sort: gnu coreutils compatibility (sort float.sh) (#9839)

* feat(sort): support international decimal separators in numeric sorting

M

fuzz/Cargo.lock

@@ -502,6 +502,17 @@ version = "0.1.6"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "645cbb3a84e60b7531617d5ae4e57f7e27308f6445f5abf653209ea76dec8dff"
 
+[[package]]
+name = "fixed_decimal"
+version = "0.7.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "35eabf480f94d69182677e37571d3be065822acfafd12f2f085db44fbbcc8e57"
+dependencies = [
+ "displaydoc",
+ "smallvec",
+ "writeable",
+]
+
 [[package]]
 name = "flate2"
 version = "1.1.5"
@@ -676,6 +687,27 @@ dependencies = [
  "zerovec",
 ]
 
+[[package]]
+name = "icu_decimal"
+version = "2.1.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a38c52231bc348f9b982c1868a2af3195199623007ba2c7650f432038f5b3e8e"
+dependencies = [
+ "fixed_decimal",
+ "icu_decimal_data",
+ "icu_locale",
+ "icu_locale_core",
+ "icu_provider",
+ "writeable",
+ "zerovec",
+]
+
+[[package]]
+name = "icu_decimal_data"
+version = "2.1.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "2905b4044eab2dd848fe84199f9195567b63ab3a93094711501363f63546fef7"
+
 [[package]]
 name = "icu_locale"
 version = "2.1.1"
@@ -1731,7 +1763,9 @@ dependencies = [
  "glob",
  "hex",
  "icu_collator",
+ "icu_decimal",
  "icu_locale",
+ "icu_provider",
  "itertools",
  "libc",
  "md-5",

src/uu/sort/Cargo.toml

@@ -34,7 +34,12 @@ self_cell = { workspace = true }
 tempfile = { workspace = true }
 thiserror = { workspace = true }
 unicode-width = { workspace = true }
-uucore = { workspace = true, features = ["fs", "parser-size", "version-cmp"] }
+uucore = { workspace = true, features = [
+  "fs",
+  "parser-size",
+  "version-cmp",
+  "i18n-decimal",
+] }
 fluent = { workspace = true }
 
 [target.'cfg(unix)'.dependencies]

src/uu/sort/src/sort.rs

@@ -47,6 +47,7 @@ use uucore::error::{FromIo, strip_errno};
 use uucore::error::{UError, UResult, USimpleError, UUsageError};
 use uucore::extendedbigdecimal::ExtendedBigDecimal;
 use uucore::format_usage;
+use uucore::i18n::decimal::locale_decimal_separator;
 use uucore::line_ending::LineEnding;
 use uucore::parser::num_parser::{ExtendedParser, ExtendedParserError};
 use uucore::parser::parse_size::{ParseSizeError, Parser};
@@ -106,6 +107,14 @@ mod options {
 
 const DECIMAL_PT: u8 = b'.';
 
+fn locale_decimal_pt() -> u8 {
+    match locale_decimal_separator().as_bytes().first().copied() {
+        Some(b'.') => b'.',
+        Some(b',') => b',',
+        _ => DECIMAL_PT,
+    }
+}
+
 const NEGATIVE: &u8 = &b'-';
 const POSITIVE: &u8 = &b'+';
 
@@ -683,8 +692,8 @@ impl<'a> Line<'a> {
                 }
                 SortMode::GeneralNumeric => {
                     let initial_selection = &self.line[selection.clone()];
-
-                    let leading = get_leading_gen(initial_selection);
+                    let decimal_pt = locale_decimal_pt();
+                    let leading = get_leading_gen(initial_selection, decimal_pt);
 
                     // Shorten selection to leading.
                     selection.start += leading.start;
@@ -1072,7 +1081,11 @@ impl FieldSelector {
             Selection::WithNumInfo(range_str, info)
         } else if self.settings.mode == SortMode::GeneralNumeric {
             // Parse this number as BigDecimal, as this is the requirement for general numeric sorting.
-            Selection::AsBigDecimal(general_bd_parse(&range_str[get_leading_gen(range_str)]))
+            let decimal_pt = locale_decimal_pt();
+            Selection::AsBigDecimal(general_bd_parse(
+                &range_str[get_leading_gen(range_str, decimal_pt)],
+                decimal_pt,
+            ))
         } else {
             // This is not a numeric sort, so we don't need a NumCache.
             Selection::Str(range_str)
@@ -2491,7 +2504,7 @@ fn ascii_case_insensitive_cmp(a: &[u8], b: &[u8]) -> Ordering {
 // scientific notation, so we strip those lines only after the end of the following numeric string.
 // For example, 5e10KFD would be 5e10 or 5x10^10 and +10000HFKJFK would become 10000.
 #[allow(clippy::cognitive_complexity)]
-fn get_leading_gen(inp: &[u8]) -> Range<usize> {
+fn get_leading_gen(inp: &[u8], decimal_pt: u8) -> Range<usize> {
     let trimmed = inp.trim_ascii_start();
     let leading_whitespace_len = inp.len() - trimmed.len();
 
@@ -2529,7 +2542,7 @@ fn get_leading_gen(inp: &[u8]) -> Range<usize> {
             continue;
         }
 
-        if c == DECIMAL_PT && !had_decimal_pt && !had_e_notation {
+        if c == decimal_pt && !had_decimal_pt && !had_e_notation {
             had_decimal_pt = true;
             continue;
         }
@@ -2572,9 +2585,16 @@ pub enum GeneralBigDecimalParseResult {
 /// Parse the beginning string into a [`GeneralBigDecimalParseResult`].
 /// Using a [`GeneralBigDecimalParseResult`] instead of [`ExtendedBigDecimal`] is necessary to correctly order floats.
 #[inline(always)]
-fn general_bd_parse(a: &[u8]) -> GeneralBigDecimalParseResult {
+fn general_bd_parse(a: &[u8], decimal_pt: u8) -> GeneralBigDecimalParseResult {
+    let parsed_bytes = (decimal_pt != DECIMAL_PT).then(|| {
+        a.iter()
+            .map(|&b| if b == decimal_pt { DECIMAL_PT } else { b })
+            .collect::<Vec<_>>()
+    });
+    let input = parsed_bytes.as_deref().unwrap_or(a);
+
     // The string should be valid ASCII to be parsed.
-    let Ok(a) = std::str::from_utf8(a) else {
+    let Ok(a) = std::str::from_utf8(input) else {
         return GeneralBigDecimalParseResult::Invalid;
     };
 

tests/by-util/test_sort.rs

@@ -1631,6 +1631,32 @@ fn test_g_float() {
         .stdout_is(output);
 }
 
+#[test]
+fn test_g_float_locale_decimal_separator() {
+    let Ok(locale_fr_utf8) = env::var("LOCALE_FR_UTF8") else {
+        return;
+    };
+    if locale_fr_utf8 == "none" {
+        return;
+    }
+
+    let ts = TestScenario::new("sort");
+
+    ts.ucmd()
+        .env("LC_ALL", &locale_fr_utf8)
+        .args(&["-g", "--stable"])
+        .pipe_in("1,9\n1,10\n")
+        .succeeds()
+        .stdout_is("1,10\n1,9\n");
+
+    ts.ucmd()
+        .env("LC_ALL", &locale_fr_utf8)
+        .args(&["-g", "--stable"])
+        .pipe_in("1.9\n1.10\n")
+        .succeeds()
+        .stdout_is("1.10\n1.9\n");
+}
+
 #[test]
 // Test misc numbers ("'a" is not interpreted as literal, trailing text is ignored...)
 fn test_g_misc() {