optimization: Eliminate Cased table

`Cased` is a derived property - it is the union of the `Lowercase`
property, the `Uppercase` property, and the `Titlecase_Letter` general
categories. We already have lookup tables for `Lowercase` and
`Uppercase`, and `Titlecase_Letter` is very small. So instead of
duplicating a lookup table for `Cased`, just test each of those
properties in turn.

This probably will be slower than the old approach, but it is not a
public API: it is only used in `string::to_lower` when deciding when a
Greek "sigma" should be mapped to `ς` or to `σ`. This is a very rare
case, so should not be performance sensitive.

Author: Kmeakin

library/alloc/src/str.rs

@@ -418,9 +418,8 @@ impl str {
         }
 
         fn case_ignorable_then_cased<I: Iterator<Item = char>>(iter: I) -> bool {
-            use core::unicode::{Case_Ignorable, Cased};
-            match iter.skip_while(|&c| Case_Ignorable(c)).next() {
-                Some(c) => Cased(c),
+            match iter.skip_while(|&c| c.is_case_ignorable()).next() {
+                Some(c) => c.is_cased(),
                 None => false,
             }
         }

library/core/src/char/methods.rs

@@ -6,7 +6,7 @@ use crate::slice;
 use crate::str::from_utf8_unchecked_mut;
 use crate::ub_checks::assert_unsafe_precondition;
 use crate::unicode::printable::is_printable;
-use crate::unicode::{self, conversions};
+use crate::unicode::{self, Case_Ignorable, conversions};
 
 impl char {
     /// The lowest valid code point a `char` can have, `'\0'`.
@@ -969,7 +969,47 @@ impl char {
     #[must_use]
     #[inline]
     pub(crate) fn is_grapheme_extended(self) -> bool {
-        unicode::Grapheme_Extend(self)
+        !self.is_ascii() && unicode::Grapheme_Extend(self)
+    }
+
+    /// Returns `true` if this `char` has the `Cased` derived property.
+    ///
+    /// `Cased` is described in Chapter 4 (Character Properties) of the [Unicode Standard] and
+    /// specified in the [Unicode Character Database][ucd] [`DerivedCoreProperties.txt`].
+    ///
+    /// [Unicode Standard]: https://www.unicode.org/versions/latest/
+    /// [ucd]: https://www.unicode.org/reports/tr44/
+    /// [`DerivedCoreProperties.txt`]: https://www.unicode.org/Public/UCD/latest/ucd/DerivedCoreProperties.txt
+    #[must_use]
+    #[inline]
+    #[doc(hidden)]
+    #[unstable(feature = "char_internals", reason = "exposed only for libstd", issue = "none")]
+    pub fn is_cased(self) -> bool {
+        if self.is_ascii() {
+            self.is_ascii_alphabetic()
+        } else {
+            unicode::Lowercase(self) || unicode::Uppercase(self) || unicode::Lt(self)
+        }
+    }
+
+    /// Returns `true` if this `char` has the `Case_Ignorable` property.
+    ///
+    /// `Case_Ignorable` is described in Chapter 4 (Character Properties) of the [Unicode Standard] and
+    /// specified in the [Unicode Character Database][ucd] [`DerivedCoreProperties.txt`].
+    ///
+    /// [Unicode Standard]: https://www.unicode.org/versions/latest/
+    /// [ucd]: https://www.unicode.org/reports/tr44/
+    /// [`DerivedCoreProperties.txt`]: https://www.unicode.org/Public/UCD/latest/ucd/DerivedCoreProperties.txt
+    #[must_use]
+    #[inline]
+    #[doc(hidden)]
+    #[unstable(feature = "char_internals", reason = "exposed only for libstd", issue = "none")]
+    pub fn is_case_ignorable(self) -> bool {
+        if self.is_ascii() {
+            matches!(self, '\'' | '.' | ':' | '^' | '`')
+        } else {
+            Case_Ignorable(self)
+        }
     }
 
     /// Returns `true` if this `char` has one of the general categories for numbers.

library/core/src/unicode/mod.rs

@@ -5,13 +5,13 @@
 // for use in alloc, not re-exported in std.
 #[rustfmt::skip]
 pub use unicode_data::case_ignorable::lookup as Case_Ignorable;
-pub use unicode_data::cased::lookup as Cased;
 pub use unicode_data::conversions;
 
 #[rustfmt::skip]
 pub(crate) use unicode_data::alphabetic::lookup as Alphabetic;
 pub(crate) use unicode_data::grapheme_extend::lookup as Grapheme_Extend;
 pub(crate) use unicode_data::lowercase::lookup as Lowercase;
+pub(crate) use unicode_data::lt::lookup as Lt;
 pub(crate) use unicode_data::n::lookup as N;
 pub(crate) use unicode_data::uppercase::lookup as Uppercase;
 pub(crate) use unicode_data::white_space::lookup as White_Space;

library/core/src/unicode/unicode_data.rs

@@ -1,16 +1,15 @@
 ///! This file is generated by `./x run src/tools/unicode-table-generator`; do not edit manually!
 // Alphabetic      :  1727 bytes, 142759 codepoints in 757 ranges (U+000041 - U+0323B0) using skiplist
 // Case_Ignorable  :  1053 bytes,   2749 codepoints in 452 ranges (U+000027 - U+0E01F0) using skiplist
-// Cased           :   407 bytes,   4578 codepoints in 159 ranges (U+000041 - U+01F18A) using skiplist
-// Cc              :     9 bytes,     65 codepoints in   2 ranges (U+000000 - U+0000A0) using skiplist
 // Grapheme_Extend :   887 bytes,   2193 codepoints in 375 ranges (U+000300 - U+0E01F0) using skiplist
 // Lowercase       :   935 bytes,   2569 codepoints in 675 ranges (U+000061 - U+01E944) using bitset
+// Lt              :    33 bytes,     31 codepoints in  10 ranges (U+0001C5 - U+001FFD) using skiplist
 // N               :   457 bytes,   1911 codepoints in 144 ranges (U+000030 - U+01FBFA) using skiplist
 // Uppercase       :   799 bytes,   1978 codepoints in 656 ranges (U+000041 - U+01F18A) using bitset
 // White_Space     :   256 bytes,     25 codepoints in  10 ranges (U+000009 - U+003001) using cascading
 // to_lower        : 11484 bytes
 // to_upper        : 13432 bytes
-// Total           : 31446 bytes
+// Total           : 31063 bytes
 
 #[inline(always)]
 const fn bitset_search<
@@ -324,52 +323,6 @@ pub mod case_ignorable {
     }
 }
 
-#[rustfmt::skip]
-pub mod cased {
-    use super::ShortOffsetRunHeader;
-
-    static SHORT_OFFSET_RUNS: [ShortOffsetRunHeader; 22] = [
-        ShortOffsetRunHeader::new(0, 4256), ShortOffsetRunHeader::new(55, 5024),
-        ShortOffsetRunHeader::new(65, 7296), ShortOffsetRunHeader::new(69, 7958),
-        ShortOffsetRunHeader::new(78, 9398), ShortOffsetRunHeader::new(153, 11264),
-        ShortOffsetRunHeader::new(155, 42560), ShortOffsetRunHeader::new(167, 43824),
-        ShortOffsetRunHeader::new(187, 64256), ShortOffsetRunHeader::new(193, 65313),
-        ShortOffsetRunHeader::new(197, 66560), ShortOffsetRunHeader::new(201, 67456),
-        ShortOffsetRunHeader::new(223, 68736), ShortOffsetRunHeader::new(231, 71840),
-        ShortOffsetRunHeader::new(239, 93760), ShortOffsetRunHeader::new(241, 119808),
-        ShortOffsetRunHeader::new(243, 120486), ShortOffsetRunHeader::new(280, 122624),
-        ShortOffsetRunHeader::new(303, 122928), ShortOffsetRunHeader::new(309, 125184),
-        ShortOffsetRunHeader::new(311, 127280), ShortOffsetRunHeader::new(313, 1241482),
-    ];
-    static OFFSETS: [u8; 319] = [
-        65, 26, 6, 26, 47, 1, 10, 1, 4, 1, 5, 23, 1, 31, 1, 195, 1, 4, 4, 208, 1, 36, 7, 2, 30, 5,
-        96, 1, 42, 4, 2, 2, 2, 4, 1, 1, 6, 1, 1, 3, 1, 1, 1, 20, 1, 83, 1, 139, 8, 166, 1, 38, 9,
-        41, 0, 38, 1, 1, 5, 1, 2, 43, 1, 4, 0, 86, 2, 6, 0, 11, 5, 43, 2, 3, 64, 192, 64, 0, 2, 6,
-        2, 38, 2, 6, 2, 8, 1, 1, 1, 1, 1, 1, 1, 31, 2, 53, 1, 7, 1, 1, 3, 3, 1, 7, 3, 4, 2, 6, 4,
-        13, 5, 3, 1, 7, 116, 1, 13, 1, 16, 13, 101, 1, 4, 1, 2, 10, 1, 1, 3, 5, 6, 1, 1, 1, 1, 1, 1,
-        4, 1, 6, 4, 1, 2, 4, 5, 5, 4, 1, 17, 32, 3, 2, 0, 52, 0, 229, 6, 4, 3, 2, 12, 38, 1, 1, 5,
-        1, 0, 46, 18, 30, 132, 102, 3, 4, 1, 62, 2, 2, 1, 1, 1, 8, 21, 5, 1, 3, 0, 43, 1, 14, 6, 80,
-        0, 7, 12, 5, 0, 26, 6, 26, 0, 80, 96, 36, 4, 36, 116, 11, 1, 15, 1, 7, 1, 2, 1, 11, 1, 15,
-        1, 7, 1, 2, 0, 1, 2, 3, 1, 42, 1, 9, 0, 51, 13, 51, 93, 22, 10, 22, 0, 64, 0, 64, 0, 85, 1,
-        71, 1, 2, 2, 1, 2, 2, 2, 4, 1, 12, 1, 1, 1, 7, 1, 65, 1, 4, 2, 8, 1, 7, 1, 28, 1, 4, 1, 5,
-        1, 1, 3, 7, 1, 0, 2, 25, 1, 25, 1, 31, 1, 25, 1, 31, 1, 25, 1, 31, 1, 25, 1, 31, 1, 25, 1,
-        8, 0, 10, 1, 20, 6, 6, 0, 62, 0, 68, 0, 26, 6, 26, 6, 26, 0,
-    ];
-    pub fn lookup(c: char) -> bool {
-        const {
-            assert!(SHORT_OFFSET_RUNS.last().unwrap().0 > char::MAX as u32);
-            let mut i = 0;
-            while i < SHORT_OFFSET_RUNS.len() {
-                assert!(SHORT_OFFSET_RUNS[i].start_index() < OFFSETS.len());
-                i += 1;
-            }
-        }
-        // SAFETY: We just ensured the last element of `SHORT_OFFSET_RUNS` is greater than `std::char::MAX`
-        // and the start indices of all elements in `SHORT_OFFSET_RUNS` are smaller than `OFFSETS.len()`.
-        unsafe { super::skip_search(c, &SHORT_OFFSET_RUNS, &OFFSETS) }
-    }
-}
-
 #[rustfmt::skip]
 pub mod grapheme_extend {
     use super::ShortOffsetRunHeader;
@@ -549,6 +502,38 @@ pub mod lowercase {
     }
 }
 
+#[rustfmt::skip]
+pub mod lt {
+    use super::ShortOffsetRunHeader;
+
+    static SHORT_OFFSET_RUNS: [ShortOffsetRunHeader; 3] = [
+        ShortOffsetRunHeader::new(0, 453), ShortOffsetRunHeader::new(1, 8072),
+        ShortOffsetRunHeader::new(9, 1122301),
+    ];
+    static OFFSETS: [u8; 21] = [
+        0, 1, 2, 1, 2, 1, 38, 1, 0, 8, 8, 8, 8, 8, 12, 1, 15, 1, 47, 1, 0,
+    ];
+    #[inline]
+    pub fn lookup(c: char) -> bool {
+        (c as u32) >= 0x1c5 && lookup_slow(c)
+    }
+
+    #[inline(never)]
+    fn lookup_slow(c: char) -> bool {
+        const {
+            assert!(SHORT_OFFSET_RUNS.last().unwrap().0 > char::MAX as u32);
+            let mut i = 0;
+            while i < SHORT_OFFSET_RUNS.len() {
+                assert!(SHORT_OFFSET_RUNS[i].start_index() < OFFSETS.len());
+                i += 1;
+            }
+        }
+        // SAFETY: We just ensured the last element of `SHORT_OFFSET_RUNS` is greater than `std::char::MAX`
+        // and the start indices of all elements in `SHORT_OFFSET_RUNS` are smaller than `OFFSETS.len()`.
+        unsafe { super::skip_search(c, &SHORT_OFFSET_RUNS, &OFFSETS) }
+    }
+}
+
 #[rustfmt::skip]
 pub mod n {
     use super::ShortOffsetRunHeader;

src/tools/unicode-table-generator/src/main.rs

@@ -90,7 +90,7 @@ static PROPERTIES: &[&str] = &[
     "Alphabetic",
     "Lowercase",
     "Uppercase",
-    "Cased",
+    "Lt",
     "Case_Ignorable",
     "Grapheme_Extend",
     "White_Space",