| 1 | //! Helper code for character escaping. |
| 2 | |
| 3 | use crate::ascii; |
| 4 | use crate::num::NonZero; |
| 5 | use crate::ops::Range; |
| 6 | |
| 7 | const HEX_DIGITS: [ascii::Char; 16] = *b"0123456789abcdef" .as_ascii().unwrap(); |
| 8 | |
| 9 | #[inline ] |
| 10 | const fn backslash<const N: usize>(a: ascii::Char) -> ([ascii::Char; N], Range<u8>) { |
| 11 | const { assert!(N >= 2) }; |
| 12 | |
| 13 | let mut output: [AsciiChar; N] = [ascii::Char::Null; N]; |
| 14 | |
| 15 | output[0] = ascii::Char::ReverseSolidus; |
| 16 | output[1] = a; |
| 17 | |
| 18 | (output, 0..2) |
| 19 | } |
| 20 | |
| 21 | #[inline ] |
| 22 | const fn hex_escape<const N: usize>(byte: u8) -> ([ascii::Char; N], Range<u8>) { |
| 23 | const { assert!(N >= 4) }; |
| 24 | |
| 25 | let mut output: [AsciiChar; N] = [ascii::Char::Null; N]; |
| 26 | |
| 27 | let hi: AsciiChar = HEX_DIGITS[(byte >> 4) as usize]; |
| 28 | let lo: AsciiChar = HEX_DIGITS[(byte & 0xf) as usize]; |
| 29 | |
| 30 | output[0] = ascii::Char::ReverseSolidus; |
| 31 | output[1] = ascii::Char::SmallX; |
| 32 | output[2] = hi; |
| 33 | output[3] = lo; |
| 34 | |
| 35 | (output, 0..4) |
| 36 | } |
| 37 | |
| 38 | #[inline ] |
| 39 | const fn verbatim<const N: usize>(a: ascii::Char) -> ([ascii::Char; N], Range<u8>) { |
| 40 | const { assert!(N >= 1) }; |
| 41 | |
| 42 | let mut output: [AsciiChar; N] = [ascii::Char::Null; N]; |
| 43 | |
| 44 | output[0] = a; |
| 45 | |
| 46 | (output, 0..1) |
| 47 | } |
| 48 | |
| 49 | /// Escapes an ASCII character. |
| 50 | /// |
| 51 | /// Returns a buffer and the length of the escaped representation. |
| 52 | const fn escape_ascii<const N: usize>(byte: u8) -> ([ascii::Char; N], Range<u8>) { |
| 53 | const { assert!(N >= 4) }; |
| 54 | |
| 55 | #[cfg (feature = "optimize_for_size" )] |
| 56 | { |
| 57 | match byte { |
| 58 | b' \t' => backslash(ascii::Char::SmallT), |
| 59 | b' \r' => backslash(ascii::Char::SmallR), |
| 60 | b' \n' => backslash(ascii::Char::SmallN), |
| 61 | b' \\' => backslash(ascii::Char::ReverseSolidus), |
| 62 | b' \'' => backslash(ascii::Char::Apostrophe), |
| 63 | b'"' => backslash(ascii::Char::QuotationMark), |
| 64 | 0x00..=0x1F | 0x7F => hex_escape(byte), |
| 65 | _ => match ascii::Char::from_u8(byte) { |
| 66 | Some(a) => verbatim(a), |
| 67 | None => hex_escape(byte), |
| 68 | }, |
| 69 | } |
| 70 | } |
| 71 | |
| 72 | #[cfg (not(feature = "optimize_for_size" ))] |
| 73 | { |
| 74 | /// Lookup table helps us determine how to display character. |
| 75 | /// |
| 76 | /// Since ASCII characters will always be 7 bits, we can exploit this to store the 8th bit to |
| 77 | /// indicate whether the result is escaped or unescaped. |
| 78 | /// |
| 79 | /// We additionally use 0x80 (escaped NUL character) to indicate hex-escaped bytes, since |
| 80 | /// escaped NUL will not occur. |
| 81 | const LOOKUP: [u8; 256] = { |
| 82 | let mut arr = [0; 256]; |
| 83 | let mut idx = 0; |
| 84 | while idx <= 255 { |
| 85 | arr[idx] = match idx as u8 { |
| 86 | // use 8th bit to indicate escaped |
| 87 | b' \t' => 0x80 | b't' , |
| 88 | b' \r' => 0x80 | b'r' , |
| 89 | b' \n' => 0x80 | b'n' , |
| 90 | b' \\' => 0x80 | b' \\' , |
| 91 | b' \'' => 0x80 | b' \'' , |
| 92 | b'"' => 0x80 | b'"' , |
| 93 | |
| 94 | // use NUL to indicate hex-escaped |
| 95 | 0x00..=0x1F | 0x7F..=0xFF => 0x80 | b' \0' , |
| 96 | |
| 97 | idx => idx, |
| 98 | }; |
| 99 | idx += 1; |
| 100 | } |
| 101 | arr |
| 102 | }; |
| 103 | |
| 104 | let lookup = LOOKUP[byte as usize]; |
| 105 | |
| 106 | // 8th bit indicates escape |
| 107 | let lookup_escaped = lookup & 0x80 != 0; |
| 108 | |
| 109 | // SAFETY: We explicitly mask out the eighth bit to get a 7-bit ASCII character. |
| 110 | let lookup_ascii = unsafe { ascii::Char::from_u8_unchecked(lookup & 0x7F) }; |
| 111 | |
| 112 | if lookup_escaped { |
| 113 | // NUL indicates hex-escaped |
| 114 | if matches!(lookup_ascii, ascii::Char::Null) { |
| 115 | hex_escape(byte) |
| 116 | } else { |
| 117 | backslash(lookup_ascii) |
| 118 | } |
| 119 | } else { |
| 120 | verbatim(lookup_ascii) |
| 121 | } |
| 122 | } |
| 123 | } |
| 124 | |
| 125 | /// Escapes a character `\u{NNNN}` representation. |
| 126 | /// |
| 127 | /// Returns a buffer and the length of the escaped representation. |
| 128 | const fn escape_unicode<const N: usize>(c: char) -> ([ascii::Char; N], Range<u8>) { |
| 129 | const { assert!(N >= 10 && N < u8::MAX as usize) }; |
| 130 | |
| 131 | let c: u32 = c as u32; |
| 132 | |
| 133 | // OR-ing `1` ensures that for `c == 0` the code computes that |
| 134 | // one digit should be printed. |
| 135 | let start: usize = (c | 1).leading_zeros() as usize / 4 - 2; |
| 136 | |
| 137 | let mut output: [AsciiChar; N] = [ascii::Char::Null; N]; |
| 138 | output[3] = HEX_DIGITS[((c >> 20) & 15) as usize]; |
| 139 | output[4] = HEX_DIGITS[((c >> 16) & 15) as usize]; |
| 140 | output[5] = HEX_DIGITS[((c >> 12) & 15) as usize]; |
| 141 | output[6] = HEX_DIGITS[((c >> 8) & 15) as usize]; |
| 142 | output[7] = HEX_DIGITS[((c >> 4) & 15) as usize]; |
| 143 | output[8] = HEX_DIGITS[((c >> 0) & 15) as usize]; |
| 144 | output[9] = ascii::Char::RightCurlyBracket; |
| 145 | output[start + 0] = ascii::Char::ReverseSolidus; |
| 146 | output[start + 1] = ascii::Char::SmallU; |
| 147 | output[start + 2] = ascii::Char::LeftCurlyBracket; |
| 148 | |
| 149 | (output, (start as u8)..(N as u8)) |
| 150 | } |
| 151 | |
| 152 | /// An iterator over an fixed-size array. |
| 153 | /// |
| 154 | /// This is essentially equivalent to array’s IntoIter except that indexes are |
| 155 | /// limited to u8 to reduce size of the structure. |
| 156 | #[derive (Clone, Debug)] |
| 157 | pub(crate) struct EscapeIterInner<const N: usize> { |
| 158 | // The element type ensures this is always ASCII, and thus also valid UTF-8. |
| 159 | data: [ascii::Char; N], |
| 160 | |
| 161 | // Invariant: `alive.start <= alive.end <= N` |
| 162 | alive: Range<u8>, |
| 163 | } |
| 164 | |
| 165 | impl<const N: usize> EscapeIterInner<N> { |
| 166 | pub(crate) const fn backslash(c: ascii::Char) -> Self { |
| 167 | let (data, range) = backslash(c); |
| 168 | Self { data, alive: range } |
| 169 | } |
| 170 | |
| 171 | pub(crate) const fn ascii(c: u8) -> Self { |
| 172 | let (data, range) = escape_ascii(c); |
| 173 | Self { data, alive: range } |
| 174 | } |
| 175 | |
| 176 | pub(crate) const fn unicode(c: char) -> Self { |
| 177 | let (data, range) = escape_unicode(c); |
| 178 | Self { data, alive: range } |
| 179 | } |
| 180 | |
| 181 | #[inline ] |
| 182 | pub(crate) const fn empty() -> Self { |
| 183 | Self { data: [ascii::Char::Null; N], alive: 0..0 } |
| 184 | } |
| 185 | |
| 186 | #[inline ] |
| 187 | pub(crate) fn as_ascii(&self) -> &[ascii::Char] { |
| 188 | // SAFETY: `self.alive` is guaranteed to be a valid range for indexing `self.data`. |
| 189 | unsafe { |
| 190 | self.data.get_unchecked(usize::from(self.alive.start)..usize::from(self.alive.end)) |
| 191 | } |
| 192 | } |
| 193 | |
| 194 | #[inline ] |
| 195 | pub(crate) fn as_str(&self) -> &str { |
| 196 | self.as_ascii().as_str() |
| 197 | } |
| 198 | |
| 199 | #[inline ] |
| 200 | pub(crate) fn len(&self) -> usize { |
| 201 | usize::from(self.alive.end - self.alive.start) |
| 202 | } |
| 203 | |
| 204 | pub(crate) fn next(&mut self) -> Option<u8> { |
| 205 | let i = self.alive.next()?; |
| 206 | |
| 207 | // SAFETY: `i` is guaranteed to be a valid index for `self.data`. |
| 208 | unsafe { Some(self.data.get_unchecked(usize::from(i)).to_u8()) } |
| 209 | } |
| 210 | |
| 211 | pub(crate) fn next_back(&mut self) -> Option<u8> { |
| 212 | let i = self.alive.next_back()?; |
| 213 | |
| 214 | // SAFETY: `i` is guaranteed to be a valid index for `self.data`. |
| 215 | unsafe { Some(self.data.get_unchecked(usize::from(i)).to_u8()) } |
| 216 | } |
| 217 | |
| 218 | pub(crate) fn advance_by(&mut self, n: usize) -> Result<(), NonZero<usize>> { |
| 219 | self.alive.advance_by(n) |
| 220 | } |
| 221 | |
| 222 | pub(crate) fn advance_back_by(&mut self, n: usize) -> Result<(), NonZero<usize>> { |
| 223 | self.alive.advance_back_by(n) |
| 224 | } |
| 225 | } |
| 226 | |