Factor out json.decoder.decode_string

This commit will soon be followed by a rollback, but it is committed
anyway so that we can refer to these numbers in the git log.

name                                    old speed     new speed     delta

wuffs_json_decode_1k/clang5             607MB/s ± 0%  543MB/s ± 1%  -10.68%  (p=0.008 n=5+5)
wuffs_json_decode_21k_formatted/clang5  330MB/s ± 2%  320MB/s ± 3%     ~     (p=0.056 n=5+5)
wuffs_json_decode_26k_compact/clang5    468MB/s ± 2%  434MB/s ± 1%   -7.26%  (p=0.008 n=5+5)
wuffs_json_decode_217k_stringy/clang5   391MB/s ± 1%  363MB/s ± 0%   -7.03%  (p=0.008 n=5+5)

wuffs_json_decode_1k/gcc7               715MB/s ± 0%  583MB/s ± 0%  -18.38%  (p=0.016 n=5+4)
wuffs_json_decode_21k_formatted/gcc7    403MB/s ± 0%  350MB/s ± 0%  -12.94%  (p=0.016 n=5+4)
wuffs_json_decode_26k_compact/gcc7      534MB/s ± 0%  420MB/s ± 1%  -21.41%  (p=0.008 n=5+5)
wuffs_json_decode_217k_stringy/gcc7     435MB/s ± 0%  359MB/s ± 0%  -17.44%  (p=0.008 n=5+5)
This commit is contained in:
Nigel Tao committed 2020-04-04 11:32:35 +11:00
1 parent 721190afff
commit 86d3b89f9a
2 files changed
+1270 -1204

No files matched your search

+576 -570
View File
@@ -110,31 +110,13 @@ pub func decoder.decode_tokens?(dst: base.token_writer, src: base.io_reader, wor
var vminor : base.u32[..= 0xFF_FFFF]
var number_length : base.u32[..= 0x3FF]
var number_status : base.u32[..= 0x3]
var string_length : base.u32[..= 0xFFFB]
var whitespace_length : base.u32[..= 0xFFFE]
var depth : base.u32[..= 1024]
var stack_byte : base.u32[..= (1024 / 32) - 1]
var stack_bit : base.u32[..= 31]
var match : base.u32[..= 2]
var c4 : base.u32
var c : base.u8
var backslash : base.u8
var char : base.u8
var class : base.u8[..= 0x0F]
var multi_byte_utf8 : base.u32
var backslash_x_length : base.u32[..= 0xFFFF]
var backslash_x_ok : base.u8
var backslash_x_string : base.u32
var uni4_ok : base.u8
var uni4_string : base.u64
var uni4_value : base.u32[..= 0xFFFF]
var uni4_high_surrogate : base.u32[..= 0x10_FC00]
var uni8_ok : base.u8
var uni8_string : base.u64
var uni8_value : base.u32[..= 0xFFFF_FFFF]
// expect is a bitmask of what the next character class can be.
//
@@ -242,558 +224,7 @@ pub func decoder.decode_tokens?(dst: base.token_writer, src: base.io_reader, wor
length: 1)
args.src.skip32_fast!(actual: 1, worst_case: 1)
while.string_loop_outer true {
if args.dst.available() <= 0 {
yield? base."$short write"
continue.string_loop_outer
}
string_length = 0
while.string_loop_inner true,
pre args.dst.available() > 0,
{
if args.src.available() <= 0 {
if string_length > 0 {
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x40_0021,
link: 0x3,
length: string_length)
string_length = 0
}
if args.src.is_closed() {
return "#bad input"
}
yield? base."$short read"
string_length = 0
continue.string_loop_outer
}
// As an optimization, consume non-special ASCII 4 bytes at
// a time.
while args.src.available() > 4,
inv args.dst.available() > 0,
inv args.src.available() > 0,
{
c4 = args.src.peek_u32le()
if 0x00 <> (lut_chars[0xFF & (c4 >> 0)] |
lut_chars[0xFF & (c4 >> 8)] |
lut_chars[0xFF & (c4 >> 16)] |
lut_chars[0xFF & (c4 >> 24)]) {
break
}
args.src.skip32_fast!(actual: 4, worst_case: 4)
if string_length > (0xFFFB - 4) {
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x40_0021,
link: 0x3,
length: string_length + 4)
string_length = 0
continue.string_loop_outer
}
string_length += 4
} endwhile
c = args.src.peek_u8()
char = lut_chars[c]
if char == 0x00 { // Non-special ASCII.
args.src.skip32_fast!(actual: 1, worst_case: 1)
if string_length >= 0xFFFB {
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x40_0021,
link: 0x3,
length: 0xFFFC)
string_length = 0
continue.string_loop_outer
}
string_length += 1
continue.string_loop_inner
} else if char == 0x01 { // '"'
if string_length <> 0 {
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x40_0021,
link: 0x3,
length: string_length)
string_length = 0
}
break.string_loop_outer
} else if char == 0x02 { // '\\'.
if string_length > 0 {
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x40_0021,
link: 0x3,
length: string_length)
string_length = 0
if args.dst.available() <= 0 {
continue.string_loop_outer
}
}
assert args.dst.available() > 0
if args.src.available() < 2 {
if args.src.is_closed() {
return "#bad backslash-escape"
}
yield? base."$short read"
string_length = 0
char = 0
continue.string_loop_outer
}
c = (args.src.peek_u16le() >> 8) as base.u8
backslash = lut_backslashes[c]
if (backslash & 0x80) <> 0 {
args.src.skip32_fast!(actual: 2, worst_case: 2)
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x60_0000 | ((backslash & 0x7F) as base.u32),
link: 0x3,
length: 2)
continue.string_loop_outer
} else if backslash <> 0 {
if this.quirk_enabled_allow_backslash_etc[backslash & 7] {
args.src.skip32_fast!(actual: 2, worst_case: 2)
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x60_0000 | (lut_quirky_backslashes[backslash & 7] as base.u32),
link: 0x3,
length: 2)
continue.string_loop_outer
}
} else if c == 'u' {
// -------- BEGIN backslash-u.
if args.src.available() < 6 {
if args.src.is_closed() {
return "#bad backslash-escape"
}
yield? base."$short read"
string_length = 0
char = 0
continue.string_loop_outer
}
uni4_string = args.src.peek_u48le_as_u64() >> 16
uni4_value = 0
uni4_ok = 0x80
c = lut_hexadecimal_digits[0xFF & (uni4_string >> 0)]
uni4_ok &= c
uni4_value |= ((c & 0x0F) as base.u32) << 12
c = lut_hexadecimal_digits[0xFF & (uni4_string >> 8)]
uni4_ok &= c
uni4_value |= ((c & 0x0F) as base.u32) << 8
c = lut_hexadecimal_digits[0xFF & (uni4_string >> 16)]
uni4_ok &= c
uni4_value |= ((c & 0x0F) as base.u32) << 4
c = lut_hexadecimal_digits[0xFF & (uni4_string >> 24)]
uni4_ok &= c
uni4_value |= ((c & 0x0F) as base.u32) << 0
if uni4_ok == 0 {
// It wasn't 4 hexadecimal digits. No-op (and
// fall through to "#bad backslash-escape").
} else if (uni4_value < 0xD800) or (0xDFFF < uni4_value) {
// Not a Unicode surrogate. We're good.
args.src.skip32_fast!(actual: 6, worst_case: 6)
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x60_0000 | uni4_value,
link: 0x3,
length: 6)
continue.string_loop_outer
} else if uni4_value >= 0xDC00 {
// Low surrogate. No-op (and fall through to
// "#bad backslash-escape").
} else {
// High surrogate, which needs to be followed
// by a "\\u1234" low surrogate. We've already
// peeked 6 bytes for the high surrogate. We
// need 12 in total: another 8 bytes at an
// offset of 4.
if args.src.available() < 12 {
if args.src.is_closed() {
if this.quirk_enabled_replace_invalid_unicode {
args.src.skip32_fast!(actual: 6, worst_case: 6)
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x60_FFFD,
link: 0x3,
length: 6)
continue.string_loop_outer
}
return "#bad backslash-escape"
}
yield? base."$short read"
string_length = 0
uni4_value = 0
char = 0
continue.string_loop_outer
}
uni4_string = args.src.peek_u64le_at(offset: 4) >> 16
// Look for the low surrogate's "\\u".
if ((0xFF & (uni4_string >> 0)) <> '\\') or
((0xFF & (uni4_string >> 8)) <> 'u') {
uni4_high_surrogate = 0
uni4_value = 0
uni4_ok = 0
} else {
uni4_high_surrogate =
0x1_0000 + ((uni4_value - 0xD800) << 10)
uni4_value = 0
uni4_ok = 0x80
uni4_string >>= 16
c = lut_hexadecimal_digits[0xFF & (uni4_string >> 0)]
uni4_ok &= c
uni4_value |= ((c & 0x0F) as base.u32) << 12
c = lut_hexadecimal_digits[0xFF & (uni4_string >> 8)]
uni4_ok &= c
uni4_value |= ((c & 0x0F) as base.u32) << 8
c = lut_hexadecimal_digits[0xFF & (uni4_string >> 16)]
uni4_ok &= c
uni4_value |= ((c & 0x0F) as base.u32) << 4
c = lut_hexadecimal_digits[0xFF & (uni4_string >> 24)]
uni4_ok &= c
uni4_value |= ((c & 0x0F) as base.u32) << 0
}
if (uni4_ok <> 0) and
(0xDC00 <= uni4_value) and (uni4_value <= 0xDFFF) {
// Emit a single token for the surrogate
// pair.
uni4_value -= 0xDC00
args.src.skip32_fast!(actual: 12, worst_case: 12)
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x60_0000 | uni4_high_surrogate | uni4_value,
link: 0x3,
length: 12)
continue.string_loop_outer
}
}
if this.quirk_enabled_replace_invalid_unicode {
if args.src.available() < 6 {
return "#internal error: inconsistent I/O"
}
args.src.skip32_fast!(actual: 6, worst_case: 6)
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x60_FFFD,
link: 0x3,
length: 6)
continue.string_loop_outer
}
// -------- END backslash-u.
} else if (c == 'U') and
this.quirk_enabled_allow_backslash_capital_u {
// -------- BEGIN backslash-capital-u.
if args.src.available() < 10 {
if args.src.is_closed() {
return "#bad backslash-escape"
}
yield? base."$short read"
string_length = 0
char = 0
continue.string_loop_outer
}
uni8_string = args.src.peek_u64le_at(offset: 2)
uni8_value = 0
uni8_ok = 0x80
c = lut_hexadecimal_digits[0xFF & (uni8_string >> 0)]
uni8_ok &= c
uni8_value |= ((c & 0x0F) as base.u32) << 28
c = lut_hexadecimal_digits[0xFF & (uni8_string >> 8)]
uni8_ok &= c
uni8_value |= ((c & 0x0F) as base.u32) << 24
c = lut_hexadecimal_digits[0xFF & (uni8_string >> 16)]
uni8_ok &= c
uni8_value |= ((c & 0x0F) as base.u32) << 20
c = lut_hexadecimal_digits[0xFF & (uni8_string >> 24)]
uni8_ok &= c
uni8_value |= ((c & 0x0F) as base.u32) << 16
c = lut_hexadecimal_digits[0xFF & (uni8_string >> 32)]
uni8_ok &= c
uni8_value |= ((c & 0x0F) as base.u32) << 12
c = lut_hexadecimal_digits[0xFF & (uni8_string >> 40)]
uni8_ok &= c
uni8_value |= ((c & 0x0F) as base.u32) << 8
c = lut_hexadecimal_digits[0xFF & (uni8_string >> 48)]
uni8_ok &= c
uni8_value |= ((c & 0x0F) as base.u32) << 4
c = lut_hexadecimal_digits[0xFF & (uni8_string >> 56)]
uni8_ok &= c
uni8_value |= ((c & 0x0F) as base.u32) << 0
if uni8_ok == 0 {
// It wasn't 8 hexadecimal digits. No-op (and
// fall through to "#bad backslash-escape").
} else if (uni8_value < 0xD800) or (
(0xDFFF < uni8_value) and (uni8_value <= 0x10_FFFF)) {
// Not a Unicode surrogate. We're good.
args.src.skip32_fast!(actual: 10, worst_case: 10)
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x60_0000 | (uni8_value & 0x1F_FFFF),
link: 0x3,
length: 10)
continue.string_loop_outer
} else if this.quirk_enabled_replace_invalid_unicode {
args.src.skip32_fast!(actual: 10, worst_case: 10)
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x60_FFFD,
link: 0x3,
length: 10)
continue.string_loop_outer
}
// -------- END backslash-capital-u.
} else if (c == 'x') and
this.quirk_enabled_allow_backslash_x {
// -------- BEGIN backslash-x
if args.src.available() < 4 {
if args.src.is_closed() {
return "#bad backslash-escape"
}
yield? base."$short read"
string_length = 0
char = 0
continue.string_loop_outer
}
backslash_x_length = 0
while (backslash_x_length <= 0xFFFB) and (args.src.available() >= 4),
inv args.dst.available() > 0,
{
backslash_x_string = args.src.peek_u32le()
backslash_x_ok = 0x80
c = lut_hexadecimal_digits[0xFF & (backslash_x_string >> 16)]
backslash_x_ok &= c
c = lut_hexadecimal_digits[0xFF & (backslash_x_string >> 24)]
backslash_x_ok &= c
if (backslash_x_ok == 0) or
((backslash_x_string & 0xFFFF) <> 0x785C) {
// It wasn't "\\x34", for some hexadecimal
// digits "34".
break
}
args.src.skip32_fast!(actual: 4, worst_case: 4)
backslash_x_length += 4
} endwhile
if backslash_x_length == 0 {
return "#bad backslash-escape"
}
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x40_0080,
link: 0x3,
length: backslash_x_length)
continue.string_loop_outer
// -------- END backslash-x
}
return "#bad backslash-escape"
} else if char == 0x03 { // 2-byte UTF-8.
if args.src.available() < 2 {
if string_length > 0 {
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x40_0021,
link: 0x3,
length: string_length)
string_length = 0
if args.dst.available() <= 0 {
continue.string_loop_outer
}
}
if args.src.is_closed() {
if this.quirk_enabled_replace_invalid_unicode {
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x60_FFFD,
link: 0x3,
length: 1)
args.src.skip32_fast!(actual: 1, worst_case: 1)
continue.string_loop_outer
}
return "#bad UTF-8"
}
yield? base."$short read"
string_length = 0
char = 0
continue.string_loop_outer
}
multi_byte_utf8 = args.src.peek_u16le_as_u32()
if (multi_byte_utf8 & 0xC000) == 0x8000 {
multi_byte_utf8 = (0x00_07C0 & (multi_byte_utf8 ~mod<< 6)) |
(0x00_003F & (multi_byte_utf8 >> 8))
args.src.skip32_fast!(actual: 2, worst_case: 2)
if string_length >= 0xFFF8 {
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x40_0021,
link: 0x3,
length: string_length + 2)
string_length = 0
continue.string_loop_outer
}
string_length += 2
continue.string_loop_inner
}
} else if char == 0x04 { // 3-byte UTF-8.
if args.src.available() < 3 {
if string_length > 0 {
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x40_0021,
link: 0x3,
length: string_length)
string_length = 0
if args.dst.available() <= 0 {
continue.string_loop_outer
}
}
if args.src.is_closed() {
if this.quirk_enabled_replace_invalid_unicode {
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x60_FFFD,
link: 0x3,
length: 1)
args.src.skip32_fast!(actual: 1, worst_case: 1)
continue.string_loop_outer
}
return "#bad UTF-8"
}
yield? base."$short read"
string_length = 0
char = 0
continue.string_loop_outer
}
multi_byte_utf8 = args.src.peek_u24le_as_u32()
if (multi_byte_utf8 & 0xC0_C000) == 0x80_8000 {
multi_byte_utf8 = (0x00_F000 & (multi_byte_utf8 ~mod<< 12)) |
(0x00_0FC0 & (multi_byte_utf8 >> 2)) |
(0x00_003F & (multi_byte_utf8 >> 16))
if (0x07FF < multi_byte_utf8) and
((multi_byte_utf8 < 0xD800) or (0xDFFF < multi_byte_utf8)) {
args.src.skip32_fast!(actual: 3, worst_case: 3)
if string_length >= 0xFFF8 {
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x40_0021,
link: 0x3,
length: string_length + 3)
string_length = 0
continue.string_loop_outer
}
string_length += 3
continue.string_loop_inner
}
}
} else if char == 0x05 { // 4-byte UTF-8.
if args.src.available() < 4 {
if string_length > 0 {
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x40_0021,
link: 0x3,
length: string_length)
string_length = 0
if args.dst.available() <= 0 {
continue.string_loop_outer
}
}
if args.src.is_closed() {
if this.quirk_enabled_replace_invalid_unicode {
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x60_FFFD,
link: 0x3,
length: 1)
args.src.skip32_fast!(actual: 1, worst_case: 1)
continue.string_loop_outer
}
return "#bad UTF-8"
}
yield? base."$short read"
string_length = 0
char = 0
continue.string_loop_outer
}
multi_byte_utf8 = args.src.peek_u32le()
if (multi_byte_utf8 & 0xC0C0_C000) == 0x8080_8000 {
multi_byte_utf8 = (0x1C_0000 & (multi_byte_utf8 ~mod<< 18)) |
(0x03_F000 & (multi_byte_utf8 ~mod<< 4)) |
(0x00_0FC0 & (multi_byte_utf8 >> 10)) |
(0x00_003F & (multi_byte_utf8 >> 24))
if (0xFFFF < multi_byte_utf8) and (multi_byte_utf8 <= 0x10_FFFF) {
args.src.skip32_fast!(actual: 4, worst_case: 4)
if string_length >= 0xFFF8 {
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x40_0021,
link: 0x3,
length: string_length + 4)
string_length = 0
continue.string_loop_outer
}
string_length += 4
continue.string_loop_inner
}
}
}
if string_length > 0 {
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x40_0021,
link: 0x3,
length: string_length)
string_length = 0
if args.dst.available() <= 0 {
continue.string_loop_outer
}
}
if char == 0x80 {
return "#bad C0 control code"
}
if this.quirk_enabled_replace_invalid_unicode {
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x60_FFFD,
link: 0x3,
length: 1)
args.src.skip32_fast!(actual: 1, worst_case: 1)
continue.string_loop_outer
}
return "#bad UTF-8"
} endwhile.string_loop_inner
} endwhile.string_loop_outer
this.decode_string?(dst: args.dst, src: args.src)
// Emit the trailing '"'.
while true {
@@ -1119,6 +550,581 @@ pub func decoder.decode_tokens?(dst: base.token_writer, src: base.io_reader, wor
this.end_of_data = true
}
pri func decoder.decode_string?(dst: base.token_writer, src: base.io_reader) {
var string_length : base.u32[..= 0xFFFB]
var c4 : base.u32
var c : base.u8
var backslash : base.u8
var char : base.u8
var multi_byte_utf8 : base.u32
var backslash_x_length : base.u32[..= 0xFFFF]
var backslash_x_ok : base.u8
var backslash_x_string : base.u32
var uni4_ok : base.u8
var uni4_string : base.u64
var uni4_value : base.u32[..= 0xFFFF]
var uni4_high_surrogate : base.u32[..= 0x10_FC00]
var uni8_ok : base.u8
var uni8_string : base.u64
var uni8_value : base.u32[..= 0xFFFF_FFFF]
while.string_loop_outer true {
if args.dst.available() <= 0 {
yield? base."$short write"
continue.string_loop_outer
}
string_length = 0
while.string_loop_inner true,
pre args.dst.available() > 0,
{
if args.src.available() <= 0 {
if string_length > 0 {
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x40_0021,
link: 0x3,
length: string_length)
string_length = 0
}
if args.src.is_closed() {
return "#bad input"
}
yield? base."$short read"
string_length = 0
continue.string_loop_outer
}
// As an optimization, consume non-special ASCII 4 bytes at
// a time.
while args.src.available() > 4,
inv args.dst.available() > 0,
inv args.src.available() > 0,
{
c4 = args.src.peek_u32le()
if 0x00 <> (lut_chars[0xFF & (c4 >> 0)] |
lut_chars[0xFF & (c4 >> 8)] |
lut_chars[0xFF & (c4 >> 16)] |
lut_chars[0xFF & (c4 >> 24)]) {
break
}
args.src.skip32_fast!(actual: 4, worst_case: 4)
if string_length > (0xFFFB - 4) {
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x40_0021,
link: 0x3,
length: string_length + 4)
string_length = 0
continue.string_loop_outer
}
string_length += 4
} endwhile
c = args.src.peek_u8()
char = lut_chars[c]
if char == 0x00 { // Non-special ASCII.
args.src.skip32_fast!(actual: 1, worst_case: 1)
if string_length >= 0xFFFB {
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x40_0021,
link: 0x3,
length: 0xFFFC)
string_length = 0
continue.string_loop_outer
}
string_length += 1
continue.string_loop_inner
} else if char == 0x01 { // '"'
if string_length <> 0 {
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x40_0021,
link: 0x3,
length: string_length)
string_length = 0
}
break.string_loop_outer
} else if char == 0x02 { // '\\'.
if string_length > 0 {
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x40_0021,
link: 0x3,
length: string_length)
string_length = 0
if args.dst.available() <= 0 {
continue.string_loop_outer
}
}
assert args.dst.available() > 0
if args.src.available() < 2 {
if args.src.is_closed() {
return "#bad backslash-escape"
}
yield? base."$short read"
string_length = 0
char = 0
continue.string_loop_outer
}
c = (args.src.peek_u16le() >> 8) as base.u8
backslash = lut_backslashes[c]
if (backslash & 0x80) <> 0 {
args.src.skip32_fast!(actual: 2, worst_case: 2)
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x60_0000 | ((backslash & 0x7F) as base.u32),
link: 0x3,
length: 2)
continue.string_loop_outer
} else if backslash <> 0 {
if this.quirk_enabled_allow_backslash_etc[backslash & 7] {
args.src.skip32_fast!(actual: 2, worst_case: 2)
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x60_0000 | (lut_quirky_backslashes[backslash & 7] as base.u32),
link: 0x3,
length: 2)
continue.string_loop_outer
}
} else if c == 'u' {
// -------- BEGIN backslash-u.
if args.src.available() < 6 {
if args.src.is_closed() {
return "#bad backslash-escape"
}
yield? base."$short read"
string_length = 0
char = 0
continue.string_loop_outer
}
uni4_string = args.src.peek_u48le_as_u64() >> 16
uni4_value = 0
uni4_ok = 0x80
c = lut_hexadecimal_digits[0xFF & (uni4_string >> 0)]
uni4_ok &= c
uni4_value |= ((c & 0x0F) as base.u32) << 12
c = lut_hexadecimal_digits[0xFF & (uni4_string >> 8)]
uni4_ok &= c
uni4_value |= ((c & 0x0F) as base.u32) << 8
c = lut_hexadecimal_digits[0xFF & (uni4_string >> 16)]
uni4_ok &= c
uni4_value |= ((c & 0x0F) as base.u32) << 4
c = lut_hexadecimal_digits[0xFF & (uni4_string >> 24)]
uni4_ok &= c
uni4_value |= ((c & 0x0F) as base.u32) << 0
if uni4_ok == 0 {
// It wasn't 4 hexadecimal digits. No-op (and
// fall through to "#bad backslash-escape").
} else if (uni4_value < 0xD800) or (0xDFFF < uni4_value) {
// Not a Unicode surrogate. We're good.
args.src.skip32_fast!(actual: 6, worst_case: 6)
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x60_0000 | uni4_value,
link: 0x3,
length: 6)
continue.string_loop_outer
} else if uni4_value >= 0xDC00 {
// Low surrogate. No-op (and fall through to
// "#bad backslash-escape").
} else {
// High surrogate, which needs to be followed
// by a "\\u1234" low surrogate. We've already
// peeked 6 bytes for the high surrogate. We
// need 12 in total: another 8 bytes at an
// offset of 4.
if args.src.available() < 12 {
if args.src.is_closed() {
if this.quirk_enabled_replace_invalid_unicode {
args.src.skip32_fast!(actual: 6, worst_case: 6)
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x60_FFFD,
link: 0x3,
length: 6)
continue.string_loop_outer
}
return "#bad backslash-escape"
}
yield? base."$short read"
string_length = 0
uni4_value = 0
char = 0
continue.string_loop_outer
}
uni4_string = args.src.peek_u64le_at(offset: 4) >> 16
// Look for the low surrogate's "\\u".
if ((0xFF & (uni4_string >> 0)) <> '\\') or
((0xFF & (uni4_string >> 8)) <> 'u') {
uni4_high_surrogate = 0
uni4_value = 0
uni4_ok = 0
} else {
uni4_high_surrogate =
0x1_0000 + ((uni4_value - 0xD800) << 10)
uni4_value = 0
uni4_ok = 0x80
uni4_string >>= 16
c = lut_hexadecimal_digits[0xFF & (uni4_string >> 0)]
uni4_ok &= c
uni4_value |= ((c & 0x0F) as base.u32) << 12
c = lut_hexadecimal_digits[0xFF & (uni4_string >> 8)]
uni4_ok &= c
uni4_value |= ((c & 0x0F) as base.u32) << 8
c = lut_hexadecimal_digits[0xFF & (uni4_string >> 16)]
uni4_ok &= c
uni4_value |= ((c & 0x0F) as base.u32) << 4
c = lut_hexadecimal_digits[0xFF & (uni4_string >> 24)]
uni4_ok &= c
uni4_value |= ((c & 0x0F) as base.u32) << 0
}
if (uni4_ok <> 0) and
(0xDC00 <= uni4_value) and (uni4_value <= 0xDFFF) {
// Emit a single token for the surrogate
// pair.
uni4_value -= 0xDC00
args.src.skip32_fast!(actual: 12, worst_case: 12)
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x60_0000 | uni4_high_surrogate | uni4_value,
link: 0x3,
length: 12)
continue.string_loop_outer
}
}
if this.quirk_enabled_replace_invalid_unicode {
if args.src.available() < 6 {
return "#internal error: inconsistent I/O"
}
args.src.skip32_fast!(actual: 6, worst_case: 6)
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x60_FFFD,
link: 0x3,
length: 6)
continue.string_loop_outer
}
// -------- END backslash-u.
} else if (c == 'U') and
this.quirk_enabled_allow_backslash_capital_u {
// -------- BEGIN backslash-capital-u.
if args.src.available() < 10 {
if args.src.is_closed() {
return "#bad backslash-escape"
}
yield? base."$short read"
string_length = 0
char = 0
continue.string_loop_outer
}
uni8_string = args.src.peek_u64le_at(offset: 2)
uni8_value = 0
uni8_ok = 0x80
c = lut_hexadecimal_digits[0xFF & (uni8_string >> 0)]
uni8_ok &= c
uni8_value |= ((c & 0x0F) as base.u32) << 28
c = lut_hexadecimal_digits[0xFF & (uni8_string >> 8)]
uni8_ok &= c
uni8_value |= ((c & 0x0F) as base.u32) << 24
c = lut_hexadecimal_digits[0xFF & (uni8_string >> 16)]
uni8_ok &= c
uni8_value |= ((c & 0x0F) as base.u32) << 20
c = lut_hexadecimal_digits[0xFF & (uni8_string >> 24)]
uni8_ok &= c
uni8_value |= ((c & 0x0F) as base.u32) << 16
c = lut_hexadecimal_digits[0xFF & (uni8_string >> 32)]
uni8_ok &= c
uni8_value |= ((c & 0x0F) as base.u32) << 12
c = lut_hexadecimal_digits[0xFF & (uni8_string >> 40)]
uni8_ok &= c
uni8_value |= ((c & 0x0F) as base.u32) << 8
c = lut_hexadecimal_digits[0xFF & (uni8_string >> 48)]
uni8_ok &= c
uni8_value |= ((c & 0x0F) as base.u32) << 4
c = lut_hexadecimal_digits[0xFF & (uni8_string >> 56)]
uni8_ok &= c
uni8_value |= ((c & 0x0F) as base.u32) << 0
if uni8_ok == 0 {
// It wasn't 8 hexadecimal digits. No-op (and
// fall through to "#bad backslash-escape").
} else if (uni8_value < 0xD800) or (
(0xDFFF < uni8_value) and (uni8_value <= 0x10_FFFF)) {
// Not a Unicode surrogate. We're good.
args.src.skip32_fast!(actual: 10, worst_case: 10)
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x60_0000 | (uni8_value & 0x1F_FFFF),
link: 0x3,
length: 10)
continue.string_loop_outer
} else if this.quirk_enabled_replace_invalid_unicode {
args.src.skip32_fast!(actual: 10, worst_case: 10)
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x60_FFFD,
link: 0x3,
length: 10)
continue.string_loop_outer
}
// -------- END backslash-capital-u.
} else if (c == 'x') and
this.quirk_enabled_allow_backslash_x {
// -------- BEGIN backslash-x
if args.src.available() < 4 {
if args.src.is_closed() {
return "#bad backslash-escape"
}
yield? base."$short read"
string_length = 0
char = 0
continue.string_loop_outer
}
backslash_x_length = 0
while (backslash_x_length <= 0xFFFB) and (args.src.available() >= 4),
inv args.dst.available() > 0,
{
backslash_x_string = args.src.peek_u32le()
backslash_x_ok = 0x80
c = lut_hexadecimal_digits[0xFF & (backslash_x_string >> 16)]
backslash_x_ok &= c
c = lut_hexadecimal_digits[0xFF & (backslash_x_string >> 24)]
backslash_x_ok &= c
if (backslash_x_ok == 0) or
((backslash_x_string & 0xFFFF) <> 0x785C) {
// It wasn't "\\x34", for some hexadecimal
// digits "34".
break
}
args.src.skip32_fast!(actual: 4, worst_case: 4)
backslash_x_length += 4
} endwhile
if backslash_x_length == 0 {
return "#bad backslash-escape"
}
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x40_0080,
link: 0x3,
length: backslash_x_length)
continue.string_loop_outer
// -------- END backslash-x
}
return "#bad backslash-escape"
} else if char == 0x03 { // 2-byte UTF-8.
if args.src.available() < 2 {
if string_length > 0 {
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x40_0021,
link: 0x3,
length: string_length)
string_length = 0
if args.dst.available() <= 0 {
continue.string_loop_outer
}
}
if args.src.is_closed() {
if this.quirk_enabled_replace_invalid_unicode {
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x60_FFFD,
link: 0x3,
length: 1)
args.src.skip32_fast!(actual: 1, worst_case: 1)
continue.string_loop_outer
}
return "#bad UTF-8"
}
yield? base."$short read"
string_length = 0
char = 0
continue.string_loop_outer
}
multi_byte_utf8 = args.src.peek_u16le_as_u32()
if (multi_byte_utf8 & 0xC000) == 0x8000 {
multi_byte_utf8 = (0x00_07C0 & (multi_byte_utf8 ~mod<< 6)) |
(0x00_003F & (multi_byte_utf8 >> 8))
args.src.skip32_fast!(actual: 2, worst_case: 2)
if string_length >= 0xFFF8 {
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x40_0021,
link: 0x3,
length: string_length + 2)
string_length = 0
continue.string_loop_outer
}
string_length += 2
continue.string_loop_inner
}
} else if char == 0x04 { // 3-byte UTF-8.
if args.src.available() < 3 {
if string_length > 0 {
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x40_0021,
link: 0x3,
length: string_length)
string_length = 0
if args.dst.available() <= 0 {
continue.string_loop_outer
}
}
if args.src.is_closed() {
if this.quirk_enabled_replace_invalid_unicode {
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x60_FFFD,
link: 0x3,
length: 1)
args.src.skip32_fast!(actual: 1, worst_case: 1)
continue.string_loop_outer
}
return "#bad UTF-8"
}
yield? base."$short read"
string_length = 0
char = 0
continue.string_loop_outer
}
multi_byte_utf8 = args.src.peek_u24le_as_u32()
if (multi_byte_utf8 & 0xC0_C000) == 0x80_8000 {
multi_byte_utf8 = (0x00_F000 & (multi_byte_utf8 ~mod<< 12)) |
(0x00_0FC0 & (multi_byte_utf8 >> 2)) |
(0x00_003F & (multi_byte_utf8 >> 16))
if (0x07FF < multi_byte_utf8) and
((multi_byte_utf8 < 0xD800) or (0xDFFF < multi_byte_utf8)) {
args.src.skip32_fast!(actual: 3, worst_case: 3)
if string_length >= 0xFFF8 {
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x40_0021,
link: 0x3,
length: string_length + 3)
string_length = 0
continue.string_loop_outer
}
string_length += 3
continue.string_loop_inner
}
}
} else if char == 0x05 { // 4-byte UTF-8.
if args.src.available() < 4 {
if string_length > 0 {
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x40_0021,
link: 0x3,
length: string_length)
string_length = 0
if args.dst.available() <= 0 {
continue.string_loop_outer
}
}
if args.src.is_closed() {
if this.quirk_enabled_replace_invalid_unicode {
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x60_FFFD,
link: 0x3,
length: 1)
args.src.skip32_fast!(actual: 1, worst_case: 1)
continue.string_loop_outer
}
return "#bad UTF-8"
}
yield? base."$short read"
string_length = 0
char = 0
continue.string_loop_outer
}
multi_byte_utf8 = args.src.peek_u32le()
if (multi_byte_utf8 & 0xC0C0_C000) == 0x8080_8000 {
multi_byte_utf8 = (0x1C_0000 & (multi_byte_utf8 ~mod<< 18)) |
(0x03_F000 & (multi_byte_utf8 ~mod<< 4)) |
(0x00_0FC0 & (multi_byte_utf8 >> 10)) |
(0x00_003F & (multi_byte_utf8 >> 24))
if (0xFFFF < multi_byte_utf8) and (multi_byte_utf8 <= 0x10_FFFF) {
args.src.skip32_fast!(actual: 4, worst_case: 4)
if string_length >= 0xFFF8 {
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x40_0021,
link: 0x3,
length: string_length + 4)
string_length = 0
continue.string_loop_outer
}
string_length += 4
continue.string_loop_inner
}
}
}
if string_length > 0 {
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x40_0021,
link: 0x3,
length: string_length)
string_length = 0
if args.dst.available() <= 0 {
continue.string_loop_outer
}
}
if char == 0x80 {
return "#bad C0 control code"
}
if this.quirk_enabled_replace_invalid_unicode {
args.dst.write_simple_token_fast!(
value_major: 0,
value_minor: 0x60_FFFD,
link: 0x3,
length: 1)
args.src.skip32_fast!(actual: 1, worst_case: 1)
continue.string_loop_outer
}
return "#bad UTF-8"
} endwhile.string_loop_inner
} endwhile.string_loop_outer
}
pri func decoder.decode_number!(src: base.io_reader) base.u32[..= 0x3FF] {
var c : base.u8
var n : base.u32[..= 0x3FF]