| 1 | use crate::error::{Error, ErrorCode, Result}; |
| 2 | use alloc::vec::Vec; |
| 3 | use core::char; |
| 4 | use core::cmp; |
| 5 | use core::ops::Deref; |
| 6 | use core::str; |
| 7 | |
| 8 | #[cfg (feature = "std" )] |
| 9 | use crate::io; |
| 10 | #[cfg (feature = "std" )] |
| 11 | use crate::iter::LineColIterator; |
| 12 | |
| 13 | #[cfg (feature = "raw_value" )] |
| 14 | use crate::raw::BorrowedRawDeserializer; |
| 15 | #[cfg (all(feature = "raw_value" , feature = "std" ))] |
| 16 | use crate::raw::OwnedRawDeserializer; |
| 17 | #[cfg (feature = "raw_value" )] |
| 18 | use serde::de::Visitor; |
| 19 | |
| 20 | /// Trait used by the deserializer for iterating over input. This is manually |
| 21 | /// "specialized" for iterating over &[u8]. Once feature(specialization) is |
| 22 | /// stable we can use actual specialization. |
| 23 | /// |
| 24 | /// This trait is sealed and cannot be implemented for types outside of |
| 25 | /// `serde_json`. |
| 26 | pub trait Read<'de>: private::Sealed { |
| 27 | #[doc (hidden)] |
| 28 | fn next(&mut self) -> Result<Option<u8>>; |
| 29 | #[doc (hidden)] |
| 30 | fn peek(&mut self) -> Result<Option<u8>>; |
| 31 | |
| 32 | /// Only valid after a call to peek(). Discards the peeked byte. |
| 33 | #[doc (hidden)] |
| 34 | fn discard(&mut self); |
| 35 | |
| 36 | /// Position of the most recent call to next(). |
| 37 | /// |
| 38 | /// The most recent call was probably next() and not peek(), but this method |
| 39 | /// should try to return a sensible result if the most recent call was |
| 40 | /// actually peek() because we don't always know. |
| 41 | /// |
| 42 | /// Only called in case of an error, so performance is not important. |
| 43 | #[doc (hidden)] |
| 44 | fn position(&self) -> Position; |
| 45 | |
| 46 | /// Position of the most recent call to peek(). |
| 47 | /// |
| 48 | /// The most recent call was probably peek() and not next(), but this method |
| 49 | /// should try to return a sensible result if the most recent call was |
| 50 | /// actually next() because we don't always know. |
| 51 | /// |
| 52 | /// Only called in case of an error, so performance is not important. |
| 53 | #[doc (hidden)] |
| 54 | fn peek_position(&self) -> Position; |
| 55 | |
| 56 | /// Offset from the beginning of the input to the next byte that would be |
| 57 | /// returned by next() or peek(). |
| 58 | #[doc (hidden)] |
| 59 | fn byte_offset(&self) -> usize; |
| 60 | |
| 61 | /// Assumes the previous byte was a quotation mark. Parses a JSON-escaped |
| 62 | /// string until the next quotation mark using the given scratch space if |
| 63 | /// necessary. The scratch space is initially empty. |
| 64 | #[doc (hidden)] |
| 65 | fn parse_str<'s>(&'s mut self, scratch: &'s mut Vec<u8>) -> Result<Reference<'de, 's, str>>; |
| 66 | |
| 67 | /// Assumes the previous byte was a quotation mark. Parses a JSON-escaped |
| 68 | /// string until the next quotation mark using the given scratch space if |
| 69 | /// necessary. The scratch space is initially empty. |
| 70 | /// |
| 71 | /// This function returns the raw bytes in the string with escape sequences |
| 72 | /// expanded but without performing unicode validation. |
| 73 | #[doc (hidden)] |
| 74 | fn parse_str_raw<'s>( |
| 75 | &'s mut self, |
| 76 | scratch: &'s mut Vec<u8>, |
| 77 | ) -> Result<Reference<'de, 's, [u8]>>; |
| 78 | |
| 79 | /// Assumes the previous byte was a quotation mark. Parses a JSON-escaped |
| 80 | /// string until the next quotation mark but discards the data. |
| 81 | #[doc (hidden)] |
| 82 | fn ignore_str(&mut self) -> Result<()>; |
| 83 | |
| 84 | /// Assumes the previous byte was a hex escape sequence ('\u') in a string. |
| 85 | /// Parses next hexadecimal sequence. |
| 86 | #[doc (hidden)] |
| 87 | fn decode_hex_escape(&mut self) -> Result<u16>; |
| 88 | |
| 89 | /// Switch raw buffering mode on. |
| 90 | /// |
| 91 | /// This is used when deserializing `RawValue`. |
| 92 | #[cfg (feature = "raw_value" )] |
| 93 | #[doc (hidden)] |
| 94 | fn begin_raw_buffering(&mut self); |
| 95 | |
| 96 | /// Switch raw buffering mode off and provides the raw buffered data to the |
| 97 | /// given visitor. |
| 98 | #[cfg (feature = "raw_value" )] |
| 99 | #[doc (hidden)] |
| 100 | fn end_raw_buffering<V>(&mut self, visitor: V) -> Result<V::Value> |
| 101 | where |
| 102 | V: Visitor<'de>; |
| 103 | |
| 104 | /// Whether StreamDeserializer::next needs to check the failed flag. True |
| 105 | /// for IoRead, false for StrRead and SliceRead which can track failure by |
| 106 | /// truncating their input slice to avoid the extra check on every next |
| 107 | /// call. |
| 108 | #[doc (hidden)] |
| 109 | const should_early_return_if_failed: bool; |
| 110 | |
| 111 | /// Mark a persistent failure of StreamDeserializer, either by setting the |
| 112 | /// flag or by truncating the input data. |
| 113 | #[doc (hidden)] |
| 114 | fn set_failed(&mut self, failed: &mut bool); |
| 115 | } |
| 116 | |
| 117 | pub struct Position { |
| 118 | pub line: usize, |
| 119 | pub column: usize, |
| 120 | } |
| 121 | |
| 122 | pub enum Reference<'b, 'c, T> |
| 123 | where |
| 124 | T: ?Sized + 'static, |
| 125 | { |
| 126 | Borrowed(&'b T), |
| 127 | Copied(&'c T), |
| 128 | } |
| 129 | |
| 130 | impl<'b, 'c, T> Deref for Reference<'b, 'c, T> |
| 131 | where |
| 132 | T: ?Sized + 'static, |
| 133 | { |
| 134 | type Target = T; |
| 135 | |
| 136 | fn deref(&self) -> &Self::Target { |
| 137 | match *self { |
| 138 | Reference::Borrowed(b) => b, |
| 139 | Reference::Copied(c) => c, |
| 140 | } |
| 141 | } |
| 142 | } |
| 143 | |
| 144 | /// JSON input source that reads from a std::io input stream. |
| 145 | #[cfg (feature = "std" )] |
| 146 | #[cfg_attr (docsrs, doc(cfg(feature = "std" )))] |
| 147 | pub struct IoRead<R> |
| 148 | where |
| 149 | R: io::Read, |
| 150 | { |
| 151 | iter: LineColIterator<io::Bytes<R>>, |
| 152 | /// Temporary storage of peeked byte. |
| 153 | ch: Option<u8>, |
| 154 | #[cfg (feature = "raw_value" )] |
| 155 | raw_buffer: Option<Vec<u8>>, |
| 156 | } |
| 157 | |
| 158 | /// JSON input source that reads from a slice of bytes. |
| 159 | // |
| 160 | // This is more efficient than other iterators because peek() can be read-only |
| 161 | // and we can compute line/col position only if an error happens. |
| 162 | pub struct SliceRead<'a> { |
| 163 | slice: &'a [u8], |
| 164 | /// Index of the *next* byte that will be returned by next() or peek(). |
| 165 | index: usize, |
| 166 | #[cfg (feature = "raw_value" )] |
| 167 | raw_buffering_start_index: usize, |
| 168 | } |
| 169 | |
| 170 | /// JSON input source that reads from a UTF-8 string. |
| 171 | // |
| 172 | // Able to elide UTF-8 checks by assuming that the input is valid UTF-8. |
| 173 | pub struct StrRead<'a> { |
| 174 | delegate: SliceRead<'a>, |
| 175 | #[cfg (feature = "raw_value" )] |
| 176 | data: &'a str, |
| 177 | } |
| 178 | |
| 179 | // Prevent users from implementing the Read trait. |
| 180 | mod private { |
| 181 | pub trait Sealed {} |
| 182 | } |
| 183 | |
| 184 | ////////////////////////////////////////////////////////////////////////////// |
| 185 | |
| 186 | #[cfg (feature = "std" )] |
| 187 | impl<R> IoRead<R> |
| 188 | where |
| 189 | R: io::Read, |
| 190 | { |
| 191 | /// Create a JSON input source to read from a std::io input stream. |
| 192 | pub fn new(reader: R) -> Self { |
| 193 | IoRead { |
| 194 | iter: LineColIterator::new(reader.bytes()), |
| 195 | ch: None, |
| 196 | #[cfg (feature = "raw_value" )] |
| 197 | raw_buffer: None, |
| 198 | } |
| 199 | } |
| 200 | } |
| 201 | |
| 202 | #[cfg (feature = "std" )] |
| 203 | impl<R> private::Sealed for IoRead<R> where R: io::Read {} |
| 204 | |
| 205 | #[cfg (feature = "std" )] |
| 206 | impl<R> IoRead<R> |
| 207 | where |
| 208 | R: io::Read, |
| 209 | { |
| 210 | fn parse_str_bytes<'s, T, F>( |
| 211 | &'s mut self, |
| 212 | scratch: &'s mut Vec<u8>, |
| 213 | validate: bool, |
| 214 | result: F, |
| 215 | ) -> Result<T> |
| 216 | where |
| 217 | T: 's, |
| 218 | F: FnOnce(&'s Self, &'s [u8]) -> Result<T>, |
| 219 | { |
| 220 | loop { |
| 221 | let ch = tri!(next_or_eof(self)); |
| 222 | if !ESCAPE[ch as usize] { |
| 223 | scratch.push(ch); |
| 224 | continue; |
| 225 | } |
| 226 | match ch { |
| 227 | b'"' => { |
| 228 | return result(self, scratch); |
| 229 | } |
| 230 | b' \\' => { |
| 231 | tri!(parse_escape(self, validate, scratch)); |
| 232 | } |
| 233 | _ => { |
| 234 | if validate { |
| 235 | return error(self, ErrorCode::ControlCharacterWhileParsingString); |
| 236 | } |
| 237 | scratch.push(ch); |
| 238 | } |
| 239 | } |
| 240 | } |
| 241 | } |
| 242 | } |
| 243 | |
| 244 | #[cfg (feature = "std" )] |
| 245 | impl<'de, R> Read<'de> for IoRead<R> |
| 246 | where |
| 247 | R: io::Read, |
| 248 | { |
| 249 | #[inline ] |
| 250 | fn next(&mut self) -> Result<Option<u8>> { |
| 251 | match self.ch.take() { |
| 252 | Some(ch) => { |
| 253 | #[cfg (feature = "raw_value" )] |
| 254 | { |
| 255 | if let Some(buf) = &mut self.raw_buffer { |
| 256 | buf.push(ch); |
| 257 | } |
| 258 | } |
| 259 | Ok(Some(ch)) |
| 260 | } |
| 261 | None => match self.iter.next() { |
| 262 | Some(Err(err)) => Err(Error::io(err)), |
| 263 | Some(Ok(ch)) => { |
| 264 | #[cfg (feature = "raw_value" )] |
| 265 | { |
| 266 | if let Some(buf) = &mut self.raw_buffer { |
| 267 | buf.push(ch); |
| 268 | } |
| 269 | } |
| 270 | Ok(Some(ch)) |
| 271 | } |
| 272 | None => Ok(None), |
| 273 | }, |
| 274 | } |
| 275 | } |
| 276 | |
| 277 | #[inline ] |
| 278 | fn peek(&mut self) -> Result<Option<u8>> { |
| 279 | match self.ch { |
| 280 | Some(ch) => Ok(Some(ch)), |
| 281 | None => match self.iter.next() { |
| 282 | Some(Err(err)) => Err(Error::io(err)), |
| 283 | Some(Ok(ch)) => { |
| 284 | self.ch = Some(ch); |
| 285 | Ok(self.ch) |
| 286 | } |
| 287 | None => Ok(None), |
| 288 | }, |
| 289 | } |
| 290 | } |
| 291 | |
| 292 | #[cfg (not(feature = "raw_value" ))] |
| 293 | #[inline ] |
| 294 | fn discard(&mut self) { |
| 295 | self.ch = None; |
| 296 | } |
| 297 | |
| 298 | #[cfg (feature = "raw_value" )] |
| 299 | fn discard(&mut self) { |
| 300 | if let Some(ch) = self.ch.take() { |
| 301 | if let Some(buf) = &mut self.raw_buffer { |
| 302 | buf.push(ch); |
| 303 | } |
| 304 | } |
| 305 | } |
| 306 | |
| 307 | fn position(&self) -> Position { |
| 308 | Position { |
| 309 | line: self.iter.line(), |
| 310 | column: self.iter.col(), |
| 311 | } |
| 312 | } |
| 313 | |
| 314 | fn peek_position(&self) -> Position { |
| 315 | // The LineColIterator updates its position during peek() so it has the |
| 316 | // right one here. |
| 317 | self.position() |
| 318 | } |
| 319 | |
| 320 | fn byte_offset(&self) -> usize { |
| 321 | match self.ch { |
| 322 | Some(_) => self.iter.byte_offset() - 1, |
| 323 | None => self.iter.byte_offset(), |
| 324 | } |
| 325 | } |
| 326 | |
| 327 | fn parse_str<'s>(&'s mut self, scratch: &'s mut Vec<u8>) -> Result<Reference<'de, 's, str>> { |
| 328 | self.parse_str_bytes(scratch, true, as_str) |
| 329 | .map(Reference::Copied) |
| 330 | } |
| 331 | |
| 332 | fn parse_str_raw<'s>( |
| 333 | &'s mut self, |
| 334 | scratch: &'s mut Vec<u8>, |
| 335 | ) -> Result<Reference<'de, 's, [u8]>> { |
| 336 | self.parse_str_bytes(scratch, false, |_, bytes| Ok(bytes)) |
| 337 | .map(Reference::Copied) |
| 338 | } |
| 339 | |
| 340 | fn ignore_str(&mut self) -> Result<()> { |
| 341 | loop { |
| 342 | let ch = tri!(next_or_eof(self)); |
| 343 | if !ESCAPE[ch as usize] { |
| 344 | continue; |
| 345 | } |
| 346 | match ch { |
| 347 | b'"' => { |
| 348 | return Ok(()); |
| 349 | } |
| 350 | b' \\' => { |
| 351 | tri!(ignore_escape(self)); |
| 352 | } |
| 353 | _ => { |
| 354 | return error(self, ErrorCode::ControlCharacterWhileParsingString); |
| 355 | } |
| 356 | } |
| 357 | } |
| 358 | } |
| 359 | |
| 360 | fn decode_hex_escape(&mut self) -> Result<u16> { |
| 361 | let mut n = 0; |
| 362 | for _ in 0..4 { |
| 363 | match decode_hex_val(tri!(next_or_eof(self))) { |
| 364 | None => return error(self, ErrorCode::InvalidEscape), |
| 365 | Some(val) => { |
| 366 | n = (n << 4) + val; |
| 367 | } |
| 368 | } |
| 369 | } |
| 370 | Ok(n) |
| 371 | } |
| 372 | |
| 373 | #[cfg (feature = "raw_value" )] |
| 374 | fn begin_raw_buffering(&mut self) { |
| 375 | self.raw_buffer = Some(Vec::new()); |
| 376 | } |
| 377 | |
| 378 | #[cfg (feature = "raw_value" )] |
| 379 | fn end_raw_buffering<V>(&mut self, visitor: V) -> Result<V::Value> |
| 380 | where |
| 381 | V: Visitor<'de>, |
| 382 | { |
| 383 | let raw = self.raw_buffer.take().unwrap(); |
| 384 | let raw = match String::from_utf8(raw) { |
| 385 | Ok(raw) => raw, |
| 386 | Err(_) => return error(self, ErrorCode::InvalidUnicodeCodePoint), |
| 387 | }; |
| 388 | visitor.visit_map(OwnedRawDeserializer { |
| 389 | raw_value: Some(raw), |
| 390 | }) |
| 391 | } |
| 392 | |
| 393 | const should_early_return_if_failed: bool = true; |
| 394 | |
| 395 | #[inline ] |
| 396 | #[cold ] |
| 397 | fn set_failed(&mut self, failed: &mut bool) { |
| 398 | *failed = true; |
| 399 | } |
| 400 | } |
| 401 | |
| 402 | ////////////////////////////////////////////////////////////////////////////// |
| 403 | |
| 404 | impl<'a> SliceRead<'a> { |
| 405 | /// Create a JSON input source to read from a slice of bytes. |
| 406 | pub fn new(slice: &'a [u8]) -> Self { |
| 407 | SliceRead { |
| 408 | slice, |
| 409 | index: 0, |
| 410 | #[cfg (feature = "raw_value" )] |
| 411 | raw_buffering_start_index: 0, |
| 412 | } |
| 413 | } |
| 414 | |
| 415 | fn position_of_index(&self, i: usize) -> Position { |
| 416 | let mut position = Position { line: 1, column: 0 }; |
| 417 | for ch in &self.slice[..i] { |
| 418 | match *ch { |
| 419 | b' \n' => { |
| 420 | position.line += 1; |
| 421 | position.column = 0; |
| 422 | } |
| 423 | _ => { |
| 424 | position.column += 1; |
| 425 | } |
| 426 | } |
| 427 | } |
| 428 | position |
| 429 | } |
| 430 | |
| 431 | /// The big optimization here over IoRead is that if the string contains no |
| 432 | /// backslash escape sequences, the returned &str is a slice of the raw JSON |
| 433 | /// data so we avoid copying into the scratch space. |
| 434 | fn parse_str_bytes<'s, T, F>( |
| 435 | &'s mut self, |
| 436 | scratch: &'s mut Vec<u8>, |
| 437 | validate: bool, |
| 438 | result: F, |
| 439 | ) -> Result<Reference<'a, 's, T>> |
| 440 | where |
| 441 | T: ?Sized + 's, |
| 442 | F: for<'f> FnOnce(&'s Self, &'f [u8]) -> Result<&'f T>, |
| 443 | { |
| 444 | // Index of the first byte not yet copied into the scratch space. |
| 445 | let mut start = self.index; |
| 446 | |
| 447 | loop { |
| 448 | while self.index < self.slice.len() && !ESCAPE[self.slice[self.index] as usize] { |
| 449 | self.index += 1; |
| 450 | } |
| 451 | if self.index == self.slice.len() { |
| 452 | return error(self, ErrorCode::EofWhileParsingString); |
| 453 | } |
| 454 | match self.slice[self.index] { |
| 455 | b'"' => { |
| 456 | if scratch.is_empty() { |
| 457 | // Fast path: return a slice of the raw JSON without any |
| 458 | // copying. |
| 459 | let borrowed = &self.slice[start..self.index]; |
| 460 | self.index += 1; |
| 461 | return result(self, borrowed).map(Reference::Borrowed); |
| 462 | } else { |
| 463 | scratch.extend_from_slice(&self.slice[start..self.index]); |
| 464 | self.index += 1; |
| 465 | return result(self, scratch).map(Reference::Copied); |
| 466 | } |
| 467 | } |
| 468 | b' \\' => { |
| 469 | scratch.extend_from_slice(&self.slice[start..self.index]); |
| 470 | self.index += 1; |
| 471 | tri!(parse_escape(self, validate, scratch)); |
| 472 | start = self.index; |
| 473 | } |
| 474 | _ => { |
| 475 | self.index += 1; |
| 476 | if validate { |
| 477 | return error(self, ErrorCode::ControlCharacterWhileParsingString); |
| 478 | } |
| 479 | } |
| 480 | } |
| 481 | } |
| 482 | } |
| 483 | } |
| 484 | |
| 485 | impl<'a> private::Sealed for SliceRead<'a> {} |
| 486 | |
| 487 | impl<'a> Read<'a> for SliceRead<'a> { |
| 488 | #[inline ] |
| 489 | fn next(&mut self) -> Result<Option<u8>> { |
| 490 | // `Ok(self.slice.get(self.index).map(|ch| { self.index += 1; *ch }))` |
| 491 | // is about 10% slower. |
| 492 | Ok(if self.index < self.slice.len() { |
| 493 | let ch = self.slice[self.index]; |
| 494 | self.index += 1; |
| 495 | Some(ch) |
| 496 | } else { |
| 497 | None |
| 498 | }) |
| 499 | } |
| 500 | |
| 501 | #[inline ] |
| 502 | fn peek(&mut self) -> Result<Option<u8>> { |
| 503 | // `Ok(self.slice.get(self.index).map(|ch| *ch))` is about 10% slower |
| 504 | // for some reason. |
| 505 | Ok(if self.index < self.slice.len() { |
| 506 | Some(self.slice[self.index]) |
| 507 | } else { |
| 508 | None |
| 509 | }) |
| 510 | } |
| 511 | |
| 512 | #[inline ] |
| 513 | fn discard(&mut self) { |
| 514 | self.index += 1; |
| 515 | } |
| 516 | |
| 517 | fn position(&self) -> Position { |
| 518 | self.position_of_index(self.index) |
| 519 | } |
| 520 | |
| 521 | fn peek_position(&self) -> Position { |
| 522 | // Cap it at slice.len() just in case the most recent call was next() |
| 523 | // and it returned the last byte. |
| 524 | self.position_of_index(cmp::min(self.slice.len(), self.index + 1)) |
| 525 | } |
| 526 | |
| 527 | fn byte_offset(&self) -> usize { |
| 528 | self.index |
| 529 | } |
| 530 | |
| 531 | fn parse_str<'s>(&'s mut self, scratch: &'s mut Vec<u8>) -> Result<Reference<'a, 's, str>> { |
| 532 | self.parse_str_bytes(scratch, true, as_str) |
| 533 | } |
| 534 | |
| 535 | fn parse_str_raw<'s>( |
| 536 | &'s mut self, |
| 537 | scratch: &'s mut Vec<u8>, |
| 538 | ) -> Result<Reference<'a, 's, [u8]>> { |
| 539 | self.parse_str_bytes(scratch, false, |_, bytes| Ok(bytes)) |
| 540 | } |
| 541 | |
| 542 | fn ignore_str(&mut self) -> Result<()> { |
| 543 | loop { |
| 544 | while self.index < self.slice.len() && !ESCAPE[self.slice[self.index] as usize] { |
| 545 | self.index += 1; |
| 546 | } |
| 547 | if self.index == self.slice.len() { |
| 548 | return error(self, ErrorCode::EofWhileParsingString); |
| 549 | } |
| 550 | match self.slice[self.index] { |
| 551 | b'"' => { |
| 552 | self.index += 1; |
| 553 | return Ok(()); |
| 554 | } |
| 555 | b' \\' => { |
| 556 | self.index += 1; |
| 557 | tri!(ignore_escape(self)); |
| 558 | } |
| 559 | _ => { |
| 560 | return error(self, ErrorCode::ControlCharacterWhileParsingString); |
| 561 | } |
| 562 | } |
| 563 | } |
| 564 | } |
| 565 | |
| 566 | fn decode_hex_escape(&mut self) -> Result<u16> { |
| 567 | if self.index + 4 > self.slice.len() { |
| 568 | self.index = self.slice.len(); |
| 569 | return error(self, ErrorCode::EofWhileParsingString); |
| 570 | } |
| 571 | |
| 572 | let mut n = 0; |
| 573 | for _ in 0..4 { |
| 574 | let ch = decode_hex_val(self.slice[self.index]); |
| 575 | self.index += 1; |
| 576 | match ch { |
| 577 | None => return error(self, ErrorCode::InvalidEscape), |
| 578 | Some(val) => { |
| 579 | n = (n << 4) + val; |
| 580 | } |
| 581 | } |
| 582 | } |
| 583 | Ok(n) |
| 584 | } |
| 585 | |
| 586 | #[cfg (feature = "raw_value" )] |
| 587 | fn begin_raw_buffering(&mut self) { |
| 588 | self.raw_buffering_start_index = self.index; |
| 589 | } |
| 590 | |
| 591 | #[cfg (feature = "raw_value" )] |
| 592 | fn end_raw_buffering<V>(&mut self, visitor: V) -> Result<V::Value> |
| 593 | where |
| 594 | V: Visitor<'a>, |
| 595 | { |
| 596 | let raw = &self.slice[self.raw_buffering_start_index..self.index]; |
| 597 | let raw = match str::from_utf8(raw) { |
| 598 | Ok(raw) => raw, |
| 599 | Err(_) => return error(self, ErrorCode::InvalidUnicodeCodePoint), |
| 600 | }; |
| 601 | visitor.visit_map(BorrowedRawDeserializer { |
| 602 | raw_value: Some(raw), |
| 603 | }) |
| 604 | } |
| 605 | |
| 606 | const should_early_return_if_failed: bool = false; |
| 607 | |
| 608 | #[inline ] |
| 609 | #[cold ] |
| 610 | fn set_failed(&mut self, _failed: &mut bool) { |
| 611 | self.slice = &self.slice[..self.index]; |
| 612 | } |
| 613 | } |
| 614 | |
| 615 | ////////////////////////////////////////////////////////////////////////////// |
| 616 | |
| 617 | impl<'a> StrRead<'a> { |
| 618 | /// Create a JSON input source to read from a UTF-8 string. |
| 619 | pub fn new(s: &'a str) -> Self { |
| 620 | StrRead { |
| 621 | delegate: SliceRead::new(s.as_bytes()), |
| 622 | #[cfg (feature = "raw_value" )] |
| 623 | data: s, |
| 624 | } |
| 625 | } |
| 626 | } |
| 627 | |
| 628 | impl<'a> private::Sealed for StrRead<'a> {} |
| 629 | |
| 630 | impl<'a> Read<'a> for StrRead<'a> { |
| 631 | #[inline ] |
| 632 | fn next(&mut self) -> Result<Option<u8>> { |
| 633 | self.delegate.next() |
| 634 | } |
| 635 | |
| 636 | #[inline ] |
| 637 | fn peek(&mut self) -> Result<Option<u8>> { |
| 638 | self.delegate.peek() |
| 639 | } |
| 640 | |
| 641 | #[inline ] |
| 642 | fn discard(&mut self) { |
| 643 | self.delegate.discard(); |
| 644 | } |
| 645 | |
| 646 | fn position(&self) -> Position { |
| 647 | self.delegate.position() |
| 648 | } |
| 649 | |
| 650 | fn peek_position(&self) -> Position { |
| 651 | self.delegate.peek_position() |
| 652 | } |
| 653 | |
| 654 | fn byte_offset(&self) -> usize { |
| 655 | self.delegate.byte_offset() |
| 656 | } |
| 657 | |
| 658 | fn parse_str<'s>(&'s mut self, scratch: &'s mut Vec<u8>) -> Result<Reference<'a, 's, str>> { |
| 659 | self.delegate.parse_str_bytes(scratch, true, |_, bytes| { |
| 660 | // The deserialization input came in as &str with a UTF-8 guarantee, |
| 661 | // and the \u-escapes are checked along the way, so don't need to |
| 662 | // check here. |
| 663 | Ok(unsafe { str::from_utf8_unchecked(bytes) }) |
| 664 | }) |
| 665 | } |
| 666 | |
| 667 | fn parse_str_raw<'s>( |
| 668 | &'s mut self, |
| 669 | scratch: &'s mut Vec<u8>, |
| 670 | ) -> Result<Reference<'a, 's, [u8]>> { |
| 671 | self.delegate.parse_str_raw(scratch) |
| 672 | } |
| 673 | |
| 674 | fn ignore_str(&mut self) -> Result<()> { |
| 675 | self.delegate.ignore_str() |
| 676 | } |
| 677 | |
| 678 | fn decode_hex_escape(&mut self) -> Result<u16> { |
| 679 | self.delegate.decode_hex_escape() |
| 680 | } |
| 681 | |
| 682 | #[cfg (feature = "raw_value" )] |
| 683 | fn begin_raw_buffering(&mut self) { |
| 684 | self.delegate.begin_raw_buffering(); |
| 685 | } |
| 686 | |
| 687 | #[cfg (feature = "raw_value" )] |
| 688 | fn end_raw_buffering<V>(&mut self, visitor: V) -> Result<V::Value> |
| 689 | where |
| 690 | V: Visitor<'a>, |
| 691 | { |
| 692 | let raw = &self.data[self.delegate.raw_buffering_start_index..self.delegate.index]; |
| 693 | visitor.visit_map(BorrowedRawDeserializer { |
| 694 | raw_value: Some(raw), |
| 695 | }) |
| 696 | } |
| 697 | |
| 698 | const should_early_return_if_failed: bool = false; |
| 699 | |
| 700 | #[inline ] |
| 701 | #[cold ] |
| 702 | fn set_failed(&mut self, failed: &mut bool) { |
| 703 | self.delegate.set_failed(failed); |
| 704 | } |
| 705 | } |
| 706 | |
| 707 | ////////////////////////////////////////////////////////////////////////////// |
| 708 | |
| 709 | impl<'a, 'de, R> private::Sealed for &'a mut R where R: Read<'de> {} |
| 710 | |
| 711 | impl<'a, 'de, R> Read<'de> for &'a mut R |
| 712 | where |
| 713 | R: Read<'de>, |
| 714 | { |
| 715 | fn next(&mut self) -> Result<Option<u8>> { |
| 716 | R::next(self) |
| 717 | } |
| 718 | |
| 719 | fn peek(&mut self) -> Result<Option<u8>> { |
| 720 | R::peek(self) |
| 721 | } |
| 722 | |
| 723 | fn discard(&mut self) { |
| 724 | R::discard(self); |
| 725 | } |
| 726 | |
| 727 | fn position(&self) -> Position { |
| 728 | R::position(self) |
| 729 | } |
| 730 | |
| 731 | fn peek_position(&self) -> Position { |
| 732 | R::peek_position(self) |
| 733 | } |
| 734 | |
| 735 | fn byte_offset(&self) -> usize { |
| 736 | R::byte_offset(self) |
| 737 | } |
| 738 | |
| 739 | fn parse_str<'s>(&'s mut self, scratch: &'s mut Vec<u8>) -> Result<Reference<'de, 's, str>> { |
| 740 | R::parse_str(self, scratch) |
| 741 | } |
| 742 | |
| 743 | fn parse_str_raw<'s>( |
| 744 | &'s mut self, |
| 745 | scratch: &'s mut Vec<u8>, |
| 746 | ) -> Result<Reference<'de, 's, [u8]>> { |
| 747 | R::parse_str_raw(self, scratch) |
| 748 | } |
| 749 | |
| 750 | fn ignore_str(&mut self) -> Result<()> { |
| 751 | R::ignore_str(self) |
| 752 | } |
| 753 | |
| 754 | fn decode_hex_escape(&mut self) -> Result<u16> { |
| 755 | R::decode_hex_escape(self) |
| 756 | } |
| 757 | |
| 758 | #[cfg (feature = "raw_value" )] |
| 759 | fn begin_raw_buffering(&mut self) { |
| 760 | R::begin_raw_buffering(self); |
| 761 | } |
| 762 | |
| 763 | #[cfg (feature = "raw_value" )] |
| 764 | fn end_raw_buffering<V>(&mut self, visitor: V) -> Result<V::Value> |
| 765 | where |
| 766 | V: Visitor<'de>, |
| 767 | { |
| 768 | R::end_raw_buffering(self, visitor) |
| 769 | } |
| 770 | |
| 771 | const should_early_return_if_failed: bool = R::should_early_return_if_failed; |
| 772 | |
| 773 | fn set_failed(&mut self, failed: &mut bool) { |
| 774 | R::set_failed(self, failed); |
| 775 | } |
| 776 | } |
| 777 | |
| 778 | ////////////////////////////////////////////////////////////////////////////// |
| 779 | |
| 780 | /// Marker for whether StreamDeserializer can implement FusedIterator. |
| 781 | pub trait Fused: private::Sealed {} |
| 782 | impl<'a> Fused for SliceRead<'a> {} |
| 783 | impl<'a> Fused for StrRead<'a> {} |
| 784 | |
| 785 | // Lookup table of bytes that must be escaped. A value of true at index i means |
| 786 | // that byte i requires an escape sequence in the input. |
| 787 | static ESCAPE: [bool; 256] = { |
| 788 | const CT: bool = true; // control character \x00..=\x1F |
| 789 | const QU: bool = true; // quote \x22 |
| 790 | const BS: bool = true; // backslash \x5C |
| 791 | const __: bool = false; // allow unescaped |
| 792 | [ |
| 793 | // 1 2 3 4 5 6 7 8 9 A B C D E F |
| 794 | CT, CT, CT, CT, CT, CT, CT, CT, CT, CT, CT, CT, CT, CT, CT, CT, // 0 |
| 795 | CT, CT, CT, CT, CT, CT, CT, CT, CT, CT, CT, CT, CT, CT, CT, CT, // 1 |
| 796 | __, __, QU, __, __, __, __, __, __, __, __, __, __, __, __, __, // 2 |
| 797 | __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // 3 |
| 798 | __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // 4 |
| 799 | __, __, __, __, __, __, __, __, __, __, __, __, BS, __, __, __, // 5 |
| 800 | __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // 6 |
| 801 | __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // 7 |
| 802 | __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // 8 |
| 803 | __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // 9 |
| 804 | __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // A |
| 805 | __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // B |
| 806 | __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // C |
| 807 | __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // D |
| 808 | __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // E |
| 809 | __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // F |
| 810 | ] |
| 811 | }; |
| 812 | |
| 813 | fn next_or_eof<'de, R>(read: &mut R) -> Result<u8> |
| 814 | where |
| 815 | R: ?Sized + Read<'de>, |
| 816 | { |
| 817 | match tri!(read.next()) { |
| 818 | Some(b) => Ok(b), |
| 819 | None => error(read, ErrorCode::EofWhileParsingString), |
| 820 | } |
| 821 | } |
| 822 | |
| 823 | fn peek_or_eof<'de, R>(read: &mut R) -> Result<u8> |
| 824 | where |
| 825 | R: ?Sized + Read<'de>, |
| 826 | { |
| 827 | match tri!(read.peek()) { |
| 828 | Some(b) => Ok(b), |
| 829 | None => error(read, ErrorCode::EofWhileParsingString), |
| 830 | } |
| 831 | } |
| 832 | |
| 833 | fn error<'de, R, T>(read: &R, reason: ErrorCode) -> Result<T> |
| 834 | where |
| 835 | R: ?Sized + Read<'de>, |
| 836 | { |
| 837 | let position = read.position(); |
| 838 | Err(Error::syntax(reason, position.line, position.column)) |
| 839 | } |
| 840 | |
| 841 | fn as_str<'de, 's, R: Read<'de>>(read: &R, slice: &'s [u8]) -> Result<&'s str> { |
| 842 | str::from_utf8(slice).or_else(|_| error(read, ErrorCode::InvalidUnicodeCodePoint)) |
| 843 | } |
| 844 | |
| 845 | /// Parses a JSON escape sequence and appends it into the scratch space. Assumes |
| 846 | /// the previous byte read was a backslash. |
| 847 | fn parse_escape<'de, R: Read<'de>>( |
| 848 | read: &mut R, |
| 849 | validate: bool, |
| 850 | scratch: &mut Vec<u8>, |
| 851 | ) -> Result<()> { |
| 852 | let ch = tri!(next_or_eof(read)); |
| 853 | |
| 854 | match ch { |
| 855 | b'"' => scratch.push(b'"' ), |
| 856 | b' \\' => scratch.push(b' \\' ), |
| 857 | b'/' => scratch.push(b'/' ), |
| 858 | b'b' => scratch.push(b' \x08' ), |
| 859 | b'f' => scratch.push(b' \x0c' ), |
| 860 | b'n' => scratch.push(b' \n' ), |
| 861 | b'r' => scratch.push(b' \r' ), |
| 862 | b't' => scratch.push(b' \t' ), |
| 863 | b'u' => { |
| 864 | fn encode_surrogate(scratch: &mut Vec<u8>, n: u16) { |
| 865 | scratch.extend_from_slice(&[ |
| 866 | (n >> 12 & 0b0000_1111) as u8 | 0b1110_0000, |
| 867 | (n >> 6 & 0b0011_1111) as u8 | 0b1000_0000, |
| 868 | (n & 0b0011_1111) as u8 | 0b1000_0000, |
| 869 | ]); |
| 870 | } |
| 871 | |
| 872 | let c = match tri!(read.decode_hex_escape()) { |
| 873 | n @ 0xDC00..=0xDFFF => { |
| 874 | return if validate { |
| 875 | error(read, ErrorCode::LoneLeadingSurrogateInHexEscape) |
| 876 | } else { |
| 877 | encode_surrogate(scratch, n); |
| 878 | Ok(()) |
| 879 | }; |
| 880 | } |
| 881 | |
| 882 | // Non-BMP characters are encoded as a sequence of two hex |
| 883 | // escapes, representing UTF-16 surrogates. If deserializing a |
| 884 | // utf-8 string the surrogates are required to be paired, |
| 885 | // whereas deserializing a byte string accepts lone surrogates. |
| 886 | n1 @ 0xD800..=0xDBFF => { |
| 887 | if tri!(peek_or_eof(read)) == b' \\' { |
| 888 | read.discard(); |
| 889 | } else { |
| 890 | return if validate { |
| 891 | read.discard(); |
| 892 | error(read, ErrorCode::UnexpectedEndOfHexEscape) |
| 893 | } else { |
| 894 | encode_surrogate(scratch, n1); |
| 895 | Ok(()) |
| 896 | }; |
| 897 | } |
| 898 | |
| 899 | if tri!(peek_or_eof(read)) == b'u' { |
| 900 | read.discard(); |
| 901 | } else { |
| 902 | return if validate { |
| 903 | read.discard(); |
| 904 | error(read, ErrorCode::UnexpectedEndOfHexEscape) |
| 905 | } else { |
| 906 | encode_surrogate(scratch, n1); |
| 907 | // The \ prior to this byte started an escape sequence, |
| 908 | // so we need to parse that now. This recursive call |
| 909 | // does not blow the stack on malicious input because |
| 910 | // the escape is not \u, so it will be handled by one |
| 911 | // of the easy nonrecursive cases. |
| 912 | parse_escape(read, validate, scratch) |
| 913 | }; |
| 914 | } |
| 915 | |
| 916 | let n2 = tri!(read.decode_hex_escape()); |
| 917 | |
| 918 | if n2 < 0xDC00 || n2 > 0xDFFF { |
| 919 | return error(read, ErrorCode::LoneLeadingSurrogateInHexEscape); |
| 920 | } |
| 921 | |
| 922 | let n = (((n1 - 0xD800) as u32) << 10 | (n2 - 0xDC00) as u32) + 0x1_0000; |
| 923 | |
| 924 | match char::from_u32(n) { |
| 925 | Some(c) => c, |
| 926 | None => { |
| 927 | return error(read, ErrorCode::InvalidUnicodeCodePoint); |
| 928 | } |
| 929 | } |
| 930 | } |
| 931 | |
| 932 | // Every u16 outside of the surrogate ranges above is guaranteed |
| 933 | // to be a legal char. |
| 934 | n => char::from_u32(n as u32).unwrap(), |
| 935 | }; |
| 936 | |
| 937 | scratch.extend_from_slice(c.encode_utf8(&mut [0_u8; 4]).as_bytes()); |
| 938 | } |
| 939 | _ => { |
| 940 | return error(read, ErrorCode::InvalidEscape); |
| 941 | } |
| 942 | } |
| 943 | |
| 944 | Ok(()) |
| 945 | } |
| 946 | |
| 947 | /// Parses a JSON escape sequence and discards the value. Assumes the previous |
| 948 | /// byte read was a backslash. |
| 949 | fn ignore_escape<'de, R>(read: &mut R) -> Result<()> |
| 950 | where |
| 951 | R: ?Sized + Read<'de>, |
| 952 | { |
| 953 | let ch = tri!(next_or_eof(read)); |
| 954 | |
| 955 | match ch { |
| 956 | b'"' | b' \\' | b'/' | b'b' | b'f' | b'n' | b'r' | b't' => {} |
| 957 | b'u' => { |
| 958 | // At this point we don't care if the codepoint is valid. We just |
| 959 | // want to consume it. We don't actually know what is valid or not |
| 960 | // at this point, because that depends on if this string will |
| 961 | // ultimately be parsed into a string or a byte buffer in the "real" |
| 962 | // parse. |
| 963 | |
| 964 | tri!(read.decode_hex_escape()); |
| 965 | } |
| 966 | _ => { |
| 967 | return error(read, ErrorCode::InvalidEscape); |
| 968 | } |
| 969 | } |
| 970 | |
| 971 | Ok(()) |
| 972 | } |
| 973 | |
| 974 | static HEX: [u8; 256] = { |
| 975 | const __: u8 = 255; // not a hex digit |
| 976 | [ |
| 977 | // 1 2 3 4 5 6 7 8 9 A B C D E F |
| 978 | __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // 0 |
| 979 | __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // 1 |
| 980 | __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // 2 |
| 981 | 00, 01, 02, 03, 04, 05, 06, 07, 08, 09, __, __, __, __, __, __, // 3 |
| 982 | __, 10, 11, 12, 13, 14, 15, __, __, __, __, __, __, __, __, __, // 4 |
| 983 | __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // 5 |
| 984 | __, 10, 11, 12, 13, 14, 15, __, __, __, __, __, __, __, __, __, // 6 |
| 985 | __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // 7 |
| 986 | __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // 8 |
| 987 | __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // 9 |
| 988 | __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // A |
| 989 | __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // B |
| 990 | __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // C |
| 991 | __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // D |
| 992 | __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // E |
| 993 | __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // F |
| 994 | ] |
| 995 | }; |
| 996 | |
| 997 | fn decode_hex_val(val: u8) -> Option<u16> { |
| 998 | let n = HEX[val as usize] as u16; |
| 999 | if n == 255 { |
| 1000 | None |
| 1001 | } else { |
| 1002 | Some(n) |
| 1003 | } |
| 1004 | } |
| 1005 | |