Merge pull request #6226 from mvo5/du-h-precision

du: give `-h` output the same precision as GNU coreutils

Cargo.lock

@@ -3325,6 +3325,7 @@ dependencies = [
  "md-5",
  "memchr",
  "nix",
+ "number_prefix",
  "once_cell",
  "os_display",
  "sha1",

src/uu/du/Cargo.toml

@@ -19,7 +19,7 @@ chrono = { workspace = true }
 # For the --exclude & --exclude-from options
 glob = { workspace = true }
 clap = { workspace = true }
-uucore = { workspace = true }
+uucore = { workspace = true, features = ["format"] }
 
 [target.'cfg(target_os = "windows")'.dependencies]
 windows-sys = { workspace = true, features = [

src/uu/du/src/du.rs

@@ -75,9 +75,6 @@ const ABOUT: &str = help_about!("du.md");
 const AFTER_HELP: &str = help_section!("after help", "du.md");
 const USAGE: &str = help_usage!("du.md");
 
-// TODO: Support Z & Y (currently limited by size of u64)
-const UNITS: [(char, u32); 6] = [('E', 6), ('P', 5), ('T', 4), ('G', 3), ('M', 2), ('K', 1)];
-
 struct TraversalOptions {
     all: bool,
     separate_dirs: bool,
@@ -117,7 +114,8 @@ enum Time {
 
 #[derive(Clone)]
 enum SizeFormat {
-    Human(u64),
+    HumanDecimal,
+    HumanBinary,
     BlockSize(u64),
 }
 
@@ -549,18 +547,14 @@ impl StatPrinter {
             return size.to_string();
         }
         match self.size_format {
-            SizeFormat::Human(multiplier) => {
-                if size == 0 {
-                    return "0".to_string();
-                }
-                for &(unit, power) in &UNITS {
-                    let limit = multiplier.pow(power);
-                    if size >= limit {
-                        return format!("{:.1}{}", (size as f64) / (limit as f64), unit);
-                    }
-                }
-                format!("{size}B")
-            }
+            SizeFormat::HumanDecimal => uucore::format::human::human_readable(
+                size,
+                uucore::format::human::SizeFormat::Decimal,
+            ),
+            SizeFormat::HumanBinary => uucore::format::human::human_readable(
+                size,
+                uucore::format::human::SizeFormat::Binary,
+            ),
             SizeFormat::BlockSize(block_size) => div_ceil(size, block_size).to_string(),
         }
     }
@@ -688,9 +682,9 @@ pub fn uumain(args: impl uucore::Args) -> UResult<()> {
     });
 
     let size_format = if matches.get_flag(options::HUMAN_READABLE) {
-        SizeFormat::Human(1024)
+        SizeFormat::HumanBinary
     } else if matches.get_flag(options::SI) {
-        SizeFormat::Human(1000)
+        SizeFormat::HumanDecimal
     } else if matches.get_flag(options::BYTES) {
         SizeFormat::BlockSize(1)
     } else if matches.get_flag(options::BLOCK_SIZE_1K) {

src/uu/ls/Cargo.toml

@@ -26,6 +26,7 @@ lscolors = { workspace = true }
 uucore = { workspace = true, features = [
   "colors",
   "entries",
+  "format",
   "fs",
   "fsxattr",
   "quoting-style",

src/uu/ls/src/ls.rs

@@ -12,7 +12,6 @@ use clap::{
 use glob::{MatchOptions, Pattern};
 use lscolors::{LsColors, Style};
 
-use number_prefix::NumberPrefix;
 use std::{cell::OnceCell, num::IntErrorKind};
 use std::{collections::HashSet, io::IsTerminal};
 
@@ -37,6 +36,7 @@ use std::{
 use term_grid::{Cell, Direction, Filling, Grid, GridOptions};
 use unicode_width::UnicodeWidthStr;
 use uucore::error::USimpleError;
+use uucore::format::human::{human_readable, SizeFormat};
 #[cfg(all(unix, not(any(target_os = "android", target_os = "macos"))))]
 use uucore::fsxattr::has_acl;
 #[cfg(any(
@@ -313,13 +313,6 @@ enum Sort {
     Width,
 }
 
-#[derive(PartialEq)]
-enum SizeFormat {
-    Bytes,
-    Binary,  // Powers of 1024, --human-readable, -h
-    Decimal, // Powers of 1000, --si
-}
-
 #[derive(PartialEq, Eq)]
 enum Files {
     All,
@@ -3038,30 +3031,6 @@ fn display_date(metadata: &Metadata, config: &Config) -> String {
     }
 }
 
-// There are a few peculiarities to how GNU formats the sizes:
-// 1. One decimal place is given if and only if the size is smaller than 10
-// 2. It rounds sizes up.
-// 3. The human-readable format uses powers for 1024, but does not display the "i"
-//    that is commonly used to denote Kibi, Mebi, etc.
-// 4. Kibi and Kilo are denoted differently ("k" and "K", respectively)
-fn format_prefixed(prefixed: &NumberPrefix<f64>) -> String {
-    match prefixed {
-        NumberPrefix::Standalone(bytes) => bytes.to_string(),
-        NumberPrefix::Prefixed(prefix, bytes) => {
-            // Remove the "i" from "Ki", "Mi", etc. if present
-            let prefix_str = prefix.symbol().trim_end_matches('i');
-
-            // Check whether we get more than 10 if we round up to the first decimal
-            // because we want do display 9.81 as "9.9", not as "10".
-            if (10.0 * bytes).ceil() >= 100.0 {
-                format!("{:.0}{}", bytes.ceil(), prefix_str)
-            } else {
-                format!("{:.1}{}", (10.0 * bytes).ceil() / 10.0, prefix_str)
-            }
-        }
-    }
-}
-
 #[allow(dead_code)]
 enum SizeOrDeviceId {
     Size(String),
@@ -3104,13 +3073,7 @@ fn display_len_or_rdev(metadata: &Metadata, config: &Config) -> SizeOrDeviceId {
 }
 
 fn display_size(size: u64, config: &Config) -> String {
-    // NOTE: The human-readable behavior deviates from the GNU ls.
-    // The GNU ls uses binary prefixes by default.
-    match config.size_format {
-        SizeFormat::Binary => format_prefixed(&NumberPrefix::binary(size as f64)),
-        SizeFormat::Decimal => format_prefixed(&NumberPrefix::decimal(size as f64)),
-        SizeFormat::Bytes => size.to_string(),
-    }
+    human_readable(size, config.size_format)
 }
 
 #[cfg(unix)]

src/uucore/Cargo.toml

@@ -20,6 +20,7 @@ path = "src/lib/lib.rs"
 [dependencies]
 clap = { workspace = true }
 uucore_procs = { workspace = true }
+number_prefix = { workspace = true }
 dns-lookup = { version = "2.0.4", optional = true }
 dunce = { version = "1.0.4", optional = true }
 wild = "2.2"

src/uucore/src/lib/features/format/human.rs

@@ -0,0 +1,65 @@
+// This file is part of the uutils coreutils package.
+//
+// For the full copyright and license information, please view the LICENSE
+// file that was distributed with this source code.
+
+// spell-checker:ignore gnulibs sfmt
+
+//! `human`-size formatting
+//!
+//! Format sizes like gnulibs human_readable() would
+
+use number_prefix::NumberPrefix;
+
+#[derive(Copy, Clone, PartialEq)]
+pub enum SizeFormat {
+    Bytes,
+    Binary,  // Powers of 1024, --human-readable, -h
+    Decimal, // Powers of 1000, --si
+}
+
+// There are a few peculiarities to how GNU formats the sizes:
+// 1. One decimal place is given if and only if the size is smaller than 10
+// 2. It rounds sizes up.
+// 3. The human-readable format uses powers for 1024, but does not display the "i"
+//    that is commonly used to denote Kibi, Mebi, etc.
+// 4. Kibi and Kilo are denoted differently ("k" and "K", respectively)
+fn format_prefixed(prefixed: &NumberPrefix<f64>) -> String {
+    match prefixed {
+        NumberPrefix::Standalone(bytes) => bytes.to_string(),
+        NumberPrefix::Prefixed(prefix, bytes) => {
+            // Remove the "i" from "Ki", "Mi", etc. if present
+            let prefix_str = prefix.symbol().trim_end_matches('i');
+
+            // Check whether we get more than 10 if we round up to the first decimal
+            // because we want do display 9.81 as "9.9", not as "10".
+            if (10.0 * bytes).ceil() >= 100.0 {
+                format!("{:.0}{}", bytes.ceil(), prefix_str)
+            } else {
+                format!("{:.1}{}", (10.0 * bytes).ceil() / 10.0, prefix_str)
+            }
+        }
+    }
+}
+
+pub fn human_readable(size: u64, sfmt: SizeFormat) -> String {
+    match sfmt {
+        SizeFormat::Binary => format_prefixed(&NumberPrefix::binary(size as f64)),
+        SizeFormat::Decimal => format_prefixed(&NumberPrefix::decimal(size as f64)),
+        SizeFormat::Bytes => size.to_string(),
+    }
+}
+
+#[cfg(test)]
+#[test]
+fn test_human_readable() {
+    let test_cases = [
+        (133456345, SizeFormat::Binary, "128M"),
+        (12 * 1024 * 1024, SizeFormat::Binary, "12M"),
+        (8500, SizeFormat::Binary, "8.4K"),
+    ];
+
+    for &(size, sfmt, expected_str) in &test_cases {
+        assert_eq!(human_readable(size, sfmt), expected_str);
+    }
+}

src/uucore/src/lib/features/format/mod.rs

@@ -32,6 +32,7 @@
 
 mod argument;
 mod escape;
+pub mod human;
 pub mod num_format;
 pub mod num_parser;
 mod spec;

tests/by-util/test_du.rs

@@ -3,7 +3,7 @@
 // For the full copyright and license information, please view the LICENSE
 // file that was distributed with this source code.
 
-// spell-checker:ignore (paths) atim sublink subwords azerty azeaze xcwww azeaz amaz azea qzerty tazerty tsublink testfile1 testfile2 filelist testdir testfile
+// spell-checker:ignore (paths) atim sublink subwords azerty azeaze xcwww azeaz amaz azea qzerty tazerty tsublink testfile1 testfile2 filelist fpath testdir testfile
 #[cfg(not(windows))]
 use regex::Regex;
 
@@ -543,6 +543,34 @@ fn test_du_h_flag_empty_file() {
         .stdout_only("0\tempty.txt\n");
 }
 
+#[test]
+fn test_du_h_precision() {
+    let test_cases = [
+        (133456345, "128M"),
+        (12 * 1024 * 1024, "12M"),
+        (8500, "8.4K"),
+    ];
+
+    for &(test_len, expected_output) in &test_cases {
+        let (at, mut ucmd) = at_and_ucmd!();
+
+        let fpath = at.plus("test.txt");
+        std::fs::File::create(&fpath)
+            .expect("cannot create test file")
+            .set_len(test_len)
+            .expect("cannot truncate test len to size");
+        ucmd.arg("-h")
+            .arg("--apparent-size")
+            .arg(&fpath)
+            .succeeds()
+            .stdout_only(format!(
+                "{}\t{}\n",
+                expected_output,
+                &fpath.to_string_lossy()
+            ));
+    }
+}
+
 #[cfg(feature = "touch")]
 #[test]
 fn test_du_time() {