763 26990 : if (lex == &failed_oom)
764 0 : return JSON_OUT_OF_MEMORY;
765 26990 : if (lex->incremental)
766 0 : return JSON_INVALID_LEXER_TYPE;
767 :
768 : /* get the initial token */
769 26990 : result = json_lex(lex);
770 26990 : if (result != JSON_SUCCESS)
771 168 : return result;
772 :
773 26822 : tok = lex_peek(lex);
774 :
775 : /* parse by recursive descent */
776 26822 : switch (tok)
777 : {
778 13859 : case JSON_TOKEN_OBJECT_START:
779 13859 : result = parse_object(lex, sem);
780 13806 : break;
781 5875 : case JSON_TOKEN_ARRAY_START:
782 5875 : result = parse_array(lex, sem);
783 5562 : break;
784 7088 : default:
785 7088 : result = parse_scalar(lex, sem); /* json can be a bare scalar */
786 : }
787 :
788 26400 : if (result == JSON_SUCCESS)
789 25762 : result = lex_expect(JSON_PARSE_END, lex, JSON_TOKEN_END);
790 :
791 26400 : return result;
792 : #endif
793 : }
794 :
795 : /*
796 : * json_count_array_elements
797 : *
798 : * Returns number of array elements in lex context at start of array token
799 : * until end of array token at same nesting level.
800 : *
801 : * Designed to be called from array_start routines.
802 : */
803 : JsonParseErrorType
804 4 : json_count_array_elements(JsonLexContext *lex, int *elements)
805 : {
806 : JsonLexContext copylex;
807 : int count;
808 : JsonParseErrorType result;
809 :
810 4 : if (lex == &failed_oom)
811 0 : return JSON_OUT_OF_MEMORY;
812 :
813 : /*
814 : * It's safe to do this with a shallow copy because the lexical routines
815 : * don't scribble on the input. They do scribble on the other pointers
816 : * etc, so doing this with a copy makes that safe.
817 : */
818 4 : memcpy(©lex, lex, sizeof(JsonLexContext));
819 4 : copylex.need_escapes = false; /* not interested in values here */
820 4 : copylex.lex_level++;
821 :
822 4 : count = 0;
823 4 : result = lex_expect(JSON_PARSE_ARRAY_START, ©lex,
824 : JSON_TOKEN_ARRAY_START);
825 4 : if (result != JSON_SUCCESS)
826 0 : return result;
827 4 : if (lex_peek(©lex) != JSON_TOKEN_ARRAY_END)
828 : {
829 : while (1)
830 : {
831 32 : count++;
832 32 : result = parse_array_element(©lex, &nullSemAction);
833 32 : if (result != JSON_SUCCESS)
834 0 : return result;
835 32 : if (copylex.token_type != JSON_TOKEN_COMMA)
836 4 : break;
837 28 : result = json_lex(©lex);
838 28 : if (result != JSON_SUCCESS)
839 0 : return result;
840 : }
841 : }
842 4 : result = lex_expect(JSON_PARSE_ARRAY_NEXT, ©lex,
843 : JSON_TOKEN_ARRAY_END);
844 4 : if (result != JSON_SUCCESS)
845 0 : return result;
846 :
847 4 : *elements = count;
848 4 : return JSON_SUCCESS;
849 : }
850 :
851 : /*
852 : * pg_parse_json_incremental
853 : *
854 : * Routine for incremental parsing of json. This uses the non-recursive top
855 : * down method of the Dragon Book Algorithm 4.3. It's somewhat slower than
856 : * the Recursive Descent pattern used above, so we only use it for incremental
857 : * parsing of JSON.
858 : *
859 : * The lexing context needs to be set up by a call to
860 : * makeJsonLexContextIncremental(). sem is a structure of function pointers
861 : * to semantic action routines, which should function exactly as those used
862 : * in the recursive descent parser.
863 : *
864 : * This routine can be called repeatedly with chunks of JSON. On the final
865 : * chunk is_last must be set to true. len is the length of the json chunk,
866 : * which does not need to be null terminated.
867 : */
868 : JsonParseErrorType
869 372954 : pg_parse_json_incremental(JsonLexContext *lex,
870 : const JsonSemAction *sem,
871 : const char *json,
872 : size_t len,
873 : bool is_last)
874 : {
875 : JsonTokenType tok;
876 : JsonParseErrorType result;
877 372954 : JsonParseContext ctx = JSON_PARSE_VALUE;
878 372954 : JsonParserStack *pstack = lex->pstack;
879 :
880 372954 : if (lex == &failed_oom || lex->inc_state == &failed_inc_oom)
881 0 : return JSON_OUT_OF_MEMORY;
882 372954 : if (!lex->incremental)
883 0 : return JSON_INVALID_LEXER_TYPE;
884 :
885 372954 : lex->input = lex->token_terminator = lex->line_start = json;
886 372954 : lex->input_length = len;
887 372954 : lex->inc_state->is_last_chunk = is_last;
888 372954 : lex->inc_state->started = true;
889 :
890 : /* get the initial token */
891 372954 : result = json_lex(lex);
892 372954 : if (result != JSON_SUCCESS)
893 71780 : return result;
894 :
895 301174 : tok = lex_peek(lex);
896 :
897 : /* use prediction stack for incremental parsing */
898 :
899 301174 : if (!have_prediction(pstack))
900 : {
901 1899 : td_entry goal = TD_ENTRY(JSON_PROD_GOAL);
902 :
903 1899 : push_prediction(pstack, goal);
904 : }
905 :
906 24222266 : while (have_prediction(pstack))
907 : {
908 24221137 : char top = pop_prediction(pstack);
909 : td_entry entry;
910 :
911 : /*
912 : * these first two branches are the guts of the Table Driven method
913 : */
914 24221137 : if (top == tok)
915 : {
916 : /*
917 : * tok can only be a terminal symbol, so top must be too. the
918 : * token matches the top of the stack, so get the next token.
919 : */
920 6251656 : if (tok < JSON_TOKEN_END)
921 : {
922 6250527 : result = json_lex(lex);
923 6250527 : if (result != JSON_SUCCESS)
924 300044 : return result;
925 5951180 : tok = lex_peek(lex);
926 : }
927 : }
928 17969481 : else if (IS_NT(top) && (entry = td_parser_table[OFS(top)][tok]).prod != NULL)
929 : {
930 : /*
931 : * the token is in the director set for a production of the
932 : * non-terminal at the top of the stack, so push the reversed RHS
933 : * of the production onto the stack.
934 : */
935 7497072 : push_prediction(pstack, entry);
936 : }
937 10472409 : else if (IS_SEM(top))
938 : {
939 : /*
940 : * top is a semantic action marker, so take action accordingly.
941 : * It's important to have these markers in the prediction stack
942 : * before any token they might need so we don't advance the token
943 : * prematurely. Note in a couple of cases we need to do something
944 : * both before and after the token.
945 : */
946 10471968 : switch (top)
947 : {
948 137783 : case JSON_SEM_OSTART:
949 : {
950 137783 : json_struct_action ostart = sem->object_start;
951 :
952 137783 : if (lex->lex_level >= JSON_TD_MAX_STACK)
953 0 : return JSON_NESTING_TOO_DEEP;
954 :
955 137783 : if (ostart != NULL)
956 : {
957 127107 : result = (*ostart) (sem->semstate);
958 127107 : if (result != JSON_SUCCESS)
959 0 : return result;
960 : }
961 :
962 137783 : if (!inc_lex_level(lex))
963 0 : return JSON_OUT_OF_MEMORY;
964 : }
965 137783 : break;
966 137418 : case JSON_SEM_OEND:
967 : {
968 137418 : json_struct_action oend = sem->object_end;
969 :
970 137418 : dec_lex_level(lex);
971 137418 : if (oend != NULL)
972 : {
973 127106 : result = (*oend) (sem->semstate);
974 127106 : if (result != JSON_SUCCESS)
975 0 : return result;
976 : }
977 : }
978 137418 : break;
979 2461282 : case JSON_SEM_ASTART:
980 : {
981 2461282 : json_struct_action astart = sem->array_start;
982 :
983 2461282 : if (lex->lex_level >= JSON_TD_MAX_STACK)
984 256 : return JSON_NESTING_TOO_DEEP;
985 :
986 2461026 : if (astart != NULL)
987 : {
988 286 : result = (*astart) (sem->semstate);
989 286 : if (result != JSON_SUCCESS)
990 0 : return result;
991 : }
992 :
993 2461026 : if (!inc_lex_level(lex))
994 0 : return JSON_OUT_OF_MEMORY;
995 : }
996 2461026 : break;
997 822490 : case JSON_SEM_AEND:
998 : {
999 822490 : json_struct_action aend = sem->array_end;
1000 :
1001 822490 : dec_lex_level(lex);
1002 822490 : if (aend != NULL)
1003 : {
1004 286 : result = (*aend) (sem->semstate);
1005 286 : if (result != JSON_SUCCESS)
1006 0 : return result;
1007 : }
1008 : }
1009 822490 : break;
1010 675490 : case JSON_SEM_OFIELD_INIT:
1011 : {
1012 : /*
1013 : * all we do here is save out the field name. We have
1014 : * to wait to get past the ':' to see if the next
1015 : * value is null so we can call the semantic routine
1016 : */
1017 675490 : char *fname = NULL;
1018 675490 : json_ofield_action ostart = sem->object_field_start;
1019 675490 : json_ofield_action oend = sem->object_field_end;
1020 :
1021 675490 : if ((ostart != NULL || oend != NULL) && lex->need_escapes)
1022 : {
1023 635110 : fname = STRDUP(lex->strval->data);
1024 635110 : if (fname == NULL)
1025 0 : return JSON_OUT_OF_MEMORY;
1026 : }
1027 675490 : set_fname(lex, fname);
1028 : }
1029 675490 : break;
1030 675362 : case JSON_SEM_OFIELD_START:
1031 : {
1032 : /*
1033 : * the current token should be the first token of the
1034 : * value
1035 : */
1036 675362 : bool isnull = tok == JSON_TOKEN_NULL;
1037 675362 : json_ofield_action ostart = sem->object_field_start;
1038 :
1039 675362 : set_fnull(lex, isnull);
1040 :
1041 675362 : if (ostart != NULL)
1042 : {
1043 635110 : char *fname = get_fname(lex);
1044 :
1045 635110 : result = (*ostart) (sem->semstate, fname, isnull);
1046 635110 : if (result != JSON_SUCCESS)
1047 0 : return result;
1048 : }
1049 : }
1050 675362 : break;
1051 675325 : case JSON_SEM_OFIELD_END:
1052 : {
1053 675325 : json_ofield_action oend = sem->object_field_end;
1054 :
1055 675325 : if (oend != NULL)
1056 : {
1057 624 : char *fname = get_fname(lex);
1058 624 : bool isnull = get_fnull(lex);
1059 :
1060 624 : result = (*oend) (sem->semstate, fname, isnull);
1061 624 : if (result != JSON_SUCCESS)
1062 0 : return result;
1063 : }
1064 : }
1065 675325 : break;
1066 2592225 : case JSON_SEM_AELEM_START:
1067 : {
1068 2592225 : json_aelem_action astart = sem->array_element_start;
1069 2592225 : bool isnull = tok == JSON_TOKEN_NULL;
1070 :
1071 2592225 : set_fnull(lex, isnull);
1072 :
1073 2592225 : if (astart != NULL)
1074 : {
1075 120 : result = (*astart) (sem->semstate, isnull);
1076 120 : if (result != JSON_SUCCESS)
1077 0 : return result;
1078 : }
1079 : }
1080 2592225 : break;
1081 953825 : case JSON_SEM_AELEM_END:
1082 : {
1083 953825 : json_aelem_action aend = sem->array_element_end;
1084 :
1085 953825 : if (aend != NULL)
1086 : {
1087 120 : bool isnull = get_fnull(lex);
1088 :
1089 120 : result = (*aend) (sem->semstate, isnull);
1090 120 : if (result != JSON_SUCCESS)
1091 0 : return result;
1092 : }
1093 : }
1094 953825 : break;
1095 670384 : case JSON_SEM_SCALAR_INIT:
1096 : {
1097 670384 : json_scalar_action sfunc = sem->scalar;
1098 :
1099 670384 : pstack->scalar_val = NULL;
1100 :
1101 670384 : if (sfunc != NULL)
1102 : {
1103 : /*
1104 : * extract the de-escaped string value, or the raw
1105 : * lexeme
1106 : */
1107 : /*
1108 : * XXX copied from RD parser but looks like a
1109 : * buglet
1110 : */
1111 634788 : if (tok == JSON_TOKEN_STRING)
1112 : {
1113 507599 : if (lex->need_escapes)
1114 : {
1115 507599 : pstack->scalar_val = STRDUP(lex->strval->data);
1116 507599 : if (pstack->scalar_val == NULL)
1117 0 : return JSON_OUT_OF_MEMORY;
1118 : }
1119 : }
1120 : else
1121 : {
1122 127189 : ptrdiff_t tlen = (lex->token_terminator - lex->token_start);
1123 :
1124 127189 : pstack->scalar_val = ALLOC(tlen + 1);
1125 127189 : if (pstack->scalar_val == NULL)
1126 0 : return JSON_OUT_OF_MEMORY;
1127 :
1128 127189 : memcpy(pstack->scalar_val, lex->token_start, tlen);
1129 127189 : pstack->scalar_val[tlen] = '\0';
1130 : }
1131 634788 : pstack->scalar_tok = tok;
1132 : }
1133 : }
1134 670384 : break;
1135 670384 : case JSON_SEM_SCALAR_CALL:
1136 : {
1137 : /*
1138 : * We'd like to be able to get rid of this business of
1139 : * two bits of scalar action, but we can't. It breaks
1140 : * certain semantic actions which expect that when
1141 : * called the lexer has consumed the item. See for
1142 : * example get_scalar() in jsonfuncs.c.
1143 : */
1144 670384 : json_scalar_action sfunc = sem->scalar;
1145 :
1146 670384 : if (sfunc != NULL)
1147 : {
1148 634788 : result = (*sfunc) (sem->semstate, pstack->scalar_val, pstack->scalar_tok);
1149 :
1150 : /*
1151 : * Either ownership of the token passed to the
1152 : * callback, or we need to free it now. Either
1153 : * way, clear our pointer to it so it doesn't get
1154 : * freed in the future.
1155 : */
1156 634787 : if (lex->flags & JSONLEX_CTX_OWNS_TOKENS)
1157 272 : FREE(pstack->scalar_val);
1158 634787 : pstack->scalar_val = NULL;
1159 :
1160 634787 : if (result != JSON_SUCCESS)
1161 0 : return result;
1162 : }
1163 : }
1164 670383 : break;
1165 0 : default:
1166 : /* should not happen */
1167 0 : break;
1168 : }
1169 : }
1170 : else
1171 : {
1172 : /*
1173 : * The token didn't match the stack top if it's a terminal nor a
1174 : * production for the stack top if it's a non-terminal.
1175 : *
1176 : * Various cases here are Asserted to be not possible, as the
1177 : * token would not appear at the top of the prediction stack
1178 : * unless the lookahead matched.
1179 : */
1180 441 : switch (top)
1181 : {
1182 76 : case JSON_TOKEN_STRING:
1183 76 : if (next_prediction(pstack) == JSON_TOKEN_COLON)
1184 76 : ctx = JSON_PARSE_STRING;
1185 : else
1186 : {
1187 : Assert(false);
1188 0 : ctx = JSON_PARSE_VALUE;
1189 : }
1190 76 : break;
1191 0 : case JSON_TOKEN_NUMBER:
1192 : case JSON_TOKEN_TRUE:
1193 : case JSON_TOKEN_FALSE:
1194 : case JSON_TOKEN_NULL:
1195 : case JSON_TOKEN_ARRAY_START:
1196 : case JSON_TOKEN_OBJECT_START:
1197 : Assert(false);
1198 0 : ctx = JSON_PARSE_VALUE;
1199 0 : break;
1200 0 : case JSON_TOKEN_ARRAY_END:
1201 : Assert(false);
1202 0 : ctx = JSON_PARSE_ARRAY_NEXT;
1203 0 : break;
1204 0 : case JSON_TOKEN_OBJECT_END:
1205 : Assert(false);
1206 0 : ctx = JSON_PARSE_OBJECT_NEXT;
1207 0 : break;
1208 0 : case JSON_TOKEN_COMMA:
1209 : Assert(false);
1210 0 : if (next_prediction(pstack) == JSON_TOKEN_STRING)
1211 0 : ctx = JSON_PARSE_OBJECT_NEXT;
1212 : else
1213 0 : ctx = JSON_PARSE_ARRAY_NEXT;
1214 0 : break;
1215 52 : case JSON_TOKEN_COLON:
1216 52 : ctx = JSON_PARSE_OBJECT_LABEL;
1217 52 : break;
1218 12 : case JSON_TOKEN_END:
1219 12 : ctx = JSON_PARSE_END;
1220 12 : break;
1221 36 : case JSON_NT_MORE_ARRAY_ELEMENTS:
1222 36 : ctx = JSON_PARSE_ARRAY_NEXT;
1223 36 : break;
1224 28 : case JSON_NT_ARRAY_ELEMENTS:
1225 28 : ctx = JSON_PARSE_ARRAY_START;
1226 28 : break;
1227 140 : case JSON_NT_MORE_KEY_PAIRS:
1228 140 : ctx = JSON_PARSE_OBJECT_NEXT;
1229 140 : break;
1230 60 : case JSON_NT_KEY_PAIRS:
1231 60 : ctx = JSON_PARSE_OBJECT_START;
1232 60 : break;
1233 37 : default:
1234 37 : ctx = JSON_PARSE_VALUE;
1235 : }
1236 441 : return report_parse_error(ctx, lex);
1237 : }
1238 : }
1239 :
1240 1129 : return JSON_SUCCESS;
1241 : }
1242 :
1243 : /*
1244 : * Recursive Descent parse routines. There is one for each structural
1245 : * element in a json document:
1246 : * - scalar (string, number, true, false, null)
1247 : * - array ( [ ] )
1248 : * - array element
1249 : * - object ( { } )
1250 : * - object field
1251 : */
1252 : static inline JsonParseErrorType
1253 224038 : parse_scalar(JsonLexContext *lex, const JsonSemAction *sem)
1254 : {
1255 224038 : char *val = NULL;
1256 224038 : json_scalar_action sfunc = sem->scalar;
1257 224038 : JsonTokenType tok = lex_peek(lex);
1258 : JsonParseErrorType result;
1259 :
1260 : /* a scalar must be a string, a number, true, false, or null */
1261 224038 : if (tok != JSON_TOKEN_STRING && tok != JSON_TOKEN_NUMBER &&
1262 22324 : tok != JSON_TOKEN_TRUE && tok != JSON_TOKEN_FALSE &&
1263 : tok != JSON_TOKEN_NULL)
1264 125 : return report_parse_error(JSON_PARSE_VALUE, lex);
1265 :
1266 : /* if no semantic function, just consume the token */
1267 223913 : if (sfunc == NULL)
1268 7989 : return json_lex(lex);
1269 :
1270 : /* extract the de-escaped string value, or the raw lexeme */
1271 215924 : if (lex_peek(lex) == JSON_TOKEN_STRING)
1272 : {
1273 49904 : if (lex->need_escapes)
1274 : {
1275 46104 : val = STRDUP(lex->strval->data);
1276 46104 : if (val == NULL)
1277 0 : return JSON_OUT_OF_MEMORY;
1278 : }
1279 : }
1280 : else
1281 : {
1282 166020 : int len = (lex->token_terminator - lex->token_start);
1283 :
1284 166020 : val = ALLOC(len + 1);
1285 166020 : if (val == NULL)
1286 0 : return JSON_OUT_OF_MEMORY;
1287 :
1288 166020 : memcpy(val, lex->token_start, len);
1289 166020 : val[len] = '\0';
1290 : }
1291 :
1292 : /* consume the token */
1293 215924 : result = json_lex(lex);
1294 215924 : if (result != JSON_SUCCESS)
1295 : {
1296 0 : FREE(val);
1297 0 : return result;
1298 : }
1299 :
1300 : /*
1301 : * invoke the callback, which may take ownership of val. For string
1302 : * values, val is NULL if need_escapes is false.
1303 : */
1304 215924 : result = (*sfunc) (sem->semstate, val, tok);
1305 :
1306 215769 : if (lex->flags & JSONLEX_CTX_OWNS_TOKENS)
1307 0 : FREE(val);
1308 :
1309 215769 : return result;
1310 : }
1311 :
1312 : static JsonParseErrorType
1313 216495 : parse_object_field(JsonLexContext *lex, const JsonSemAction *sem)
1314 : {
1315 : /*
1316 : * An object field is "fieldname" : value where value can be a scalar,
1317 : * object or array. Note: in user-facing docs and error messages, we
1318 : * generally call a field name a "key".
1319 : */
1320 :
1321 216495 : char *fname = NULL;
1322 216495 : json_ofield_action ostart = sem->object_field_start;
1323 216495 : json_ofield_action oend = sem->object_field_end;
1324 : bool isnull;
1325 : JsonTokenType tok;
1326 : JsonParseErrorType result;
1327 :
1328 216495 : if (lex_peek(lex) != JSON_TOKEN_STRING)
1329 8 : return report_parse_error(JSON_PARSE_STRING, lex);
1330 216487 : if ((ostart != NULL || oend != NULL) && lex->need_escapes)
1331 : {
1332 : /* fname is NULL if need_escapes is false */
1333 171247 : fname = STRDUP(lex->strval->data);
1334 171247 : if (fname == NULL)
1335 0 : return JSON_OUT_OF_MEMORY;
1336 : }
1337 216487 : result = json_lex(lex);
1338 216487 : if (result != JSON_SUCCESS)
1339 : {
1340 8 : FREE(fname);
1341 8 : return result;
1342 : }
1343 :
1344 216479 : result = lex_expect(JSON_PARSE_OBJECT_LABEL, lex, JSON_TOKEN_COLON);
1345 216479 : if (result != JSON_SUCCESS)
1346 : {
1347 96 : FREE(fname);
1348 96 : return result;
1349 : }
1350 :
1351 216383 : tok = lex_peek(lex);
1352 216383 : isnull = tok == JSON_TOKEN_NULL;
1353 :
1354 216383 : if (ostart != NULL)
1355 : {
1356 171155 : result = (*ostart) (sem->semstate, fname, isnull);
1357 171115 : if (result != JSON_SUCCESS)
1358 36 : goto ofield_cleanup;
1359 : }
1360 :
1361 216307 : switch (tok)
1362 : {
1363 7514 : case JSON_TOKEN_OBJECT_START:
1364 7514 : result = parse_object(lex, sem);
1365 2943 : break;
1366 10917 : case JSON_TOKEN_ARRAY_START:
1367 10917 : result = parse_array(lex, sem);
1368 10820 : break;
1369 197876 : default:
1370 197876 : result = parse_scalar(lex, sem);
1371 : }
1372 211584 : if (result != JSON_SUCCESS)
1373 196 : goto ofield_cleanup;
1374 :
1375 211388 : if (oend != NULL)
1376 : {
1377 122698 : result = (*oend) (sem->semstate, fname, isnull);
1378 122698 : if (result != JSON_SUCCESS)
1379 0 : goto ofield_cleanup;
1380 : }
1381 :
1382 211388 : ofield_cleanup:
1383 211620 : if (lex->flags & JSONLEX_CTX_OWNS_TOKENS)
1384 0 : FREE(fname);
1385 211620 : return result;
1386 : }
1387 :
1388 : static JsonParseErrorType
1389 34205 : parse_object(JsonLexContext *lex, const JsonSemAction *sem)
1390 : {
1391 : /*
1392 : * an object is a possibly empty sequence of object fields, separated by
1393 : * commas and surrounded by curly braces.
1394 : */
1395 34205 : json_struct_action ostart = sem->object_start;
1396 34205 : json_struct_action oend = sem->object_end;
1397 : JsonTokenType tok;
1398 : JsonParseErrorType result;
1399 :
1400 : #ifndef FRONTEND
1401 :
1402 : /*
1403 : * TODO: clients need some way to put a bound on stack growth. Parse level
1404 : * limits maybe?
1405 : */
1406 31060 : check_stack_depth();
1407 : #endif
1408 :
1409 34197 : if (ostart != NULL)
1410 : {
1411 21073 : result = (*ostart) (sem->semstate);
1412 21016 : if (result != JSON_SUCCESS)
1413 44 : return result;
1414 : }
1415 :
1416 : /*
1417 : * Data inside an object is at a higher nesting level than the object
1418 : * itself. Note that we increment this after we call the semantic routine
1419 : * for the object start and restore it before we call the routine for the
1420 : * object end.
1421 : */
1422 34096 : lex->lex_level++;
1423 :
1424 : Assert(lex_peek(lex) == JSON_TOKEN_OBJECT_START);
1425 34096 : result = json_lex(lex);
1426 34096 : if (result != JSON_SUCCESS)
1427 56 : return result;
1428 :
1429 34040 : tok = lex_peek(lex);
1430 34040 : switch (tok)
1431 : {
1432 32049 : case JSON_TOKEN_STRING:
1433 32049 : result = parse_object_field(lex, sem);
1434 211732 : while (result == JSON_SUCCESS && lex_peek(lex) == JSON_TOKEN_COMMA)
1435 : {
1436 184454 : result = json_lex(lex);
1437 184454 : if (result != JSON_SUCCESS)
1438 8 : break;
1439 184446 : result = parse_object_field(lex, sem);
1440 : }
1441 27286 : break;
1442 1982 : case JSON_TOKEN_OBJECT_END:
1443 1982 : break;
1444 9 : default:
1445 : /* case of an invalid initial token inside the object */
1446 9 : result = report_parse_error(JSON_PARSE_OBJECT_START, lex);
1447 : }
1448 29277 : if (result != JSON_SUCCESS)
1449 361 : return result;
1450 :
1451 28916 : result = lex_expect(JSON_PARSE_OBJECT_NEXT, lex, JSON_TOKEN_OBJECT_END);
1452 28916 : if (result != JSON_SUCCESS)
1453 24 : return result;
1454 :
1455 28892 : lex->lex_level--;
1456 :
1457 28892 : if (oend != NULL)
1458 : {
1459 16594 : result = (*oend) (sem->semstate);
1460 16529 : if (result != JSON_SUCCESS)
1461 40 : return result;
1462 : }
1463 :
1464 28787 : return JSON_SUCCESS;
1465 : }
1466 :
1467 : static JsonParseErrorType
1468 40378 : parse_array_element(JsonLexContext *lex, const JsonSemAction *sem)
1469 : {
1470 40378 : json_aelem_action astart = sem->array_element_start;
1471 40378 : json_aelem_action aend = sem->array_element_end;
1472 40378 : JsonTokenType tok = lex_peek(lex);
1473 : JsonParseErrorType result;
1474 : bool isnull;
1475 :
1476 40378 : isnull = tok == JSON_TOKEN_NULL;
1477 :
1478 40378 : if (astart != NULL)
1479 : {
1480 7561 : result = (*astart) (sem->semstate, isnull);
1481 7545 : if (result != JSON_SUCCESS)
1482 16 : return result;
1483 : }
1484 :
1485 : /* an array element is any object, array or scalar */
1486 40346 : switch (tok)
1487 : {
1488 12832 : case JSON_TOKEN_OBJECT_START:
1489 12832 : result = parse_object(lex, sem);
1490 12563 : break;
1491 8440 : case JSON_TOKEN_ARRAY_START:
1492 8440 : result = parse_array(lex, sem);
1493 2736 : break;
1494 19074 : default:
1495 19074 : result = parse_scalar(lex, sem);
1496 : }
1497 :
1498 34329 : if (result != JSON_SUCCESS)
1499 384 : return result;
1500 :
1501 33945 : if (aend != NULL)
1502 : {
1503 4876 : result = (*aend) (sem->semstate, isnull);
1504 4868 : if (result != JSON_SUCCESS)
1505 0 : return result;
1506 : }
1507 :
1508 33937 : return JSON_SUCCESS;
1509 : }
1510 :
1511 : static JsonParseErrorType
1512 25232 : parse_array(JsonLexContext *lex, const JsonSemAction *sem)
1513 : {
1514 : /*
1515 : * an array is a possibly empty sequence of array elements, separated by
1516 : * commas and surrounded by square brackets.
1517 : */
1518 25232 : json_struct_action astart = sem->array_start;
1519 25232 : json_struct_action aend = sem->array_end;
1520 : JsonParseErrorType result;
1521 :
1522 : #ifndef FRONTEND
1523 25204 : check_stack_depth();
1524 : #endif
1525 :
1526 25224 : if (astart != NULL)
1527 : {
1528 13255 : result = (*astart) (sem->semstate);
1529 13222 : if (result != JSON_SUCCESS)
1530 24 : return result;
1531 : }
1532 :
1533 : /*
1534 : * Data inside an array is at a higher nesting level than the array
1535 : * itself. Note that we increment this after we call the semantic routine
1536 : * for the array start and restore it before we call the routine for the
1537 : * array end.
1538 : */
1539 25167 : lex->lex_level++;
1540 :
1541 25167 : result = lex_expect(JSON_PARSE_ARRAY_START, lex, JSON_TOKEN_ARRAY_START);
1542 25167 : if (result == JSON_SUCCESS && lex_peek(lex) != JSON_TOKEN_ARRAY_END)
1543 : {
1544 20021 : result = parse_array_element(lex, sem);
1545 :
1546 34305 : while (result == JSON_SUCCESS && lex_peek(lex) == JSON_TOKEN_COMMA)
1547 : {
1548 20325 : result = json_lex(lex);
1549 20325 : if (result != JSON_SUCCESS)
1550 0 : break;
1551 20325 : result = parse_array_element(lex, sem);
1552 : }
1553 : }
1554 19126 : if (result != JSON_SUCCESS)
1555 416 : return result;
1556 :
1557 18710 : result = lex_expect(JSON_PARSE_ARRAY_NEXT, lex, JSON_TOKEN_ARRAY_END);
1558 18710 : if (result != JSON_SUCCESS)
1559 16 : return result;
1560 :
1561 18694 : lex->lex_level--;
1562 :
1563 18694 : if (aend != NULL)
1564 : {
1565 8220 : result = (*aend) (sem->semstate);
1566 8188 : if (result != JSON_SUCCESS)
1567 16 : return result;
1568 : }
1569 :
1570 18646 : return JSON_SUCCESS;
1571 : }
1572 :
1573 : /*
1574 : * Lex one token from the input stream.
1575 : *
1576 : * When doing incremental parsing, we can reach the end of the input string
1577 : * without having (or knowing we have) a complete token. If it's not the
1578 : * final chunk of input, the partial token is then saved to the lex
1579 : * structure's ptok StringInfo. On subsequent calls input is appended to this
1580 : * buffer until we have something that we think is a complete token,
1581 : * which is then lexed using a recursive call to json_lex. Processing then
1582 : * continues as normal on subsequent calls.
1583 : *
1584 : * Note than when doing incremental processing, the lex.prev_token_terminator
1585 : * should not be relied on. It could point into a previous input chunk or
1586 : * worse.
1587 : */
1588 : JsonParseErrorType
1589 7679962 : json_lex(JsonLexContext *lex)
1590 : {
1591 : const char *s;
1592 7679962 : const char *const end = lex->input + lex->input_length;
1593 : JsonParseErrorType result;
1594 :
1595 7679962 : if (lex == &failed_oom || lex->inc_state == &failed_inc_oom)
1596 0 : return JSON_OUT_OF_MEMORY;
1597 :
1598 7679962 : if (lex->incremental)
1599 : {
1600 6623481 : if (lex->inc_state->partial_completed)
1601 : {
1602 : /*
1603 : * We just lexed a completed partial token on the last call, so
1604 : * reset everything
1605 : */
1606 34432 : jsonapi_resetStringInfo(&(lex->inc_state->partial_token));
1607 34432 : lex->token_terminator = lex->input;
1608 34432 : lex->inc_state->partial_completed = false;
1609 : }
1610 :
1611 : #ifdef JSONAPI_USE_PQEXPBUFFER
1612 : /* Make sure our partial token buffer is valid before using it below. */
1613 1915760 : if (PQExpBufferDataBroken(lex->inc_state->partial_token))
1614 0 : return JSON_OUT_OF_MEMORY;
1615 : #endif
1616 : }
1617 :
1618 7679962 : s = lex->token_terminator;
1619 :
1620 7679962 : if (lex->incremental && lex->inc_state->partial_token.len)
1621 : {
1622 : /*
1623 : * We have a partial token. Extend it and if completed lex it by a
1624 : * recursive call
1625 : */
1626 83204 : jsonapi_StrValType *ptok = &(lex->inc_state->partial_token);
1627 83204 : size_t added = 0;
1628 83204 : bool tok_done = false;
1629 83204 : JsonLexContext dummy_lex = {0};
1630 : JsonParseErrorType partial_result;
1631 :
1632 83204 : if (ptok->data[0] == '"')
1633 : {
1634 : /*
1635 : * It's a string. Accumulate characters until we reach an
1636 : * unescaped '"'.
1637 : */
1638 80079 : int escapes = 0;
1639 :
1640 81363 : for (int i = ptok->len - 1; i > 0; i--)
1641 : {
1642 : /* count the trailing backslashes on the partial token */
1643 75674 : if (ptok->data[i] == '\\')
1644 1284 : escapes++;
1645 : else
1646 74390 : break;
1647 : }
1648 :
1649 607283 : for (size_t i = 0; i < lex->input_length; i++)
1650 : {
1651 559711 : char c = lex->input[i];
1652 :
1653 559711 : jsonapi_appendStringInfoCharMacro(ptok, c);
1654 559711 : added++;
1655 559711 : if (c == '"' && escapes % 2 == 0)
1656 : {
1657 32507 : tok_done = true;
1658 32507 : break;
1659 : }
1660 527204 : if (c == '\\')
1661 2048 : escapes++;
1662 : else
1663 525156 : escapes = 0;
1664 : }
1665 : }
1666 : else
1667 : {
1668 : /* not a string */
1669 3125 : char c = ptok->data[0];
1670 :
1671 3125 : if (c == '-' || (c >= '0' && c <= '9'))
1672 : {
1673 : /*
1674 : * Accumulate numeric continuations, respecting JSON number
1675 : * grammar: -? int [frac] [exp]
1676 : *
1677 : * We must track what parts of the number we've already seen
1678 : * so we don't over-consume. '.' is valid only once and not
1679 : * after 'e'/'E'; 'e'/'E' is valid only once; '+'/'-' are
1680 : * valid only immediately after 'e'/'E'.
1681 : */
1682 657 : bool numend = false;
1683 657 : bool seen_dot = false;
1684 657 : bool seen_exp = false;
1685 : char prev;
1686 :
1687 : /* Scan existing partial token for state */
1688 2518 : for (int j = 0; j < ptok->len; j++)
1689 : {
1690 1861 : char pc = ptok->data[j];
1691 :
1692 1861 : if (pc == '.')
1693 0 : seen_dot = true;
1694 1861 : else if (pc == 'e' || pc == 'E')
1695 0 : seen_exp = true;
1696 : }
1697 657 : prev = ptok->data[ptok->len - 1];
1698 :
1699 1842 : for (size_t i = 0; i < lex->input_length && !numend; i++)
1700 : {
1701 1185 : char cc = lex->input[i];
1702 :
1703 1185 : switch (cc)
1704 : {
1705 0 : case '+':
1706 : case '-':
1707 0 : if (prev != 'e' && prev != 'E')
1708 : {
1709 0 : numend = true;
1710 0 : break;
1711 : }
1712 0 : jsonapi_appendStringInfoCharMacro(ptok, cc);
1713 0 : added++;
1714 0 : break;
1715 0 : case '.':
1716 0 : if (seen_dot || seen_exp)
1717 : {
1718 0 : numend = true;
1719 0 : break;
1720 : }
1721 0 : seen_dot = true;
1722 0 : jsonapi_appendStringInfoCharMacro(ptok, cc);
1723 0 : added++;
1724 0 : break;
1725 0 : case 'e':
1726 : case 'E':
1727 0 : if (seen_exp)
1728 : {
1729 0 : numend = true;
1730 0 : break;
1731 : }
1732 0 : seen_exp = true;
1733 0 : jsonapi_appendStringInfoCharMacro(ptok, cc);
1734 0 : added++;
1735 0 : break;
1736 796 : case '0':
1737 : case '1':
1738 : case '2':
1739 : case '3':
1740 : case '4':
1741 : case '5':
1742 : case '6':
1743 : case '7':
1744 : case '8':
1745 : case '9':
1746 796 : jsonapi_appendStringInfoCharMacro(ptok, cc);
1747 796 : added++;
1748 796 : break;
1749 389 : default:
1750 389 : numend = true;
1751 : }
1752 1185 : if (!numend)
1753 796 : prev = cc;
1754 : }
1755 : }
1756 :
1757 : /*
1758 : * Add any remaining alphanumeric chars. This takes care of the
1759 : * {null, false, true} literals as well as any trailing
1760 : * alphanumeric junk on non-string tokens.
1761 : */
1762 6449 : for (size_t i = added; i < lex->input_length; i++)
1763 : {
1764 5469 : char cc = lex->input[i];
1765 :
1766 5469 : if (JSON_ALPHANUMERIC_CHAR(cc))
1767 : {
1768 3324 : jsonapi_appendStringInfoCharMacro(ptok, cc);
1769 3324 : added++;
1770 : }
1771 : else
1772 : {
1773 2145 : tok_done = true;
1774 2145 : break;
1775 : }
1776 : }
1777 3125 : if (added == lex->input_length &&
1778 980 : lex->inc_state->is_last_chunk)
1779 : {
1780 76 : tok_done = true;
1781 : }
1782 : }
1783 :
1784 83204 : if (!tok_done)
1785 : {
1786 : /* We should have consumed the whole chunk in this case. */
1787 : Assert(added == lex->input_length);
1788 :
1789 48476 : if (!lex->inc_state->is_last_chunk)
1790 48436 : return JSON_INCOMPLETE;
1791 :
1792 : /* json_errdetail() needs access to the accumulated token. */
1793 40 : lex->token_start = ptok->data;
1794 40 : lex->token_terminator = ptok->data + ptok->len;
1795 40 : return JSON_INVALID_TOKEN;
1796 : }
1797 :
1798 : /*
1799 : * Everything up to lex->input[added] has been added to the partial
1800 : * token, so move the input past it.
1801 : */
1802 34728 : lex->input += added;
1803 34728 : lex->input_length -= added;
1804 :
1805 34728 : dummy_lex.input = dummy_lex.token_terminator =
1806 34728 : dummy_lex.line_start = ptok->data;
1807 34728 : dummy_lex.line_number = lex->line_number;
1808 34728 : dummy_lex.input_length = ptok->len;
1809 34728 : dummy_lex.input_encoding = lex->input_encoding;
1810 34728 : dummy_lex.incremental = false;
1811 34728 : dummy_lex.need_escapes = lex->need_escapes;
1812 34728 : dummy_lex.strval = lex->strval;
1813 :
1814 34728 : partial_result = json_lex(&dummy_lex);
1815 :
1816 : /*
1817 : * We either have a complete token or an error. In either case we need
1818 : * to point to the partial token data for the semantic or error
1819 : * routines. If it's not an error we'll readjust on the next call to
1820 : * json_lex.
1821 : */
1822 34728 : lex->token_type = dummy_lex.token_type;
1823 34728 : lex->line_number = dummy_lex.line_number;
1824 :
1825 : /*
1826 : * We know the prev_token_terminator must be back in some previous
1827 : * piece of input, so we just make it NULL.
1828 : */
1829 34728 : lex->prev_token_terminator = NULL;
1830 :
1831 : /*
1832 : * Normally token_start would be ptok->data, but it could be later,
1833 : * see json_lex_string's handling of invalid escapes.
1834 : */
1835 34728 : lex->token_start = dummy_lex.token_start;
1836 34728 : lex->token_terminator = dummy_lex.token_terminator;
1837 34728 : if (partial_result == JSON_SUCCESS)
1838 : {
1839 : /* make sure we've used all the input */
1840 34620 : if (lex->token_terminator - lex->token_start != ptok->len)
1841 : {
1842 : Assert(false);
1843 0 : return JSON_INVALID_TOKEN;
1844 : }
1845 :
1846 34620 : lex->inc_state->partial_completed = true;
1847 : }
1848 34728 : return partial_result;
1849 : /* end of partial token processing */
1850 : }
1851 :
1852 : /* Skip leading whitespace. */
1853 12937929 : while (s < end && (*s == ' ' || *s == '\t' || *s == '\n' || *s == '\r'))
1854 : {
1855 5341171 : if (*s++ == '\n')
1856 : {
1857 344706 : ++lex->line_number;
1858 344706 : lex->line_start = s;
1859 : }
1860 : }
1861 7596758 : lex->token_start = s;
1862 :
1863 : /* Determine token type. */
1864 7596758 : if (s >= end)
1865 : {
1866 340541 : lex->token_start = NULL;
1867 340541 : lex->prev_token_terminator = lex->token_terminator;
1868 340541 : lex->token_terminator = s;
1869 340541 : lex->token_type = JSON_TOKEN_END;
1870 : }
1871 : else
1872 : {
1873 7256217 : switch (*s)
1874 : {
1875 : /* Single-character token, some kind of punctuation mark. */
1876 172214 : case '{':
1877 172214 : lex->prev_token_terminator = lex->token_terminator;
1878 172214 : lex->token_terminator = s + 1;
1879 172214 : lex->token_type = JSON_TOKEN_OBJECT_START;
1880 172214 : break;
1881 166421 : case '}':
1882 166421 : lex->prev_token_terminator = lex->token_terminator;
1883 166421 : lex->token_terminator = s + 1;
1884 166421 : lex->token_type = JSON_TOKEN_OBJECT_END;
1885 166421 : break;
1886 2486638 : case '[':
1887 2486638 : lex->prev_token_terminator = lex->token_terminator;
1888 2486638 : lex->token_terminator = s + 1;
1889 2486638 : lex->token_type = JSON_TOKEN_ARRAY_START;
1890 2486638 : break;
1891 841419 : case ']':
1892 841419 : lex->prev_token_terminator = lex->token_terminator;
1893 841419 : lex->token_terminator = s + 1;
1894 841419 : lex->token_type = JSON_TOKEN_ARRAY_END;
1895 841419 : break;
1896 875606 : case ',':
1897 875606 : lex->prev_token_terminator = lex->token_terminator;
1898 875606 : lex->token_terminator = s + 1;
1899 875606 : lex->token_type = JSON_TOKEN_COMMA;
1900 875606 : break;
1901 891841 : case ':':
1902 891841 : lex->prev_token_terminator = lex->token_terminator;
1903 891841 : lex->token_terminator = s + 1;
1904 891841 : lex->token_type = JSON_TOKEN_COLON;
1905 891841 : break;
1906 1511958 : case '"':
1907 : /* string */
1908 1511958 : result = json_lex_string(lex);
1909 1511958 : if (result != JSON_SUCCESS)
1910 32824 : return result;
1911 1479134 : lex->token_type = JSON_TOKEN_STRING;
1912 1479134 : break;
1913 425 : case '-':
1914 : /* Negative number. */
1915 425 : result = json_lex_number(lex, s + 1, NULL, NULL);
1916 425 : if (result != JSON_SUCCESS)
1917 4 : return result;
1918 421 : lex->token_type = JSON_TOKEN_NUMBER;
1919 421 : break;
1920 272509 : case '0':
1921 : case '1':
1922 : case '2':
1923 : case '3':
1924 : case '4':
1925 : case '5':
1926 : case '6':
1927 : case '7':
1928 : case '8':
1929 : case '9':
1930 : /* Positive number. */
1931 272509 : result = json_lex_number(lex, s, NULL, NULL);
1932 272509 : if (result != JSON_SUCCESS)
1933 461 : return result;
1934 272048 : lex->token_type = JSON_TOKEN_NUMBER;
1935 272048 : break;
1936 37186 : default:
1937 : {
1938 : const char *p;
1939 :
1940 : /*
1941 : * We're not dealing with a string, number, legal
1942 : * punctuation mark, or end of string. The only legal
1943 : * tokens we might find here are true, false, and null,
1944 : * but for error reporting purposes we scan until we see a
1945 : * non-alphanumeric character. That way, we can report
1946 : * the whole word as an unexpected token, rather than just
1947 : * some unintuitive prefix thereof.
1948 : */
1949 202519 : for (p = s; p < end && JSON_ALPHANUMERIC_CHAR(*p); p++)
1950 : /* skip */ ;
1951 :
1952 : /*
1953 : * We got some sort of unexpected punctuation or an
1954 : * otherwise unexpected character, so just complain about
1955 : * that one character.
1956 : */
1957 37186 : if (p == s)
1958 : {
1959 44 : lex->prev_token_terminator = lex->token_terminator;
1960 44 : lex->token_terminator = s + 1;
1961 44 : return JSON_INVALID_TOKEN;
1962 : }
1963 :
1964 37142 : if (lex->incremental && !lex->inc_state->is_last_chunk &&
1965 7832 : p == lex->input + lex->input_length)
1966 : {
1967 1832 : jsonapi_appendBinaryStringInfo(&(lex->inc_state->partial_token), s, end - s);
1968 1832 : return JSON_INCOMPLETE;
1969 : }
1970 :
1971 : /*
1972 : * We've got a real alphanumeric token here. If it
1973 : * happens to be true, false, or null, all is well. If
1974 : * not, error out.
1975 : */
1976 35310 : lex->prev_token_terminator = lex->token_terminator;
1977 35310 : lex->token_terminator = p;
1978 35310 : if (p - s == 4)
1979 : {
1980 15495 : if (memcmp(s, "true", 4) == 0)
1981 4972 : lex->token_type = JSON_TOKEN_TRUE;
1982 10523 : else if (memcmp(s, "null", 4) == 0)
1983 10515 : lex->token_type = JSON_TOKEN_NULL;
1984 : else
1985 8 : return JSON_INVALID_TOKEN;
1986 : }
1987 19815 : else if (p - s == 5 && memcmp(s, "false", 5) == 0)
1988 19656 : lex->token_type = JSON_TOKEN_FALSE;
1989 : else
1990 159 : return JSON_INVALID_TOKEN;
1991 : }
1992 : } /* end of switch */
1993 : }
1994 :
1995 7561426 : if (lex->incremental && lex->token_type == JSON_TOKEN_END && !lex->inc_state->is_last_chunk)
1996 287667 : return JSON_INCOMPLETE;
1997 : else
1998 7273759 : return JSON_SUCCESS;
1999 : }
2000 :
2001 : /*
2002 : * The next token in the input stream is known to be a string; lex it.
2003 : *
2004 : * If lex->strval isn't NULL, fill it with the decoded string.
2005 : * Set lex->token_terminator to the end of the decoded input, and in
2006 : * success cases, transfer its previous value to lex->prev_token_terminator.
2007 : * Return JSON_SUCCESS or an error code.
2008 : *
2009 : * Note: be careful that all error exits advance lex->token_terminator
2010 : * to the point after the character we detected the error on.
2011 : */
2012 : static inline JsonParseErrorType
2013 1511958 : json_lex_string(JsonLexContext *lex)
2014 : {
2015 : const char *s;
2016 1511958 : const char *const end = lex->input + lex->input_length;
2017 1511958 : int hi_surrogate = -1;
2018 :
2019 : /* Convenience macros for error exits */
2020 : #define FAIL_OR_INCOMPLETE_AT_CHAR_START(code) \
2021 : do { \
2022 : if (lex->incremental && !lex->inc_state->is_last_chunk) \
2023 : { \
2024 : jsonapi_appendBinaryStringInfo(&lex->inc_state->partial_token, \
2025 : lex->token_start, \
2026 : end - lex->token_start); \
2027 : return JSON_INCOMPLETE; \
2028 : } \
2029 : lex->token_terminator = s; \
2030 : return code; \
2031 : } while (0)
2032 : #define FAIL_AT_CHAR_END(code) \
2033 : do { \
2034 : ptrdiff_t remaining = end - s; \
2035 : int charlen; \
2036 : charlen = pg_encoding_mblen_or_incomplete(lex->input_encoding, \
2037 : s, remaining); \
2038 : lex->token_terminator = (charlen <= remaining) ? s + charlen : end; \
2039 : return code; \
2040 : } while (0)
2041 :
2042 1511958 : if (lex->need_escapes)
2043 : {
2044 : #ifdef JSONAPI_USE_PQEXPBUFFER
2045 : /* make sure initialization succeeded */
2046 668 : if (lex->strval == NULL)
2047 0 : return JSON_OUT_OF_MEMORY;
2048 : #endif
2049 1369084 : jsonapi_resetStringInfo(lex->strval);
2050 : }
2051 :
2052 : Assert(lex->input_length > 0);
2053 1511958 : s = lex->token_start;
2054 : for (;;)
2055 : {
2056 3024610 : s++;
2057 : /* Premature end of the string. */
2058 3024610 : if (s >= end)
2059 32335 : FAIL_OR_INCOMPLETE_AT_CHAR_START(JSON_INVALID_TOKEN);
2060 2992275 : else if (*s == '"')
2061 1479198 : break;
2062 1513077 : else if (*s == '\\')
2063 : {
2064 : /* OK, we have an escape character. */
2065 5498 : s++;
2066 5498 : if (s >= end)
2067 96 : FAIL_OR_INCOMPLETE_AT_CHAR_START(JSON_INVALID_TOKEN);
2068 5402 : else if (*s == 'u')
2069 : {
2070 : int i;
2071 2071 : int ch = 0;
2072 :
2073 9967 : for (i = 1; i <= 4; i++)
2074 : {
2075 8049 : s++;
2076 8049 : if (s >= end)
2077 128 : FAIL_OR_INCOMPLETE_AT_CHAR_START(JSON_INVALID_TOKEN);
2078 7921 : else if (*s >= '0' && *s <= '9')
2079 4899 : ch = (ch * 16) + (*s - '0');
2080 3022 : else if (*s >= 'a' && *s <= 'f')
2081 2981 : ch = (ch * 16) + (*s - 'a') + 10;
2082 41 : else if (*s >= 'A' && *s <= 'F')
2083 16 : ch = (ch * 16) + (*s - 'A') + 10;
2084 : else
2085 25 : FAIL_AT_CHAR_END(JSON_UNICODE_ESCAPE_FORMAT);
2086 : }
2087 1918 : if (lex->need_escapes)
2088 : {
2089 : /*
2090 : * Combine surrogate pairs.
2091 : */
2092 234 : if (is_utf16_surrogate_first(ch))
2093 : {
2094 114 : if (hi_surrogate != -1)
2095 8 : FAIL_AT_CHAR_END(JSON_UNICODE_HIGH_SURROGATE);
2096 106 : hi_surrogate = ch;
2097 106 : continue;
2098 : }
2099 120 : else if (is_utf16_surrogate_second(ch))
2100 : {
2101 42 : if (hi_surrogate == -1)
2102 16 : FAIL_AT_CHAR_END(JSON_UNICODE_LOW_SURROGATE);
2103 26 : ch = surrogate_pair_to_codepoint(hi_surrogate, ch);
2104 26 : hi_surrogate = -1;
2105 : }
2106 :
2107 104 : if (hi_surrogate != -1)
2108 0 : FAIL_AT_CHAR_END(JSON_UNICODE_LOW_SURROGATE);
2109 :
2110 : /*
2111 : * Reject invalid cases. We can't have a value above
2112 : * 0xFFFF here (since we only accepted 4 hex digits
2113 : * above), so no need to test for out-of-range chars.
2114 : */
2115 104 : if (ch == 0)
2116 : {
2117 : /* We can't allow this, since our TEXT type doesn't */
2118 16 : FAIL_AT_CHAR_END(JSON_UNICODE_CODE_POINT_ZERO);
2119 : }
2120 :
2121 : /*
2122 : * Add the represented character to lex->strval. In the
2123 : * backend, we can let pg_unicode_to_server_noerror()
2124 : * handle any required character set conversion; in
2125 : * frontend, we can only deal with trivial conversions.
2126 : */
2127 : #ifndef FRONTEND
2128 : {
2129 : char cbuf[MAX_UNICODE_EQUIVALENT_STRING + 1];
2130 :
2131 60 : if (!pg_unicode_to_server_noerror(ch, (unsigned char *) cbuf))
2132 0 : FAIL_AT_CHAR_END(JSON_UNICODE_UNTRANSLATABLE);
2133 60 : appendStringInfoString(lex->strval, cbuf);
2134 : }
2135 : #else
2136 28 : if (lex->input_encoding == PG_UTF8)
2137 : {
2138 : /* OK, we can map the code point to UTF8 easily */
2139 : char utf8str[5];
2140 : int utf8len;
2141 :
2142 28 : unicode_to_utf8(ch, (unsigned char *) utf8str);
2143 28 : utf8len = pg_utf_mblen((unsigned char *) utf8str);
2144 28 : jsonapi_appendBinaryStringInfo(lex->strval, utf8str, utf8len);
2145 : }
2146 0 : else if (ch <= 0x007f)
2147 : {
2148 : /* The ASCII range is the same in all encodings */
2149 0 : jsonapi_appendStringInfoChar(lex->strval, (char) ch);
2150 : }
2151 : else
2152 0 : FAIL_AT_CHAR_END(JSON_UNICODE_HIGH_ESCAPE);
2153 : #endif /* FRONTEND */
2154 : }
2155 : }
2156 3331 : else if (lex->need_escapes)
2157 : {
2158 518 : if (hi_surrogate != -1)
2159 0 : FAIL_AT_CHAR_END(JSON_UNICODE_LOW_SURROGATE);
2160 :
2161 518 : switch (*s)
2162 : {
2163 378 : case '"':
2164 : case '\\':
2165 : case '/':
2166 378 : jsonapi_appendStringInfoChar(lex->strval, *s);
2167 378 : break;
2168 28 : case 'b':
2169 28 : jsonapi_appendStringInfoChar(lex->strval, '\b');
2170 28 : break;
2171 4 : case 'f':
2172 4 : jsonapi_appendStringInfoChar(lex->strval, '\f');
2173 4 : break;
2174 40 : case 'n':
2175 40 : jsonapi_appendStringInfoChar(lex->strval, '\n');
2176 40 : break;
2177 4 : case 'r':
2178 4 : jsonapi_appendStringInfoChar(lex->strval, '\r');
2179 4 : break;
2180 60 : case 't':
2181 60 : jsonapi_appendStringInfoChar(lex->strval, '\t');
2182 60 : break;
2183 4 : default:
2184 :
2185 : /*
2186 : * Not a valid string escape, so signal error. We
2187 : * adjust token_start so that just the escape sequence
2188 : * is reported, not the whole string.
2189 : */
2190 4 : lex->token_start = s;
2191 4 : FAIL_AT_CHAR_END(JSON_ESCAPING_INVALID);
2192 : }
2193 : }
2194 2813 : else if (strchr("\"\\/bfnrt", *s) == NULL)
2195 : {
2196 : /*
2197 : * Simpler processing if we're not bothered about de-escaping
2198 : *
2199 : * It's very tempting to remove the strchr() call here and
2200 : * replace it with a switch statement, but testing so far has
2201 : * shown it's not a performance win.
2202 : */
2203 64 : lex->token_start = s;
2204 64 : FAIL_AT_CHAR_END(JSON_ESCAPING_INVALID);
2205 : }
2206 : }
2207 : else
2208 : {
2209 1507579 : const char *p = s;
2210 :
2211 1507579 : if (hi_surrogate != -1)
2212 8 : FAIL_AT_CHAR_END(JSON_UNICODE_LOW_SURROGATE);
2213 :
2214 : /*
2215 : * Skip to the first byte that requires special handling, so we
2216 : * can batch calls to jsonapi_appendBinaryStringInfo.
2217 : */
2218 1507571 : while (p < end - sizeof(Vector8) &&
2219 1856127 : !pg_lfind8('\\', (const uint8 *) p, sizeof(Vector8)) &&
2220 3806621 : !pg_lfind8('"', (const uint8 *) p, sizeof(Vector8)) &&
2221 445000 : !pg_lfind8_le(31, (const uint8 *) p, sizeof(Vector8)))
2222 445000 : p += sizeof(Vector8);
2223 :
2224 11606826 : for (; p < end; p++)
2225 : {
2226 11580084 : if (*p == '\\' || *p == '"')
2227 : break;
2228 10099315 : else if ((unsigned char) *p <= 31)
2229 : {
2230 : /* Per RFC4627, these characters MUST be escaped. */
2231 : /*
2232 : * Since *p isn't printable, exclude it from the context
2233 : * string
2234 : */
2235 60 : lex->token_terminator = p;
2236 60 : return JSON_ESCAPING_REQUIRED;
2237 : }
2238 : }
2239 :
2240 1507511 : if (lex->need_escapes)
2241 1369266 : jsonapi_appendBinaryStringInfo(lex->strval, s, p - s);
2242 :
2243 : /*
2244 : * s will be incremented at the top of the loop, so set it to just
2245 : * behind our lookahead position
2246 : */
2247 1507511 : s = p - 1;
2248 : }
2249 : }
2250 :
2251 1479198 : if (hi_surrogate != -1)
2252 : {
2253 64 : lex->token_terminator = s + 1;
2254 64 : return JSON_UNICODE_LOW_SURROGATE;
2255 : }
2256 :
2257 : #ifdef JSONAPI_USE_PQEXPBUFFER
2258 34448 : if (lex->need_escapes && PQExpBufferBroken(lex->strval))
2259 0 : return JSON_OUT_OF_MEMORY;
2260 : #endif
2261 :
2262 : /* Hooray, we found the end of the string! */
2263 1479134 : lex->prev_token_terminator = lex->token_terminator;
2264 1479134 : lex->token_terminator = s + 1;
2265 1479134 : return JSON_SUCCESS;
2266 :
2267 : #undef FAIL_OR_INCOMPLETE_AT_CHAR_START
2268 : #undef FAIL_AT_CHAR_END
2269 : }
2270 :
2271 : /*
2272 : * The next token in the input stream is known to be a number; lex it.
2273 : *
2274 : * In JSON, a number consists of four parts:
2275 : *
2276 : * (1) An optional minus sign ('-').
2277 : *
2278 : * (2) Either a single '0', or a string of one or more digits that does not
2279 : * begin with a '0'.
2280 : *
2281 : * (3) An optional decimal part, consisting of a period ('.') followed by
2282 : * one or more digits. (Note: While this part can be omitted
2283 : * completely, it's not OK to have only the decimal point without
2284 : * any digits afterwards.)
2285 : *
2286 : * (4) An optional exponent part, consisting of 'e' or 'E', optionally
2287 : * followed by '+' or '-', followed by one or more digits. (Note:
2288 : * As with the decimal part, if 'e' or 'E' is present, it must be
2289 : * followed by at least one digit.)
2290 : *
2291 : * The 's' argument to this function points to the ostensible beginning
2292 : * of part 2 - i.e. the character after any optional minus sign, or the
2293 : * first character of the string if there is none.
2294 : *
2295 : * If num_err is not NULL, we return an error flag to *num_err rather than
2296 : * raising an error for a badly-formed number. Also, if total_len is not NULL
2297 : * the distance from lex->input to the token end+1 is returned to *total_len.
2298 : */
2299 : static inline JsonParseErrorType
2300 272957 : json_lex_number(JsonLexContext *lex, const char *s,
2301 : bool *num_err, size_t *total_len)
2302 : {
2303 272957 : bool error = false;
2304 272957 : int len = s - lex->input;
2305 :
2306 : /* Part (1): leading sign indicator. */
2307 : /* Caller already did this for us; so do nothing. */
2308 :
2309 : /* Part (2): parse main digit string. */
2310 272957 : if (len < lex->input_length && *s == '0')
2311 : {
2312 82102 : s++;
2313 82102 : len++;
2314 : }
2315 190855 : else if (len < lex->input_length && *s >= '1' && *s <= '9')
2316 : {
2317 : do
2318 : {
2319 625381 : s++;
2320 625381 : len++;
2321 625381 : } while (len < lex->input_length && *s >= '0' && *s <= '9');
2322 : }
2323 : else
2324 5 : error = true;
2325 :
2326 : /* Part (3): parse optional decimal portion. */
2327 272957 : if (len < lex->input_length && *s == '.')
2328 : {
2329 30435 : s++;
2330 30435 : len++;
2331 30435 : if (len == lex->input_length || *s < '0' || *s > '9')
2332 8 : error = true;
2333 : else
2334 : {
2335 : do
2336 : {
2337 72449 : s++;
2338 72449 : len++;
2339 72449 : } while (len < lex->input_length && *s >= '0' && *s <= '9');
2340 : }
2341 : }
2342 :
2343 : /* Part (4): parse optional exponent. */
2344 272957 : if (len < lex->input_length && (*s == 'e' || *s == 'E'))
2345 : {
2346 61 : s++;
2347 61 : len++;
2348 61 : if (len < lex->input_length && (*s == '+' || *s == '-'))
2349 : {
2350 5 : s++;
2351 5 : len++;
2352 : }
2353 61 : if (len == lex->input_length || *s < '0' || *s > '9')
2354 8 : error = true;
2355 : else
2356 : {
2357 : do
2358 : {
2359 186 : s++;
2360 186 : len++;
2361 186 : } while (len < lex->input_length && *s >= '0' && *s <= '9');
2362 : }
2363 : }
2364 :
2365 : /*
2366 : * Check for trailing garbage. As in json_lex(), any alphanumeric stuff
2367 : * here should be considered part of the token for error-reporting
2368 : * purposes.
2369 : */
2370 273181 : for (; len < lex->input_length && JSON_ALPHANUMERIC_CHAR(*s); s++, len++)
2371 224 : error = true;
2372 :
2373 272957 : if (total_len != NULL)
2374 23 : *total_len = len;
2375 :
2376 272957 : if (lex->incremental && !lex->inc_state->is_last_chunk &&
2377 69039 : len >= lex->input_length)
2378 : {
2379 389 : jsonapi_appendBinaryStringInfo(&lex->inc_state->partial_token,
2380 389 : lex->token_start, s - lex->token_start);
2381 389 : if (num_err != NULL)
2382 0 : *num_err = error;
2383 :
2384 389 : return JSON_INCOMPLETE;
2385 : }
2386 272568 : else if (num_err != NULL)
2387 : {
2388 : /* let the caller handle any error */
2389 23 : *num_err = error;
2390 : }
2391 : else
2392 : {
2393 : /* return token endpoint */
2394 272545 : lex->prev_token_terminator = lex->token_terminator;
2395 272545 : lex->token_terminator = s;
2396 : /* handle error if any */
2397 272545 : if (error)
2398 76 : return JSON_INVALID_TOKEN;
2399 : }
2400 :
2401 272492 : return JSON_SUCCESS;
2402 : }
2403 :
2404 : /*
2405 : * Report a parse error.
2406 : *
2407 : * lex->token_start and lex->token_terminator must identify the current token.
2408 : */
2409 : static JsonParseErrorType
2410 655 : report_parse_error(JsonParseContext ctx, JsonLexContext *lex)
2411 : {
2412 : /* Handle case where the input ended prematurely. */
2413 655 : if (lex->token_start == NULL || lex->token_type == JSON_TOKEN_END)
2414 155 : return JSON_EXPECTED_MORE;
2415 :
2416 : /* Otherwise choose the error type based on the parsing context. */
2417 500 : switch (ctx)
2418 : {
2419 28 : case JSON_PARSE_END:
2420 28 : return JSON_EXPECTED_END;
2421 104 : case JSON_PARSE_VALUE:
2422 104 : return JSON_EXPECTED_JSON;
2423 84 : case JSON_PARSE_STRING:
2424 84 : return JSON_EXPECTED_STRING;
2425 28 : case JSON_PARSE_ARRAY_START:
2426 28 : return JSON_EXPECTED_ARRAY_FIRST;
2427 36 : case JSON_PARSE_ARRAY_NEXT:
2428 36 : return JSON_EXPECTED_ARRAY_NEXT;
2429 64 : case JSON_PARSE_OBJECT_START:
2430 64 : return JSON_EXPECTED_OBJECT_FIRST;
2431 68 : case JSON_PARSE_OBJECT_LABEL:
2432 68 : return JSON_EXPECTED_COLON;
2433 88 : case JSON_PARSE_OBJECT_NEXT:
2434 88 : return JSON_EXPECTED_OBJECT_NEXT;
2435 0 : case JSON_PARSE_OBJECT_COMMA:
2436 0 : return JSON_EXPECTED_STRING;
2437 : }
2438 :
2439 : /*
2440 : * We don't use a default: case, so that the compiler will warn about
2441 : * unhandled enum values.
2442 : */
2443 : Assert(false);
2444 0 : return JSON_SUCCESS; /* silence stupider compilers */
2445 : }
2446 :
2447 : /*
2448 : * Construct an (already translated) detail message for a JSON error.
2449 : *
2450 : * The returned pointer should not be freed, the allocation is either static
2451 : * or owned by the JsonLexContext.
2452 : */
2453 : char *
2454 1257 : json_errdetail(JsonParseErrorType error, JsonLexContext *lex)
2455 : {
2456 1257 : if (error == JSON_OUT_OF_MEMORY || lex == &failed_oom)
2457 : {
2458 : /* Short circuit. Allocating anything for this case is unhelpful. */
2459 0 : return _("out of memory");
2460 : }
2461 :
2462 1257 : if (lex->errormsg)
2463 0 : jsonapi_resetStringInfo(lex->errormsg);
2464 : else
2465 1257 : lex->errormsg = jsonapi_makeStringInfo();
2466 :
2467 : /*
2468 : * A helper for error messages that should print the current token. The
2469 : * format must contain exactly one %.*s specifier.
2470 : */
2471 : #define json_token_error(lex, format) \
2472 : jsonapi_appendStringInfo((lex)->errormsg, _(format), \
2473 : (int) ((lex)->token_terminator - (lex)->token_start), \
2474 : (lex)->token_start);
2475 :
2476 1257 : switch (error)
2477 : {
2478 0 : case JSON_INCOMPLETE:
2479 : case JSON_SUCCESS:
2480 : /* fall through to the error code after switch */
2481 0 : break;
2482 0 : case JSON_INVALID_LEXER_TYPE:
2483 0 : if (lex->incremental)
2484 0 : return _("Recursive descent parser cannot use incremental lexer.");
2485 : else
2486 0 : return _("Incremental parser requires incremental lexer.");
2487 256 : case JSON_NESTING_TOO_DEEP:
2488 256 : return (_("JSON nested too deep, maximum permitted depth is 6400."));
2489 68 : case JSON_ESCAPING_INVALID:
2490 68 : json_token_error(lex, "Escape sequence \"\\%.*s\" is invalid.");
2491 68 : break;
2492 60 : case JSON_ESCAPING_REQUIRED:
2493 60 : jsonapi_appendStringInfo(lex->errormsg,
2494 60 : _("Character with value 0x%02x must be escaped."),
2495 60 : (unsigned char) *(lex->token_terminator));
2496 60 : break;
2497 28 : case JSON_EXPECTED_END:
2498 28 : json_token_error(lex, "Expected end of input, but found \"%.*s\".");
2499 28 : break;
2500 28 : case JSON_EXPECTED_ARRAY_FIRST:
2501 28 : json_token_error(lex, "Expected array element or \"]\", but found \"%.*s\".");
2502 28 : break;
2503 36 : case JSON_EXPECTED_ARRAY_NEXT:
2504 36 : json_token_error(lex, "Expected \",\" or \"]\", but found \"%.*s\".");
2505 36 : break;
2506 68 : case JSON_EXPECTED_COLON:
2507 68 : json_token_error(lex, "Expected \":\", but found \"%.*s\".");
2508 68 : break;
2509 68 : case JSON_EXPECTED_JSON:
2510 68 : json_token_error(lex, "Expected JSON value, but found \"%.*s\".");
2511 68 : break;
2512 105 : case JSON_EXPECTED_MORE:
2513 105 : return _("The input string ended unexpectedly.");
2514 64 : case JSON_EXPECTED_OBJECT_FIRST:
2515 64 : json_token_error(lex, "Expected string or \"}\", but found \"%.*s\".");
2516 64 : break;
2517 88 : case JSON_EXPECTED_OBJECT_NEXT:
2518 88 : json_token_error(lex, "Expected \",\" or \"}\", but found \"%.*s\".");
2519 88 : break;
2520 84 : case JSON_EXPECTED_STRING:
2521 84 : json_token_error(lex, "Expected string, but found \"%.*s\".");
2522 84 : break;
2523 231 : case JSON_INVALID_TOKEN:
2524 231 : json_token_error(lex, "Token \"%.*s\" is invalid.");
2525 231 : break;
2526 0 : case JSON_OUT_OF_MEMORY:
2527 : /* should have been handled above; use the error path */
2528 0 : break;
2529 16 : case JSON_UNICODE_CODE_POINT_ZERO:
2530 16 : return _("\\u0000 cannot be converted to text.");
2531 25 : case JSON_UNICODE_ESCAPE_FORMAT:
2532 25 : return _("\"\\u\" must be followed by four hexadecimal digits.");
2533 0 : case JSON_UNICODE_HIGH_ESCAPE:
2534 : /* note: this case is only reachable in frontend not backend */
2535 0 : return _("Unicode escape values cannot be used for code point values above 007F when the encoding is not UTF8.");
2536 0 : case JSON_UNICODE_UNTRANSLATABLE:
2537 :
2538 : /*
2539 : * Note: this case is only reachable in backend and not frontend.
2540 : * #ifdef it away so the frontend doesn't try to link against
2541 : * backend functionality.
2542 : */
2543 : #ifndef FRONTEND
2544 0 : return psprintf(_("Unicode escape value could not be translated to the server's encoding %s."),
2545 : GetDatabaseEncodingName());
2546 : #else
2547 : Assert(false);
2548 0 : break;
2549 : #endif
2550 8 : case JSON_UNICODE_HIGH_SURROGATE:
2551 8 : return _("Unicode high surrogate must not follow a high surrogate.");
2552 24 : case JSON_UNICODE_LOW_SURROGATE:
2553 24 : return _("Unicode low surrogate must follow a high surrogate.");
2554 0 : case JSON_SEM_ACTION_FAILED:
2555 : /* fall through to the error code after switch */
2556 0 : break;
2557 : }
2558 : #undef json_token_error
2559 :
2560 : /* Note that lex->errormsg can be NULL in shlib code. */
2561 823 : if (lex->errormsg && lex->errormsg->len == 0)
2562 : {
2563 : /*
2564 : * We don't use a default: case, so that the compiler will warn about
2565 : * unhandled enum values. But this needs to be here anyway to cover
2566 : * the possibility of an incorrect input.
2567 : */
2568 0 : jsonapi_appendStringInfo(lex->errormsg,
2569 : "unexpected json parse error type: %d",
2570 : (int) error);
2571 : }
2572 :
2573 : #ifdef JSONAPI_USE_PQEXPBUFFER
2574 316 : if (PQExpBufferBroken(lex->errormsg))
2575 0 : return _("out of memory while constructing error description");
2576 : #endif
2577 :
2578 823 : return lex->errormsg->data;
2579 : }
|
