Merge pull request #7782 from drinkcat/seq-perf-more-use-cat

Move seq's `fast_inc` to `uucore`, use it in `cat`

src/uu/cat/Cargo.toml

@@ -21,7 +21,7 @@ path = "src/cat.rs"
 clap = { workspace = true }
 memchr = { workspace = true }
 thiserror = { workspace = true }
-uucore = { workspace = true, features = ["fs", "pipes"] }
+uucore = { workspace = true, features = ["fast-inc", "fs", "pipes"] }
 
 [target.'cfg(unix)'.dependencies]
 nix = { workspace = true }

src/uu/cat/src/cat.rs

@@ -24,7 +24,7 @@ use thiserror::Error;
 use uucore::display::Quotable;
 use uucore::error::UResult;
 use uucore::fs::FileInformation;
-use uucore::{format_usage, help_about, help_usage};
+use uucore::{fast_inc::fast_inc_one, format_usage, help_about, help_usage};
 
 /// Linux splice support
 #[cfg(any(target_os = "linux", target_os = "android"))]
@@ -33,61 +33,55 @@ mod splice;
 const USAGE: &str = help_usage!("cat.md");
 const ABOUT: &str = help_about!("cat.md");
 
+// Allocate 32 digits for the line number.
+// An estimate is that we can print about 1e8 lines/seconds, so 32 digits
+// would be enough for billions of universe lifetimes.
+const LINE_NUMBER_BUF_SIZE: usize = 32;
+
 struct LineNumber {
-    buf: Vec<u8>,
+    buf: [u8; LINE_NUMBER_BUF_SIZE],
+    print_start: usize,
+    num_start: usize,
+    num_end: usize,
 }
 
 // Logic to store a string for the line number. Manually incrementing the value
 // represented in a buffer like this is significantly faster than storing
 // a `usize` and using the standard Rust formatting macros to format a `usize`
 // to a string each time it's needed.
-// String is initialized to "     1\t" and incremented each time `increment` is
+// Buffer is initialized to "     1\t" and incremented each time `increment` is
-// called. When the value overflows the range storable in the buffer, a b'1' is
+// called, using uucore's fast_inc function that operates on strings.
-// prepended and the counting continues.
 impl LineNumber {
     fn new() -> Self {
+        let mut buf = [b'0'; LINE_NUMBER_BUF_SIZE];
+
+        let init_str = "     1\t";
+        let print_start = buf.len() - init_str.len();
+        let num_start = buf.len() - 2;
+        let num_end = buf.len() - 1;
+
+        buf[print_start..].copy_from_slice(init_str.as_bytes());
+
         LineNumber {
-            // Initialize buf to b"     1\t"
+            buf,
-            buf: Vec::from(b"     1\t"),
+            print_start,
+            num_start,
+            num_end,
         }
     }
 
     fn increment(&mut self) {
-        // skip(1) to avoid the \t in the last byte.
+        fast_inc_one(&mut self.buf, &mut self.num_start, self.num_end);
-        for ascii_digit in self.buf.iter_mut().rev().skip(1) {
+        self.print_start = self.print_start.min(self.num_start);
-            // Working from the least-significant digit, increment the number in the buffer.
-            // If we hit anything other than a b'9' we can break since the next digit is
-            // unaffected.
-            // Also note that if we hit a b' ', we can think of that as a 0 and increment to b'1'.
-            // If/else here is faster than match (as measured with some benchmarking Apr-2025),
-            // probably since we can prioritize most likely digits first.
-            if (b'0'..=b'8').contains(ascii_digit) {
-                *ascii_digit += 1;
-                break;
-            } else if b'9' == *ascii_digit {
-                *ascii_digit = b'0';
-            } else {
-                assert_eq!(*ascii_digit, b' ');
-                *ascii_digit = b'1';
-                break;
     }
-        }
-        if self.buf[0] == b'0' {
-            // This implies we've overflowed. In this case the buffer will be
-            // [b'0', b'0', ..., b'0', b'\t'].
-            // For debugging, the following logic would assert that to be the case.
-            // assert_eq!(*self.buf.last().unwrap(), b'\t');
-            // for ascii_digit in self.buf.iter_mut().rev().skip(1) {
-            //     assert_eq!(*ascii_digit, b'0');
-            // }
 
-            // All we need to do is prepend a b'1' and we're good.
+    #[inline]
-            self.buf.insert(0, b'1');
+    fn to_str(&self) -> &[u8] {
-        }
+        &self.buf[self.print_start..]
     }
 
     fn write(&self, writer: &mut impl Write) -> io::Result<()> {
-        writer.write_all(&self.buf)
+        writer.write_all(self.to_str())
     }
 }
 
@@ -804,21 +798,21 @@ mod tests {
     #[test]
     fn test_incrementing_string() {
         let mut incrementing_string = super::LineNumber::new();
-        assert_eq!(b"     1\t", incrementing_string.buf.as_slice());
+        assert_eq!(b"     1\t", incrementing_string.to_str());
         incrementing_string.increment();
-        assert_eq!(b"     2\t", incrementing_string.buf.as_slice());
+        assert_eq!(b"     2\t", incrementing_string.to_str());
         // Run through to 100
         for _ in 3..=100 {
             incrementing_string.increment();
         }
-        assert_eq!(b"   100\t", incrementing_string.buf.as_slice());
+        assert_eq!(b"   100\t", incrementing_string.to_str());
         // Run through until we overflow the original size.
         for _ in 101..=1_000_000 {
             incrementing_string.increment();
         }
-        // Confirm that the buffer expands when we overflow the original size.
+        // Confirm that the start position moves when we overflow the original size.
-        assert_eq!(b"1000000\t", incrementing_string.buf.as_slice());
+        assert_eq!(b"1000000\t", incrementing_string.to_str());
         incrementing_string.increment();
-        assert_eq!(b"1000001\t", incrementing_string.buf.as_slice());
+        assert_eq!(b"1000001\t", incrementing_string.to_str());
     }
 }

src/uu/seq/BENCHMARKING.md

@@ -76,5 +76,20 @@ write!(stdout, "{separator}")?
 
 The change above resulted in a ~10% speedup.
 
+### Fast increment path
+
+When dealing with positive integer values (first/increment/last), and
+the default format is used, we use a custom fast path that does arithmetic
+on u8 arrays (i.e. strings), instead of repeatedly calling into
+formatting format.
+
+This provides _massive_ performance gains, in the order of 10-20x compared
+with the default implementation, at the expense of some added code complexity.
+
+Just from performance numbers, it is clear that GNU `seq` uses similar
+tricks, but we are more liberal on when we use our fast path (e.g. large
+increments are supported, equal width is supported). Our fast path
+implementation gets within ~10% of `seq` performance when its fast
+path is activated.
 
 [0]: https://github.com/sharkdp/hyperfine

src/uu/seq/Cargo.toml

@@ -23,6 +23,7 @@ num-traits = { workspace = true }
 thiserror = { workspace = true }
 uucore = { workspace = true, features = [
   "extendedbigdecimal",
+  "fast-inc",
   "format",
   "parser",
   "quoting-style",

src/uu/seq/src/seq.rs

@@ -2,18 +2,20 @@
 //
 // For the full copyright and license information, please view the LICENSE
 // file that was distributed with this source code.
-// spell-checker:ignore (ToDO) bigdecimal extendedbigdecimal numberparse hexadecimalfloat
+// spell-checker:ignore (ToDO) bigdecimal extendedbigdecimal numberparse hexadecimalfloat biguint
 use std::ffi::OsString;
 use std::io::{BufWriter, ErrorKind, Write, stdout};
 
 use clap::{Arg, ArgAction, Command};
+use num_bigint::BigUint;
+use num_traits::ToPrimitive;
 use num_traits::Zero;
 
 use uucore::error::{FromIo, UResult};
 use uucore::extendedbigdecimal::ExtendedBigDecimal;
 use uucore::format::num_format::FloatVariant;
 use uucore::format::{Format, num_format};
-use uucore::{format_usage, help_about, help_usage};
+use uucore::{fast_inc::fast_inc, format_usage, help_about, help_usage};
 
 mod error;
 
@@ -149,13 +151,17 @@ pub fn uumain(args: impl uucore::Args) -> UResult<()> {
         }
     };
 
-    let precision = select_precision(&first, &increment, &last);
-
     // If a format was passed on the command line, use that.
     // If not, use some default format based on parameters precision.
-    let format = match options.format {
+    let (format, padding, fast_allowed) = match options.format {
-        Some(str) => Format::<num_format::Float, &ExtendedBigDecimal>::parse(str)?,
+        Some(str) => (
+            Format::<num_format::Float, &ExtendedBigDecimal>::parse(str)?,
+            0,
+            false,
+        ),
         None => {
+            let precision = select_precision(&first, &increment, &last);
+
             let padding = if options.equal_width {
                 let precision_value = precision.unwrap_or(0);
                 first
@@ -186,7 +192,12 @@ pub fn uumain(args: impl uucore::Args) -> UResult<()> {
                     ..Default::default()
                 },
             };
-            Format::from_formatter(formatter)
+            // Allow fast printing if precision is 0 (integer inputs), `print_seq` will do further checks.
+            (
+                Format::from_formatter(formatter),
+                padding,
+                precision == Some(0),
+            )
         }
     };
 
@@ -195,7 +206,10 @@ pub fn uumain(args: impl uucore::Args) -> UResult<()> {
         &options.separator,
         &options.terminator,
         &format,
+        fast_allowed,
+        padding,
     );
+
     match result {
         Ok(()) => Ok(()),
         Err(err) if err.kind() == ErrorKind::BrokenPipe => Ok(()),
@@ -245,6 +259,72 @@ pub fn uu_app() -> Command {
         )
 }
 
+/// Integer print, default format, positive increment: fast code path
+/// that avoids reformating digit at all iterations.
+fn fast_print_seq(
+    mut stdout: impl Write,
+    first: &BigUint,
+    increment: u64,
+    last: &BigUint,
+    separator: &str,
+    terminator: &str,
+    padding: usize,
+) -> std::io::Result<()> {
+    // Nothing to do, just return.
+    if last < first {
+        return Ok(());
+    }
+
+    // Do at most u64::MAX loops. We can print in the order of 1e8 digits per second,
+    // u64::MAX is 1e19, so it'd take hundreds of years for this to complete anyway.
+    // TODO: we can move this test to `print_seq` if we care about this case.
+    let loop_cnt = ((last - first) / increment).to_u64().unwrap_or(u64::MAX);
+
+    // Format the first number.
+    let first_str = first.to_string();
+
+    // Makeshift log10.ceil
+    let last_length = last.to_string().len();
+
+    // Allocate a large u8 buffer, that contains a preformatted string
+    // of the number followed by the `separator`.
+    //
+    // | ... head space ... | number | separator |
+    // ^0                   ^ start  ^ num_end   ^ size (==buf.len())
+    //
+    // We keep track of start in this buffer, as the number grows.
+    // When printing, we take a slice between start and end.
+    let size = last_length.max(padding) + separator.len();
+    // Fill with '0', this is needed for equal_width, and harmless otherwise.
+    let mut buf = vec![b'0'; size];
+    let buf = buf.as_mut_slice();
+
+    let num_end = buf.len() - separator.len();
+    let mut start = num_end - first_str.len();
+
+    // Initialize buf with first and separator.
+    buf[start..num_end].copy_from_slice(first_str.as_bytes());
+    buf[num_end..].copy_from_slice(separator.as_bytes());
+
+    // Normally, if padding is > 0, it should be equal to last_length,
+    // so start would be == 0, but there are corner cases.
+    start = start.min(num_end - padding);
+
+    // Prepare the number to increment with as a string
+    let inc_str = increment.to_string();
+    let inc_str = inc_str.as_bytes();
+
+    for _ in 0..loop_cnt {
+        stdout.write_all(&buf[start..])?;
+        fast_inc(buf, &mut start, num_end, inc_str);
+    }
+    // Write the last number without separator, but with terminator.
+    stdout.write_all(&buf[start..num_end])?;
+    write!(stdout, "{terminator}")?;
+    stdout.flush()?;
+    Ok(())
+}
+
 fn done_printing<T: Zero + PartialOrd>(next: &T, increment: &T, last: &T) -> bool {
     if increment >= &T::zero() {
         next > last
@@ -253,16 +333,42 @@ fn done_printing<T: Zero + PartialOrd>(next: &T, increment: &T, last: &T) -> boo
     }
 }
 
-/// Floating point based code path
+/// Arbitrary precision decimal number code path ("slow" path)
 fn print_seq(
     range: RangeFloat,
     separator: &str,
     terminator: &str,
     format: &Format<num_format::Float, &ExtendedBigDecimal>,
+    fast_allowed: bool,
+    padding: usize, // Used by fast path only
 ) -> std::io::Result<()> {
     let stdout = stdout().lock();
     let mut stdout = BufWriter::new(stdout);
     let (first, increment, last) = range;
+
+    if fast_allowed {
+        // Test if we can use fast code path.
+        // First try to convert the range to BigUint (u64 for the increment).
+        let (first_bui, increment_u64, last_bui) = (
+            first.to_biguint(),
+            increment.to_biguint().and_then(|x| x.to_u64()),
+            last.to_biguint(),
+        );
+        if let (Some(first_bui), Some(increment_u64), Some(last_bui)) =
+            (first_bui, increment_u64, last_bui)
+        {
+            return fast_print_seq(
+                stdout,
+                &first_bui,
+                increment_u64,
+                &last_bui,
+                separator,
+                terminator,
+                padding,
+            );
+        }
+    }
+
     let mut value = first;
 
     let mut is_first_iteration = true;

src/uucore/Cargo.toml

@@ -91,6 +91,7 @@ checksum = ["data-encoding", "thiserror", "sum"]
 encoding = ["data-encoding", "data-encoding-macro", "z85"]
 entries = ["libc"]
 extendedbigdecimal = ["bigdecimal", "num-traits"]
+fast-inc = []
 fs = ["dunce", "libc", "winapi-util", "windows-sys"]
 fsext = ["libc", "windows-sys"]
 fsxattr = ["xattr"]

src/uucore/src/lib/features.rs

@@ -20,6 +20,8 @@ pub mod custom_tz_fmt;
 pub mod encoding;
 #[cfg(feature = "extendedbigdecimal")]
 pub mod extendedbigdecimal;
+#[cfg(feature = "fast-inc")]
+pub mod fast_inc;
 #[cfg(feature = "format")]
 pub mod format;
 #[cfg(feature = "fs")]

src/uucore/src/lib/features/extendedbigdecimal.rs

@@ -2,7 +2,7 @@
 //
 // For the full copyright and license information, please view the LICENSE
 // file that was distributed with this source code.
-// spell-checker:ignore bigdecimal extendedbigdecimal
+// spell-checker:ignore bigdecimal extendedbigdecimal biguint
 //! An arbitrary precision float that can also represent infinity, NaN, etc.
 //!
 //! The finite values are stored as [`BigDecimal`] instances. Because
@@ -25,7 +25,9 @@ use std::ops::Add;
 use std::ops::Neg;
 
 use bigdecimal::BigDecimal;
+use bigdecimal::num_bigint::BigUint;
 use num_traits::FromPrimitive;
+use num_traits::Signed;
 use num_traits::Zero;
 
 #[derive(Debug, Clone)]
@@ -107,6 +109,20 @@ impl ExtendedBigDecimal {
     pub fn one() -> Self {
         Self::BigDecimal(1.into())
     }
+
+    pub fn to_biguint(&self) -> Option<BigUint> {
+        match self {
+            ExtendedBigDecimal::BigDecimal(big_decimal) => {
+                let (bi, scale) = big_decimal.as_bigint_and_scale();
+                if bi.is_negative() || scale > 0 || scale < -(u32::MAX as i64) {
+                    return None;
+                }
+                bi.to_biguint()
+                    .map(|bi| bi * BigUint::from(10u32).pow(-scale as u32))
+            }
+            _ => None,
+        }
+    }
 }
 
 impl Zero for ExtendedBigDecimal {

src/uucore/src/lib/features/fast_inc.rs

@@ -0,0 +1,209 @@
+// 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.
+
+/// Fast increment function, operating on ASCII strings.
+///
+/// Add inc to the string val[start..end]. This operates on ASCII digits, assuming
+/// val and inc are well formed.
+///
+/// Updates `start` if we have a carry, or if inc > start.
+///
+/// We also assume that there is enough space in val to expand if start needs
+/// to be updated.
+/// ```
+/// use uucore::fast_inc::fast_inc;
+///
+/// // Start with a buffer containing "0", with one byte of head space
+/// let mut val = Vec::from(".0".as_bytes());
+/// let mut start = val.len()-1;
+/// let end = val.len();
+/// let inc = "6".as_bytes();
+/// assert_eq!(&val[start..end], "0".as_bytes());
+/// fast_inc(val.as_mut(), &mut start, end, inc);
+/// assert_eq!(&val[start..end], "6".as_bytes());
+/// fast_inc(val.as_mut(), &mut start, end, inc);
+/// assert_eq!(&val[start..end], "12".as_bytes());
+/// ```
+#[inline]
+pub fn fast_inc(val: &mut [u8], start: &mut usize, end: usize, inc: &[u8]) {
+    // To avoid a lot of casts to signed integers, we make sure to decrement pos
+    // as late as possible, so that it does not ever go negative.
+    let mut pos = end;
+    let mut carry = 0u8;
+
+    // First loop, add all digits of inc into val.
+    for inc_pos in (0..inc.len()).rev() {
+        // The decrement operation would also panic in debug mode, print a message for developer convenience.
+        debug_assert!(
+            pos > 0,
+            "Buffer overflowed, make sure you allocate val with enough headroom."
+        );
+        pos -= 1;
+
+        let mut new_val = inc[inc_pos] + carry;
+        // Be careful here, only add existing digit of val.
+        if pos >= *start {
+            new_val += val[pos] - b'0';
+        }
+        if new_val > b'9' {
+            carry = 1;
+            new_val -= 10;
+        } else {
+            carry = 0;
+        }
+        val[pos] = new_val;
+    }
+
+    // Done, now, if we have a carry, add that to the upper digits of val.
+    if carry == 0 {
+        *start = (*start).min(pos);
+        return;
+    }
+
+    fast_inc_one(val, start, pos)
+}
+
+/// Fast increment by one function, operating on ASCII strings.
+///
+/// Add 1 to the string val[start..end]. This operates on ASCII digits, assuming
+/// val is well formed.
+///
+/// Updates `start` if we have a carry, or if inc > start.
+///
+/// We also assume that there is enough space in val to expand if start needs
+/// to be updated.
+/// ```
+/// use uucore::fast_inc::fast_inc_one;
+///
+/// // Start with a buffer containing "8", with one byte of head space
+/// let mut val = Vec::from(".8".as_bytes());
+/// let mut start = val.len()-1;
+/// let end = val.len();
+/// assert_eq!(&val[start..end], "8".as_bytes());
+/// fast_inc_one(val.as_mut(), &mut start, end);
+/// assert_eq!(&val[start..end], "9".as_bytes());
+/// fast_inc_one(val.as_mut(), &mut start, end);
+/// assert_eq!(&val[start..end], "10".as_bytes());
+/// ```
+#[inline]
+pub fn fast_inc_one(val: &mut [u8], start: &mut usize, end: usize) {
+    let mut pos = end;
+
+    while pos > *start {
+        pos -= 1;
+
+        if val[pos] == b'9' {
+            // 9+1 = 10. Carry propagating, keep going.
+            val[pos] = b'0';
+        } else {
+            // Carry stopped propagating, return unchanged start.
+            val[pos] += 1;
+            return;
+        }
+    }
+
+    // The following decrement operation would also panic in debug mode, print a message for developer convenience.
+    debug_assert!(
+        *start > 0,
+        "Buffer overflowed, make sure you allocate val with enough headroom."
+    );
+    // The carry propagated so far that a new digit was added.
+    val[*start - 1] = b'1';
+    *start -= 1;
+}
+
+#[cfg(test)]
+mod tests {
+    use crate::fast_inc::fast_inc;
+    use crate::fast_inc::fast_inc_one;
+
+    #[test]
+    fn test_fast_inc_simple() {
+        let mut val = Vec::from("...0_".as_bytes());
+        let mut start: usize = 3;
+        let inc = "4".as_bytes();
+        fast_inc(val.as_mut(), &mut start, 4, inc);
+        assert_eq!(start, 3);
+        assert_eq!(val, "...4_".as_bytes());
+        fast_inc(val.as_mut(), &mut start, 4, inc);
+        assert_eq!(start, 3);
+        assert_eq!(val, "...8_".as_bytes());
+        fast_inc(val.as_mut(), &mut start, 4, inc);
+        assert_eq!(start, 2); // carried 1 more digit
+        assert_eq!(val, "..12_".as_bytes());
+
+        let mut val = Vec::from("0_".as_bytes());
+        let mut start: usize = 0;
+        let inc = "2".as_bytes();
+        fast_inc(val.as_mut(), &mut start, 1, inc);
+        assert_eq!(start, 0);
+        assert_eq!(val, "2_".as_bytes());
+        fast_inc(val.as_mut(), &mut start, 1, inc);
+        assert_eq!(start, 0);
+        assert_eq!(val, "4_".as_bytes());
+        fast_inc(val.as_mut(), &mut start, 1, inc);
+        assert_eq!(start, 0);
+        assert_eq!(val, "6_".as_bytes());
+    }
+
+    // Check that we handle increment > val correctly.
+    #[test]
+    fn test_fast_inc_large_inc() {
+        let mut val = Vec::from("...7_".as_bytes());
+        let mut start: usize = 3;
+        let inc = "543".as_bytes();
+        fast_inc(val.as_mut(), &mut start, 4, inc);
+        assert_eq!(start, 1); // carried 2 more digits
+        assert_eq!(val, ".550_".as_bytes());
+        fast_inc(val.as_mut(), &mut start, 4, inc);
+        assert_eq!(start, 0); // carried 1 more digit
+        assert_eq!(val, "1093_".as_bytes());
+    }
+
+    // Check that we handle longer carries
+    #[test]
+    fn test_fast_inc_carry() {
+        let mut val = Vec::from(".999_".as_bytes());
+        let mut start: usize = 1;
+        let inc = "1".as_bytes();
+        fast_inc(val.as_mut(), &mut start, 4, inc);
+        assert_eq!(start, 0);
+        assert_eq!(val, "1000_".as_bytes());
+
+        let mut val = Vec::from(".999_".as_bytes());
+        let mut start: usize = 1;
+        let inc = "11".as_bytes();
+        fast_inc(val.as_mut(), &mut start, 4, inc);
+        assert_eq!(start, 0);
+        assert_eq!(val, "1010_".as_bytes());
+    }
+
+    #[test]
+    fn test_fast_inc_one_simple() {
+        let mut val = Vec::from("...8_".as_bytes());
+        let mut start: usize = 3;
+        fast_inc_one(val.as_mut(), &mut start, 4);
+        assert_eq!(start, 3);
+        assert_eq!(val, "...9_".as_bytes());
+        fast_inc_one(val.as_mut(), &mut start, 4);
+        assert_eq!(start, 2); // carried 1 more digit
+        assert_eq!(val, "..10_".as_bytes());
+        fast_inc_one(val.as_mut(), &mut start, 4);
+        assert_eq!(start, 2);
+        assert_eq!(val, "..11_".as_bytes());
+
+        let mut val = Vec::from("0_".as_bytes());
+        let mut start: usize = 0;
+        fast_inc_one(val.as_mut(), &mut start, 1);
+        assert_eq!(start, 0);
+        assert_eq!(val, "1_".as_bytes());
+        fast_inc_one(val.as_mut(), &mut start, 1);
+        assert_eq!(start, 0);
+        assert_eq!(val, "2_".as_bytes());
+        fast_inc_one(val.as_mut(), &mut start, 1);
+        assert_eq!(start, 0);
+        assert_eq!(val, "3_".as_bytes());
+    }
+}

src/uucore/src/lib/lib.rs

@@ -45,6 +45,8 @@ pub use crate::features::custom_tz_fmt;
 pub use crate::features::encoding;
 #[cfg(feature = "extendedbigdecimal")]
 pub use crate::features::extendedbigdecimal;
+#[cfg(feature = "fast-inc")]
+pub use crate::features::fast_inc;
 #[cfg(feature = "format")]
 pub use crate::features::format;
 #[cfg(feature = "fs")]

tests/by-util/test_seq.rs

@@ -216,6 +216,10 @@ fn test_separator_and_terminator() {
         .args(&["-s", ",", "2", "6"])
         .succeeds()
         .stdout_is("2,3,4,5,6\n");
+    new_ucmd!()
+        .args(&["-s", "", "2", "6"])
+        .succeeds()
+        .stdout_is("23456\n");
     new_ucmd!()
         .args(&["-s", "\n", "2", "6"])
         .succeeds()
@@ -286,6 +290,10 @@ fn test_separator_and_terminator_floats() {
         .args(&["-s", ",", "-t", "!", "2.0", "6"])
         .succeeds()
         .stdout_is("2.0,3.0,4.0,5.0,6.0!");
+    new_ucmd!()
+        .args(&["-s", "", "-t", "!", "2.0", "6"])
+        .succeeds()
+        .stdout_is("2.03.04.05.06.0!");
 }
 
 #[test]