Line data Source code
1 : /*-------------------------------------------------------------------------
2 : *
3 : * jsonapi.c
4 : * JSON parser and lexer interfaces
5 : *
6 : * Portions Copyright (c) 1996-2025, 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 14315882 : lex_peek(JsonLexContext *lex)
306 : {
307 14315882 : 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 483278 : lex_expect(JsonParseContext ctx, JsonLexContext *lex, JsonTokenType token)
318 : {
319 483278 : if (lex_peek(lex) == token)
320 483170 : return json_lex(lex);
321 : else
322 108 : 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 46 : IsValidJsonNumber(const char *str, size_t len)
340 : {
341 : bool numeric_error;
342 : size_t total_len;
343 46 : JsonLexContext dummy_lex = {0};
344 :
345 46 : 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 46 : if (*str == '-')
355 : {
356 4 : dummy_lex.input = str + 1;
357 4 : dummy_lex.input_length = len - 1;
358 : }
359 : else
360 : {
361 42 : dummy_lex.input = str;
362 42 : dummy_lex.input_length = len;
363 : }
364 :
365 46 : dummy_lex.token_start = dummy_lex.input;
366 :
367 46 : json_lex_number(&dummy_lex, dummy_lex.input, &numeric_error, &total_len);
368 :
369 46 : 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 41362 : makeJsonLexContextCstringLen(JsonLexContext *lex, const char *json,
393 : size_t len, int encoding, bool need_escapes)
394 : {
395 41362 : if (lex == NULL)
396 : {
397 5530 : lex = ALLOC0(sizeof(JsonLexContext));
398 5530 : if (!lex)
399 0 : return &failed_oom;
400 5530 : lex->flags |= JSONLEX_FREE_STRUCT;
401 : }
402 : else
403 35832 : memset(lex, 0, sizeof(JsonLexContext));
404 :
405 41362 : lex->errormsg = NULL;
406 41362 : lex->input = lex->token_terminator = lex->line_start = json;
407 41362 : lex->line_number = 1;
408 41362 : lex->input_length = len;
409 41362 : lex->input_encoding = encoding;
410 41362 : lex->need_escapes = need_escapes;
411 41362 : 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 32778 : lex->strval = jsonapi_makeStringInfo();
419 32778 : lex->flags |= JSONLEX_FREE_STRVAL;
420 : }
421 :
422 41362 : 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 4154 : allocate_incremental_state(JsonLexContext *lex)
435 : {
436 : void *pstack,
437 : *prediction,
438 : *fnames,
439 : *fnull;
440 :
441 4154 : lex->inc_state = ALLOC0(sizeof(JsonIncrementalState));
442 4154 : pstack = ALLOC0(sizeof(JsonParserStack));
443 4154 : prediction = ALLOC(JS_STACK_CHUNK_SIZE * JS_MAX_PROD_LEN);
444 4154 : fnames = ALLOC(JS_STACK_CHUNK_SIZE * sizeof(char *));
445 4154 : fnull = ALLOC(JS_STACK_CHUNK_SIZE * sizeof(bool));
446 :
447 : #ifdef JSONAPI_USE_PQEXPBUFFER
448 1960 : if (!lex->inc_state
449 1960 : || !pstack
450 1960 : || !prediction
451 1960 : || !fnames
452 1960 : || !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 4154 : jsonapi_initStringInfo(&(lex->inc_state->partial_token));
466 4154 : lex->pstack = pstack;
467 4154 : lex->pstack->stack_size = JS_STACK_CHUNK_SIZE;
468 4154 : lex->pstack->prediction = prediction;
469 4154 : lex->pstack->fnames = fnames;
470 4154 : 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 4154 : lex->pstack->fnames[0] = NULL;
479 :
480 4154 : lex->incremental = true;
481 4154 : 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 4154 : makeJsonLexContextIncremental(JsonLexContext *lex, int encoding,
499 : bool need_escapes)
500 : {
501 4154 : if (lex == NULL)
502 : {
503 2 : lex = ALLOC0(sizeof(JsonLexContext));
504 2 : if (!lex)
505 0 : return &failed_oom;
506 :
507 2 : lex->flags |= JSONLEX_FREE_STRUCT;
508 : }
509 : else
510 4152 : memset(lex, 0, sizeof(JsonLexContext));
511 :
512 4154 : lex->line_number = 1;
513 4154 : lex->input_encoding = encoding;
514 :
515 4154 : 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 4154 : lex->need_escapes = need_escapes;
528 4154 : 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 : */
535 240 : lex->strval = jsonapi_makeStringInfo();
536 240 : lex->flags |= JSONLEX_FREE_STRVAL;
537 : }
538 :
539 4154 : return lex;
540 : }
541 :
542 : void
543 3920 : setJsonLexContextOwnsTokens(JsonLexContext *lex, bool owned_by_context)
544 : {
545 3920 : 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 3920 : if (owned_by_context)
556 1960 : lex->flags |= JSONLEX_CTX_OWNS_TOKENS;
557 : else
558 1960 : lex->flags &= ~JSONLEX_CTX_OWNS_TOKENS;
559 : }
560 :
561 : static inline bool
562 5170906 : inc_lex_level(JsonLexContext *lex)
563 : {
564 5170906 : 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 76800 : new_stack_size = lex->pstack->stack_size + JS_STACK_CHUNK_SIZE;
572 :
573 76800 : new_prediction = REALLOC(lex->pstack->prediction,
574 : new_stack_size * JS_MAX_PROD_LEN);
575 : #ifdef JSONAPI_USE_PQEXPBUFFER
576 38400 : if (!new_prediction)
577 0 : return false;
578 : #endif
579 76800 : lex->pstack->prediction = new_prediction;
580 :
581 76800 : new_fnames = REALLOC(lex->pstack->fnames,
582 : new_stack_size * sizeof(char *));
583 : #ifdef JSONAPI_USE_PQEXPBUFFER
584 38400 : if (!new_fnames)
585 0 : return false;
586 : #endif
587 76800 : lex->pstack->fnames = new_fnames;
588 :
589 76800 : new_fnull = REALLOC(lex->pstack->fnull, new_stack_size * sizeof(bool));
590 : #ifdef JSONAPI_USE_PQEXPBUFFER
591 38400 : if (!new_fnull)
592 0 : return false;
593 : #endif
594 76800 : lex->pstack->fnull = new_fnull;
595 :
596 76800 : lex->pstack->stack_size = new_stack_size;
597 : }
598 :
599 5170906 : lex->lex_level += 1;
600 :
601 5170906 : if (lex->incremental)
602 : {
603 : /*
604 : * Ensure freeJsonLexContext() remains safe even if no fname is
605 : * assigned at this level.
606 : */
607 5170906 : lex->pstack->fnames[lex->lex_level] = NULL;
608 : }
609 :
610 5170906 : return true;
611 : }
612 :
613 : static inline void
614 1893104 : dec_lex_level(JsonLexContext *lex)
615 : {
616 1893104 : set_fname(lex, NULL); /* free the current level's fname, if needed */
617 1893104 : lex->lex_level -= 1;
618 1893104 : }
619 :
620 : static inline void
621 14651022 : push_prediction(JsonParserStack *pstack, td_entry entry)
622 : {
623 14651022 : memcpy(pstack->prediction + pstack->pred_index, entry.prod, entry.len);
624 14651022 : pstack->pred_index += entry.len;
625 14651022 : }
626 :
627 : static inline char
628 46734518 : pop_prediction(JsonParserStack *pstack)
629 : {
630 : Assert(pstack->pred_index > 0);
631 46734518 : return pstack->prediction[--pstack->pred_index];
632 : }
633 :
634 : static inline char
635 152 : next_prediction(JsonParserStack *pstack)
636 : {
637 : Assert(pstack->pred_index > 0);
638 152 : return pstack->prediction[pstack->pred_index - 1];
639 : }
640 :
641 : static inline bool
642 47339084 : have_prediction(JsonParserStack *pstack)
643 : {
644 47339084 : return pstack->pred_index > 0;
645 : }
646 :
647 : static inline void
648 3110668 : set_fname(JsonLexContext *lex, char *fname)
649 : {
650 3110668 : 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 873724 : FREE(lex->pstack->fnames[lex->lex_level]);
657 : }
658 :
659 3110668 : lex->pstack->fnames[lex->lex_level] = fname;
660 3110668 : }
661 :
662 : static inline char *
663 1138052 : get_fname(JsonLexContext *lex)
664 : {
665 1138052 : return lex->pstack->fnames[lex->lex_level];
666 : }
667 :
668 : static inline void
669 6375082 : set_fnull(JsonLexContext *lex, bool fnull)
670 : {
671 6375082 : lex->pstack->fnull[lex->lex_level] = fnull;
672 6375082 : }
673 :
674 : static inline bool
675 1488 : get_fnull(JsonLexContext *lex)
676 : {
677 1488 : 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 10428 : freeJsonLexContext(JsonLexContext *lex)
689 : {
690 : static const JsonLexContext empty = {0};
691 :
692 10428 : if (!lex || lex == &failed_oom)
693 0 : return;
694 :
695 10428 : if (lex->flags & JSONLEX_FREE_STRVAL)
696 6126 : jsonapi_destroyStringInfo(lex->strval);
697 :
698 10428 : if (lex->errormsg)
699 1906 : jsonapi_destroyStringInfo(lex->errormsg);
700 :
701 10428 : if (lex->incremental)
702 : {
703 4148 : jsonapi_termStringInfo(&lex->inc_state->partial_token);
704 4148 : FREE(lex->inc_state);
705 4148 : FREE(lex->pstack->prediction);
706 :
707 4148 : if (lex->flags & JSONLEX_CTX_OWNS_TOKENS)
708 : {
709 : int i;
710 :
711 : /* Clean up any tokens that were left behind. */
712 1642820 : for (i = 0; i <= lex->lex_level; i++)
713 1640860 : FREE(lex->pstack->fnames[i]);
714 : }
715 :
716 4148 : FREE(lex->pstack->fnames);
717 4148 : FREE(lex->pstack->fnull);
718 4148 : FREE(lex->pstack->scalar_val);
719 4148 : FREE(lex->pstack);
720 : }
721 :
722 10428 : if (lex->flags & JSONLEX_FREE_STRUCT)
723 5310 : FREE(lex);
724 : else
725 5118 : *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 40624 : 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 40624 : if (lex == &failed_oom)
764 0 : return JSON_OUT_OF_MEMORY;
765 40624 : if (lex->incremental)
766 0 : return JSON_INVALID_LEXER_TYPE;
767 :
768 : /* get the initial token */
769 40624 : result = json_lex(lex);
770 40624 : if (result != JSON_SUCCESS)
771 246 : return result;
772 :
773 40378 : tok = lex_peek(lex);
774 :
775 : /* parse by recursive descent */
776 40378 : switch (tok)
777 : {
778 22806 : case JSON_TOKEN_OBJECT_START:
779 22806 : result = parse_object(lex, sem);
780 22738 : break;
781 7388 : case JSON_TOKEN_ARRAY_START:
782 7388 : result = parse_array(lex, sem);
783 7284 : break;
784 10184 : default:
785 10184 : result = parse_scalar(lex, sem); /* json can be a bare scalar */
786 : }
787 :
788 40134 : if (result == JSON_SUCCESS)
789 39690 : result = lex_expect(JSON_PARSE_END, lex, JSON_TOKEN_END);
790 :
791 40134 : 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 6 : json_count_array_elements(JsonLexContext *lex, int *elements)
805 : {
806 : JsonLexContext copylex;
807 : int count;
808 : JsonParseErrorType result;
809 :
810 6 : 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 6 : memcpy(©lex, lex, sizeof(JsonLexContext));
819 6 : copylex.need_escapes = false; /* not interested in values here */
820 6 : copylex.lex_level++;
821 :
822 6 : count = 0;
823 6 : result = lex_expect(JSON_PARSE_ARRAY_START, ©lex,
824 : JSON_TOKEN_ARRAY_START);
825 6 : if (result != JSON_SUCCESS)
826 0 : return result;
827 6 : if (lex_peek(©lex) != JSON_TOKEN_ARRAY_END)
828 : {
829 : while (1)
830 : {
831 48 : count++;
832 48 : result = parse_array_element(©lex, &nullSemAction);
833 48 : if (result != JSON_SUCCESS)
834 0 : return result;
835 48 : if (copylex.token_type != JSON_TOKEN_COMMA)
836 6 : break;
837 42 : result = json_lex(©lex);
838 42 : if (result != JSON_SUCCESS)
839 0 : return result;
840 : }
841 : }
842 6 : result = lex_expect(JSON_PARSE_ARRAY_NEXT, ©lex,
843 : JSON_TOKEN_ARRAY_END);
844 6 : if (result != JSON_SUCCESS)
845 0 : return result;
846 :
847 6 : *elements = count;
848 6 : 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 745880 : 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 745880 : JsonParseContext ctx = JSON_PARSE_VALUE;
878 745880 : JsonParserStack *pstack = lex->pstack;
879 :
880 745880 : if (lex == &failed_oom || lex->inc_state == &failed_inc_oom)
881 0 : return JSON_OUT_OF_MEMORY;
882 745880 : if (!lex->incremental)
883 0 : return JSON_INVALID_LEXER_TYPE;
884 :
885 745880 : lex->input = lex->token_terminator = lex->line_start = json;
886 745880 : lex->input_length = len;
887 745880 : lex->inc_state->is_last_chunk = is_last;
888 745880 : lex->inc_state->started = true;
889 :
890 : /* get the initial token */
891 745880 : result = json_lex(lex);
892 745880 : if (result != JSON_SUCCESS)
893 143560 : return result;
894 :
895 602320 : tok = lex_peek(lex);
896 :
897 : /* use prediction stack for incremental parsing */
898 :
899 602320 : if (!have_prediction(pstack))
900 : {
901 3786 : td_entry goal = TD_ENTRY(JSON_PROD_GOAL);
902 :
903 3786 : push_prediction(pstack, goal);
904 : }
905 :
906 46736764 : while (have_prediction(pstack))
907 : {
908 46734518 : 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 46734518 : 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 11916272 : if (tok < JSON_TOKEN_END)
921 : {
922 11914026 : result = json_lex(lex);
923 11914026 : if (result != JSON_SUCCESS)
924 600072 : return result;
925 11315348 : tok = lex_peek(lex);
926 : }
927 : }
928 34818246 : 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 14647236 : push_prediction(pstack, entry);
936 : }
937 20171010 : 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 20170128 : switch (top)
947 : {
948 248878 : case JSON_SEM_OSTART:
949 : {
950 248878 : json_struct_action ostart = sem->object_start;
951 :
952 248878 : if (lex->lex_level >= JSON_TD_MAX_STACK)
953 0 : return JSON_NESTING_TOO_DEEP;
954 :
955 248878 : if (ostart != NULL)
956 : {
957 227526 : result = (*ostart) (sem->semstate);
958 227526 : if (result != JSON_SUCCESS)
959 0 : return result;
960 : }
961 :
962 248878 : if (!inc_lex_level(lex))
963 0 : return JSON_OUT_OF_MEMORY;
964 : }
965 248878 : break;
966 248148 : case JSON_SEM_OEND:
967 : {
968 248148 : json_struct_action oend = sem->object_end;
969 :
970 248148 : dec_lex_level(lex);
971 248148 : if (oend != NULL)
972 : {
973 227524 : result = (*oend) (sem->semstate);
974 227524 : if (result != JSON_SUCCESS)
975 0 : return result;
976 : }
977 : }
978 248148 : break;
979 4922540 : case JSON_SEM_ASTART:
980 : {
981 4922540 : json_struct_action astart = sem->array_start;
982 :
983 4922540 : if (lex->lex_level >= JSON_TD_MAX_STACK)
984 512 : return JSON_NESTING_TOO_DEEP;
985 :
986 4922028 : if (astart != NULL)
987 : {
988 548 : result = (*astart) (sem->semstate);
989 548 : if (result != JSON_SUCCESS)
990 0 : return result;
991 : }
992 :
993 4922028 : if (!inc_lex_level(lex))
994 0 : return JSON_OUT_OF_MEMORY;
995 : }
996 4922028 : break;
997 1644956 : case JSON_SEM_AEND:
998 : {
999 1644956 : json_struct_action aend = sem->array_end;
1000 :
1001 1644956 : dec_lex_level(lex);
1002 1644956 : if (aend != NULL)
1003 : {
1004 548 : result = (*aend) (sem->semstate);
1005 548 : if (result != JSON_SUCCESS)
1006 0 : return result;
1007 : }
1008 : }
1009 1644956 : break;
1010 1217564 : 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 1217564 : char *fname = NULL;
1018 1217564 : json_ofield_action ostart = sem->object_field_start;
1019 1217564 : json_ofield_action oend = sem->object_field_end;
1020 :
1021 1217564 : if ((ostart != NULL || oend != NULL) && lex->need_escapes)
1022 : {
1023 1136804 : fname = STRDUP(lex->strval->data);
1024 1136804 : if (fname == NULL)
1025 0 : return JSON_OUT_OF_MEMORY;
1026 : }
1027 1217564 : set_fname(lex, fname);
1028 : }
1029 1217564 : break;
1030 1217308 : case JSON_SEM_OFIELD_START:
1031 : {
1032 : /*
1033 : * the current token should be the first token of the
1034 : * value
1035 : */
1036 1217308 : bool isnull = tok == JSON_TOKEN_NULL;
1037 1217308 : json_ofield_action ostart = sem->object_field_start;
1038 :
1039 1217308 : set_fnull(lex, isnull);
1040 :
1041 1217308 : if (ostart != NULL)
1042 : {
1043 1136804 : char *fname = get_fname(lex);
1044 :
1045 1136804 : result = (*ostart) (sem->semstate, fname, isnull);
1046 1136804 : if (result != JSON_SUCCESS)
1047 0 : return result;
1048 : }
1049 : }
1050 1217308 : break;
1051 1217234 : case JSON_SEM_OFIELD_END:
1052 : {
1053 1217234 : json_ofield_action oend = sem->object_field_end;
1054 :
1055 1217234 : if (oend != NULL)
1056 : {
1057 1248 : char *fname = get_fname(lex);
1058 1248 : bool isnull = get_fnull(lex);
1059 :
1060 1248 : result = (*oend) (sem->semstate, fname, isnull);
1061 1248 : if (result != JSON_SUCCESS)
1062 0 : return result;
1063 : }
1064 : }
1065 1217234 : break;
1066 5157774 : case JSON_SEM_AELEM_START:
1067 : {
1068 5157774 : json_aelem_action astart = sem->array_element_start;
1069 5157774 : bool isnull = tok == JSON_TOKEN_NULL;
1070 :
1071 5157774 : set_fnull(lex, isnull);
1072 :
1073 5157774 : if (astart != NULL)
1074 : {
1075 240 : result = (*astart) (sem->semstate, isnull);
1076 240 : if (result != JSON_SUCCESS)
1077 0 : return result;
1078 : }
1079 : }
1080 5157774 : break;
1081 1880974 : case JSON_SEM_AELEM_END:
1082 : {
1083 1880974 : json_aelem_action aend = sem->array_element_end;
1084 :
1085 1880974 : if (aend != NULL)
1086 : {
1087 240 : bool isnull = get_fnull(lex);
1088 :
1089 240 : result = (*aend) (sem->semstate, isnull);
1090 240 : if (result != JSON_SUCCESS)
1091 0 : return result;
1092 : }
1093 : }
1094 1880974 : break;
1095 1207376 : case JSON_SEM_SCALAR_INIT:
1096 : {
1097 1207376 : json_scalar_action sfunc = sem->scalar;
1098 :
1099 1207376 : pstack->scalar_val = NULL;
1100 :
1101 1207376 : 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 1136184 : if (tok == JSON_TOKEN_STRING)
1112 : {
1113 908506 : if (lex->need_escapes)
1114 : {
1115 908506 : pstack->scalar_val = STRDUP(lex->strval->data);
1116 908506 : if (pstack->scalar_val == NULL)
1117 0 : return JSON_OUT_OF_MEMORY;
1118 : }
1119 : }
1120 : else
1121 : {
1122 227678 : ptrdiff_t tlen = (lex->token_terminator - lex->token_start);
1123 :
1124 227678 : pstack->scalar_val = ALLOC(tlen + 1);
1125 227678 : if (pstack->scalar_val == NULL)
1126 0 : return JSON_OUT_OF_MEMORY;
1127 :
1128 227678 : memcpy(pstack->scalar_val, lex->token_start, tlen);
1129 227678 : pstack->scalar_val[tlen] = '\0';
1130 : }
1131 1136184 : pstack->scalar_tok = tok;
1132 : }
1133 : }
1134 1207376 : break;
1135 1207376 : 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 1207376 : json_scalar_action sfunc = sem->scalar;
1145 :
1146 1207376 : if (sfunc != NULL)
1147 : {
1148 1136184 : 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 1136182 : if (lex->flags & JSONLEX_CTX_OWNS_TOKENS)
1157 544 : FREE(pstack->scalar_val);
1158 1136182 : pstack->scalar_val = NULL;
1159 :
1160 1136182 : if (result != JSON_SUCCESS)
1161 0 : return result;
1162 : }
1163 : }
1164 1207374 : 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 882 : switch (top)
1181 : {
1182 152 : case JSON_TOKEN_STRING:
1183 152 : if (next_prediction(pstack) == JSON_TOKEN_COLON)
1184 152 : ctx = JSON_PARSE_STRING;
1185 : else
1186 : {
1187 : Assert(false);
1188 0 : ctx = JSON_PARSE_VALUE;
1189 : }
1190 152 : 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 104 : case JSON_TOKEN_COLON:
1216 104 : ctx = JSON_PARSE_OBJECT_LABEL;
1217 104 : break;
1218 24 : case JSON_TOKEN_END:
1219 24 : ctx = JSON_PARSE_END;
1220 24 : break;
1221 72 : case JSON_NT_MORE_ARRAY_ELEMENTS:
1222 72 : ctx = JSON_PARSE_ARRAY_NEXT;
1223 72 : break;
1224 56 : case JSON_NT_ARRAY_ELEMENTS:
1225 56 : ctx = JSON_PARSE_ARRAY_START;
1226 56 : break;
1227 280 : case JSON_NT_MORE_KEY_PAIRS:
1228 280 : ctx = JSON_PARSE_OBJECT_NEXT;
1229 280 : break;
1230 120 : case JSON_NT_KEY_PAIRS:
1231 120 : ctx = JSON_PARSE_OBJECT_START;
1232 120 : break;
1233 74 : default:
1234 74 : ctx = JSON_PARSE_VALUE;
1235 : }
1236 882 : return report_parse_error(ctx, lex);
1237 : }
1238 : }
1239 :
1240 2246 : 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 345504 : parse_scalar(JsonLexContext *lex, const JsonSemAction *sem)
1254 : {
1255 345504 : char *val = NULL;
1256 345504 : json_scalar_action sfunc = sem->scalar;
1257 345504 : JsonTokenType tok = lex_peek(lex);
1258 : JsonParseErrorType result;
1259 :
1260 : /* a scalar must be a string, a number, true, false, or null */
1261 345504 : if (tok != JSON_TOKEN_STRING && tok != JSON_TOKEN_NUMBER &&
1262 34030 : tok != JSON_TOKEN_TRUE && tok != JSON_TOKEN_FALSE &&
1263 : tok != JSON_TOKEN_NULL)
1264 188 : return report_parse_error(JSON_PARSE_VALUE, lex);
1265 :
1266 : /* if no semantic function, just consume the token */
1267 345316 : if (sfunc == NULL)
1268 11896 : return json_lex(lex);
1269 :
1270 : /* extract the de-escaped string value, or the raw lexeme */
1271 333420 : if (lex_peek(lex) == JSON_TOKEN_STRING)
1272 : {
1273 80220 : if (lex->need_escapes)
1274 : {
1275 74544 : val = STRDUP(lex->strval->data);
1276 74544 : if (val == NULL)
1277 0 : return JSON_OUT_OF_MEMORY;
1278 : }
1279 : }
1280 : else
1281 : {
1282 253200 : int len = (lex->token_terminator - lex->token_start);
1283 :
1284 253200 : val = ALLOC(len + 1);
1285 253200 : if (val == NULL)
1286 0 : return JSON_OUT_OF_MEMORY;
1287 :
1288 253200 : memcpy(val, lex->token_start, len);
1289 253200 : val[len] = '\0';
1290 : }
1291 :
1292 : /* consume the token */
1293 333420 : result = json_lex(lex);
1294 333420 : 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 333420 : result = (*sfunc) (sem->semstate, val, tok);
1305 :
1306 333324 : if (lex->flags & JSONLEX_CTX_OWNS_TOKENS)
1307 0 : FREE(val);
1308 :
1309 333324 : return result;
1310 : }
1311 :
1312 : static JsonParseErrorType
1313 335366 : 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 335366 : char *fname = NULL;
1322 335366 : json_ofield_action ostart = sem->object_field_start;
1323 335366 : json_ofield_action oend = sem->object_field_end;
1324 : bool isnull;
1325 : JsonTokenType tok;
1326 : JsonParseErrorType result;
1327 :
1328 335366 : if (lex_peek(lex) != JSON_TOKEN_STRING)
1329 12 : return report_parse_error(JSON_PARSE_STRING, lex);
1330 335354 : if ((ostart != NULL || oend != NULL) && lex->need_escapes)
1331 : {
1332 : /* fname is NULL if need_escapes is false */
1333 267134 : fname = STRDUP(lex->strval->data);
1334 267134 : if (fname == NULL)
1335 0 : return JSON_OUT_OF_MEMORY;
1336 : }
1337 335354 : result = json_lex(lex);
1338 335354 : if (result != JSON_SUCCESS)
1339 : {
1340 12 : FREE(fname);
1341 12 : return result;
1342 : }
1343 :
1344 335342 : result = lex_expect(JSON_PARSE_OBJECT_LABEL, lex, JSON_TOKEN_COLON);
1345 335342 : if (result != JSON_SUCCESS)
1346 : {
1347 90 : FREE(fname);
1348 90 : return result;
1349 : }
1350 :
1351 335252 : tok = lex_peek(lex);
1352 335252 : isnull = tok == JSON_TOKEN_NULL;
1353 :
1354 335252 : if (ostart != NULL)
1355 : {
1356 267050 : result = (*ostart) (sem->semstate, fname, isnull);
1357 267042 : if (result != JSON_SUCCESS)
1358 0 : goto ofield_cleanup;
1359 : }
1360 :
1361 335244 : switch (tok)
1362 : {
1363 11284 : case JSON_TOKEN_OBJECT_START:
1364 11284 : result = parse_object(lex, sem);
1365 4420 : break;
1366 15338 : case JSON_TOKEN_ARRAY_START:
1367 15338 : result = parse_array(lex, sem);
1368 15304 : break;
1369 308622 : default:
1370 308622 : result = parse_scalar(lex, sem);
1371 : }
1372 328340 : if (result != JSON_SUCCESS)
1373 42 : goto ofield_cleanup;
1374 :
1375 328298 : if (oend != NULL)
1376 : {
1377 183770 : result = (*oend) (sem->semstate, fname, isnull);
1378 183770 : if (result != JSON_SUCCESS)
1379 0 : goto ofield_cleanup;
1380 : }
1381 :
1382 328298 : ofield_cleanup:
1383 328340 : if (lex->flags & JSONLEX_CTX_OWNS_TOKENS)
1384 0 : FREE(fname);
1385 328340 : return result;
1386 : }
1387 :
1388 : static JsonParseErrorType
1389 53030 : 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 53030 : json_struct_action ostart = sem->object_start;
1396 53030 : 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 47106 : check_stack_depth();
1407 : #endif
1408 :
1409 53018 : if (ostart != NULL)
1410 : {
1411 33460 : result = (*ostart) (sem->semstate);
1412 33440 : if (result != JSON_SUCCESS)
1413 0 : 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 52998 : lex->lex_level++;
1423 :
1424 : Assert(lex_peek(lex) == JSON_TOKEN_OBJECT_START);
1425 52998 : result = json_lex(lex);
1426 52998 : if (result != JSON_SUCCESS)
1427 62 : return result;
1428 :
1429 52936 : tok = lex_peek(lex);
1430 52936 : switch (tok)
1431 : {
1432 49740 : case JSON_TOKEN_STRING:
1433 49740 : result = parse_object_field(lex, sem);
1434 328454 : while (result == JSON_SUCCESS && lex_peek(lex) == JSON_TOKEN_COMMA)
1435 : {
1436 285626 : result = json_lex(lex);
1437 285626 : if (result != JSON_SUCCESS)
1438 0 : break;
1439 285626 : result = parse_object_field(lex, sem);
1440 : }
1441 42828 : break;
1442 3182 : case JSON_TOKEN_OBJECT_END:
1443 3182 : break;
1444 14 : default:
1445 : /* case of an invalid initial token inside the object */
1446 14 : result = report_parse_error(JSON_PARSE_OBJECT_START, lex);
1447 : }
1448 46024 : if (result != JSON_SUCCESS)
1449 170 : return result;
1450 :
1451 45854 : result = lex_expect(JSON_PARSE_OBJECT_NEXT, lex, JSON_TOKEN_OBJECT_END);
1452 45854 : if (result != JSON_SUCCESS)
1453 36 : return result;
1454 :
1455 45818 : lex->lex_level--;
1456 :
1457 45818 : if (oend != NULL)
1458 : {
1459 27518 : result = (*oend) (sem->semstate);
1460 27472 : if (result != JSON_SUCCESS)
1461 0 : return result;
1462 : }
1463 :
1464 45772 : return JSON_SUCCESS;
1465 : }
1466 :
1467 : static JsonParseErrorType
1468 58506 : parse_array_element(JsonLexContext *lex, const JsonSemAction *sem)
1469 : {
1470 58506 : json_aelem_action astart = sem->array_element_start;
1471 58506 : json_aelem_action aend = sem->array_element_end;
1472 58506 : JsonTokenType tok = lex_peek(lex);
1473 : JsonParseErrorType result;
1474 : bool isnull;
1475 :
1476 58506 : isnull = tok == JSON_TOKEN_NULL;
1477 :
1478 58506 : if (astart != NULL)
1479 : {
1480 7804 : result = (*astart) (sem->semstate, isnull);
1481 7804 : if (result != JSON_SUCCESS)
1482 0 : return result;
1483 : }
1484 :
1485 : /* an array element is any object, array or scalar */
1486 58506 : switch (tok)
1487 : {
1488 18940 : case JSON_TOKEN_OBJECT_START:
1489 18940 : result = parse_object(lex, sem);
1490 18882 : break;
1491 12868 : case JSON_TOKEN_ARRAY_START:
1492 12868 : result = parse_array(lex, sem);
1493 4266 : break;
1494 26698 : default:
1495 26698 : result = parse_scalar(lex, sem);
1496 : }
1497 :
1498 49828 : if (result != JSON_SUCCESS)
1499 66 : return result;
1500 :
1501 49762 : if (aend != NULL)
1502 : {
1503 7228 : result = (*aend) (sem->semstate, isnull);
1504 7216 : if (result != JSON_SUCCESS)
1505 0 : return result;
1506 : }
1507 :
1508 49750 : return JSON_SUCCESS;
1509 : }
1510 :
1511 : static JsonParseErrorType
1512 35594 : 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 35594 : json_struct_action astart = sem->array_start;
1519 35594 : json_struct_action aend = sem->array_end;
1520 : JsonParseErrorType result;
1521 :
1522 : #ifndef FRONTEND
1523 35538 : check_stack_depth();
1524 : #endif
1525 :
1526 35582 : if (astart != NULL)
1527 : {
1528 17682 : result = (*astart) (sem->semstate);
1529 17668 : if (result != JSON_SUCCESS)
1530 0 : 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 35568 : lex->lex_level++;
1540 :
1541 35568 : result = lex_expect(JSON_PARSE_ARRAY_START, lex, JSON_TOKEN_ARRAY_START);
1542 35568 : if (result == JSON_SUCCESS && lex_peek(lex) != JSON_TOKEN_ARRAY_END)
1543 : {
1544 28094 : result = parse_array_element(lex, sem);
1545 :
1546 49768 : while (result == JSON_SUCCESS && lex_peek(lex) == JSON_TOKEN_COMMA)
1547 : {
1548 30364 : result = json_lex(lex);
1549 30364 : if (result != JSON_SUCCESS)
1550 0 : break;
1551 30364 : result = parse_array_element(lex, sem);
1552 : }
1553 : }
1554 26878 : if (result != JSON_SUCCESS)
1555 66 : return result;
1556 :
1557 26812 : result = lex_expect(JSON_PARSE_ARRAY_NEXT, lex, JSON_TOKEN_ARRAY_END);
1558 26812 : if (result != JSON_SUCCESS)
1559 24 : return result;
1560 :
1561 26788 : lex->lex_level--;
1562 :
1563 26788 : if (aend != NULL)
1564 : {
1565 11176 : result = (*aend) (sem->semstate);
1566 11152 : if (result != JSON_SUCCESS)
1567 0 : return result;
1568 : }
1569 :
1570 26764 : 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 14303646 : json_lex(JsonLexContext *lex)
1590 : {
1591 : const char *s;
1592 14303646 : const char *const end = lex->input + lex->input_length;
1593 : JsonParseErrorType result;
1594 :
1595 14303646 : if (lex == &failed_oom || lex->inc_state == &failed_inc_oom)
1596 0 : return JSON_OUT_OF_MEMORY;
1597 :
1598 14303646 : if (lex->incremental)
1599 : {
1600 12659906 : 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 68916 : jsonapi_resetStringInfo(&(lex->inc_state->partial_token));
1607 68916 : lex->token_terminator = lex->input;
1608 68916 : 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 3831520 : if (PQExpBufferDataBroken(lex->inc_state->partial_token))
1614 0 : return JSON_OUT_OF_MEMORY;
1615 : #endif
1616 : }
1617 :
1618 14303646 : s = lex->token_terminator;
1619 :
1620 14303646 : 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 166460 : jsonapi_StrValType *ptok = &(lex->inc_state->partial_token);
1627 166460 : size_t added = 0;
1628 166460 : bool tok_done = false;
1629 166460 : JsonLexContext dummy_lex = {0};
1630 : JsonParseErrorType partial_result;
1631 :
1632 166460 : if (ptok->data[0] == '"')
1633 : {
1634 : /*
1635 : * It's a string. Accumulate characters until we reach an
1636 : * unescaped '"'.
1637 : */
1638 160212 : int escapes = 0;
1639 :
1640 162780 : for (int i = ptok->len - 1; i > 0; i--)
1641 : {
1642 : /* count the trailing backslashes on the partial token */
1643 151404 : if (ptok->data[i] == '\\')
1644 2568 : escapes++;
1645 : else
1646 148836 : break;
1647 : }
1648 :
1649 1214332 : for (size_t i = 0; i < lex->input_length; i++)
1650 : {
1651 1119188 : char c = lex->input[i];
1652 :
1653 1119188 : jsonapi_appendStringInfoCharMacro(ptok, c);
1654 1119188 : added++;
1655 1119188 : if (c == '"' && escapes % 2 == 0)
1656 : {
1657 65068 : tok_done = true;
1658 65068 : break;
1659 : }
1660 1054120 : if (c == '\\')
1661 4096 : escapes++;
1662 : else
1663 1050024 : escapes = 0;
1664 : }
1665 : }
1666 : else
1667 : {
1668 : /* not a string */
1669 6248 : char c = ptok->data[0];
1670 :
1671 6248 : if (c == '-' || (c >= '0' && c <= '9'))
1672 : {
1673 : /* for numbers look for possible numeric continuations */
1674 :
1675 1312 : bool numend = false;
1676 :
1677 3680 : for (size_t i = 0; i < lex->input_length && !numend; i++)
1678 : {
1679 2368 : char cc = lex->input[i];
1680 :
1681 2368 : switch (cc)
1682 : {
1683 1592 : case '+':
1684 : case '-':
1685 : case 'e':
1686 : case 'E':
1687 : case '0':
1688 : case '1':
1689 : case '2':
1690 : case '3':
1691 : case '4':
1692 : case '5':
1693 : case '6':
1694 : case '7':
1695 : case '8':
1696 : case '9':
1697 : {
1698 1592 : jsonapi_appendStringInfoCharMacro(ptok, cc);
1699 1592 : added++;
1700 : }
1701 1592 : break;
1702 776 : default:
1703 776 : numend = true;
1704 : }
1705 : }
1706 : }
1707 :
1708 : /*
1709 : * Add any remaining alphanumeric chars. This takes care of the
1710 : * {null, false, true} literals as well as any trailing
1711 : * alphanumeric junk on non-string tokens.
1712 : */
1713 12896 : for (size_t i = added; i < lex->input_length; i++)
1714 : {
1715 10936 : char cc = lex->input[i];
1716 :
1717 10936 : if (JSON_ALPHANUMERIC_CHAR(cc))
1718 : {
1719 6648 : jsonapi_appendStringInfoCharMacro(ptok, cc);
1720 6648 : added++;
1721 : }
1722 : else
1723 : {
1724 4288 : tok_done = true;
1725 4288 : break;
1726 : }
1727 : }
1728 6248 : if (added == lex->input_length &&
1729 1960 : lex->inc_state->is_last_chunk)
1730 : {
1731 152 : tok_done = true;
1732 : }
1733 : }
1734 :
1735 166460 : if (!tok_done)
1736 : {
1737 : /* We should have consumed the whole chunk in this case. */
1738 : Assert(added == lex->input_length);
1739 :
1740 96952 : if (!lex->inc_state->is_last_chunk)
1741 96872 : return JSON_INCOMPLETE;
1742 :
1743 : /* json_errdetail() needs access to the accumulated token. */
1744 80 : lex->token_start = ptok->data;
1745 80 : lex->token_terminator = ptok->data + ptok->len;
1746 80 : return JSON_INVALID_TOKEN;
1747 : }
1748 :
1749 : /*
1750 : * Everything up to lex->input[added] has been added to the partial
1751 : * token, so move the input past it.
1752 : */
1753 69508 : lex->input += added;
1754 69508 : lex->input_length -= added;
1755 :
1756 69508 : dummy_lex.input = dummy_lex.token_terminator =
1757 69508 : dummy_lex.line_start = ptok->data;
1758 69508 : dummy_lex.line_number = lex->line_number;
1759 69508 : dummy_lex.input_length = ptok->len;
1760 69508 : dummy_lex.input_encoding = lex->input_encoding;
1761 69508 : dummy_lex.incremental = false;
1762 69508 : dummy_lex.need_escapes = lex->need_escapes;
1763 69508 : dummy_lex.strval = lex->strval;
1764 :
1765 69508 : partial_result = json_lex(&dummy_lex);
1766 :
1767 : /*
1768 : * We either have a complete token or an error. In either case we need
1769 : * to point to the partial token data for the semantic or error
1770 : * routines. If it's not an error we'll readjust on the next call to
1771 : * json_lex.
1772 : */
1773 69508 : lex->token_type = dummy_lex.token_type;
1774 69508 : lex->line_number = dummy_lex.line_number;
1775 :
1776 : /*
1777 : * We know the prev_token_terminator must be back in some previous
1778 : * piece of input, so we just make it NULL.
1779 : */
1780 69508 : lex->prev_token_terminator = NULL;
1781 :
1782 : /*
1783 : * Normally token_start would be ptok->data, but it could be later,
1784 : * see json_lex_string's handling of invalid escapes.
1785 : */
1786 69508 : lex->token_start = dummy_lex.token_start;
1787 69508 : lex->token_terminator = dummy_lex.token_terminator;
1788 69508 : if (partial_result == JSON_SUCCESS)
1789 : {
1790 : /* make sure we've used all the input */
1791 69292 : if (lex->token_terminator - lex->token_start != ptok->len)
1792 : {
1793 : Assert(false);
1794 0 : return JSON_INVALID_TOKEN;
1795 : }
1796 :
1797 69292 : lex->inc_state->partial_completed = true;
1798 : }
1799 69508 : return partial_result;
1800 : /* end of partial token processing */
1801 : }
1802 :
1803 : /* Skip leading whitespace. */
1804 22797998 : while (s < end && (*s == ' ' || *s == '\t' || *s == '\n' || *s == '\r'))
1805 : {
1806 8660812 : if (*s++ == '\n')
1807 : {
1808 573150 : ++lex->line_number;
1809 573150 : lex->line_start = s;
1810 : }
1811 : }
1812 14137186 : lex->token_start = s;
1813 :
1814 : /* Determine token type. */
1815 14137186 : if (s >= end)
1816 : {
1817 657210 : lex->token_start = NULL;
1818 657210 : lex->prev_token_terminator = lex->token_terminator;
1819 657210 : lex->token_terminator = s;
1820 657210 : lex->token_type = JSON_TOKEN_END;
1821 : }
1822 : else
1823 : {
1824 13479976 : switch (*s)
1825 : {
1826 : /* Single-character token, some kind of punctuation mark. */
1827 302260 : case '{':
1828 302260 : lex->prev_token_terminator = lex->token_terminator;
1829 302260 : lex->token_terminator = s + 1;
1830 302260 : lex->token_type = JSON_TOKEN_OBJECT_START;
1831 302260 : break;
1832 294130 : case '}':
1833 294130 : lex->prev_token_terminator = lex->token_terminator;
1834 294130 : lex->token_terminator = s + 1;
1835 294130 : lex->token_type = JSON_TOKEN_OBJECT_END;
1836 294130 : break;
1837 4958272 : case '[':
1838 4958272 : lex->prev_token_terminator = lex->token_terminator;
1839 4958272 : lex->token_terminator = s + 1;
1840 4958272 : lex->token_type = JSON_TOKEN_ARRAY_START;
1841 4958272 : break;
1842 1671918 : case ']':
1843 1671918 : lex->prev_token_terminator = lex->token_terminator;
1844 1671918 : lex->token_terminator = s + 1;
1845 1671918 : lex->token_type = JSON_TOKEN_ARRAY_END;
1846 1671918 : break;
1847 1524008 : case ',':
1848 1524008 : lex->prev_token_terminator = lex->token_terminator;
1849 1524008 : lex->token_terminator = s + 1;
1850 1524008 : lex->token_type = JSON_TOKEN_COMMA;
1851 1524008 : break;
1852 1552650 : case ':':
1853 1552650 : lex->prev_token_terminator = lex->token_terminator;
1854 1552650 : lex->token_terminator = s + 1;
1855 1552650 : lex->token_type = JSON_TOKEN_COLON;
1856 1552650 : break;
1857 2665378 : case '"':
1858 : /* string */
1859 2665378 : result = json_lex_string(lex);
1860 2665378 : if (result != JSON_SUCCESS)
1861 65526 : return result;
1862 2599852 : lex->token_type = JSON_TOKEN_STRING;
1863 2599852 : break;
1864 350 : case '-':
1865 : /* Negative number. */
1866 350 : result = json_lex_number(lex, s + 1, NULL, NULL);
1867 350 : if (result != JSON_SUCCESS)
1868 0 : return result;
1869 350 : lex->token_type = JSON_TOKEN_NUMBER;
1870 350 : break;
1871 448926 : case '0':
1872 : case '1':
1873 : case '2':
1874 : case '3':
1875 : case '4':
1876 : case '5':
1877 : case '6':
1878 : case '7':
1879 : case '8':
1880 : case '9':
1881 : /* Positive number. */
1882 448926 : result = json_lex_number(lex, s, NULL, NULL);
1883 448926 : if (result != JSON_SUCCESS)
1884 904 : return result;
1885 448022 : lex->token_type = JSON_TOKEN_NUMBER;
1886 448022 : break;
1887 62084 : default:
1888 : {
1889 : const char *p;
1890 :
1891 : /*
1892 : * We're not dealing with a string, number, legal
1893 : * punctuation mark, or end of string. The only legal
1894 : * tokens we might find here are true, false, and null,
1895 : * but for error reporting purposes we scan until we see a
1896 : * non-alphanumeric character. That way, we can report
1897 : * the whole word as an unexpected token, rather than just
1898 : * some unintuitive prefix thereof.
1899 : */
1900 335042 : for (p = s; p < end && JSON_ALPHANUMERIC_CHAR(*p); p++)
1901 : /* skip */ ;
1902 :
1903 : /*
1904 : * We got some sort of unexpected punctuation or an
1905 : * otherwise unexpected character, so just complain about
1906 : * that one character.
1907 : */
1908 62084 : if (p == s)
1909 : {
1910 80 : lex->prev_token_terminator = lex->token_terminator;
1911 80 : lex->token_terminator = s + 1;
1912 80 : return JSON_INVALID_TOKEN;
1913 : }
1914 :
1915 62004 : if (lex->incremental && !lex->inc_state->is_last_chunk &&
1916 15664 : p == lex->input + lex->input_length)
1917 : {
1918 3664 : jsonapi_appendBinaryStringInfo(&(lex->inc_state->partial_token), s, end - s);
1919 3664 : return JSON_INCOMPLETE;
1920 : }
1921 :
1922 : /*
1923 : * We've got a real alphanumeric token here. If it
1924 : * happens to be true, false, or null, all is well. If
1925 : * not, error out.
1926 : */
1927 58340 : lex->prev_token_terminator = lex->token_terminator;
1928 58340 : lex->token_terminator = p;
1929 58340 : if (p - s == 4)
1930 : {
1931 27014 : if (memcmp(s, "true", 4) == 0)
1932 8394 : lex->token_type = JSON_TOKEN_TRUE;
1933 18620 : else if (memcmp(s, "null", 4) == 0)
1934 18608 : lex->token_type = JSON_TOKEN_NULL;
1935 : else
1936 12 : return JSON_INVALID_TOKEN;
1937 : }
1938 31326 : else if (p - s == 5 && memcmp(s, "false", 5) == 0)
1939 31088 : lex->token_type = JSON_TOKEN_FALSE;
1940 : else
1941 238 : return JSON_INVALID_TOKEN;
1942 : }
1943 : } /* end of switch */
1944 : }
1945 :
1946 14066762 : if (lex->incremental && lex->token_type == JSON_TOKEN_END && !lex->inc_state->is_last_chunk)
1947 575266 : return JSON_INCOMPLETE;
1948 : else
1949 13491496 : return JSON_SUCCESS;
1950 : }
1951 :
1952 : /*
1953 : * The next token in the input stream is known to be a string; lex it.
1954 : *
1955 : * If lex->strval isn't NULL, fill it with the decoded string.
1956 : * Set lex->token_terminator to the end of the decoded input, and in
1957 : * success cases, transfer its previous value to lex->prev_token_terminator.
1958 : * Return JSON_SUCCESS or an error code.
1959 : *
1960 : * Note: be careful that all error exits advance lex->token_terminator
1961 : * to the point after the character we detected the error on.
1962 : */
1963 : static inline JsonParseErrorType
1964 2665378 : json_lex_string(JsonLexContext *lex)
1965 : {
1966 : const char *s;
1967 2665378 : const char *const end = lex->input + lex->input_length;
1968 2665378 : int hi_surrogate = -1;
1969 :
1970 : /* Convenience macros for error exits */
1971 : #define FAIL_OR_INCOMPLETE_AT_CHAR_START(code) \
1972 : do { \
1973 : if (lex->incremental && !lex->inc_state->is_last_chunk) \
1974 : { \
1975 : jsonapi_appendBinaryStringInfo(&lex->inc_state->partial_token, \
1976 : lex->token_start, \
1977 : end - lex->token_start); \
1978 : return JSON_INCOMPLETE; \
1979 : } \
1980 : lex->token_terminator = s; \
1981 : return code; \
1982 : } while (0)
1983 : #define FAIL_AT_CHAR_END(code) \
1984 : do { \
1985 : ptrdiff_t remaining = end - s; \
1986 : int charlen; \
1987 : charlen = pg_encoding_mblen_or_incomplete(lex->input_encoding, \
1988 : s, remaining); \
1989 : lex->token_terminator = (charlen <= remaining) ? s + charlen : end; \
1990 : return code; \
1991 : } while (0)
1992 :
1993 2665378 : if (lex->need_escapes)
1994 : {
1995 : #ifdef JSONAPI_USE_PQEXPBUFFER
1996 : /* make sure initialization succeeded */
1997 1336 : if (lex->strval == NULL)
1998 0 : return JSON_OUT_OF_MEMORY;
1999 : #endif
2000 2401070 : jsonapi_resetStringInfo(lex->strval);
2001 : }
2002 :
2003 : Assert(lex->input_length > 0);
2004 2665378 : s = lex->token_start;
2005 : for (;;)
2006 : {
2007 5331258 : s++;
2008 : /* Premature end of the string. */
2009 5331258 : if (s >= end)
2010 64720 : FAIL_OR_INCOMPLETE_AT_CHAR_START(JSON_INVALID_TOKEN);
2011 5266538 : else if (*s == '"')
2012 2599852 : break;
2013 2666686 : else if (*s == '\\')
2014 : {
2015 : /* OK, we have an escape character. */
2016 10258 : s++;
2017 10258 : if (s >= end)
2018 192 : FAIL_OR_INCOMPLETE_AT_CHAR_START(JSON_INVALID_TOKEN);
2019 10066 : else if (*s == 'u')
2020 : {
2021 : int i;
2022 3886 : int ch = 0;
2023 :
2024 18682 : for (i = 1; i <= 4; i++)
2025 : {
2026 15090 : s++;
2027 15090 : if (s >= end)
2028 256 : FAIL_OR_INCOMPLETE_AT_CHAR_START(JSON_INVALID_TOKEN);
2029 14834 : else if (*s >= '0' && *s <= '9')
2030 9230 : ch = (ch * 16) + (*s - '0');
2031 5604 : else if (*s >= 'a' && *s <= 'f')
2032 5542 : ch = (ch * 16) + (*s - 'a') + 10;
2033 62 : else if (*s >= 'A' && *s <= 'F')
2034 24 : ch = (ch * 16) + (*s - 'A') + 10;
2035 : else
2036 38 : FAIL_AT_CHAR_END(JSON_UNICODE_ESCAPE_FORMAT);
2037 : }
2038 3592 : if (lex->need_escapes)
2039 : {
2040 : /*
2041 : * Combine surrogate pairs.
2042 : */
2043 260 : if (is_utf16_surrogate_first(ch))
2044 : {
2045 72 : if (hi_surrogate != -1)
2046 12 : FAIL_AT_CHAR_END(JSON_UNICODE_HIGH_SURROGATE);
2047 60 : hi_surrogate = ch;
2048 60 : continue;
2049 : }
2050 188 : else if (is_utf16_surrogate_second(ch))
2051 : {
2052 60 : if (hi_surrogate == -1)
2053 24 : FAIL_AT_CHAR_END(JSON_UNICODE_LOW_SURROGATE);
2054 36 : ch = surrogate_pair_to_codepoint(hi_surrogate, ch);
2055 36 : hi_surrogate = -1;
2056 : }
2057 :
2058 164 : if (hi_surrogate != -1)
2059 0 : FAIL_AT_CHAR_END(JSON_UNICODE_LOW_SURROGATE);
2060 :
2061 : /*
2062 : * Reject invalid cases. We can't have a value above
2063 : * 0xFFFF here (since we only accepted 4 hex digits
2064 : * above), so no need to test for out-of-range chars.
2065 : */
2066 164 : if (ch == 0)
2067 : {
2068 : /* We can't allow this, since our TEXT type doesn't */
2069 24 : FAIL_AT_CHAR_END(JSON_UNICODE_CODE_POINT_ZERO);
2070 : }
2071 :
2072 : /*
2073 : * Add the represented character to lex->strval. In the
2074 : * backend, we can let pg_unicode_to_server_noerror()
2075 : * handle any required character set conversion; in
2076 : * frontend, we can only deal with trivial conversions.
2077 : */
2078 : #ifndef FRONTEND
2079 : {
2080 : char cbuf[MAX_UNICODE_EQUIVALENT_STRING + 1];
2081 :
2082 84 : if (!pg_unicode_to_server_noerror(ch, (unsigned char *) cbuf))
2083 0 : FAIL_AT_CHAR_END(JSON_UNICODE_UNTRANSLATABLE);
2084 84 : appendStringInfoString(lex->strval, cbuf);
2085 : }
2086 : #else
2087 56 : if (lex->input_encoding == PG_UTF8)
2088 : {
2089 : /* OK, we can map the code point to UTF8 easily */
2090 : char utf8str[5];
2091 : int utf8len;
2092 :
2093 56 : unicode_to_utf8(ch, (unsigned char *) utf8str);
2094 56 : utf8len = pg_utf_mblen((unsigned char *) utf8str);
2095 56 : jsonapi_appendBinaryStringInfo(lex->strval, utf8str, utf8len);
2096 : }
2097 0 : else if (ch <= 0x007f)
2098 : {
2099 : /* The ASCII range is the same in all encodings */
2100 0 : jsonapi_appendStringInfoChar(lex->strval, (char) ch);
2101 : }
2102 : else
2103 0 : FAIL_AT_CHAR_END(JSON_UNICODE_HIGH_ESCAPE);
2104 : #endif /* FRONTEND */
2105 : }
2106 : }
2107 6180 : else if (lex->need_escapes)
2108 : {
2109 606 : if (hi_surrogate != -1)
2110 0 : FAIL_AT_CHAR_END(JSON_UNICODE_LOW_SURROGATE);
2111 :
2112 606 : switch (*s)
2113 : {
2114 386 : case '"':
2115 : case '\\':
2116 : case '/':
2117 386 : jsonapi_appendStringInfoChar(lex->strval, *s);
2118 386 : break;
2119 44 : case 'b':
2120 44 : jsonapi_appendStringInfoChar(lex->strval, '\b');
2121 44 : break;
2122 8 : case 'f':
2123 8 : jsonapi_appendStringInfoChar(lex->strval, '\f');
2124 8 : break;
2125 62 : case 'n':
2126 62 : jsonapi_appendStringInfoChar(lex->strval, '\n');
2127 62 : break;
2128 8 : case 'r':
2129 8 : jsonapi_appendStringInfoChar(lex->strval, '\r');
2130 8 : break;
2131 92 : case 't':
2132 92 : jsonapi_appendStringInfoChar(lex->strval, '\t');
2133 92 : break;
2134 6 : default:
2135 :
2136 : /*
2137 : * Not a valid string escape, so signal error. We
2138 : * adjust token_start so that just the escape sequence
2139 : * is reported, not the whole string.
2140 : */
2141 6 : lex->token_start = s;
2142 6 : FAIL_AT_CHAR_END(JSON_ESCAPING_INVALID);
2143 : }
2144 : }
2145 5574 : else if (strchr("\"\\/bfnrt", *s) == NULL)
2146 : {
2147 : /*
2148 : * Simpler processing if we're not bothered about de-escaping
2149 : *
2150 : * It's very tempting to remove the strchr() call here and
2151 : * replace it with a switch statement, but testing so far has
2152 : * shown it's not a performance win.
2153 : */
2154 126 : lex->token_start = s;
2155 126 : FAIL_AT_CHAR_END(JSON_ESCAPING_INVALID);
2156 : }
2157 : }
2158 : else
2159 : {
2160 2656428 : const char *p = s;
2161 :
2162 2656428 : if (hi_surrogate != -1)
2163 12 : FAIL_AT_CHAR_END(JSON_UNICODE_LOW_SURROGATE);
2164 :
2165 : /*
2166 : * Skip to the first byte that requires special handling, so we
2167 : * can batch calls to jsonapi_appendBinaryStringInfo.
2168 : */
2169 3439220 : while (p < end - sizeof(Vector8) &&
2170 3256218 : !pg_lfind8('\\', (uint8 *) p, sizeof(Vector8)) &&
2171 3252600 : !pg_lfind8('"', (uint8 *) p, sizeof(Vector8)) &&
2172 782804 : !pg_lfind8_le(31, (uint8 *) p, sizeof(Vector8)))
2173 782804 : p += sizeof(Vector8);
2174 :
2175 20515516 : for (; p < end; p++)
2176 : {
2177 20461980 : if (*p == '\\' || *p == '"')
2178 : break;
2179 17859216 : else if ((unsigned char) *p <= 31)
2180 : {
2181 : /* Per RFC4627, these characters MUST be escaped. */
2182 : /*
2183 : * Since *p isn't printable, exclude it from the context
2184 : * string
2185 : */
2186 116 : lex->token_terminator = p;
2187 116 : return JSON_ESCAPING_REQUIRED;
2188 : }
2189 : }
2190 :
2191 2656300 : if (lex->need_escapes)
2192 2401262 : jsonapi_appendBinaryStringInfo(lex->strval, s, p - s);
2193 :
2194 : /*
2195 : * s will be incremented at the top of the loop, so set it to just
2196 : * behind our lookahead position
2197 : */
2198 2656300 : s = p - 1;
2199 : }
2200 : }
2201 :
2202 2599852 : if (hi_surrogate != -1)
2203 : {
2204 0 : lex->token_terminator = s + 1;
2205 0 : return JSON_UNICODE_LOW_SURROGATE;
2206 : }
2207 :
2208 : #ifdef JSONAPI_USE_PQEXPBUFFER
2209 68896 : if (lex->need_escapes && PQExpBufferBroken(lex->strval))
2210 0 : return JSON_OUT_OF_MEMORY;
2211 : #endif
2212 :
2213 : /* Hooray, we found the end of the string! */
2214 2599852 : lex->prev_token_terminator = lex->token_terminator;
2215 2599852 : lex->token_terminator = s + 1;
2216 2599852 : return JSON_SUCCESS;
2217 :
2218 : #undef FAIL_OR_INCOMPLETE_AT_CHAR_START
2219 : #undef FAIL_AT_CHAR_END
2220 : }
2221 :
2222 : /*
2223 : * The next token in the input stream is known to be a number; lex it.
2224 : *
2225 : * In JSON, a number consists of four parts:
2226 : *
2227 : * (1) An optional minus sign ('-').
2228 : *
2229 : * (2) Either a single '0', or a string of one or more digits that does not
2230 : * begin with a '0'.
2231 : *
2232 : * (3) An optional decimal part, consisting of a period ('.') followed by
2233 : * one or more digits. (Note: While this part can be omitted
2234 : * completely, it's not OK to have only the decimal point without
2235 : * any digits afterwards.)
2236 : *
2237 : * (4) An optional exponent part, consisting of 'e' or 'E', optionally
2238 : * followed by '+' or '-', followed by one or more digits. (Note:
2239 : * As with the decimal part, if 'e' or 'E' is present, it must be
2240 : * followed by at least one digit.)
2241 : *
2242 : * The 's' argument to this function points to the ostensible beginning
2243 : * of part 2 - i.e. the character after any optional minus sign, or the
2244 : * first character of the string if there is none.
2245 : *
2246 : * If num_err is not NULL, we return an error flag to *num_err rather than
2247 : * raising an error for a badly-formed number. Also, if total_len is not NULL
2248 : * the distance from lex->input to the token end+1 is returned to *total_len.
2249 : */
2250 : static inline JsonParseErrorType
2251 449322 : json_lex_number(JsonLexContext *lex, const char *s,
2252 : bool *num_err, size_t *total_len)
2253 : {
2254 449322 : bool error = false;
2255 449322 : int len = s - lex->input;
2256 :
2257 : /* Part (1): leading sign indicator. */
2258 : /* Caller already did this for us; so do nothing. */
2259 :
2260 : /* Part (2): parse main digit string. */
2261 449322 : if (len < lex->input_length && *s == '0')
2262 : {
2263 127500 : s++;
2264 127500 : len++;
2265 : }
2266 321822 : else if (len < lex->input_length && *s >= '1' && *s <= '9')
2267 : {
2268 : do
2269 : {
2270 1096260 : s++;
2271 1096260 : len++;
2272 1096260 : } while (len < lex->input_length && *s >= '0' && *s <= '9');
2273 : }
2274 : else
2275 2 : error = true;
2276 :
2277 : /* Part (3): parse optional decimal portion. */
2278 449322 : if (len < lex->input_length && *s == '.')
2279 : {
2280 45534 : s++;
2281 45534 : len++;
2282 45534 : if (len == lex->input_length || *s < '0' || *s > '9')
2283 12 : error = true;
2284 : else
2285 : {
2286 : do
2287 : {
2288 108028 : s++;
2289 108028 : len++;
2290 108028 : } while (len < lex->input_length && *s >= '0' && *s <= '9');
2291 : }
2292 : }
2293 :
2294 : /* Part (4): parse optional exponent. */
2295 449322 : if (len < lex->input_length && (*s == 'e' || *s == 'E'))
2296 : {
2297 94 : s++;
2298 94 : len++;
2299 94 : if (len < lex->input_length && (*s == '+' || *s == '-'))
2300 : {
2301 10 : s++;
2302 10 : len++;
2303 : }
2304 94 : if (len == lex->input_length || *s < '0' || *s > '9')
2305 12 : error = true;
2306 : else
2307 : {
2308 : do
2309 : {
2310 284 : s++;
2311 284 : len++;
2312 284 : } while (len < lex->input_length && *s >= '0' && *s <= '9');
2313 : }
2314 : }
2315 :
2316 : /*
2317 : * Check for trailing garbage. As in json_lex(), any alphanumeric stuff
2318 : * here should be considered part of the token for error-reporting
2319 : * purposes.
2320 : */
2321 449670 : for (; len < lex->input_length && JSON_ALPHANUMERIC_CHAR(*s); s++, len++)
2322 348 : error = true;
2323 :
2324 449322 : if (total_len != NULL)
2325 46 : *total_len = len;
2326 :
2327 449322 : if (lex->incremental && !lex->inc_state->is_last_chunk &&
2328 123542 : len >= lex->input_length)
2329 : {
2330 776 : jsonapi_appendBinaryStringInfo(&lex->inc_state->partial_token,
2331 776 : lex->token_start, s - lex->token_start);
2332 776 : if (num_err != NULL)
2333 0 : *num_err = error;
2334 :
2335 776 : return JSON_INCOMPLETE;
2336 : }
2337 448546 : else if (num_err != NULL)
2338 : {
2339 : /* let the caller handle any error */
2340 46 : *num_err = error;
2341 : }
2342 : else
2343 : {
2344 : /* return token endpoint */
2345 448500 : lex->prev_token_terminator = lex->token_terminator;
2346 448500 : lex->token_terminator = s;
2347 : /* handle error if any */
2348 448500 : if (error)
2349 128 : return JSON_INVALID_TOKEN;
2350 : }
2351 :
2352 448418 : return JSON_SUCCESS;
2353 : }
2354 :
2355 : /*
2356 : * Report a parse error.
2357 : *
2358 : * lex->token_start and lex->token_terminator must identify the current token.
2359 : */
2360 : static JsonParseErrorType
2361 1204 : report_parse_error(JsonParseContext ctx, JsonLexContext *lex)
2362 : {
2363 : /* Handle case where the input ended prematurely. */
2364 1204 : if (lex->token_start == NULL || lex->token_type == JSON_TOKEN_END)
2365 266 : return JSON_EXPECTED_MORE;
2366 :
2367 : /* Otherwise choose the error type based on the parsing context. */
2368 938 : switch (ctx)
2369 : {
2370 48 : case JSON_PARSE_END:
2371 48 : return JSON_EXPECTED_END;
2372 174 : case JSON_PARSE_VALUE:
2373 174 : return JSON_EXPECTED_JSON;
2374 164 : case JSON_PARSE_STRING:
2375 164 : return JSON_EXPECTED_STRING;
2376 56 : case JSON_PARSE_ARRAY_START:
2377 56 : return JSON_EXPECTED_ARRAY_FIRST;
2378 72 : case JSON_PARSE_ARRAY_NEXT:
2379 72 : return JSON_EXPECTED_ARRAY_NEXT;
2380 124 : case JSON_PARSE_OBJECT_START:
2381 124 : return JSON_EXPECTED_OBJECT_FIRST;
2382 128 : case JSON_PARSE_OBJECT_LABEL:
2383 128 : return JSON_EXPECTED_COLON;
2384 172 : case JSON_PARSE_OBJECT_NEXT:
2385 172 : return JSON_EXPECTED_OBJECT_NEXT;
2386 0 : case JSON_PARSE_OBJECT_COMMA:
2387 0 : return JSON_EXPECTED_STRING;
2388 : }
2389 :
2390 : /*
2391 : * We don't use a default: case, so that the compiler will warn about
2392 : * unhandled enum values.
2393 : */
2394 : Assert(false);
2395 0 : return JSON_SUCCESS; /* silence stupider compilers */
2396 : }
2397 :
2398 : /*
2399 : * Construct an (already translated) detail message for a JSON error.
2400 : *
2401 : * The returned pointer should not be freed, the allocation is either static
2402 : * or owned by the JsonLexContext.
2403 : */
2404 : char *
2405 2358 : json_errdetail(JsonParseErrorType error, JsonLexContext *lex)
2406 : {
2407 2358 : if (error == JSON_OUT_OF_MEMORY || lex == &failed_oom)
2408 : {
2409 : /* Short circuit. Allocating anything for this case is unhelpful. */
2410 0 : return _("out of memory");
2411 : }
2412 :
2413 2358 : if (lex->errormsg)
2414 0 : jsonapi_resetStringInfo(lex->errormsg);
2415 : else
2416 2358 : lex->errormsg = jsonapi_makeStringInfo();
2417 :
2418 : /*
2419 : * A helper for error messages that should print the current token. The
2420 : * format must contain exactly one %.*s specifier.
2421 : */
2422 : #define json_token_error(lex, format) \
2423 : jsonapi_appendStringInfo((lex)->errormsg, _(format), \
2424 : (int) ((lex)->token_terminator - (lex)->token_start), \
2425 : (lex)->token_start);
2426 :
2427 2358 : switch (error)
2428 : {
2429 0 : case JSON_INCOMPLETE:
2430 : case JSON_SUCCESS:
2431 : /* fall through to the error code after switch */
2432 0 : break;
2433 0 : case JSON_INVALID_LEXER_TYPE:
2434 0 : if (lex->incremental)
2435 0 : return _("Recursive descent parser cannot use incremental lexer.");
2436 : else
2437 0 : return _("Incremental parser requires incremental lexer.");
2438 512 : case JSON_NESTING_TOO_DEEP:
2439 512 : return (_("JSON nested too deep, maximum permitted depth is 6400."));
2440 132 : case JSON_ESCAPING_INVALID:
2441 132 : json_token_error(lex, "Escape sequence \"\\%.*s\" is invalid.");
2442 132 : break;
2443 116 : case JSON_ESCAPING_REQUIRED:
2444 116 : jsonapi_appendStringInfo(lex->errormsg,
2445 116 : _("Character with value 0x%02x must be escaped."),
2446 116 : (unsigned char) *(lex->token_terminator));
2447 116 : break;
2448 48 : case JSON_EXPECTED_END:
2449 48 : json_token_error(lex, "Expected end of input, but found \"%.*s\".");
2450 48 : break;
2451 56 : case JSON_EXPECTED_ARRAY_FIRST:
2452 56 : json_token_error(lex, "Expected array element or \"]\", but found \"%.*s\".");
2453 56 : break;
2454 72 : case JSON_EXPECTED_ARRAY_NEXT:
2455 72 : json_token_error(lex, "Expected \",\" or \"]\", but found \"%.*s\".");
2456 72 : break;
2457 128 : case JSON_EXPECTED_COLON:
2458 128 : json_token_error(lex, "Expected \":\", but found \"%.*s\".");
2459 128 : break;
2460 120 : case JSON_EXPECTED_JSON:
2461 120 : json_token_error(lex, "Expected JSON value, but found \"%.*s\".");
2462 120 : break;
2463 190 : case JSON_EXPECTED_MORE:
2464 190 : return _("The input string ended unexpectedly.");
2465 124 : case JSON_EXPECTED_OBJECT_FIRST:
2466 124 : json_token_error(lex, "Expected string or \"}\", but found \"%.*s\".");
2467 124 : break;
2468 172 : case JSON_EXPECTED_OBJECT_NEXT:
2469 172 : json_token_error(lex, "Expected \",\" or \"}\", but found \"%.*s\".");
2470 172 : break;
2471 164 : case JSON_EXPECTED_STRING:
2472 164 : json_token_error(lex, "Expected string, but found \"%.*s\".");
2473 164 : break;
2474 414 : case JSON_INVALID_TOKEN:
2475 414 : json_token_error(lex, "Token \"%.*s\" is invalid.");
2476 414 : break;
2477 0 : case JSON_OUT_OF_MEMORY:
2478 : /* should have been handled above; use the error path */
2479 0 : break;
2480 24 : case JSON_UNICODE_CODE_POINT_ZERO:
2481 24 : return _("\\u0000 cannot be converted to text.");
2482 38 : case JSON_UNICODE_ESCAPE_FORMAT:
2483 38 : return _("\"\\u\" must be followed by four hexadecimal digits.");
2484 0 : case JSON_UNICODE_HIGH_ESCAPE:
2485 : /* note: this case is only reachable in frontend not backend */
2486 0 : return _("Unicode escape values cannot be used for code point values above 007F when the encoding is not UTF8.");
2487 0 : case JSON_UNICODE_UNTRANSLATABLE:
2488 :
2489 : /*
2490 : * Note: this case is only reachable in backend and not frontend.
2491 : * #ifdef it away so the frontend doesn't try to link against
2492 : * backend functionality.
2493 : */
2494 : #ifndef FRONTEND
2495 0 : return psprintf(_("Unicode escape value could not be translated to the server's encoding %s."),
2496 : GetDatabaseEncodingName());
2497 : #else
2498 : Assert(false);
2499 0 : break;
2500 : #endif
2501 12 : case JSON_UNICODE_HIGH_SURROGATE:
2502 12 : return _("Unicode high surrogate must not follow a high surrogate.");
2503 36 : case JSON_UNICODE_LOW_SURROGATE:
2504 36 : return _("Unicode low surrogate must follow a high surrogate.");
2505 0 : case JSON_SEM_ACTION_FAILED:
2506 : /* fall through to the error code after switch */
2507 0 : break;
2508 : }
2509 : #undef json_token_error
2510 :
2511 : /* Note that lex->errormsg can be NULL in shlib code. */
2512 1546 : if (lex->errormsg && lex->errormsg->len == 0)
2513 : {
2514 : /*
2515 : * We don't use a default: case, so that the compiler will warn about
2516 : * unhandled enum values. But this needs to be here anyway to cover
2517 : * the possibility of an incorrect input.
2518 : */
2519 0 : jsonapi_appendStringInfo(lex->errormsg,
2520 : "unexpected json parse error type: %d",
2521 : (int) error);
2522 : }
2523 :
2524 : #ifdef JSONAPI_USE_PQEXPBUFFER
2525 632 : if (PQExpBufferBroken(lex->errormsg))
2526 0 : return _("out of memory while constructing error description");
2527 : #endif
2528 :
2529 1546 : return lex->errormsg->data;
2530 : }
|
