#numeric #runtime #format #formatting


Format numbers according to a format spec determined at runtime

3 releases

0.1.2 Feb 24, 2021
0.1.1 Feb 23, 2021
0.1.0 Feb 23, 2021

#632 in Rust patterns

Download history 266/week @ 2023-02-02 289/week @ 2023-02-09 210/week @ 2023-02-16 92/week @ 2023-02-23 22/week @ 2023-03-02 124/week @ 2023-03-09 212/week @ 2023-03-16 81/week @ 2023-03-23 151/week @ 2023-03-30 122/week @ 2023-04-06 244/week @ 2023-04-13 95/week @ 2023-04-20 195/week @ 2023-04-27 110/week @ 2023-05-04 129/week @ 2023-05-11 91/week @ 2023-05-18

536 downloads per month
Used in calc


965 lines


Format numbers according to runtime specifications.

Build and Test Status

Why another numeric formatting crate?

This crate implements numeric formatting with a combination of properties not found elsewhere:

  • Runtime: The format specifiers do not need to be known in advance
  • Flexible: The format specification supports nearly a superset of the features of the standard library.
  • Focused: Keeps things simple by offering a very simple interface which can format a single number at a time. More complicated formatting jobs can be accomplished by passing the output into standard formatting machinery.


  • std::fmt is expressive and built into the standard library. However, you must provide the format string at compile time, and it can't handle digit separation.
  • num-format is fairly comprehensive and supports no-std applications. However, it doesn't support non-decimal output modes or arbitrary groupings.
  • runtime-fmt allows for runtime format strings which support everything the standard library does, but it's nightly-only.
  • fomat-macros provides alternate formatting macros with their own syntax, but which appear to be somewhat less powerful than those of the stdlib.

Format String Reference

The formatters in this crate implement a near superset of features available in the format macros provided by std::fmt. However, they are exclusively concerned with formatting a single numeric value. Therefore, the specification language is somewhat truncated: it omits both braces and the colon which precedes the format specification. Therefore, where in the standard formatting machinery you might write:

let hex_digit = format!("{:02x}", 0xf0);

With this library, the equivalent would be:

let hex_digit = NumFmt::from_str("02x")?.fmt(0xf0);

Note: though these formatters support a superset of features of the standard ones, in that anything possible with the standard formatters is possible with this library, they do not have a superset of the syntax: while the intersection is large, there are a few syntax elements legal in the standard formatter which are not legal for this formatter. In particular, the standard formatter supports named width parameters. Dynamic width parameters are legal in this crate, but they cannot be named.


The gramar for the format string derives substantially from the standard library's:

format_spec := [[fill]align][sign]['#'][['0']width]['.' precision][format][separator[spacing]]
fill := character
align := '<' | '^' | '>' | 'v'
sign := '+' | '-'
width := integer not beginning with '0'
precision := integer
format := 'b' | 'o' | 'd' | 'x' | 'X'
separator := '_', | ',' | ' '
spacing := integer

Note: there is no special syntax for dynamic insertion of with, precision and spacing. Simply use NumFmt::format_with; the dynamic values there always override any values for those fields, whether set or unset in the format string.


Any single char which precedes an align specifier is construed as the fill character: when width is greater than the actual rendered width of the number, the excess is padded with this character.

Note: Wide characters are counted according to their bit width, not their quantity.

let heart = '🖤';
assert_eq!(heart.len_utf8(), 4);
let fmt = NumFmt::builder().fill(heart).width(6).build();
// Note that this renders as two characters: we requested a width of 6.
// The number renders as a single character. The heart fills up the next 4 for a total of 5.
// Adding an extra heart would exceed the requested width, so it only renders one.
assert_eq!(fmt.fmt(1), "🖤1");


  • >: the output is right-aligned in width columns (default).
  • ^: the output is centered in width columns.
  • <: the output is left-aligned in width columns.
  • v: attempt to align the decimal point at column index width. For integers, equivalent to >.


  • -: print a leading - for negative numbers, and nothing in particular for positive (default)
  • +: print a leading + for positive numbers


If a # character is present, print a base specification before the number according to its format (see format below).

  • binary: 0b
  • octal: 0o
  • decimal: 0d
  • hex: 0x

This base specification counts toward the width of the number:

assert_eq!(NumFmt::from_str("#04b").unwrap().fmt(2), "0b10");


Engage the zero handler.

The zero handler overrides the padding specification to 0, and treats pad characters as part of the number, in contrast to the default behavior which treats them as arbitrary spacing.


// sign handling
assert_eq!(NumFmt::from_str("-03").unwrap().fmt(-1).unwrap(),   "-01");
assert_eq!(NumFmt::from_str("0>-3").unwrap().fmt(-1).unwrap(), "-001");
// separator handling
assert_eq!(NumFmt::from_str("0>7,").fmt(1).unwrap(), "0000001");
assert_eq!(NumFmt::from_str("07,").fmt(1).unwrap(),  "000,001");


This is a parameter for the "minimum width" that the format should take up. If the value's string does not fill up this many characters, then the padding specified by fill/alignment will be used to take up the required space (see fill above).

When using the $ sigil instead of an explicit width, the width can be set dynamically:

assert_eq!(NumFmt::from_str("-^$").unwrap().fmt_with(1, Dynamic::width(5)), "--1--");

If an explicit width is not provided, defaults to 0.


How many digits after the decimal point are printed. Note that integers can be forced to emit decimal places with this modifier.

If an explicit precision is not provided, defaults to emitting all post-decimal digits emitted by the underlying type.

assert_eq!(NumFmt::from_str(".2").unwrap().fmt(3.14159).unwrap(), "3.14");
assert_eq!(NumFmt::from_str(".7").unwrap().fmt(3.14159).unwrap(), "3.1415900");

If the requested precision exceeds the native precision available to this number, the remainder is always filled with '0', even if fill is specified:

assert_eq!(NumFmt::from_str("-<6.2").unwrap().fmt(1.0_f32).unwrap(), "1.00--");


  • b: Emit this number's binary representation
  • o: Emit this number's octal representation
  • d: Emit this number's decimal representation (default)
  • x: Emit this number's hexadecimal representation with lowercase letters
  • X: Emit this number's hexadecimal representation with uppercase letters

Note: This is one of a few areas where the standard library has capabilities this library does not: it supports some other numeric formats. Pull requests welcomed to bring this up to parity.


A separator is a (typically non-numeric) character inserted between groups of digits to make it easier for humans to parse the number when reading. Different separators may be desirable in different contexts.

  • _: Separate numeric groups with an underscore
  • ,: Separate numeric groups with a comma
  • (space char): Separate numeric groups with a space

By default, numeric groups are not separated. It is not possible to explicitly specify that numeric groups are not separated when using a format string. However, this can be specified when building the formatter via builder.

Wyhen using the builder to explicitly set formatter options, it is also possible to separate numeric groups with an arbitrary char. This can be desirable to i.e. support German number formats, which use a . to separate numeric groups and a , as a decimal separator.


Spacing determines the number of characters in each character group. It is only of interest when the separator is set. The default spacing is 3.

Apparently some cultures separate numeric digits with a non-constant group size. Please file an issue if this feature is important to you.

Decimal separator

When using the builder to explicitly set formatter options, it is possible to set the decimal separator to any char. This can be desirable to i.e. support German number formats, which use a . to separate numeric groups and a , as a decimal separator.


~67K SLoC