json 2.12.0 → 2.13.2
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- checksums.yaml +4 -4
- data/CHANGES.md +23 -0
- data/README.md +13 -0
- data/ext/json/ext/fbuffer/fbuffer.h +38 -4
- data/ext/json/ext/generator/extconf.rb +2 -25
- data/ext/json/ext/generator/generator.c +80 -107
- data/ext/json/ext/parser/extconf.rb +5 -1
- data/ext/json/ext/parser/parser.c +214 -66
- data/ext/json/ext/simd/conf.rb +24 -0
- data/ext/json/ext/simd/simd.h +188 -0
- data/ext/json/ext/vendor/fpconv.c +5 -5
- data/json.gemspec +2 -3
- data/lib/json/common.rb +23 -22
- data/lib/json/ext.rb +2 -2
- data/lib/json/version.rb +1 -1
- data/lib/json.rb +33 -0
- metadata +4 -3
- data/ext/json/ext/generator/simd.h +0 -112
checksums.yaml
CHANGED
@@ -1,7 +1,7 @@
|
|
1
1
|
---
|
2
2
|
SHA256:
|
3
|
-
metadata.gz:
|
4
|
-
data.tar.gz:
|
3
|
+
metadata.gz: 71499860706a6f27871853ec88fe26d5bd6a53f85fba2e9764b9ef0aadc170d9
|
4
|
+
data.tar.gz: c07cd26190c4f36864490465890c162e1b9ce0ae1f262069e51c94d7cf74b117
|
5
5
|
SHA512:
|
6
|
-
metadata.gz:
|
7
|
-
data.tar.gz:
|
6
|
+
metadata.gz: f1c626d30c67e99c56d9f411b7944b6263261676567c02f6d57ba7566087743b26835d3c6c4c1636a0dc76cc2b8c4e2739a1130f286cd069406c38c66138df97
|
7
|
+
data.tar.gz: eb4fd62079fc730962359e4128f7abd36ca04c17b580506c3e8e3ebdd0e8e0c54c3cf75d1ae899dc53e13697bfa0c11c6c04277d701cfc83fa9ee611785ac85e
|
data/CHANGES.md
CHANGED
@@ -2,6 +2,29 @@
|
|
2
2
|
|
3
3
|
### Unreleased
|
4
4
|
|
5
|
+
### 2025-07-28 (2.13.2)
|
6
|
+
|
7
|
+
* Improve duplicate key warning and errors to include the key name and point to the right caller.
|
8
|
+
|
9
|
+
### 2025-07-24 (2.13.1)
|
10
|
+
|
11
|
+
* Fix support for older compilers without `__builtin_cpu_supports`.
|
12
|
+
|
13
|
+
### 2025-07-17 (2.13.0)
|
14
|
+
|
15
|
+
* Add new `allow_duplicate_key` parsing options. By default a warning is now emitted when a duplicated key is encountered.
|
16
|
+
In `json 3.0` an error will be raised.
|
17
|
+
* Optimize parsing further using SIMD to scan strings.
|
18
|
+
|
19
|
+
### 2025-05-23 (2.12.2)
|
20
|
+
|
21
|
+
* Fix compiler optimization level.
|
22
|
+
|
23
|
+
### 2025-05-23 (2.12.1)
|
24
|
+
|
25
|
+
* Fix a potential crash in large negative floating point number generation.
|
26
|
+
* Fix for JSON.pretty_generate to use passed state object's generate instead of state class as the required parameters aren't available.
|
27
|
+
|
5
28
|
### 2025-05-12 (2.12.0)
|
6
29
|
|
7
30
|
* Improve floating point generation to not use scientific notation as much.
|
data/README.md
CHANGED
@@ -233,6 +233,19 @@ the `pp` library's `pp` methods.
|
|
233
233
|
|
234
234
|
## Development
|
235
235
|
|
236
|
+
### Prerequisites
|
237
|
+
|
238
|
+
1. Clone the repository
|
239
|
+
2. Install dependencies with `bundle install`
|
240
|
+
|
241
|
+
### Testing
|
242
|
+
|
243
|
+
The full test suite can be run with:
|
244
|
+
|
245
|
+
```bash
|
246
|
+
bundle exec rake test
|
247
|
+
```
|
248
|
+
|
236
249
|
### Release
|
237
250
|
|
238
251
|
Update the `lib/json/version.rb` file.
|
@@ -36,6 +36,12 @@ typedef unsigned char _Bool;
|
|
36
36
|
# define MAYBE_UNUSED(x) x
|
37
37
|
#endif
|
38
38
|
|
39
|
+
#ifdef RUBY_DEBUG
|
40
|
+
#ifndef JSON_DEBUG
|
41
|
+
#define JSON_DEBUG RUBY_DEBUG
|
42
|
+
#endif
|
43
|
+
#endif
|
44
|
+
|
39
45
|
enum fbuffer_type {
|
40
46
|
FBUFFER_HEAP_ALLOCATED = 0,
|
41
47
|
FBUFFER_STACK_ALLOCATED = 1,
|
@@ -46,6 +52,9 @@ typedef struct FBufferStruct {
|
|
46
52
|
unsigned long initial_length;
|
47
53
|
unsigned long len;
|
48
54
|
unsigned long capa;
|
55
|
+
#ifdef JSON_DEBUG
|
56
|
+
unsigned long requested;
|
57
|
+
#endif
|
49
58
|
char *ptr;
|
50
59
|
VALUE io;
|
51
60
|
} FBuffer;
|
@@ -74,6 +83,20 @@ static void fbuffer_stack_init(FBuffer *fb, unsigned long initial_length, char *
|
|
74
83
|
fb->ptr = stack_buffer;
|
75
84
|
fb->capa = stack_buffer_size;
|
76
85
|
}
|
86
|
+
#ifdef JSON_DEBUG
|
87
|
+
fb->requested = 0;
|
88
|
+
#endif
|
89
|
+
}
|
90
|
+
|
91
|
+
static inline void fbuffer_consumed(FBuffer *fb, unsigned long consumed)
|
92
|
+
{
|
93
|
+
#ifdef JSON_DEBUG
|
94
|
+
if (consumed > fb->requested) {
|
95
|
+
rb_bug("fbuffer: Out of bound write");
|
96
|
+
}
|
97
|
+
fb->requested = 0;
|
98
|
+
#endif
|
99
|
+
fb->len += consumed;
|
77
100
|
}
|
78
101
|
|
79
102
|
static void fbuffer_free(FBuffer *fb)
|
@@ -137,6 +160,10 @@ static void fbuffer_do_inc_capa(FBuffer *fb, unsigned long requested)
|
|
137
160
|
|
138
161
|
static inline void fbuffer_inc_capa(FBuffer *fb, unsigned long requested)
|
139
162
|
{
|
163
|
+
#ifdef JSON_DEBUG
|
164
|
+
fb->requested = requested;
|
165
|
+
#endif
|
166
|
+
|
140
167
|
if (RB_UNLIKELY(requested > fb->capa - fb->len)) {
|
141
168
|
fbuffer_do_inc_capa(fb, requested);
|
142
169
|
}
|
@@ -147,15 +174,22 @@ static void fbuffer_append(FBuffer *fb, const char *newstr, unsigned long len)
|
|
147
174
|
if (len > 0) {
|
148
175
|
fbuffer_inc_capa(fb, len);
|
149
176
|
MEMCPY(fb->ptr + fb->len, newstr, char, len);
|
150
|
-
fb
|
177
|
+
fbuffer_consumed(fb, len);
|
151
178
|
}
|
152
179
|
}
|
153
180
|
|
154
181
|
/* Appends a character into a buffer. The buffer needs to have sufficient capacity, via fbuffer_inc_capa(...). */
|
155
182
|
static inline void fbuffer_append_reserved_char(FBuffer *fb, char chr)
|
156
183
|
{
|
184
|
+
#ifdef JSON_DEBUG
|
185
|
+
if (fb->requested < 1) {
|
186
|
+
rb_bug("fbuffer: unreserved write");
|
187
|
+
}
|
188
|
+
fb->requested--;
|
189
|
+
#endif
|
190
|
+
|
157
191
|
fb->ptr[fb->len] = chr;
|
158
|
-
fb->len
|
192
|
+
fb->len++;
|
159
193
|
}
|
160
194
|
|
161
195
|
static void fbuffer_append_str(FBuffer *fb, VALUE str)
|
@@ -172,7 +206,7 @@ static inline void fbuffer_append_char(FBuffer *fb, char newchr)
|
|
172
206
|
{
|
173
207
|
fbuffer_inc_capa(fb, 1);
|
174
208
|
*(fb->ptr + fb->len) = newchr;
|
175
|
-
fb
|
209
|
+
fbuffer_consumed(fb, 1);
|
176
210
|
}
|
177
211
|
|
178
212
|
static inline char *fbuffer_cursor(FBuffer *fb)
|
@@ -182,7 +216,7 @@ static inline char *fbuffer_cursor(FBuffer *fb)
|
|
182
216
|
|
183
217
|
static inline void fbuffer_advance_to(FBuffer *fb, char *end)
|
184
218
|
{
|
185
|
-
fb
|
219
|
+
fbuffer_consumed(fb, (end - fb->ptr) - fb->len);
|
186
220
|
}
|
187
221
|
|
188
222
|
/*
|
@@ -6,33 +6,10 @@ if RUBY_ENGINE == 'truffleruby'
|
|
6
6
|
else
|
7
7
|
append_cflags("-std=c99")
|
8
8
|
$defs << "-DJSON_GENERATOR"
|
9
|
+
$defs << "-DJSON_DEBUG" if ENV["JSON_DEBUG"]
|
9
10
|
|
10
11
|
if enable_config('generator-use-simd', default=!ENV["JSON_DISABLE_SIMD"])
|
11
|
-
|
12
|
-
# Try to compile a small program using NEON instructions
|
13
|
-
if have_header('arm_neon.h')
|
14
|
-
have_type('uint8x16_t', headers=['arm_neon.h']) && try_compile(<<~'SRC')
|
15
|
-
#include <arm_neon.h>
|
16
|
-
int main() {
|
17
|
-
uint8x16_t test = vdupq_n_u8(32);
|
18
|
-
return 0;
|
19
|
-
}
|
20
|
-
SRC
|
21
|
-
$defs.push("-DJSON_ENABLE_SIMD")
|
22
|
-
end
|
23
|
-
end
|
24
|
-
|
25
|
-
if have_header('x86intrin.h') && have_type('__m128i', headers=['x86intrin.h']) && try_compile(<<~'SRC')
|
26
|
-
#include <x86intrin.h>
|
27
|
-
int main() {
|
28
|
-
__m128i test = _mm_set1_epi8(32);
|
29
|
-
return 0;
|
30
|
-
}
|
31
|
-
SRC
|
32
|
-
$defs.push("-DJSON_ENABLE_SIMD")
|
33
|
-
end
|
34
|
-
|
35
|
-
have_header('cpuid.h')
|
12
|
+
load __dir__ + "/../simd/conf.rb"
|
36
13
|
end
|
37
14
|
|
38
15
|
create_makefile 'json/ext/generator'
|
@@ -5,7 +5,7 @@
|
|
5
5
|
#include <math.h>
|
6
6
|
#include <ctype.h>
|
7
7
|
|
8
|
-
#include "simd.h"
|
8
|
+
#include "../simd/simd.h"
|
9
9
|
|
10
10
|
/* ruby api and some helpers */
|
11
11
|
|
@@ -304,28 +304,6 @@ static inline FORCE_INLINE unsigned char neon_next_match(search_state *search)
|
|
304
304
|
return 1;
|
305
305
|
}
|
306
306
|
|
307
|
-
// See: https://community.arm.com/arm-community-blogs/b/servers-and-cloud-computing-blog/posts/porting-x86-vector-bitmask-optimizations-to-arm-neon
|
308
|
-
static inline FORCE_INLINE uint64_t neon_match_mask(uint8x16_t matches)
|
309
|
-
{
|
310
|
-
const uint8x8_t res = vshrn_n_u16(vreinterpretq_u16_u8(matches), 4);
|
311
|
-
const uint64_t mask = vget_lane_u64(vreinterpret_u64_u8(res), 0);
|
312
|
-
return mask & 0x8888888888888888ull;
|
313
|
-
}
|
314
|
-
|
315
|
-
static inline FORCE_INLINE uint64_t neon_rules_update(const char *ptr)
|
316
|
-
{
|
317
|
-
uint8x16_t chunk = vld1q_u8((const unsigned char *)ptr);
|
318
|
-
|
319
|
-
// Trick: c < 32 || c == 34 can be factored as c ^ 2 < 33
|
320
|
-
// https://lemire.me/blog/2025/04/13/detect-control-characters-quotes-and-backslashes-efficiently-using-swar/
|
321
|
-
const uint8x16_t too_low_or_dbl_quote = vcltq_u8(veorq_u8(chunk, vdupq_n_u8(2)), vdupq_n_u8(33));
|
322
|
-
|
323
|
-
uint8x16_t has_backslash = vceqq_u8(chunk, vdupq_n_u8('\\'));
|
324
|
-
uint8x16_t needs_escape = vorrq_u8(too_low_or_dbl_quote, has_backslash);
|
325
|
-
|
326
|
-
return neon_match_mask(needs_escape);
|
327
|
-
}
|
328
|
-
|
329
307
|
static inline unsigned char search_escape_basic_neon(search_state *search)
|
330
308
|
{
|
331
309
|
if (RB_UNLIKELY(search->has_matches)) {
|
@@ -333,7 +311,7 @@ static inline unsigned char search_escape_basic_neon(search_state *search)
|
|
333
311
|
if (search->matches_mask > 0) {
|
334
312
|
return neon_next_match(search);
|
335
313
|
} else {
|
336
|
-
// neon_next_match will only advance search->ptr up to the last matching character.
|
314
|
+
// neon_next_match will only advance search->ptr up to the last matching character.
|
337
315
|
// Skip over any characters in the last chunk that occur after the last match.
|
338
316
|
search->has_matches = false;
|
339
317
|
search->ptr = search->chunk_end;
|
@@ -342,69 +320,63 @@ static inline unsigned char search_escape_basic_neon(search_state *search)
|
|
342
320
|
|
343
321
|
/*
|
344
322
|
* The code below implements an SIMD-based algorithm to determine if N bytes at a time
|
345
|
-
* need to be escaped.
|
346
|
-
*
|
323
|
+
* need to be escaped.
|
324
|
+
*
|
347
325
|
* Assume the ptr = "Te\sting!" (the double quotes are included in the string)
|
348
|
-
*
|
326
|
+
*
|
349
327
|
* The explanation will be limited to the first 8 bytes of the string for simplicity. However
|
350
328
|
* the vector insructions may work on larger vectors.
|
351
|
-
*
|
329
|
+
*
|
352
330
|
* First, we load three constants 'lower_bound', 'backslash' and 'dblquote" in vector registers.
|
353
|
-
*
|
354
|
-
* lower_bound: [20 20 20 20 20 20 20 20]
|
355
|
-
* backslash: [5C 5C 5C 5C 5C 5C 5C 5C]
|
356
|
-
* dblquote: [22 22 22 22 22 22 22 22]
|
357
|
-
*
|
358
|
-
* Next we load the first chunk of the ptr:
|
331
|
+
*
|
332
|
+
* lower_bound: [20 20 20 20 20 20 20 20]
|
333
|
+
* backslash: [5C 5C 5C 5C 5C 5C 5C 5C]
|
334
|
+
* dblquote: [22 22 22 22 22 22 22 22]
|
335
|
+
*
|
336
|
+
* Next we load the first chunk of the ptr:
|
359
337
|
* [22 54 65 5C 73 74 69 6E] (" T e \ s t i n)
|
360
|
-
*
|
338
|
+
*
|
361
339
|
* First we check if any byte in chunk is less than 32 (0x20). This returns the following vector
|
362
340
|
* as no bytes are less than 32 (0x20):
|
363
341
|
* [0 0 0 0 0 0 0 0]
|
364
|
-
*
|
342
|
+
*
|
365
343
|
* Next, we check if any byte in chunk is equal to a backslash:
|
366
344
|
* [0 0 0 FF 0 0 0 0]
|
367
|
-
*
|
345
|
+
*
|
368
346
|
* Finally we check if any byte in chunk is equal to a double quote:
|
369
|
-
* [FF 0 0 0 0 0 0 0]
|
370
|
-
*
|
347
|
+
* [FF 0 0 0 0 0 0 0]
|
348
|
+
*
|
371
349
|
* Now we have three vectors where each byte indicates if the corresponding byte in chunk
|
372
350
|
* needs to be escaped. We combine these vectors with a series of logical OR instructions.
|
373
351
|
* This is the needs_escape vector and it is equal to:
|
374
|
-
* [FF 0 0 FF 0 0 0 0]
|
375
|
-
*
|
352
|
+
* [FF 0 0 FF 0 0 0 0]
|
353
|
+
*
|
376
354
|
* Next we compute the bitwise AND between each byte and 0x1 and compute the horizontal sum of
|
377
355
|
* the values in the vector. This computes how many bytes need to be escaped within this chunk.
|
378
|
-
*
|
356
|
+
*
|
379
357
|
* Finally we compute a mask that indicates which bytes need to be escaped. If the mask is 0 then,
|
380
358
|
* no bytes need to be escaped and we can continue to the next chunk. If the mask is not 0 then we
|
381
359
|
* have at least one byte that needs to be escaped.
|
382
360
|
*/
|
383
|
-
while (search->ptr + sizeof(uint8x16_t) <= search->end) {
|
384
|
-
uint64_t mask = neon_rules_update(search->ptr);
|
385
361
|
|
386
|
-
|
387
|
-
search->ptr += sizeof(uint8x16_t);
|
388
|
-
continue;
|
389
|
-
}
|
390
|
-
search->matches_mask = mask;
|
362
|
+
if (string_scan_simd_neon(&search->ptr, search->end, &search->matches_mask)) {
|
391
363
|
search->has_matches = true;
|
392
364
|
search->chunk_base = search->ptr;
|
393
365
|
search->chunk_end = search->ptr + sizeof(uint8x16_t);
|
394
366
|
return neon_next_match(search);
|
395
367
|
}
|
396
368
|
|
397
|
-
// There are fewer than 16 bytes left.
|
369
|
+
// There are fewer than 16 bytes left.
|
398
370
|
unsigned long remaining = (search->end - search->ptr);
|
399
371
|
if (remaining >= SIMD_MINIMUM_THRESHOLD) {
|
400
372
|
char *s = copy_remaining_bytes(search, sizeof(uint8x16_t), remaining);
|
401
373
|
|
402
|
-
uint64_t mask =
|
374
|
+
uint64_t mask = compute_chunk_mask_neon(s);
|
403
375
|
|
404
376
|
if (!mask) {
|
405
|
-
// Nothing to escape, ensure search_flush doesn't do anything by setting
|
377
|
+
// Nothing to escape, ensure search_flush doesn't do anything by setting
|
406
378
|
// search->cursor to search->ptr.
|
407
|
-
search->buffer
|
379
|
+
fbuffer_consumed(search->buffer, remaining);
|
408
380
|
search->ptr = search->end;
|
409
381
|
search->cursor = search->end;
|
410
382
|
return 0;
|
@@ -428,11 +400,6 @@ static inline unsigned char search_escape_basic_neon(search_state *search)
|
|
428
400
|
|
429
401
|
#ifdef HAVE_SIMD_SSE2
|
430
402
|
|
431
|
-
#define _mm_cmpge_epu8(a, b) _mm_cmpeq_epi8(_mm_max_epu8(a, b), a)
|
432
|
-
#define _mm_cmple_epu8(a, b) _mm_cmpge_epu8(b, a)
|
433
|
-
#define _mm_cmpgt_epu8(a, b) _mm_xor_si128(_mm_cmple_epu8(a, b), _mm_set1_epi8(-1))
|
434
|
-
#define _mm_cmplt_epu8(a, b) _mm_cmpgt_epu8(b, a)
|
435
|
-
|
436
403
|
static inline FORCE_INLINE unsigned char sse2_next_match(search_state *search)
|
437
404
|
{
|
438
405
|
int mask = search->matches_mask;
|
@@ -457,18 +424,6 @@ static inline FORCE_INLINE unsigned char sse2_next_match(search_state *search)
|
|
457
424
|
#define TARGET_SSE2
|
458
425
|
#endif
|
459
426
|
|
460
|
-
static inline TARGET_SSE2 FORCE_INLINE int sse2_update(const char *ptr)
|
461
|
-
{
|
462
|
-
__m128i chunk = _mm_loadu_si128((__m128i const*)ptr);
|
463
|
-
|
464
|
-
// Trick: c < 32 || c == 34 can be factored as c ^ 2 < 33
|
465
|
-
// https://lemire.me/blog/2025/04/13/detect-control-characters-quotes-and-backslashes-efficiently-using-swar/
|
466
|
-
__m128i too_low_or_dbl_quote = _mm_cmplt_epu8(_mm_xor_si128(chunk, _mm_set1_epi8(2)), _mm_set1_epi8(33));
|
467
|
-
__m128i has_backslash = _mm_cmpeq_epi8(chunk, _mm_set1_epi8('\\'));
|
468
|
-
__m128i needs_escape = _mm_or_si128(too_low_or_dbl_quote, has_backslash);
|
469
|
-
return _mm_movemask_epi8(needs_escape);
|
470
|
-
}
|
471
|
-
|
472
427
|
static inline TARGET_SSE2 FORCE_INLINE unsigned char search_escape_basic_sse2(search_state *search)
|
473
428
|
{
|
474
429
|
if (RB_UNLIKELY(search->has_matches)) {
|
@@ -476,7 +431,7 @@ static inline TARGET_SSE2 FORCE_INLINE unsigned char search_escape_basic_sse2(se
|
|
476
431
|
if (search->matches_mask > 0) {
|
477
432
|
return sse2_next_match(search);
|
478
433
|
} else {
|
479
|
-
// sse2_next_match will only advance search->ptr up to the last matching character.
|
434
|
+
// sse2_next_match will only advance search->ptr up to the last matching character.
|
480
435
|
// Skip over any characters in the last chunk that occur after the last match.
|
481
436
|
search->has_matches = false;
|
482
437
|
if (RB_UNLIKELY(search->chunk_base + sizeof(__m128i) >= search->end)) {
|
@@ -487,31 +442,24 @@ static inline TARGET_SSE2 FORCE_INLINE unsigned char search_escape_basic_sse2(se
|
|
487
442
|
}
|
488
443
|
}
|
489
444
|
|
490
|
-
|
491
|
-
int needs_escape_mask = sse2_update(search->ptr);
|
492
|
-
|
493
|
-
if (needs_escape_mask == 0) {
|
494
|
-
search->ptr += sizeof(__m128i);
|
495
|
-
continue;
|
496
|
-
}
|
497
|
-
|
445
|
+
if (string_scan_simd_sse2(&search->ptr, search->end, &search->matches_mask)) {
|
498
446
|
search->has_matches = true;
|
499
|
-
search->matches_mask = needs_escape_mask;
|
500
447
|
search->chunk_base = search->ptr;
|
448
|
+
search->chunk_end = search->ptr + sizeof(__m128i);
|
501
449
|
return sse2_next_match(search);
|
502
450
|
}
|
503
451
|
|
504
|
-
// There are fewer than 16 bytes left.
|
452
|
+
// There are fewer than 16 bytes left.
|
505
453
|
unsigned long remaining = (search->end - search->ptr);
|
506
454
|
if (remaining >= SIMD_MINIMUM_THRESHOLD) {
|
507
455
|
char *s = copy_remaining_bytes(search, sizeof(__m128i), remaining);
|
508
456
|
|
509
|
-
int needs_escape_mask =
|
457
|
+
int needs_escape_mask = compute_chunk_mask_sse2(s);
|
510
458
|
|
511
459
|
if (needs_escape_mask == 0) {
|
512
|
-
// Nothing to escape, ensure search_flush doesn't do anything by setting
|
460
|
+
// Nothing to escape, ensure search_flush doesn't do anything by setting
|
513
461
|
// search->cursor to search->ptr.
|
514
|
-
search->buffer
|
462
|
+
fbuffer_consumed(search->buffer, remaining);
|
515
463
|
search->ptr = search->end;
|
516
464
|
search->cursor = search->end;
|
517
465
|
return 0;
|
@@ -638,7 +586,8 @@ static inline unsigned char search_ascii_only_escape(search_state *search, const
|
|
638
586
|
return 0;
|
639
587
|
}
|
640
588
|
|
641
|
-
static inline void full_escape_UTF8_char(search_state *search, unsigned char ch_len)
|
589
|
+
static inline void full_escape_UTF8_char(search_state *search, unsigned char ch_len)
|
590
|
+
{
|
642
591
|
const unsigned char ch = (unsigned char)*search->ptr;
|
643
592
|
switch (ch_len) {
|
644
593
|
case 1: {
|
@@ -668,7 +617,7 @@ static inline void full_escape_UTF8_char(search_state *search, unsigned char ch_
|
|
668
617
|
|
669
618
|
uint32_t wchar = 0;
|
670
619
|
|
671
|
-
switch(ch_len) {
|
620
|
+
switch (ch_len) {
|
672
621
|
case 2:
|
673
622
|
wchar = ch & 0x1F;
|
674
623
|
break;
|
@@ -828,7 +777,8 @@ static VALUE mHash_to_json(int argc, VALUE *argv, VALUE self)
|
|
828
777
|
* _state_ is a JSON::State object, that can also be used to configure the
|
829
778
|
* produced JSON string output further.
|
830
779
|
*/
|
831
|
-
static VALUE mArray_to_json(int argc, VALUE *argv, VALUE self)
|
780
|
+
static VALUE mArray_to_json(int argc, VALUE *argv, VALUE self)
|
781
|
+
{
|
832
782
|
rb_check_arity(argc, 0, 1);
|
833
783
|
VALUE Vstate = cState_from_state_s(cState, argc == 1 ? argv[0] : Qnil);
|
834
784
|
return cState_partial_generate(Vstate, self, generate_json_array, Qfalse);
|
@@ -890,7 +840,8 @@ static VALUE mFloat_to_json(int argc, VALUE *argv, VALUE self)
|
|
890
840
|
*
|
891
841
|
* Extends _modul_ with the String::Extend module.
|
892
842
|
*/
|
893
|
-
static VALUE mString_included_s(VALUE self, VALUE modul)
|
843
|
+
static VALUE mString_included_s(VALUE self, VALUE modul)
|
844
|
+
{
|
894
845
|
VALUE result = rb_funcall(modul, i_extend, 1, mString_Extend);
|
895
846
|
rb_call_super(1, &modul);
|
896
847
|
return result;
|
@@ -1135,7 +1086,7 @@ json_object_i(VALUE key, VALUE val, VALUE _arg)
|
|
1135
1086
|
}
|
1136
1087
|
|
1137
1088
|
VALUE key_to_s;
|
1138
|
-
switch(rb_type(key)) {
|
1089
|
+
switch (rb_type(key)) {
|
1139
1090
|
case T_STRING:
|
1140
1091
|
if (RB_LIKELY(RBASIC_CLASS(key) == rb_cString)) {
|
1141
1092
|
key_to_s = key;
|
@@ -1219,7 +1170,7 @@ static void generate_json_array(FBuffer *buffer, struct generate_json_data *data
|
|
1219
1170
|
|
1220
1171
|
fbuffer_append_char(buffer, '[');
|
1221
1172
|
if (RB_UNLIKELY(data->state->array_nl)) fbuffer_append_str(buffer, data->state->array_nl);
|
1222
|
-
for(i = 0; i < RARRAY_LEN(obj); i++) {
|
1173
|
+
for (i = 0; i < RARRAY_LEN(obj); i++) {
|
1223
1174
|
if (i > 0) {
|
1224
1175
|
fbuffer_append_char(buffer, ',');
|
1225
1176
|
if (RB_UNLIKELY(data->state->array_nl)) fbuffer_append_str(buffer, data->state->array_nl);
|
@@ -1304,7 +1255,7 @@ static void generate_json_string(FBuffer *buffer, struct generate_json_data *dat
|
|
1304
1255
|
search.chunk_base = NULL;
|
1305
1256
|
#endif /* HAVE_SIMD */
|
1306
1257
|
|
1307
|
-
switch(rb_enc_str_coderange(obj)) {
|
1258
|
+
switch (rb_enc_str_coderange(obj)) {
|
1308
1259
|
case ENC_CODERANGE_7BIT:
|
1309
1260
|
case ENC_CODERANGE_VALID:
|
1310
1261
|
if (RB_UNLIKELY(data->state->ascii_only)) {
|
@@ -1406,17 +1357,16 @@ static void generate_json_float(FBuffer *buffer, struct generate_json_data *data
|
|
1406
1357
|
}
|
1407
1358
|
|
1408
1359
|
/* This implementation writes directly into the buffer. We reserve
|
1409
|
-
* the
|
1410
|
-
* 2 more characters for the potential ".0" suffix.
|
1360
|
+
* the 28 characters that fpconv_dtoa states as its maximum.
|
1411
1361
|
*/
|
1412
|
-
fbuffer_inc_capa(buffer,
|
1362
|
+
fbuffer_inc_capa(buffer, 28);
|
1413
1363
|
char* d = buffer->ptr + buffer->len;
|
1414
1364
|
int len = fpconv_dtoa(value, d);
|
1415
1365
|
|
1416
1366
|
/* fpconv_dtoa converts a float to its shortest string representation,
|
1417
1367
|
* but it adds a ".0" if this is a plain integer.
|
1418
1368
|
*/
|
1419
|
-
buffer
|
1369
|
+
fbuffer_consumed(buffer, len);
|
1420
1370
|
}
|
1421
1371
|
|
1422
1372
|
static void generate_json_fragment(FBuffer *buffer, struct generate_json_data *data, VALUE obj)
|
@@ -1957,15 +1907,30 @@ static VALUE cState_buffer_initial_length_set(VALUE self, VALUE buffer_initial_l
|
|
1957
1907
|
return Qnil;
|
1958
1908
|
}
|
1959
1909
|
|
1910
|
+
struct configure_state_data {
|
1911
|
+
JSON_Generator_State *state;
|
1912
|
+
VALUE vstate; // Ruby object that owns the state, or Qfalse if stack-allocated
|
1913
|
+
};
|
1914
|
+
|
1915
|
+
static inline void state_write_value(struct configure_state_data *data, VALUE *field, VALUE value)
|
1916
|
+
{
|
1917
|
+
if (RTEST(data->vstate)) {
|
1918
|
+
RB_OBJ_WRITE(data->vstate, field, value);
|
1919
|
+
} else {
|
1920
|
+
*field = value;
|
1921
|
+
}
|
1922
|
+
}
|
1923
|
+
|
1960
1924
|
static int configure_state_i(VALUE key, VALUE val, VALUE _arg)
|
1961
1925
|
{
|
1962
|
-
|
1926
|
+
struct configure_state_data *data = (struct configure_state_data *)_arg;
|
1927
|
+
JSON_Generator_State *state = data->state;
|
1963
1928
|
|
1964
|
-
if (key == sym_indent) { state->indent
|
1965
|
-
else if (key == sym_space) { state->space
|
1966
|
-
else if (key == sym_space_before) { state->space_before
|
1967
|
-
else if (key == sym_object_nl) { state->object_nl
|
1968
|
-
else if (key == sym_array_nl) { state->array_nl
|
1929
|
+
if (key == sym_indent) { state_write_value(data, &state->indent, string_config(val)); }
|
1930
|
+
else if (key == sym_space) { state_write_value(data, &state->space, string_config(val)); }
|
1931
|
+
else if (key == sym_space_before) { state_write_value(data, &state->space_before, string_config(val)); }
|
1932
|
+
else if (key == sym_object_nl) { state_write_value(data, &state->object_nl, string_config(val)); }
|
1933
|
+
else if (key == sym_array_nl) { state_write_value(data, &state->array_nl, string_config(val)); }
|
1969
1934
|
else if (key == sym_max_nesting) { state->max_nesting = long_config(val); }
|
1970
1935
|
else if (key == sym_allow_nan) { state->allow_nan = RTEST(val); }
|
1971
1936
|
else if (key == sym_ascii_only) { state->ascii_only = RTEST(val); }
|
@@ -1974,11 +1939,14 @@ static int configure_state_i(VALUE key, VALUE val, VALUE _arg)
|
|
1974
1939
|
else if (key == sym_script_safe) { state->script_safe = RTEST(val); }
|
1975
1940
|
else if (key == sym_escape_slash) { state->script_safe = RTEST(val); }
|
1976
1941
|
else if (key == sym_strict) { state->strict = RTEST(val); }
|
1977
|
-
else if (key == sym_as_json) {
|
1942
|
+
else if (key == sym_as_json) {
|
1943
|
+
VALUE proc = RTEST(val) ? rb_convert_type(val, T_DATA, "Proc", "to_proc") : Qfalse;
|
1944
|
+
state_write_value(data, &state->as_json, proc);
|
1945
|
+
}
|
1978
1946
|
return ST_CONTINUE;
|
1979
1947
|
}
|
1980
1948
|
|
1981
|
-
static void configure_state(JSON_Generator_State *state, VALUE config)
|
1949
|
+
static void configure_state(JSON_Generator_State *state, VALUE vstate, VALUE config)
|
1982
1950
|
{
|
1983
1951
|
if (!RTEST(config)) return;
|
1984
1952
|
|
@@ -1986,15 +1954,20 @@ static void configure_state(JSON_Generator_State *state, VALUE config)
|
|
1986
1954
|
|
1987
1955
|
if (!RHASH_SIZE(config)) return;
|
1988
1956
|
|
1957
|
+
struct configure_state_data data = {
|
1958
|
+
.state = state,
|
1959
|
+
.vstate = vstate
|
1960
|
+
};
|
1961
|
+
|
1989
1962
|
// We assume in most cases few keys are set so it's faster to go over
|
1990
1963
|
// the provided keys than to check all possible keys.
|
1991
|
-
rb_hash_foreach(config, configure_state_i, (VALUE)
|
1964
|
+
rb_hash_foreach(config, configure_state_i, (VALUE)&data);
|
1992
1965
|
}
|
1993
1966
|
|
1994
1967
|
static VALUE cState_configure(VALUE self, VALUE opts)
|
1995
1968
|
{
|
1996
1969
|
GET_STATE(self);
|
1997
|
-
configure_state(state, opts);
|
1970
|
+
configure_state(state, self, opts);
|
1998
1971
|
return self;
|
1999
1972
|
}
|
2000
1973
|
|
@@ -2002,7 +1975,7 @@ static VALUE cState_m_generate(VALUE klass, VALUE obj, VALUE opts, VALUE io)
|
|
2002
1975
|
{
|
2003
1976
|
JSON_Generator_State state = {0};
|
2004
1977
|
state_init(&state);
|
2005
|
-
configure_state(&state, opts);
|
1978
|
+
configure_state(&state, Qfalse, opts);
|
2006
1979
|
|
2007
1980
|
char stack_buffer[FBUFFER_STACK_SIZE];
|
2008
1981
|
FBuffer buffer = {
|
@@ -2169,7 +2142,7 @@ void Init_generator(void)
|
|
2169
2142
|
rb_require("json/ext/generator/state");
|
2170
2143
|
|
2171
2144
|
|
2172
|
-
switch(find_simd_implementation()) {
|
2145
|
+
switch (find_simd_implementation()) {
|
2173
2146
|
#ifdef HAVE_SIMD
|
2174
2147
|
#ifdef HAVE_SIMD_NEON
|
2175
2148
|
case SIMD_NEON:
|
@@ -1,11 +1,15 @@
|
|
1
1
|
# frozen_string_literal: true
|
2
2
|
require 'mkmf'
|
3
3
|
|
4
|
-
have_func("rb_enc_interned_str", "ruby.h") # RUBY_VERSION >= 3.0
|
4
|
+
have_func("rb_enc_interned_str", "ruby/encoding.h") # RUBY_VERSION >= 3.0
|
5
5
|
have_func("rb_hash_new_capa", "ruby.h") # RUBY_VERSION >= 3.2
|
6
6
|
have_func("rb_hash_bulk_insert", "ruby.h") # Missing on TruffleRuby
|
7
7
|
have_func("strnlen", "string.h") # Missing on Solaris 10
|
8
8
|
|
9
9
|
append_cflags("-std=c99")
|
10
10
|
|
11
|
+
if enable_config('parser-use-simd', default=!ENV["JSON_DISABLE_SIMD"])
|
12
|
+
load __dir__ + "/../simd/conf.rb"
|
13
|
+
end
|
14
|
+
|
11
15
|
create_makefile 'json/ext/parser'
|