v0.3.0

Cargo.toml

@@ -1,28 +1,34 @@
 [package]
 name = "median-accumulator"
-version = "0.2.0"
+version = "0.4.0"
 edition = "2021"
 authors = ["tuxmain <tuxmain@zettascript.org>"]
 license = "AGPL-3.0-only"
 repository = "https://git.txmn.tk/tuxmain/median-accumulator"
 documentation = "https://docs.rs/median-accumulator/"
-description = "Simple, fast, space-efficient accumulator for computing median"
+description = "Simple, fast, space-efficient, generic accumulator for computing median"
 categories = ["algorithms", "data-structures", "no-std"]
+keywords = ["median"]
 
 [dependencies]
-cc-traits = "1.0.0"
+cc-traits = { version = "2.0.0", default_features = false }
 smallvec = { version = "^1.6", optional = true }
 
 [features]
-nostd = ["cc-traits/nostd"]
+std = ["cc-traits/alloc", "cc-traits/std"]
 smallvec = ["dep:smallvec", "cc-traits/smallvec"]
 
+default = ["std"]
+
 [dev-dependencies]
-criterion = { version = "0.4.0", features = ["html_reports"] }
-medianheap = "0.3.0"
+criterion = { version = "0.5.1", features = ["html_reports"] }
+medianheap = "0.4.1"
 rand = "0.8.5"
 smallvec = "^1.6"
 
 [[bench]]
 name = "comparison"
 harness = false
+
+[package.metadata.docs.rs]
+features = ["std"]

README.md

@@ -7,11 +7,10 @@ Simple, space-efficient algorithm to compute the median of an accumulation of el
 * **Space-efficient**: `O(D)` space, D being the number of _different_ samples, not the _total_ number of samples
 * **Time-efficient**: push is `O(log(N))`
 * **Generic**: `T: Clone + Ord`
-* **Tested**
-* **No unsafe**, no deps
+* **No unsafe**
 * **no_std** (optional): supports generic collections
 
-Faster than other implementations if there are samples having the same value. If this is not your case, you should use another implementation.
+Faster than other implementations if lots of samples have the same value. If this is not your case, you should use another implementation.
 
 ## Use
 
@@ -45,7 +44,7 @@ For other collections than `Vec` or `SmallVec`, you must implement [cc-traits](h
 
 ## License
 
-CopyLeft 2022-2023 Pascal Engélibert [(why copyleft?)](https://txmn.tk/blog/why-copyleft/)
+CopyLeft 2022-2024 Pascal Engélibert [(why copyleft?)](https://txmn.tk/blog/why-copyleft/)
 
 This program is free software: you can redistribute it and/or modify it under the terms of the GNU Affero General Public License as published by the Free Software Foundation, version 3 of the License.
 

benches/comparison.rs

@@ -1,54 +1,34 @@
-use criterion::{criterion_group, criterion_main, BenchmarkId, Criterion};
+use criterion::{black_box, criterion_group, criterion_main, BenchmarkId, Criterion};
 use rand::Rng;
 
+static ITERS: u32 = 10_000;
+
 fn compare_crates(c: &mut Criterion) {
 	let mut rng = rand::thread_rng();
 	let mut group = c.benchmark_group("Comparison");
-	for len in [10, 50, 100, 500, 1000] {
-		let samples: Vec<u32> = (0..len).map(|_| rng.gen_range(0..len / 5)).collect();
+	for redundancy in [1, 5, 10, 20, 40] {
+		let samples: Vec<u32> = (0..ITERS)
+			.map(|_| rng.gen_range(0..ITERS / redundancy))
+			.collect();
 		group.bench_with_input(
-			BenchmarkId::new("median_accumulator 1:5", len),
+			BenchmarkId::new("median_accumulator", redundancy),
 			&samples,
 			|b, _i| {
 				b.iter(|| {
 					let mut median = median_accumulator::vec::MedianAcc::new();
 					samples.iter().for_each(|s| median.push(*s));
-					median.get_median()
+					black_box(median.get_median());
 				})
 			},
 		);
 		group.bench_with_input(
-			BenchmarkId::new("medianheap 1:5", len),
+			BenchmarkId::new("medianheap", redundancy),
 			&samples,
 			|b, _i| {
 				b.iter(|| {
 					let mut median = medianheap::MedianHeap::new();
 					samples.iter().for_each(|s| median.push(*s));
-					median.median()
-				})
-			},
-		);
-
-		let samples: Vec<u32> = (0..len).map(|_| rng.gen_range(0..len)).collect();
-		group.bench_with_input(
-			BenchmarkId::new("median_accumulator 1:1", len),
-			&samples,
-			|b, _i| {
-				b.iter(|| {
-					let mut median = median_accumulator::vec::MedianAcc::new();
-					samples.iter().for_each(|s| median.push(*s));
-					median.get_median()
-				})
-			},
-		);
-		group.bench_with_input(
-			BenchmarkId::new("medianheap 1:1", len),
-			&samples,
-			|b, _i| {
-				b.iter(|| {
-					let mut median = medianheap::MedianHeap::new();
-					samples.iter().for_each(|s| median.push(*s));
-					median.median()
+					black_box(median.median());
 				})
 			},
 		);

src/lib.rs

@@ -14,7 +14,7 @@
 //!
 //! In doc comments, _N_ represents the number of samples, _D_ represents the number of different values taken by the samples.
 
-#![cfg_attr(feature = "nostd", no_std)]
+#![cfg_attr(not(feature = "std"), no_std)]
 
 mod traits;
 
@@ -34,7 +34,7 @@ pub struct MedianAcc<
 	_t: core::marker::PhantomData<T>,
 }
 
-#[cfg(not(feature = "nostd"))]
+#[cfg(feature = "std")]
 pub mod vec {
 	pub type MedianAcc<T> = crate::MedianAcc<T, Vec<(T, u32)>>;
 }
@@ -213,6 +213,7 @@ mod tests {
 
 	use rand::Rng;
 
+	#[cfg(feature = "std")]
 	fn naive_median<T: Clone + Ord>(samples: &mut [T]) -> Option<MedianResult<T>> {
 		if samples.is_empty() {
 			None
@@ -232,6 +233,7 @@ mod tests {
 		}
 	}
 
+	#[cfg(feature = "std")]
 	#[test]
 	fn correctness() {
 		let mut rng = rand::thread_rng();
@@ -249,6 +251,7 @@ mod tests {
 		}
 	}
 
+	#[cfg(feature = "smallvec")]
 	#[test]
 	fn correctness_smallvec() {
 		let mut rng = rand::thread_rng();

src/traits.rs

@@ -9,7 +9,7 @@ pub trait InsertIndex: cc_traits::Collection {
 	) -> Self::Output;
 }
 
-#[cfg(not(feature = "nostd"))]
+#[cfg(feature = "std")]
 impl<T> InsertIndex for Vec<T> {
 	type Output = ();