sort: support percent arguments to -S option

Add support for parsing percent arguments to the `-S` option. The given
percentage specifies a percentage of the total physical memory. For
Linux, the total physical memory is read from `/proc/meminfo`. The
feature is not yet implemented for other systems.

In order to implement the feature, the `uucore::parser::parse_size`
function was updated to recognize strings of the form `NNN%`.

Fixes #3500

Author: jfinkels

src/uu/sort/src/sort.rs

@@ -288,7 +288,7 @@ impl GlobalSettings {
         // GNU sort (8.32) invalid:  b, B, 1B,                         p, e, z, y
         let size = Parser::default()
             .with_allow_list(&[
-                "b", "k", "K", "m", "M", "g", "G", "t", "T", "P", "E", "Z", "Y",
+                "b", "k", "K", "m", "M", "g", "G", "t", "T", "P", "E", "Z", "Y", "%",
             ])
             .with_default_unit("K")
             .with_b_byte_count(true)

src/uucore/src/lib/parser/parse_size.rs

@@ -8,10 +8,70 @@
 
 use std::error::Error;
 use std::fmt;
-use std::num::IntErrorKind;
+#[cfg(target_os = "linux")]
+use std::io::BufRead;
+use std::num::{IntErrorKind, ParseIntError};
 
 use crate::display::Quotable;
 
+/// Error arising from trying to compute system memory.
+enum SystemError {
+    IOError,
+    ParseError,
+    NotFound,
+}
+
+impl From<std::io::Error> for SystemError {
+    fn from(_: std::io::Error) -> Self {
+        Self::IOError
+    }
+}
+
+impl From<ParseIntError> for SystemError {
+    fn from(_: ParseIntError) -> Self {
+        Self::ParseError
+    }
+}
+
+/// Get the total number of bytes of physical memory.
+///
+/// The information is read from the `/proc/meminfo` file.
+///
+/// # Errors
+///
+/// If there is a problem reading the file or finding the appropriate
+/// entry in the file.
+#[cfg(target_os = "linux")]
+fn total_physical_memory() -> Result<u128, SystemError> {
+    // On Linux, the `/proc/meminfo` file has a table with information
+    // about memory usage. For example,
+    //
+    //     MemTotal:        7811500 kB
+    //     MemFree:         1487876 kB
+    //     MemAvailable:    3857232 kB
+    //     ...
+    //
+    // We just need to extract the number of `MemTotal`
+    let table = std::fs::read("/proc/meminfo")?;
+    for line in table.lines() {
+        let line = line?;
+        if line.starts_with("MemTotal:") && line.ends_with("kB") {
+            let num_kilobytes: u128 = line[9..line.len() - 2].trim().parse()?;
+            let num_bytes = 1024 * num_kilobytes;
+            return Ok(num_bytes);
+        }
+    }
+    Err(SystemError::NotFound)
+}
+
+/// Get the total number of bytes of physical memory.
+///
+/// TODO Implement this for non-Linux systems.
+#[cfg(not(target_os = "linux"))]
+fn total_physical_memory() -> Result<u128, SystemError> {
+    Err(SystemError::NotFound)
+}
+
 /// Parser for sizes in SI or IEC units (multiples of 1000 or 1024 bytes).
 ///
 /// The [`Parser::parse`] function performs the parse.
@@ -133,6 +193,16 @@ impl<'parser> Parser<'parser> {
             }
         }
 
+        // Special case: for percentage, just compute the given fraction
+        // of the total physical memory on the machine, if possible.
+        if unit == "%" {
+            let number: u128 = Self::parse_number(&numeric_string, 10, size)?;
+            return match total_physical_memory() {
+                Ok(total) => Ok((number / 100) * total),
+                Err(_) => Err(ParseSizeError::PhysicalMem(size.to_string())),
+            };
+        }
+
         // Compute the factor the unit represents.
         // empty string means the factor is 1.
         //
@@ -688,4 +758,16 @@ mod tests {
         assert_eq!(Ok(94722), parse_size_u64("0x17202"));
         assert_eq!(Ok(44251 * 1024), parse_size_u128("0xACDBK"));
     }
+
+    #[test]
+    #[cfg(target_os = "linux")]
+    fn parse_percent() {
+        assert!(parse_size_u64("0%").is_ok());
+        assert!(parse_size_u64("50%").is_ok());
+        assert!(parse_size_u64("100%").is_ok());
+        assert!(parse_size_u64("100000%").is_ok());
+        assert!(parse_size_u64("-1%").is_err());
+        assert!(parse_size_u64("1.0%").is_err());
+        assert!(parse_size_u64("0x1%").is_err());
+    }
 }

tests/by-util/test_sort.rs

@@ -29,6 +29,10 @@ fn test_helper(file_name: &str, possible_args: &[&str]) {
 
 #[test]
 fn test_buffer_sizes() {
+    #[cfg(target_os = "linux")]
+    let buffer_sizes = ["0", "50K", "50k", "1M", "100M", "0%", "10%"];
+    // TODO Percentage sizes are not yet supported beyond Linux.
+    #[cfg(not(target_os = "linux"))]
     let buffer_sizes = ["0", "50K", "50k", "1M", "100M"];
     for buffer_size in &buffer_sizes {
         TestScenario::new(util_name!())
@@ -73,6 +77,15 @@ fn test_invalid_buffer_size() {
         .code_is(2)
         .stderr_only("sort: invalid suffix in --buffer-size argument '100f'\n");
 
+    // TODO Percentage sizes are not yet supported beyond Linux.
+    #[cfg(target_os = "linux")]
+    new_ucmd!()
+        .arg("-S")
+        .arg("0x123%")
+        .fails()
+        .code_is(2)
+        .stderr_only("sort: invalid --buffer-size argument '0x123%'\n");
+
     new_ucmd!()
         .arg("-n")
         .arg("-S")