itsscb/json
1
0
Fork 0
mirror of https://github.com/serde-rs/json.git synced 2025-10-04 00:06:04 +00:00
Code Issues Packages Projects Releases Wiki Activity

Simplify unicode escape handling

Browse Source 
This does not affect performance.
This commit is contained in:
Alisa Sireneva 2024-08-12 19:53:51 +03:00
parent 2f28d106e6
commit 236cc8247d
1 changed files with 53 additions and 54 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

107
src/read.rs
View File

@ -898,67 +898,66 @@ fn parse_unicode_escape<'de, R: Read<'de>>(
validate: bool, validate: bool,
scratch: &mut Vec<u8>, scratch: &mut Vec<u8>,
) -> Result<()> { ) -> Result<()> {
let c = match tri!(read.decode_hex_escape()) { let n = tri!(read.decode_hex_escape());
n @ 0xDC00..=0xDFFF => {
return if validate {
error(read, ErrorCode::LoneLeadingSurrogateInHexEscape)
} else {
push_wtf8_codepoint(n as u32, scratch);
Ok(())
};
}
// Non-BMP characters are encoded as a sequence of two hex // Non-BMP characters are encoded as a sequence of two hex
// escapes, representing UTF-16 surrogates. If deserializing a // escapes, representing UTF-16 surrogates. If deserializing a
// utf-8 string the surrogates are required to be paired, // utf-8 string the surrogates are required to be paired,
// whereas deserializing a byte string accepts lone surrogates. // whereas deserializing a byte string accepts lone surrogates.
n1 @ 0xD800..=0xDBFF => { if validate && n >= 0xDC00 && n <= 0xDFFF {
if tri!(peek_or_eof(read)) == b'\\' { // XXX: This is actually a trailing surrogate.
read.discard(); return error(read, ErrorCode::LoneLeadingSurrogateInHexEscape);
} else { }
return if validate {
read.discard();
error(read, ErrorCode::UnexpectedEndOfHexEscape)
} else {
push_wtf8_codepoint(n1 as u32, scratch);
Ok(())
};
}
if tri!(peek_or_eof(read)) == b'u' { if n < 0xD800 || n > 0xDBFF {
read.discard(); // Every u16 outside of the surrogate ranges is guaranteed to be a
} else { // legal char.
return if validate { push_wtf8_codepoint(n as u32, scratch);
read.discard(); return Ok(());
error(read, ErrorCode::UnexpectedEndOfHexEscape) }
} else {
push_wtf8_codepoint(n1 as u32, scratch);
// The \ prior to this byte started an escape sequence,
// so we need to parse that now. This recursive call
// does not blow the stack on malicious input because
// the escape is not \u, so it will be handled by one
// of the easy nonrecursive cases.
parse_escape(read, validate, scratch)
};
}
let n2 = tri!(read.decode_hex_escape()); // n is a leading surrogate, we now expect a trailing surrogate.
let n1 = n;
if n2 < 0xDC00 || n2 > 0xDFFF { if tri!(peek_or_eof(read)) == b'\\' {
return error(read, ErrorCode::LoneLeadingSurrogateInHexEscape); read.discard();
} } else {
return if validate {
read.discard();
error(read, ErrorCode::UnexpectedEndOfHexEscape)
} else {
push_wtf8_codepoint(n1 as u32, scratch);
Ok(())
};
}
// This value is in range U+10000..=U+10FFFF, which is always a if tri!(peek_or_eof(read)) == b'u' {
// valid codepoint. read.discard();
(((n1 - 0xD800) as u32) << 10 | (n2 - 0xDC00) as u32) + 0x1_0000 } else {
} return if validate {
read.discard();
error(read, ErrorCode::UnexpectedEndOfHexEscape)
} else {
push_wtf8_codepoint(n1 as u32, scratch);
// The \ prior to this byte started an escape sequence,
// so we need to parse that now. This recursive call
// does not blow the stack on malicious input because
// the escape is not \u, so it will be handled by one
// of the easy nonrecursive cases.
parse_escape(read, validate, scratch)
};
}
// Every u16 outside of the surrogate ranges above is guaranteed let n2 = tri!(read.decode_hex_escape());
// to be a legal char.
n => n as u32,
};
push_wtf8_codepoint(c, scratch); if n2 < 0xDC00 || n2 > 0xDFFF {
return error(read, ErrorCode::LoneLeadingSurrogateInHexEscape);
}
// This value is in range U+10000..=U+10FFFF, which is always a
// valid codepoint.
let n = (((n1 - 0xD800) as u32) << 10 | (n2 - 0xDC00) as u32) + 0x1_0000;
push_wtf8_codepoint(n, scratch);
Ok(()) Ok(())
} }