| 1 | use std::ffi::OsStr; |
| 2 | |
| 3 | pub trait OsStrExt: private::Sealed { |
| 4 | /// Converts to a string slice. |
| 5 | fn try_str(&self) -> Result<&str, std::str::Utf8Error>; |
| 6 | /// Returns `true` if the given pattern matches a sub-slice of |
| 7 | /// this string slice. |
| 8 | /// |
| 9 | /// Returns `false` if it does not. |
| 10 | /// |
| 11 | /// # Examples |
| 12 | /// |
| 13 | /// ```rust |
| 14 | /// use clap_lex::OsStrExt as _; |
| 15 | /// let bananas = std::ffi::OsStr::new("bananas" ); |
| 16 | /// |
| 17 | /// assert!(bananas.contains("nana" )); |
| 18 | /// assert!(!bananas.contains("apples" )); |
| 19 | /// ``` |
| 20 | fn contains(&self, needle: &str) -> bool; |
| 21 | /// Returns the byte index of the first character of this string slice that |
| 22 | /// matches the pattern. |
| 23 | /// |
| 24 | /// Returns [`None`] if the pattern doesn't match. |
| 25 | /// |
| 26 | /// # Examples |
| 27 | /// |
| 28 | /// ```rust |
| 29 | /// use clap_lex::OsStrExt as _; |
| 30 | /// let s = std::ffi::OsStr::new("Löwe 老虎 Léopard Gepardi" ); |
| 31 | /// |
| 32 | /// assert_eq!(s.find("L" ), Some(0)); |
| 33 | /// assert_eq!(s.find("é" ), Some(14)); |
| 34 | /// assert_eq!(s.find("par" ), Some(17)); |
| 35 | /// ``` |
| 36 | /// |
| 37 | /// Not finding the pattern: |
| 38 | /// |
| 39 | /// ```rust |
| 40 | /// use clap_lex::OsStrExt as _; |
| 41 | /// let s = std::ffi::OsStr::new("Löwe 老虎 Léopard" ); |
| 42 | /// |
| 43 | /// assert_eq!(s.find("1" ), None); |
| 44 | /// ``` |
| 45 | fn find(&self, needle: &str) -> Option<usize>; |
| 46 | /// Returns a string slice with the prefix removed. |
| 47 | /// |
| 48 | /// If the string starts with the pattern `prefix`, returns substring after the prefix, wrapped |
| 49 | /// in `Some`. |
| 50 | /// |
| 51 | /// If the string does not start with `prefix`, returns `None`. |
| 52 | /// |
| 53 | /// # Examples |
| 54 | /// |
| 55 | /// ``` |
| 56 | /// use std::ffi::OsStr; |
| 57 | /// use clap_lex::OsStrExt as _; |
| 58 | /// assert_eq!(OsStr::new("foo:bar" ).strip_prefix("foo:" ), Some(OsStr::new("bar" ))); |
| 59 | /// assert_eq!(OsStr::new("foo:bar" ).strip_prefix("bar" ), None); |
| 60 | /// assert_eq!(OsStr::new("foofoo" ).strip_prefix("foo" ), Some(OsStr::new("foo" ))); |
| 61 | /// ``` |
| 62 | fn strip_prefix(&self, prefix: &str) -> Option<&OsStr>; |
| 63 | /// Returns `true` if the given pattern matches a prefix of this |
| 64 | /// string slice. |
| 65 | /// |
| 66 | /// Returns `false` if it does not. |
| 67 | /// |
| 68 | /// # Examples |
| 69 | /// |
| 70 | /// ``` |
| 71 | /// use clap_lex::OsStrExt as _; |
| 72 | /// let bananas = std::ffi::OsStr::new("bananas" ); |
| 73 | /// |
| 74 | /// assert!(bananas.starts_with("bana" )); |
| 75 | /// assert!(!bananas.starts_with("nana" )); |
| 76 | /// ``` |
| 77 | fn starts_with(&self, prefix: &str) -> bool; |
| 78 | /// An iterator over substrings of this string slice, separated by |
| 79 | /// characters matched by a pattern. |
| 80 | /// |
| 81 | /// # Examples |
| 82 | /// |
| 83 | /// Simple patterns: |
| 84 | /// |
| 85 | /// ``` |
| 86 | /// use std::ffi::OsStr; |
| 87 | /// use clap_lex::OsStrExt as _; |
| 88 | /// let v: Vec<_> = OsStr::new("Mary had a little lamb" ).split(" " ).collect(); |
| 89 | /// assert_eq!(v, [OsStr::new("Mary" ), OsStr::new("had" ), OsStr::new("a" ), OsStr::new("little" ), OsStr::new("lamb" )]); |
| 90 | /// |
| 91 | /// let v: Vec<_> = OsStr::new("" ).split("X" ).collect(); |
| 92 | /// assert_eq!(v, [OsStr::new("" )]); |
| 93 | /// |
| 94 | /// let v: Vec<_> = OsStr::new("lionXXtigerXleopard" ).split("X" ).collect(); |
| 95 | /// assert_eq!(v, [OsStr::new("lion" ), OsStr::new("" ), OsStr::new("tiger" ), OsStr::new("leopard" )]); |
| 96 | /// |
| 97 | /// let v: Vec<_> = OsStr::new("lion::tiger::leopard" ).split("::" ).collect(); |
| 98 | /// assert_eq!(v, [OsStr::new("lion" ), OsStr::new("tiger" ), OsStr::new("leopard" )]); |
| 99 | /// ``` |
| 100 | /// |
| 101 | /// If a string contains multiple contiguous separators, you will end up |
| 102 | /// with empty strings in the output: |
| 103 | /// |
| 104 | /// ``` |
| 105 | /// use std::ffi::OsStr; |
| 106 | /// use clap_lex::OsStrExt as _; |
| 107 | /// let x = OsStr::new("||||a||b|c" ); |
| 108 | /// let d: Vec<_> = x.split("|" ).collect(); |
| 109 | /// |
| 110 | /// assert_eq!(d, &[OsStr::new("" ), OsStr::new("" ), OsStr::new("" ), OsStr::new("" ), OsStr::new("a" ), OsStr::new("" ), OsStr::new("b" ), OsStr::new("c" )]); |
| 111 | /// ``` |
| 112 | /// |
| 113 | /// Contiguous separators are separated by the empty string. |
| 114 | /// |
| 115 | /// ``` |
| 116 | /// use std::ffi::OsStr; |
| 117 | /// use clap_lex::OsStrExt as _; |
| 118 | /// let x = OsStr::new("(///)" ); |
| 119 | /// let d: Vec<_> = x.split("/" ).collect(); |
| 120 | /// |
| 121 | /// assert_eq!(d, &[OsStr::new("(" ), OsStr::new("" ), OsStr::new("" ), OsStr::new(")" )]); |
| 122 | /// ``` |
| 123 | /// |
| 124 | /// Separators at the start or end of a string are neighbored |
| 125 | /// by empty strings. |
| 126 | /// |
| 127 | /// ``` |
| 128 | /// use std::ffi::OsStr; |
| 129 | /// use clap_lex::OsStrExt as _; |
| 130 | /// let d: Vec<_> = OsStr::new("010" ).split("0" ).collect(); |
| 131 | /// assert_eq!(d, &[OsStr::new("" ), OsStr::new("1" ), OsStr::new("" )]); |
| 132 | /// ``` |
| 133 | /// |
| 134 | /// When the empty string is used as a separator, it panics |
| 135 | /// |
| 136 | /// ```should_panic |
| 137 | /// use std::ffi::OsStr; |
| 138 | /// use clap_lex::OsStrExt as _; |
| 139 | /// let f: Vec<_> = OsStr::new("rust" ).split("" ).collect(); |
| 140 | /// assert_eq!(f, &[OsStr::new("" ), OsStr::new("r" ), OsStr::new("u" ), OsStr::new("s" ), OsStr::new("t" ), OsStr::new("" )]); |
| 141 | /// ``` |
| 142 | /// |
| 143 | /// Contiguous separators can lead to possibly surprising behavior |
| 144 | /// when whitespace is used as the separator. This code is correct: |
| 145 | /// |
| 146 | /// ``` |
| 147 | /// use std::ffi::OsStr; |
| 148 | /// use clap_lex::OsStrExt as _; |
| 149 | /// let x = OsStr::new(" a b c" ); |
| 150 | /// let d: Vec<_> = x.split(" " ).collect(); |
| 151 | /// |
| 152 | /// assert_eq!(d, &[OsStr::new("" ), OsStr::new("" ), OsStr::new("" ), OsStr::new("" ), OsStr::new("a" ), OsStr::new("" ), OsStr::new("b" ), OsStr::new("c" )]); |
| 153 | /// ``` |
| 154 | /// |
| 155 | /// It does _not_ give you: |
| 156 | /// |
| 157 | /// ```,ignore |
| 158 | /// assert_eq!(d, &[OsStr::new("a" ), OsStr::new("b" ), OsStr::new("c" )]); |
| 159 | /// ``` |
| 160 | /// |
| 161 | /// Use [`split_whitespace`] for this behavior. |
| 162 | /// |
| 163 | /// [`split_whitespace`]: str::split_whitespace |
| 164 | fn split<'s, 'n>(&'s self, needle: &'n str) -> Split<'s, 'n>; |
| 165 | /// Splits the string on the first occurrence of the specified delimiter and |
| 166 | /// returns prefix before delimiter and suffix after delimiter. |
| 167 | /// |
| 168 | /// # Examples |
| 169 | /// |
| 170 | /// ``` |
| 171 | /// use std::ffi::OsStr; |
| 172 | /// use clap_lex::OsStrExt as _; |
| 173 | /// assert_eq!(OsStr::new("cfg" ).split_once("=" ), None); |
| 174 | /// assert_eq!(OsStr::new("cfg=" ).split_once("=" ), Some((OsStr::new("cfg" ), OsStr::new("" )))); |
| 175 | /// assert_eq!(OsStr::new("cfg=foo" ).split_once("=" ), Some((OsStr::new("cfg" ), OsStr::new("foo" )))); |
| 176 | /// assert_eq!(OsStr::new("cfg=foo=bar" ).split_once("=" ), Some((OsStr::new("cfg" ), OsStr::new("foo=bar" )))); |
| 177 | /// ``` |
| 178 | fn split_once(&self, needle: &'_ str) -> Option<(&OsStr, &OsStr)>; |
| 179 | } |
| 180 | |
| 181 | impl OsStrExt for OsStr { |
| 182 | fn try_str(&self) -> Result<&str, std::str::Utf8Error> { |
| 183 | let bytes = to_bytes(self); |
| 184 | std::str::from_utf8(bytes) |
| 185 | } |
| 186 | |
| 187 | fn contains(&self, needle: &str) -> bool { |
| 188 | self.find(needle).is_some() |
| 189 | } |
| 190 | |
| 191 | fn find(&self, needle: &str) -> Option<usize> { |
| 192 | let bytes = to_bytes(self); |
| 193 | (0..=self.len().checked_sub(needle.len())?) |
| 194 | .find(|&x| bytes[x..].starts_with(needle.as_bytes())) |
| 195 | } |
| 196 | |
| 197 | fn strip_prefix(&self, prefix: &str) -> Option<&OsStr> { |
| 198 | let bytes = to_bytes(self); |
| 199 | bytes.strip_prefix(prefix.as_bytes()).map(|s| { |
| 200 | // SAFETY: |
| 201 | // - This came from `to_bytes` |
| 202 | // - Since `prefix` is `&str`, any split will be along UTF-8 boundarie |
| 203 | unsafe { to_os_str_unchecked(s) } |
| 204 | }) |
| 205 | } |
| 206 | fn starts_with(&self, prefix: &str) -> bool { |
| 207 | let bytes = to_bytes(self); |
| 208 | bytes.starts_with(prefix.as_bytes()) |
| 209 | } |
| 210 | |
| 211 | fn split<'s, 'n>(&'s self, needle: &'n str) -> Split<'s, 'n> { |
| 212 | assert_ne!(needle, "" ); |
| 213 | Split { |
| 214 | haystack: Some(self), |
| 215 | needle, |
| 216 | } |
| 217 | } |
| 218 | |
| 219 | fn split_once(&self, needle: &'_ str) -> Option<(&OsStr, &OsStr)> { |
| 220 | let start = self.find(needle)?; |
| 221 | let end = start + needle.len(); |
| 222 | let haystack = to_bytes(self); |
| 223 | let first = &haystack[0..start]; |
| 224 | let second = &haystack[end..]; |
| 225 | // SAFETY: |
| 226 | // - This came from `to_bytes` |
| 227 | // - Since `needle` is `&str`, any split will be along UTF-8 boundarie |
| 228 | unsafe { Some((to_os_str_unchecked(first), to_os_str_unchecked(second))) } |
| 229 | } |
| 230 | } |
| 231 | |
| 232 | mod private { |
| 233 | pub trait Sealed {} |
| 234 | |
| 235 | impl Sealed for std::ffi::OsStr {} |
| 236 | } |
| 237 | |
| 238 | /// Allow access to raw bytes |
| 239 | /// |
| 240 | /// As the non-UTF8 encoding is not defined, the bytes only make sense when compared with |
| 241 | /// 7-bit ASCII or `&str` |
| 242 | /// |
| 243 | /// # Compatibility |
| 244 | /// |
| 245 | /// There is no guarantee how non-UTF8 bytes will be encoded, even within versions of this crate |
| 246 | /// (since its dependent on rustc) |
| 247 | fn to_bytes(s: &OsStr) -> &[u8] { |
| 248 | // SAFETY: |
| 249 | // - Lifetimes are the same |
| 250 | // - Types are compatible (`OsStr` is effectively a transparent wrapper for `[u8]`) |
| 251 | // - The primary contract is that the encoding for invalid surrogate code points is not |
| 252 | // guaranteed which isn't a problem here |
| 253 | // |
| 254 | // There is a proposal to support this natively (https://github.com/rust-lang/rust/pull/95290) |
| 255 | // but its in limbo |
| 256 | unsafe { std::mem::transmute(s) } |
| 257 | } |
| 258 | |
| 259 | /// Restore raw bytes as `OsStr` |
| 260 | /// |
| 261 | /// # Safety |
| 262 | /// |
| 263 | /// - `&[u8]` must either by a `&str` or originated with `to_bytes` within the same binary |
| 264 | /// - Any splits of the original `&[u8]` must be done along UTF-8 boundaries |
| 265 | unsafe fn to_os_str_unchecked(s: &[u8]) -> &OsStr { |
| 266 | // SAFETY: |
| 267 | // - Lifetimes are the same |
| 268 | // - Types are compatible (`OsStr` is effectively a transparent wrapper for `[u8]`) |
| 269 | // - The primary contract is that the encoding for invalid surrogate code points is not |
| 270 | // guaranteed which isn't a problem here |
| 271 | // |
| 272 | // There is a proposal to support this natively (https://github.com/rust-lang/rust/pull/95290) |
| 273 | // but its in limbo |
| 274 | std::mem::transmute(s) |
| 275 | } |
| 276 | |
| 277 | pub struct Split<'s, 'n> { |
| 278 | haystack: Option<&'s OsStr>, |
| 279 | needle: &'n str, |
| 280 | } |
| 281 | |
| 282 | impl<'s, 'n> Iterator for Split<'s, 'n> { |
| 283 | type Item = &'s OsStr; |
| 284 | |
| 285 | fn next(&mut self) -> Option<Self::Item> { |
| 286 | let haystack = self.haystack?; |
| 287 | match haystack.split_once(self.needle) { |
| 288 | Some((first, second)) => { |
| 289 | if !haystack.is_empty() { |
| 290 | debug_assert_ne!(haystack, second); |
| 291 | } |
| 292 | self.haystack = Some(second); |
| 293 | Some(first) |
| 294 | } |
| 295 | None => { |
| 296 | self.haystack = None; |
| 297 | Some(haystack) |
| 298 | } |
| 299 | } |
| 300 | } |
| 301 | } |
| 302 | |
| 303 | /// Split an `OsStr` |
| 304 | /// |
| 305 | /// # Safety |
| 306 | /// |
| 307 | /// `index` must be at a valid UTF-8 boundary |
| 308 | pub(crate) unsafe fn split_at(os: &OsStr, index: usize) -> (&OsStr, &OsStr) { |
| 309 | let bytes = to_bytes(os); |
| 310 | let (first, second) = bytes.split_at(index); |
| 311 | (to_os_str_unchecked(first), to_os_str_unchecked(second)) |
| 312 | } |
| 313 | |