This is a library for formatting numbers with numeric prefixes, such as turning “3000 metres” into “3 kilometres”, or “8705 bytes” into “8.5 KiB”.
This crate works with Cargo. Add the following to your Cargo.toml
dependencies section:
[dependencies]
number_prefix = "0.4"
This crate has no_std
support. To activate it, disable the std
Cargo feature.
The earliest version of Rust that this crate is tested against is Rust v1.31.0.
The function NumberPrefix::decimal
returns either a pair of the resulting number and its prefix, or a notice that the number was too small to have any prefix applied to it.
For example:
use number_prefix::NumberPrefix;
let amount = 8542_f32;
let result = match NumberPrefix::decimal(amount) {
NumberPrefix::Standalone(bytes) => {
format!("The file is {} bytes in size", bytes)
}
NumberPrefix::Prefixed(prefix, n) => {
format!("The file is {:.1} {}B in size", n, prefix)
}
};
assert_eq!("The file is 8.5 kB in size", result);
The {:.1}
part of the formatting string tells it to restrict the output to only one decimal place.
This value is calculated by repeatedly dividing the number by 1000 until it becomes less than that, which in this case results in 8.542, which gets rounded down.
Because only one division had to take place, the function also returns the decimal prefix Kilo
, which gets converted to its internationally-recognised symbol when formatted as a string.
If the value is too small to have any prefixes applied to it — in this case, if it’s under 1000 — then the standalone value will be returned:
use number_prefix::NumberPrefix;
let amount = 705_f32;
let result = match NumberPrefix::decimal(amount) {
NumberPrefix::Standalone(bytes) => {
format!("The file is {} bytes in size", bytes)
}
NumberPrefix::Prefixed(prefix, n) => {
format!("The file is {:.1} {}B in size", n, prefix)
}
};
assert_eq!("The file is 705 bytes in size", result);
In this particular example, the user expects different formatting for both bytes and kilobytes: while prefixed values are given more precision, there’s no point using anything other than whole numbers for just byte amounts. This is why the function pays attention to values without any prefixes — they often need to be special-cased.
This library also allows you to use the binary prefixes, which use the number 1024 (210) as the multiplier, rather than the more common 1000 (103).
This uses the NumberPrefix::binary
function. For example:
use number_prefix::NumberPrefix;
let amount = 8542_f32;
let result = match NumberPrefix::binary(amount) {
NumberPrefix::Standalone(bytes) => {
format!("The file is {} bytes in size", bytes)
}
NumberPrefix::Prefixed(prefix, n) => {
format!("The file is {:.1} {}B in size", n, prefix)
}
};
assert_eq!("The file is 8.3 KiB in size", result);
A kibibyte is slightly larger than a kilobyte, so the number is smaller in the result; but other than that, it works in exactly the same way, with the binary prefix being converted to a symbol automatically.
There is no correct answer to this question! Current practice is to use the binary prefixes for numbers of bytes, while still using the decimal prefixes for everything else. Computers work with powers of two, rather than powers of ten, and by using the binary prefixes, you get a more accurate representation of the amount of data.
If you need to describe your unit in actual words, rather than just with the symbol, use one of the upper
, caps
, lower
, or symbol
, which output the prefix in a variety of formats. For example:
use number_prefix::NumberPrefix;
let amount = 8542_f32;
let result = match NumberPrefix::decimal(amount) {
NumberPrefix::Standalone(bytes) => {
format!("The file is {} bytes in size", bytes)
}
NumberPrefix::Prefixed(prefix, n) => {
format!("The file is {:.1} {}bytes in size", n, prefix.lower())
}
};
assert_eq!("The file is 8.5 kilobytes in size", result);
There is a FromStr
implementation for NumberPrefix
that parses strings containing numbers and trailing prefixes, such as 7.5E
.
Currently, the only supported units are b
and B
for bytes, and m
for metres.
Whitespace is allowed between the number and the rest of the string.
use number_prefix::{NumberPrefix, Prefix};
assert_eq!("7.05E".parse::<NumberPrefix<_>>(),
Ok(NumberPrefix::Prefixed(Prefix::Exa, 7.05_f64)));
assert_eq!("7.05".parse::<NumberPrefix<_>>(),
Ok(NumberPrefix::Standalone(7.05_f64)));
assert_eq!("7.05 GiB".parse::<NumberPrefix<_>>(),
Ok(NumberPrefix::Prefixed(Prefix::Gibi, 7.05_f64)));