| 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(all(feature = "raw_value", feature = "std"))] |
| 18 |
use alloc::string::String; |
| 19 |
#[cfg(feature = "raw_value")] |
| 20 |
use serde::de::Visitor; |
| 21 |
|
| 22 |
/// Trait used by the deserializer for iterating over input. This is manually |
| 23 |
/// "specialized" for iterating over `&[u8]`. Once feature(specialization) is |
| 24 |
/// stable we can use actual specialization. |
| 25 |
/// |
| 26 |
/// This trait is sealed and cannot be implemented for types outside of |
| 27 |
/// `serde_json`. |
| 28 |
pub trait Read<'de>: private::Sealed { |
| 29 |
#[doc(hidden)] |
| 30 |
fn next(&mut self) -> Result<Option<u8>>; |
| 31 |
#[doc(hidden)] |
| 32 |
fn peek(&mut self) -> Result<Option<u8>>; |
| 33 |
|
| 34 |
/// Only valid after a call to peek(). Discards the peeked byte. |
| 35 |
#[doc(hidden)] |
| 36 |
fn discard(&mut self); |
| 37 |
|
| 38 |
/// Position of the most recent call to next(). |
| 39 |
/// |
| 40 |
/// The most recent call was probably next() and not peek(), but this method |
| 41 |
/// should try to return a sensible result if the most recent call was |
| 42 |
/// actually peek() because we don't always know. |
| 43 |
/// |
| 44 |
/// Only called in case of an error, so performance is not important. |
| 45 |
#[doc(hidden)] |
| 46 |
fn position(&self) -> Position; |
| 47 |
|
| 48 |
/// Position of the most recent call to peek(). |
| 49 |
/// |
| 50 |
/// The most recent call was probably peek() and not next(), but this method |
| 51 |
/// should try to return a sensible result if the most recent call was |
| 52 |
/// actually next() because we don't always know. |
| 53 |
/// |
| 54 |
/// Only called in case of an error, so performance is not important. |
| 55 |
#[doc(hidden)] |
| 56 |
fn peek_position(&self) -> Position; |
| 57 |
|
| 58 |
/// Offset from the beginning of the input to the next byte that would be |
| 59 |
/// returned by next() or peek(). |
| 60 |
#[doc(hidden)] |
| 61 |
fn byte_offset(&self) -> usize; |
| 62 |
|
| 63 |
/// Assumes the previous byte was a quotation mark. Parses a JSON-escaped |
| 64 |
/// string until the next quotation mark using the given scratch space if |
| 65 |
/// necessary. The scratch space is initially empty. |
| 66 |
#[doc(hidden)] |
| 67 |
fn parse_str<'s>(&'s mut self, scratch: &'s mut Vec<u8>) -> Result<Reference<'de, 's, str>>; |
| 68 |
|
| 69 |
/// Assumes the previous byte was a quotation mark. Parses a JSON-escaped |
| 70 |
/// string until the next quotation mark using the given scratch space if |
| 71 |
/// necessary. The scratch space is initially empty. |
| 72 |
/// |
| 73 |
/// This function returns the raw bytes in the string with escape sequences |
| 74 |
/// expanded but without performing unicode validation. |
| 75 |
#[doc(hidden)] |
| 76 |
fn parse_str_raw<'s>( |
| 77 |
&'s mut self, |
| 78 |
scratch: &'s mut Vec<u8>, |
| 79 |
) -> Result<Reference<'de, 's, [u8]>>; |
| 80 |
|
| 81 |
/// Assumes the previous byte was a quotation mark. Parses a JSON-escaped |
| 82 |
/// string until the next quotation mark but discards the data. |
| 83 |
#[doc(hidden)] |
| 84 |
fn ignore_str(&mut self) -> Result<()>; |
| 85 |
|
| 86 |
/// Assumes the previous byte was a hex escape sequence ('\u') in a string. |
| 87 |
/// Parses next hexadecimal sequence. |
| 88 |
#[doc(hidden)] |
| 89 |
fn decode_hex_escape(&mut self) -> Result<u16>; |
| 90 |
|
| 91 |
/// Switch raw buffering mode on. |
| 92 |
/// |
| 93 |
/// This is used when deserializing `RawValue`. |
| 94 |
#[cfg(feature = "raw_value")] |
| 95 |
#[doc(hidden)] |
| 96 |
fn begin_raw_buffering(&mut self); |
| 97 |
|
| 98 |
/// Switch raw buffering mode off and provides the raw buffered data to the |
| 99 |
/// given visitor. |
| 100 |
#[cfg(feature = "raw_value")] |
| 101 |
#[doc(hidden)] |
| 102 |
fn end_raw_buffering<V>(&mut self, visitor: V) -> Result<V::Value> |
| 103 |
where |
| 104 |
V: Visitor<'de>; |
| 105 |
|
| 106 |
/// Whether StreamDeserializer::next needs to check the failed flag. True |
| 107 |
/// for IoRead, false for StrRead and SliceRead which can track failure by |
| 108 |
/// truncating their input slice to avoid the extra check on every next |
| 109 |
/// call. |
| 110 |
#[doc(hidden)] |
| 111 |
const should_early_return_if_failed: bool; |
| 112 |
|
| 113 |
/// Mark a persistent failure of StreamDeserializer, either by setting the |
| 114 |
/// flag or by truncating the input data. |
| 115 |
#[doc(hidden)] |
| 116 |
fn set_failed(&mut self, failed: &mut bool); |
| 117 |
} |
| 118 |
|
| 119 |
pub struct Position { |
| 120 |
pub line: usize, |
| 121 |
pub column: usize, |
| 122 |
} |
| 123 |
|
| 124 |
pub enum Reference<'b, 'c, T> |
| 125 |
where |
| 126 |
T: ?Sized + 'static, |
| 127 |
{ |
| 128 |
Borrowed(&'b T), |
| 129 |
Copied(&'c T), |
| 130 |
} |
| 131 |
|
| 132 |
impl<'b, 'c, T> Deref for Reference<'b, 'c, T> |
| 133 |
where |
| 134 |
T: ?Sized + 'static, |
| 135 |
{ |
| 136 |
type Target = T; |
| 137 |
|
| 138 |
fn deref(&self) -> &Self::Target { |
| 139 |
match *self { |
| 140 |
Reference::Borrowed(b) => b, |
| 141 |
Reference::Copied(c) => c, |
| 142 |
} |
| 143 |
} |
| 144 |
} |
| 145 |
|
| 146 |
/// JSON input source that reads from a std::io input stream. |
| 147 |
#[cfg(feature = "std")] |
| 148 |
#[cfg_attr(docsrs, doc(cfg(feature = "std")))] |
| 149 |
pub struct IoRead<R> |
| 150 |
where |
| 151 |
R: io::Read, |
| 152 |
{ |
| 153 |
iter: LineColIterator<io::Bytes<R>>, |
| 154 |
/// Temporary storage of peeked byte. |
| 155 |
ch: Option<u8>, |
| 156 |
#[cfg(feature = "raw_value")] |
| 157 |
raw_buffer: Option<Vec<u8>>, |
| 158 |
} |
| 159 |
|
| 160 |
/// JSON input source that reads from a slice of bytes. |
| 161 |
// |
| 162 |
// This is more efficient than other iterators because peek() can be read-only |
| 163 |
// and we can compute line/col position only if an error happens. |
| 164 |
pub struct SliceRead<'a> { |
| 165 |
slice: &'a [u8], |
| 166 |
/// Index of the *next* byte that will be returned by next() or peek(). |
| 167 |
index: usize, |
| 168 |
#[cfg(feature = "raw_value")] |
| 169 |
raw_buffering_start_index: usize, |
| 170 |
} |
| 171 |
|
| 172 |
/// JSON input source that reads from a UTF-8 string. |
| 173 |
// |
| 174 |
// Able to elide UTF-8 checks by assuming that the input is valid UTF-8. |
| 175 |
pub struct StrRead<'a> { |
| 176 |
delegate: SliceRead<'a>, |
| 177 |
#[cfg(feature = "raw_value")] |
| 178 |
data: &'a str, |
| 179 |
} |
| 180 |
|
| 181 |
// Prevent users from implementing the Read trait. |
| 182 |
mod private { |
| 183 |
pub trait Sealed {} |
| 184 |
} |
| 185 |
|
| 186 |
////////////////////////////////////////////////////////////////////////////// |
| 187 |
|
| 188 |
#[cfg(feature = "std")] |
| 189 |
impl<R> IoRead<R> |
| 190 |
where |
| 191 |
R: io::Read, |
| 192 |
{ |
| 193 |
/// Create a JSON input source to read from a std::io input stream. |
| 194 |
pub fn new(reader: R) -> Self { |
| 195 |
IoRead { |
| 196 |
iter: LineColIterator::new(reader.bytes()), |
| 197 |
ch: None, |
| 198 |
#[cfg(feature = "raw_value")] |
| 199 |
raw_buffer: None, |
| 200 |
} |
| 201 |
} |
| 202 |
} |
| 203 |
|
| 204 |
#[cfg(feature = "std")] |
| 205 |
impl<R> private::Sealed for IoRead<R> where R: io::Read {} |
| 206 |
|
| 207 |
#[cfg(feature = "std")] |
| 208 |
impl<R> IoRead<R> |
| 209 |
where |
| 210 |
R: io::Read, |
| 211 |
{ |
| 212 |
fn parse_str_bytes<'s, T, F>( |
| 213 |
&'s mut self, |
| 214 |
scratch: &'s mut Vec<u8>, |
| 215 |
validate: bool, |
| 216 |
result: F, |
| 217 |
) -> Result<T> |
| 218 |
where |
| 219 |
T: 's, |
| 220 |
F: FnOnce(&'s Self, &'s [u8]) -> Result<T>, |
| 221 |
{ |
| 222 |
loop { |
| 223 |
let ch = tri!(next_or_eof(self)); |
| 224 |
if !ESCAPE[ch as usize] { |
| 225 |
scratch.push(ch); |
| 226 |
continue; |
| 227 |
} |
| 228 |
match ch { |
| 229 |
b'"' => { |
| 230 |
return result(self, scratch); |
| 231 |
} |
| 232 |
b'\\' => { |
| 233 |
tri!(parse_escape(self, validate, scratch)); |
| 234 |
} |
| 235 |
_ => { |
| 236 |
if validate { |
| 237 |
return error(self, ErrorCode::ControlCharacterWhileParsingString); |
| 238 |
} |
| 239 |
scratch.push(ch); |
| 240 |
} |
| 241 |
} |
| 242 |
} |
| 243 |
} |
| 244 |
} |
| 245 |
|
| 246 |
#[cfg(feature = "std")] |
| 247 |
impl<'de, R> Read<'de> for IoRead<R> |
| 248 |
where |
| 249 |
R: io::Read, |
| 250 |
{ |
| 251 |
#[inline] |
| 252 |
fn next(&mut self) -> Result<Option<u8>> { |
| 253 |
match self.ch.take() { |
| 254 |
Some(ch) => { |
| 255 |
#[cfg(feature = "raw_value")] |
| 256 |
{ |
| 257 |
if let Some(buf) = &mut self.raw_buffer { |
| 258 |
buf.push(ch); |
| 259 |
} |
| 260 |
} |
| 261 |
Ok(Some(ch)) |
| 262 |
} |
| 263 |
None => match self.iter.next() { |
| 264 |
Some(Err(err)) => Err(Error::io(err)), |
| 265 |
Some(Ok(ch)) => { |
| 266 |
#[cfg(feature = "raw_value")] |
| 267 |
{ |
| 268 |
if let Some(buf) = &mut self.raw_buffer { |
| 269 |
buf.push(ch); |
| 270 |
} |
| 271 |
} |
| 272 |
Ok(Some(ch)) |
| 273 |
} |
| 274 |
None => Ok(None), |
| 275 |
}, |
| 276 |
} |
| 277 |
} |
| 278 |
|
| 279 |
#[inline] |
| 280 |
fn peek(&mut self) -> Result<Option<u8>> { |
| 281 |
match self.ch { |
| 282 |
Some(ch) => Ok(Some(ch)), |
| 283 |
None => match self.iter.next() { |
| 284 |
Some(Err(err)) => Err(Error::io(err)), |
| 285 |
Some(Ok(ch)) => { |
| 286 |
self.ch = Some(ch); |
| 287 |
Ok(self.ch) |
| 288 |
} |
| 289 |
None => Ok(None), |
| 290 |
}, |
| 291 |
} |
| 292 |
} |
| 293 |
|
| 294 |
#[cfg(not(feature = "raw_value"))] |
| 295 |
#[inline] |
| 296 |
fn discard(&mut self) { |
| 297 |
self.ch = None; |
| 298 |
} |
| 299 |
|
| 300 |
#[cfg(feature = "raw_value")] |
| 301 |
fn discard(&mut self) { |
| 302 |
if let Some(ch) = self.ch.take() { |
| 303 |
if let Some(buf) = &mut self.raw_buffer { |
| 304 |
buf.push(ch); |
| 305 |
} |
| 306 |
} |
| 307 |
} |
| 308 |
|
| 309 |
fn position(&self) -> Position { |
| 310 |
Position { |
| 311 |
line: self.iter.line(), |
| 312 |
column: self.iter.col(), |
| 313 |
} |
| 314 |
} |
| 315 |
|
| 316 |
fn peek_position(&self) -> Position { |
| 317 |
// The LineColIterator updates its position during peek() so it has the |
| 318 |
// right one here. |
| 319 |
self.position() |
| 320 |
} |
| 321 |
|
| 322 |
fn byte_offset(&self) -> usize { |
| 323 |
match self.ch { |
| 324 |
Some(_) => self.iter.byte_offset() - 1, |
| 325 |
None => self.iter.byte_offset(), |
| 326 |
} |
| 327 |
} |
| 328 |
|
| 329 |
fn parse_str<'s>(&'s mut self, scratch: &'s mut Vec<u8>) -> Result<Reference<'de, 's, str>> { |
| 330 |
self.parse_str_bytes(scratch, true, as_str) |
| 331 |
.map(Reference::Copied) |
| 332 |
} |
| 333 |
|
| 334 |
fn parse_str_raw<'s>( |
| 335 |
&'s mut self, |
| 336 |
scratch: &'s mut Vec<u8>, |
| 337 |
) -> Result<Reference<'de, 's, [u8]>> { |
| 338 |
self.parse_str_bytes(scratch, false, |_, bytes| Ok(bytes)) |
| 339 |
.map(Reference::Copied) |
| 340 |
} |
| 341 |
|
| 342 |
fn ignore_str(&mut self) -> Result<()> { |
| 343 |
loop { |
| 344 |
let ch = tri!(next_or_eof(self)); |
| 345 |
if !ESCAPE[ch as usize] { |
| 346 |
continue; |
| 347 |
} |
| 348 |
match ch { |
| 349 |
b'"' => { |
| 350 |
return Ok(()); |
| 351 |
} |
| 352 |
b'\\' => { |
| 353 |
tri!(ignore_escape(self)); |
| 354 |
} |
| 355 |
_ => { |
| 356 |
return error(self, ErrorCode::ControlCharacterWhileParsingString); |
| 357 |
} |
| 358 |
} |
| 359 |
} |
| 360 |
} |
| 361 |
|
| 362 |
fn decode_hex_escape(&mut self) -> Result<u16> { |
| 363 |
let mut n = 0; |
| 364 |
for _ in 0..4 { |
| 365 |
match decode_hex_val(tri!(next_or_eof(self))) { |
| 366 |
None => return error(self, ErrorCode::InvalidEscape), |
| 367 |
Some(val) => { |
| 368 |
n = (n << 4) + val; |
| 369 |
} |
| 370 |
} |
| 371 |
} |
| 372 |
Ok(n) |
| 373 |
} |
| 374 |
|
| 375 |
#[cfg(feature = "raw_value")] |
| 376 |
fn begin_raw_buffering(&mut self) { |
| 377 |
self.raw_buffer = Some(Vec::new()); |
| 378 |
} |
| 379 |
|
| 380 |
#[cfg(feature = "raw_value")] |
| 381 |
fn end_raw_buffering<V>(&mut self, visitor: V) -> Result<V::Value> |
| 382 |
where |
| 383 |
V: Visitor<'de>, |
| 384 |
{ |
| 385 |
let raw = self.raw_buffer.take().unwrap(); |
| 386 |
let raw = match String::from_utf8(raw) { |
| 387 |
Ok(raw) => raw, |
| 388 |
Err(_) => return error(self, ErrorCode::InvalidUnicodeCodePoint), |
| 389 |
}; |
| 390 |
visitor.visit_map(OwnedRawDeserializer { |
| 391 |
raw_value: Some(raw), |
| 392 |
}) |
| 393 |
} |
| 394 |
|
| 395 |
const should_early_return_if_failed: bool = true; |
| 396 |
|
| 397 |
#[inline] |
| 398 |
#[cold] |
| 399 |
fn set_failed(&mut self, failed: &mut bool) { |
| 400 |
*failed = true; |
| 401 |
} |
| 402 |
} |
| 403 |
|
| 404 |
////////////////////////////////////////////////////////////////////////////// |
| 405 |
|
| 406 |
impl<'a> SliceRead<'a> { |
| 407 |
/// Create a JSON input source to read from a slice of bytes. |
| 408 |
pub fn new(slice: &'a [u8]) -> Self { |
| 409 |
SliceRead { |
| 410 |
slice, |
| 411 |
index: 0, |
| 412 |
#[cfg(feature = "raw_value")] |
| 413 |
raw_buffering_start_index: 0, |
| 414 |
} |
| 415 |
} |
| 416 |
|
| 417 |
fn position_of_index(&self, i: usize) -> Position { |
| 418 |
let mut position = Position { line: 1, column: 0 }; |
| 419 |
for ch in &self.slice[..i] { |
| 420 |
match *ch { |
| 421 |
b'\n' => { |
| 422 |
position.line += 1; |
| 423 |
position.column = 0; |
| 424 |
} |
| 425 |
_ => { |
| 426 |
position.column += 1; |
| 427 |
} |
| 428 |
} |
| 429 |
} |
| 430 |
position |
| 431 |
} |
| 432 |
|
| 433 |
/// The big optimization here over IoRead is that if the string contains no |
| 434 |
/// backslash escape sequences, the returned &str is a slice of the raw JSON |
| 435 |
/// data so we avoid copying into the scratch space. |
| 436 |
fn parse_str_bytes<'s, T, F>( |
| 437 |
&'s mut self, |
| 438 |
scratch: &'s mut Vec<u8>, |
| 439 |
validate: bool, |
| 440 |
result: F, |
| 441 |
) -> Result<Reference<'a, 's, T>> |
| 442 |
where |
| 443 |
T: ?Sized + 's, |
| 444 |
F: for<'f> FnOnce(&'s Self, &'f [u8]) -> Result<&'f T>, |
| 445 |
{ |
| 446 |
// Index of the first byte not yet copied into the scratch space. |
| 447 |
let mut start = self.index; |
| 448 |
|
| 449 |
loop { |
| 450 |
while self.index < self.slice.len() && !ESCAPE[self.slice[self.index] as usize] { |
| 451 |
self.index += 1; |
| 452 |
} |
| 453 |
if self.index == self.slice.len() { |
| 454 |
return error(self, ErrorCode::EofWhileParsingString); |
| 455 |
} |
| 456 |
match self.slice[self.index] { |
| 457 |
b'"' => { |
| 458 |
if scratch.is_empty() { |
| 459 |
// Fast path: return a slice of the raw JSON without any |
| 460 |
// copying. |
| 461 |
let borrowed = &self.slice[start..self.index]; |
| 462 |
self.index += 1; |
| 463 |
return result(self, borrowed).map(Reference::Borrowed); |
| 464 |
} else { |
| 465 |
scratch.extend_from_slice(&self.slice[start..self.index]); |
| 466 |
self.index += 1; |
| 467 |
return result(self, scratch).map(Reference::Copied); |
| 468 |
} |
| 469 |
} |
| 470 |
b'\\' => { |
| 471 |
scratch.extend_from_slice(&self.slice[start..self.index]); |
| 472 |
self.index += 1; |
| 473 |
tri!(parse_escape(self, validate, scratch)); |
| 474 |
start = self.index; |
| 475 |
} |
| 476 |
_ => { |
| 477 |
self.index += 1; |
| 478 |
if validate { |
| 479 |
return error(self, ErrorCode::ControlCharacterWhileParsingString); |
| 480 |
} |
| 481 |
} |
| 482 |
} |
| 483 |
} |
| 484 |
} |
| 485 |
} |
| 486 |
|
| 487 |
impl<'a> private::Sealed for SliceRead<'a> {} |
| 488 |
|
| 489 |
impl<'a> Read<'a> for SliceRead<'a> { |
| 490 |
#[inline] |
| 491 |
fn next(&mut self) -> Result<Option<u8>> { |
| 492 |
// `Ok(self.slice.get(self.index).map(|ch| { self.index += 1; *ch }))` |
| 493 |
// is about 10% slower. |
| 494 |
Ok(if self.index < self.slice.len() { |
| 495 |
let ch = self.slice[self.index]; |
| 496 |
self.index += 1; |
| 497 |
Some(ch) |
| 498 |
} else { |
| 499 |
None |
| 500 |
}) |
| 501 |
} |
| 502 |
|
| 503 |
#[inline] |
| 504 |
fn peek(&mut self) -> Result<Option<u8>> { |
| 505 |
// `Ok(self.slice.get(self.index).map(|ch| *ch))` is about 10% slower |
| 506 |
// for some reason. |
| 507 |
Ok(if self.index < self.slice.len() { |
| 508 |
Some(self.slice[self.index]) |
| 509 |
} else { |
| 510 |
None |
| 511 |
}) |
| 512 |
} |
| 513 |
|
| 514 |
#[inline] |
| 515 |
fn discard(&mut self) { |
| 516 |
self.index += 1; |
| 517 |
} |
| 518 |
|
| 519 |
fn position(&self) -> Position { |
| 520 |
self.position_of_index(self.index) |
| 521 |
} |
| 522 |
|
| 523 |
fn peek_position(&self) -> Position { |
| 524 |
// Cap it at slice.len() just in case the most recent call was next() |
| 525 |
// and it returned the last byte. |
| 526 |
self.position_of_index(cmp::min(self.slice.len(), self.index + 1)) |
| 527 |
} |
| 528 |
|
| 529 |
fn byte_offset(&self) -> usize { |
| 530 |
self.index |
| 531 |
} |
| 532 |
|
| 533 |
fn parse_str<'s>(&'s mut self, scratch: &'s mut Vec<u8>) -> Result<Reference<'a, 's, str>> { |
| 534 |
self.parse_str_bytes(scratch, true, as_str) |
| 535 |
} |
| 536 |
|
| 537 |
fn parse_str_raw<'s>( |
| 538 |
&'s mut self, |
| 539 |
scratch: &'s mut Vec<u8>, |
| 540 |
) -> Result<Reference<'a, 's, [u8]>> { |
| 541 |
self.parse_str_bytes(scratch, false, |_, bytes| Ok(bytes)) |
| 542 |
} |
| 543 |
|
| 544 |
fn ignore_str(&mut self) -> Result<()> { |
| 545 |
loop { |
| 546 |
while self.index < self.slice.len() && !ESCAPE[self.slice[self.index] as usize] { |
| 547 |
self.index += 1; |
| 548 |
} |
| 549 |
if self.index == self.slice.len() { |
| 550 |
return error(self, ErrorCode::EofWhileParsingString); |
| 551 |
} |
| 552 |
match self.slice[self.index] { |
| 553 |
b'"' => { |
| 554 |
self.index += 1; |
| 555 |
return Ok(()); |
| 556 |
} |
| 557 |
b'\\' => { |
| 558 |
self.index += 1; |
| 559 |
tri!(ignore_escape(self)); |
| 560 |
} |
| 561 |
_ => { |
| 562 |
return error(self, ErrorCode::ControlCharacterWhileParsingString); |
| 563 |
} |
| 564 |
} |
| 565 |
} |
| 566 |
} |
| 567 |
|
| 568 |
fn decode_hex_escape(&mut self) -> Result<u16> { |
| 569 |
if self.index + 4 > self.slice.len() { |
| 570 |
self.index = self.slice.len(); |
| 571 |
return error(self, ErrorCode::EofWhileParsingString); |
| 572 |
} |
| 573 |
|
| 574 |
let mut n = 0; |
| 575 |
for _ in 0..4 { |
| 576 |
let ch = decode_hex_val(self.slice[self.index]); |
| 577 |
self.index += 1; |
| 578 |
match ch { |
| 579 |
None => return error(self, ErrorCode::InvalidEscape), |
| 580 |
Some(val) => { |
| 581 |
n = (n << 4) + val; |
| 582 |
} |
| 583 |
} |
| 584 |
} |
| 585 |
Ok(n) |
| 586 |
} |
| 587 |
|
| 588 |
#[cfg(feature = "raw_value")] |
| 589 |
fn begin_raw_buffering(&mut self) { |
| 590 |
self.raw_buffering_start_index = self.index; |
| 591 |
} |
| 592 |
|
| 593 |
#[cfg(feature = "raw_value")] |
| 594 |
fn end_raw_buffering<V>(&mut self, visitor: V) -> Result<V::Value> |
| 595 |
where |
| 596 |
V: Visitor<'a>, |
| 597 |
{ |
| 598 |
let raw = &self.slice[self.raw_buffering_start_index..self.index]; |
| 599 |
let raw = match str::from_utf8(raw) { |
| 600 |
Ok(raw) => raw, |
| 601 |
Err(_) => return error(self, ErrorCode::InvalidUnicodeCodePoint), |
| 602 |
}; |
| 603 |
visitor.visit_map(BorrowedRawDeserializer { |
| 604 |
raw_value: Some(raw), |
| 605 |
}) |
| 606 |
} |
| 607 |
|
| 608 |
const should_early_return_if_failed: bool = false; |
| 609 |
|
| 610 |
#[inline] |
| 611 |
#[cold] |
| 612 |
fn set_failed(&mut self, _failed: &mut bool) { |
| 613 |
self.slice = &self.slice[..self.index]; |
| 614 |
} |
| 615 |
} |
| 616 |
|
| 617 |
////////////////////////////////////////////////////////////////////////////// |
| 618 |
|
| 619 |
impl<'a> StrRead<'a> { |
| 620 |
/// Create a JSON input source to read from a UTF-8 string. |
| 621 |
pub fn new(s: &'a str) -> Self { |
| 622 |
StrRead { |
| 623 |
delegate: SliceRead::new(s.as_bytes()), |
| 624 |
#[cfg(feature = "raw_value")] |
| 625 |
data: s, |
| 626 |
} |
| 627 |
} |
| 628 |
} |
| 629 |
|
| 630 |
impl<'a> private::Sealed for StrRead<'a> {} |
| 631 |
|
| 632 |
impl<'a> Read<'a> for StrRead<'a> { |
| 633 |
#[inline] |
| 634 |
fn next(&mut self) -> Result<Option<u8>> { |
| 635 |
self.delegate.next() |
| 636 |
} |
| 637 |
|
| 638 |
#[inline] |
| 639 |
fn peek(&mut self) -> Result<Option<u8>> { |
| 640 |
self.delegate.peek() |
| 641 |
} |
| 642 |
|
| 643 |
#[inline] |
| 644 |
fn discard(&mut self) { |
| 645 |
self.delegate.discard(); |
| 646 |
} |
| 647 |
|
| 648 |
fn position(&self) -> Position { |
| 649 |
self.delegate.position() |
| 650 |
} |
| 651 |
|
| 652 |
fn peek_position(&self) -> Position { |
| 653 |
self.delegate.peek_position() |
| 654 |
} |
| 655 |
|
| 656 |
fn byte_offset(&self) -> usize { |
| 657 |
self.delegate.byte_offset() |
| 658 |
} |
| 659 |
|
| 660 |
fn parse_str<'s>(&'s mut self, scratch: &'s mut Vec<u8>) -> Result<Reference<'a, 's, str>> { |
| 661 |
self.delegate.parse_str_bytes(scratch, true, |_, bytes| { |
| 662 |
// The deserialization input came in as &str with a UTF-8 guarantee, |
| 663 |
// and the \u-escapes are checked along the way, so don't need to |
| 664 |
// check here. |
| 665 |
Ok(unsafe { str::from_utf8_unchecked(bytes) }) |
| 666 |
}) |
| 667 |
} |
| 668 |
|
| 669 |
fn parse_str_raw<'s>( |
| 670 |
&'s mut self, |
| 671 |
scratch: &'s mut Vec<u8>, |
| 672 |
) -> Result<Reference<'a, 's, [u8]>> { |
| 673 |
self.delegate.parse_str_raw(scratch) |
| 674 |
} |
| 675 |
|
| 676 |
fn ignore_str(&mut self) -> Result<()> { |
| 677 |
self.delegate.ignore_str() |
| 678 |
} |
| 679 |
|
| 680 |
fn decode_hex_escape(&mut self) -> Result<u16> { |
| 681 |
self.delegate.decode_hex_escape() |
| 682 |
} |
| 683 |
|
| 684 |
#[cfg(feature = "raw_value")] |
| 685 |
fn begin_raw_buffering(&mut self) { |
| 686 |
self.delegate.begin_raw_buffering(); |
| 687 |
} |
| 688 |
|
| 689 |
#[cfg(feature = "raw_value")] |
| 690 |
fn end_raw_buffering<V>(&mut self, visitor: V) -> Result<V::Value> |
| 691 |
where |
| 692 |
V: Visitor<'a>, |
| 693 |
{ |
| 694 |
let raw = &self.data[self.delegate.raw_buffering_start_index..self.delegate.index]; |
| 695 |
visitor.visit_map(BorrowedRawDeserializer { |
| 696 |
raw_value: Some(raw), |
| 697 |
}) |
| 698 |
} |
| 699 |
|
| 700 |
const should_early_return_if_failed: bool = false; |
| 701 |
|
| 702 |
#[inline] |
| 703 |
#[cold] |
| 704 |
fn set_failed(&mut self, failed: &mut bool) { |
| 705 |
self.delegate.set_failed(failed); |
| 706 |
} |
| 707 |
} |
| 708 |
|
| 709 |
////////////////////////////////////////////////////////////////////////////// |
| 710 |
|
| 711 |
impl<'a, 'de, R> private::Sealed for &'a mut R where R: Read<'de> {} |
| 712 |
|
| 713 |
impl<'a, 'de, R> Read<'de> for &'a mut R |
| 714 |
where |
| 715 |
R: Read<'de>, |
| 716 |
{ |
| 717 |
fn next(&mut self) -> Result<Option<u8>> { |
| 718 |
R::next(self) |
| 719 |
} |
| 720 |
|
| 721 |
fn peek(&mut self) -> Result<Option<u8>> { |
| 722 |
R::peek(self) |
| 723 |
} |
| 724 |
|
| 725 |
fn discard(&mut self) { |
| 726 |
R::discard(self); |
| 727 |
} |
| 728 |
|
| 729 |
fn position(&self) -> Position { |
| 730 |
R::position(self) |
| 731 |
} |
| 732 |
|
| 733 |
fn peek_position(&self) -> Position { |
| 734 |
R::peek_position(self) |
| 735 |
} |
| 736 |
|
| 737 |
fn byte_offset(&self) -> usize { |
| 738 |
R::byte_offset(self) |
| 739 |
} |
| 740 |
|
| 741 |
fn parse_str<'s>(&'s mut self, scratch: &'s mut Vec<u8>) -> Result<Reference<'de, 's, str>> { |
| 742 |
R::parse_str(self, scratch) |
| 743 |
} |
| 744 |
|
| 745 |
fn parse_str_raw<'s>( |
| 746 |
&'s mut self, |
| 747 |
scratch: &'s mut Vec<u8>, |
| 748 |
) -> Result<Reference<'de, 's, [u8]>> { |
| 749 |
R::parse_str_raw(self, scratch) |
| 750 |
} |
| 751 |
|
| 752 |
fn ignore_str(&mut self) -> Result<()> { |
| 753 |
R::ignore_str(self) |
| 754 |
} |
| 755 |
|
| 756 |
fn decode_hex_escape(&mut self) -> Result<u16> { |
| 757 |
R::decode_hex_escape(self) |
| 758 |
} |
| 759 |
|
| 760 |
#[cfg(feature = "raw_value")] |
| 761 |
fn begin_raw_buffering(&mut self) { |
| 762 |
R::begin_raw_buffering(self); |
| 763 |
} |
| 764 |
|
| 765 |
#[cfg(feature = "raw_value")] |
| 766 |
fn end_raw_buffering<V>(&mut self, visitor: V) -> Result<V::Value> |
| 767 |
where |
| 768 |
V: Visitor<'de>, |
| 769 |
{ |
| 770 |
R::end_raw_buffering(self, visitor) |
| 771 |
} |
| 772 |
|
| 773 |
const should_early_return_if_failed: bool = R::should_early_return_if_failed; |
| 774 |
|
| 775 |
fn set_failed(&mut self, failed: &mut bool) { |
| 776 |
R::set_failed(self, failed); |
| 777 |
} |
| 778 |
} |
| 779 |
|
| 780 |
////////////////////////////////////////////////////////////////////////////// |
| 781 |
|
| 782 |
/// Marker for whether StreamDeserializer can implement FusedIterator. |
| 783 |
pub trait Fused: private::Sealed {} |
| 784 |
impl<'a> Fused for SliceRead<'a> {} |
| 785 |
impl<'a> Fused for StrRead<'a> {} |
| 786 |
|
| 787 |
// Lookup table of bytes that must be escaped. A value of true at index i means |
| 788 |
// that byte i requires an escape sequence in the input. |
| 789 |
static ESCAPE: [bool; 256] = { |
| 790 |
const CT: bool = true; // control character \x00..=\x1F |
| 791 |
const QU: bool = true; // quote \x22 |
| 792 |
const BS: bool = true; // backslash \x5C |
| 793 |
const __: bool = false; // allow unescaped |
| 794 |
[ |
| 795 |
// 1 2 3 4 5 6 7 8 9 A B C D E F |
| 796 |
CT, CT, CT, CT, CT, CT, CT, CT, CT, CT, CT, CT, CT, CT, CT, CT, // 0 |
| 797 |
CT, CT, CT, CT, CT, CT, CT, CT, CT, CT, CT, CT, CT, CT, CT, CT, // 1 |
| 798 |
__, __, QU, __, __, __, __, __, __, __, __, __, __, __, __, __, // 2 |
| 799 |
__, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // 3 |
| 800 |
__, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // 4 |
| 801 |
__, __, __, __, __, __, __, __, __, __, __, __, BS, __, __, __, // 5 |
| 802 |
__, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // 6 |
| 803 |
__, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // 7 |
| 804 |
__, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // 8 |
| 805 |
__, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // 9 |
| 806 |
__, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // A |
| 807 |
__, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // B |
| 808 |
__, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // C |
| 809 |
__, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // D |
| 810 |
__, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // E |
| 811 |
__, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // F |
| 812 |
] |
| 813 |
}; |
| 814 |
|
| 815 |
fn next_or_eof<'de, R>(read: &mut R) -> Result<u8> |
| 816 |
where |
| 817 |
R: ?Sized + Read<'de>, |
| 818 |
{ |
| 819 |
match tri!(read.next()) { |
| 820 |
Some(b) => Ok(b), |
| 821 |
None => error(read, ErrorCode::EofWhileParsingString), |
| 822 |
} |
| 823 |
} |
| 824 |
|
| 825 |
fn peek_or_eof<'de, R>(read: &mut R) -> Result<u8> |
| 826 |
where |
| 827 |
R: ?Sized + Read<'de>, |
| 828 |
{ |
| 829 |
match tri!(read.peek()) { |
| 830 |
Some(b) => Ok(b), |
| 831 |
None => error(read, ErrorCode::EofWhileParsingString), |
| 832 |
} |
| 833 |
} |
| 834 |
|
| 835 |
fn error<'de, R, T>(read: &R, reason: ErrorCode) -> Result<T> |
| 836 |
where |
| 837 |
R: ?Sized + Read<'de>, |
| 838 |
{ |
| 839 |
let position = read.position(); |
| 840 |
Err(Error::syntax(reason, position.line, position.column)) |
| 841 |
} |
| 842 |
|
| 843 |
fn as_str<'de, 's, R: Read<'de>>(read: &R, slice: &'s [u8]) -> Result<&'s str> { |
| 844 |
str::from_utf8(slice).or_else(|_| error(read, ErrorCode::InvalidUnicodeCodePoint)) |
| 845 |
} |
| 846 |
|
| 847 |
/// Parses a JSON escape sequence and appends it into the scratch space. Assumes |
| 848 |
/// the previous byte read was a backslash. |
| 849 |
fn parse_escape<'de, R: Read<'de>>( |
| 850 |
read: &mut R, |
| 851 |
validate: bool, |
| 852 |
scratch: &mut Vec<u8>, |
| 853 |
) -> Result<()> { |
| 854 |
let ch = tri!(next_or_eof(read)); |
| 855 |
|
| 856 |
match ch { |
| 857 |
b'"' => scratch.push(b'"'), |
| 858 |
b'\\' => scratch.push(b'\\'), |
| 859 |
b'/' => scratch.push(b'/'), |
| 860 |
b'b' => scratch.push(b'\x08'), |
| 861 |
b'f' => scratch.push(b'\x0c'), |
| 862 |
b'n' => scratch.push(b'\n'), |
| 863 |
b'r' => scratch.push(b'\r'), |
| 864 |
b't' => scratch.push(b'\t'), |
| 865 |
b'u' => { |
| 866 |
fn encode_surrogate(scratch: &mut Vec<u8>, n: u16) { |
| 867 |
scratch.extend_from_slice(&[ |
| 868 |
(n >> 12 & 0b0000_1111) as u8 | 0b1110_0000, |
| 869 |
(n >> 6 & 0b0011_1111) as u8 | 0b1000_0000, |
| 870 |
(n & 0b0011_1111) as u8 | 0b1000_0000, |
| 871 |
]); |
| 872 |
} |
| 873 |
|
| 874 |
let c = match tri!(read.decode_hex_escape()) { |
| 875 |
n @ 0xDC00..=0xDFFF => { |
| 876 |
return if validate { |
| 877 |
error(read, ErrorCode::LoneLeadingSurrogateInHexEscape) |
| 878 |
} else { |
| 879 |
encode_surrogate(scratch, n); |
| 880 |
Ok(()) |
| 881 |
}; |
| 882 |
} |
| 883 |
|
| 884 |
// Non-BMP characters are encoded as a sequence of two hex |
| 885 |
// escapes, representing UTF-16 surrogates. If deserializing a |
| 886 |
// utf-8 string the surrogates are required to be paired, |
| 887 |
// whereas deserializing a byte string accepts lone surrogates. |
| 888 |
n1 @ 0xD800..=0xDBFF => { |
| 889 |
if tri!(peek_or_eof(read)) == b'\\' { |
| 890 |
read.discard(); |
| 891 |
} else { |
| 892 |
return if validate { |
| 893 |
read.discard(); |
| 894 |
error(read, ErrorCode::UnexpectedEndOfHexEscape) |
| 895 |
} else { |
| 896 |
encode_surrogate(scratch, n1); |
| 897 |
Ok(()) |
| 898 |
}; |
| 899 |
} |
| 900 |
|
| 901 |
if tri!(peek_or_eof(read)) == b'u' { |
| 902 |
read.discard(); |
| 903 |
} else { |
| 904 |
return if validate { |
| 905 |
read.discard(); |
| 906 |
error(read, ErrorCode::UnexpectedEndOfHexEscape) |
| 907 |
} else { |
| 908 |
encode_surrogate(scratch, n1); |
| 909 |
// The \ prior to this byte started an escape sequence, |
| 910 |
// so we need to parse that now. This recursive call |
| 911 |
// does not blow the stack on malicious input because |
| 912 |
// the escape is not \u, so it will be handled by one |
| 913 |
// of the easy nonrecursive cases. |
| 914 |
parse_escape(read, validate, scratch) |
| 915 |
}; |
| 916 |
} |
| 917 |
|
| 918 |
let n2 = tri!(read.decode_hex_escape()); |
| 919 |
|
| 920 |
if n2 < 0xDC00 || n2 > 0xDFFF { |
| 921 |
return error(read, ErrorCode::LoneLeadingSurrogateInHexEscape); |
| 922 |
} |
| 923 |
|
| 924 |
let n = (((n1 - 0xD800) as u32) << 10 | (n2 - 0xDC00) as u32) + 0x1_0000; |
| 925 |
|
| 926 |
match char::from_u32(n) { |
| 927 |
Some(c) => c, |
| 928 |
None => { |
| 929 |
return error(read, ErrorCode::InvalidUnicodeCodePoint); |
| 930 |
} |
| 931 |
} |
| 932 |
} |
| 933 |
|
| 934 |
// Every u16 outside of the surrogate ranges above is guaranteed |
| 935 |
// to be a legal char. |
| 936 |
n => char::from_u32(n as u32).unwrap(), |
| 937 |
}; |
| 938 |
|
| 939 |
scratch.extend_from_slice(c.encode_utf8(&mut [0_u8; 4]).as_bytes()); |
| 940 |
} |
| 941 |
_ => { |
| 942 |
return error(read, ErrorCode::InvalidEscape); |
| 943 |
} |
| 944 |
} |
| 945 |
|
| 946 |
Ok(()) |
| 947 |
} |
| 948 |
|
| 949 |
/// Parses a JSON escape sequence and discards the value. Assumes the previous |
| 950 |
/// byte read was a backslash. |
| 951 |
fn ignore_escape<'de, R>(read: &mut R) -> Result<()> |
| 952 |
where |
| 953 |
R: ?Sized + Read<'de>, |
| 954 |
{ |
| 955 |
let ch = tri!(next_or_eof(read)); |
| 956 |
|
| 957 |
match ch { |
| 958 |
b'"' | b'\\' | b'/' | b'b' | b'f' | b'n' | b'r' | b't' => {} |
| 959 |
b'u' => { |
| 960 |
// At this point we don't care if the codepoint is valid. We just |
| 961 |
// want to consume it. We don't actually know what is valid or not |
| 962 |
// at this point, because that depends on if this string will |
| 963 |
// ultimately be parsed into a string or a byte buffer in the "real" |
| 964 |
// parse. |
| 965 |
|
| 966 |
tri!(read.decode_hex_escape()); |
| 967 |
} |
| 968 |
_ => { |
| 969 |
return error(read, ErrorCode::InvalidEscape); |
| 970 |
} |
| 971 |
} |
| 972 |
|
| 973 |
Ok(()) |
| 974 |
} |
| 975 |
|
| 976 |
static HEX: [u8; 256] = { |
| 977 |
const __: u8 = 255; // not a hex digit |
| 978 |
[ |
| 979 |
// 1 2 3 4 5 6 7 8 9 A B C D E F |
| 980 |
__, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // 0 |
| 981 |
__, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // 1 |
| 982 |
__, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // 2 |
| 983 |
00, 01, 02, 03, 04, 05, 06, 07, 08, 09, __, __, __, __, __, __, // 3 |
| 984 |
__, 10, 11, 12, 13, 14, 15, __, __, __, __, __, __, __, __, __, // 4 |
| 985 |
__, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // 5 |
| 986 |
__, 10, 11, 12, 13, 14, 15, __, __, __, __, __, __, __, __, __, // 6 |
| 987 |
__, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // 7 |
| 988 |
__, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // 8 |
| 989 |
__, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // 9 |
| 990 |
__, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // A |
| 991 |
__, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // B |
| 992 |
__, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // C |
| 993 |
__, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // D |
| 994 |
__, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // E |
| 995 |
__, __, __, __, __, __, __, __, __, __, __, __, __, __, __, __, // F |
| 996 |
] |
| 997 |
}; |
| 998 |
|
| 999 |
fn decode_hex_val(val: u8) -> Option<u16> { |
| 1000 |
let n = HEX[val as usize] as u16; |
| 1001 |
if n == 255 { |
| 1002 |
None |
| 1003 |
} else { |
| 1004 |
Some(n) |
| 1005 |
} |
| 1006 |
} |