Rename N to Numeric

Workaround for issue #148387

Author: joboet

library/core/src/unicode/mod.rs

@@ -12,7 +12,7 @@ pub use unicode_data::conversions;
 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::n::lookup as N;
+pub(crate) use unicode_data::numeric::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

@@ -4,7 +4,7 @@
 // Cased           :   401 bytes,   4580 codepoints in 156 ranges (U+0000AA - U+01F18A) using skiplist
 // Grapheme_Extend :   899 bytes,   2232 codepoints in 383 ranges (U+000300 - U+0E01F0) using skiplist
 // Lowercase       :   943 bytes,   2569 codepoints in 676 ranges (U+0000AA - U+01E944) using bitset
-// N               :   463 bytes,   1914 codepoints in 145 ranges (U+0000B2 - U+01FBFA) using skiplist
+// Numeric         :   463 bytes,   1914 codepoints in 145 ranges (U+0000B2 - U+01FBFA) using skiplist
 // Uppercase       :   799 bytes,   1980 codepoints in 659 ranges (U+0000C0 - U+01F18A) using bitset
 // White_Space     :   256 bytes,     19 codepoints in   8 ranges (U+000085 - U+003001) using cascading
 // to_lower        : 11708 bytes
@@ -539,7 +539,7 @@ pub mod lowercase {
     }
 }
 
-pub mod n {
+pub mod numeric {
     use super::ShortOffsetRunHeader;
 
     static SHORT_OFFSET_RUNS: [ShortOffsetRunHeader; 43] = [

library/coretests/tests/unicode.rs

@@ -73,8 +73,8 @@ fn lowercase() {
 
 #[test]
 #[cfg_attr(miri, ignore)] // Miri is too slow
-fn n() {
-    test_boolean_property(test_data::N, unicode_data::n::lookup);
+fn numeric() {
+    test_boolean_property(test_data::NUMERIC, unicode_data::numeric::lookup);
 }
 
 #[test]

library/coretests/tests/unicode/test_data.rs

@@ -777,7 +777,7 @@ pub(super) static LOWERCASE: &[RangeInclusive<char>; 676] = &[
 ];
 
 #[rustfmt::skip]
-pub(super) static N: &[RangeInclusive<char>; 145] = &[
+pub(super) static NUMERIC: &[RangeInclusive<char>; 145] = &[
     '\u{b2}'..='\u{b3}', '\u{b9}'..='\u{b9}', '\u{bc}'..='\u{be}', '\u{660}'..='\u{669}',
     '\u{6f0}'..='\u{6f9}', '\u{7c0}'..='\u{7c9}', '\u{966}'..='\u{96f}', '\u{9e6}'..='\u{9ef}',
     '\u{9f4}'..='\u{9f9}', '\u{a66}'..='\u{a6f}', '\u{ae6}'..='\u{aef}', '\u{b66}'..='\u{b6f}',

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

@@ -93,7 +93,7 @@ static PROPERTIES: &[&str] = &[
     "Case_Ignorable",
     "Grapheme_Extend",
     "White_Space",
-    "N",
+    "Numeric",
 ];
 
 struct UnicodeData {
@@ -148,10 +148,13 @@ fn load_data() -> UnicodeData {
     for row in ucd_parse::UnicodeDataExpander::new(
         ucd_parse::parse::<_, ucd_parse::UnicodeData>(&UNICODE_DIRECTORY).unwrap(),
     ) {
-        let general_category = if ["Nd", "Nl", "No"].contains(&row.general_category.as_str()) {
-            "N"
-        } else {
-            row.general_category.as_str()
+        // FIXME: this used to map `Nd`, `Nl` and `No` to `N`, but the generated
+        // `unicode_data` is `pub`, so the module names can show up in
+        // recommendations (see issue #148387). While that's a rustc issue, we
+        // choose a better name for the `N` property to avoid bad diagnostics.
+        let general_category = match row.general_category.as_str() {
+            "Nd" | "Nl" | "No" => "Numeric",
+            category => category,
         };
         if let Some(name) = PROPERTIES.iter().find(|prop| **prop == general_category) {
             properties
@@ -241,7 +244,9 @@ fn main() {
             emit_codepoints(&mut emitter, ranges);
         }
 
-        modules.push((property.to_lowercase().to_string(), emitter.file));
+        let module_name = property.to_lowercase();
+
+        modules.push((module_name, emitter.file));
         writeln!(
             table_file,
             "// {:16}: {:5} bytes, {:6} codepoints in {:3} ranges (U+{:06X} - U+{:06X}) using {}",