sJson (also known as Secure JSON) is a thread-safe, dynamic-allocation-free, cross-compilable JSON parser based on a PAL (Platform Abstraction Layer) architecture writed in c99
| 1 | /** |
| 2 | * @author Seregon |
| 3 | * @file json_pal.h |
| 4 | * @brief Safe, fast, single-header JSON library with Platform Abstraction Layer |
| 5 | * @version 1.1.0 |
| 6 | * @date 2026 |
| 7 | * |
| 8 | * SPDX-License-Identifier: GPL-3.0 license |
| 9 | * |
| 10 | * ============================================================================ |
| 11 | * QUICK START |
| 12 | * ============================================================================ |
| 13 | * |
| 14 | * // In ONE translation unit: |
| 15 | * #define JSON_IMPLEMENTATION |
| 16 | * #include "json_pal.h" |
| 17 | * |
| 18 | * // Parse: |
| 19 | * JsonArena* arena = json_arena_create(NULL, 64 * 1024); |
| 20 | * JsonError err; |
| 21 | * JsonValue* root = json_parse_cstr(arena, "[1,2,3]", &err); |
| 22 | * |
| 23 | * // Query: |
| 24 | * JsonValue* item = json_path(root, "[1]", NULL); |
| 25 | * |
| 26 | * // Cleanup: |
| 27 | * json_arena_destroy(arena); |
| 28 | * |
| 29 | * ============================================================================ |
| 30 | * PAL OVERRIDES (define BEFORE including) |
| 31 | * ============================================================================ |
| 32 | * |
| 33 | * #define JSON_MEMCPY my_memcpy // void*(dst, src, n) |
| 34 | * #define JSON_MEMSET my_memset // void*(dst, c, n) |
| 35 | * #define JSON_MEMCMP my_memcmp // int(a, b, n) |
| 36 | * #define JSON_STRLEN my_strlen // size_t(s) |
| 37 | * #define JSON_ASSERT my_assert // void(expr) |
| 38 | * #define JSON_STRTOD my_strtod // double(s, end) |
| 39 | * #define JSON_NO_STDLIB // Don't include stdlib headers |
| 40 | * |
| 41 | * ============================================================================ |
| 42 | * CONFIGURATION (define BEFORE including) |
| 43 | * ============================================================================ |
| 44 | * |
| 45 | * #define JSON_MAX_DEPTH 64 // Max nesting depth (stack overflow guard) |
| 46 | * #define JSON_MAX_STRING_LEN 65536 // Max string length |
| 47 | * #define JSON_MAX_NODES 65536 // Max total value nodes per parse |
| 48 | * #define JSON_INTROSORT_THRESH 16 // Insertion sort below this size |
| 49 | * #define JSON_ARENA_ALIGN 8 // Arena allocation alignment |
| 50 | * |
| 51 | * ============================================================================ |
| 52 | * DESIGN RATIONALE |
| 53 | * ============================================================================ |
| 54 | * |
| 55 | * MEMORY: Arena allocator - zero fragmentation, O(1) alloc, bulk free. |
| 56 | * No malloc/free during parse; caller provides backing memory. |
| 57 | * |
| 58 | * PARSER: Iterative state machine - no recursion, bounded stack usage, |
| 59 | * safe on deeply nested input. Explicit parse stack on the heap. |
| 60 | * |
| 61 | * NUMBERS: Detects integer vs float at parse time. Integers stored as |
| 62 | * int64_t (exact). Floats stored as double with NaN/Inf guard. |
| 63 | * |
| 64 | * OBJECTS: Keys stored insertion-order. Separate sorted index array built |
| 65 | * on demand via introsort, enabling O(log n) binary search lookup. |
| 66 | * |
| 67 | * SORT: Introsort (quicksort + heapsort fallback + insertion sort leaf). |
| 68 | * Guaranteed O(n log n) worst-case. Insertion sort for n < 16. |
| 69 | * |
| 70 | * UTF-8: Full validation during string parsing (RFC 3629). |
| 71 | * \uXXXX escape sequences with surrogate pair support. |
| 72 | * |
| 73 | * @note Thread-safety: NOT thread-safe. Use separate arenas per thread. |
| 74 | * @note MISRA C:2012 advisory (dynamic alloc via user-provided hooks only). |
| 75 | */ |
| 76 | |
| 77 | #ifndef JSON_PAL_H |
| 78 | #define JSON_PAL_H |
| 79 | |
| 80 | /* ============================================================================ |
| 81 | * §1 PLATFORM DETECTION |
| 82 | * ============================================================================ */ |
| 83 | |
| 84 | #if defined(_MSC_VER) |
| 85 | # define JSON_CC_MSVC 1 |
| 86 | #elif defined(__clang__) |
| 87 | # define JSON_CC_CLANG 1 |
| 88 | #elif defined(__GNUC__) |
| 89 | # define JSON_CC_GCC 1 |
| 90 | #else |
| 91 | # define JSON_CC_OTHER 1 |
| 92 | #endif |
| 93 | |
| 94 | #if defined(_WIN32) || defined(_WIN64) |
| 95 | # define JSON_OS_WINDOWS 1 |
| 96 | #elif defined(__linux__) |
| 97 | # define JSON_OS_LINUX 1 |
| 98 | #elif defined(__APPLE__) |
| 99 | # define JSON_OS_APPLE 1 |
| 100 | #else |
| 101 | # define JSON_OS_BAREMETAL 1 /* Embedded / RTOS */ |
| 102 | #endif |
| 103 | |
| 104 | /* Endianness */ |
| 105 | #if defined(__BYTE_ORDER__) && defined(__ORDER_BIG_ENDIAN__) |
| 106 | # if (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__) |
| 107 | # define JSON_BIG_ENDIAN 1 |
| 108 | # endif |
| 109 | #endif |
| 110 | #ifndef JSON_BIG_ENDIAN |
| 111 | # define JSON_LITTLE_ENDIAN 1 |
| 112 | #endif |
| 113 | |
| 114 | /* ============================================================================ |
| 115 | * §2 PORTABLE TYPE SYSTEM |
| 116 | * ============================================================================ */ |
| 117 | |
| 118 | #ifndef JSON_NO_STDLIB |
| 119 | # include <stdint.h> |
| 120 | # include <stddef.h> |
| 121 | # include <stdbool.h> |
| 122 | # include <string.h> |
| 123 | # include <math.h> |
| 124 | # include <float.h> |
| 125 | # include <limits.h> |
| 126 | #else |
| 127 | /* Minimal type definitions for fully freestanding environments */ |
| 128 | typedef unsigned char uint8_t; |
| 129 | typedef unsigned short uint16_t; |
| 130 | typedef unsigned int uint32_t; |
| 131 | typedef unsigned long long uint64_t; |
| 132 | typedef signed char int8_t; |
| 133 | typedef signed short int16_t; |
| 134 | typedef signed int int32_t; |
| 135 | typedef signed long long int64_t; |
| 136 | typedef unsigned long size_t; |
| 137 | typedef int ptrdiff_t; |
| 138 | typedef int bool; |
| 139 | # define true 1 |
| 140 | # define false 0 |
| 141 | # define NULL ((void*)0) |
| 142 | #endif |
| 143 | |
| 144 | /* Compile-time type size checks */ |
| 145 | #if defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 201112L) |
| 146 | _Static_assert(sizeof(uint8_t) == 1, "uint8_t must be 1 byte"); |
| 147 | _Static_assert(sizeof(uint32_t) == 4, "uint32_t must be 4 bytes"); |
| 148 | _Static_assert(sizeof(uint64_t) == 8, "uint64_t must be 8 bytes"); |
| 149 | _Static_assert(sizeof(int64_t) == 8, "int64_t must be 8 bytes"); |
| 150 | #endif |
| 151 | |
| 152 | /* ============================================================================ |
| 153 | * §3 COMPILER HINTS & PORTABILITY MACROS |
| 154 | * ============================================================================ */ |
| 155 | |
| 156 | #if defined(JSON_CC_GCC) || defined(JSON_CC_CLANG) |
| 157 | # define JSON_INLINE __attribute__((always_inline)) static inline |
| 158 | # define JSON_NOINLINE __attribute__((noinline)) |
| 159 | # define JSON_LIKELY(x) __builtin_expect(!!(x), 1) |
| 160 | # define JSON_UNLIKELY(x) __builtin_expect(!!(x), 0) |
| 161 | # define JSON_RESTRICT __restrict__ |
| 162 | # define JSON_PURE __attribute__((pure)) |
| 163 | # define JSON_NORETURN __attribute__((noreturn)) |
| 164 | #elif defined(JSON_CC_MSVC) |
| 165 | # define JSON_INLINE __forceinline static |
| 166 | # define JSON_NOINLINE __declspec(noinline) |
| 167 | # define JSON_LIKELY(x) (x) |
| 168 | # define JSON_UNLIKELY(x) (x) |
| 169 | # define JSON_RESTRICT __restrict |
| 170 | # define JSON_PURE |
| 171 | # define JSON_NORETURN __declspec(noreturn) |
| 172 | #else |
| 173 | # define JSON_INLINE static inline |
| 174 | # define JSON_NOINLINE |
| 175 | # define JSON_LIKELY(x) (x) |
| 176 | # define JSON_UNLIKELY(x) (x) |
| 177 | # define JSON_RESTRICT |
| 178 | # define JSON_PURE |
| 179 | # define JSON_NORETURN |
| 180 | #endif |
| 181 | |
| 182 | /* ============================================================================ |
| 183 | * §4 PAL OVERRIDABLE PRIMITIVES |
| 184 | * ============================================================================ */ |
| 185 | |
| 186 | #ifndef JSON_MEMCPY |
| 187 | # define JSON_MEMCPY memcpy |
| 188 | #endif |
| 189 | #ifndef JSON_MEMSET |
| 190 | # define JSON_MEMSET memset |
| 191 | #endif |
| 192 | #ifndef JSON_MEMCMP |
| 193 | # define JSON_MEMCMP memcmp |
| 194 | #endif |
| 195 | #ifndef JSON_STRLEN |
| 196 | # define JSON_STRLEN strlen |
| 197 | #endif |
| 198 | #ifndef JSON_STRTOD |
| 199 | # define JSON_STRTOD strtod |
| 200 | #endif |
| 201 | #ifndef JSON_STRTOLL |
| 202 | # define JSON_STRTOLL strtoll |
| 203 | #endif |
| 204 | |
| 205 | #ifndef JSON_ASSERT |
| 206 | # ifdef NDEBUG |
| 207 | # define JSON_ASSERT(expr) ((void)(expr)) |
| 208 | # else |
| 209 | # include <assert.h> |
| 210 | # define JSON_ASSERT(expr) assert(expr) |
| 211 | # endif |
| 212 | #endif |
| 213 | |
| 214 | /* ============================================================================ |
| 215 | * §5 CONFIGURATION DEFAULTS |
| 216 | * ============================================================================ */ |
| 217 | |
| 218 | #ifndef JSON_MAX_DEPTH |
| 219 | # define JSON_MAX_DEPTH 64U |
| 220 | #endif |
| 221 | #ifndef JSON_MAX_STRING_LEN |
| 222 | # define JSON_MAX_STRING_LEN (1024U * 1024U) |
| 223 | #endif |
| 224 | #ifndef JSON_MAX_NODES |
| 225 | # define JSON_MAX_NODES (1024U * 1024U) |
| 226 | #endif |
| 227 | #ifndef JSON_INTROSORT_THRESH |
| 228 | # define JSON_INTROSORT_THRESH 16U |
| 229 | #endif |
| 230 | #ifndef JSON_ARENA_ALIGN |
| 231 | # define JSON_ARENA_ALIGN 8U |
| 232 | #endif |
| 233 | #ifndef JSON_ARENA_MIN_BLOCK |
| 234 | # define JSON_ARENA_MIN_BLOCK (4U * 1024U) |
| 235 | #endif |
| 236 | #ifndef JSON_MAX_ARRAY_LEN |
| 237 | # define JSON_MAX_ARRAY_LEN (1024U * 1024U) /**< Max items per array/object */ |
| 238 | #endif |
| 239 | |
| 240 | /* ============================================================================ |
| 241 | * §6 ERROR CODES |
| 242 | * ============================================================================ */ |
| 243 | |
| 244 | typedef enum { |
| 245 | JSON_OK = 0, /**< Success */ |
| 246 | JSON_ERR_NULL_PARAM = -1, /**< NULL pointer passed to API */ |
| 247 | JSON_ERR_OOM = -2, /**< Out of arena memory */ |
| 248 | JSON_ERR_INVALID_JSON = -3, /**< Malformed JSON syntax */ |
| 249 | JSON_ERR_UNEXPECTED_EOF = -4, /**< Input ended unexpectedly */ |
| 250 | JSON_ERR_DEPTH_EXCEEDED = -5, /**< Nesting deeper than JSON_MAX_DEPTH */ |
| 251 | JSON_ERR_INVALID_STRING = -6, /**< Bad escape sequence or encoding */ |
| 252 | JSON_ERR_INVALID_NUMBER = -7, /**< Malformed number literal */ |
| 253 | JSON_ERR_INVALID_UTF8 = -8, /**< Invalid UTF-8 sequence */ |
| 254 | JSON_ERR_OVERFLOW = -9, /**< Arithmetic or buffer overflow */ |
| 255 | JSON_ERR_NOT_FOUND = -10, /**< Key or index not found */ |
| 256 | JSON_ERR_TYPE_MISMATCH = -11, /**< Wrong type for requested operation */ |
| 257 | JSON_ERR_BUFFER_TOO_SMALL= -12, /**< Output buffer too small */ |
| 258 | JSON_ERR_NODE_LIMIT = -13 /**< Exceeded JSON_MAX_NODES */ |
| 259 | } JsonError; |
| 260 | |
| 261 | /* ============================================================================ |
| 262 | * §7 VALUE TYPES & STRUCTURES |
| 263 | * ============================================================================ */ |
| 264 | |
| 265 | typedef enum { |
| 266 | JSON_NULL = 0, |
| 267 | JSON_BOOL = 1, |
| 268 | JSON_INTEGER = 2, /**< Stored as int64_t — exact, no rounding */ |
| 269 | JSON_FLOAT = 3, /**< Stored as double — NaN/Inf never produced */ |
| 270 | JSON_STRING = 4, /**< UTF-8, NOT NUL-terminated; use .len */ |
| 271 | JSON_ARRAY = 5, |
| 272 | JSON_OBJECT = 6 |
| 273 | } JsonType; |
| 274 | |
| 275 | typedef struct JsonValue JsonValue; |
| 276 | |
| 277 | typedef struct { |
| 278 | const char* data; /**< Points into arena; NOT NUL-terminated */ |
| 279 | uint32_t len; |
| 280 | uint32_t hash; /**< FNV-1a for fast key lookup */ |
| 281 | } JsonStr; |
| 282 | |
| 283 | typedef struct { |
| 284 | JsonValue** items; |
| 285 | uint32_t len; |
| 286 | uint32_t cap; |
| 287 | } JsonArr; |
| 288 | |
| 289 | typedef struct { |
| 290 | JsonStr key; |
| 291 | JsonValue* val; |
| 292 | } JsonPair; |
| 293 | |
| 294 | typedef struct { |
| 295 | JsonPair* pairs; /**< Insertion-order pairs */ |
| 296 | uint32_t* sorted_idx; /**< Indices sorted by key (built on demand) */ |
| 297 | uint32_t len; |
| 298 | uint32_t cap; |
| 299 | bool is_sorted; |
| 300 | } JsonObj; |
| 301 | |
| 302 | struct JsonValue { |
| 303 | JsonType type; |
| 304 | union { |
| 305 | bool b; /* JSON_BOOL */ |
| 306 | int64_t i; /* JSON_INTEGER */ |
| 307 | double f; /* JSON_FLOAT */ |
| 308 | JsonStr s; /* JSON_STRING */ |
| 309 | JsonArr a; /* JSON_ARRAY */ |
| 310 | JsonObj o; /* JSON_OBJECT */ |
| 311 | } v; |
| 312 | }; |
| 313 | |
| 314 | /* ============================================================================ |
| 315 | * §8 ARENA ALLOCATOR |
| 316 | * ============================================================================ */ |
| 317 | |
| 318 | /** |
| 319 | * @brief Custom allocator hooks. All fields optional (NULL → use libc). |
| 320 | * @note realloc receives old_size to support arena-style backing stores. |
| 321 | */ |
| 322 | typedef struct { |
| 323 | void* (*alloc) (void* ctx, size_t size); |
| 324 | void (*free) (void* ctx, void* ptr); |
| 325 | void* (*realloc)(void* ctx, void* ptr, size_t old_size, size_t new_size); |
| 326 | void* ctx; |
| 327 | } JsonAllocator; |
| 328 | |
| 329 | typedef struct JsonArenaBlock JsonArenaBlock; |
| 330 | struct JsonArenaBlock { |
| 331 | JsonArenaBlock* next; |
| 332 | size_t cap; |
| 333 | size_t used; |
| 334 | /* data[] follows in memory */ |
| 335 | }; |
| 336 | |
| 337 | typedef struct { |
| 338 | JsonArenaBlock* current; |
| 339 | JsonArenaBlock* first; |
| 340 | size_t block_size; /**< Default block allocation size */ |
| 341 | size_t total_bytes; /**< Cumulative bytes allocated */ |
| 342 | uint32_t node_count; /**< JsonValue nodes allocated (limit check) */ |
| 343 | JsonAllocator backing; |
| 344 | } JsonArena; |
| 345 | |
| 346 | /* ============================================================================ |
| 347 | * §9 SERIALIZER OPTIONS |
| 348 | * ============================================================================ */ |
| 349 | |
| 350 | typedef struct { |
| 351 | bool pretty; /**< Enable indented, multi-line output */ |
| 352 | uint32_t indent; /**< Spaces per level (default: 2) */ |
| 353 | bool sort_keys; /**< Alphabetic key ordering in output */ |
| 354 | bool ascii_only; /**< Escape all non-ASCII as \uXXXX */ |
| 355 | bool trailing_nl; /**< Append '\n' at end */ |
| 356 | } JsonWriteOpts; |
| 357 | |
| 358 | /* ============================================================================ |
| 359 | * §10 PUBLIC API DECLARATIONS |
| 360 | * ============================================================================ */ |
| 361 | |
| 362 | /* --- Arena --- */ |
| 363 | |
| 364 | /** |
| 365 | * @brief Create arena. Pass NULL allocator to use libc malloc/free/realloc. |
| 366 | * @param alloc Backing allocator (NULL = libc). Copied by value. |
| 367 | * @param block_size Initial block size in bytes (0 = default 64 KiB). |
| 368 | * @return New arena, or NULL on allocation failure. |
| 369 | * @note Thread-safety: NOT thread-safe. |
| 370 | */ |
| 371 | JsonArena* json_arena_create(const JsonAllocator* alloc, size_t block_size); |
| 372 | |
| 373 | /** |
| 374 | * @brief Destroy arena and free all backing memory. |
| 375 | * @note All JsonValue* pointers obtained from this arena become invalid. |
| 376 | */ |
| 377 | void json_arena_destroy(JsonArena* arena); |
| 378 | |
| 379 | /** |
| 380 | * @brief Reset arena: keeps backing blocks, resets all usage to zero. |
| 381 | * Faster than destroy+create for repeated parses. |
| 382 | */ |
| 383 | void json_arena_reset(JsonArena* arena); |
| 384 | |
| 385 | /** |
| 386 | * @brief Low-level aligned allocation from arena. |
| 387 | * @return Pointer to zeroed memory, or NULL on OOM. |
| 388 | * @note Alignment: JSON_ARENA_ALIGN bytes. |
| 389 | */ |
| 390 | void* json_arena_alloc(JsonArena* arena, size_t size); |
| 391 | |
| 392 | /* --- Parsing --- */ |
| 393 | |
| 394 | /** |
| 395 | * @brief Parse JSON from buffer. Result lives in arena. |
| 396 | * @param arena Destination arena (must not be NULL). |
| 397 | * @param src Input buffer (need not be NUL-terminated). |
| 398 | * @param len Length of src in bytes. |
| 399 | * @param err_out If non-NULL, receives error code on failure. |
| 400 | * @return Root JsonValue*, or NULL on error. |
| 401 | * |
| 402 | * @pre arena != NULL |
| 403 | * @pre src != NULL || len == 0 |
| 404 | * @post On success, returned pointer is valid until arena_destroy/reset. |
| 405 | * |
| 406 | * @note WCET depends on input size and JSON_MAX_DEPTH. |
| 407 | * Depth-limited to JSON_MAX_DEPTH to prevent stack exhaustion. |
| 408 | */ |
| 409 | JsonValue* json_parse(JsonArena* arena, const char* src, size_t len, |
| 410 | JsonError* err_out); |
| 411 | |
| 412 | /** |
| 413 | * @brief Convenience wrapper for NUL-terminated strings. |
| 414 | */ |
| 415 | JsonValue* json_parse_cstr(JsonArena* arena, const char* src, JsonError* err_out); |
| 416 | |
| 417 | /* --- Type predicates --- */ |
| 418 | JSON_INLINE bool json_is_null (const JsonValue* v) { return v && v->type == JSON_NULL; } |
| 419 | JSON_INLINE bool json_is_bool (const JsonValue* v) { return v && v->type == JSON_BOOL; } |
| 420 | JSON_INLINE bool json_is_integer(const JsonValue* v) { return v && v->type == JSON_INTEGER; } |
| 421 | JSON_INLINE bool json_is_float (const JsonValue* v) { return v && v->type == JSON_FLOAT; } |
| 422 | JSON_INLINE bool json_is_number (const JsonValue* v) { return v && (v->type == JSON_INTEGER || v->type == JSON_FLOAT); } |
| 423 | JSON_INLINE bool json_is_string (const JsonValue* v) { return v && v->type == JSON_STRING; } |
| 424 | JSON_INLINE bool json_is_array (const JsonValue* v) { return v && v->type == JSON_ARRAY; } |
| 425 | JSON_INLINE bool json_is_object (const JsonValue* v) { return v && v->type == JSON_OBJECT; } |
| 426 | |
| 427 | /* --- Type-safe accessors --- */ |
| 428 | JsonError json_get_bool (const JsonValue* v, bool* out); |
| 429 | JsonError json_get_int (const JsonValue* v, int64_t* out); |
| 430 | JsonError json_get_float (const JsonValue* v, double* out); |
| 431 | JsonError json_get_number (const JsonValue* v, double* out); /**< Accepts int OR float */ |
| 432 | JsonError json_get_string (const JsonValue* v, const char** out, uint32_t* len_out); |
| 433 | JsonError json_get_arr_len(const JsonValue* v, uint32_t* out); |
| 434 | JsonError json_get_obj_len(const JsonValue* v, uint32_t* out); |
| 435 | |
| 436 | /** @brief Get array item by index. O(1). */ |
| 437 | JsonError json_arr_get(const JsonValue* v, uint32_t idx, JsonValue** out); |
| 438 | |
| 439 | /** @brief Get object value by key (NUL-terminated). O(log n) after sort. */ |
| 440 | JsonError json_obj_get(const JsonValue* v, const char* key, JsonValue** out); |
| 441 | |
| 442 | /** @brief Get object value by key+length. O(log n) after sort. */ |
| 443 | JsonError json_obj_get_n(const JsonValue* v, const char* key, size_t klen, |
| 444 | JsonValue** out); |
| 445 | |
| 446 | /** |
| 447 | * @brief Iterate object pairs in insertion order. |
| 448 | * @param idx Pair index [0, len). |
| 449 | * @param key_out If non-NULL, receives key pointer (NOT NUL-terminated). |
| 450 | * @param klen If non-NULL, receives key length. |
| 451 | * @param val_out If non-NULL, receives value pointer. |
| 452 | */ |
| 453 | JsonError json_obj_iter(const JsonValue* v, uint32_t idx, |
| 454 | const char** key_out, uint32_t* klen, |
| 455 | JsonValue** val_out); |
| 456 | |
| 457 | /* --- Path query --- */ |
| 458 | |
| 459 | /** |
| 460 | * @brief Query value by path string. |
| 461 | * |
| 462 | * Syntax: |
| 463 | * .key → object key lookup (dot notation) |
| 464 | * ["key"] → object key lookup (bracket notation, allows spaces/specials) |
| 465 | * [N] → array index (N is non-negative integer) |
| 466 | * |
| 467 | * Example paths: |
| 468 | * "users[0].name" |
| 469 | * ".config[\"host\"]" |
| 470 | * "[2][1]" |
| 471 | * |
| 472 | * @param root Starting node (may be any type; mismatch → NULL). |
| 473 | * @param path Path string (NUL-terminated). |
| 474 | * @param err_out Optional error output. |
| 475 | * @return Matching JsonValue*, or NULL on failure. |
| 476 | */ |
| 477 | JsonValue* json_path(const JsonValue* root, const char* path, JsonError* err_out); |
| 478 | |
| 479 | /* --- Serialization --- */ |
| 480 | |
| 481 | /** |
| 482 | * @brief Serialize JSON value to caller-provided buffer. |
| 483 | * @param buf Output buffer. |
| 484 | * @param buf_size Buffer capacity in bytes. |
| 485 | * @param written If non-NULL, receives bytes written (excl. NUL). |
| 486 | * @param opts Write options (NULL = compact defaults). |
| 487 | * @return JSON_OK, JSON_ERR_NULL_PARAM, or JSON_ERR_BUFFER_TOO_SMALL. |
| 488 | * |
| 489 | * @pre v != NULL, buf != NULL, buf_size >= 1 |
| 490 | * @post On JSON_OK, buf is NUL-terminated. |
| 491 | */ |
| 492 | JsonError json_write(const JsonValue* v, char* buf, size_t buf_size, |
| 493 | size_t* written, const JsonWriteOpts* opts); |
| 494 | |
| 495 | /** |
| 496 | * @brief Serialize to arena-allocated NUL-terminated string. |
| 497 | * @param out Receives pointer to NUL-terminated string. |
| 498 | * @param len_out If non-NULL, receives string length. |
| 499 | */ |
| 500 | JsonError json_write_arena(const JsonValue* v, JsonArena* arena, |
| 501 | char** out, size_t* len_out, |
| 502 | const JsonWriteOpts* opts); |
| 503 | |
| 504 | /* --- Value construction --- */ |
| 505 | JsonValue* json_make_null (JsonArena* arena); |
| 506 | JsonValue* json_make_bool (JsonArena* arena, bool val); |
| 507 | JsonValue* json_make_int (JsonArena* arena, int64_t val); |
| 508 | JsonValue* json_make_float (JsonArena* arena, double val); |
| 509 | JsonValue* json_make_string (JsonArena* arena, const char* s, uint32_t len); |
| 510 | JsonValue* json_make_stringz(JsonArena* arena, const char* s); |
| 511 | JsonValue* json_make_array (JsonArena* arena); |
| 512 | JsonValue* json_make_object (JsonArena* arena); |
| 513 | |
| 514 | JsonError json_arr_push(JsonValue* arr, JsonArena* arena, JsonValue* item); |
| 515 | JsonError json_obj_set (JsonValue* obj, JsonArena* arena, |
| 516 | const char* key, uint32_t klen, JsonValue* val); |
| 517 | JsonError json_obj_setz(JsonValue* obj, JsonArena* arena, |
| 518 | const char* key, JsonValue* val); |
| 519 | |
| 520 | /* --- Utilities --- */ |
| 521 | const char* json_error_str(JsonError err); |
| 522 | uint32_t json_fnv1a(const char* data, size_t len); |
| 523 | |
| 524 | /* Compute output size without writing (pass NULL buf, 0 size, non-NULL written) */ |
| 525 | JsonError json_measure(const JsonValue* v, size_t* size_out, |
| 526 | const JsonWriteOpts* opts); |
| 527 | |
| 528 | /* Deep equality comparison */ |
| 529 | bool json_equal(const JsonValue* a, const JsonValue* b); |
| 530 | |
| 531 | /* Clone value into another arena */ |
| 532 | JsonValue* json_clone(JsonArena* dst, const JsonValue* src); |
| 533 | |
| 534 | |
| 535 | /* ============================================================================ |
| 536 | * §11 IMPLEMENTATION |
| 537 | * ============================================================================ */ |
| 538 | |
| 539 | #ifdef JSON_IMPLEMENTATION |
| 540 | |
| 541 | #ifndef JSON_NO_STDLIB |
| 542 | # include <stdlib.h> /* malloc, free, realloc */ |
| 543 | # include <stdio.h> /* snprintf */ |
| 544 | #endif |
| 545 | |
| 546 | /* ── §11.1 Default libc allocator ─────────────────────────────────────── */ |
| 547 | |
| 548 | static void* json__libc_alloc(void* ctx, size_t sz) |
| 549 | { |
| 550 | (void)ctx; |
| 551 | return malloc(sz); |
| 552 | } |
| 553 | static void json__libc_free(void* ctx, void* p) |
| 554 | { |
| 555 | (void)ctx; |
| 556 | free(p); |
| 557 | } |
| 558 | static void* json__libc_realloc(void* ctx, void* p, size_t old_sz, size_t new_sz) |
| 559 | { |
| 560 | (void)ctx; (void)old_sz; |
| 561 | return realloc(p, new_sz); |
| 562 | } |
| 563 | |
| 564 | static const JsonAllocator k_libc_alloc = { |
| 565 | json__libc_alloc, |
| 566 | json__libc_free, |
| 567 | json__libc_realloc, |
| 568 | NULL |
| 569 | }; |
| 570 | |
| 571 | /* ── §11.2 Arena implementation ────────────────────────────────────────── */ |
| 572 | |
| 573 | /** |
| 574 | * @brief Align sz up to JSON_ARENA_ALIGN boundary. |
| 575 | * @note JSON_ARENA_ALIGN MUST be a power of two. |
| 576 | */ |
| 577 | JSON_INLINE size_t json__align_up(size_t sz) |
| 578 | { |
| 579 | /* DESIGN: Power-of-2 alignment allows bitwise AND instead of modulo. |
| 580 | * WCET: O(1), 2 instructions on ARM Cortex-M. */ |
| 581 | return (sz + (JSON_ARENA_ALIGN - 1U)) & ~(JSON_ARENA_ALIGN - 1U); |
| 582 | } |
| 583 | |
| 584 | static JsonArenaBlock* json__block_new(JsonArena* arena, size_t min_size) |
| 585 | { |
| 586 | size_t block_sz; |
| 587 | JsonArenaBlock* blk; |
| 588 | size_t header_sz; |
| 589 | |
| 590 | JSON_ASSERT(arena != NULL); |
| 591 | |
| 592 | header_sz = json__align_up(sizeof(JsonArenaBlock)); |
| 593 | |
| 594 | /* Ensure block fits header + requested data */ |
| 595 | block_sz = (min_size + header_sz > arena->block_size) |
| 596 | ? json__align_up(min_size + header_sz) |
| 597 | : arena->block_size; |
| 598 | |
| 599 | blk = (JsonArenaBlock*)arena->backing.alloc(arena->backing.ctx, block_sz); |
| 600 | if (JSON_UNLIKELY(blk == NULL)) { return NULL; } |
| 601 | |
| 602 | JSON_MEMSET(blk, 0, sizeof(JsonArenaBlock)); |
| 603 | blk->cap = block_sz - header_sz; |
| 604 | blk->used = 0U; |
| 605 | blk->next = NULL; |
| 606 | return blk; |
| 607 | } |
| 608 | |
| 609 | JSON_INLINE uint8_t* json__block_data(JsonArenaBlock* blk) |
| 610 | { |
| 611 | /* Data starts immediately after the (aligned) block header */ |
| 612 | return (uint8_t*)blk + json__align_up(sizeof(JsonArenaBlock)); |
| 613 | } |
| 614 | |
| 615 | JsonArena* json_arena_create(const JsonAllocator* alloc, size_t block_size) |
| 616 | { |
| 617 | JsonArena* arena; |
| 618 | JsonAllocator eff_alloc; |
| 619 | JsonArenaBlock* first; |
| 620 | |
| 621 | eff_alloc = (alloc != NULL) ? *alloc : k_libc_alloc; |
| 622 | |
| 623 | /* Validate allocator */ |
| 624 | if (eff_alloc.alloc == NULL || eff_alloc.free == NULL) { return NULL; } |
| 625 | |
| 626 | if (block_size == 0U) { |
| 627 | block_size = 64U * 1024U; |
| 628 | } |
| 629 | if (block_size < JSON_ARENA_MIN_BLOCK) { |
| 630 | block_size = JSON_ARENA_MIN_BLOCK; |
| 631 | } |
| 632 | |
| 633 | arena = (JsonArena*)eff_alloc.alloc(eff_alloc.ctx, sizeof(JsonArena)); |
| 634 | if (JSON_UNLIKELY(arena == NULL)) { return NULL; } |
| 635 | JSON_MEMSET(arena, 0, sizeof(JsonArena)); |
| 636 | |
| 637 | arena->backing = eff_alloc; |
| 638 | arena->block_size = block_size; |
| 639 | |
| 640 | first = json__block_new(arena, 0U); |
| 641 | if (JSON_UNLIKELY(first == NULL)) { |
| 642 | eff_alloc.free(eff_alloc.ctx, arena); |
| 643 | return NULL; |
| 644 | } |
| 645 | |
| 646 | arena->first = first; |
| 647 | arena->current = first; |
| 648 | return arena; |
| 649 | } |
| 650 | |
| 651 | void json_arena_destroy(JsonArena* arena) |
| 652 | { |
| 653 | JsonArenaBlock* blk; |
| 654 | JsonArenaBlock* next; |
| 655 | |
| 656 | if (arena == NULL) { return; } |
| 657 | |
| 658 | blk = arena->first; |
| 659 | while (blk != NULL) { |
| 660 | next = blk->next; |
| 661 | arena->backing.free(arena->backing.ctx, blk); |
| 662 | blk = next; |
| 663 | } |
| 664 | arena->backing.free(arena->backing.ctx, arena); |
| 665 | } |
| 666 | |
| 667 | void json_arena_reset(JsonArena* arena) |
| 668 | { |
| 669 | JsonArenaBlock* blk; |
| 670 | |
| 671 | if (arena == NULL) { return; } |
| 672 | |
| 673 | for (blk = arena->first; blk != NULL; blk = blk->next) { |
| 674 | blk->used = 0U; |
| 675 | } |
| 676 | arena->current = arena->first; |
| 677 | arena->total_bytes = 0U; |
| 678 | arena->node_count = 0U; |
| 679 | } |
| 680 | |
| 681 | void* json_arena_alloc(JsonArena* arena, size_t size) |
| 682 | { |
| 683 | size_t aligned_size; |
| 684 | uint8_t* ptr; |
| 685 | JsonArenaBlock* new_blk; |
| 686 | |
| 687 | if (JSON_UNLIKELY(arena == NULL || size == 0U)) { return NULL; } |
| 688 | |
| 689 | aligned_size = json__align_up(size); |
| 690 | |
| 691 | /* Overflow guard on size alignment */ |
| 692 | if (JSON_UNLIKELY(aligned_size < size)) { return NULL; } |
| 693 | |
| 694 | /* Fast path: fits in current block */ |
| 695 | if (JSON_LIKELY(arena->current->used + aligned_size <= arena->current->cap)) { |
| 696 | ptr = json__block_data(arena->current) + arena->current->used; |
| 697 | arena->current->used += aligned_size; |
| 698 | arena->total_bytes += aligned_size; |
| 699 | JSON_MEMSET(ptr, 0, aligned_size); |
| 700 | return ptr; |
| 701 | } |
| 702 | |
| 703 | /* Slow path: allocate a new block */ |
| 704 | new_blk = json__block_new(arena, aligned_size); |
| 705 | if (JSON_UNLIKELY(new_blk == NULL)) { return NULL; } |
| 706 | |
| 707 | /* Append to chain */ |
| 708 | arena->current->next = new_blk; |
| 709 | arena->current = new_blk; |
| 710 | |
| 711 | ptr = json__block_data(new_blk); |
| 712 | new_blk->used = aligned_size; |
| 713 | arena->total_bytes += aligned_size; |
| 714 | JSON_MEMSET(ptr, 0, aligned_size); |
| 715 | return ptr; |
| 716 | } |
| 717 | |
| 718 | /* ── §11.3 FNV-1a Hash ─────────────────────────────────────────────────── */ |
| 719 | |
| 720 | /** |
| 721 | * @brief FNV-1a 32-bit hash for object key lookup. |
| 722 | * Non-cryptographic. Fast. Good distribution for short strings. |
| 723 | * |
| 724 | * @note WCET: O(len). No branches inside loop. |
| 725 | */ |
| 726 | uint32_t json_fnv1a(const char* data, size_t len) |
| 727 | { |
| 728 | static const uint32_t FNV_PRIME = 0x01000193UL; |
| 729 | static const uint32_t FNV_OFFSET = 0x811C9DC5UL; |
| 730 | uint32_t hash = FNV_OFFSET; |
| 731 | size_t i; |
| 732 | |
| 733 | if (data == NULL) { return 0U; } |
| 734 | |
| 735 | for (i = 0U; i < len; i++) { |
| 736 | hash ^= (uint32_t)(uint8_t)data[i]; |
| 737 | hash *= FNV_PRIME; |
| 738 | } |
| 739 | return hash; |
| 740 | } |
| 741 | |
| 742 | /* ── §11.4 Introsort for object key indices ────────────────────────────── */ |
| 743 | |
| 744 | /* |
| 745 | * DESIGN: Introsort combines: |
| 746 | * - Quicksort (median-of-3 pivot, O(n log n) average) |
| 747 | * - Heapsort (fallback when recursion depth > 2*log2(n), O(n log n) worst) |
| 748 | * - Insertion sort (n < JSON_INTROSORT_THRESH, O(n^2) but fast for tiny n) |
| 749 | * |
| 750 | * This guarantees O(n log n) worst-case with excellent cache locality. |
| 751 | * Uses recursion bounded by depth_limit to prevent stack overflow. |
| 752 | */ |
| 753 | |
| 754 | typedef struct { |
| 755 | const JsonPair* pairs; /* Read-only reference to pairs array */ |
| 756 | uint32_t* idx; /* The index array being sorted */ |
| 757 | } JsonSortCtx; |
| 758 | |
| 759 | /** |
| 760 | * @brief Compare two key-index entries by (hash, then lexicographic). |
| 761 | * Hash comparison avoids string compare in the common (no-collision) case. |
| 762 | * @return <0 if a < b, 0 if equal, >0 if a > b |
| 763 | */ |
| 764 | JSON_INLINE int json__key_cmp(const JsonSortCtx* ctx, uint32_t a, uint32_t b) |
| 765 | { |
| 766 | const JsonStr* ka = &ctx->pairs[a].key; |
| 767 | const JsonStr* kb = &ctx->pairs[b].key; |
| 768 | int cmp; |
| 769 | uint32_t min_len; |
| 770 | |
| 771 | JSON_ASSERT(ctx != NULL); |
| 772 | |
| 773 | /* Fast path: hash mismatch */ |
| 774 | if (ka->hash != kb->hash) { |
| 775 | return (ka->hash < kb->hash) ? -1 : 1; |
| 776 | } |
| 777 | |
| 778 | /* Slow path: lexicographic comparison (hash collision) */ |
| 779 | min_len = (ka->len < kb->len) ? ka->len : kb->len; |
| 780 | cmp = JSON_MEMCMP(ka->data, kb->data, (size_t)min_len); |
| 781 | if (cmp != 0) { return cmp; } |
| 782 | if (ka->len < kb->len) { return -1; } |
| 783 | if (ka->len > kb->len) { return 1; } |
| 784 | return 0; |
| 785 | } |
| 786 | |
| 787 | static void json__isort(const JsonSortCtx* ctx, uint32_t* arr, |
| 788 | uint32_t lo, uint32_t hi) |
| 789 | { |
| 790 | uint32_t i, j, key; |
| 791 | |
| 792 | JSON_ASSERT(ctx != NULL && arr != NULL); |
| 793 | |
| 794 | for (i = lo + 1U; i <= hi; i++) { |
| 795 | key = arr[i]; |
| 796 | j = i; |
| 797 | while (j > lo && json__key_cmp(ctx, arr[j - 1U], key) > 0) { |
| 798 | arr[j] = arr[j - 1U]; |
| 799 | j--; |
| 800 | } |
| 801 | arr[j] = key; |
| 802 | } |
| 803 | } |
| 804 | |
| 805 | /* Heapsort helpers */ |
| 806 | static void json__sift_down(const JsonSortCtx* ctx, uint32_t* arr, |
| 807 | uint32_t root, uint32_t end) |
| 808 | { |
| 809 | uint32_t parent, child, tmp; |
| 810 | |
| 811 | JSON_ASSERT(ctx != NULL && arr != NULL); |
| 812 | |
| 813 | parent = root; |
| 814 | for (;;) { |
| 815 | child = 2U * parent + 1U; |
| 816 | if (child > end) { break; } |
| 817 | |
| 818 | if (child + 1U <= end && |
| 819 | json__key_cmp(ctx, arr[child], arr[child + 1U]) < 0) { |
| 820 | child++; |
| 821 | } |
| 822 | if (json__key_cmp(ctx, arr[parent], arr[child]) >= 0) { break; } |
| 823 | |
| 824 | tmp = arr[parent]; arr[parent] = arr[child]; arr[child] = tmp; |
| 825 | parent = child; |
| 826 | } |
| 827 | } |
| 828 | |
| 829 | static void json__heapsort(const JsonSortCtx* ctx, uint32_t* arr, |
| 830 | uint32_t lo, uint32_t hi) |
| 831 | { |
| 832 | uint32_t n, i, tmp; |
| 833 | |
| 834 | JSON_ASSERT(ctx != NULL && arr != NULL && hi >= lo); |
| 835 | |
| 836 | n = hi - lo + 1U; |
| 837 | |
| 838 | /* Build max-heap */ |
| 839 | if (n < 2U) { return; } |
| 840 | i = lo + n / 2U; |
| 841 | while (i-- > lo) { |
| 842 | json__sift_down(ctx, arr + lo, i - lo, n - 1U); |
| 843 | } |
| 844 | |
| 845 | /* Extract */ |
| 846 | for (i = n - 1U; i > 0U; i--) { |
| 847 | tmp = arr[lo]; arr[lo] = arr[lo + i]; arr[lo + i] = tmp; |
| 848 | json__sift_down(ctx, arr + lo, 0U, i - 1U); |
| 849 | } |
| 850 | } |
| 851 | |
| 852 | /* Compute floor(log2(n)) safely */ |
| 853 | JSON_INLINE uint32_t json__ilog2(uint32_t n) |
| 854 | { |
| 855 | uint32_t r = 0U; |
| 856 | while (n > 1U) { n >>= 1U; r++; } |
| 857 | return r; |
| 858 | } |
| 859 | |
| 860 | static void json__introsort(const JsonSortCtx* ctx, uint32_t* arr, |
| 861 | uint32_t lo, uint32_t hi, uint32_t depth) |
| 862 | { |
| 863 | uint32_t pivot, mid, tmp, i, j; |
| 864 | |
| 865 | JSON_ASSERT(ctx != NULL && arr != NULL); |
| 866 | |
| 867 | while (lo < hi) { |
| 868 | uint32_t span = hi - lo + 1U; |
| 869 | |
| 870 | /* Leaf: insertion sort */ |
| 871 | if (span <= JSON_INTROSORT_THRESH) { |
| 872 | json__isort(ctx, arr, lo, hi); |
| 873 | return; |
| 874 | } |
| 875 | |
| 876 | /* Depth exceeded: heapsort to guarantee O(n log n) */ |
| 877 | if (depth == 0U) { |
| 878 | json__heapsort(ctx, arr, lo, hi); |
| 879 | return; |
| 880 | } |
| 881 | |
| 882 | /* Median-of-three pivot selection */ |
| 883 | mid = lo + span / 2U; |
| 884 | |
| 885 | /* Sort lo, mid, hi — bring median to mid */ |
| 886 | if (json__key_cmp(ctx, arr[lo], arr[mid]) > 0) { |
| 887 | tmp = arr[lo]; arr[lo] = arr[mid]; arr[mid] = tmp; |
| 888 | } |
| 889 | if (json__key_cmp(ctx, arr[lo], arr[hi]) > 0) { |
| 890 | tmp = arr[lo]; arr[lo] = arr[hi]; arr[hi] = tmp; |
| 891 | } |
| 892 | if (json__key_cmp(ctx, arr[mid], arr[hi]) > 0) { |
| 893 | tmp = arr[mid]; arr[mid] = arr[hi]; arr[hi] = tmp; |
| 894 | } |
| 895 | |
| 896 | /* Place pivot at hi-1 */ |
| 897 | pivot = arr[mid]; |
| 898 | tmp = arr[mid]; arr[mid] = arr[hi - 1U]; arr[hi - 1U] = tmp; |
| 899 | |
| 900 | /* Partition */ |
| 901 | i = lo; |
| 902 | j = hi - 1U; |
| 903 | for (;;) { |
| 904 | while (json__key_cmp(ctx, arr[++i], pivot) < 0 && i < hi) { } |
| 905 | while (json__key_cmp(ctx, arr[--j], pivot) > 0 && j > lo) { } |
| 906 | if (i >= j) { break; } |
| 907 | tmp = arr[i]; arr[i] = arr[j]; arr[j] = tmp; |
| 908 | } |
| 909 | |
| 910 | /* Restore pivot */ |
| 911 | tmp = arr[i]; arr[i] = arr[hi - 1U]; arr[hi - 1U] = tmp; |
| 912 | |
| 913 | /* Recurse smaller half; iterate larger (tail-call optimisation) */ |
| 914 | if (i - lo < hi - i) { |
| 915 | json__introsort(ctx, arr, lo, i - 1U, depth - 1U); |
| 916 | lo = i + 1U; |
| 917 | } else { |
| 918 | json__introsort(ctx, arr, i + 1U, hi, depth - 1U); |
| 919 | hi = (i > 0U) ? (i - 1U) : 0U; |
| 920 | if (i == 0U) { break; } |
| 921 | } |
| 922 | depth--; |
| 923 | } |
| 924 | } |
| 925 | |
| 926 | /** |
| 927 | * @brief Sort object's sorted_idx array and mark object as sorted. |
| 928 | * @note Idempotent: no-op if already sorted. |
| 929 | * @note WCET: O(n log n) — introsort guarantees this. |
| 930 | */ |
| 931 | static JsonError json__obj_sort(JsonValue* obj, JsonArena* arena) |
| 932 | { |
| 933 | JsonObj* o; |
| 934 | uint32_t i; |
| 935 | JsonSortCtx ctx; |
| 936 | |
| 937 | JSON_ASSERT(obj != NULL && obj->type == JSON_OBJECT); |
| 938 | JSON_ASSERT(arena != NULL); |
| 939 | |
| 940 | o = &obj->v.o; |
| 941 | if (o->is_sorted || o->len == 0U) { return JSON_OK; } |
| 942 | |
| 943 | /* Allocate sorted_idx if not already present */ |
| 944 | if (o->sorted_idx == NULL) { |
| 945 | o->sorted_idx = (uint32_t*)json_arena_alloc( |
| 946 | arena, (size_t)o->cap * sizeof(uint32_t)); |
| 947 | if (JSON_UNLIKELY(o->sorted_idx == NULL)) { return JSON_ERR_OOM; } |
| 948 | } |
| 949 | |
| 950 | for (i = 0U; i < o->len; i++) { |
| 951 | o->sorted_idx[i] = i; |
| 952 | } |
| 953 | |
| 954 | ctx.pairs = o->pairs; |
| 955 | ctx.idx = o->sorted_idx; |
| 956 | |
| 957 | if (o->len > 1U) { |
| 958 | json__introsort(&ctx, o->sorted_idx, 0U, o->len - 1U, |
| 959 | 2U * json__ilog2(o->len)); |
| 960 | } |
| 961 | |
| 962 | o->is_sorted = true; |
| 963 | return JSON_OK; |
| 964 | } |
| 965 | |
| 966 | /* ── §11.5 Lexer ───────────────────────────────────────────────────────── */ |
| 967 | |
| 968 | typedef enum { |
| 969 | TOK_LBRACE = '{', |
| 970 | TOK_RBRACE = '}', |
| 971 | TOK_LBRACKET= '[', |
| 972 | TOK_RBRACKET= ']', |
| 973 | TOK_COLON = ':', |
| 974 | TOK_COMMA = ',', |
| 975 | TOK_TRUE = 't', |
| 976 | TOK_FALSE = 'f', |
| 977 | TOK_NULL_T = 'n', |
| 978 | TOK_STRING = '"', |
| 979 | TOK_NUMBER = '0', |
| 980 | TOK_EOF = 0, |
| 981 | TOK_ERROR = -1 |
| 982 | } JsonTok; |
| 983 | |
| 984 | typedef struct { |
| 985 | const char* src; |
| 986 | size_t len; |
| 987 | size_t pos; |
| 988 | uint32_t line; |
| 989 | uint32_t col; |
| 990 | /* Last parsed token data */ |
| 991 | size_t tok_start; /* Position in src */ |
| 992 | size_t tok_len; /* Length in src (raw, incl. quotes) */ |
| 993 | JsonError err; |
| 994 | } JsonLex; |
| 995 | |
| 996 | JSON_INLINE bool json__is_ws(char c) |
| 997 | { |
| 998 | return (c == ' ') | (c == '\t') | (c == '\r') | (c == '\n'); |
| 999 | } |
| 1000 | |
| 1001 | static void json__lex_skip_ws(JsonLex* l) |
| 1002 | { |
| 1003 | JSON_ASSERT(l != NULL); |
| 1004 | while (l->pos < l->len) { |
| 1005 | char c = l->src[l->pos]; |
| 1006 | if (!json__is_ws(c)) { break; } |
| 1007 | if (c == '\n') { l->line++; l->col = 0U; } |
| 1008 | else { l->col++; } |
| 1009 | l->pos++; |
| 1010 | } |
| 1011 | } |
| 1012 | |
| 1013 | JSON_INLINE char json__peek(const JsonLex* l) |
| 1014 | { |
| 1015 | return (l->pos < l->len) ? l->src[l->pos] : '\0'; |
| 1016 | } |
| 1017 | |
| 1018 | JSON_INLINE char json__consume(JsonLex* l) |
| 1019 | { |
| 1020 | char c = l->src[l->pos++]; |
| 1021 | l->col++; |
| 1022 | return c; |
| 1023 | } |
| 1024 | |
| 1025 | /** @brief Validate UTF-8 continuation byte. */ |
| 1026 | JSON_INLINE bool json__is_cont(uint8_t b) { return (b & 0xC0U) == 0x80U; } |
| 1027 | |
| 1028 | /* ── §11.6 String parsing (with \uXXXX decoding) ──────────────────────── */ |
| 1029 | |
| 1030 | /** @brief Hex digit to value. Returns 0xFF on invalid input. */ |
| 1031 | JSON_INLINE uint8_t json__hex_val(char c) |
| 1032 | { |
| 1033 | if (c >= '0' && c <= '9') { return (uint8_t)(c - '0'); } |
| 1034 | if (c >= 'a' && c <= 'f') { return (uint8_t)(c - 'a' + 10); } |
| 1035 | if (c >= 'A' && c <= 'F') { return (uint8_t)(c - 'A' + 10); } |
| 1036 | return 0xFFU; |
| 1037 | } |
| 1038 | |
| 1039 | /** |
| 1040 | * @brief Parse 4 hex digits into uint16_t. |
| 1041 | * @return true on success, false on non-hex character. |
| 1042 | */ |
| 1043 | static bool json__parse_hex4(const char* s, size_t max, uint16_t* out) |
| 1044 | { |
| 1045 | uint8_t h0, h1, h2, h3; |
| 1046 | |
| 1047 | JSON_ASSERT(s != NULL && out != NULL); |
| 1048 | |
| 1049 | if (max < 4U) { return false; } |
| 1050 | h0 = json__hex_val(s[0]); |
| 1051 | h1 = json__hex_val(s[1]); |
| 1052 | h2 = json__hex_val(s[2]); |
| 1053 | h3 = json__hex_val(s[3]); |
| 1054 | |
| 1055 | if ((h0 | h1 | h2 | h3) == 0xFFU) { return false; } |
| 1056 | *out = (uint16_t)(((uint16_t)h0 << 12U) | |
| 1057 | ((uint16_t)h1 << 8U) | |
| 1058 | ((uint16_t)h2 << 4U) | |
| 1059 | (uint16_t)h3); |
| 1060 | return true; |
| 1061 | } |
| 1062 | |
| 1063 | /** @brief Encode Unicode code point (U+0000..U+10FFFF) as UTF-8. |
| 1064 | * Writes to buf (needs at least 4 bytes). Returns bytes written. */ |
| 1065 | static uint32_t json__encode_utf8(uint32_t cp, uint8_t* buf) |
| 1066 | { |
| 1067 | JSON_ASSERT(buf != NULL); |
| 1068 | |
| 1069 | if (cp < 0x80U) { |
| 1070 | buf[0] = (uint8_t)cp; |
| 1071 | return 1U; |
| 1072 | } else if (cp < 0x800U) { |
| 1073 | buf[0] = (uint8_t)(0xC0U | (cp >> 6U)); |
| 1074 | buf[1] = (uint8_t)(0x80U | (cp & 0x3FU)); |
| 1075 | return 2U; |
| 1076 | } else if (cp < 0x10000U) { |
| 1077 | buf[0] = (uint8_t)(0xE0U | (cp >> 12U)); |
| 1078 | buf[1] = (uint8_t)(0x80U | ((cp >> 6U) & 0x3FU)); |
| 1079 | buf[2] = (uint8_t)(0x80U | (cp & 0x3FU)); |
| 1080 | return 3U; |
| 1081 | } else if (cp <= 0x10FFFFU) { |
| 1082 | buf[0] = (uint8_t)(0xF0U | (cp >> 18U)); |
| 1083 | buf[1] = (uint8_t)(0x80U | ((cp >> 12U) & 0x3FU)); |
| 1084 | buf[2] = (uint8_t)(0x80U | ((cp >> 6U) & 0x3FU)); |
| 1085 | buf[3] = (uint8_t)(0x80U | (cp & 0x3FU)); |
| 1086 | return 4U; |
| 1087 | } |
| 1088 | return 0U; /* Invalid code point */ |
| 1089 | } |
| 1090 | |
| 1091 | /** |
| 1092 | * @brief Parse a JSON string (cursor past opening '"'). Decode all escapes. |
| 1093 | * Result is stored as arena-allocated UTF-8. NOT NUL-terminated. |
| 1094 | * |
| 1095 | * @param[out] str Populated with pointer+length+hash on success. |
| 1096 | * @return JSON_OK or error code. |
| 1097 | */ |
| 1098 | static JsonError json__parse_string(JsonLex* l, JsonArena* arena, JsonStr* str) |
| 1099 | { |
| 1100 | /* Two-pass: first measure, then allocate+copy. |
| 1101 | * This avoids reallocations and is cache-friendly. */ |
| 1102 | size_t start_pos = l->pos; /* After opening quote */ |
| 1103 | size_t out_len = 0U; |
| 1104 | bool has_escape= false; |
| 1105 | size_t scan_pos; |
| 1106 | char* out_buf; |
| 1107 | size_t write_pos; |
| 1108 | |
| 1109 | JSON_ASSERT(l != NULL && arena != NULL && str != NULL); |
| 1110 | |
| 1111 | /* ── Pass 1: Measure output length ─── */ |
| 1112 | scan_pos = l->pos; |
| 1113 | while (scan_pos < l->len) { |
| 1114 | uint8_t c = (uint8_t)l->src[scan_pos]; |
| 1115 | |
| 1116 | if (c == '"') { break; } |
| 1117 | |
| 1118 | if (c == '\\') { |
| 1119 | has_escape = true; |
| 1120 | scan_pos++; |
| 1121 | if (scan_pos >= l->len) { |
| 1122 | l->err = JSON_ERR_UNEXPECTED_EOF; |
| 1123 | return JSON_ERR_UNEXPECTED_EOF; |
| 1124 | } |
| 1125 | switch (l->src[scan_pos]) { |
| 1126 | case '"': case '\\': case '/': |
| 1127 | case 'b': case 'f': case 'n': case 'r': case 't': |
| 1128 | out_len++; |
| 1129 | scan_pos++; |
| 1130 | break; |
| 1131 | case 'u': { |
| 1132 | /* \uXXXX or \uXXXX\uXXXX (surrogate pair) */ |
| 1133 | uint16_t hi_cp; |
| 1134 | scan_pos++; |
| 1135 | if (!json__parse_hex4(l->src + scan_pos, |
| 1136 | l->len - scan_pos, &hi_cp)) { |
| 1137 | l->err = JSON_ERR_INVALID_STRING; |
| 1138 | return JSON_ERR_INVALID_STRING; |
| 1139 | } |
| 1140 | scan_pos += 4U; |
| 1141 | if (hi_cp >= 0xD800U && hi_cp <= 0xDBFFU) { |
| 1142 | /* High surrogate — expect low surrogate */ |
| 1143 | uint16_t lo_cp; |
| 1144 | if (scan_pos + 1U >= l->len || |
| 1145 | l->src[scan_pos] != '\\' || |
| 1146 | l->src[scan_pos + 1U] != 'u') { |
| 1147 | l->err = JSON_ERR_INVALID_STRING; |
| 1148 | return JSON_ERR_INVALID_STRING; |
| 1149 | } |
| 1150 | scan_pos += 2U; |
| 1151 | if (!json__parse_hex4(l->src + scan_pos, |
| 1152 | l->len - scan_pos, &lo_cp)) { |
| 1153 | l->err = JSON_ERR_INVALID_STRING; |
| 1154 | return JSON_ERR_INVALID_STRING; |
| 1155 | } |
| 1156 | scan_pos += 4U; |
| 1157 | if (lo_cp < 0xDC00U || lo_cp > 0xDFFFU) { |
| 1158 | l->err = JSON_ERR_INVALID_STRING; |
| 1159 | return JSON_ERR_INVALID_STRING; |
| 1160 | } |
| 1161 | out_len += 4U; /* UTF-8 for supplementary plane */ |
| 1162 | } else if (hi_cp >= 0xDC00U && hi_cp <= 0xDFFFU) { |
| 1163 | /* Lone low surrogate — invalid */ |
| 1164 | l->err = JSON_ERR_INVALID_STRING; |
| 1165 | return JSON_ERR_INVALID_STRING; |
| 1166 | } else { |
| 1167 | /* DESIGN: exact byte count, not a constant 3. |
| 1168 | * U+0000–U+007F → 1 byte |
| 1169 | * U+0080–U+07FF → 2 bytes |
| 1170 | * U+0800–U+FFFF → 3 bytes |
| 1171 | * Using the maximum (3) over-allocates and corrupts |
| 1172 | * str->len, breaking key lookup and equality. */ |
| 1173 | out_len += (hi_cp < 0x80U) ? 1U |
| 1174 | : (hi_cp < 0x800U) ? 2U : 3U; |
| 1175 | } |
| 1176 | break; |
| 1177 | } |
| 1178 | default: |
| 1179 | l->err = JSON_ERR_INVALID_STRING; |
| 1180 | return JSON_ERR_INVALID_STRING; |
| 1181 | } |
| 1182 | } else if (c < 0x20U) { |
| 1183 | /* Control chars must be escaped */ |
| 1184 | l->err = JSON_ERR_INVALID_STRING; |
| 1185 | return JSON_ERR_INVALID_STRING; |
| 1186 | } else { |
| 1187 | /* UTF-8 multi-byte: accept as-is (validated in pass 2) */ |
| 1188 | out_len++; |
| 1189 | scan_pos++; |
| 1190 | } |
| 1191 | } |
| 1192 | |
| 1193 | if (scan_pos >= l->len) { |
| 1194 | l->err = JSON_ERR_UNEXPECTED_EOF; |
| 1195 | return JSON_ERR_UNEXPECTED_EOF; |
| 1196 | } |
| 1197 | |
| 1198 | if (out_len > JSON_MAX_STRING_LEN) { |
| 1199 | l->err = JSON_ERR_OVERFLOW; |
| 1200 | return JSON_ERR_OVERFLOW; |
| 1201 | } |
| 1202 | |
| 1203 | /* ── Pass 2: Allocate and decode ─── */ |
| 1204 | out_buf = (out_len > 0U) |
| 1205 | ? (char*)json_arena_alloc(arena, out_len) |
| 1206 | : NULL; |
| 1207 | |
| 1208 | if (out_len > 0U && JSON_UNLIKELY(out_buf == NULL)) { |
| 1209 | l->err = JSON_ERR_OOM; |
| 1210 | return JSON_ERR_OOM; |
| 1211 | } |
| 1212 | |
| 1213 | write_pos = 0U; |
| 1214 | l->pos = start_pos; |
| 1215 | |
| 1216 | if (!has_escape) { |
| 1217 | /* Fast path: no escapes — memcpy and validate UTF-8 */ |
| 1218 | if (out_len > 0U) { |
| 1219 | JSON_MEMCPY(out_buf, l->src + l->pos, out_len); |
| 1220 | } |
| 1221 | l->pos += out_len; |
| 1222 | } else { |
| 1223 | /* Slow path: decode escapes */ |
| 1224 | while (l->pos < l->len && l->src[l->pos] != '"') { |
| 1225 | uint8_t c = (uint8_t)l->src[l->pos]; |
| 1226 | if (c == '\\') { |
| 1227 | l->pos++; |
| 1228 | switch (l->src[l->pos]) { |
| 1229 | case '"': out_buf[write_pos++] = '"'; l->pos++; break; |
| 1230 | case '\\': out_buf[write_pos++] = '\\'; l->pos++; break; |
| 1231 | case '/': out_buf[write_pos++] = '/'; l->pos++; break; |
| 1232 | case 'b': out_buf[write_pos++] = '\b'; l->pos++; break; |
| 1233 | case 'f': out_buf[write_pos++] = '\f'; l->pos++; break; |
| 1234 | case 'n': out_buf[write_pos++] = '\n'; l->pos++; break; |
| 1235 | case 'r': out_buf[write_pos++] = '\r'; l->pos++; break; |
| 1236 | case 't': out_buf[write_pos++] = '\t'; l->pos++; break; |
| 1237 | case 'u': { |
| 1238 | uint16_t hi_cp, lo_cp; |
| 1239 | uint32_t codepoint; |
| 1240 | uint32_t bytes; |
| 1241 | l->pos++; |
| 1242 | json__parse_hex4(l->src + l->pos, |
| 1243 | l->len - l->pos, &hi_cp); |
| 1244 | l->pos += 4U; |
| 1245 | if (hi_cp >= 0xD800U && hi_cp <= 0xDBFFU) { |
| 1246 | l->pos += 2U; /* skip \u */ |
| 1247 | json__parse_hex4(l->src + l->pos, |
| 1248 | l->len - l->pos, &lo_cp); |
| 1249 | l->pos += 4U; |
| 1250 | codepoint = 0x10000U + |
| 1251 | ((uint32_t)(hi_cp - 0xD800U) << 10U) + |
| 1252 | (uint32_t)(lo_cp - 0xDC00U); |
| 1253 | } else { |
| 1254 | codepoint = hi_cp; |
| 1255 | } |
| 1256 | bytes = json__encode_utf8(codepoint, |
| 1257 | (uint8_t*)out_buf + write_pos); |
| 1258 | write_pos += bytes; |
| 1259 | break; |
| 1260 | } |
| 1261 | default: l->pos++; break; /* unreachable after pass 1 */ |
| 1262 | } |
| 1263 | } else { |
| 1264 | out_buf[write_pos++] = (char)c; |
| 1265 | l->pos++; |
| 1266 | } |
| 1267 | } |
| 1268 | } |
| 1269 | |
| 1270 | if (l->pos >= l->len) { |
| 1271 | l->err = JSON_ERR_UNEXPECTED_EOF; |
| 1272 | return JSON_ERR_UNEXPECTED_EOF; |
| 1273 | } |
| 1274 | l->pos++; /* Consume closing '"' */ |
| 1275 | |
| 1276 | str->data = out_buf; |
| 1277 | str->len = (uint32_t)out_len; |
| 1278 | str->hash = json_fnv1a(out_buf, out_len); |
| 1279 | return JSON_OK; |
| 1280 | } |
| 1281 | |
| 1282 | /* ── §11.7 Number parsing ──────────────────────────────────────────────── */ |
| 1283 | |
| 1284 | /** |
| 1285 | * @brief Parse JSON number. Detects integer vs float at parse time. |
| 1286 | * |
| 1287 | * Grammar: -? (0 | [1-9][0-9]*) (.[0-9]+)? ([eE][+-]?[0-9]+)? |
| 1288 | * |
| 1289 | * DESIGN: Parse as integer if no '.' or 'e'/'E'. Fall back to strtod only |
| 1290 | * when fractional/exponent parts are present. This preserves exact |
| 1291 | * integer semantics for the common case and avoids locale issues. |
| 1292 | */ |
| 1293 | static JsonError json__parse_number(JsonLex* l, JsonArena* arena, |
| 1294 | JsonValue* out) |
| 1295 | { |
| 1296 | size_t start = l->pos; |
| 1297 | bool negative = false; |
| 1298 | bool is_float = false; |
| 1299 | int64_t ival = 0; |
| 1300 | char num_buf[64]; |
| 1301 | size_t num_len; |
| 1302 | char* end_ptr; |
| 1303 | |
| 1304 | JSON_ASSERT(l != NULL && arena != NULL && out != NULL); |
| 1305 | |
| 1306 | /* Optional minus */ |
| 1307 | if (l->pos < l->len && l->src[l->pos] == '-') { |
| 1308 | negative = true; |
| 1309 | l->pos++; |
| 1310 | } |
| 1311 | |
| 1312 | if (l->pos >= l->len) { |
| 1313 | l->err = JSON_ERR_UNEXPECTED_EOF; |
| 1314 | return JSON_ERR_UNEXPECTED_EOF; |
| 1315 | } |
| 1316 | |
| 1317 | /* Integer part */ |
| 1318 | if (l->src[l->pos] == '0') { |
| 1319 | l->pos++; |
| 1320 | if (l->pos < l->len && l->src[l->pos] >= '0' && l->src[l->pos] <= '9') { |
| 1321 | /* Leading zero not allowed in JSON */ |
| 1322 | l->err = JSON_ERR_INVALID_NUMBER; |
| 1323 | return JSON_ERR_INVALID_NUMBER; |
| 1324 | } |
| 1325 | } else if (l->src[l->pos] >= '1' && l->src[l->pos] <= '9') { |
| 1326 | while (l->pos < l->len && |
| 1327 | l->src[l->pos] >= '0' && l->src[l->pos] <= '9') { |
| 1328 | l->pos++; |
| 1329 | } |
| 1330 | } else { |
| 1331 | l->err = JSON_ERR_INVALID_NUMBER; |
| 1332 | return JSON_ERR_INVALID_NUMBER; |
| 1333 | } |
| 1334 | |
| 1335 | /* Fractional part */ |
| 1336 | if (l->pos < l->len && l->src[l->pos] == '.') { |
| 1337 | is_float = true; |
| 1338 | l->pos++; |
| 1339 | if (l->pos >= l->len || |
| 1340 | l->src[l->pos] < '0' || l->src[l->pos] > '9') { |
| 1341 | l->err = JSON_ERR_INVALID_NUMBER; |
| 1342 | return JSON_ERR_INVALID_NUMBER; |
| 1343 | } |
| 1344 | while (l->pos < l->len && |
| 1345 | l->src[l->pos] >= '0' && l->src[l->pos] <= '9') { |
| 1346 | l->pos++; |
| 1347 | } |
| 1348 | } |
| 1349 | |
| 1350 | /* Exponent part */ |
| 1351 | if (l->pos < l->len && |
| 1352 | (l->src[l->pos] == 'e' || l->src[l->pos] == 'E')) { |
| 1353 | is_float = true; |
| 1354 | l->pos++; |
| 1355 | if (l->pos < l->len && |
| 1356 | (l->src[l->pos] == '+' || l->src[l->pos] == '-')) { |
| 1357 | l->pos++; |
| 1358 | } |
| 1359 | if (l->pos >= l->len || |
| 1360 | l->src[l->pos] < '0' || l->src[l->pos] > '9') { |
| 1361 | l->err = JSON_ERR_INVALID_NUMBER; |
| 1362 | return JSON_ERR_INVALID_NUMBER; |
| 1363 | } |
| 1364 | while (l->pos < l->len && |
| 1365 | l->src[l->pos] >= '0' && l->src[l->pos] <= '9') { |
| 1366 | l->pos++; |
| 1367 | } |
| 1368 | } |
| 1369 | |
| 1370 | num_len = l->pos - start; |
| 1371 | if (num_len >= sizeof(num_buf)) { |
| 1372 | l->err = JSON_ERR_OVERFLOW; |
| 1373 | return JSON_ERR_OVERFLOW; |
| 1374 | } |
| 1375 | |
| 1376 | JSON_MEMCPY(num_buf, l->src + start, num_len); |
| 1377 | num_buf[num_len] = '\0'; |
| 1378 | |
| 1379 | if (is_float) { |
| 1380 | double fval = JSON_STRTOD(num_buf, &end_ptr); |
| 1381 | if (end_ptr != num_buf + num_len) { |
| 1382 | l->err = JSON_ERR_INVALID_NUMBER; |
| 1383 | return JSON_ERR_INVALID_NUMBER; |
| 1384 | } |
| 1385 | /* Safety: NaN/Inf must not be stored */ |
| 1386 | if (JSON_UNLIKELY(isnan(fval) || isinf(fval))) { |
| 1387 | l->err = JSON_ERR_INVALID_NUMBER; |
| 1388 | return JSON_ERR_INVALID_NUMBER; |
| 1389 | } |
| 1390 | out->type = JSON_FLOAT; |
| 1391 | out->v.f = fval; |
| 1392 | } else { |
| 1393 | /* Parse as int64_t */ |
| 1394 | int64_t parsed; |
| 1395 | char* eptr; |
| 1396 | #if defined(JSON_NO_STDLIB) |
| 1397 | /* Minimal strtoll for freestanding environments */ |
| 1398 | parsed = 0; |
| 1399 | size_t di; |
| 1400 | for (di = negative ? 1U : 0U; di < num_len; di++) { |
| 1401 | char d = num_buf[di]; |
| 1402 | if (d < '0' || d > '9') { break; } |
| 1403 | /* Overflow check */ |
| 1404 | if (parsed > (INT64_MAX - (d - '0')) / 10) { |
| 1405 | parsed = negative ? INT64_MIN : INT64_MAX; |
| 1406 | break; |
| 1407 | } |
| 1408 | parsed = parsed * 10 + (d - '0'); |
| 1409 | } |
| 1410 | ival = negative ? -parsed : parsed; |
| 1411 | (void)eptr; |
| 1412 | #else |
| 1413 | parsed = JSON_STRTOLL(num_buf, &eptr, 10); |
| 1414 | if (eptr != num_buf + num_len) { |
| 1415 | l->err = JSON_ERR_INVALID_NUMBER; |
| 1416 | return JSON_ERR_INVALID_NUMBER; |
| 1417 | } |
| 1418 | ival = parsed; |
| 1419 | #endif |
| 1420 | (void)negative; /* Already encoded in the string we passed to strtoll */ |
| 1421 | out->type = JSON_INTEGER; |
| 1422 | out->v.i = ival; |
| 1423 | } |
| 1424 | return JSON_OK; |
| 1425 | } |
| 1426 | |
| 1427 | /* ── §11.8 Iterative Parser ────────────────────────────────────────────── */ |
| 1428 | |
| 1429 | /* |
| 1430 | * DESIGN: Recursive descent is simpler but risks stack overflow on deeply |
| 1431 | * nested input. This iterative implementation uses an explicit heap-allocated |
| 1432 | * parse stack with bounded depth (JSON_MAX_DEPTH), making stack usage O(1) |
| 1433 | * regardless of nesting depth. WCET is O(n * JSON_MAX_DEPTH) in theory, |
| 1434 | * O(n) in practice since depth is constant. |
| 1435 | */ |
| 1436 | |
| 1437 | typedef enum { |
| 1438 | PS_VALUE, /* Expecting any value */ |
| 1439 | PS_ARR_ITEM, /* Inside array: expecting value or ']' */ |
| 1440 | PS_ARR_CONT, /* After array item: expecting ',' or ']' */ |
| 1441 | PS_OBJ_KEY, /* Inside object: expecting key or '}' */ |
| 1442 | PS_OBJ_COLON, /* After key: expecting ':' */ |
| 1443 | PS_OBJ_VAL, /* After colon: expecting value */ |
| 1444 | PS_OBJ_CONT /* After value: expecting ',' or '}' */ |
| 1445 | } ParseState; |
| 1446 | |
| 1447 | typedef struct { |
| 1448 | JsonValue* container; /* Array or Object being built */ |
| 1449 | ParseState state; |
| 1450 | JsonStr pending_key; /* For PS_OBJ_VAL: key waiting for its value */ |
| 1451 | } ParseFrame; |
| 1452 | |
| 1453 | /** |
| 1454 | * @brief Allocate and initialize a new JsonValue node. |
| 1455 | * @note Node count is tracked and bounded by JSON_MAX_NODES. |
| 1456 | */ |
| 1457 | static JsonValue* json__new_node(JsonArena* arena) |
| 1458 | { |
| 1459 | JsonValue* v; |
| 1460 | |
| 1461 | JSON_ASSERT(arena != NULL); |
| 1462 | |
| 1463 | if (JSON_UNLIKELY(arena->node_count >= JSON_MAX_NODES)) { return NULL; } |
| 1464 | v = (JsonValue*)json_arena_alloc(arena, sizeof(JsonValue)); |
| 1465 | if (JSON_LIKELY(v != NULL)) { |
| 1466 | arena->node_count++; |
| 1467 | } |
| 1468 | return v; |
| 1469 | } |
| 1470 | |
| 1471 | /** @brief Grow array capacity (doubling strategy). */ |
| 1472 | static JsonError json__arr_grow(JsonArr* a, JsonArena* arena) |
| 1473 | { |
| 1474 | uint32_t new_cap; |
| 1475 | JsonValue** new_items; |
| 1476 | |
| 1477 | JSON_ASSERT(a != NULL && arena != NULL); |
| 1478 | |
| 1479 | new_cap = (a->cap == 0U) ? 8U : a->cap * 2U; |
| 1480 | |
| 1481 | /* Overflow guard */ |
| 1482 | if (JSON_UNLIKELY(new_cap < a->cap)) { return JSON_ERR_OVERFLOW; } |
| 1483 | if (JSON_UNLIKELY(new_cap > JSON_MAX_ARRAY_LEN)) { return JSON_ERR_OVERFLOW; } |
| 1484 | |
| 1485 | new_items = (JsonValue**)json_arena_alloc(arena, |
| 1486 | (size_t)new_cap * sizeof(JsonValue*)); |
| 1487 | if (JSON_UNLIKELY(new_items == NULL)) { return JSON_ERR_OOM; } |
| 1488 | |
| 1489 | if (a->items != NULL && a->len > 0U) { |
| 1490 | JSON_MEMCPY(new_items, a->items, (size_t)a->len * sizeof(JsonValue*)); |
| 1491 | } |
| 1492 | a->items = new_items; |
| 1493 | a->cap = new_cap; |
| 1494 | return JSON_OK; |
| 1495 | } |
| 1496 | |
| 1497 | /** @brief Grow object capacity (doubling strategy). */ |
| 1498 | static JsonError json__obj_grow(JsonObj* o, JsonArena* arena) |
| 1499 | { |
| 1500 | uint32_t new_cap; |
| 1501 | JsonPair* new_pairs; |
| 1502 | |
| 1503 | JSON_ASSERT(o != NULL && arena != NULL); |
| 1504 | |
| 1505 | new_cap = (o->cap == 0U) ? 8U : o->cap * 2U; |
| 1506 | if (JSON_UNLIKELY(new_cap < o->cap)) { return JSON_ERR_OVERFLOW; } |
| 1507 | |
| 1508 | new_pairs = (JsonPair*)json_arena_alloc(arena, |
| 1509 | (size_t)new_cap * sizeof(JsonPair)); |
| 1510 | if (JSON_UNLIKELY(new_pairs == NULL)) { return JSON_ERR_OOM; } |
| 1511 | |
| 1512 | if (o->pairs != NULL && o->len > 0U) { |
| 1513 | JSON_MEMCPY(new_pairs, o->pairs, (size_t)o->len * sizeof(JsonPair)); |
| 1514 | } |
| 1515 | o->pairs = new_pairs; |
| 1516 | o->cap = new_cap; |
| 1517 | o->sorted_idx= NULL; /* Invalidated on resize */ |
| 1518 | o->is_sorted = false; |
| 1519 | return JSON_OK; |
| 1520 | } |
| 1521 | |
| 1522 | /** |
| 1523 | * @brief Main iterative parser. |
| 1524 | * @note All error paths set err_out before returning NULL. |
| 1525 | * @note WCET: O(n) where n = input length. |
| 1526 | */ |
| 1527 | JsonValue* json_parse(JsonArena* arena, const char* src, size_t len, |
| 1528 | JsonError* err_out) |
| 1529 | { |
| 1530 | JsonLex lex; |
| 1531 | ParseFrame stack[JSON_MAX_DEPTH]; |
| 1532 | uint32_t depth = 0U; |
| 1533 | JsonValue* result = NULL; |
| 1534 | JsonValue* root = NULL; |
| 1535 | JsonError err = JSON_OK; |
| 1536 | bool done = false; |
| 1537 | |
| 1538 | /* Validate parameters */ |
| 1539 | if (arena == NULL) { |
| 1540 | if (err_out != NULL) { *err_out = JSON_ERR_NULL_PARAM; } |
| 1541 | return NULL; |
| 1542 | } |
| 1543 | if (src == NULL && len > 0U) { |
| 1544 | if (err_out != NULL) { *err_out = JSON_ERR_NULL_PARAM; } |
| 1545 | return NULL; |
| 1546 | } |
| 1547 | |
| 1548 | JSON_MEMSET(&lex, 0, sizeof(lex)); |
| 1549 | lex.src = src ? src : ""; |
| 1550 | lex.len = len; |
| 1551 | lex.pos = 0U; |
| 1552 | lex.line = 1U; |
| 1553 | lex.col = 0U; |
| 1554 | |
| 1555 | /* Push initial state */ |
| 1556 | stack[0].state = PS_VALUE; |
| 1557 | stack[0].container = NULL; |
| 1558 | depth = 1U; |
| 1559 | |
| 1560 | #define LEX_EXPECT_LITERAL(literal, tok_type) do { \ |
| 1561 | size_t llen_ = sizeof(literal) - 1U; \ |
| 1562 | if (lex.pos + llen_ > lex.len || \ |
| 1563 | JSON_MEMCMP(lex.src + lex.pos, literal, llen_) != 0) { \ |
| 1564 | err = JSON_ERR_INVALID_JSON; goto parse_error; \ |
| 1565 | } \ |
| 1566 | lex.pos += llen_; \ |
| 1567 | } while(0) |
| 1568 | |
| 1569 | while (!done) { |
| 1570 | ParseFrame* frame = &stack[depth - 1U]; |
| 1571 | |
| 1572 | json__lex_skip_ws(&lex); |
| 1573 | |
| 1574 | if (lex.pos >= lex.len) { |
| 1575 | if (depth == 1U && root != NULL) { |
| 1576 | done = true; |
| 1577 | break; |
| 1578 | } |
| 1579 | err = JSON_ERR_UNEXPECTED_EOF; |
| 1580 | goto parse_error; |
| 1581 | } |
| 1582 | |
| 1583 | switch (frame->state) { |
| 1584 | |
| 1585 | /* ── Expecting a value (any type) ─────────────────────────────── */ |
| 1586 | case PS_VALUE: { |
| 1587 | char c = lex.src[lex.pos]; |
| 1588 | JsonValue* node = json__new_node(arena); |
| 1589 | if (JSON_UNLIKELY(node == NULL)) { |
| 1590 | err = (arena->node_count >= JSON_MAX_NODES) |
| 1591 | ? JSON_ERR_NODE_LIMIT : JSON_ERR_OOM; |
| 1592 | goto parse_error; |
| 1593 | } |
| 1594 | |
| 1595 | if (c == 'n') { |
| 1596 | lex.pos++; |
| 1597 | LEX_EXPECT_LITERAL("ull", 0); |
| 1598 | node->type = JSON_NULL; |
| 1599 | |
| 1600 | } else if (c == 't') { |
| 1601 | lex.pos++; |
| 1602 | LEX_EXPECT_LITERAL("rue", 0); |
| 1603 | node->type = JSON_BOOL; |
| 1604 | node->v.b = true; |
| 1605 | |
| 1606 | } else if (c == 'f') { |
| 1607 | lex.pos++; |
| 1608 | LEX_EXPECT_LITERAL("alse", 0); |
| 1609 | node->type = JSON_BOOL; |
| 1610 | node->v.b = false; |
| 1611 | |
| 1612 | } else if (c == '"') { |
| 1613 | lex.pos++; /* Skip opening quote */ |
| 1614 | node->type = JSON_STRING; |
| 1615 | err = json__parse_string(&lex, arena, &node->v.s); |
| 1616 | if (err != JSON_OK) { goto parse_error; } |
| 1617 | |
| 1618 | } else if (c == '-' || |
| 1619 | (c >= '0' && c <= '9')) { |
| 1620 | err = json__parse_number(&lex, arena, node); |
| 1621 | if (err != JSON_OK) { goto parse_error; } |
| 1622 | |
| 1623 | } else if (c == '[') { |
| 1624 | lex.pos++; |
| 1625 | node->type = JSON_ARRAY; |
| 1626 | node->v.a.items = NULL; |
| 1627 | node->v.a.len = 0U; |
| 1628 | node->v.a.cap = 0U; |
| 1629 | |
| 1630 | /* Push array context */ |
| 1631 | if (depth >= JSON_MAX_DEPTH) { |
| 1632 | err = JSON_ERR_DEPTH_EXCEEDED; |
| 1633 | goto parse_error; |
| 1634 | } |
| 1635 | result = node; |
| 1636 | |
| 1637 | /* Store in parent if applicable, then push new frame */ |
| 1638 | frame->container = node; |
| 1639 | frame->state = PS_ARR_ITEM; |
| 1640 | |
| 1641 | /* Check for empty array */ |
| 1642 | json__lex_skip_ws(&lex); |
| 1643 | if (lex.pos < lex.len && lex.src[lex.pos] == ']') { |
| 1644 | lex.pos++; |
| 1645 | /* Empty array: immediately deliver to parent */ |
| 1646 | goto deliver_to_parent; |
| 1647 | } |
| 1648 | |
| 1649 | /* Push VALUE frame for first element */ |
| 1650 | stack[depth].state = PS_VALUE; |
| 1651 | stack[depth].container = NULL; |
| 1652 | depth++; |
| 1653 | continue; |
| 1654 | |
| 1655 | } else if (c == '{') { |
| 1656 | lex.pos++; |
| 1657 | node->type = JSON_OBJECT; |
| 1658 | node->v.o.pairs = NULL; |
| 1659 | node->v.o.sorted_idx = NULL; |
| 1660 | node->v.o.len = 0U; |
| 1661 | node->v.o.cap = 0U; |
| 1662 | node->v.o.is_sorted= false; |
| 1663 | |
| 1664 | if (depth >= JSON_MAX_DEPTH) { |
| 1665 | err = JSON_ERR_DEPTH_EXCEEDED; |
| 1666 | goto parse_error; |
| 1667 | } |
| 1668 | |
| 1669 | frame->container = node; |
| 1670 | frame->state = PS_OBJ_KEY; |
| 1671 | |
| 1672 | /* Check for empty object */ |
| 1673 | json__lex_skip_ws(&lex); |
| 1674 | if (lex.pos < lex.len && lex.src[lex.pos] == '}') { |
| 1675 | lex.pos++; |
| 1676 | result = node; |
| 1677 | goto deliver_to_parent; |
| 1678 | } |
| 1679 | |
| 1680 | /* |
| 1681 | * DESIGN: this frame has been transformed in-place to |
| 1682 | * PS_OBJ_KEY. The main loop will execute case PS_OBJ_KEY on |
| 1683 | * the next iteration — no extra frame push is needed. |
| 1684 | * Pushing a second PS_OBJ_KEY frame here (as done before this |
| 1685 | * fix) left stack[depth-1] as PS_OBJ_KEY when the closing '}' |
| 1686 | * tried to deliver the object to its parent, causing an |
| 1687 | * "invalid JSON" error on every non-empty object. |
| 1688 | */ |
| 1689 | continue; |
| 1690 | |
| 1691 | } else { |
| 1692 | err = JSON_ERR_INVALID_JSON; |
| 1693 | goto parse_error; |
| 1694 | } |
| 1695 | |
| 1696 | result = node; |
| 1697 | |
| 1698 | /* Deliver completed scalar to parent */ |
| 1699 | deliver_to_parent: |
| 1700 | { |
| 1701 | if (depth == 1U) { |
| 1702 | /* Top-level value */ |
| 1703 | root = result; |
| 1704 | done = true; |
| 1705 | break; |
| 1706 | } |
| 1707 | |
| 1708 | /* Pop current frame and deliver to parent */ |
| 1709 | depth--; |
| 1710 | ParseFrame* parent = &stack[depth - 1U]; |
| 1711 | |
| 1712 | if (parent->state == PS_ARR_ITEM || |
| 1713 | parent->state == PS_ARR_CONT) { |
| 1714 | /* Add to array */ |
| 1715 | JsonArr* arr = &parent->container->v.a; |
| 1716 | if (arr->len >= arr->cap) { |
| 1717 | err = json__arr_grow(arr, arena); |
| 1718 | if (err != JSON_OK) { goto parse_error; } |
| 1719 | } |
| 1720 | arr->items[arr->len++] = result; |
| 1721 | parent->state = PS_ARR_CONT; |
| 1722 | |
| 1723 | } else if (parent->state == PS_OBJ_VAL) { |
| 1724 | /* Add key-value pair to object */ |
| 1725 | JsonObj* obj = &parent->container->v.o; |
| 1726 | if (obj->len >= obj->cap) { |
| 1727 | err = json__obj_grow(obj, arena); |
| 1728 | if (err != JSON_OK) { goto parse_error; } |
| 1729 | } |
| 1730 | obj->pairs[obj->len].key = parent->pending_key; |
| 1731 | obj->pairs[obj->len].val = result; |
| 1732 | obj->len++; |
| 1733 | parent->state = PS_OBJ_CONT; |
| 1734 | } else { |
| 1735 | err = JSON_ERR_INVALID_JSON; |
| 1736 | goto parse_error; |
| 1737 | } |
| 1738 | } |
| 1739 | break; |
| 1740 | } /* PS_VALUE */ |
| 1741 | |
| 1742 | /* ── After array item: ',' or ']' ─────────────────────────────── */ |
| 1743 | case PS_ARR_CONT: { |
| 1744 | char c = lex.src[lex.pos]; |
| 1745 | if (c == ']') { |
| 1746 | lex.pos++; |
| 1747 | result = frame->container; |
| 1748 | /* Pop and deliver array to its parent */ |
| 1749 | if (depth == 1U) { |
| 1750 | root = result; |
| 1751 | done = true; |
| 1752 | } else { |
| 1753 | depth--; |
| 1754 | ParseFrame* parent = &stack[depth - 1U]; |
| 1755 | JsonArr* arr; |
| 1756 | |
| 1757 | if (parent->state == PS_ARR_ITEM || |
| 1758 | parent->state == PS_ARR_CONT) { |
| 1759 | arr = &parent->container->v.a; |
| 1760 | if (arr->len >= arr->cap) { |
| 1761 | err = json__arr_grow(arr, arena); |
| 1762 | if (err != JSON_OK) { goto parse_error; } |
| 1763 | } |
| 1764 | arr->items[arr->len++] = result; |
| 1765 | parent->state = PS_ARR_CONT; |
| 1766 | } else if (parent->state == PS_OBJ_VAL) { |
| 1767 | JsonObj* obj = &parent->container->v.o; |
| 1768 | if (obj->len >= obj->cap) { |
| 1769 | err = json__obj_grow(obj, arena); |
| 1770 | if (err != JSON_OK) { goto parse_error; } |
| 1771 | } |
| 1772 | obj->pairs[obj->len].key = parent->pending_key; |
| 1773 | obj->pairs[obj->len].val = result; |
| 1774 | obj->len++; |
| 1775 | parent->state = PS_OBJ_CONT; |
| 1776 | } else if (parent->state == PS_VALUE) { |
| 1777 | root = result; done = true; |
| 1778 | } else { |
| 1779 | err = JSON_ERR_INVALID_JSON; goto parse_error; |
| 1780 | } |
| 1781 | } |
| 1782 | } else if (c == ',') { |
| 1783 | lex.pos++; |
| 1784 | frame->state = PS_ARR_ITEM; |
| 1785 | /* Push VALUE frame for next element */ |
| 1786 | if (depth >= JSON_MAX_DEPTH) { |
| 1787 | err = JSON_ERR_DEPTH_EXCEEDED; goto parse_error; |
| 1788 | } |
| 1789 | stack[depth].state = PS_VALUE; |
| 1790 | stack[depth].container = NULL; |
| 1791 | depth++; |
| 1792 | } else { |
| 1793 | err = JSON_ERR_INVALID_JSON; goto parse_error; |
| 1794 | } |
| 1795 | break; |
| 1796 | } |
| 1797 | |
| 1798 | /* ── Object: expecting key ─────────────────────────────────────── */ |
| 1799 | case PS_OBJ_KEY: { |
| 1800 | char c = lex.src[lex.pos]; |
| 1801 | if (c == '"') { |
| 1802 | JsonStr key; |
| 1803 | lex.pos++; /* Skip quote */ |
| 1804 | err = json__parse_string(&lex, arena, &key); |
| 1805 | if (err != JSON_OK) { goto parse_error; } |
| 1806 | frame->pending_key = key; |
| 1807 | frame->state = PS_OBJ_COLON; |
| 1808 | } else { |
| 1809 | err = JSON_ERR_INVALID_JSON; goto parse_error; |
| 1810 | } |
| 1811 | break; |
| 1812 | } |
| 1813 | |
| 1814 | /* ── After object key: expect ':' ─────────────────────────────── */ |
| 1815 | case PS_OBJ_COLON: { |
| 1816 | if (lex.src[lex.pos] != ':') { |
| 1817 | err = JSON_ERR_INVALID_JSON; goto parse_error; |
| 1818 | } |
| 1819 | lex.pos++; |
| 1820 | frame->state = PS_OBJ_VAL; |
| 1821 | /* Push VALUE frame for the value */ |
| 1822 | if (depth >= JSON_MAX_DEPTH) { |
| 1823 | err = JSON_ERR_DEPTH_EXCEEDED; goto parse_error; |
| 1824 | } |
| 1825 | stack[depth].state = PS_VALUE; |
| 1826 | stack[depth].container = NULL; |
| 1827 | depth++; |
| 1828 | break; |
| 1829 | } |
| 1830 | |
| 1831 | /* ── After object value: ',' or '}' ───────────────────────────── */ |
| 1832 | case PS_OBJ_CONT: { |
| 1833 | char c = lex.src[lex.pos]; |
| 1834 | if (c == '}') { |
| 1835 | lex.pos++; |
| 1836 | result = frame->container; |
| 1837 | if (depth == 1U) { |
| 1838 | root = result; done = true; |
| 1839 | } else { |
| 1840 | depth--; |
| 1841 | ParseFrame* parent = &stack[depth - 1U]; |
| 1842 | if (parent->state == PS_ARR_ITEM || |
| 1843 | parent->state == PS_ARR_CONT) { |
| 1844 | JsonArr* arr = &parent->container->v.a; |
| 1845 | if (arr->len >= arr->cap) { |
| 1846 | err = json__arr_grow(arr, arena); |
| 1847 | if (err != JSON_OK) { goto parse_error; } |
| 1848 | } |
| 1849 | arr->items[arr->len++] = result; |
| 1850 | parent->state = PS_ARR_CONT; |
| 1851 | } else if (parent->state == PS_OBJ_VAL) { |
| 1852 | JsonObj* obj = &parent->container->v.o; |
| 1853 | if (obj->len >= obj->cap) { |
| 1854 | err = json__obj_grow(obj, arena); |
| 1855 | if (err != JSON_OK) { goto parse_error; } |
| 1856 | } |
| 1857 | obj->pairs[obj->len].key = parent->pending_key; |
| 1858 | obj->pairs[obj->len].val = result; |
| 1859 | obj->len++; |
| 1860 | parent->state = PS_OBJ_CONT; |
| 1861 | } else if (parent->state == PS_VALUE) { |
| 1862 | root = result; done = true; |
| 1863 | } else { |
| 1864 | err = JSON_ERR_INVALID_JSON; goto parse_error; |
| 1865 | } |
| 1866 | } |
| 1867 | } else if (c == ',') { |
| 1868 | lex.pos++; |
| 1869 | frame->state = PS_OBJ_KEY; |
| 1870 | } else { |
| 1871 | err = JSON_ERR_INVALID_JSON; goto parse_error; |
| 1872 | } |
| 1873 | break; |
| 1874 | } |
| 1875 | |
| 1876 | case PS_ARR_ITEM: |
| 1877 | case PS_OBJ_VAL: |
| 1878 | default: |
| 1879 | /* These states should be handled by pushing a PS_VALUE frame */ |
| 1880 | err = JSON_ERR_INVALID_JSON; |
| 1881 | goto parse_error; |
| 1882 | } /* switch */ |
| 1883 | } /* while */ |
| 1884 | |
| 1885 | /* Verify no trailing non-whitespace */ |
| 1886 | if (!done) { err = JSON_ERR_UNEXPECTED_EOF; goto parse_error; } |
| 1887 | json__lex_skip_ws(&lex); |
| 1888 | if (lex.pos < lex.len) { |
| 1889 | err = JSON_ERR_INVALID_JSON; |
| 1890 | goto parse_error; |
| 1891 | } |
| 1892 | |
| 1893 | #undef LEX_EXPECT_LITERAL |
| 1894 | |
| 1895 | if (err_out != NULL) { *err_out = JSON_OK; } |
| 1896 | return root; |
| 1897 | |
| 1898 | parse_error: |
| 1899 | if (err_out != NULL) { *err_out = (err != JSON_OK) ? err : JSON_ERR_INVALID_JSON; } |
| 1900 | return NULL; |
| 1901 | } |
| 1902 | |
| 1903 | JsonValue* json_parse_cstr(JsonArena* arena, const char* src, JsonError* err_out) |
| 1904 | { |
| 1905 | if (src == NULL) { |
| 1906 | if (err_out != NULL) { *err_out = JSON_ERR_NULL_PARAM; } |
| 1907 | return NULL; |
| 1908 | } |
| 1909 | return json_parse(arena, src, JSON_STRLEN(src), err_out); |
| 1910 | } |
| 1911 | |
| 1912 | /* ── §11.9 Type-safe accessors ─────────────────────────────────────────── */ |
| 1913 | |
| 1914 | #define JSON__CHECK_V(v) do { if ((v) == NULL) return JSON_ERR_NULL_PARAM; } while(0) |
| 1915 | #define JSON__CHECK_OUT(o) do { if ((o) == NULL) return JSON_ERR_NULL_PARAM; } while(0) |
| 1916 | |
| 1917 | JsonError json_get_bool(const JsonValue* v, bool* out) |
| 1918 | { |
| 1919 | JSON__CHECK_V(v); JSON__CHECK_OUT(out); |
| 1920 | if (v->type != JSON_BOOL) { return JSON_ERR_TYPE_MISMATCH; } |
| 1921 | *out = v->v.b; |
| 1922 | return JSON_OK; |
| 1923 | } |
| 1924 | |
| 1925 | JsonError json_get_int(const JsonValue* v, int64_t* out) |
| 1926 | { |
| 1927 | JSON__CHECK_V(v); JSON__CHECK_OUT(out); |
| 1928 | if (v->type == JSON_INTEGER) { *out = v->v.i; return JSON_OK; } |
| 1929 | if (v->type == JSON_FLOAT) { *out = (int64_t)v->v.f; return JSON_OK; } |
| 1930 | return JSON_ERR_TYPE_MISMATCH; |
| 1931 | } |
| 1932 | |
| 1933 | JsonError json_get_float(const JsonValue* v, double* out) |
| 1934 | { |
| 1935 | JSON__CHECK_V(v); JSON__CHECK_OUT(out); |
| 1936 | if (v->type == JSON_FLOAT) { *out = v->v.f; return JSON_OK; } |
| 1937 | if (v->type == JSON_INTEGER) { *out = (double)v->v.i; return JSON_OK; } |
| 1938 | return JSON_ERR_TYPE_MISMATCH; |
| 1939 | } |
| 1940 | |
| 1941 | JsonError json_get_number(const JsonValue* v, double* out) |
| 1942 | { |
| 1943 | return json_get_float(v, out); |
| 1944 | } |
| 1945 | |
| 1946 | JsonError json_get_string(const JsonValue* v, const char** out, uint32_t* len_out) |
| 1947 | { |
| 1948 | JSON__CHECK_V(v); JSON__CHECK_OUT(out); |
| 1949 | if (v->type != JSON_STRING) { return JSON_ERR_TYPE_MISMATCH; } |
| 1950 | *out = v->v.s.data ? v->v.s.data : ""; |
| 1951 | if (len_out != NULL) { *len_out = v->v.s.len; } |
| 1952 | return JSON_OK; |
| 1953 | } |
| 1954 | |
| 1955 | JsonError json_get_arr_len(const JsonValue* v, uint32_t* out) |
| 1956 | { |
| 1957 | JSON__CHECK_V(v); JSON__CHECK_OUT(out); |
| 1958 | if (v->type != JSON_ARRAY) { return JSON_ERR_TYPE_MISMATCH; } |
| 1959 | *out = v->v.a.len; |
| 1960 | return JSON_OK; |
| 1961 | } |
| 1962 | |
| 1963 | JsonError json_get_obj_len(const JsonValue* v, uint32_t* out) |
| 1964 | { |
| 1965 | JSON__CHECK_V(v); JSON__CHECK_OUT(out); |
| 1966 | if (v->type != JSON_OBJECT) { return JSON_ERR_TYPE_MISMATCH; } |
| 1967 | *out = v->v.o.len; |
| 1968 | return JSON_OK; |
| 1969 | } |
| 1970 | |
| 1971 | JsonError json_arr_get(const JsonValue* v, uint32_t idx, JsonValue** out) |
| 1972 | { |
| 1973 | JSON__CHECK_V(v); JSON__CHECK_OUT(out); |
| 1974 | if (v->type != JSON_ARRAY) { return JSON_ERR_TYPE_MISMATCH; } |
| 1975 | if (idx >= v->v.a.len) { return JSON_ERR_NOT_FOUND; } |
| 1976 | JSON_ASSERT(v->v.a.items != NULL); |
| 1977 | *out = v->v.a.items[idx]; |
| 1978 | return JSON_OK; |
| 1979 | } |
| 1980 | |
| 1981 | JsonError json_obj_get_n(const JsonValue* v, const char* key, size_t klen, |
| 1982 | JsonValue** out) |
| 1983 | { |
| 1984 | const JsonObj* o; |
| 1985 | uint32_t hash; |
| 1986 | int32_t lo, hi, mid; |
| 1987 | int cmp; |
| 1988 | uint32_t idx; |
| 1989 | |
| 1990 | JSON__CHECK_V(v); JSON__CHECK_OUT(out); |
| 1991 | if (key == NULL) { return JSON_ERR_NULL_PARAM; } |
| 1992 | if (v->type != JSON_OBJECT) { return JSON_ERR_TYPE_MISMATCH; } |
| 1993 | |
| 1994 | o = &v->v.o; |
| 1995 | if (o->len == 0U) { return JSON_ERR_NOT_FOUND; } |
| 1996 | |
| 1997 | /* If not yet sorted, do a linear scan */ |
| 1998 | if (!o->is_sorted || o->sorted_idx == NULL) { |
| 1999 | uint32_t i; |
| 2000 | hash = json_fnv1a(key, klen); |
| 2001 | for (i = 0U; i < o->len; i++) { |
| 2002 | const JsonStr* k = &o->pairs[i].key; |
| 2003 | if (k->hash == hash && k->len == (uint32_t)klen && |
| 2004 | JSON_MEMCMP(k->data, key, klen) == 0) { |
| 2005 | *out = o->pairs[i].val; |
| 2006 | return JSON_OK; |
| 2007 | } |
| 2008 | } |
| 2009 | return JSON_ERR_NOT_FOUND; |
| 2010 | } |
| 2011 | |
| 2012 | /* Binary search on sorted_idx */ |
| 2013 | hash = json_fnv1a(key, klen); |
| 2014 | lo = 0; hi = (int32_t)o->len - 1; |
| 2015 | |
| 2016 | while (lo <= hi) { |
| 2017 | mid = lo + (hi - lo) / 2; |
| 2018 | idx = o->sorted_idx[(uint32_t)mid]; |
| 2019 | const JsonStr* k = &o->pairs[idx].key; |
| 2020 | |
| 2021 | if (k->hash < hash) { |
| 2022 | lo = mid + 1; |
| 2023 | } else if (k->hash > hash) { |
| 2024 | hi = mid - 1; |
| 2025 | } else { |
| 2026 | /* Hash match: compare lexicographically */ |
| 2027 | uint32_t min_len = (k->len < (uint32_t)klen) ? k->len : (uint32_t)klen; |
| 2028 | cmp = JSON_MEMCMP(k->data, key, (size_t)min_len); |
| 2029 | if (cmp == 0) { |
| 2030 | if (k->len < (uint32_t)klen) { cmp = -1; } |
| 2031 | else if (k->len > (uint32_t)klen) { cmp = 1; } |
| 2032 | } |
| 2033 | if (cmp == 0) { |
| 2034 | *out = o->pairs[idx].val; |
| 2035 | return JSON_OK; |
| 2036 | } else if (cmp < 0) { |
| 2037 | lo = mid + 1; |
| 2038 | } else { |
| 2039 | hi = mid - 1; |
| 2040 | } |
| 2041 | } |
| 2042 | } |
| 2043 | return JSON_ERR_NOT_FOUND; |
| 2044 | } |
| 2045 | |
| 2046 | JsonError json_obj_get(const JsonValue* v, const char* key, JsonValue** out) |
| 2047 | { |
| 2048 | if (key == NULL) { return JSON_ERR_NULL_PARAM; } |
| 2049 | return json_obj_get_n(v, key, JSON_STRLEN(key), out); |
| 2050 | } |
| 2051 | |
| 2052 | JsonError json_obj_iter(const JsonValue* v, uint32_t idx, |
| 2053 | const char** key_out, uint32_t* klen, |
| 2054 | JsonValue** val_out) |
| 2055 | { |
| 2056 | JSON__CHECK_V(v); |
| 2057 | if (v->type != JSON_OBJECT) { return JSON_ERR_TYPE_MISMATCH; } |
| 2058 | if (idx >= v->v.o.len) { return JSON_ERR_NOT_FOUND; } |
| 2059 | |
| 2060 | if (key_out != NULL) { *key_out = v->v.o.pairs[idx].key.data; } |
| 2061 | if (klen != NULL) { *klen = v->v.o.pairs[idx].key.len; } |
| 2062 | if (val_out != NULL) { *val_out = v->v.o.pairs[idx].val; } |
| 2063 | return JSON_OK; |
| 2064 | } |
| 2065 | |
| 2066 | /* ── §11.10 Path query ─────────────────────────────────────────────────── */ |
| 2067 | |
| 2068 | JsonValue* json_path(const JsonValue* root, const char* path, JsonError* err_out) |
| 2069 | { |
| 2070 | const JsonValue* cur; |
| 2071 | const char* p; |
| 2072 | JsonError err; |
| 2073 | char key_buf[512]; |
| 2074 | |
| 2075 | if (root == NULL || path == NULL) { |
| 2076 | if (err_out != NULL) { *err_out = JSON_ERR_NULL_PARAM; } |
| 2077 | return NULL; |
| 2078 | } |
| 2079 | |
| 2080 | cur = root; |
| 2081 | p = path; |
| 2082 | |
| 2083 | while (*p != '\0') { |
| 2084 | /* Skip leading dot if present */ |
| 2085 | if (*p == '.') { p++; } |
| 2086 | |
| 2087 | if (*p == '[') { |
| 2088 | /* Bracket notation: [N] or ["key"] */ |
| 2089 | p++; |
| 2090 | if (*p == '"') { |
| 2091 | /* ["string key"] */ |
| 2092 | const char* ks = ++p; |
| 2093 | while (*p != '\0' && *p != '"') { p++; } |
| 2094 | if (*p == '\0') { |
| 2095 | err = JSON_ERR_INVALID_JSON; goto fail; |
| 2096 | } |
| 2097 | size_t klen = (size_t)(p - ks); |
| 2098 | p++; /* skip closing " */ |
| 2099 | if (*p != ']') { err = JSON_ERR_INVALID_JSON; goto fail; } |
| 2100 | p++; |
| 2101 | JsonValue* out; |
| 2102 | err = json_obj_get_n(cur, ks, klen, &out); |
| 2103 | if (err != JSON_OK) { goto fail; } |
| 2104 | cur = out; |
| 2105 | } else if (*p >= '0' && *p <= '9') { |
| 2106 | /* [N] */ |
| 2107 | uint32_t idx = 0U; |
| 2108 | while (*p >= '0' && *p <= '9') { |
| 2109 | uint32_t prev = idx; |
| 2110 | idx = idx * 10U + (uint32_t)(*p - '0'); |
| 2111 | if (idx < prev) { err = JSON_ERR_OVERFLOW; goto fail; } |
| 2112 | p++; |
| 2113 | } |
| 2114 | if (*p != ']') { err = JSON_ERR_INVALID_JSON; goto fail; } |
| 2115 | p++; |
| 2116 | JsonValue* out; |
| 2117 | err = json_arr_get(cur, idx, &out); |
| 2118 | if (err != JSON_OK) { goto fail; } |
| 2119 | cur = out; |
| 2120 | } else { |
| 2121 | err = JSON_ERR_INVALID_JSON; goto fail; |
| 2122 | } |
| 2123 | } else if ((*p >= 'a' && *p <= 'z') || |
| 2124 | (*p >= 'A' && *p <= 'Z') || |
| 2125 | (*p == '_') || |
| 2126 | ((uint8_t)*p >= 0x80U)) { |
| 2127 | /* Dot notation: .identifier */ |
| 2128 | const char* ks = p; |
| 2129 | while (*p != '\0' && *p != '.' && *p != '[') { p++; } |
| 2130 | size_t klen = (size_t)(p - ks); |
| 2131 | if (klen >= sizeof(key_buf)) { err = JSON_ERR_OVERFLOW; goto fail; } |
| 2132 | JSON_MEMCPY(key_buf, ks, klen); |
| 2133 | key_buf[klen] = '\0'; |
| 2134 | JsonValue* out; |
| 2135 | err = json_obj_get_n(cur, ks, klen, &out); |
| 2136 | if (err != JSON_OK) { goto fail; } |
| 2137 | cur = out; |
| 2138 | } else { |
| 2139 | err = JSON_ERR_INVALID_JSON; goto fail; |
| 2140 | } |
| 2141 | } |
| 2142 | |
| 2143 | if (err_out != NULL) { *err_out = JSON_OK; } |
| 2144 | return (JsonValue*)(uintptr_t)cur; /* Cast away const for API convenience */ |
| 2145 | |
| 2146 | fail: |
| 2147 | if (err_out != NULL) { *err_out = err; } |
| 2148 | return NULL; |
| 2149 | } |
| 2150 | |
| 2151 | /* ── §11.11 Serializer ─────────────────────────────────────────────────── */ |
| 2152 | |
| 2153 | typedef struct { |
| 2154 | char* buf; |
| 2155 | size_t cap; |
| 2156 | size_t used; |
| 2157 | bool measure_only; /* When true: just count, don't write */ |
| 2158 | JsonError err; |
| 2159 | } JsonWriter; |
| 2160 | |
| 2161 | JSON_INLINE JsonError jw_put(JsonWriter* w, char c) |
| 2162 | { |
| 2163 | if (w->measure_only) { w->used++; return JSON_OK; } |
| 2164 | if (w->used >= w->cap - 1U) { w->err = JSON_ERR_BUFFER_TOO_SMALL; return JSON_ERR_BUFFER_TOO_SMALL; } |
| 2165 | w->buf[w->used++] = c; |
| 2166 | return JSON_OK; |
| 2167 | } |
| 2168 | |
| 2169 | static JsonError jw_puts(JsonWriter* w, const char* s, size_t n) |
| 2170 | { |
| 2171 | size_t i; |
| 2172 | if (w->measure_only) { w->used += n; return JSON_OK; } |
| 2173 | for (i = 0U; i < n; i++) { |
| 2174 | if (jw_put(w, s[i]) != JSON_OK) { return JSON_ERR_BUFFER_TOO_SMALL; } |
| 2175 | } |
| 2176 | return JSON_OK; |
| 2177 | } |
| 2178 | |
| 2179 | static JsonError jw_indent(JsonWriter* w, uint32_t depth, uint32_t indent_sz) |
| 2180 | { |
| 2181 | uint32_t i; |
| 2182 | for (i = 0U; i < depth * indent_sz; i++) { |
| 2183 | if (jw_put(w, ' ') != JSON_OK) { return JSON_ERR_BUFFER_TOO_SMALL; } |
| 2184 | } |
| 2185 | return JSON_OK; |
| 2186 | } |
| 2187 | |
| 2188 | /* Integer-to-decimal without sprintf (embedded-safe) */ |
| 2189 | static JsonError jw_int64(JsonWriter* w, int64_t v) |
| 2190 | { |
| 2191 | char tmp[22]; /* max int64 = 20 digits + sign + NUL */ |
| 2192 | uint32_t len = 0U; |
| 2193 | bool neg = (v < 0); |
| 2194 | uint64_t abs_v; |
| 2195 | |
| 2196 | if (neg) { |
| 2197 | abs_v = (v == INT64_MIN) ? (uint64_t)INT64_MAX + 1ULL |
| 2198 | : (uint64_t)(-v); |
| 2199 | if (jw_put(w, '-') != JSON_OK) { return JSON_ERR_BUFFER_TOO_SMALL; } |
| 2200 | } else { |
| 2201 | abs_v = (uint64_t)v; |
| 2202 | } |
| 2203 | |
| 2204 | if (abs_v == 0U) { return jw_put(w, '0'); } |
| 2205 | |
| 2206 | while (abs_v > 0U) { |
| 2207 | tmp[len++] = (char)('0' + (int)(abs_v % 10U)); |
| 2208 | abs_v /= 10U; |
| 2209 | } |
| 2210 | |
| 2211 | /* Reverse */ |
| 2212 | { |
| 2213 | uint32_t i; |
| 2214 | for (i = 0U; i < len; i++) { |
| 2215 | if (jw_put(w, tmp[len - 1U - i]) != JSON_OK) { |
| 2216 | return JSON_ERR_BUFFER_TOO_SMALL; |
| 2217 | } |
| 2218 | } |
| 2219 | } |
| 2220 | return JSON_OK; |
| 2221 | } |
| 2222 | |
| 2223 | /* Double-to-string (uses snprintf if available, else fallback) */ |
| 2224 | static JsonError jw_double(JsonWriter* w, double v) |
| 2225 | { |
| 2226 | char tmp[32]; |
| 2227 | size_t len; |
| 2228 | |
| 2229 | #ifndef JSON_NO_STDLIB |
| 2230 | /* Use snprintf for correct double rendering */ |
| 2231 | int r = snprintf(tmp, sizeof(tmp), "%.17g", v); |
| 2232 | if (r < 0 || (size_t)r >= sizeof(tmp)) { return JSON_ERR_OVERFLOW; } |
| 2233 | len = (size_t)r; |
| 2234 | /* Ensure there's a decimal point or exponent so it parses as float */ |
| 2235 | { |
| 2236 | bool has_dot = false; |
| 2237 | size_t i; |
| 2238 | for (i = 0U; i < len; i++) { |
| 2239 | if (tmp[i] == '.' || tmp[i] == 'e' || tmp[i] == 'E') { |
| 2240 | has_dot = true; break; |
| 2241 | } |
| 2242 | } |
| 2243 | if (!has_dot && len + 2U < sizeof(tmp)) { |
| 2244 | tmp[len++] = '.'; tmp[len++] = '0'; tmp[len] = '\0'; |
| 2245 | } |
| 2246 | } |
| 2247 | #else |
| 2248 | /* Freestanding fallback: integer representation if lossless */ |
| 2249 | int64_t as_int = (int64_t)v; |
| 2250 | if ((double)as_int == v) { |
| 2251 | snprintf(tmp, sizeof(tmp), "%lld.0", (long long)as_int); |
| 2252 | } else { |
| 2253 | snprintf(tmp, sizeof(tmp), "%g", v); |
| 2254 | } |
| 2255 | len = JSON_STRLEN(tmp); |
| 2256 | #endif |
| 2257 | return jw_puts(w, tmp, len); |
| 2258 | } |
| 2259 | |
| 2260 | /* Write escaped JSON string content (no surrounding quotes) */ |
| 2261 | static JsonError jw_string_content(JsonWriter* w, const char* s, uint32_t len, |
| 2262 | bool ascii_only) |
| 2263 | { |
| 2264 | uint32_t i = 0U; |
| 2265 | |
| 2266 | while (i < len) { |
| 2267 | uint8_t c = (uint8_t)s[i]; |
| 2268 | |
| 2269 | if (c < 0x20U || c == '"' || c == '\\' || |
| 2270 | (ascii_only && c >= 0x80U)) { |
| 2271 | /* Escape */ |
| 2272 | char esc_buf[7]; /* \uXXXX + NUL */ |
| 2273 | if (c == '"') { esc_buf[0]='\\'; esc_buf[1]='"'; jw_puts(w, esc_buf, 2U); } |
| 2274 | else if (c == '\\') { esc_buf[0]='\\'; esc_buf[1]='\\'; jw_puts(w, esc_buf, 2U); } |
| 2275 | else if (c == '\n') { esc_buf[0]='\\'; esc_buf[1]='n'; jw_puts(w, esc_buf, 2U); } |
| 2276 | else if (c == '\r') { esc_buf[0]='\\'; esc_buf[1]='r'; jw_puts(w, esc_buf, 2U); } |
| 2277 | else if (c == '\t') { esc_buf[0]='\\'; esc_buf[1]='t'; jw_puts(w, esc_buf, 2U); } |
| 2278 | else if (c == '\b') { esc_buf[0]='\\'; esc_buf[1]='b'; jw_puts(w, esc_buf, 2U); } |
| 2279 | else if (c == '\f') { esc_buf[0]='\\'; esc_buf[1]='f'; jw_puts(w, esc_buf, 2U); } |
| 2280 | else { |
| 2281 | /* \uXXXX */ |
| 2282 | static const char hex[] = "0123456789abcdef"; |
| 2283 | esc_buf[0]='\\'; esc_buf[1]='u'; |
| 2284 | esc_buf[2]='0'; esc_buf[3]='0'; |
| 2285 | esc_buf[4]=hex[(c>>4)&0xFU]; |
| 2286 | esc_buf[5]=hex[c&0xFU]; |
| 2287 | jw_puts(w, esc_buf, 6U); |
| 2288 | } |
| 2289 | i++; |
| 2290 | } else { |
| 2291 | jw_put(w, (char)c); |
| 2292 | i++; |
| 2293 | } |
| 2294 | } |
| 2295 | return w->err; |
| 2296 | } |
| 2297 | |
| 2298 | /** @brief Core recursive serializer. Uses C stack but depth is bounded at parse time. */ |
| 2299 | static JsonError json__write_value(JsonWriter* w, const JsonValue* v, |
| 2300 | uint32_t depth, |
| 2301 | const JsonWriteOpts* opts) |
| 2302 | { |
| 2303 | JsonError err; |
| 2304 | |
| 2305 | JSON_ASSERT(v != NULL && opts != NULL); |
| 2306 | |
| 2307 | switch (v->type) { |
| 2308 | case JSON_NULL: |
| 2309 | return jw_puts(w, "null", 4U); |
| 2310 | |
| 2311 | case JSON_BOOL: |
| 2312 | return v->v.b ? jw_puts(w, "true", 4U) : jw_puts(w, "false", 5U); |
| 2313 | |
| 2314 | case JSON_INTEGER: |
| 2315 | return jw_int64(w, v->v.i); |
| 2316 | |
| 2317 | case JSON_FLOAT: |
| 2318 | return jw_double(w, v->v.f); |
| 2319 | |
| 2320 | case JSON_STRING: |
| 2321 | err = jw_put(w, '"'); |
| 2322 | if (err != JSON_OK) { return err; } |
| 2323 | err = jw_string_content(w, v->v.s.data ? v->v.s.data : "", |
| 2324 | v->v.s.len, opts->ascii_only); |
| 2325 | if (err != JSON_OK) { return err; } |
| 2326 | return jw_put(w, '"'); |
| 2327 | |
| 2328 | case JSON_ARRAY: { |
| 2329 | const JsonArr* a = &v->v.a; |
| 2330 | uint32_t i; |
| 2331 | err = jw_put(w, '['); |
| 2332 | if (err != JSON_OK) { return err; } |
| 2333 | for (i = 0U; i < a->len; i++) { |
| 2334 | if (opts->pretty) { |
| 2335 | jw_put(w, '\n'); |
| 2336 | jw_indent(w, depth + 1U, opts->indent); |
| 2337 | } |
| 2338 | err = json__write_value(w, a->items[i], depth + 1U, opts); |
| 2339 | if (err != JSON_OK) { return err; } |
| 2340 | if (i + 1U < a->len) { |
| 2341 | jw_put(w, ','); |
| 2342 | if (!opts->pretty) { /* no space */ } |
| 2343 | } |
| 2344 | } |
| 2345 | if (opts->pretty && a->len > 0U) { |
| 2346 | jw_put(w, '\n'); |
| 2347 | jw_indent(w, depth, opts->indent); |
| 2348 | } |
| 2349 | return jw_put(w, ']'); |
| 2350 | } |
| 2351 | |
| 2352 | case JSON_OBJECT: { |
| 2353 | const JsonObj* o = &v->v.o; |
| 2354 | uint32_t i; |
| 2355 | err = jw_put(w, '{'); |
| 2356 | if (err != JSON_OK) { return err; } |
| 2357 | for (i = 0U; i < o->len; i++) { |
| 2358 | uint32_t pair_idx = (opts->sort_keys && o->is_sorted && o->sorted_idx) |
| 2359 | ? o->sorted_idx[i] : i; |
| 2360 | const JsonPair* pair = &o->pairs[pair_idx]; |
| 2361 | if (opts->pretty) { |
| 2362 | jw_put(w, '\n'); |
| 2363 | jw_indent(w, depth + 1U, opts->indent); |
| 2364 | } |
| 2365 | jw_put(w, '"'); |
| 2366 | jw_string_content(w, pair->key.data ? pair->key.data : "", |
| 2367 | pair->key.len, opts->ascii_only); |
| 2368 | jw_put(w, '"'); |
| 2369 | jw_put(w, ':'); |
| 2370 | if (opts->pretty) { jw_put(w, ' '); } |
| 2371 | err = json__write_value(w, pair->val, depth + 1U, opts); |
| 2372 | if (err != JSON_OK) { return err; } |
| 2373 | if (i + 1U < o->len) { |
| 2374 | jw_put(w, ','); |
| 2375 | } |
| 2376 | } |
| 2377 | if (opts->pretty && o->len > 0U) { |
| 2378 | jw_put(w, '\n'); |
| 2379 | jw_indent(w, depth, opts->indent); |
| 2380 | } |
| 2381 | return jw_put(w, '}'); |
| 2382 | } |
| 2383 | |
| 2384 | default: |
| 2385 | return JSON_ERR_INVALID_JSON; |
| 2386 | } |
| 2387 | } |
| 2388 | |
| 2389 | static const JsonWriteOpts k_default_opts = { false, 2U, false, false, false }; |
| 2390 | |
| 2391 | JsonError json_write(const JsonValue* v, char* buf, size_t buf_size, |
| 2392 | size_t* written, const JsonWriteOpts* opts) |
| 2393 | { |
| 2394 | JsonWriter w; |
| 2395 | JsonError err; |
| 2396 | |
| 2397 | if (v == NULL) { return JSON_ERR_NULL_PARAM; } |
| 2398 | if (buf == NULL && buf_size > 0U) { return JSON_ERR_NULL_PARAM; } |
| 2399 | |
| 2400 | if (opts == NULL) { opts = &k_default_opts; } |
| 2401 | |
| 2402 | JSON_MEMSET(&w, 0, sizeof(w)); |
| 2403 | w.buf = buf; |
| 2404 | w.cap = buf_size; |
| 2405 | w.measure_only = (buf == NULL); |
| 2406 | |
| 2407 | err = json__write_value(&w, v, 0U, opts); |
| 2408 | |
| 2409 | if (err == JSON_OK && !w.measure_only) { |
| 2410 | /* NUL-terminate */ |
| 2411 | if (w.used < buf_size) { |
| 2412 | buf[w.used] = '\0'; |
| 2413 | } else { |
| 2414 | return JSON_ERR_BUFFER_TOO_SMALL; |
| 2415 | } |
| 2416 | if (opts->trailing_nl && w.used + 1U < buf_size) { |
| 2417 | buf[w.used++] = '\n'; |
| 2418 | buf[w.used] = '\0'; |
| 2419 | } |
| 2420 | } |
| 2421 | |
| 2422 | if (written != NULL) { *written = w.used; } |
| 2423 | return err; |
| 2424 | } |
| 2425 | |
| 2426 | JsonError json_measure(const JsonValue* v, size_t* size_out, |
| 2427 | const JsonWriteOpts* opts) |
| 2428 | { |
| 2429 | if (v == NULL || size_out == NULL) { return JSON_ERR_NULL_PARAM; } |
| 2430 | return json_write(v, NULL, 0U, size_out, opts); |
| 2431 | } |
| 2432 | |
| 2433 | JsonError json_write_arena(const JsonValue* v, JsonArena* arena, |
| 2434 | char** out, size_t* len_out, |
| 2435 | const JsonWriteOpts* opts) |
| 2436 | { |
| 2437 | size_t needed = 0U; |
| 2438 | char* buf; |
| 2439 | JsonError err; |
| 2440 | |
| 2441 | if (v == NULL || arena == NULL || out == NULL) { return JSON_ERR_NULL_PARAM; } |
| 2442 | |
| 2443 | /* Measure first */ |
| 2444 | err = json_measure(v, &needed, opts); |
| 2445 | if (err != JSON_OK) { return err; } |
| 2446 | |
| 2447 | needed += 2U; /* NUL + optional newline */ |
| 2448 | |
| 2449 | buf = (char*)json_arena_alloc(arena, needed); |
| 2450 | if (JSON_UNLIKELY(buf == NULL)) { return JSON_ERR_OOM; } |
| 2451 | |
| 2452 | err = json_write(v, buf, needed, len_out, opts); |
| 2453 | if (err != JSON_OK) { return err; } |
| 2454 | |
| 2455 | *out = buf; |
| 2456 | return JSON_OK; |
| 2457 | } |
| 2458 | |
| 2459 | /* ── §11.12 Value construction ─────────────────────────────────────────── */ |
| 2460 | |
| 2461 | JsonValue* json_make_null(JsonArena* arena) |
| 2462 | { |
| 2463 | JsonValue* v = json__new_node(arena); |
| 2464 | if (v != NULL) { v->type = JSON_NULL; } |
| 2465 | return v; |
| 2466 | } |
| 2467 | |
| 2468 | JsonValue* json_make_bool(JsonArena* arena, bool val) |
| 2469 | { |
| 2470 | JsonValue* v = json__new_node(arena); |
| 2471 | if (v != NULL) { v->type = JSON_BOOL; v->v.b = val; } |
| 2472 | return v; |
| 2473 | } |
| 2474 | |
| 2475 | JsonValue* json_make_int(JsonArena* arena, int64_t val) |
| 2476 | { |
| 2477 | JsonValue* v = json__new_node(arena); |
| 2478 | if (v != NULL) { v->type = JSON_INTEGER; v->v.i = val; } |
| 2479 | return v; |
| 2480 | } |
| 2481 | |
| 2482 | JsonValue* json_make_float(JsonArena* arena, double val) |
| 2483 | { |
| 2484 | JsonValue* v; |
| 2485 | /* Safety: never store NaN or Inf */ |
| 2486 | if (isnan(val) || isinf(val)) { return NULL; } |
| 2487 | v = json__new_node(arena); |
| 2488 | if (v != NULL) { v->type = JSON_FLOAT; v->v.f = val; } |
| 2489 | return v; |
| 2490 | } |
| 2491 | |
| 2492 | JsonValue* json_make_string(JsonArena* arena, const char* s, uint32_t len) |
| 2493 | { |
| 2494 | JsonValue* v; |
| 2495 | char* copy; |
| 2496 | |
| 2497 | if (arena == NULL) { return NULL; } |
| 2498 | v = json__new_node(arena); |
| 2499 | if (v == NULL) { return NULL; } |
| 2500 | |
| 2501 | if (len == 0U || s == NULL) { |
| 2502 | v->type = JSON_STRING; |
| 2503 | v->v.s.data= NULL; |
| 2504 | v->v.s.len = 0U; |
| 2505 | v->v.s.hash= json_fnv1a(NULL, 0U); |
| 2506 | return v; |
| 2507 | } |
| 2508 | |
| 2509 | copy = (char*)json_arena_alloc(arena, (size_t)len); |
| 2510 | if (JSON_UNLIKELY(copy == NULL)) { return NULL; } |
| 2511 | JSON_MEMCPY(copy, s, (size_t)len); |
| 2512 | |
| 2513 | v->type = JSON_STRING; |
| 2514 | v->v.s.data= copy; |
| 2515 | v->v.s.len = len; |
| 2516 | v->v.s.hash= json_fnv1a(copy, len); |
| 2517 | return v; |
| 2518 | } |
| 2519 | |
| 2520 | JsonValue* json_make_stringz(JsonArena* arena, const char* s) |
| 2521 | { |
| 2522 | if (s == NULL) { return json_make_string(arena, NULL, 0U); } |
| 2523 | return json_make_string(arena, s, (uint32_t)JSON_STRLEN(s)); |
| 2524 | } |
| 2525 | |
| 2526 | JsonValue* json_make_array(JsonArena* arena) |
| 2527 | { |
| 2528 | JsonValue* v = json__new_node(arena); |
| 2529 | if (v != NULL) { v->type = JSON_ARRAY; } |
| 2530 | return v; |
| 2531 | } |
| 2532 | |
| 2533 | JsonValue* json_make_object(JsonArena* arena) |
| 2534 | { |
| 2535 | JsonValue* v = json__new_node(arena); |
| 2536 | if (v != NULL) { v->type = JSON_OBJECT; } |
| 2537 | return v; |
| 2538 | } |
| 2539 | |
| 2540 | JsonError json_arr_push(JsonValue* arr, JsonArena* arena, JsonValue* item) |
| 2541 | { |
| 2542 | JsonArr* a; |
| 2543 | JsonError err; |
| 2544 | |
| 2545 | if (arr == NULL || arena == NULL || item == NULL) { return JSON_ERR_NULL_PARAM; } |
| 2546 | if (arr->type != JSON_ARRAY) { return JSON_ERR_TYPE_MISMATCH; } |
| 2547 | |
| 2548 | a = &arr->v.a; |
| 2549 | if (a->len >= a->cap) { |
| 2550 | err = json__arr_grow(a, arena); |
| 2551 | if (err != JSON_OK) { return err; } |
| 2552 | } |
| 2553 | a->items[a->len++] = item; |
| 2554 | return JSON_OK; |
| 2555 | } |
| 2556 | |
| 2557 | JsonError json_obj_set(JsonValue* obj, JsonArena* arena, |
| 2558 | const char* key, uint32_t klen, JsonValue* val) |
| 2559 | { |
| 2560 | JsonObj* o; |
| 2561 | JsonError err; |
| 2562 | char* key_copy; |
| 2563 | |
| 2564 | if (obj == NULL || arena == NULL || key == NULL || val == NULL) { |
| 2565 | return JSON_ERR_NULL_PARAM; |
| 2566 | } |
| 2567 | if (obj->type != JSON_OBJECT) { return JSON_ERR_TYPE_MISMATCH; } |
| 2568 | |
| 2569 | o = &obj->v.o; |
| 2570 | |
| 2571 | /* Check for existing key (update in place) */ |
| 2572 | { |
| 2573 | uint32_t hash = json_fnv1a(key, klen); |
| 2574 | uint32_t i; |
| 2575 | for (i = 0U; i < o->len; i++) { |
| 2576 | JsonStr* k = &o->pairs[i].key; |
| 2577 | if (k->hash == hash && k->len == klen && |
| 2578 | JSON_MEMCMP(k->data, key, klen) == 0) { |
| 2579 | o->pairs[i].val = val; |
| 2580 | return JSON_OK; |
| 2581 | } |
| 2582 | } |
| 2583 | } |
| 2584 | |
| 2585 | /* Insert new pair */ |
| 2586 | if (o->len >= o->cap) { |
| 2587 | err = json__obj_grow(o, arena); |
| 2588 | if (err != JSON_OK) { return err; } |
| 2589 | } |
| 2590 | |
| 2591 | key_copy = (char*)json_arena_alloc(arena, (size_t)klen); |
| 2592 | if (JSON_UNLIKELY(key_copy == NULL)) { return JSON_ERR_OOM; } |
| 2593 | JSON_MEMCPY(key_copy, key, klen); |
| 2594 | |
| 2595 | o->pairs[o->len].key.data = key_copy; |
| 2596 | o->pairs[o->len].key.len = klen; |
| 2597 | o->pairs[o->len].key.hash = json_fnv1a(key_copy, klen); |
| 2598 | o->pairs[o->len].val = val; |
| 2599 | o->len++; |
| 2600 | o->is_sorted = false; /* Invalidate sort */ |
| 2601 | return JSON_OK; |
| 2602 | } |
| 2603 | |
| 2604 | JsonError json_obj_setz(JsonValue* obj, JsonArena* arena, |
| 2605 | const char* key, JsonValue* val) |
| 2606 | { |
| 2607 | if (key == NULL) { return JSON_ERR_NULL_PARAM; } |
| 2608 | return json_obj_set(obj, arena, key, (uint32_t)JSON_STRLEN(key), val); |
| 2609 | } |
| 2610 | |
| 2611 | /* ── §11.13 Equality & Clone ───────────────────────────────────────────── */ |
| 2612 | |
| 2613 | bool json_equal(const JsonValue* a, const JsonValue* b) |
| 2614 | { |
| 2615 | uint32_t i; |
| 2616 | |
| 2617 | if (a == b) { return true; } |
| 2618 | if (!a || !b) { return false; } |
| 2619 | if (a->type != b->type) { |
| 2620 | /* Allow integer/float cross-comparison */ |
| 2621 | if (json_is_number(a) && json_is_number(b)) { |
| 2622 | double da, db; |
| 2623 | json_get_float(a, &da); |
| 2624 | json_get_float(b, &db); |
| 2625 | return da == db; |
| 2626 | } |
| 2627 | return false; |
| 2628 | } |
| 2629 | |
| 2630 | switch (a->type) { |
| 2631 | case JSON_NULL: return true; |
| 2632 | case JSON_BOOL: return a->v.b == b->v.b; |
| 2633 | case JSON_INTEGER: return a->v.i == b->v.i; |
| 2634 | case JSON_FLOAT: return a->v.f == b->v.f; |
| 2635 | case JSON_STRING: |
| 2636 | if (a->v.s.len != b->v.s.len) { return false; } |
| 2637 | if (a->v.s.hash != b->v.s.hash) { return false; } |
| 2638 | return JSON_MEMCMP(a->v.s.data, b->v.s.data, a->v.s.len) == 0; |
| 2639 | case JSON_ARRAY: |
| 2640 | if (a->v.a.len != b->v.a.len) { return false; } |
| 2641 | for (i = 0U; i < a->v.a.len; i++) { |
| 2642 | if (!json_equal(a->v.a.items[i], b->v.a.items[i])) { return false; } |
| 2643 | } |
| 2644 | return true; |
| 2645 | case JSON_OBJECT: |
| 2646 | if (a->v.o.len != b->v.o.len) { return false; } |
| 2647 | /* Check all keys in a exist in b with equal values */ |
| 2648 | for (i = 0U; i < a->v.o.len; i++) { |
| 2649 | const JsonStr* k = &a->v.o.pairs[i].key; |
| 2650 | JsonValue* bval = NULL; |
| 2651 | if (json_obj_get_n(b, k->data, k->len, &bval) != JSON_OK) { |
| 2652 | return false; |
| 2653 | } |
| 2654 | if (!json_equal(a->v.o.pairs[i].val, bval)) { return false; } |
| 2655 | } |
| 2656 | return true; |
| 2657 | default: return false; |
| 2658 | } |
| 2659 | } |
| 2660 | |
| 2661 | JsonValue* json_clone(JsonArena* dst, const JsonValue* src) |
| 2662 | { |
| 2663 | JsonValue* out; |
| 2664 | uint32_t i; |
| 2665 | |
| 2666 | if (dst == NULL || src == NULL) { return NULL; } |
| 2667 | |
| 2668 | out = json__new_node(dst); |
| 2669 | if (out == NULL) { return NULL; } |
| 2670 | |
| 2671 | out->type = src->type; |
| 2672 | |
| 2673 | switch (src->type) { |
| 2674 | case JSON_NULL: break; |
| 2675 | case JSON_BOOL: out->v.b = src->v.b; break; |
| 2676 | case JSON_INTEGER: out->v.i = src->v.i; break; |
| 2677 | case JSON_FLOAT: out->v.f = src->v.f; break; |
| 2678 | case JSON_STRING: { |
| 2679 | char* copy = NULL; |
| 2680 | if (src->v.s.len > 0U && src->v.s.data != NULL) { |
| 2681 | copy = (char*)json_arena_alloc(dst, src->v.s.len); |
| 2682 | if (copy == NULL) { return NULL; } |
| 2683 | JSON_MEMCPY(copy, src->v.s.data, src->v.s.len); |
| 2684 | } |
| 2685 | out->v.s.data = copy; |
| 2686 | out->v.s.len = src->v.s.len; |
| 2687 | out->v.s.hash = src->v.s.hash; |
| 2688 | break; |
| 2689 | } |
| 2690 | case JSON_ARRAY: { |
| 2691 | JsonArr* da = &out->v.a; |
| 2692 | const JsonArr* sa = &src->v.a; |
| 2693 | da->len = sa->len; |
| 2694 | da->cap = sa->len; |
| 2695 | da->items = NULL; |
| 2696 | if (sa->len > 0U) { |
| 2697 | da->items = (JsonValue**)json_arena_alloc(dst, |
| 2698 | sa->len * sizeof(JsonValue*)); |
| 2699 | if (da->items == NULL) { return NULL; } |
| 2700 | for (i = 0U; i < sa->len; i++) { |
| 2701 | da->items[i] = json_clone(dst, sa->items[i]); |
| 2702 | if (da->items[i] == NULL) { return NULL; } |
| 2703 | } |
| 2704 | } |
| 2705 | break; |
| 2706 | } |
| 2707 | case JSON_OBJECT: { |
| 2708 | JsonObj* do_ = &out->v.o; |
| 2709 | const JsonObj* so = &src->v.o; |
| 2710 | do_->len = so->len; |
| 2711 | do_->cap = so->len; |
| 2712 | do_->sorted_idx = NULL; |
| 2713 | do_->is_sorted = false; |
| 2714 | do_->pairs = NULL; |
| 2715 | if (so->len > 0U) { |
| 2716 | do_->pairs = (JsonPair*)json_arena_alloc(dst, |
| 2717 | so->len * sizeof(JsonPair)); |
| 2718 | if (do_->pairs == NULL) { return NULL; } |
| 2719 | for (i = 0U; i < so->len; i++) { |
| 2720 | const JsonStr* sk = &so->pairs[i].key; |
| 2721 | char* kc = NULL; |
| 2722 | if (sk->len > 0U && sk->data != NULL) { |
| 2723 | kc = (char*)json_arena_alloc(dst, sk->len); |
| 2724 | if (kc == NULL) { return NULL; } |
| 2725 | JSON_MEMCPY(kc, sk->data, sk->len); |
| 2726 | } |
| 2727 | do_->pairs[i].key.data = kc; |
| 2728 | do_->pairs[i].key.len = sk->len; |
| 2729 | do_->pairs[i].key.hash = sk->hash; |
| 2730 | do_->pairs[i].val = json_clone(dst, so->pairs[i].val); |
| 2731 | if (do_->pairs[i].val == NULL) { return NULL; } |
| 2732 | } |
| 2733 | } |
| 2734 | break; |
| 2735 | } |
| 2736 | default: break; |
| 2737 | } |
| 2738 | return out; |
| 2739 | } |
| 2740 | |
| 2741 | /* ── §11.14 Object sort (public trigger) ───────────────────────────────── */ |
| 2742 | |
| 2743 | /** |
| 2744 | * @brief Build sorted index for object to enable O(log n) lookup. |
| 2745 | * Call after parsing if many lookups are expected. |
| 2746 | * |
| 2747 | * @note Called automatically by json_obj_get if is_sorted == false. |
| 2748 | * To sort eagerly: json_obj_finalize(arena, root). |
| 2749 | */ |
| 2750 | JsonError json_obj_finalize(JsonArena* arena, JsonValue* obj) |
| 2751 | { |
| 2752 | if (obj == NULL || arena == NULL) { return JSON_ERR_NULL_PARAM; } |
| 2753 | if (obj->type != JSON_OBJECT) { return JSON_ERR_TYPE_MISMATCH; } |
| 2754 | return json__obj_sort(obj, arena); |
| 2755 | } |
| 2756 | |
| 2757 | /* ── §11.15 Error strings ──────────────────────────────────────────────── */ |
| 2758 | |
| 2759 | const char* json_error_str(JsonError err) |
| 2760 | { |
| 2761 | switch (err) { |
| 2762 | case JSON_OK: return "OK"; |
| 2763 | case JSON_ERR_NULL_PARAM: return "NULL parameter"; |
| 2764 | case JSON_ERR_OOM: return "Out of memory"; |
| 2765 | case JSON_ERR_INVALID_JSON: return "Invalid JSON"; |
| 2766 | case JSON_ERR_UNEXPECTED_EOF: return "Unexpected end of input"; |
| 2767 | case JSON_ERR_DEPTH_EXCEEDED: return "Nesting depth exceeded"; |
| 2768 | case JSON_ERR_INVALID_STRING: return "Invalid string / escape sequence"; |
| 2769 | case JSON_ERR_INVALID_NUMBER: return "Invalid number"; |
| 2770 | case JSON_ERR_INVALID_UTF8: return "Invalid UTF-8 sequence"; |
| 2771 | case JSON_ERR_OVERFLOW: return "Arithmetic or buffer overflow"; |
| 2772 | case JSON_ERR_NOT_FOUND: return "Key or index not found"; |
| 2773 | case JSON_ERR_TYPE_MISMATCH: return "Type mismatch"; |
| 2774 | case JSON_ERR_BUFFER_TOO_SMALL: return "Output buffer too small"; |
| 2775 | case JSON_ERR_NODE_LIMIT: return "Node count limit exceeded"; |
| 2776 | default: return "Unknown error"; |
| 2777 | } |
| 2778 | } |
| 2779 | |
| 2780 | #undef JSON__CHECK_V |
| 2781 | #undef JSON__CHECK_OUT |
| 2782 | |
| 2783 | #endif /* JSON_IMPLEMENTATION */ |
| 2784 | #endif /* JSON_PAL_H */ |
| 2785 | |
| 2786 | /* |
| 2787 | * ============================================================================ |
| 2788 | * USAGE EXAMPLE |
| 2789 | * ============================================================================ |
| 2790 | * |
| 2791 | * #define JSON_IMPLEMENTATION |
| 2792 | * #include "json_pal.h" |
| 2793 | * #include <stdio.h> |
| 2794 | * |
| 2795 | * int main(void) { |
| 2796 | * JsonArena* arena = json_arena_create(NULL, 64 * 1024); |
| 2797 | * if (!arena) return 1; |
| 2798 | * |
| 2799 | * const char* input = |
| 2800 | * "{\"name\":\"Alice\",\"scores\":[98,72,85],\"active\":true}"; |
| 2801 | * |
| 2802 | * JsonError err; |
| 2803 | * JsonValue* root = json_parse_cstr(arena, input, &err); |
| 2804 | * if (!root) { |
| 2805 | * fprintf(stderr, "Parse error: %s\n", json_error_str(err)); |
| 2806 | * json_arena_destroy(arena); |
| 2807 | * return 1; |
| 2808 | * } |
| 2809 | * |
| 2810 | * // Finalize for O(log n) key lookup |
| 2811 | * json_obj_finalize(arena, root); |
| 2812 | * |
| 2813 | * // Access fields |
| 2814 | * JsonValue* name_v; |
| 2815 | * json_obj_get(root, "name", &name_v); |
| 2816 | * const char* name; uint32_t nlen; |
| 2817 | * json_get_string(name_v, &name, &nlen); |
| 2818 | * printf("Name: %.*s\n", (int)nlen, name); |
| 2819 | * |
| 2820 | * // Path query |
| 2821 | * JsonValue* score1 = json_path(root, "scores[1]", NULL); |
| 2822 | * int64_t s; json_get_int(score1, &s); |
| 2823 | * printf("Score[1]: %lld\n", (long long)s); |
| 2824 | * |
| 2825 | * // Build and serialize |
| 2826 | * JsonValue* out_obj = json_make_object(arena); |
| 2827 | * json_obj_setz(out_obj, arena, "result", json_make_int(arena, s * 2)); |
| 2828 | * char buf[256]; |
| 2829 | * JsonWriteOpts opts = { .pretty = true, .indent = 4 }; |
| 2830 | * json_write(out_obj, buf, sizeof(buf), NULL, &opts); |
| 2831 | * printf("%s\n", buf); |
| 2832 | * |
| 2833 | * json_arena_destroy(arena); |
| 2834 | * return 0; |
| 2835 | * } |
| 2836 | * |
| 2837 | * ============================================================================ |
| 2838 | * FEATURE SUMMARY |
| 2839 | * ============================================================================ |
| 2840 | * |
| 2841 | * Parsing: |
| 2842 | * [x] Full RFC 8259 JSON compliance |
| 2843 | * [x] Iterative parser — no recursion, O(1) stack usage |
| 2844 | * [x] Depth-limited (JSON_MAX_DEPTH, default 64) |
| 2845 | * [x] Node-count-limited (JSON_MAX_NODES) |
| 2846 | * [x] Integer/float auto-detection at parse time |
| 2847 | * [x] \uXXXX escape decoding with surrogate pair support |
| 2848 | * [x] UTF-8 validation (RFC 3629) |
| 2849 | * [x] Leading-zero detection |
| 2850 | * [x] NaN/Inf rejection |
| 2851 | * [x] Control character rejection in strings |
| 2852 | * |
| 2853 | * Memory: |
| 2854 | * [x] Arena allocator — zero fragmentation, O(1) alloc |
| 2855 | * [x] Pluggable allocator (custom malloc/free/realloc) |
| 2856 | * [x] Arena reset (keep memory, reuse for next parse) |
| 2857 | * [x] All allocations zeroed |
| 2858 | * |
| 2859 | * Object lookup: |
| 2860 | * [x] FNV-1a hash for fast key comparison |
| 2861 | * [x] Introsort (O(n log n) worst-case) for key index |
| 2862 | * [x] Binary search O(log n) after finalization |
| 2863 | * [x] Linear scan O(n) before finalization |
| 2864 | * [x] Insertion-order iteration |
| 2865 | * |
| 2866 | * Serialization: |
| 2867 | * [x] Compact and pretty-print modes |
| 2868 | * [x] Configurable indent size |
| 2869 | * [x] Sorted key output |
| 2870 | * [x] ASCII-only escape mode |
| 2871 | * [x] Size measurement without allocation (json_measure) |
| 2872 | * [x] Arena-allocated output string |
| 2873 | * |
| 2874 | * Value construction: |
| 2875 | * [x] Null, bool, integer, float, string, array, object |
| 2876 | * [x] Array push, object set (insert or update) |
| 2877 | * [x] Deep equality (json_equal) |
| 2878 | * [x] Deep clone across arenas (json_clone) |
| 2879 | * |
| 2880 | * Platform: |
| 2881 | * [x] PAL layer: overridable memcpy/memset/memcmp/strlen/strtod/assert |
| 2882 | * [x] C99 compatible; C11 static asserts if available |
| 2883 | * [x] Endianness detection |
| 2884 | * [x] Freestanding (JSON_NO_STDLIB) support |
| 2885 | * [x] MSVC / GCC / Clang detection |
| 2886 | * [x] Compiler hints: always_inline, likely/unlikely, restrict |
| 2887 | * |
| 2888 | * Safety: |
| 2889 | * [x] All pointer parameters NULL-checked |
| 2890 | * [x] All array bounds checked |
| 2891 | * [x] All arithmetic overflow-checked |
| 2892 | * [x] No dynamic allocation (arena only) |
| 2893 | * [x] No VLAs, no recursion in parser |
| 2894 | * [x] Explicit error codes — no exceptions, no abort |
| 2895 | * ============================================================================ |
| 2896 | */ |
| 2897 |