The afsort crate implements a sorting algorithm based on
American Flag sort. The implementation is
currently limited to sort byte slices, e.g. Strings. The main motivation is to sort strings of
text, so most of the benchmarks are based on English text strings. When sorting English words,
this implementation seems to be about 40% faster than
sort_unstable from the Rust standard
For small input, this method falls back to the standard library.
Add the depndency to your
[dependencies] afsort = "0.1"
In your crate root:
ignore extern crate afsort;
You can now use afsort to e.g. sort arrays of strings or string slices.
use afsort; let mut strings = vec!("red", "green", "blue"); afsort::sort_unstable(&mut strings); assert_eq!(strings, vec!["blue", "green", "red"]);
You can also sort by an extractor function, e.g.:
use afsort; let mut tuples = vec![("b", 2), ("a", 1)]; afsort::sort_unstable_by(&mut tuples, |t: &(&str, _) | t.0.as_bytes()); assert_eq!(tuples, vec![("a", 1), ("b", 2)]);
Essentially, I noticed that sorting of strings took a long time when using the fst crate, since it requires the input to be ordered. Since sorting strings is a general problem, this is now a crate.
As mentioned, this implementation seems to be about 40% faster than the sort in the standard library, when sorting strings of random English words. The implementation is fairly naive, so I would not be surprised if it could be improved further.
You can run the benchmark tests using
cargo bench (currently requires nightly rust), like this:
% cargo bench Compiling afsort v0.1.0 (file:///home/anton/dev/off/afsort) Finished release [optimized] target(s) in 5.66 secs Running target/release/deps/afsort-ca28db3ba0643253 running 12 tests test tests::sorts_strings_same_as_unstable ... ignored test tests::sorts_tuples_same_as_unstable ... ignored test tests::sort_100000_en_af ... bench: 20,402,968 ns/iter (+/- 1,512,907) test tests::sort_100000_en_std ... bench: 30,132,067 ns/iter (+/- 1,698,223) test tests::sort_10000_en_af ... bench: 1,377,303 ns/iter (+/- 114,481) test tests::sort_10000_en_lower_af ... bench: 1,371,022 ns/iter (+/- 95,391) test tests::sort_10000_en_lower_std ... bench: 2,227,486 ns/iter (+/- 127,281) test tests::sort_10000_en_sorted_af ... bench: 878,665 ns/iter (+/- 545,256) test tests::sort_10000_en_sorted_std ... bench: 618,329 ns/iter (+/- 536,338) test tests::sort_10000_en_std ... bench: 2,221,089 ns/iter (+/- 157,461) test tests::sort_1000_en_af ... bench: 101,625 ns/iter (+/- 6,946) test tests::sort_1000_en_std ... bench: 171,655 ns/iter (+/- 9,844) test result: ok. 0 passed; 0 failed; 2 ignored; 10 measured; 0 filtered out
Currently, this crate only supports sorting elements of
AsRef<[u8]>. I think that this is a
good first step, since it supports sorting of Strings. There is however no reason why American
Flag sorting should be limited to this data type. Any kind of element that can deliver a radix
to a certain digit/depth can be sorted using this technique.
The American Flag algorithm is unstable, in the same way that sort_unstable in the standard library. That is, equal elements might be re-ordered.
This crate can only sort strings based on their
utf-8 byte values. For many problems, this
is fine. However, if you want to sort strings for display to a user, Locale might matter. This
crate does not try to address this issue.
Testing is done using the quickcheck crate. We run about 50k different variations of input strings. We treat the standard library's sort_unstable as the gold standard.
Enhances slices of e.g. Strings to have a
Main sort method.
Sort method which accepts function to convert elements to &[u8].