optimization: Eliminate Cased table

`Cased` is a derived property - it is the union of the `Lowercase`
property, the `Uppercase` property, and the `Titlecase_Letter` generaral
category. 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/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'`.
@@ -968,7 +968,7 @@ impl char {
         !self.is_ascii() && unicode::Grapheme_Extend(self)
     }
 
-    /// Returns `true` if this `char` has the `Cased` property.
+    /// 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`].
@@ -981,7 +981,11 @@ impl char {
     #[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::Cased(self) }
+        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.
@@ -1000,7 +1004,7 @@ impl char {
         if self.is_ascii() {
             matches!(self, '\'' | '.' | ':' | '^' | '`')
         } else {
-            unicode::Case_Ignorable(self)
+            Case_Ignorable(self)
         }
     }
 

library/core/src/unicode/mod.rs

@@ -8,10 +8,10 @@ pub use unicode_data::conversions;
 #[rustfmt::skip]
 pub(crate) use unicode_data::alphabetic::lookup as Alphabetic;
 pub(crate) use unicode_data::case_ignorable::lookup as Case_Ignorable;
-pub(crate) use unicode_data::cased::lookup as Cased;
 pub(crate) use unicode_data::cc::lookup as Cc;
 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 +1,16 @@
 ///! This file is generated by `./x run src/tools/unicode-table-generator`; do not edit manually!
 // Alphabetic      :  1723 bytes, 142707 codepoints in 755 ranges (U+0000AA - U+0323B0) using skiplist
 // Case_Ignorable  :  1043 bytes,   2744 codepoints in 447 ranges (U+0000A8 - U+0E01F0) using skiplist
-// Cased           :   403 bytes,   4526 codepoints in 157 ranges (U+0000AA - U+01F18A) using skiplist
 // Cc              :     7 bytes,     32 codepoints in   1 ranges (U+000080 - U+0000A0) using skiplist
 // Grapheme_Extend :   887 bytes,   2193 codepoints in 375 ranges (U+000300 - U+0E01F0) using skiplist
 // Lowercase       :   933 bytes,   2543 codepoints in 674 ranges (U+0000AA - U+01E944) using bitset
+// Lt              :    33 bytes,     31 codepoints in  10 ranges (U+0001C5 - U+001FFD) using skiplist
 // N               :   455 bytes,   1901 codepoints in 143 ranges (U+0000B2 - U+01FBFA) using skiplist
 // Uppercase       :   797 bytes,   1952 codepoints in 655 ranges (U+0000C0 - U+01F18A) using bitset
 // White_Space     :   256 bytes,     19 codepoints in   8 ranges (U+000085 - U+003001) using cascading
 // to_lower        : 11484 bytes
 // to_upper        : 13432 bytes
-// Total           : 31420 bytes
+// Total           : 31050 bytes
 
 #[inline(always)]
 const fn bitset_search<
@@ -338,59 +338,6 @@ pub mod case_ignorable {
     }
 }
 
-#[rustfmt::skip]
-pub mod cased {
-    use super::ShortOffsetRunHeader;
-
-    static SHORT_OFFSET_RUNS: [ShortOffsetRunHeader; 22] = [
-        ShortOffsetRunHeader::new(0, 4256), ShortOffsetRunHeader::new(51, 5024),
-        ShortOffsetRunHeader::new(61, 7296), ShortOffsetRunHeader::new(65, 7958),
-        ShortOffsetRunHeader::new(74, 9398), ShortOffsetRunHeader::new(149, 11264),
-        ShortOffsetRunHeader::new(151, 42560), ShortOffsetRunHeader::new(163, 43824),
-        ShortOffsetRunHeader::new(183, 64256), ShortOffsetRunHeader::new(189, 65313),
-        ShortOffsetRunHeader::new(193, 66560), ShortOffsetRunHeader::new(197, 67456),
-        ShortOffsetRunHeader::new(219, 68736), ShortOffsetRunHeader::new(227, 71840),
-        ShortOffsetRunHeader::new(235, 93760), ShortOffsetRunHeader::new(237, 119808),
-        ShortOffsetRunHeader::new(239, 120486), ShortOffsetRunHeader::new(276, 122624),
-        ShortOffsetRunHeader::new(299, 122928), ShortOffsetRunHeader::new(305, 125184),
-        ShortOffsetRunHeader::new(307, 127280), ShortOffsetRunHeader::new(309, 1241482),
-    ];
-    static OFFSETS: [u8; 315] = [
-        170, 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,
-    ];
-    #[inline]
-    pub fn lookup(c: char) -> bool {
-        debug_assert!(!c.is_ascii());
-        (c as u32) >= 0xaa && 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 cc {
     use super::ShortOffsetRunHeader;
@@ -605,6 +552,39 @@ 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 {
+        debug_assert!(!c.is_ascii());
+        (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",