Parallel sorting of SCCS is now actually parallel; Zipped lenders are sorted only if the components are sorted

Author: vigna

README.md

@@ -4,17 +4,17 @@
 [![Line count](https://tokei.rs/b1/github/vigna/webgraph-rs)](https://github.com/vigna/webgraph-rs)
 
 A Rust implementation of the [WebGraph
-framework](http://webgraph.di.unimi.it/) for graph compression.
+framework](https://webgraph.di.unimi.it/) for graph compression.
 
 ## At a Glance
 
 - Compressed graph representation (start from here):
-  [`webgraph`](http://crates.io/crates/webgraph) ([repo](/webgraph))
+  [`webgraph`](https://crates.io/crates/webgraph) ([repo](/webgraph))
 
-- Algorithms: [`webgraph_algo`](http://crates.io/crates/webgraph_algo)
+- Algorithms: [`webgraph-algo`](https://crates.io/crates/webgraph_algo)
   ([repo](/algo))
 
-- CLI commands: [`webgraph_cli`](http://crates.io/crates/webgraph_cli)
+- CLI commands: [`webgraph-cli`](https://crates.io/crates/webgraph_cli)
   ([repo](/cli))
 
 ## Acknowledgments

algo/src/sccs/mod.rs

@@ -142,7 +142,7 @@ impl<C: AsMut<[usize]> + AsRef<[usize]>> Sccs<C> {
             .as_mut()
             .par_iter_mut()
             .for_each(|node_component| *node_component = inv_perm[*node_component]);
-        sizes.sort_by(|&x, &y| y.cmp(&x));
+        sizes.par_sort_by(|&x, &y| y.cmp(&x));
         sizes
     }
 }

cli/CHANGELOG.md

@@ -13,9 +13,6 @@
 - Floats without a specified precision are now printed using the `zmij` crate,
   which brings a 3-4x speed improvement.
 
-- We use the crate `zmij` for saving floats in ASCII when no precision is
-  specified, which brings a 3-4x speed improvement.
-
 ## [0.4.0] - 2026-02-18
 
 ### New

cli/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "webgraph-cli"
-version = "0.4.0"
+version = "0.4.1"
 edition = "2024"
 authors = [
 	"Tommaso Fontana <tommaso.fontana.96@gmail.com>",

webgraph/CHANGELOG.md

@@ -10,6 +10,11 @@
 
 - `BvGraphSeq` was sub-preallocating the successor vector (+1% speed).
 
+### Improved
+
+- Temporary compression files are now immediately removed when they
+  are no longer necessary.
+
 ## [0.6.0] - 2026-02-18
 
 ### New

webgraph/src/graphs/arc_list_graph.rs

@@ -27,7 +27,7 @@ pub struct ArcListGraph<I> {
     into_iter: I,
 }
 
-impl<L: Clone + Copy + 'static, I: Iterator<Item = ((usize, usize), L)> + Clone> ArcListGraph<I> {
+impl<L: Copy + 'static, I: Iterator<Item = ((usize, usize), L)> + Clone> ArcListGraph<I> {
     /// Creates a new arc-list graph from the given [`IntoIterator`].
     #[inline(always)]
     pub fn new_labeled(num_nodes: usize, iter: I) -> Self {

webgraph/src/graphs/btree_graph.rs

@@ -24,7 +24,7 @@ use std::collections::BTreeMap;
 #[derive(Clone, Debug, PartialEq, Eq)]
 pub struct LabeledBTreeGraph<L: Clone + 'static = ()> {
     /// The number of arcs in the graph.
-    number_of_arcs: u64,
+    num_arcs: u64,
     /// For each node, its list of successors.
     succ: Vec<BTreeMap<usize, L>>,
 }
@@ -39,15 +39,15 @@ impl<L: Clone + 'static> LabeledBTreeGraph<L> {
     /// Creates a new empty graph.
     pub fn new() -> Self {
         Self {
-            number_of_arcs: 0,
+            num_arcs: 0,
             succ: vec![],
         }
     }
 
     /// Creates a new empty graph with `n` nodes.
     pub fn empty(n: usize) -> Self {
         Self {
-            number_of_arcs: 0,
+            num_arcs: 0,
             succ: Vec::from_iter((0..n).map(|_| BTreeMap::new())),
         }
     }
@@ -73,7 +73,7 @@ impl<L: Clone + 'static> LabeledBTreeGraph<L> {
             );
         }
         let is_new_arc = self.succ[u].insert(v, l).is_none();
-        self.number_of_arcs += is_new_arc as u64;
+        self.num_arcs += is_new_arc as u64;
         is_new_arc
     }
 
@@ -88,7 +88,7 @@ impl<L: Clone + 'static> LabeledBTreeGraph<L> {
             );
         }
         let arc_existed = self.succ[u].remove(&v).is_some();
-        self.number_of_arcs -= arc_existed as u64;
+        self.num_arcs -= arc_existed as u64;
         arc_existed
     }
 
@@ -198,7 +198,7 @@ impl<L: Clone + 'static> RandomAccessLabeling for LabeledBTreeGraph<L> {
         L: 'succ;
     #[inline(always)]
     fn num_arcs(&self) -> u64 {
-        self.number_of_arcs
+        self.num_arcs
     }
 
     #[inline(always)]
@@ -363,7 +363,7 @@ impl RandomAccessLabeling for BTreeGraph {
 
     #[inline(always)]
     fn num_arcs(&self) -> u64 {
-        self.0.number_of_arcs
+        self.0.num_arcs
     }
 
     #[inline(always)]

webgraph/src/graphs/csr_graph.rs

@@ -29,8 +29,6 @@ pub type CompressedCsrGraph = CsrGraph<EF, BitFieldVec>;
 pub type CompressedCsrSortedGraph = CsrSortedGraph<EF, BitFieldVec>;
 
 /// A compressed sparse-row graph.
-#[derive(Debug, Clone, Epserde)]
-#[cfg_attr(feature = "serde", derive(serde::Serialize, serde::Deserialize))]
 ///
 /// It is a graph representation that stores the degree cumulative function
 /// (DCF) and the successors in a compressed format. The DCF is a sequence of
@@ -51,6 +49,8 @@ pub type CompressedCsrSortedGraph = CsrSortedGraph<EF, BitFieldVec>;
 /// using a [`BitFieldVec`]. There is also a [version with sorted
 /// successors](CompressedCsrSortedGraph). Their construction requires a
 /// sequential graph providing the number of arcs.
+#[derive(Debug, Clone, Epserde)]
+#[cfg_attr(feature = "serde", derive(serde::Serialize, serde::Deserialize))]
 pub struct CsrGraph<DCF = Box<[usize]>, S = Box<[usize]>> {
     dcf: DCF,
     successors: S,
@@ -268,7 +268,7 @@ where
     DCF: SliceByValue + IterateByValueFrom<Item = usize>,
     S: SliceByValue + IterateByValueFrom<Item = usize>,
 {
-    type Lender = LenderImpl<IterFrom<'a, DCF>, IterFrom<'a, S>>;
+    type Lender = NodeLabels<IterFrom<'a, DCF>, IterFrom<'a, S>>;
 
     #[inline(always)]
     fn into_lender(self) -> Self::Lender {
@@ -299,7 +299,7 @@ where
 {
     type Label = usize;
     type Lender<'a>
-        = LenderImpl<IterFrom<'a, DCF>, IterFrom<'a, S>>
+        = NodeLabels<IterFrom<'a, DCF>, IterFrom<'a, S>>
     where
         Self: 'a;
 
@@ -320,7 +320,7 @@ where
         // because it might not exist
         let offset = offsets_iter.next().unwrap_or(0);
 
-        LenderImpl {
+        NodeLabels {
             node: from,
             last_offset: offset,
             current_offset: offset,
@@ -462,7 +462,7 @@ where
 
 /// Sequential Lender for the CSR graph.
 #[derive(Debug, Clone)]
-pub struct LenderImpl<O: Iterator<Item = usize>, S: Iterator<Item = usize>> {
+pub struct NodeLabels<O: Iterator<Item = usize>, S: Iterator<Item = usize>> {
     /// The next node to lend labels for.
     node: usize,
     /// This is the offset of the last successor of the previous node.
@@ -477,11 +477,11 @@ pub struct LenderImpl<O: Iterator<Item = usize>, S: Iterator<Item = usize>> {
 }
 
 unsafe impl<O: Iterator<Item = usize>, S: Iterator<Item = usize>> SortedLender
-    for LenderImpl<O, S>
+    for NodeLabels<O, S>
 {
 }
 
-impl<'succ, I, D> NodeLabelsLender<'succ> for LenderImpl<I, D>
+impl<'succ, I, D> NodeLabelsLender<'succ> for NodeLabels<I, D>
 where
     I: Iterator<Item = usize>,
     D: Iterator<Item = usize>,
@@ -490,15 +490,15 @@ where
     type IntoIterator = SeqSucc<'succ, D>;
 }
 
-impl<'succ, I, D> Lending<'succ> for LenderImpl<I, D>
+impl<'succ, I, D> Lending<'succ> for NodeLabels<I, D>
 where
     I: Iterator<Item = usize>,
     D: Iterator<Item = usize>,
 {
     type Lend = (usize, SeqSucc<'succ, D>);
 }
 
-impl<I, D> Lender for LenderImpl<I, D>
+impl<I, D> Lender for NodeLabels<I, D>
 where
     I: Iterator<Item = usize>,
     D: Iterator<Item = usize>,
@@ -535,7 +535,7 @@ where
 /// Sequential Lender for the CSR graph.
 #[derive(Debug, Clone)]
 #[repr(transparent)]
-pub struct LenderSortedImpl<O: Iterator<Item = usize>, S: Iterator<Item = usize>>(LenderImpl<O, S>);
+pub struct LenderSortedImpl<O: Iterator<Item = usize>, S: Iterator<Item = usize>>(NodeLabels<O, S>);
 
 unsafe impl<O: Iterator<Item = usize>, S: Iterator<Item = usize>> SortedLender
     for LenderSortedImpl<O, S>

webgraph/src/graphs/union_graph.rs

@@ -213,8 +213,6 @@ impl<I: Iterator<Item = usize>, J: Iterator<Item = usize>> std::iter::FusedItera
 
 #[cfg(test)]
 mod tests {
-    use core::panic;
-
     use super::*;
 
     #[test]

webgraph/src/labels/zip.rs

@@ -173,10 +173,10 @@ impl<G: RandomAccessGraph, L: RandomAccessLabeling> LabeledRandomAccessGraph<L::
 {
 }
 
-unsafe impl<L, R> SortedLender for NodeLabels<L, R>
-where
-    L: Lender + for<'next> NodeLabelsLender<'next>,
-    R: Lender + for<'next> NodeLabelsLender<'next>,
+unsafe impl<
+    L: SortedLender + for<'next> NodeLabelsLender<'next>,
+    R: SortedLender + for<'next> NodeLabelsLender<'next>,
+> SortedLender for NodeLabels<L, R>
 {
 }