Skip to content

Commit 2aefa41

Browse files
committed
Add a fast path for ASCII strings
This optimization is based on a few assumptions: - Most strings are ASCII only. - Most strings had their coderange scanned already. If the above is true, then by checking the string coderange, we can use a much more streamlined function to encode ASCII strings. Before: ``` == Encoding twitter.json (466906 bytes) ruby 3.4.0preview2 (2024-10-07 master 32c733f57b) +YJIT +PRISM [arm64-darwin23] Warming up -------------------------------------- json 140.000 i/100ms oj 230.000 i/100ms rapidjson 108.000 i/100ms Calculating ------------------------------------- json 1.464k (± 1.4%) i/s (682.83 μs/i) - 7.420k in 5.067573s oj 2.338k (± 1.5%) i/s (427.64 μs/i) - 11.730k in 5.017336s rapidjson 1.075k (± 1.6%) i/s (930.40 μs/i) - 5.400k in 5.025469s Comparison: json: 1464.5 i/s oj: 2338.4 i/s - 1.60x faster rapidjson: 1074.8 i/s - 1.36x slower ``` After: ``` == Encoding twitter.json (466906 bytes) ruby 3.4.0preview2 (2024-10-07 master 32c733f57b) +YJIT +PRISM [arm64-darwin23] Warming up -------------------------------------- json 189.000 i/100ms oj 228.000 i/100ms rapidjson 108.000 i/100ms Calculating ------------------------------------- json 1.903k (± 1.2%) i/s (525.55 μs/i) - 9.639k in 5.066521s oj 2.306k (± 1.3%) i/s (433.71 μs/i) - 11.628k in 5.044096s rapidjson 1.069k (± 2.4%) i/s (935.38 μs/i) - 5.400k in 5.053794s Comparison: json: 1902.8 i/s oj: 2305.7 i/s - 1.21x faster rapidjson: 1069.1 i/s - 1.78x slower ```
1 parent 92a2d54 commit 2aefa41

File tree

2 files changed

+185
-31
lines changed

2 files changed

+185
-31
lines changed

ext/json/ext/generator/generator.c

Lines changed: 185 additions & 30 deletions
Original file line numberDiff line numberDiff line change
@@ -25,14 +25,13 @@ static ID i_to_s, i_to_json, i_new, i_pack, i_unpack, i_create_id, i_extend;
2525
* Everything else (should be UTF-8) is just passed through and
2626
* appended to the result.
2727
*/
28-
static void convert_UTF8_to_JSON(FBuffer *out_buffer, VALUE in_string, bool out_ascii_only, bool out_script_safe)
28+
static void convert_UTF8_to_JSON(FBuffer *out_buffer, VALUE in_string, bool out_script_safe)
2929
{
3030
const char *hexdig = "0123456789abcdef";
3131
char scratch[12] = { '\\', 'u', 0, 0, 0, 0, '\\', 'u' };
3232

3333
const char *in_utf8_str = RSTRING_PTR(in_string);
3434
unsigned long in_utf8_len = RSTRING_LEN(in_string);
35-
bool in_is_ascii_only = rb_enc_str_asciionly_p(in_string);
3635

3736
unsigned long beg = 0, pos;
3837

@@ -42,46 +41,183 @@ static void convert_UTF8_to_JSON(FBuffer *out_buffer, VALUE in_string, bool out_
4241
bool should_escape;
4342

4443
/* UTF-8 decoding */
45-
if (in_is_ascii_only) {
46-
ch = in_utf8_str[pos];
47-
ch_len = 1;
48-
} else {
49-
short i;
50-
if ((in_utf8_str[pos] & 0x80) == 0x00) { ch_len = 1; ch = in_utf8_str[pos]; } /* leading 1 bit is 0b0 */
51-
else if ((in_utf8_str[pos] & 0xE0) == 0xC0) { ch_len = 2; ch = in_utf8_str[pos] & 0x1F; } /* leading 3 bits are 0b110 */
52-
else if ((in_utf8_str[pos] & 0xF0) == 0xE0) { ch_len = 3; ch = in_utf8_str[pos] & 0x0F; } /* leading 4 bits are 0b1110 */
53-
else if ((in_utf8_str[pos] & 0xF8) == 0xF0) { ch_len = 4; ch = in_utf8_str[pos] & 0x07; } /* leading 5 bits are 0b11110 */
54-
else
55-
rb_raise(rb_path2class("JSON::GeneratorError"),
56-
"source sequence is illegal/malformed utf-8");
57-
if ((pos+ch_len) > in_utf8_len)
58-
rb_raise(rb_path2class("JSON::GeneratorError"),
59-
"partial character in source, but hit end");
60-
for (i = 1; i < ch_len; i++) {
61-
if ((in_utf8_str[pos+i] & 0xC0) != 0x80) /* leading 2 bits should be 0b10 */
62-
rb_raise(rb_path2class("JSON::GeneratorError"),
63-
"source sequence is illegal/malformed utf-8");
64-
ch = (ch<<6) | (in_utf8_str[pos+i] & 0x3F);
44+
short i;
45+
if ((in_utf8_str[pos] & 0x80) == 0x00) { ch_len = 1; ch = in_utf8_str[pos]; } /* leading 1 bit is 0b0 */
46+
else if ((in_utf8_str[pos] & 0xE0) == 0xC0) { ch_len = 2; ch = in_utf8_str[pos] & 0x1F; } /* leading 3 bits are 0b110 */
47+
else if ((in_utf8_str[pos] & 0xF0) == 0xE0) { ch_len = 3; ch = in_utf8_str[pos] & 0x0F; } /* leading 4 bits are 0b1110 */
48+
else if ((in_utf8_str[pos] & 0xF8) == 0xF0) { ch_len = 4; ch = in_utf8_str[pos] & 0x07; } /* leading 5 bits are 0b11110 */
49+
else {
50+
rb_raise(rb_path2class("JSON::GeneratorError"), "source sequence is illegal/malformed utf-8");
51+
}
52+
53+
for (i = 1; i < ch_len; i++) {
54+
ch = (ch<<6) | (in_utf8_str[pos+i] & 0x3F);
55+
}
56+
57+
/* JSON policy */
58+
should_escape =
59+
(ch < 0x20) ||
60+
(ch == '"') ||
61+
(ch == '\\') ||
62+
(out_script_safe && (ch == '/')) ||
63+
(out_script_safe && (ch == 0x2028)) ||
64+
(out_script_safe && (ch == 0x2029));
65+
66+
/* JSON encoding */
67+
if (should_escape) {
68+
if (pos > beg) {
69+
fbuffer_append(out_buffer, &in_utf8_str[beg], pos - beg);
70+
}
71+
72+
beg = pos + ch_len;
73+
switch (ch) {
74+
case '"': fbuffer_append(out_buffer, "\\\"", 2); break;
75+
case '\\': fbuffer_append(out_buffer, "\\\\", 2); break;
76+
case '/': fbuffer_append(out_buffer, "\\/", 2); break;
77+
case '\b': fbuffer_append(out_buffer, "\\b", 2); break;
78+
case '\f': fbuffer_append(out_buffer, "\\f", 2); break;
79+
case '\n': fbuffer_append(out_buffer, "\\n", 2); break;
80+
case '\r': fbuffer_append(out_buffer, "\\r", 2); break;
81+
case '\t': fbuffer_append(out_buffer, "\\t", 2); break;
82+
default:
83+
if (ch <= 0xFFFF) {
84+
scratch[2] = hexdig[ch >> 12];
85+
scratch[3] = hexdig[(ch >> 8) & 0xf];
86+
scratch[4] = hexdig[(ch >> 4) & 0xf];
87+
scratch[5] = hexdig[ch & 0xf];
88+
fbuffer_append(out_buffer, scratch, 6);
89+
} else {
90+
uint16_t hi, lo;
91+
ch -= 0x10000;
92+
hi = 0xD800 + (uint16_t)(ch >> 10);
93+
lo = 0xDC00 + (uint16_t)(ch & 0x3FF);
94+
95+
scratch[2] = hexdig[hi >> 12];
96+
scratch[3] = hexdig[(hi >> 8) & 0xf];
97+
scratch[4] = hexdig[(hi >> 4) & 0xf];
98+
scratch[5] = hexdig[hi & 0xf];
99+
100+
scratch[8] = hexdig[lo >> 12];
101+
scratch[9] = hexdig[(lo >> 8) & 0xf];
102+
scratch[10] = hexdig[(lo >> 4) & 0xf];
103+
scratch[11] = hexdig[lo & 0xf];
104+
105+
fbuffer_append(out_buffer, scratch, 12);
106+
}
65107
}
66-
if (ch > 0x10FFFF)
67-
rb_raise(rb_path2class("JSON::GeneratorError"),
68-
"source sequence is illegal/malformed utf-8");
69108
}
70109

110+
pos += ch_len;
111+
}
112+
113+
if (beg < in_utf8_len) {
114+
fbuffer_append(out_buffer, &in_utf8_str[beg], in_utf8_len - beg);
115+
}
116+
117+
RB_GC_GUARD(in_string);
118+
}
119+
120+
static void convert_ASCII_to_JSON(FBuffer *out_buffer, VALUE str, bool out_script_safe)
121+
{
122+
const char *hexdig = "0123456789abcdef";
123+
char scratch[12] = { '\\', 'u', 0, 0, 0, 0, '\\', 'u' };
124+
125+
const char *ptr = RSTRING_PTR(str);
126+
unsigned long len = RSTRING_LEN(str);
127+
128+
unsigned long beg = 0, pos;
129+
130+
for (pos = 0; pos < len;) {
131+
unsigned char ch = ptr[pos];
132+
bool should_escape;
133+
71134
/* JSON policy */
72135
should_escape =
73136
(ch < 0x20) ||
74137
(ch == '"') ||
75138
(ch == '\\') ||
76-
(out_ascii_only && (ch > 0x7F)) ||
139+
(out_script_safe && (ch == '/'));
140+
141+
/* JSON encoding */
142+
if (should_escape) {
143+
if (pos > beg) {
144+
fbuffer_append(out_buffer, &ptr[beg], pos - beg);
145+
}
146+
147+
beg = pos + 1;
148+
switch (ch) {
149+
case '"': fbuffer_append(out_buffer, "\\\"", 2); break;
150+
case '\\': fbuffer_append(out_buffer, "\\\\", 2); break;
151+
case '/': fbuffer_append(out_buffer, "\\/", 2); break;
152+
case '\b': fbuffer_append(out_buffer, "\\b", 2); break;
153+
case '\f': fbuffer_append(out_buffer, "\\f", 2); break;
154+
case '\n': fbuffer_append(out_buffer, "\\n", 2); break;
155+
case '\r': fbuffer_append(out_buffer, "\\r", 2); break;
156+
case '\t': fbuffer_append(out_buffer, "\\t", 2); break;
157+
default:
158+
scratch[2] = hexdig[ch >> 12];
159+
scratch[3] = hexdig[(ch >> 8) & 0xf];
160+
scratch[4] = hexdig[(ch >> 4) & 0xf];
161+
scratch[5] = hexdig[ch & 0xf];
162+
fbuffer_append(out_buffer, scratch, 6);
163+
}
164+
}
165+
166+
pos++;
167+
}
168+
169+
if (beg < len) {
170+
fbuffer_append(out_buffer, &ptr[beg], len - beg);
171+
}
172+
173+
RB_GC_GUARD(str);
174+
}
175+
176+
static void convert_UTF8_to_ASCII_only_JSON(FBuffer *out_buffer, VALUE in_string, bool out_script_safe)
177+
{
178+
const char *hexdig = "0123456789abcdef";
179+
char scratch[12] = { '\\', 'u', 0, 0, 0, 0, '\\', 'u' };
180+
181+
const char *in_utf8_str = RSTRING_PTR(in_string);
182+
unsigned long in_utf8_len = RSTRING_LEN(in_string);
183+
184+
unsigned long beg = 0, pos;
185+
186+
for (pos = 0; pos < in_utf8_len;) {
187+
uint32_t ch;
188+
short ch_len;
189+
bool should_escape;
190+
191+
/* UTF-8 decoding */
192+
short i;
193+
if ((in_utf8_str[pos] & 0x80) == 0x00) { ch_len = 1; ch = in_utf8_str[pos]; } /* leading 1 bit is 0b0 */
194+
else if ((in_utf8_str[pos] & 0xE0) == 0xC0) { ch_len = 2; ch = in_utf8_str[pos] & 0x1F; } /* leading 3 bits are 0b110 */
195+
else if ((in_utf8_str[pos] & 0xF0) == 0xE0) { ch_len = 3; ch = in_utf8_str[pos] & 0x0F; } /* leading 4 bits are 0b1110 */
196+
else if ((in_utf8_str[pos] & 0xF8) == 0xF0) { ch_len = 4; ch = in_utf8_str[pos] & 0x07; } /* leading 5 bits are 0b11110 */
197+
else {
198+
rb_raise(rb_path2class("JSON::GeneratorError"), "source sequence is illegal/malformed utf-8");
199+
}
200+
201+
for (i = 1; i < ch_len; i++) {
202+
ch = (ch<<6) | (in_utf8_str[pos+i] & 0x3F);
203+
}
204+
205+
/* JSON policy */
206+
should_escape =
207+
(ch < 0x20) ||
208+
(ch == '"') ||
209+
(ch == '\\') ||
210+
(ch > 0x7F) ||
77211
(out_script_safe && (ch == '/')) ||
78212
(out_script_safe && (ch == 0x2028)) ||
79213
(out_script_safe && (ch == 0x2029));
80214

81215
/* JSON encoding */
82216
if (should_escape) {
83-
if (pos > beg)
217+
if (pos > beg) {
84218
fbuffer_append(out_buffer, &in_utf8_str[beg], pos - beg);
219+
}
220+
85221
beg = pos + ch_len;
86222
switch (ch) {
87223
case '"': fbuffer_append(out_buffer, "\\\"", 2); break;
@@ -122,8 +258,11 @@ static void convert_UTF8_to_JSON(FBuffer *out_buffer, VALUE in_string, bool out_
122258

123259
pos += ch_len;
124260
}
125-
if (beg < in_utf8_len)
261+
262+
if (beg < in_utf8_len) {
126263
fbuffer_append(out_buffer, &in_utf8_str[beg], in_utf8_len - beg);
264+
}
265+
127266
RB_GC_GUARD(in_string);
128267
}
129268

@@ -570,11 +709,27 @@ static int enc_utf8_compatible_p(int enc_idx)
570709

571710
static void generate_json_string(FBuffer *buffer, VALUE Vstate, JSON_Generator_State *state, VALUE obj)
572711
{
573-
fbuffer_append_char(buffer, '"');
574712
if (!enc_utf8_compatible_p(RB_ENCODING_GET(obj))) {
575713
obj = rb_str_export_to_enc(obj, rb_utf8_encoding());
576714
}
577-
convert_UTF8_to_JSON(buffer, obj, state->ascii_only, state->script_safe);
715+
716+
fbuffer_append_char(buffer, '"');
717+
718+
switch(rb_enc_str_coderange(obj)) {
719+
case ENC_CODERANGE_7BIT:
720+
convert_ASCII_to_JSON(buffer, obj, state->script_safe);
721+
break;
722+
case ENC_CODERANGE_VALID:
723+
if (RB_UNLIKELY(state->ascii_only)) {
724+
convert_UTF8_to_ASCII_only_JSON(buffer, obj, state->script_safe);
725+
} else {
726+
convert_UTF8_to_JSON(buffer, obj, state->script_safe);
727+
}
728+
break;
729+
default:
730+
rb_raise(rb_path2class("JSON::GeneratorError"), "source sequence is illegal/malformed utf-8");
731+
break;
732+
}
578733
fbuffer_append_char(buffer, '"');
579734
}
580735

ext/json/ext/generator/generator.h

Lines changed: 0 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -23,7 +23,6 @@ typedef unsigned char _Bool;
2323
#endif
2424
#endif
2525

26-
static void convert_UTF8_to_JSON(FBuffer *out_buffer, VALUE in_string, bool out_ascii_only, bool out_script_safe);
2726
static char *fstrndup(const char *ptr, unsigned long len);
2827

2928
/* ruby api and some helpers */

0 commit comments

Comments
 (0)