From e3fc97be2b90add55bbe48362cba902c6a0675be Mon Sep 17 00:00:00 2001 From: Markus Reiter Date: Tue, 23 Apr 2024 22:16:29 +0200 Subject: [PATCH 1/3] Inline `EscapeDebug::size_hint`. --- library/core/src/char/mod.rs | 1 + 1 file changed, 1 insertion(+) diff --git a/library/core/src/char/mod.rs b/library/core/src/char/mod.rs index a860c7c6aaadc..bf9f633eef370 100644 --- a/library/core/src/char/mod.rs +++ b/library/core/src/char/mod.rs @@ -339,6 +339,7 @@ impl Iterator for EscapeDebug { } } + #[inline] fn size_hint(&self) -> (usize, Option) { let n = self.len(); (n, Some(n)) From 16981ba40607ba87bc3767c0e810622455dc0d85 Mon Sep 17 00:00:00 2001 From: Markus Reiter Date: Thu, 25 Apr 2024 01:26:32 +0200 Subject: [PATCH 2/3] Avoid panicking branch in `EscapeIterInner`. --- library/core/src/ascii.rs | 14 +-- library/core/src/char/methods.rs | 8 +- library/core/src/char/mod.rs | 48 +++++----- library/core/src/escape.rs | 148 ++++++++++++++++++++----------- 4 files changed, 137 insertions(+), 81 deletions(-) diff --git a/library/core/src/ascii.rs b/library/core/src/ascii.rs index c29e5565d514a..17fbb19d0397a 100644 --- a/library/core/src/ascii.rs +++ b/library/core/src/ascii.rs @@ -91,17 +91,21 @@ pub struct EscapeDefault(escape::EscapeIterInner<4>); /// ``` #[stable(feature = "rust1", since = "1.0.0")] pub fn escape_default(c: u8) -> EscapeDefault { - let mut data = [Char::Null; 4]; - let range = escape::escape_ascii_into(&mut data, c); - EscapeDefault(escape::EscapeIterInner::new(data, range)) + EscapeDefault::new(c) } impl EscapeDefault { + #[inline] + pub(crate) const fn new(c: u8) -> Self { + Self(escape::EscapeIterInner::ascii(c)) + } + + #[inline] pub(crate) fn empty() -> Self { - let data = [Char::Null; 4]; - EscapeDefault(escape::EscapeIterInner::new(data, 0..0)) + EscapeDefault(escape::EscapeIterInner::empty()) } + #[inline] pub(crate) fn as_str(&self) -> &str { self.0.as_str() } diff --git a/library/core/src/char/methods.rs b/library/core/src/char/methods.rs index a93b94867ce4c..458be49fb152a 100644 --- a/library/core/src/char/methods.rs +++ b/library/core/src/char/methods.rs @@ -449,10 +449,10 @@ impl char { '\"' if args.escape_double_quote => EscapeDebug::backslash(ascii::Char::QuotationMark), '\'' if args.escape_single_quote => EscapeDebug::backslash(ascii::Char::Apostrophe), _ if args.escape_grapheme_extended && self.is_grapheme_extended() => { - EscapeDebug::from_unicode(self.escape_unicode()) + EscapeDebug::unicode(self) } _ if is_printable(self) => EscapeDebug::printable(self), - _ => EscapeDebug::from_unicode(self.escape_unicode()), + _ => EscapeDebug::unicode(self), } } @@ -555,9 +555,9 @@ impl char { '\t' => EscapeDefault::backslash(ascii::Char::SmallT), '\r' => EscapeDefault::backslash(ascii::Char::SmallR), '\n' => EscapeDefault::backslash(ascii::Char::SmallN), - '\\' | '\'' | '"' => EscapeDefault::backslash(self.as_ascii().unwrap()), + '\\' | '\'' | '\"' => EscapeDefault::backslash(self.as_ascii().unwrap()), '\x20'..='\x7e' => EscapeDefault::printable(self.as_ascii().unwrap()), - _ => EscapeDefault::from_unicode(self.escape_unicode()), + _ => EscapeDefault::unicode(self), } } diff --git a/library/core/src/char/mod.rs b/library/core/src/char/mod.rs index bf9f633eef370..d99ac226da005 100644 --- a/library/core/src/char/mod.rs +++ b/library/core/src/char/mod.rs @@ -152,10 +152,9 @@ pub const fn from_digit(num: u32, radix: u32) -> Option { pub struct EscapeUnicode(escape::EscapeIterInner<10>); impl EscapeUnicode { - fn new(chr: char) -> Self { - let mut data = [ascii::Char::Null; 10]; - let range = escape::escape_unicode_into(&mut data, chr); - Self(escape::EscapeIterInner::new(data, range)) + #[inline] + const fn new(c: char) -> Self { + Self(escape::EscapeIterInner::unicode(c)) } } @@ -219,18 +218,24 @@ impl fmt::Display for EscapeUnicode { pub struct EscapeDefault(escape::EscapeIterInner<10>); impl EscapeDefault { - fn printable(chr: ascii::Char) -> Self { - let data = [chr]; - Self(escape::EscapeIterInner::from_array(data)) + #[inline] + const fn printable(c: ascii::Char) -> Self { + Self::ascii(c.to_u8()) } - fn backslash(chr: ascii::Char) -> Self { - let data = [ascii::Char::ReverseSolidus, chr]; - Self(escape::EscapeIterInner::from_array(data)) + #[inline] + const fn backslash(c: ascii::Char) -> Self { + Self(escape::EscapeIterInner::backslash(c)) } - fn from_unicode(esc: EscapeUnicode) -> Self { - Self(esc.0) + #[inline] + const fn ascii(c: u8) -> Self { + Self(escape::EscapeIterInner::ascii(c)) + } + + #[inline] + const fn unicode(c: char) -> Self { + Self(escape::EscapeIterInner::unicode(c)) } } @@ -304,23 +309,24 @@ enum EscapeDebugInner { } impl EscapeDebug { - fn printable(chr: char) -> Self { + #[inline] + const fn printable(chr: char) -> Self { Self(EscapeDebugInner::Char(chr)) } - fn backslash(chr: ascii::Char) -> Self { - let data = [ascii::Char::ReverseSolidus, chr]; - let iter = escape::EscapeIterInner::from_array(data); - Self(EscapeDebugInner::Bytes(iter)) + #[inline] + const fn backslash(c: ascii::Char) -> Self { + Self(EscapeDebugInner::Bytes(escape::EscapeIterInner::backslash(c))) } - fn from_unicode(esc: EscapeUnicode) -> Self { - Self(EscapeDebugInner::Bytes(esc.0)) + #[inline] + const fn unicode(c: char) -> Self { + Self(EscapeDebugInner::Bytes(escape::EscapeIterInner::unicode(c))) } + #[inline] fn clear(&mut self) { - let bytes = escape::EscapeIterInner::from_array([]); - self.0 = EscapeDebugInner::Bytes(bytes); + self.0 = EscapeDebugInner::Bytes(escape::EscapeIterInner::empty()); } } diff --git a/library/core/src/escape.rs b/library/core/src/escape.rs index 143e277283e2c..a71fe59e50a13 100644 --- a/library/core/src/escape.rs +++ b/library/core/src/escape.rs @@ -6,56 +6,85 @@ use crate::ops::Range; const HEX_DIGITS: [ascii::Char; 16] = *b"0123456789abcdef".as_ascii().unwrap(); -/// Escapes a byte into provided buffer; returns length of escaped -/// representation. -pub(crate) fn escape_ascii_into(output: &mut [ascii::Char; 4], byte: u8) -> Range { - #[inline] - fn backslash(a: ascii::Char) -> ([ascii::Char; 4], u8) { - ([ascii::Char::ReverseSolidus, a, ascii::Char::Null, ascii::Char::Null], 2) - } +#[inline] +const fn backslash(a: ascii::Char) -> ([ascii::Char; N], u8) { + const { assert!(N >= 2) }; + + let mut output = [ascii::Char::Null; N]; + + output[0] = ascii::Char::ReverseSolidus; + output[1] = a; + + (output, 2) +} + +/// Escapes an ASCII character. +/// +/// Returns a buffer and the length of the escaped representation. +const fn escape_ascii(byte: u8) -> ([ascii::Char; N], u8) { + const { assert!(N >= 4) }; - let (data, len) = match byte { + match byte { b'\t' => backslash(ascii::Char::SmallT), b'\r' => backslash(ascii::Char::SmallR), b'\n' => backslash(ascii::Char::SmallN), b'\\' => backslash(ascii::Char::ReverseSolidus), b'\'' => backslash(ascii::Char::Apostrophe), b'\"' => backslash(ascii::Char::QuotationMark), - _ => { - if let Some(a) = byte.as_ascii() + byte => { + let mut output = [ascii::Char::Null; N]; + + if let Some(c) = byte.as_ascii() && !byte.is_ascii_control() { - ([a, ascii::Char::Null, ascii::Char::Null, ascii::Char::Null], 1) + output[0] = c; + (output, 1) } else { - let hi = HEX_DIGITS[usize::from(byte >> 4)]; - let lo = HEX_DIGITS[usize::from(byte & 0xf)]; - ([ascii::Char::ReverseSolidus, ascii::Char::SmallX, hi, lo], 4) + let hi = HEX_DIGITS[(byte >> 4) as usize]; + let lo = HEX_DIGITS[(byte & 0xf) as usize]; + + output[0] = ascii::Char::ReverseSolidus; + output[1] = ascii::Char::SmallX; + output[2] = hi; + output[3] = lo; + + (output, 4) } } - }; - *output = data; - 0..len + } } -/// Escapes a character into provided buffer using `\u{NNNN}` representation. -pub(crate) fn escape_unicode_into(output: &mut [ascii::Char; 10], ch: char) -> Range { - output[9] = ascii::Char::RightCurlyBracket; - - let ch = ch as u32; - output[3] = HEX_DIGITS[((ch >> 20) & 15) as usize]; - output[4] = HEX_DIGITS[((ch >> 16) & 15) as usize]; - output[5] = HEX_DIGITS[((ch >> 12) & 15) as usize]; - output[6] = HEX_DIGITS[((ch >> 8) & 15) as usize]; - output[7] = HEX_DIGITS[((ch >> 4) & 15) as usize]; - output[8] = HEX_DIGITS[((ch >> 0) & 15) as usize]; - - // or-ing 1 ensures that for ch==0 the code computes that one digit should - // be printed. - let start = (ch | 1).leading_zeros() as usize / 4 - 2; - const UNICODE_ESCAPE_PREFIX: &[ascii::Char; 3] = b"\\u{".as_ascii().unwrap(); - output[start..][..3].copy_from_slice(UNICODE_ESCAPE_PREFIX); - - (start as u8)..10 +/// Escapes a character `\u{NNNN}` representation. +/// +/// Returns a buffer and the length of the escaped representation. +const fn escape_unicode(c: char) -> ([ascii::Char; N], u8) { + const { assert!(N >= 10) }; + + let c = c as u32; + + // OR-ing `1` ensures that for `c == 0` the code computes that + // one digit should be printed. + let u_len = (8 - (c | 1).leading_zeros() / 4) as usize; + + let closing_paren_offset = 3 + u_len; + + let mut output = [ascii::Char::Null; N]; + + output[0] = ascii::Char::ReverseSolidus; + output[1] = ascii::Char::SmallU; + output[2] = ascii::Char::LeftCurlyBracket; + + output[3 + u_len.saturating_sub(6)] = HEX_DIGITS[((c >> 20) & 0x0f) as usize]; + output[3 + u_len.saturating_sub(5)] = HEX_DIGITS[((c >> 16) & 0x0f) as usize]; + output[3 + u_len.saturating_sub(4)] = HEX_DIGITS[((c >> 12) & 0x0f) as usize]; + output[3 + u_len.saturating_sub(3)] = HEX_DIGITS[((c >> 8) & 0x0f) as usize]; + output[3 + u_len.saturating_sub(2)] = HEX_DIGITS[((c >> 4) & 0x0f) as usize]; + output[3 + u_len.saturating_sub(1)] = HEX_DIGITS[((c >> 0) & 0x0f) as usize]; + + output[closing_paren_offset] = ascii::Char::RightCurlyBracket; + + let len = (closing_paren_offset + 1) as u8; + (output, len) } /// An iterator over an fixed-size array. @@ -65,45 +94,62 @@ pub(crate) fn escape_unicode_into(output: &mut [ascii::Char; 10], ch: char) -> R #[derive(Clone, Debug)] pub(crate) struct EscapeIterInner { // The element type ensures this is always ASCII, and thus also valid UTF-8. - pub(crate) data: [ascii::Char; N], + data: [ascii::Char; N], - // Invariant: alive.start <= alive.end <= N. - pub(crate) alive: Range, + // Invariant: `alive.start <= alive.end <= N` + alive: Range, } impl EscapeIterInner { - pub fn new(data: [ascii::Char; N], alive: Range) -> Self { - const { assert!(N < 256) }; - debug_assert!(alive.start <= alive.end && usize::from(alive.end) <= N, "{alive:?}"); - Self { data, alive } + pub const fn backslash(c: ascii::Char) -> Self { + let (data, len) = backslash(c); + Self { data, alive: 0..len } } - pub fn from_array(array: [ascii::Char; M]) -> Self { - const { assert!(M <= N) }; + pub const fn ascii(c: u8) -> Self { + let (data, len) = escape_ascii(c); + Self { data, alive: 0..len } + } - let mut data = [ascii::Char::Null; N]; - data[..M].copy_from_slice(&array); - Self::new(data, 0..M as u8) + pub const fn unicode(c: char) -> Self { + let (data, len) = escape_unicode(c); + Self { data, alive: 0..len } + } + + #[inline] + pub const fn empty() -> Self { + Self { data: [ascii::Char::Null; N], alive: 0..0 } } pub fn as_ascii(&self) -> &[ascii::Char] { - &self.data[usize::from(self.alive.start)..usize::from(self.alive.end)] + // SAFETY: `self.alive` is guaranteed to be a valid range for indexing `self.data`. + unsafe { + self.data.get_unchecked(usize::from(self.alive.start)..usize::from(self.alive.end)) + } } + #[inline] pub fn as_str(&self) -> &str { self.as_ascii().as_str() } + #[inline] pub fn len(&self) -> usize { usize::from(self.alive.end - self.alive.start) } pub fn next(&mut self) -> Option { - self.alive.next().map(|i| self.data[usize::from(i)].to_u8()) + let i = self.alive.next()?; + + // SAFETY: `i` is guaranteed to be a valid index for `self.data`. + unsafe { Some(self.data.get_unchecked(usize::from(i)).to_u8()) } } pub fn next_back(&mut self) -> Option { - self.alive.next_back().map(|i| self.data[usize::from(i)].to_u8()) + let i = self.alive.next_back()?; + + // SAFETY: `i` is guaranteed to be a valid index for `self.data`. + unsafe { Some(self.data.get_unchecked(usize::from(i)).to_u8()) } } pub fn advance_by(&mut self, n: usize) -> Result<(), NonZero> { From 4edf12d33e0aa67a82e6587e94469f3761cb1f53 Mon Sep 17 00:00:00 2001 From: Markus Reiter Date: Wed, 1 May 2024 16:14:37 +0200 Subject: [PATCH 3/3] Improve escape methods. --- library/core/src/ascii.rs | 2 +- library/core/src/char/mod.rs | 7 +---- library/core/src/escape.rs | 61 +++++++++++++++++------------------- 3 files changed, 30 insertions(+), 40 deletions(-) diff --git a/library/core/src/ascii.rs b/library/core/src/ascii.rs index 17fbb19d0397a..e9f4d0f93ed49 100644 --- a/library/core/src/ascii.rs +++ b/library/core/src/ascii.rs @@ -102,7 +102,7 @@ impl EscapeDefault { #[inline] pub(crate) fn empty() -> Self { - EscapeDefault(escape::EscapeIterInner::empty()) + Self(escape::EscapeIterInner::empty()) } #[inline] diff --git a/library/core/src/char/mod.rs b/library/core/src/char/mod.rs index d99ac226da005..f3683fe3f9c83 100644 --- a/library/core/src/char/mod.rs +++ b/library/core/src/char/mod.rs @@ -220,7 +220,7 @@ pub struct EscapeDefault(escape::EscapeIterInner<10>); impl EscapeDefault { #[inline] const fn printable(c: ascii::Char) -> Self { - Self::ascii(c.to_u8()) + Self(escape::EscapeIterInner::ascii(c.to_u8())) } #[inline] @@ -228,11 +228,6 @@ impl EscapeDefault { Self(escape::EscapeIterInner::backslash(c)) } - #[inline] - const fn ascii(c: u8) -> Self { - Self(escape::EscapeIterInner::ascii(c)) - } - #[inline] const fn unicode(c: char) -> Self { Self(escape::EscapeIterInner::unicode(c)) diff --git a/library/core/src/escape.rs b/library/core/src/escape.rs index a71fe59e50a13..f6ec30b9f793a 100644 --- a/library/core/src/escape.rs +++ b/library/core/src/escape.rs @@ -7,7 +7,7 @@ use crate::ops::Range; const HEX_DIGITS: [ascii::Char; 16] = *b"0123456789abcdef".as_ascii().unwrap(); #[inline] -const fn backslash(a: ascii::Char) -> ([ascii::Char; N], u8) { +const fn backslash(a: ascii::Char) -> ([ascii::Char; N], Range) { const { assert!(N >= 2) }; let mut output = [ascii::Char::Null; N]; @@ -15,13 +15,13 @@ const fn backslash(a: ascii::Char) -> ([ascii::Char; N], u8) { output[0] = ascii::Char::ReverseSolidus; output[1] = a; - (output, 2) + (output, 0..2) } /// Escapes an ASCII character. /// /// Returns a buffer and the length of the escaped representation. -const fn escape_ascii(byte: u8) -> ([ascii::Char; N], u8) { +const fn escape_ascii(byte: u8) -> ([ascii::Char; N], Range) { const { assert!(N >= 4) }; match byte { @@ -38,7 +38,7 @@ const fn escape_ascii(byte: u8) -> ([ascii::Char; N], u8) { && !byte.is_ascii_control() { output[0] = c; - (output, 1) + (output, 0..1) } else { let hi = HEX_DIGITS[(byte >> 4) as usize]; let lo = HEX_DIGITS[(byte & 0xf) as usize]; @@ -48,7 +48,7 @@ const fn escape_ascii(byte: u8) -> ([ascii::Char; N], u8) { output[2] = hi; output[3] = lo; - (output, 4) + (output, 0..4) } } } @@ -57,34 +57,28 @@ const fn escape_ascii(byte: u8) -> ([ascii::Char; N], u8) { /// Escapes a character `\u{NNNN}` representation. /// /// Returns a buffer and the length of the escaped representation. -const fn escape_unicode(c: char) -> ([ascii::Char; N], u8) { - const { assert!(N >= 10) }; +const fn escape_unicode(c: char) -> ([ascii::Char; N], Range) { + const { assert!(N >= 10 && N < u8::MAX as usize) }; - let c = c as u32; + let c = u32::from(c); // OR-ing `1` ensures that for `c == 0` the code computes that // one digit should be printed. - let u_len = (8 - (c | 1).leading_zeros() / 4) as usize; - - let closing_paren_offset = 3 + u_len; + let start = (c | 1).leading_zeros() as usize / 4 - 2; let mut output = [ascii::Char::Null; N]; - - output[0] = ascii::Char::ReverseSolidus; - output[1] = ascii::Char::SmallU; - output[2] = ascii::Char::LeftCurlyBracket; - - output[3 + u_len.saturating_sub(6)] = HEX_DIGITS[((c >> 20) & 0x0f) as usize]; - output[3 + u_len.saturating_sub(5)] = HEX_DIGITS[((c >> 16) & 0x0f) as usize]; - output[3 + u_len.saturating_sub(4)] = HEX_DIGITS[((c >> 12) & 0x0f) as usize]; - output[3 + u_len.saturating_sub(3)] = HEX_DIGITS[((c >> 8) & 0x0f) as usize]; - output[3 + u_len.saturating_sub(2)] = HEX_DIGITS[((c >> 4) & 0x0f) as usize]; - output[3 + u_len.saturating_sub(1)] = HEX_DIGITS[((c >> 0) & 0x0f) as usize]; - - output[closing_paren_offset] = ascii::Char::RightCurlyBracket; - - let len = (closing_paren_offset + 1) as u8; - (output, len) + output[3] = HEX_DIGITS[((c >> 20) & 15) as usize]; + output[4] = HEX_DIGITS[((c >> 16) & 15) as usize]; + output[5] = HEX_DIGITS[((c >> 12) & 15) as usize]; + output[6] = HEX_DIGITS[((c >> 8) & 15) as usize]; + output[7] = HEX_DIGITS[((c >> 4) & 15) as usize]; + output[8] = HEX_DIGITS[((c >> 0) & 15) as usize]; + output[9] = ascii::Char::RightCurlyBracket; + output[start + 0] = ascii::Char::ReverseSolidus; + output[start + 1] = ascii::Char::SmallU; + output[start + 2] = ascii::Char::LeftCurlyBracket; + + (output, (start as u8)..(N as u8)) } /// An iterator over an fixed-size array. @@ -102,18 +96,18 @@ pub(crate) struct EscapeIterInner { impl EscapeIterInner { pub const fn backslash(c: ascii::Char) -> Self { - let (data, len) = backslash(c); - Self { data, alive: 0..len } + let (data, range) = backslash(c); + Self { data, alive: range } } pub const fn ascii(c: u8) -> Self { - let (data, len) = escape_ascii(c); - Self { data, alive: 0..len } + let (data, range) = escape_ascii(c); + Self { data, alive: range } } pub const fn unicode(c: char) -> Self { - let (data, len) = escape_unicode(c); - Self { data, alive: 0..len } + let (data, range) = escape_unicode(c); + Self { data, alive: range } } #[inline] @@ -121,6 +115,7 @@ impl EscapeIterInner { Self { data: [ascii::Char::Null; N], alive: 0..0 } } + #[inline] pub fn as_ascii(&self) -> &[ascii::Char] { // SAFETY: `self.alive` is guaranteed to be a valid range for indexing `self.data`. unsafe {