Skip to content

Commit

Permalink
convert_UTF8_to_ASCII_only_JSON: apply the same optimization pass
Browse files Browse the repository at this point in the history
  • Loading branch information
@byroot
byroot committed Oct 19, 2024
1 parent 1a33853 commit 42edaf7
Showing 1 changed file with 71 additions and 60 deletions.
131 changes: 71 additions & 60 deletions ext/json/ext/generator/generator.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -199,73 +199,80 @@ static void convert_ASCII_to_JSON(FBuffer *out_buffer, VALUE str, const char esc
RB_GC_GUARD(str);
}

static void convert_UTF8_to_ASCII_only_JSON(FBuffer *out_buffer, VALUE in_string, bool out_script_safe)
static void convert_UTF8_to_ASCII_only_JSON(FBuffer *out_buffer, VALUE str, const char escape_table[256], bool out_script_safe)
{
const char *hexdig = "0123456789abcdef";
char scratch[12] = { '\\', 'u', 0, 0, 0, 0, '\\', 'u' };

const char *in_utf8_str = RSTRING_PTR(in_string);
unsigned long in_utf8_len = RSTRING_LEN(in_string);
const char *ptr = RSTRING_PTR(str);
unsigned long len = RSTRING_LEN(str);

unsigned long beg = 0, pos;
unsigned long beg = 0, pos = 0;

for (pos = 0; pos < in_utf8_len;) {
uint32_t ch;
short ch_len;
bool should_escape;

/* UTF-8 decoding */
short i;
if ((in_utf8_str[pos] & 0x80) == 0x00) { ch_len = 1; ch = in_utf8_str[pos]; } /* leading 1 bit is 0b0 */
else if ((in_utf8_str[pos] & 0xE0) == 0xC0) { ch_len = 2; ch = in_utf8_str[pos] & 0x1F; } /* leading 3 bits are 0b110 */
else if ((in_utf8_str[pos] & 0xF0) == 0xE0) { ch_len = 3; ch = in_utf8_str[pos] & 0x0F; } /* leading 4 bits are 0b1110 */
else if ((in_utf8_str[pos] & 0xF8) == 0xF0) { ch_len = 4; ch = in_utf8_str[pos] & 0x07; } /* leading 5 bits are 0b11110 */
else {
rb_raise(rb_path2class("JSON::GeneratorError"), "source sequence is illegal/malformed utf-8");
}
#define FLUSH_POS(bytes) if (pos > beg) { fbuffer_append(out_buffer, &ptr[beg], pos - beg); } pos += bytes; beg = pos;

for (i = 1; i < ch_len; i++) {
ch = (ch<<6) | (in_utf8_str[pos+i] & 0x3F);
}
while (pos < len) {
unsigned char ch = ptr[pos];
unsigned char ch_len = escape_table[ch];

/* JSON policy */
should_escape =
(ch < 0x20) ||
(ch == '"') ||
(ch == '\\') ||
(ch > 0x7F) ||
(out_script_safe && (ch == '/')) ||
(out_script_safe && (ch == 0x2028)) ||
(out_script_safe && (ch == 0x2029));
if (RB_UNLIKELY(ch_len)) {
switch (ch_len) {
case 0:
pos++;
break;
case 1: {
FLUSH_POS(1);
switch (ch) {
case '"': fbuffer_append(out_buffer, "\\\"", 2); break;
case '\\': fbuffer_append(out_buffer, "\\\\", 2); break;
case '/': fbuffer_append(out_buffer, "\\/", 2); break;
case '\b': fbuffer_append(out_buffer, "\\b", 2); break;
case '\f': fbuffer_append(out_buffer, "\\f", 2); break;
case '\n': fbuffer_append(out_buffer, "\\n", 2); break;
case '\r': fbuffer_append(out_buffer, "\\r", 2); break;
case '\t': fbuffer_append(out_buffer, "\\t", 2); break;
default: {
scratch[2] = hexdig[ch >> 12];
scratch[3] = hexdig[(ch >> 8) & 0xf];
scratch[4] = hexdig[(ch >> 4) & 0xf];
scratch[5] = hexdig[ch & 0xf];
fbuffer_append(out_buffer, scratch, 6);
break;
}
}
break;
}
default: {
uint32_t wchar = 0;
switch(ch_len) {
case 2:
wchar = ptr[pos] & 0x1F;
break;
case 3:
wchar = ptr[pos] & 0x0F;
break;
case 4:
wchar = ptr[pos] & 0x07;
break;
}

/* JSON encoding */
if (should_escape) {
if (pos > beg) {
fbuffer_append(out_buffer, &in_utf8_str[beg], pos - beg);
}
for (short i = 1; i < ch_len; i++) {
wchar = (wchar<<6) | (ptr[pos+i] & 0x3F);
}

beg = pos + ch_len;
switch (ch) {
case '"': fbuffer_append(out_buffer, "\\\"", 2); break;
case '\\': fbuffer_append(out_buffer, "\\\\", 2); break;
case '/': fbuffer_append(out_buffer, "\\/", 2); break;
case '\b': fbuffer_append(out_buffer, "\\b", 2); break;
case '\f': fbuffer_append(out_buffer, "\\f", 2); break;
case '\n': fbuffer_append(out_buffer, "\\n", 2); break;
case '\r': fbuffer_append(out_buffer, "\\r", 2); break;
case '\t': fbuffer_append(out_buffer, "\\t", 2); break;
default:
if (ch <= 0xFFFF) {
scratch[2] = hexdig[ch >> 12];
scratch[3] = hexdig[(ch >> 8) & 0xf];
scratch[4] = hexdig[(ch >> 4) & 0xf];
scratch[5] = hexdig[ch & 0xf];
FLUSH_POS(ch_len);

if (wchar <= 0xFFFF) {
scratch[2] = hexdig[wchar >> 12];
scratch[3] = hexdig[(wchar >> 8) & 0xf];
scratch[4] = hexdig[(wchar >> 4) & 0xf];
scratch[5] = hexdig[wchar & 0xf];
fbuffer_append(out_buffer, scratch, 6);
} else {
uint16_t hi, lo;
ch -= 0x10000;
hi = 0xD800 + (uint16_t)(ch >> 10);
lo = 0xDC00 + (uint16_t)(ch & 0x3FF);
wchar -= 0x10000;
hi = 0xD800 + (uint16_t)(wchar >> 10);
lo = 0xDC00 + (uint16_t)(wchar & 0x3FF);

scratch[2] = hexdig[hi >> 12];
scratch[3] = hexdig[(hi >> 8) & 0xf];
Expand All @@ -279,17 +286,21 @@ static void convert_UTF8_to_ASCII_only_JSON(FBuffer *out_buffer, VALUE in_string

fbuffer_append(out_buffer, scratch, 12);
}

break;
}
}
} else {
pos++;
}

pos += ch_len;
}
#undef FLUSH_POS

if (beg < in_utf8_len) {
fbuffer_append(out_buffer, &in_utf8_str[beg], in_utf8_len - beg);
if (beg < len) {
fbuffer_append(out_buffer, &ptr[beg], len - beg);
}

RB_GC_GUARD(in_string);
RB_GC_GUARD(str);
}

static char *fstrndup(const char *ptr, unsigned long len) {
Expand Down Expand Up @@ -747,7 +758,7 @@ static void generate_json_string(FBuffer *buffer, VALUE Vstate, JSON_Generator_S
break;
case ENC_CODERANGE_VALID:
if (RB_UNLIKELY(state->ascii_only)) {
convert_UTF8_to_ASCII_only_JSON(buffer, obj, state->script_safe);
convert_UTF8_to_ASCII_only_JSON(buffer, obj, state->script_safe ? script_safe_escape_table : escape_table, state->script_safe);
} else {
convert_UTF8_to_JSON(buffer, obj, state->script_safe ? script_safe_escape_table : escape_table, state->script_safe);
}
Expand Down

0 comments on commit 42edaf7

Please sign in to comment.