chore: consolidated features for disk-io

Earthfile

@@ -56,7 +56,7 @@ build:
 # This target formats the project.
 fmt:
     FROM +chef-cook
-    RUN cargo fmt --all --all-features -- --check && rye fmt --all --check
+    RUN cargo fmt --all -- --check && rye fmt --all --check
 
 # This target lints the project.
 lint:
@@ -77,7 +77,7 @@ fix:
 # This target runs the tests.
 test:
     FROM +chef-cook
-    RUN cargo test --release --lib --bins --examples --tests --all-features
+    RUN cargo test -r
     # TODO: switch to --all, blocked on https://github.com/astral-sh/rye/issues/853
     RUN rye test --package abd-distances
 

crates/abd-clam/Cargo.toml

@@ -10,7 +10,7 @@ authors = [
     "Oliver McLaughlin <[email protected]>",
 ]
 edition = "2021"
-rust-version = "1.79"
+rust-version = "1.83"
 description = "Clustering, Learning and Approximation with Manifolds"
 license = "MIT"
 readme = "./README.md"
@@ -43,8 +43,7 @@ ndarray-npy = { workspace = true, optional = true }
 
 # For:
 #  - CHAODA
-smartcore = { git = "https://github.com/smartcorelib/smartcore.git", rev = "239c00428f7448d30b78bf8653923f6bc0e2c29b", features = ["serde"], optional = true }
-ordered-float = { version = "4.2.2", optional = true }
+smartcore = { version = "0.4", features = ["serde"], optional = true }
 
 # For:
 #  - MSA
@@ -58,20 +57,18 @@ flate2 = { workspace = true, optional = true }
 [dev-dependencies]
 symagen = { workspace = true }
 bitcode = { workspace = true }
-criterion = { version = "0.5.1", features = ["html_reports"] }
+criterion = { version = "0.5", features = ["html_reports"] }
 tempdir = "0.3.7"
 float-cmp = "0.10.0"
 test-case = "3.2.1"
 statistical = "1.0.0"
 
 [features]
-csv = ["dep:csv"]
-bitcode = ["dep:bitcode", "dep:flate2"]
-ndarray-bindings = ["dep:ndarray", "dep:ndarray-npy"]
-chaoda = ["dep:smartcore", "dep:ordered-float", "bitcode"]
+disk-io = ["dep:csv", "dep:bitcode", "dep:flate2", "dep:ndarray", "dep:ndarray-npy"]
+chaoda = ["dep:smartcore"]
 mbed = ["chaoda"]
 msa = ["dep:stringzilla"]
-all = ["csv", "bitcode", "ndarray-bindings", "chaoda", "mbed", "msa"]
+all = ["disk-io", "chaoda", "mbed", "msa"]
 
 [[bench]]
 name = "vector_search"

crates/abd-clam/src/cakes/cluster/permuted_ball.rs

@@ -21,8 +21,8 @@ use crate::{
 /// - `T`: The type of the distance values.
 /// - `S`: The `Cluster` type that the `PermutedBall` is based on.
 #[derive(Clone, Serialize, Deserialize)]
-#[cfg_attr(feature = "bitcode", derive(bitcode::Encode, bitcode::Decode))]
-#[cfg_attr(feature = "bitcode", bitcode(recursive))]
+#[cfg_attr(feature = "disk-io", derive(bitcode::Encode, bitcode::Decode))]
+#[cfg_attr(feature = "disk-io", bitcode(recursive))]
 pub struct PermutedBall<T: Number, S: Cluster<T>> {
     /// The `Cluster` type that the `PermutedBall` is based on.
     source: S,
@@ -155,7 +155,7 @@ impl<T: Number, S: ParCluster<T>> ParCluster<T> for PermutedBall<T, S> {
 
 /// Parameters for adapting the `PermutedBall`.
 #[derive(Debug, Default, Copy, Clone, Serialize, Deserialize)]
-#[cfg_attr(feature = "bitcode", derive(bitcode::Encode, bitcode::Decode))]
+#[cfg_attr(feature = "disk-io", derive(bitcode::Encode, bitcode::Decode))]
 pub struct Offset {
     /// The offset of the slice of indices of the `Cluster` in the reordered
     /// dataset.
@@ -271,7 +271,7 @@ impl<I: Send + Sync, T: Number, D: ParDataset<I> + Permutable, S: ParCluster<T>>
     }
 }
 
-#[cfg(feature = "csv")]
+#[cfg(feature = "disk-io")]
 impl<T: Number, S: crate::cluster::Csv<T>> crate::cluster::Csv<T> for PermutedBall<T, S> {
     fn header(&self) -> Vec<String> {
         let mut header = self.source.header();
@@ -290,13 +290,13 @@ impl<T: Number, S: crate::cluster::Csv<T>> crate::cluster::Csv<T> for PermutedBa
     }
 }
 
-#[cfg(feature = "csv")]
+#[cfg(feature = "disk-io")]
 impl<T: Number, S: crate::cluster::ParCsv<T>> crate::cluster::ParCsv<T> for PermutedBall<T, S> {}
 
-#[cfg(feature = "bitcode")]
+#[cfg(feature = "disk-io")]
 impl<T: Number, S: crate::cluster::ClusterIO<T>> crate::cluster::ClusterIO<T> for PermutedBall<T, S> {}
 
-#[cfg(feature = "bitcode")]
+#[cfg(feature = "disk-io")]
 impl<T: Number, S: crate::cluster::ParClusterIO<T>> crate::cluster::ParClusterIO<T> for PermutedBall<T, S> {}
 
 #[cfg(test)]

crates/abd-clam/src/chaoda/graph/mod.rs

@@ -3,13 +3,12 @@
 
 use core::cmp::Reverse;
 
-use std::collections::{BinaryHeap, HashMap};
+use std::collections::HashMap;
 
 use distances::Number;
-use ordered_float::OrderedFloat;
 use rayon::prelude::*;
 
-use crate::{cluster::ParCluster, dataset::ParDataset, metric::ParMetric, Cluster, Dataset, Metric};
+use crate::{cluster::ParCluster, dataset::ParDataset, metric::ParMetric, Cluster, Dataset, Metric, SizedHeap};
 
 use super::Vertex;
 
@@ -90,10 +89,10 @@ impl<'a, T: Number, S: Cluster<T>> Graph<'a, T, S> {
         // `Vertex`es are selected by highest score and then by shallowest depth.
         let mut candidates = clusters
             .into_iter()
-            .zip(scores.into_iter().map(OrderedFloat))
+            .zip(scores)
             .filter(|(c, _)| c.is_leaf() || c.depth() >= min_depth)
             .map(|(c, s)| (s, Reverse(c)))
-            .collect::<BinaryHeap<_>>();
+            .collect::<SizedHeap<_>>();
 
         let mut clusters = vec![];
         while let Some((_, Reverse(v))) = candidates.pop() {
@@ -239,10 +238,10 @@ impl<'a, T: Number, S: ParCluster<T>> Graph<'a, T, S> {
         // `Vertex`es are selected by highest score and then by shallowest depth.
         let mut candidates = clusters
             .into_iter()
-            .zip(scores.into_iter().map(OrderedFloat))
+            .zip(scores)
             .filter(|(c, _)| c.is_leaf() || c.depth() >= min_depth)
             .map(|(c, s)| (s, Reverse(c)))
-            .collect::<BinaryHeap<_>>();
+            .collect::<SizedHeap<_>>();
 
         let mut clusters = vec![];
         while let Some((_, Reverse(v))) = candidates.pop() {

crates/abd-clam/src/core/cluster/ball.rs

@@ -19,8 +19,8 @@ use super::{partition::ParPartition, Cluster, ParCluster, Partition, LFD};
 /// A metric-`Ball` is a collection of items that are within a certain distance
 /// of a center.
 #[derive(Clone, Serialize, Deserialize)]
-#[cfg_attr(feature = "bitcode", derive(bitcode::Encode, bitcode::Decode))]
-#[cfg_attr(feature = "bitcode", bitcode(recursive))]
+#[cfg_attr(feature = "disk-io", derive(bitcode::Encode, bitcode::Decode))]
+#[cfg_attr(feature = "disk-io", bitcode(recursive))]
 pub struct Ball<T: Number> {
     /// Parameters used for creating the `Ball`.
     depth: usize,
@@ -268,7 +268,7 @@ impl<T: Number> ParPartition<T> for Ball<T> {
     }
 }
 
-#[cfg(feature = "csv")]
+#[cfg(feature = "disk-io")]
 impl<T: Number> super::Csv<T> for Ball<T> {
     fn header(&self) -> Vec<String> {
         vec![
@@ -295,13 +295,13 @@ impl<T: Number> super::Csv<T> for Ball<T> {
     }
 }
 
-#[cfg(feature = "csv")]
+#[cfg(feature = "disk-io")]
 impl<T: Number> super::ParCsv<T> for Ball<T> {}
 
-#[cfg(feature = "bitcode")]
+#[cfg(feature = "disk-io")]
 impl<T: Number> super::ClusterIO<T> for Ball<T> {}
 
-#[cfg(feature = "bitcode")]
+#[cfg(feature = "disk-io")]
 impl<T: Number> super::ParClusterIO<T> for Ball<T> {}
 
 #[cfg(test)]

crates/abd-clam/src/core/cluster/io.rs

@@ -7,7 +7,7 @@ use rayon::prelude::*;
 
 use super::{Cluster, ParCluster};
 
-#[cfg(feature = "csv")]
+#[cfg(feature = "disk-io")]
 /// Write a tree to a CSV file.
 pub trait Csv<T: Number>: Cluster<T> {
     /// Returns the names of the columns in the CSV file.
@@ -45,7 +45,7 @@ pub trait Csv<T: Number>: Cluster<T> {
     }
 }
 
-#[cfg(feature = "csv")]
+#[cfg(feature = "disk-io")]
 /// Parallel version of `Csv`.
 pub trait ParCsv<T: Number>: Csv<T> + ParCluster<T> {
     /// Parallel version of `Csv::write_to_csv`.
@@ -84,7 +84,7 @@ pub trait ParCsv<T: Number>: Csv<T> + ParCluster<T> {
     }
 }
 
-#[cfg(feature = "bitcode")]
+#[cfg(feature = "disk-io")]
 /// Reading and writing `Cluster` trees to disk using `bitcode`.
 pub trait ClusterIO<T: Number>: Cluster<T> {
     /// Writes the `Cluster` to disk in binary format using `bitcode`.
@@ -116,7 +116,7 @@ pub trait ClusterIO<T: Number>: Cluster<T> {
     }
 }
 
-#[cfg(feature = "bitcode")]
+#[cfg(feature = "disk-io")]
 /// Parallel version of `ClusterIO`.
 pub trait ParClusterIO<T: Number>: ParCluster<T> + ClusterIO<T> {
     /// Parallel version of `ClusterIO::write_to`.

crates/abd-clam/src/core/cluster/mod.rs

@@ -11,7 +11,6 @@ use super::{
 pub mod adapter;
 mod balanced_ball;
 mod ball;
-mod io;
 mod lfd;
 mod partition;
 
@@ -20,12 +19,12 @@ pub use ball::Ball;
 pub use lfd::LFD;
 pub use partition::{ParPartition, Partition};
 
-#[cfg(feature = "csv")]
-pub use io::{Csv, ParCsv};
+#[cfg(feature = "disk-io")]
+mod io;
 
-#[cfg(feature = "bitcode")]
+#[cfg(feature = "disk-io")]
 #[allow(clippy::module_name_repetitions)]
-pub use io::{ClusterIO, ParClusterIO};
+pub use io::{ClusterIO, Csv, ParClusterIO, ParCsv};
 
 /// A `Cluster` is a collection of "similar" items in a dataset.
 ///

crates/abd-clam/src/core/dataset/flat_vec.rs

@@ -11,8 +11,8 @@ use super::{AssociatesMetadata, AssociatesMetadataMut, Dataset, ParDataset, Perm
 /// - `I`: The type of the items in the dataset.
 /// - `Me`: The type of the metadata associated with the items.
 #[derive(Clone, Serialize, Deserialize)]
-#[cfg_attr(feature = "bitcode", derive(bitcode::Encode, bitcode::Decode))]
-#[cfg_attr(feature = "bitcode", bitcode(recursive))]
+#[cfg_attr(feature = "disk-io", derive(bitcode::Encode, bitcode::Decode))]
+#[cfg_attr(feature = "disk-io", bitcode(recursive))]
 pub struct FlatVec<I, Me> {
     /// The items in the dataset.
     items: Vec<I>,
@@ -223,13 +223,13 @@ impl<I, Me> Permutable for FlatVec<I, Me> {
     }
 }
 
-#[cfg(feature = "bitcode")]
+#[cfg(feature = "disk-io")]
 impl<I, Me> super::DatasetIO<I> for FlatVec<I, Me> {}
 
-#[cfg(feature = "bitcode")]
+#[cfg(feature = "disk-io")]
 impl<I: Send + Sync, Me: Send + Sync> super::ParDatasetIO<I> for FlatVec<I, Me> {}
 
-#[cfg(feature = "ndarray-bindings")]
+#[cfg(feature = "disk-io")]
 impl<T: ndarray_npy::ReadableElement + Copy> FlatVec<Vec<T>, usize> {
     /// Reads a `FlatVec` from a `.npy` file.
     ///
@@ -250,7 +250,7 @@ impl<T: ndarray_npy::ReadableElement + Copy> FlatVec<Vec<T>, usize> {
     }
 }
 
-#[cfg(feature = "ndarray-bindings")]
+#[cfg(feature = "disk-io")]
 impl<T: ndarray_npy::WritableElement + Copy> FlatVec<Vec<T>, usize> {
     /// Writes the `FlatVec` to a `.npy` file in the given directory.
     ///
@@ -282,7 +282,7 @@ impl<T: ndarray_npy::WritableElement + Copy> FlatVec<Vec<T>, usize> {
     }
 }
 
-#[cfg(feature = "csv")]
+#[cfg(feature = "disk-io")]
 impl<T: std::str::FromStr + Copy> FlatVec<Vec<T>, usize> {
     /// Reads a `FlatVec` from a `.csv` file.
     ///
@@ -321,7 +321,7 @@ impl<T: std::str::FromStr + Copy> FlatVec<Vec<T>, usize> {
     }
 }
 
-#[cfg(feature = "csv")]
+#[cfg(feature = "disk-io")]
 impl<T: std::string::ToString + Copy, M> FlatVec<Vec<T>, M> {
     /// Writes the `FlatVec` to a `.csv` file with the given path.
     ///
@@ -376,7 +376,7 @@ mod tests {
         Ok(())
     }
 
-    #[cfg(feature = "ndarray-bindings")]
+    #[cfg(feature = "disk-io")]
     #[test]
     fn npy_io() -> Result<(), String> {
         let items = vec![vec![1, 2], vec![3, 4], vec![5, 6]];

crates/abd-clam/src/core/dataset/io.rs

@@ -2,7 +2,7 @@
 
 use super::{Dataset, ParDataset};
 
-#[cfg(feature = "bitcode")]
+#[cfg(feature = "disk-io")]
 /// For writing and reading datasets to and from disk.
 pub trait DatasetIO<I>: Dataset<I> {
     /// Writes the `Dataset` to disk in binary format using `bitcode`.
@@ -34,7 +34,7 @@ pub trait DatasetIO<I>: Dataset<I> {
     }
 }
 
-#[cfg(feature = "bitcode")]
+#[cfg(feature = "disk-io")]
 /// Parallel version of `DatasetIO`.
 pub trait ParDatasetIO<I: Send + Sync>: DatasetIO<I> + ParDataset<I> {
     /// Parallel version of `DatasetIO::write_to`.
@@ -45,7 +45,6 @@ pub trait ParDatasetIO<I: Send + Sync>: DatasetIO<I> + ParDataset<I> {
     ///
     /// - If the dataset cannot be encoded.
     /// - If the file cannot be written.
-    #[cfg(feature = "bitcode")]
     fn par_write_to<P: AsRef<std::path::Path>>(&self, path: &P) -> Result<(), String>
     where
         Self: bitcode::Encode,
@@ -61,7 +60,6 @@ pub trait ParDatasetIO<I: Send + Sync>: DatasetIO<I> + ParDataset<I> {
     ///
     /// - If the file cannot be read.
     /// - If the dataset cannot be decoded.
-    #[cfg(feature = "bitcode")]
     fn par_read_from<P: AsRef<std::path::Path>>(path: &P) -> Result<Self, String>
     where
         Self: bitcode::Decode,

crates/abd-clam/src/core/dataset/mod.rs

@@ -8,7 +8,6 @@ use super::{metric::ParMetric, Metric};
 
 mod associates_metadata;
 mod flat_vec;
-mod io;
 mod permutable;
 mod sized_heap;
 
@@ -17,7 +16,10 @@ pub use flat_vec::FlatVec;
 pub use permutable::Permutable;
 pub use sized_heap::SizedHeap;
 
-#[cfg(feature = "bitcode")]
+#[cfg(feature = "disk-io")]
+mod io;
+
+#[cfg(feature = "disk-io")]
 #[allow(clippy::module_name_repetitions)]
 pub use io::{DatasetIO, ParDatasetIO};