Line data Source code
1 : /*-------------------------------------------------------------------------
2 : *
3 : * jsonapi.c
4 : * JSON parser and lexer interfaces
5 : *
6 : * Portions Copyright (c) 1996-2024, 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 13579536 : lex_peek(JsonLexContext *lex)
306 : {
307 13579536 : 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 375092 : lex_expect(JsonParseContext ctx, JsonLexContext *lex, JsonTokenType token)
318 : {
319 375092 : if (lex_peek(lex) == token)
320 374984 : 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 38012 : makeJsonLexContextCstringLen(JsonLexContext *lex, const char *json,
393 : size_t len, int encoding, bool need_escapes)
394 : {
395 38012 : if (lex == NULL)
396 : {
397 5528 : lex = ALLOC0(sizeof(JsonLexContext));
398 5528 : if (!lex)
399 0 : return &failed_oom;
400 5528 : lex->flags |= JSONLEX_FREE_STRUCT;
401 : }
402 : else
403 32484 : memset(lex, 0, sizeof(JsonLexContext));
404 :
405 38012 : lex->errormsg = NULL;
406 38012 : lex->input = lex->token_terminator = lex->line_start = json;
407 38012 : lex->line_number = 1;
408 38012 : lex->input_length = len;
409 38012 : lex->input_encoding = encoding;
410 38012 : lex->need_escapes = need_escapes;
411 38012 : 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 29474 : lex->strval = jsonapi_makeStringInfo();
419 29474 : lex->flags |= JSONLEX_FREE_STRVAL;
420 : }
421 :
422 38012 : 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 4148 : allocate_incremental_state(JsonLexContext *lex)
435 : {
436 : void *pstack,
437 : *prediction,
438 : *fnames,
439 : *fnull;
440 :
441 4148 : lex->inc_state = ALLOC0(sizeof(JsonIncrementalState));
442 4148 : pstack = ALLOC0(sizeof(JsonParserStack));
443 4148 : prediction = ALLOC(JS_STACK_CHUNK_SIZE * JS_MAX_PROD_LEN);
444 4148 : fnames = ALLOC(JS_STACK_CHUNK_SIZE * sizeof(char *));
445 4148 : 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 4148 : jsonapi_initStringInfo(&(lex->inc_state->partial_token));
466 4148 : lex->pstack = pstack;
467 4148 : lex->pstack->stack_size = JS_STACK_CHUNK_SIZE;
468 4148 : lex->pstack->prediction = prediction;
469 4148 : lex->pstack->fnames = fnames;
470 4148 : 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 4148 : lex->pstack->fnames[0] = NULL;
479 :
480 4148 : lex->incremental = true;
481 4148 : 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 4148 : makeJsonLexContextIncremental(JsonLexContext *lex, int encoding,
499 : bool need_escapes)
500 : {
501 4148 : 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 4146 : memset(lex, 0, sizeof(JsonLexContext));
511 :
512 4148 : lex->line_number = 1;
513 4148 : lex->input_encoding = encoding;
514 :
515 4148 : 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 4148 : lex->need_escapes = need_escapes;
528 4148 : 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 : */
535 234 : lex->strval = jsonapi_makeStringInfo();
536 234 : lex->flags |= JSONLEX_FREE_STRVAL;
537 : }
538 :
539 4148 : 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 5165486 : inc_lex_level(JsonLexContext *lex)
563 : {
564 5165486 : 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 5165486 : lex->lex_level += 1;
600 :
601 5165486 : if (lex->incremental)
602 : {
603 : /*
604 : * Ensure freeJsonLexContext() remains safe even if no fname is
605 : * assigned at this level.
606 : */
607 5165486 : lex->pstack->fnames[lex->lex_level] = NULL;
608 : }
609 :
610 5165486 : return true;
611 : }
612 :
613 : static inline void
614 1887684 : dec_lex_level(JsonLexContext *lex)
615 : {
616 1887684 : set_fname(lex, NULL); /* free the current level's fname, if needed */
617 1887684 : lex->lex_level -= 1;
618 1887684 : }
619 :
620 : static inline void
621 14580722 : push_prediction(JsonParserStack *pstack, td_entry entry)
622 : {
623 14580722 : memcpy(pstack->prediction + pstack->pred_index, entry.prod, entry.len);
624 14580722 : pstack->pred_index += entry.len;
625 14580722 : }
626 :
627 : static inline char
628 46388500 : pop_prediction(JsonParserStack *pstack)
629 : {
630 : Assert(pstack->pred_index > 0);
631 46388500 : 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 46993048 : have_prediction(JsonParserStack *pstack)
643 : {
644 46993048 : return pstack->pred_index > 0;
645 : }
646 :
647 : static inline void
648 3078216 : set_fname(JsonLexContext *lex, char *fname)
649 : {
650 3078216 : 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 3078216 : lex->pstack->fnames[lex->lex_level] = fname;
660 3078216 : }
661 :
662 : static inline char *
663 1111020 : get_fname(JsonLexContext *lex)
664 : {
665 1111020 : return lex->pstack->fnames[lex->lex_level];
666 : }
667 :
668 : static inline void
669 6342648 : set_fnull(JsonLexContext *lex, bool fnull)
670 : {
671 6342648 : lex->pstack->fnull[lex->lex_level] = fnull;
672 6342648 : }
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 10420 : freeJsonLexContext(JsonLexContext *lex)
689 : {
690 : static const JsonLexContext empty = {0};
691 :
692 10420 : if (!lex || lex == &failed_oom)
693 0 : return;
694 :
695 10420 : if (lex->flags & JSONLEX_FREE_STRVAL)
696 6120 : jsonapi_destroyStringInfo(lex->strval);
697 :
698 10420 : if (lex->errormsg)
699 1904 : jsonapi_destroyStringInfo(lex->errormsg);
700 :
701 10420 : if (lex->incremental)
702 : {
703 4142 : jsonapi_termStringInfo(&lex->inc_state->partial_token);
704 4142 : FREE(lex->inc_state);
705 4142 : FREE(lex->pstack->prediction);
706 :
707 4142 : 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 4142 : FREE(lex->pstack->fnames);
717 4142 : FREE(lex->pstack->fnull);
718 4142 : FREE(lex->pstack->scalar_val);
719 4142 : FREE(lex->pstack);
720 : }
721 :
722 10420 : if (lex->flags & JSONLEX_FREE_STRUCT)
723 5308 : FREE(lex);
724 : else
725 5112 : *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 37274 : 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 37274 : if (lex == &failed_oom)
764 0 : return JSON_OUT_OF_MEMORY;
765 37274 : if (lex->incremental)
766 0 : return JSON_INVALID_LEXER_TYPE;
767 :
768 : /* get the initial token */
769 37274 : result = json_lex(lex);
770 37274 : if (result != JSON_SUCCESS)
771 246 : return result;
772 :
773 37028 : tok = lex_peek(lex);
774 :
775 : /* parse by recursive descent */
776 37028 : switch (tok)
777 : {
778 20344 : case JSON_TOKEN_OBJECT_START:
779 20344 : result = parse_object(lex, sem);
780 20276 : break;
781 6856 : case JSON_TOKEN_ARRAY_START:
782 6856 : result = parse_array(lex, sem);
783 6752 : break;
784 9828 : default:
785 9828 : result = parse_scalar(lex, sem); /* json can be a bare scalar */
786 : }
787 :
788 36784 : if (result == JSON_SUCCESS)
789 36342 : result = lex_expect(JSON_PARSE_END, lex, JSON_TOKEN_END);
790 :
791 36784 : 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 745868 : 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 745868 : JsonParseContext ctx = JSON_PARSE_VALUE;
878 745868 : JsonParserStack *pstack = lex->pstack;
879 :
880 745868 : if (lex == &failed_oom || lex->inc_state == &failed_inc_oom)
881 0 : return JSON_OUT_OF_MEMORY;
882 745868 : if (!lex->incremental)
883 0 : return JSON_INVALID_LEXER_TYPE;
884 :
885 745868 : lex->input = lex->token_terminator = lex->line_start = json;
886 745868 : lex->input_length = len;
887 745868 : lex->inc_state->is_last_chunk = is_last;
888 745868 : lex->inc_state->started = true;
889 :
890 : /* get the initial token */
891 745868 : result = json_lex(lex);
892 745868 : if (result != JSON_SUCCESS)
893 143560 : return result;
894 :
895 602308 : tok = lex_peek(lex);
896 :
897 : /* use prediction stack for incremental parsing */
898 :
899 602308 : if (!have_prediction(pstack))
900 : {
901 3780 : td_entry goal = TD_ENTRY(JSON_PROD_GOAL);
902 :
903 3780 : push_prediction(pstack, goal);
904 : }
905 :
906 46390740 : while (have_prediction(pstack))
907 : {
908 46388500 : 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 46388500 : 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 11797328 : if (tok < JSON_TOKEN_END)
921 : {
922 11795088 : result = json_lex(lex);
923 11795088 : if (result != JSON_SUCCESS)
924 600066 : return result;
925 11196416 : tok = lex_peek(lex);
926 : }
927 : }
928 34591172 : 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 14576942 : push_prediction(pstack, entry);
936 : }
937 20014230 : 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 20013348 : switch (top)
947 : {
948 243470 : case JSON_SEM_OSTART:
949 : {
950 243470 : json_struct_action ostart = sem->object_start;
951 :
952 243470 : if (lex->lex_level >= JSON_TD_MAX_STACK)
953 0 : return JSON_NESTING_TOO_DEEP;
954 :
955 243470 : if (ostart != NULL)
956 : {
957 222118 : result = (*ostart) (sem->semstate);
958 222118 : if (result != JSON_SUCCESS)
959 0 : return result;
960 : }
961 :
962 243470 : if (!inc_lex_level(lex))
963 0 : return JSON_OUT_OF_MEMORY;
964 : }
965 243470 : break;
966 242740 : case JSON_SEM_OEND:
967 : {
968 242740 : json_struct_action oend = sem->object_end;
969 :
970 242740 : dec_lex_level(lex);
971 242740 : if (oend != NULL)
972 : {
973 222116 : result = (*oend) (sem->semstate);
974 222116 : if (result != JSON_SUCCESS)
975 0 : return result;
976 : }
977 : }
978 242740 : break;
979 4922528 : case JSON_SEM_ASTART:
980 : {
981 4922528 : json_struct_action astart = sem->array_start;
982 :
983 4922528 : if (lex->lex_level >= JSON_TD_MAX_STACK)
984 512 : return JSON_NESTING_TOO_DEEP;
985 :
986 4922016 : if (astart != NULL)
987 : {
988 536 : result = (*astart) (sem->semstate);
989 536 : if (result != JSON_SUCCESS)
990 0 : return result;
991 : }
992 :
993 4922016 : if (!inc_lex_level(lex))
994 0 : return JSON_OUT_OF_MEMORY;
995 : }
996 4922016 : break;
997 1644944 : case JSON_SEM_AEND:
998 : {
999 1644944 : json_struct_action aend = sem->array_end;
1000 :
1001 1644944 : dec_lex_level(lex);
1002 1644944 : if (aend != NULL)
1003 : {
1004 536 : result = (*aend) (sem->semstate);
1005 536 : if (result != JSON_SUCCESS)
1006 0 : return result;
1007 : }
1008 : }
1009 1644944 : break;
1010 1190532 : 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 1190532 : char *fname = NULL;
1018 1190532 : json_ofield_action ostart = sem->object_field_start;
1019 1190532 : json_ofield_action oend = sem->object_field_end;
1020 :
1021 1190532 : if ((ostart != NULL || oend != NULL) && lex->need_escapes)
1022 : {
1023 1109772 : fname = STRDUP(lex->strval->data);
1024 1109772 : if (fname == NULL)
1025 0 : return JSON_OUT_OF_MEMORY;
1026 : }
1027 1190532 : set_fname(lex, fname);
1028 : }
1029 1190532 : break;
1030 1190276 : case JSON_SEM_OFIELD_START:
1031 : {
1032 : /*
1033 : * the current token should be the first token of the
1034 : * value
1035 : */
1036 1190276 : bool isnull = tok == JSON_TOKEN_NULL;
1037 1190276 : json_ofield_action ostart = sem->object_field_start;
1038 :
1039 1190276 : set_fnull(lex, isnull);
1040 :
1041 1190276 : if (ostart != NULL)
1042 : {
1043 1109772 : char *fname = get_fname(lex);
1044 :
1045 1109772 : result = (*ostart) (sem->semstate, fname, isnull);
1046 1109772 : if (result != JSON_SUCCESS)
1047 0 : return result;
1048 : }
1049 : }
1050 1190276 : break;
1051 1190202 : case JSON_SEM_OFIELD_END:
1052 : {
1053 1190202 : json_ofield_action oend = sem->object_field_end;
1054 :
1055 1190202 : 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 1190202 : break;
1066 5152372 : case JSON_SEM_AELEM_START:
1067 : {
1068 5152372 : json_aelem_action astart = sem->array_element_start;
1069 5152372 : bool isnull = tok == JSON_TOKEN_NULL;
1070 :
1071 5152372 : set_fnull(lex, isnull);
1072 :
1073 5152372 : if (astart != NULL)
1074 : {
1075 240 : result = (*astart) (sem->semstate, isnull);
1076 240 : if (result != JSON_SUCCESS)
1077 0 : return result;
1078 : }
1079 : }
1080 5152372 : break;
1081 1875572 : case JSON_SEM_AELEM_END:
1082 : {
1083 1875572 : json_aelem_action aend = sem->array_element_end;
1084 :
1085 1875572 : 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 1875572 : break;
1095 1180356 : case JSON_SEM_SCALAR_INIT:
1096 : {
1097 1180356 : json_scalar_action sfunc = sem->scalar;
1098 :
1099 1180356 : pstack->scalar_val = NULL;
1100 :
1101 1180356 : 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 1109164 : if (tok == JSON_TOKEN_STRING)
1112 : {
1113 886900 : if (lex->need_escapes)
1114 : {
1115 886900 : pstack->scalar_val = STRDUP(lex->strval->data);
1116 886900 : if (pstack->scalar_val == NULL)
1117 0 : return JSON_OUT_OF_MEMORY;
1118 : }
1119 : }
1120 : else
1121 : {
1122 222264 : ptrdiff_t tlen = (lex->token_terminator - lex->token_start);
1123 :
1124 222264 : pstack->scalar_val = ALLOC(tlen + 1);
1125 222264 : if (pstack->scalar_val == NULL)
1126 0 : return JSON_OUT_OF_MEMORY;
1127 :
1128 222264 : memcpy(pstack->scalar_val, lex->token_start, tlen);
1129 222264 : pstack->scalar_val[tlen] = '\0';
1130 : }
1131 1109164 : pstack->scalar_tok = tok;
1132 : }
1133 : }
1134 1180356 : break;
1135 1180356 : 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 1180356 : json_scalar_action sfunc = sem->scalar;
1145 :
1146 1180356 : if (sfunc != NULL)
1147 : {
1148 1109164 : 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 1109162 : if (lex->flags & JSONLEX_CTX_OWNS_TOKENS)
1157 544 : FREE(pstack->scalar_val);
1158 1109162 : pstack->scalar_val = NULL;
1159 :
1160 1109162 : if (result != JSON_SUCCESS)
1161 0 : return result;
1162 : }
1163 : }
1164 1180354 : 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 2240 : 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 245744 : parse_scalar(JsonLexContext *lex, const JsonSemAction *sem)
1254 : {
1255 245744 : char *val = NULL;
1256 245744 : json_scalar_action sfunc = sem->scalar;
1257 245744 : JsonTokenType tok = lex_peek(lex);
1258 : JsonParseErrorType result;
1259 :
1260 : /* a scalar must be a string, a number, true, false, or null */
1261 245744 : if (tok != JSON_TOKEN_STRING && tok != JSON_TOKEN_NUMBER &&
1262 32556 : 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 245556 : if (sfunc == NULL)
1268 11762 : return json_lex(lex);
1269 :
1270 : /* extract the de-escaped string value, or the raw lexeme */
1271 233794 : if (lex_peek(lex) == JSON_TOKEN_STRING)
1272 : {
1273 74938 : if (lex->need_escapes)
1274 : {
1275 69262 : val = STRDUP(lex->strval->data);
1276 69262 : if (val == NULL)
1277 0 : return JSON_OUT_OF_MEMORY;
1278 : }
1279 : }
1280 : else
1281 : {
1282 158856 : int len = (lex->token_terminator - lex->token_start);
1283 :
1284 158856 : val = ALLOC(len + 1);
1285 158856 : if (val == NULL)
1286 0 : return JSON_OUT_OF_MEMORY;
1287 :
1288 158856 : memcpy(val, lex->token_start, len);
1289 158856 : val[len] = '\0';
1290 : }
1291 :
1292 : /* consume the token */
1293 233794 : result = json_lex(lex);
1294 233794 : if (result != JSON_SUCCESS)
1295 : {
1296 0 : FREE(val);
1297 0 : return result;
1298 : }
1299 :
1300 : /* invoke the callback, which may take ownership of val */
1301 233794 : result = (*sfunc) (sem->semstate, val, tok);
1302 :
1303 233698 : if (lex->flags & JSONLEX_CTX_OWNS_TOKENS)
1304 0 : FREE(val);
1305 :
1306 233698 : return result;
1307 : }
1308 :
1309 : static JsonParseErrorType
1310 238036 : parse_object_field(JsonLexContext *lex, const JsonSemAction *sem)
1311 : {
1312 : /*
1313 : * An object field is "fieldname" : value where value can be a scalar,
1314 : * object or array. Note: in user-facing docs and error messages, we
1315 : * generally call a field name a "key".
1316 : */
1317 :
1318 238036 : char *fname = NULL;
1319 238036 : json_ofield_action ostart = sem->object_field_start;
1320 238036 : json_ofield_action oend = sem->object_field_end;
1321 : bool isnull;
1322 : JsonTokenType tok;
1323 : JsonParseErrorType result;
1324 :
1325 238036 : if (lex_peek(lex) != JSON_TOKEN_STRING)
1326 12 : return report_parse_error(JSON_PARSE_STRING, lex);
1327 238024 : if ((ostart != NULL || oend != NULL) && lex->need_escapes)
1328 : {
1329 190896 : fname = STRDUP(lex->strval->data);
1330 190896 : if (fname == NULL)
1331 0 : return JSON_OUT_OF_MEMORY;
1332 : }
1333 238024 : result = json_lex(lex);
1334 238024 : if (result != JSON_SUCCESS)
1335 : {
1336 12 : FREE(fname);
1337 12 : return result;
1338 : }
1339 :
1340 238012 : result = lex_expect(JSON_PARSE_OBJECT_LABEL, lex, JSON_TOKEN_COLON);
1341 238012 : if (result != JSON_SUCCESS)
1342 : {
1343 90 : FREE(fname);
1344 90 : return result;
1345 : }
1346 :
1347 237922 : tok = lex_peek(lex);
1348 237922 : isnull = tok == JSON_TOKEN_NULL;
1349 :
1350 237922 : if (ostart != NULL)
1351 : {
1352 190812 : result = (*ostart) (sem->semstate, fname, isnull);
1353 190804 : if (result != JSON_SUCCESS)
1354 0 : goto ofield_cleanup;
1355 : }
1356 :
1357 237914 : switch (tok)
1358 : {
1359 11052 : case JSON_TOKEN_OBJECT_START:
1360 11052 : result = parse_object(lex, sem);
1361 4012 : break;
1362 14822 : case JSON_TOKEN_ARRAY_START:
1363 14822 : result = parse_array(lex, sem);
1364 14788 : break;
1365 212040 : default:
1366 212040 : result = parse_scalar(lex, sem);
1367 : }
1368 230834 : if (result != JSON_SUCCESS)
1369 42 : goto ofield_cleanup;
1370 :
1371 230792 : if (oend != NULL)
1372 : {
1373 118602 : result = (*oend) (sem->semstate, fname, isnull);
1374 118602 : if (result != JSON_SUCCESS)
1375 0 : goto ofield_cleanup;
1376 : }
1377 :
1378 230792 : ofield_cleanup:
1379 230834 : if (lex->flags & JSONLEX_CTX_OWNS_TOKENS)
1380 0 : FREE(fname);
1381 230834 : return result;
1382 : }
1383 :
1384 : static JsonParseErrorType
1385 49520 : parse_object(JsonLexContext *lex, const JsonSemAction *sem)
1386 : {
1387 : /*
1388 : * an object is a possibly empty sequence of object fields, separated by
1389 : * commas and surrounded by curly braces.
1390 : */
1391 49520 : json_struct_action ostart = sem->object_start;
1392 49520 : json_struct_action oend = sem->object_end;
1393 : JsonTokenType tok;
1394 : JsonParseErrorType result;
1395 :
1396 : #ifndef FRONTEND
1397 :
1398 : /*
1399 : * TODO: clients need some way to put a bound on stack growth. Parse level
1400 : * limits maybe?
1401 : */
1402 43586 : check_stack_depth();
1403 : #endif
1404 :
1405 49508 : if (ostart != NULL)
1406 : {
1407 30250 : result = (*ostart) (sem->semstate);
1408 30230 : if (result != JSON_SUCCESS)
1409 0 : return result;
1410 : }
1411 :
1412 : /*
1413 : * Data inside an object is at a higher nesting level than the object
1414 : * itself. Note that we increment this after we call the semantic routine
1415 : * for the object start and restore it before we call the routine for the
1416 : * object end.
1417 : */
1418 49488 : lex->lex_level++;
1419 :
1420 : Assert(lex_peek(lex) == JSON_TOKEN_OBJECT_START);
1421 49488 : result = json_lex(lex);
1422 49488 : if (result != JSON_SUCCESS)
1423 60 : return result;
1424 :
1425 49428 : tok = lex_peek(lex);
1426 49428 : switch (tok)
1427 : {
1428 46566 : case JSON_TOKEN_STRING:
1429 46566 : result = parse_object_field(lex, sem);
1430 230948 : while (result == JSON_SUCCESS && lex_peek(lex) == JSON_TOKEN_COMMA)
1431 : {
1432 191470 : result = json_lex(lex);
1433 191470 : if (result != JSON_SUCCESS)
1434 0 : break;
1435 191470 : result = parse_object_field(lex, sem);
1436 : }
1437 39478 : break;
1438 2848 : case JSON_TOKEN_OBJECT_END:
1439 2848 : break;
1440 14 : default:
1441 : /* case of an invalid initial token inside the object */
1442 14 : result = report_parse_error(JSON_PARSE_OBJECT_START, lex);
1443 : }
1444 42340 : if (result != JSON_SUCCESS)
1445 170 : return result;
1446 :
1447 42170 : result = lex_expect(JSON_PARSE_OBJECT_NEXT, lex, JSON_TOKEN_OBJECT_END);
1448 42170 : if (result != JSON_SUCCESS)
1449 36 : return result;
1450 :
1451 42134 : lex->lex_level--;
1452 :
1453 42134 : if (oend != NULL)
1454 : {
1455 24220 : result = (*oend) (sem->semstate);
1456 24174 : if (result != JSON_SUCCESS)
1457 0 : return result;
1458 : }
1459 :
1460 42088 : return JSON_SUCCESS;
1461 : }
1462 :
1463 : static JsonParseErrorType
1464 54114 : parse_array_element(JsonLexContext *lex, const JsonSemAction *sem)
1465 : {
1466 54114 : json_aelem_action astart = sem->array_element_start;
1467 54114 : json_aelem_action aend = sem->array_element_end;
1468 54114 : JsonTokenType tok = lex_peek(lex);
1469 : JsonParseErrorType result;
1470 : bool isnull;
1471 :
1472 54114 : isnull = tok == JSON_TOKEN_NULL;
1473 :
1474 54114 : if (astart != NULL)
1475 : {
1476 7738 : result = (*astart) (sem->semstate, isnull);
1477 7738 : if (result != JSON_SUCCESS)
1478 0 : return result;
1479 : }
1480 :
1481 : /* an array element is any object, array or scalar */
1482 54114 : switch (tok)
1483 : {
1484 18124 : case JSON_TOKEN_OBJECT_START:
1485 18124 : result = parse_object(lex, sem);
1486 18066 : break;
1487 12114 : case JSON_TOKEN_ARRAY_START:
1488 12114 : result = parse_array(lex, sem);
1489 3292 : break;
1490 23876 : default:
1491 23876 : result = parse_scalar(lex, sem);
1492 : }
1493 :
1494 45216 : if (result != JSON_SUCCESS)
1495 66 : return result;
1496 :
1497 45150 : if (aend != NULL)
1498 : {
1499 7228 : result = (*aend) (sem->semstate, isnull);
1500 7216 : if (result != JSON_SUCCESS)
1501 0 : return result;
1502 : }
1503 :
1504 45138 : return JSON_SUCCESS;
1505 : }
1506 :
1507 : static JsonParseErrorType
1508 33792 : parse_array(JsonLexContext *lex, const JsonSemAction *sem)
1509 : {
1510 : /*
1511 : * an array is a possibly empty sequence of array elements, separated by
1512 : * commas and surrounded by square brackets.
1513 : */
1514 33792 : json_struct_action astart = sem->array_start;
1515 33792 : json_struct_action aend = sem->array_end;
1516 : JsonParseErrorType result;
1517 :
1518 : #ifndef FRONTEND
1519 33736 : check_stack_depth();
1520 : #endif
1521 :
1522 33780 : if (astart != NULL)
1523 : {
1524 15790 : result = (*astart) (sem->semstate);
1525 15776 : if (result != JSON_SUCCESS)
1526 0 : return result;
1527 : }
1528 :
1529 : /*
1530 : * Data inside an array is at a higher nesting level than the array
1531 : * itself. Note that we increment this after we call the semantic routine
1532 : * for the array start and restore it before we call the routine for the
1533 : * array end.
1534 : */
1535 33766 : lex->lex_level++;
1536 :
1537 33766 : result = lex_expect(JSON_PARSE_ARRAY_START, lex, JSON_TOKEN_ARRAY_START);
1538 33766 : if (result == JSON_SUCCESS && lex_peek(lex) != JSON_TOKEN_ARRAY_END)
1539 : {
1540 26292 : result = parse_array_element(lex, sem);
1541 :
1542 45156 : while (result == JSON_SUCCESS && lex_peek(lex) == JSON_TOKEN_COMMA)
1543 : {
1544 27774 : result = json_lex(lex);
1545 27774 : if (result != JSON_SUCCESS)
1546 0 : break;
1547 27774 : result = parse_array_element(lex, sem);
1548 : }
1549 : }
1550 24856 : if (result != JSON_SUCCESS)
1551 66 : return result;
1552 :
1553 24790 : result = lex_expect(JSON_PARSE_ARRAY_NEXT, lex, JSON_TOKEN_ARRAY_END);
1554 24790 : if (result != JSON_SUCCESS)
1555 24 : return result;
1556 :
1557 24766 : lex->lex_level--;
1558 :
1559 24766 : if (aend != NULL)
1560 : {
1561 9172 : result = (*aend) (sem->semstate);
1562 9148 : if (result != JSON_SUCCESS)
1563 0 : return result;
1564 : }
1565 :
1566 24742 : return JSON_SUCCESS;
1567 : }
1568 :
1569 : /*
1570 : * Lex one token from the input stream.
1571 : *
1572 : * When doing incremental parsing, we can reach the end of the input string
1573 : * without having (or knowing we have) a complete token. If it's not the
1574 : * final chunk of input, the partial token is then saved to the lex
1575 : * structure's ptok StringInfo. On subsequent calls input is appended to this
1576 : * buffer until we have something that we think is a complete token,
1577 : * which is then lexed using a recursive call to json_lex. Processing then
1578 : * continues as normal on subsequent calls.
1579 : *
1580 : * Note than when doing incremental processing, the lex.prev_token_terminator
1581 : * should not be relied on. It could point into a previous input chunk or
1582 : * worse.
1583 : */
1584 : JsonParseErrorType
1585 13775662 : json_lex(JsonLexContext *lex)
1586 : {
1587 : const char *s;
1588 13775662 : const char *const end = lex->input + lex->input_length;
1589 : JsonParseErrorType result;
1590 :
1591 13775662 : if (lex == &failed_oom || lex->inc_state == &failed_inc_oom)
1592 0 : return JSON_OUT_OF_MEMORY;
1593 :
1594 13775662 : if (lex->incremental)
1595 : {
1596 12540956 : if (lex->inc_state->partial_completed)
1597 : {
1598 : /*
1599 : * We just lexed a completed partial token on the last call, so
1600 : * reset everything
1601 : */
1602 68764 : jsonapi_resetStringInfo(&(lex->inc_state->partial_token));
1603 68764 : lex->token_terminator = lex->input;
1604 68764 : lex->inc_state->partial_completed = false;
1605 : }
1606 :
1607 : #ifdef JSONAPI_USE_PQEXPBUFFER
1608 : /* Make sure our partial token buffer is valid before using it below. */
1609 3831520 : if (PQExpBufferDataBroken(lex->inc_state->partial_token))
1610 0 : return JSON_OUT_OF_MEMORY;
1611 : #endif
1612 : }
1613 :
1614 13775662 : s = lex->token_terminator;
1615 :
1616 13775662 : if (lex->incremental && lex->inc_state->partial_token.len)
1617 : {
1618 : /*
1619 : * We have a partial token. Extend it and if completed lex it by a
1620 : * recursive call
1621 : */
1622 166308 : jsonapi_StrValType *ptok = &(lex->inc_state->partial_token);
1623 166308 : size_t added = 0;
1624 166308 : bool tok_done = false;
1625 166308 : JsonLexContext dummy_lex = {0};
1626 : JsonParseErrorType partial_result;
1627 :
1628 166308 : if (ptok->data[0] == '"')
1629 : {
1630 : /*
1631 : * It's a string. Accumulate characters until we reach an
1632 : * unescaped '"'.
1633 : */
1634 160060 : int escapes = 0;
1635 :
1636 162628 : for (int i = ptok->len - 1; i > 0; i--)
1637 : {
1638 : /* count the trailing backslashes on the partial token */
1639 151242 : if (ptok->data[i] == '\\')
1640 2568 : escapes++;
1641 : else
1642 148674 : break;
1643 : }
1644 :
1645 1213838 : for (size_t i = 0; i < lex->input_length; i++)
1646 : {
1647 1118694 : char c = lex->input[i];
1648 :
1649 1118694 : jsonapi_appendStringInfoCharMacro(ptok, c);
1650 1118694 : added++;
1651 1118694 : if (c == '"' && escapes % 2 == 0)
1652 : {
1653 64916 : tok_done = true;
1654 64916 : break;
1655 : }
1656 1053778 : if (c == '\\')
1657 4096 : escapes++;
1658 : else
1659 1049682 : escapes = 0;
1660 : }
1661 : }
1662 : else
1663 : {
1664 : /* not a string */
1665 6248 : char c = ptok->data[0];
1666 :
1667 6248 : if (c == '-' || (c >= '0' && c <= '9'))
1668 : {
1669 : /* for numbers look for possible numeric continuations */
1670 :
1671 1312 : bool numend = false;
1672 :
1673 3680 : for (size_t i = 0; i < lex->input_length && !numend; i++)
1674 : {
1675 2368 : char cc = lex->input[i];
1676 :
1677 2368 : switch (cc)
1678 : {
1679 1592 : case '+':
1680 : case '-':
1681 : case 'e':
1682 : case 'E':
1683 : case '0':
1684 : case '1':
1685 : case '2':
1686 : case '3':
1687 : case '4':
1688 : case '5':
1689 : case '6':
1690 : case '7':
1691 : case '8':
1692 : case '9':
1693 : {
1694 1592 : jsonapi_appendStringInfoCharMacro(ptok, cc);
1695 1592 : added++;
1696 : }
1697 1592 : break;
1698 776 : default:
1699 776 : numend = true;
1700 : }
1701 : }
1702 : }
1703 :
1704 : /*
1705 : * Add any remaining alphanumeric chars. This takes care of the
1706 : * {null, false, true} literals as well as any trailing
1707 : * alphanumeric junk on non-string tokens.
1708 : */
1709 12896 : for (size_t i = added; i < lex->input_length; i++)
1710 : {
1711 10936 : char cc = lex->input[i];
1712 :
1713 10936 : if (JSON_ALPHANUMERIC_CHAR(cc))
1714 : {
1715 6648 : jsonapi_appendStringInfoCharMacro(ptok, cc);
1716 6648 : added++;
1717 : }
1718 : else
1719 : {
1720 4288 : tok_done = true;
1721 4288 : break;
1722 : }
1723 : }
1724 6248 : if (added == lex->input_length &&
1725 1960 : lex->inc_state->is_last_chunk)
1726 : {
1727 152 : tok_done = true;
1728 : }
1729 : }
1730 :
1731 166308 : if (!tok_done)
1732 : {
1733 : /* We should have consumed the whole chunk in this case. */
1734 : Assert(added == lex->input_length);
1735 :
1736 96952 : if (!lex->inc_state->is_last_chunk)
1737 96872 : return JSON_INCOMPLETE;
1738 :
1739 : /* json_errdetail() needs access to the accumulated token. */
1740 80 : lex->token_start = ptok->data;
1741 80 : lex->token_terminator = ptok->data + ptok->len;
1742 80 : return JSON_INVALID_TOKEN;
1743 : }
1744 :
1745 : /*
1746 : * Everything up to lex->input[added] has been added to the partial
1747 : * token, so move the input past it.
1748 : */
1749 69356 : lex->input += added;
1750 69356 : lex->input_length -= added;
1751 :
1752 69356 : dummy_lex.input = dummy_lex.token_terminator =
1753 69356 : dummy_lex.line_start = ptok->data;
1754 69356 : dummy_lex.line_number = lex->line_number;
1755 69356 : dummy_lex.input_length = ptok->len;
1756 69356 : dummy_lex.input_encoding = lex->input_encoding;
1757 69356 : dummy_lex.incremental = false;
1758 69356 : dummy_lex.need_escapes = lex->need_escapes;
1759 69356 : dummy_lex.strval = lex->strval;
1760 :
1761 69356 : partial_result = json_lex(&dummy_lex);
1762 :
1763 : /*
1764 : * We either have a complete token or an error. In either case we need
1765 : * to point to the partial token data for the semantic or error
1766 : * routines. If it's not an error we'll readjust on the next call to
1767 : * json_lex.
1768 : */
1769 69356 : lex->token_type = dummy_lex.token_type;
1770 69356 : lex->line_number = dummy_lex.line_number;
1771 :
1772 : /*
1773 : * We know the prev_token_terminator must be back in some previous
1774 : * piece of input, so we just make it NULL.
1775 : */
1776 69356 : lex->prev_token_terminator = NULL;
1777 :
1778 : /*
1779 : * Normally token_start would be ptok->data, but it could be later,
1780 : * see json_lex_string's handling of invalid escapes.
1781 : */
1782 69356 : lex->token_start = dummy_lex.token_start;
1783 69356 : lex->token_terminator = dummy_lex.token_terminator;
1784 69356 : if (partial_result == JSON_SUCCESS)
1785 : {
1786 : /* make sure we've used all the input */
1787 69140 : if (lex->token_terminator - lex->token_start != ptok->len)
1788 : {
1789 : Assert(false);
1790 0 : return JSON_INVALID_TOKEN;
1791 : }
1792 :
1793 69140 : lex->inc_state->partial_completed = true;
1794 : }
1795 69356 : return partial_result;
1796 : /* end of partial token processing */
1797 : }
1798 :
1799 : /* Skip leading whitespace. */
1800 20553448 : while (s < end && (*s == ' ' || *s == '\t' || *s == '\n' || *s == '\r'))
1801 : {
1802 6944094 : if (*s++ == '\n')
1803 : {
1804 481102 : ++lex->line_number;
1805 481102 : lex->line_start = s;
1806 : }
1807 : }
1808 13609354 : lex->token_start = s;
1809 :
1810 : /* Determine token type. */
1811 13609354 : if (s >= end)
1812 : {
1813 650654 : lex->token_start = NULL;
1814 650654 : lex->prev_token_terminator = lex->token_terminator;
1815 650654 : lex->token_terminator = s;
1816 650654 : lex->token_type = JSON_TOKEN_END;
1817 : }
1818 : else
1819 : {
1820 12958700 : switch (*s)
1821 : {
1822 : /* Single-character token, some kind of punctuation mark. */
1823 293342 : case '{':
1824 293342 : lex->prev_token_terminator = lex->token_terminator;
1825 293342 : lex->token_terminator = s + 1;
1826 293342 : lex->token_type = JSON_TOKEN_OBJECT_START;
1827 293342 : break;
1828 285038 : case '}':
1829 285038 : lex->prev_token_terminator = lex->token_terminator;
1830 285038 : lex->token_terminator = s + 1;
1831 285038 : lex->token_type = JSON_TOKEN_OBJECT_END;
1832 285038 : break;
1833 4956458 : case '[':
1834 4956458 : lex->prev_token_terminator = lex->token_terminator;
1835 4956458 : lex->token_terminator = s + 1;
1836 4956458 : lex->token_type = JSON_TOKEN_ARRAY_START;
1837 4956458 : break;
1838 1669884 : case ']':
1839 1669884 : lex->prev_token_terminator = lex->token_terminator;
1840 1669884 : lex->token_terminator = s + 1;
1841 1669884 : lex->token_type = JSON_TOKEN_ARRAY_END;
1842 1669884 : break;
1843 1400248 : case ',':
1844 1400248 : lex->prev_token_terminator = lex->token_terminator;
1845 1400248 : lex->token_terminator = s + 1;
1846 1400248 : lex->token_type = JSON_TOKEN_COMMA;
1847 1400248 : break;
1848 1428288 : case ':':
1849 1428288 : lex->prev_token_terminator = lex->token_terminator;
1850 1428288 : lex->token_terminator = s + 1;
1851 1428288 : lex->token_type = JSON_TOKEN_COLON;
1852 1428288 : break;
1853 2513966 : case '"':
1854 : /* string */
1855 2513966 : result = json_lex_string(lex);
1856 2513966 : if (result != JSON_SUCCESS)
1857 65372 : return result;
1858 2448594 : lex->token_type = JSON_TOKEN_STRING;
1859 2448594 : break;
1860 184 : case '-':
1861 : /* Negative number. */
1862 184 : result = json_lex_number(lex, s + 1, NULL, NULL);
1863 184 : if (result != JSON_SUCCESS)
1864 0 : return result;
1865 184 : lex->token_type = JSON_TOKEN_NUMBER;
1866 184 : break;
1867 351598 : case '0':
1868 : case '1':
1869 : case '2':
1870 : case '3':
1871 : case '4':
1872 : case '5':
1873 : case '6':
1874 : case '7':
1875 : case '8':
1876 : case '9':
1877 : /* Positive number. */
1878 351598 : result = json_lex_number(lex, s, NULL, NULL);
1879 351598 : if (result != JSON_SUCCESS)
1880 904 : return result;
1881 350694 : lex->token_type = JSON_TOKEN_NUMBER;
1882 350694 : break;
1883 59694 : default:
1884 : {
1885 : const char *p;
1886 :
1887 : /*
1888 : * We're not dealing with a string, number, legal
1889 : * punctuation mark, or end of string. The only legal
1890 : * tokens we might find here are true, false, and null,
1891 : * but for error reporting purposes we scan until we see a
1892 : * non-alphanumeric character. That way, we can report
1893 : * the whole word as an unexpected token, rather than just
1894 : * some unintuitive prefix thereof.
1895 : */
1896 322004 : for (p = s; p < end && JSON_ALPHANUMERIC_CHAR(*p); p++)
1897 : /* skip */ ;
1898 :
1899 : /*
1900 : * We got some sort of unexpected punctuation or an
1901 : * otherwise unexpected character, so just complain about
1902 : * that one character.
1903 : */
1904 59694 : if (p == s)
1905 : {
1906 80 : lex->prev_token_terminator = lex->token_terminator;
1907 80 : lex->token_terminator = s + 1;
1908 80 : return JSON_INVALID_TOKEN;
1909 : }
1910 :
1911 59614 : if (lex->incremental && !lex->inc_state->is_last_chunk &&
1912 15664 : p == lex->input + lex->input_length)
1913 : {
1914 3664 : jsonapi_appendBinaryStringInfo(&(lex->inc_state->partial_token), s, end - s);
1915 3664 : return JSON_INCOMPLETE;
1916 : }
1917 :
1918 : /*
1919 : * We've got a real alphanumeric token here. If it
1920 : * happens to be true, false, or null, all is well. If
1921 : * not, error out.
1922 : */
1923 55950 : lex->prev_token_terminator = lex->token_terminator;
1924 55950 : lex->token_terminator = p;
1925 55950 : if (p - s == 4)
1926 : {
1927 25712 : if (memcmp(s, "true", 4) == 0)
1928 7478 : lex->token_type = JSON_TOKEN_TRUE;
1929 18234 : else if (memcmp(s, "null", 4) == 0)
1930 18222 : lex->token_type = JSON_TOKEN_NULL;
1931 : else
1932 12 : return JSON_INVALID_TOKEN;
1933 : }
1934 30238 : else if (p - s == 5 && memcmp(s, "false", 5) == 0)
1935 30000 : lex->token_type = JSON_TOKEN_FALSE;
1936 : else
1937 238 : return JSON_INVALID_TOKEN;
1938 : }
1939 : } /* end of switch */
1940 : }
1941 :
1942 13539084 : if (lex->incremental && lex->token_type == JSON_TOKEN_END && !lex->inc_state->is_last_chunk)
1943 575412 : return JSON_INCOMPLETE;
1944 : else
1945 12963672 : return JSON_SUCCESS;
1946 : }
1947 :
1948 : /*
1949 : * The next token in the input stream is known to be a string; lex it.
1950 : *
1951 : * If lex->strval isn't NULL, fill it with the decoded string.
1952 : * Set lex->token_terminator to the end of the decoded input, and in
1953 : * success cases, transfer its previous value to lex->prev_token_terminator.
1954 : * Return JSON_SUCCESS or an error code.
1955 : *
1956 : * Note: be careful that all error exits advance lex->token_terminator
1957 : * to the point after the character we detected the error on.
1958 : */
1959 : static inline JsonParseErrorType
1960 2513966 : json_lex_string(JsonLexContext *lex)
1961 : {
1962 : const char *s;
1963 2513966 : const char *const end = lex->input + lex->input_length;
1964 2513966 : int hi_surrogate = -1;
1965 :
1966 : /* Convenience macros for error exits */
1967 : #define FAIL_OR_INCOMPLETE_AT_CHAR_START(code) \
1968 : do { \
1969 : if (lex->incremental && !lex->inc_state->is_last_chunk) \
1970 : { \
1971 : jsonapi_appendBinaryStringInfo(&lex->inc_state->partial_token, \
1972 : lex->token_start, \
1973 : end - lex->token_start); \
1974 : return JSON_INCOMPLETE; \
1975 : } \
1976 : lex->token_terminator = s; \
1977 : return code; \
1978 : } while (0)
1979 : #define FAIL_AT_CHAR_END(code) \
1980 : do { \
1981 : const char *term = s + pg_encoding_mblen(lex->input_encoding, s); \
1982 : lex->token_terminator = (term <= end) ? term : end; \
1983 : return code; \
1984 : } while (0)
1985 :
1986 2513966 : if (lex->need_escapes)
1987 : {
1988 : #ifdef JSONAPI_USE_PQEXPBUFFER
1989 : /* make sure initialization succeeded */
1990 1336 : if (lex->strval == NULL)
1991 0 : return JSON_OUT_OF_MEMORY;
1992 : #endif
1993 2266220 : jsonapi_resetStringInfo(lex->strval);
1994 : }
1995 :
1996 : Assert(lex->input_length > 0);
1997 2513966 : s = lex->token_start;
1998 : for (;;)
1999 : {
2000 5028426 : s++;
2001 : /* Premature end of the string. */
2002 5028426 : if (s >= end)
2003 64568 : FAIL_OR_INCOMPLETE_AT_CHAR_START(JSON_INVALID_TOKEN);
2004 4963858 : else if (*s == '"')
2005 2448594 : break;
2006 2515264 : else if (*s == '\\')
2007 : {
2008 : /* OK, we have an escape character. */
2009 10256 : s++;
2010 10256 : if (s >= end)
2011 192 : FAIL_OR_INCOMPLETE_AT_CHAR_START(JSON_INVALID_TOKEN);
2012 10064 : else if (*s == 'u')
2013 : {
2014 : int i;
2015 3884 : int ch = 0;
2016 :
2017 18680 : for (i = 1; i <= 4; i++)
2018 : {
2019 15088 : s++;
2020 15088 : if (s >= end)
2021 256 : FAIL_OR_INCOMPLETE_AT_CHAR_START(JSON_INVALID_TOKEN);
2022 14832 : else if (*s >= '0' && *s <= '9')
2023 9230 : ch = (ch * 16) + (*s - '0');
2024 5602 : else if (*s >= 'a' && *s <= 'f')
2025 5542 : ch = (ch * 16) + (*s - 'a') + 10;
2026 60 : else if (*s >= 'A' && *s <= 'F')
2027 24 : ch = (ch * 16) + (*s - 'A') + 10;
2028 : else
2029 36 : FAIL_AT_CHAR_END(JSON_UNICODE_ESCAPE_FORMAT);
2030 : }
2031 3592 : if (lex->need_escapes)
2032 : {
2033 : /*
2034 : * Combine surrogate pairs.
2035 : */
2036 260 : if (is_utf16_surrogate_first(ch))
2037 : {
2038 72 : if (hi_surrogate != -1)
2039 12 : FAIL_AT_CHAR_END(JSON_UNICODE_HIGH_SURROGATE);
2040 60 : hi_surrogate = ch;
2041 60 : continue;
2042 : }
2043 188 : else if (is_utf16_surrogate_second(ch))
2044 : {
2045 60 : if (hi_surrogate == -1)
2046 24 : FAIL_AT_CHAR_END(JSON_UNICODE_LOW_SURROGATE);
2047 36 : ch = surrogate_pair_to_codepoint(hi_surrogate, ch);
2048 36 : hi_surrogate = -1;
2049 : }
2050 :
2051 164 : if (hi_surrogate != -1)
2052 0 : FAIL_AT_CHAR_END(JSON_UNICODE_LOW_SURROGATE);
2053 :
2054 : /*
2055 : * Reject invalid cases. We can't have a value above
2056 : * 0xFFFF here (since we only accepted 4 hex digits
2057 : * above), so no need to test for out-of-range chars.
2058 : */
2059 164 : if (ch == 0)
2060 : {
2061 : /* We can't allow this, since our TEXT type doesn't */
2062 24 : FAIL_AT_CHAR_END(JSON_UNICODE_CODE_POINT_ZERO);
2063 : }
2064 :
2065 : /*
2066 : * Add the represented character to lex->strval. In the
2067 : * backend, we can let pg_unicode_to_server_noerror()
2068 : * handle any required character set conversion; in
2069 : * frontend, we can only deal with trivial conversions.
2070 : */
2071 : #ifndef FRONTEND
2072 : {
2073 : char cbuf[MAX_UNICODE_EQUIVALENT_STRING + 1];
2074 :
2075 84 : if (!pg_unicode_to_server_noerror(ch, (unsigned char *) cbuf))
2076 0 : FAIL_AT_CHAR_END(JSON_UNICODE_UNTRANSLATABLE);
2077 84 : appendStringInfoString(lex->strval, cbuf);
2078 : }
2079 : #else
2080 56 : if (lex->input_encoding == PG_UTF8)
2081 : {
2082 : /* OK, we can map the code point to UTF8 easily */
2083 : char utf8str[5];
2084 : int utf8len;
2085 :
2086 56 : unicode_to_utf8(ch, (unsigned char *) utf8str);
2087 56 : utf8len = pg_utf_mblen((unsigned char *) utf8str);
2088 56 : jsonapi_appendBinaryStringInfo(lex->strval, utf8str, utf8len);
2089 : }
2090 0 : else if (ch <= 0x007f)
2091 : {
2092 : /* The ASCII range is the same in all encodings */
2093 0 : jsonapi_appendStringInfoChar(lex->strval, (char) ch);
2094 : }
2095 : else
2096 0 : FAIL_AT_CHAR_END(JSON_UNICODE_HIGH_ESCAPE);
2097 : #endif /* FRONTEND */
2098 : }
2099 : }
2100 6180 : else if (lex->need_escapes)
2101 : {
2102 606 : if (hi_surrogate != -1)
2103 0 : FAIL_AT_CHAR_END(JSON_UNICODE_LOW_SURROGATE);
2104 :
2105 606 : switch (*s)
2106 : {
2107 386 : case '"':
2108 : case '\\':
2109 : case '/':
2110 386 : jsonapi_appendStringInfoChar(lex->strval, *s);
2111 386 : break;
2112 44 : case 'b':
2113 44 : jsonapi_appendStringInfoChar(lex->strval, '\b');
2114 44 : break;
2115 8 : case 'f':
2116 8 : jsonapi_appendStringInfoChar(lex->strval, '\f');
2117 8 : break;
2118 62 : case 'n':
2119 62 : jsonapi_appendStringInfoChar(lex->strval, '\n');
2120 62 : break;
2121 8 : case 'r':
2122 8 : jsonapi_appendStringInfoChar(lex->strval, '\r');
2123 8 : break;
2124 92 : case 't':
2125 92 : jsonapi_appendStringInfoChar(lex->strval, '\t');
2126 92 : break;
2127 6 : default:
2128 :
2129 : /*
2130 : * Not a valid string escape, so signal error. We
2131 : * adjust token_start so that just the escape sequence
2132 : * is reported, not the whole string.
2133 : */
2134 6 : lex->token_start = s;
2135 6 : FAIL_AT_CHAR_END(JSON_ESCAPING_INVALID);
2136 : }
2137 : }
2138 5574 : else if (strchr("\"\\/bfnrt", *s) == NULL)
2139 : {
2140 : /*
2141 : * Simpler processing if we're not bothered about de-escaping
2142 : *
2143 : * It's very tempting to remove the strchr() call here and
2144 : * replace it with a switch statement, but testing so far has
2145 : * shown it's not a performance win.
2146 : */
2147 126 : lex->token_start = s;
2148 126 : FAIL_AT_CHAR_END(JSON_ESCAPING_INVALID);
2149 : }
2150 : }
2151 : else
2152 : {
2153 2505008 : const char *p = s;
2154 :
2155 2505008 : if (hi_surrogate != -1)
2156 12 : FAIL_AT_CHAR_END(JSON_UNICODE_LOW_SURROGATE);
2157 :
2158 : /*
2159 : * Skip to the first byte that requires special handling, so we
2160 : * can batch calls to jsonapi_appendBinaryStringInfo.
2161 : */
2162 3190374 : while (p < end - sizeof(Vector8) &&
2163 3010616 : !pg_lfind8('\\', (uint8 *) p, sizeof(Vector8)) &&
2164 3006998 : !pg_lfind8('"', (uint8 *) p, sizeof(Vector8)) &&
2165 685378 : !pg_lfind8_le(31, (uint8 *) p, sizeof(Vector8)))
2166 685378 : p += sizeof(Vector8);
2167 :
2168 19736850 : for (; p < end; p++)
2169 : {
2170 19683476 : if (*p == '\\' || *p == '"')
2171 : break;
2172 17231970 : else if ((unsigned char) *p <= 31)
2173 : {
2174 : /* Per RFC4627, these characters MUST be escaped. */
2175 : /*
2176 : * Since *p isn't printable, exclude it from the context
2177 : * string
2178 : */
2179 116 : lex->token_terminator = p;
2180 116 : return JSON_ESCAPING_REQUIRED;
2181 : }
2182 : }
2183 :
2184 2504880 : if (lex->need_escapes)
2185 2266402 : jsonapi_appendBinaryStringInfo(lex->strval, s, p - s);
2186 :
2187 : /*
2188 : * s will be incremented at the top of the loop, so set it to just
2189 : * behind our lookahead position
2190 : */
2191 2504880 : s = p - 1;
2192 : }
2193 : }
2194 :
2195 2448594 : if (hi_surrogate != -1)
2196 : {
2197 0 : lex->token_terminator = s + 1;
2198 0 : return JSON_UNICODE_LOW_SURROGATE;
2199 : }
2200 :
2201 : #ifdef JSONAPI_USE_PQEXPBUFFER
2202 68896 : if (lex->need_escapes && PQExpBufferBroken(lex->strval))
2203 0 : return JSON_OUT_OF_MEMORY;
2204 : #endif
2205 :
2206 : /* Hooray, we found the end of the string! */
2207 2448594 : lex->prev_token_terminator = lex->token_terminator;
2208 2448594 : lex->token_terminator = s + 1;
2209 2448594 : return JSON_SUCCESS;
2210 :
2211 : #undef FAIL_OR_INCOMPLETE_AT_CHAR_START
2212 : #undef FAIL_AT_CHAR_END
2213 : }
2214 :
2215 : /*
2216 : * The next token in the input stream is known to be a number; lex it.
2217 : *
2218 : * In JSON, a number consists of four parts:
2219 : *
2220 : * (1) An optional minus sign ('-').
2221 : *
2222 : * (2) Either a single '0', or a string of one or more digits that does not
2223 : * begin with a '0'.
2224 : *
2225 : * (3) An optional decimal part, consisting of a period ('.') followed by
2226 : * one or more digits. (Note: While this part can be omitted
2227 : * completely, it's not OK to have only the decimal point without
2228 : * any digits afterwards.)
2229 : *
2230 : * (4) An optional exponent part, consisting of 'e' or 'E', optionally
2231 : * followed by '+' or '-', followed by one or more digits. (Note:
2232 : * As with the decimal part, if 'e' or 'E' is present, it must be
2233 : * followed by at least one digit.)
2234 : *
2235 : * The 's' argument to this function points to the ostensible beginning
2236 : * of part 2 - i.e. the character after any optional minus sign, or the
2237 : * first character of the string if there is none.
2238 : *
2239 : * If num_err is not NULL, we return an error flag to *num_err rather than
2240 : * raising an error for a badly-formed number. Also, if total_len is not NULL
2241 : * the distance from lex->input to the token end+1 is returned to *total_len.
2242 : */
2243 : static inline JsonParseErrorType
2244 351828 : json_lex_number(JsonLexContext *lex, const char *s,
2245 : bool *num_err, size_t *total_len)
2246 : {
2247 351828 : bool error = false;
2248 351828 : int len = s - lex->input;
2249 :
2250 : /* Part (1): leading sign indicator. */
2251 : /* Caller already did this for us; so do nothing. */
2252 :
2253 : /* Part (2): parse main digit string. */
2254 351828 : if (len < lex->input_length && *s == '0')
2255 : {
2256 56066 : s++;
2257 56066 : len++;
2258 : }
2259 295762 : else if (len < lex->input_length && *s >= '1' && *s <= '9')
2260 : {
2261 : do
2262 : {
2263 1028828 : s++;
2264 1028828 : len++;
2265 1028828 : } while (len < lex->input_length && *s >= '0' && *s <= '9');
2266 : }
2267 : else
2268 2 : error = true;
2269 :
2270 : /* Part (3): parse optional decimal portion. */
2271 351828 : if (len < lex->input_length && *s == '.')
2272 : {
2273 37054 : s++;
2274 37054 : len++;
2275 37054 : if (len == lex->input_length || *s < '0' || *s > '9')
2276 12 : error = true;
2277 : else
2278 : {
2279 : do
2280 : {
2281 91148 : s++;
2282 91148 : len++;
2283 91148 : } while (len < lex->input_length && *s >= '0' && *s <= '9');
2284 : }
2285 : }
2286 :
2287 : /* Part (4): parse optional exponent. */
2288 351828 : if (len < lex->input_length && (*s == 'e' || *s == 'E'))
2289 : {
2290 94 : s++;
2291 94 : len++;
2292 94 : if (len < lex->input_length && (*s == '+' || *s == '-'))
2293 : {
2294 10 : s++;
2295 10 : len++;
2296 : }
2297 94 : if (len == lex->input_length || *s < '0' || *s > '9')
2298 12 : error = true;
2299 : else
2300 : {
2301 : do
2302 : {
2303 284 : s++;
2304 284 : len++;
2305 284 : } while (len < lex->input_length && *s >= '0' && *s <= '9');
2306 : }
2307 : }
2308 :
2309 : /*
2310 : * Check for trailing garbage. As in json_lex(), any alphanumeric stuff
2311 : * here should be considered part of the token for error-reporting
2312 : * purposes.
2313 : */
2314 352176 : for (; len < lex->input_length && JSON_ALPHANUMERIC_CHAR(*s); s++, len++)
2315 348 : error = true;
2316 :
2317 351828 : if (total_len != NULL)
2318 46 : *total_len = len;
2319 :
2320 351828 : if (lex->incremental && !lex->inc_state->is_last_chunk &&
2321 120008 : len >= lex->input_length)
2322 : {
2323 776 : jsonapi_appendBinaryStringInfo(&lex->inc_state->partial_token,
2324 776 : lex->token_start, s - lex->token_start);
2325 776 : if (num_err != NULL)
2326 0 : *num_err = error;
2327 :
2328 776 : return JSON_INCOMPLETE;
2329 : }
2330 351052 : else if (num_err != NULL)
2331 : {
2332 : /* let the caller handle any error */
2333 46 : *num_err = error;
2334 : }
2335 : else
2336 : {
2337 : /* return token endpoint */
2338 351006 : lex->prev_token_terminator = lex->token_terminator;
2339 351006 : lex->token_terminator = s;
2340 : /* handle error if any */
2341 351006 : if (error)
2342 128 : return JSON_INVALID_TOKEN;
2343 : }
2344 :
2345 350924 : return JSON_SUCCESS;
2346 : }
2347 :
2348 : /*
2349 : * Report a parse error.
2350 : *
2351 : * lex->token_start and lex->token_terminator must identify the current token.
2352 : */
2353 : static JsonParseErrorType
2354 1204 : report_parse_error(JsonParseContext ctx, JsonLexContext *lex)
2355 : {
2356 : /* Handle case where the input ended prematurely. */
2357 1204 : if (lex->token_start == NULL || lex->token_type == JSON_TOKEN_END)
2358 266 : return JSON_EXPECTED_MORE;
2359 :
2360 : /* Otherwise choose the error type based on the parsing context. */
2361 938 : switch (ctx)
2362 : {
2363 48 : case JSON_PARSE_END:
2364 48 : return JSON_EXPECTED_END;
2365 174 : case JSON_PARSE_VALUE:
2366 174 : return JSON_EXPECTED_JSON;
2367 164 : case JSON_PARSE_STRING:
2368 164 : return JSON_EXPECTED_STRING;
2369 56 : case JSON_PARSE_ARRAY_START:
2370 56 : return JSON_EXPECTED_ARRAY_FIRST;
2371 72 : case JSON_PARSE_ARRAY_NEXT:
2372 72 : return JSON_EXPECTED_ARRAY_NEXT;
2373 124 : case JSON_PARSE_OBJECT_START:
2374 124 : return JSON_EXPECTED_OBJECT_FIRST;
2375 128 : case JSON_PARSE_OBJECT_LABEL:
2376 128 : return JSON_EXPECTED_COLON;
2377 172 : case JSON_PARSE_OBJECT_NEXT:
2378 172 : return JSON_EXPECTED_OBJECT_NEXT;
2379 0 : case JSON_PARSE_OBJECT_COMMA:
2380 0 : return JSON_EXPECTED_STRING;
2381 : }
2382 :
2383 : /*
2384 : * We don't use a default: case, so that the compiler will warn about
2385 : * unhandled enum values.
2386 : */
2387 : Assert(false);
2388 0 : return JSON_SUCCESS; /* silence stupider compilers */
2389 : }
2390 :
2391 : /*
2392 : * Construct an (already translated) detail message for a JSON error.
2393 : *
2394 : * The returned pointer should not be freed, the allocation is either static
2395 : * or owned by the JsonLexContext.
2396 : */
2397 : char *
2398 2356 : json_errdetail(JsonParseErrorType error, JsonLexContext *lex)
2399 : {
2400 2356 : if (error == JSON_OUT_OF_MEMORY || lex == &failed_oom)
2401 : {
2402 : /* Short circuit. Allocating anything for this case is unhelpful. */
2403 0 : return _("out of memory");
2404 : }
2405 :
2406 2356 : if (lex->errormsg)
2407 0 : jsonapi_resetStringInfo(lex->errormsg);
2408 : else
2409 2356 : lex->errormsg = jsonapi_makeStringInfo();
2410 :
2411 : /*
2412 : * A helper for error messages that should print the current token. The
2413 : * format must contain exactly one %.*s specifier.
2414 : */
2415 : #define json_token_error(lex, format) \
2416 : jsonapi_appendStringInfo((lex)->errormsg, _(format), \
2417 : (int) ((lex)->token_terminator - (lex)->token_start), \
2418 : (lex)->token_start);
2419 :
2420 2356 : switch (error)
2421 : {
2422 0 : case JSON_INCOMPLETE:
2423 : case JSON_SUCCESS:
2424 : /* fall through to the error code after switch */
2425 0 : break;
2426 0 : case JSON_INVALID_LEXER_TYPE:
2427 0 : if (lex->incremental)
2428 0 : return _("Recursive descent parser cannot use incremental lexer.");
2429 : else
2430 0 : return _("Incremental parser requires incremental lexer.");
2431 512 : case JSON_NESTING_TOO_DEEP:
2432 512 : return (_("JSON nested too deep, maximum permitted depth is 6400."));
2433 132 : case JSON_ESCAPING_INVALID:
2434 132 : json_token_error(lex, "Escape sequence \"\\%.*s\" is invalid.");
2435 132 : break;
2436 116 : case JSON_ESCAPING_REQUIRED:
2437 116 : jsonapi_appendStringInfo(lex->errormsg,
2438 116 : _("Character with value 0x%02x must be escaped."),
2439 116 : (unsigned char) *(lex->token_terminator));
2440 116 : break;
2441 48 : case JSON_EXPECTED_END:
2442 48 : json_token_error(lex, "Expected end of input, but found \"%.*s\".");
2443 48 : break;
2444 56 : case JSON_EXPECTED_ARRAY_FIRST:
2445 56 : json_token_error(lex, "Expected array element or \"]\", but found \"%.*s\".");
2446 56 : break;
2447 72 : case JSON_EXPECTED_ARRAY_NEXT:
2448 72 : json_token_error(lex, "Expected \",\" or \"]\", but found \"%.*s\".");
2449 72 : break;
2450 128 : case JSON_EXPECTED_COLON:
2451 128 : json_token_error(lex, "Expected \":\", but found \"%.*s\".");
2452 128 : break;
2453 120 : case JSON_EXPECTED_JSON:
2454 120 : json_token_error(lex, "Expected JSON value, but found \"%.*s\".");
2455 120 : break;
2456 190 : case JSON_EXPECTED_MORE:
2457 190 : return _("The input string ended unexpectedly.");
2458 124 : case JSON_EXPECTED_OBJECT_FIRST:
2459 124 : json_token_error(lex, "Expected string or \"}\", but found \"%.*s\".");
2460 124 : break;
2461 172 : case JSON_EXPECTED_OBJECT_NEXT:
2462 172 : json_token_error(lex, "Expected \",\" or \"}\", but found \"%.*s\".");
2463 172 : break;
2464 164 : case JSON_EXPECTED_STRING:
2465 164 : json_token_error(lex, "Expected string, but found \"%.*s\".");
2466 164 : break;
2467 414 : case JSON_INVALID_TOKEN:
2468 414 : json_token_error(lex, "Token \"%.*s\" is invalid.");
2469 414 : break;
2470 0 : case JSON_OUT_OF_MEMORY:
2471 : /* should have been handled above; use the error path */
2472 0 : break;
2473 24 : case JSON_UNICODE_CODE_POINT_ZERO:
2474 24 : return _("\\u0000 cannot be converted to text.");
2475 36 : case JSON_UNICODE_ESCAPE_FORMAT:
2476 36 : return _("\"\\u\" must be followed by four hexadecimal digits.");
2477 0 : case JSON_UNICODE_HIGH_ESCAPE:
2478 : /* note: this case is only reachable in frontend not backend */
2479 0 : return _("Unicode escape values cannot be used for code point values above 007F when the encoding is not UTF8.");
2480 0 : case JSON_UNICODE_UNTRANSLATABLE:
2481 :
2482 : /*
2483 : * Note: this case is only reachable in backend and not frontend.
2484 : * #ifdef it away so the frontend doesn't try to link against
2485 : * backend functionality.
2486 : */
2487 : #ifndef FRONTEND
2488 0 : return psprintf(_("Unicode escape value could not be translated to the server's encoding %s."),
2489 : GetDatabaseEncodingName());
2490 : #else
2491 : Assert(false);
2492 0 : break;
2493 : #endif
2494 12 : case JSON_UNICODE_HIGH_SURROGATE:
2495 12 : return _("Unicode high surrogate must not follow a high surrogate.");
2496 36 : case JSON_UNICODE_LOW_SURROGATE:
2497 36 : return _("Unicode low surrogate must follow a high surrogate.");
2498 0 : case JSON_SEM_ACTION_FAILED:
2499 : /* fall through to the error code after switch */
2500 0 : break;
2501 : }
2502 : #undef json_token_error
2503 :
2504 : /* Note that lex->errormsg can be NULL in shlib code. */
2505 1546 : if (lex->errormsg && lex->errormsg->len == 0)
2506 : {
2507 : /*
2508 : * We don't use a default: case, so that the compiler will warn about
2509 : * unhandled enum values. But this needs to be here anyway to cover
2510 : * the possibility of an incorrect input.
2511 : */
2512 0 : jsonapi_appendStringInfo(lex->errormsg,
2513 : "unexpected json parse error type: %d",
2514 : (int) error);
2515 : }
2516 :
2517 : #ifdef JSONAPI_USE_PQEXPBUFFER
2518 632 : if (PQExpBufferBroken(lex->errormsg))
2519 0 : return _("out of memory while constructing error description");
2520 : #endif
2521 :
2522 1546 : return lex->errormsg->data;
2523 : }
|
