| 1 | use std::borrow::{Borrow, Cow}; |
| 2 | use std::fmt; |
| 3 | use std::hash::{Hash, Hasher}; |
| 4 | use std::ops::Deref; |
| 5 | use std::str::from_utf8; |
| 6 | |
| 7 | const MAX_INLINE_STR_LEN: usize = 3 * std::mem::size_of::<isize>() - 2; |
| 8 | |
| 9 | /// Returned when trying to convert a `&str` into a `InlineStr` |
| 10 | /// but it fails because it doesn't fit. |
| 11 | #[derive (Debug)] |
| 12 | pub struct StringTooLongError; |
| 13 | |
| 14 | /// An inline string that can contain almost three words |
| 15 | /// of utf-8 text. |
| 16 | #[derive (Debug, Clone, Copy, Eq)] |
| 17 | pub struct InlineStr { |
| 18 | inner: [u8; MAX_INLINE_STR_LEN], |
| 19 | len: u8, |
| 20 | } |
| 21 | |
| 22 | impl AsRef<str> for InlineStr { |
| 23 | fn as_ref(&self) -> &str { |
| 24 | self.deref() |
| 25 | } |
| 26 | } |
| 27 | |
| 28 | impl Hash for InlineStr { |
| 29 | fn hash<H: Hasher>(&self, state: &mut H) { |
| 30 | self.deref().hash(state); |
| 31 | } |
| 32 | } |
| 33 | |
| 34 | impl From<char> for InlineStr { |
| 35 | fn from(c: char) -> Self { |
| 36 | let mut inner: [u8; 22] = [0u8; MAX_INLINE_STR_LEN]; |
| 37 | c.encode_utf8(&mut inner); |
| 38 | let len: u8 = c.len_utf8() as u8; |
| 39 | Self { inner, len } |
| 40 | } |
| 41 | } |
| 42 | |
| 43 | impl std::cmp::PartialEq<InlineStr> for InlineStr { |
| 44 | fn eq(&self, other: &InlineStr) -> bool { |
| 45 | self.deref() == other.deref() |
| 46 | } |
| 47 | } |
| 48 | |
| 49 | impl TryFrom<&str> for InlineStr { |
| 50 | type Error = StringTooLongError; |
| 51 | |
| 52 | fn try_from(s: &str) -> Result<InlineStr, StringTooLongError> { |
| 53 | let len: usize = s.len(); |
| 54 | if len <= MAX_INLINE_STR_LEN { |
| 55 | let mut inner: [u8; 22] = [0u8; MAX_INLINE_STR_LEN]; |
| 56 | inner[..len].copy_from_slice(src:s.as_bytes()); |
| 57 | let len: u8 = len as u8; |
| 58 | Ok(Self { inner, len }) |
| 59 | } else { |
| 60 | Err(StringTooLongError) |
| 61 | } |
| 62 | } |
| 63 | } |
| 64 | |
| 65 | impl Deref for InlineStr { |
| 66 | type Target = str; |
| 67 | |
| 68 | fn deref(&self) -> &str { |
| 69 | let len: usize = self.len as usize; |
| 70 | from_utf8(&self.inner[..len]).unwrap() |
| 71 | } |
| 72 | } |
| 73 | |
| 74 | impl fmt::Display for InlineStr { |
| 75 | fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { |
| 76 | write!(f, " {}" , self.as_ref()) |
| 77 | } |
| 78 | } |
| 79 | |
| 80 | /// A copy-on-write string that can be owned, borrowed |
| 81 | /// or inlined. |
| 82 | /// |
| 83 | /// It is three words long. |
| 84 | #[derive (Debug, Eq)] |
| 85 | pub enum CowStr<'a> { |
| 86 | /// An owned, immutable string. |
| 87 | Boxed(Box<str>), |
| 88 | /// A borrowed string. |
| 89 | Borrowed(&'a str), |
| 90 | /// A short inline string. |
| 91 | Inlined(InlineStr), |
| 92 | } |
| 93 | |
| 94 | #[cfg (feature = "serde" )] |
| 95 | mod serde_impl { |
| 96 | use super::CowStr; |
| 97 | use serde::{de, Deserialize, Deserializer, Serialize, Serializer}; |
| 98 | use std::fmt; |
| 99 | |
| 100 | impl<'a> Serialize for CowStr<'a> { |
| 101 | fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error> |
| 102 | where |
| 103 | S: Serializer, |
| 104 | { |
| 105 | serializer.serialize_str(self.as_ref()) |
| 106 | } |
| 107 | } |
| 108 | |
| 109 | struct CowStrVisitor; |
| 110 | |
| 111 | impl<'de> de::Visitor<'de> for CowStrVisitor { |
| 112 | type Value = CowStr<'de>; |
| 113 | |
| 114 | fn expecting(&self, formatter: &mut fmt::Formatter) -> fmt::Result { |
| 115 | formatter.write_str("a string" ) |
| 116 | } |
| 117 | |
| 118 | fn visit_borrowed_str<E>(self, v: &'de str) -> Result<Self::Value, E> |
| 119 | where |
| 120 | E: de::Error, |
| 121 | { |
| 122 | Ok(CowStr::Borrowed(v)) |
| 123 | } |
| 124 | |
| 125 | fn visit_str<E>(self, v: &str) -> Result<Self::Value, E> |
| 126 | where |
| 127 | E: de::Error, |
| 128 | { |
| 129 | match v.try_into() { |
| 130 | Ok(it) => Ok(CowStr::Inlined(it)), |
| 131 | Err(_) => Ok(CowStr::Boxed(String::from(v).into_boxed_str())), |
| 132 | } |
| 133 | } |
| 134 | |
| 135 | fn visit_string<E>(self, v: String) -> Result<Self::Value, E> |
| 136 | where |
| 137 | E: de::Error, |
| 138 | { |
| 139 | Ok(CowStr::Boxed(v.into_boxed_str())) |
| 140 | } |
| 141 | } |
| 142 | |
| 143 | impl<'a, 'de: 'a> Deserialize<'de> for CowStr<'a> { |
| 144 | fn deserialize<D>(deserializer: D) -> Result<Self, D::Error> |
| 145 | where |
| 146 | D: Deserializer<'de>, |
| 147 | { |
| 148 | deserializer.deserialize_str(CowStrVisitor) |
| 149 | } |
| 150 | } |
| 151 | } |
| 152 | |
| 153 | impl<'a> AsRef<str> for CowStr<'a> { |
| 154 | fn as_ref(&self) -> &str { |
| 155 | self.deref() |
| 156 | } |
| 157 | } |
| 158 | |
| 159 | impl<'a> Hash for CowStr<'a> { |
| 160 | fn hash<H: Hasher>(&self, state: &mut H) { |
| 161 | self.deref().hash(state); |
| 162 | } |
| 163 | } |
| 164 | |
| 165 | impl<'a> std::clone::Clone for CowStr<'a> { |
| 166 | fn clone(&self) -> Self { |
| 167 | match self { |
| 168 | CowStr::Boxed(s: &Box) => match InlineStr::try_from(&**s) { |
| 169 | Ok(inline: InlineStr) => CowStr::Inlined(inline), |
| 170 | Err(..) => CowStr::Boxed(s.clone()), |
| 171 | }, |
| 172 | CowStr::Borrowed(s: &&str) => CowStr::Borrowed(s), |
| 173 | CowStr::Inlined(s: &InlineStr) => CowStr::Inlined(*s), |
| 174 | } |
| 175 | } |
| 176 | } |
| 177 | |
| 178 | impl<'a> std::cmp::PartialEq<CowStr<'a>> for CowStr<'a> { |
| 179 | fn eq(&self, other: &CowStr<'_>) -> bool { |
| 180 | self.deref() == other.deref() |
| 181 | } |
| 182 | } |
| 183 | |
| 184 | impl<'a> From<&'a str> for CowStr<'a> { |
| 185 | fn from(s: &'a str) -> Self { |
| 186 | CowStr::Borrowed(s) |
| 187 | } |
| 188 | } |
| 189 | |
| 190 | impl<'a> From<String> for CowStr<'a> { |
| 191 | fn from(s: String) -> Self { |
| 192 | CowStr::Boxed(s.into_boxed_str()) |
| 193 | } |
| 194 | } |
| 195 | |
| 196 | impl<'a> From<char> for CowStr<'a> { |
| 197 | fn from(c: char) -> Self { |
| 198 | CowStr::Inlined(c.into()) |
| 199 | } |
| 200 | } |
| 201 | |
| 202 | impl<'a> From<Cow<'a, str>> for CowStr<'a> { |
| 203 | fn from(s: Cow<'a, str>) -> Self { |
| 204 | match s { |
| 205 | Cow::Borrowed(s: &str) => CowStr::Borrowed(s), |
| 206 | Cow::Owned(s: String) => CowStr::Boxed(s.into_boxed_str()), |
| 207 | } |
| 208 | } |
| 209 | } |
| 210 | |
| 211 | impl<'a> From<CowStr<'a>> for Cow<'a, str> { |
| 212 | fn from(s: CowStr<'a>) -> Self { |
| 213 | match s { |
| 214 | CowStr::Boxed(s: Box) => Cow::Owned(s.to_string()), |
| 215 | CowStr::Inlined(s: InlineStr) => Cow::Owned(s.to_string()), |
| 216 | CowStr::Borrowed(s: &str) => Cow::Borrowed(s), |
| 217 | } |
| 218 | } |
| 219 | } |
| 220 | |
| 221 | impl<'a> From<Cow<'a, char>> for CowStr<'a> { |
| 222 | fn from(s: Cow<'a, char>) -> Self { |
| 223 | CowStr::Inlined(InlineStr::from(*s)) |
| 224 | } |
| 225 | } |
| 226 | |
| 227 | impl<'a> Deref for CowStr<'a> { |
| 228 | type Target = str; |
| 229 | |
| 230 | fn deref(&self) -> &str { |
| 231 | match self { |
| 232 | CowStr::Boxed(ref b: &Box) => b, |
| 233 | CowStr::Borrowed(b: &&str) => b, |
| 234 | CowStr::Inlined(ref s: &InlineStr) => s.deref(), |
| 235 | } |
| 236 | } |
| 237 | } |
| 238 | |
| 239 | impl<'a> Borrow<str> for CowStr<'a> { |
| 240 | fn borrow(&self) -> &str { |
| 241 | self.deref() |
| 242 | } |
| 243 | } |
| 244 | |
| 245 | impl<'a> CowStr<'a> { |
| 246 | pub fn into_string(self) -> String { |
| 247 | match self { |
| 248 | CowStr::Boxed(b: Box) => b.into(), |
| 249 | CowStr::Borrowed(b: &str) => b.to_owned(), |
| 250 | CowStr::Inlined(s: InlineStr) => s.deref().to_owned(), |
| 251 | } |
| 252 | } |
| 253 | } |
| 254 | |
| 255 | impl<'a> fmt::Display for CowStr<'a> { |
| 256 | fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { |
| 257 | write!(f, " {}" , self.as_ref()) |
| 258 | } |
| 259 | } |
| 260 | |
| 261 | #[cfg (test)] |
| 262 | mod test_special_string { |
| 263 | use super::*; |
| 264 | |
| 265 | #[test ] |
| 266 | fn inlinestr_ascii() { |
| 267 | let s: InlineStr = 'a' .into(); |
| 268 | assert_eq!("a" , s.deref()); |
| 269 | } |
| 270 | |
| 271 | #[test ] |
| 272 | fn inlinestr_unicode() { |
| 273 | let s: InlineStr = '🍔' .into(); |
| 274 | assert_eq!("🍔" , s.deref()); |
| 275 | } |
| 276 | |
| 277 | #[test ] |
| 278 | fn cowstr_size() { |
| 279 | let size = std::mem::size_of::<CowStr>(); |
| 280 | let word_size = std::mem::size_of::<isize>(); |
| 281 | assert_eq!(3 * word_size, size); |
| 282 | } |
| 283 | |
| 284 | #[test ] |
| 285 | fn cowstr_char_to_string() { |
| 286 | let c = '藏' ; |
| 287 | let smort: CowStr = c.into(); |
| 288 | let owned: String = smort.to_string(); |
| 289 | let expected = "藏" .to_owned(); |
| 290 | assert_eq!(expected, owned); |
| 291 | } |
| 292 | |
| 293 | #[test ] |
| 294 | fn max_inline_str_len_atleast_four() { |
| 295 | // we need 4 bytes to store a char |
| 296 | assert!(MAX_INLINE_STR_LEN >= 4); |
| 297 | } |
| 298 | |
| 299 | #[test ] |
| 300 | #[cfg (target_pointer_width = "64" )] |
| 301 | fn inlinestr_fits_twentytwo() { |
| 302 | let s = "0123456789abcdefghijkl" ; |
| 303 | let stack_str = InlineStr::try_from(s).unwrap(); |
| 304 | assert_eq!(stack_str.deref(), s); |
| 305 | } |
| 306 | |
| 307 | #[test ] |
| 308 | #[cfg (target_pointer_width = "64" )] |
| 309 | fn inlinestr_not_fits_twentythree() { |
| 310 | let s = "0123456789abcdefghijklm" ; |
| 311 | let _stack_str = InlineStr::try_from(s).unwrap_err(); |
| 312 | } |
| 313 | |
| 314 | #[test ] |
| 315 | #[cfg (target_pointer_width = "64" )] |
| 316 | fn small_boxed_str_clones_to_stack() { |
| 317 | let s = "0123456789abcde" .to_owned(); |
| 318 | let smort: CowStr = s.into(); |
| 319 | let smort_clone = smort.clone(); |
| 320 | |
| 321 | if let CowStr::Inlined(..) = smort_clone { |
| 322 | } else { |
| 323 | panic!("Expected a Inlined variant!" ); |
| 324 | } |
| 325 | } |
| 326 | |
| 327 | #[test ] |
| 328 | fn cow_to_cow_str() { |
| 329 | let s = "some text" ; |
| 330 | let cow = Cow::Borrowed(s); |
| 331 | let actual = CowStr::from(cow); |
| 332 | let expected = CowStr::Borrowed(s); |
| 333 | assert_eq!(actual, expected); |
| 334 | assert!(variant_eq(&actual, &expected)); |
| 335 | |
| 336 | let s = "some text" .to_string(); |
| 337 | let cow: Cow<str> = Cow::Owned(s.clone()); |
| 338 | let actual = CowStr::from(cow); |
| 339 | let expected = CowStr::Boxed(s.into_boxed_str()); |
| 340 | assert_eq!(actual, expected); |
| 341 | assert!(variant_eq(&actual, &expected)); |
| 342 | } |
| 343 | |
| 344 | #[test ] |
| 345 | fn cow_str_to_cow() { |
| 346 | let s = "some text" ; |
| 347 | let cow_str = CowStr::Borrowed(s); |
| 348 | let actual = Cow::from(cow_str); |
| 349 | let expected = Cow::Borrowed(s); |
| 350 | assert_eq!(actual, expected); |
| 351 | assert!(variant_eq(&actual, &expected)); |
| 352 | |
| 353 | let s = "s" ; |
| 354 | let inline_str: InlineStr = InlineStr::try_from(s).unwrap(); |
| 355 | let cow_str = CowStr::Inlined(inline_str); |
| 356 | let actual = Cow::from(cow_str); |
| 357 | let expected: Cow<str> = Cow::Owned(s.to_string()); |
| 358 | assert_eq!(actual, expected); |
| 359 | assert!(variant_eq(&actual, &expected)); |
| 360 | |
| 361 | let s = "s" ; |
| 362 | let cow_str = CowStr::Boxed(s.to_string().into_boxed_str()); |
| 363 | let actual = Cow::from(cow_str); |
| 364 | let expected: Cow<str> = Cow::Owned(s.to_string()); |
| 365 | assert_eq!(actual, expected); |
| 366 | assert!(variant_eq(&actual, &expected)); |
| 367 | } |
| 368 | |
| 369 | #[test ] |
| 370 | fn cow_char_to_cow_str() { |
| 371 | let c = 'c' ; |
| 372 | let cow: Cow<char> = Cow::Owned(c); |
| 373 | let actual = CowStr::from(cow); |
| 374 | let expected = CowStr::Inlined(InlineStr::from(c)); |
| 375 | assert_eq!(actual, expected); |
| 376 | assert!(variant_eq(&actual, &expected)); |
| 377 | |
| 378 | let c = 'c' ; |
| 379 | let cow: Cow<char> = Cow::Borrowed(&c); |
| 380 | let actual = CowStr::from(cow); |
| 381 | let expected = CowStr::Inlined(InlineStr::from(c)); |
| 382 | assert_eq!(actual, expected); |
| 383 | assert!(variant_eq(&actual, &expected)); |
| 384 | } |
| 385 | |
| 386 | fn variant_eq<T>(a: &T, b: &T) -> bool { |
| 387 | std::mem::discriminant(a) == std::mem::discriminant(b) |
| 388 | } |
| 389 | } |
| 390 | |