19278972970

Committed 11 Nov 2025 09:25PM UTC coverage: 62.316% (+14.3%) from 48.052%

Build # 19278972970

Build Type

push

github

Committed by

zommiommy

Commit Message

BatchCodec print stats and encoding time

Run Details

40 of 53 new or added lines in 4 files covered. (75.47%)

814 existing lines in 46 files now uncovered.

5252 of 8428 relevant lines covered (62.32%)

29580067.92 hits per line

Source File
Press 'n' to go to next uncovered line, 'b' for previous

64.91

/webgraph/src/transform/simplify.rs

/*
 * SPDX-FileCopyrightText: 2023 Inria
 *
 * SPDX-License-Identifier: Apache-2.0 OR LGPL-2.1-or-later
 */

use crate::graphs::{
    arc_list_graph, no_selfloops_graph::NoSelfLoopsGraph, union_graph::UnionGraph,
};
use crate::labels::Left;
use crate::traits::{LenderIntoIter, SequentialGraph, SortedIterator, SortedLender, SplitLabeling};
use crate::utils::sort_pairs::{KMergeIters, SortPairs};
use crate::utils::{CodecIter, DefaultBatchCodec, MemoryUsage};
use anyhow::{Context, Result};
use dsi_progress_logger::prelude::*;
use itertools::Itertools;
use lender::*;
use rayon::ThreadPool;
use tempfile::Builder;

use super::transpose;

/// Returns a simplified (i.e., undirected and loopless) version of the provided
/// sorted (both on nodes and successors) graph as a [sequential
/// graph](crate::traits::SequentialGraph).
///
/// This method exploits the fact that the input graph is already sorted,
/// sorting half the number of arcs of
/// [`simplify`](crate::transform::simplify::simplify).
pub fn simplify_sorted<G: SequentialGraph>(
    graph: G,
    memory_usage: MemoryUsage,
) -> Result<
    NoSelfLoopsGraph<
        UnionGraph<
            G,
            Left<arc_list_graph::ArcListGraph<KMergeIters<CodecIter<DefaultBatchCodec>, ()>>>,
        >,
    >,
>
where
    for<'a> G::Lender<'a>: SortedLender,
    for<'a, 'b> LenderIntoIter<'a, G::Lender<'b>>: SortedIterator,
{
    let transpose = transpose(&graph, memory_usage).context("Could not transpose the graph")?;
    Ok(NoSelfLoopsGraph(UnionGraph(graph, transpose)))
}

/// Returns a simplified (i.e., undirected and loopless) version of the provided
/// graph as a [sequential graph](crate::traits::SequentialGraph).
///
/// Note that if the graph is sorted (both on nodes and successors), it is
/// recommended to use [`simplify_sorted`](crate::transform::simplify::simplify_sorted).
///
/// For the meaning of the additional parameter, see
/// [`SortPairs`](crate::prelude::sort_pairs::SortPairs).
pub fn simplify(
    graph: &impl SequentialGraph,
    memory_usage: MemoryUsage,
) -> Result<
    Left<
        arc_list_graph::ArcListGraph<
            impl Iterator<Item = ((usize, usize), ())> + Clone + Send + Sync + 'static,
        >,
    >,
> {
    let dir = Builder::new().prefix("simplify_").tempdir()?;
    let mut sorted = SortPairs::new(memory_usage, dir.path())?;

    let mut pl = ProgressLogger::default();
    pl.item_name("node")
        .expected_updates(Some(graph.num_nodes()));
    pl.start("Creating batches...");
    // create batches of sorted edges
    let mut iter = graph.iter();
    while let Some((src, succ)) = iter.next() {
        for dst in succ {
            if src != dst {
                sorted.push(src, dst)?;
                sorted.push(dst, src)?;
            }
        }
        pl.light_update();
    }
    // merge the batches
    let map: fn(((usize, usize), ())) -> (usize, usize) = |(pair, _)| pair;
    let filter: fn(&(usize, usize)) -> bool = |(src, dst)| src != dst;
    let iter = Itertools::dedup(sorted.iter()?.map(map).filter(filter));
    let sorted = arc_list_graph::ArcListGraph::new(graph.num_nodes(), iter);
    pl.done();

    Ok(sorted)
}

/// Returns a simplified (i.e., undirected and loopless) version of the provided
/// graph as a [sequential graph](crate::traits::SequentialGraph).
///
/// This method uses splitting to sort in parallel different parts of the graph.
///
/// For the meaning of the additional parameter, see
/// [`SortPairs`](crate::prelude::sort_pairs::SortPairs).
pub fn simplify_split<S>(
    graph: &S,
    memory_usage: MemoryUsage,
    threads: &ThreadPool,
) -> Result<
    Left<
        arc_list_graph::ArcListGraph<
            itertools::Dedup<KMergeIters<CodecIter<DefaultBatchCodec>, ()>>,
        >,
    >,
>
where
    S: SequentialGraph + SplitLabeling,
{
    let num_threads = threads.current_num_threads();
    let (tx, rx) = std::sync::mpsc::channel();

    let mut dirs = vec![];

    threads.in_place_scope(|scope| {
        let mut thread_id = 0;
        #[allow(clippy::explicit_counter_loop)] // enumerate requires some extra bounds here
        for iter in graph.split_iter(num_threads) {
            let tx = tx.clone();
            let dir = Builder::new()
                .prefix(&format!("simplify_split_{thread_id}_"))
                .tempdir()
                .expect("Could not create a temporary directory");
            let dir_path = dir.path().to_path_buf();
            dirs.push(dir);
            scope.spawn(move |_| {
                log::debug!("Spawned thread {thread_id}");
                let mut sorted = SortPairs::new(memory_usage, dir_path).unwrap();
                for_!( (src, succ) in iter {
                    for dst in succ {
                        if src != dst {
                            sorted.push(src, dst).unwrap();
                            sorted.push(dst, src).unwrap();
                        }
                    }
                });
                let result = sorted.iter().context("Could not read arcs").unwrap();
                tx.send(result).expect("Could not send the sorted pairs");
                log::debug!("Thread {thread_id} finished");
            });
            thread_id += 1;
        }
    });
    drop(tx);

    // get a graph on the sorted data
    log::debug!("Waiting for threads to finish");
    let edges: KMergeIters<CodecIter<DefaultBatchCodec>> = rx.iter().sum();
    let edges = edges.dedup();
    log::debug!("All threads finished");
    let sorted = arc_list_graph::ArcListGraph::new_labeled(graph.num_nodes(), edges);

    drop(dirs);
    Ok(Left(sorted))
}

1	/*
2	* SPDX-FileCopyrightText: 2023 Inria
3	*
4	* SPDX-License-Identifier: Apache-2.0 OR LGPL-2.1-or-later
5	*/
6
7	use crate::graphs::{
8	arc_list_graph, no_selfloops_graph::NoSelfLoopsGraph, union_graph::UnionGraph,
9	};
10	use crate::labels::Left;
11	use crate::traits::{LenderIntoIter, SequentialGraph, SortedIterator, SortedLender, SplitLabeling};
12	use crate::utils::sort_pairs::{KMergeIters, SortPairs};
13	use crate::utils::{CodecIter, DefaultBatchCodec, MemoryUsage};
14	use anyhow::{Context, Result};
15	use dsi_progress_logger::prelude::*;
16	use itertools::Itertools;
17	use lender::*;
18	use rayon::ThreadPool;
19	use tempfile::Builder;
20
21	use super::transpose;
22
23	/// Returns a simplified (i.e., undirected and loopless) version of the provided
24	/// sorted (both on nodes and successors) graph as a [sequential
25	/// graph](crate::traits::SequentialGraph).
26	///
27	/// This method exploits the fact that the input graph is already sorted,
28	/// sorting half the number of arcs of
29	/// [`simplify`](crate::transform::simplify::simplify).
30	pub fn simplify_sorted<G: SequentialGraph>(	1✔
31	graph: G,
32	memory_usage: MemoryUsage,
33	) -> Result<
34	NoSelfLoopsGraph<
35	UnionGraph<
36	G,
37	Left<arc_list_graph::ArcListGraph<KMergeIters<CodecIter<DefaultBatchCodec>, ()>>>,
38	>,
39	>,
40	>
41	where
42	for<'a> G::Lender<'a>: SortedLender,
43	for<'a, 'b> LenderIntoIter<'a, G::Lender<'b>>: SortedIterator,
44	{
45	let transpose = transpose(&graph, memory_usage).context("Could not transpose the graph")?;	5✔
46	Ok(NoSelfLoopsGraph(UnionGraph(graph, transpose)))	1✔
47	}
48
49	/// Returns a simplified (i.e., undirected and loopless) version of the provided
50	/// graph as a [sequential graph](crate::traits::SequentialGraph).
51	///
52	/// Note that if the graph is sorted (both on nodes and successors), it is
53	/// recommended to use [`simplify_sorted`](crate::transform::simplify::simplify_sorted).
54	///
55	/// For the meaning of the additional parameter, see
56	/// [`SortPairs`](crate::prelude::sort_pairs::SortPairs).
UNCOV 57	pub fn simplify(	×
58	graph: &impl SequentialGraph,
59	memory_usage: MemoryUsage,
60	) -> Result<
61	Left<
62	arc_list_graph::ArcListGraph<
63	impl Iterator<Item = ((usize, usize), ())> + Clone + Send + Sync + 'static,
64	>,
65	>,
66	> {
UNCOV 67	let dir = Builder::new().prefix("simplify_").tempdir()?;	×
UNCOV 68	let mut sorted = SortPairs::new(memory_usage, dir.path())?;	×
69
70	let mut pl = ProgressLogger::default();	×
UNCOV 71	pl.item_name("node")	×
72	.expected_updates(Some(graph.num_nodes()));	×
73	pl.start("Creating batches...");	×
74	// create batches of sorted edges
75	let mut iter = graph.iter();	×
UNCOV 76	while let Some((src, succ)) = iter.next() {	×
77	for dst in succ {	×
78	if src != dst {	×
79	sorted.push(src, dst)?;	×
80	sorted.push(dst, src)?;	×
81	}
82	}
UNCOV 83	pl.light_update();	×
84	}
85	// merge the batches
UNCOV 86	let map: fn(((usize, usize), ())) -> (usize, usize) = \|(pair, _)\| pair;	×
UNCOV 87	let filter: fn(&(usize, usize)) -> bool = \|(src, dst)\| src != dst;	×
88	let iter = Itertools::dedup(sorted.iter()?.map(map).filter(filter));	×
89	let sorted = arc_list_graph::ArcListGraph::new(graph.num_nodes(), iter);	×
90	pl.done();	×
91
92	Ok(sorted)	×
93	}
94
95	/// Returns a simplified (i.e., undirected and loopless) version of the provided
96	/// graph as a [sequential graph](crate::traits::SequentialGraph).
97	///
98	/// This method uses splitting to sort in parallel different parts of the graph.
99	///
100	/// For the meaning of the additional parameter, see
101	/// [`SortPairs`](crate::prelude::sort_pairs::SortPairs).
102	pub fn simplify_split<S>(	4✔
103	graph: &S,
104	memory_usage: MemoryUsage,
105	threads: &ThreadPool,
106	) -> Result<
107	Left<
108	arc_list_graph::ArcListGraph<
109	itertools::Dedup<KMergeIters<CodecIter<DefaultBatchCodec>, ()>>,
110	>,
111	>,
112	>
113	where
114	S: SequentialGraph + SplitLabeling,
115	{
116	let num_threads = threads.current_num_threads();	12✔
117	let (tx, rx) = std::sync::mpsc::channel();	12✔
118
119	let mut dirs = vec![];	8✔
120
121	threads.in_place_scope(\|scope\| {	12✔
122	let mut thread_id = 0;	8✔
123	#[allow(clippy::explicit_counter_loop)] // enumerate requires some extra bounds here
124	for iter in graph.split_iter(num_threads) {	28✔
125	let tx = tx.clone();	48✔
126	let dir = Builder::new()	48✔
127	.prefix(&format!("simplify_split_{thread_id}_"))	32✔
128	.tempdir()	16✔
129	.expect("Could not create a temporary directory");	32✔
130	let dir_path = dir.path().to_path_buf();	48✔
131	dirs.push(dir);	48✔
132	scope.spawn(move \|_\| {	48✔
133	log::debug!("Spawned thread {thread_id}");	16✔
134	let mut sorted = SortPairs::new(memory_usage, dir_path).unwrap();	80✔
135	for_!( (src, succ) in iter {	2,604,472✔
136	for dst in succ {	27,031,444✔
137	if src != dst {	25,379,448✔
138	sorted.push(src, dst).unwrap();	75,089,040✔
139	sorted.push(dst, src).unwrap();	50,059,360✔
140	}
141	}
142	});
143	let result = sorted.iter().context("Could not read arcs").unwrap();	80✔
144	tx.send(result).expect("Could not send the sorted pairs");	80✔
145	log::debug!("Thread {thread_id} finished");	16✔
146	});
147	thread_id += 1;	16✔
148	}
149	});
150	drop(tx);	8✔
151
152	// get a graph on the sorted data
153	log::debug!("Waiting for threads to finish");	4✔
154	let edges: KMergeIters<CodecIter<DefaultBatchCodec>> = rx.iter().sum();	20✔
155	let edges = edges.dedup();	12✔
156	log::debug!("All threads finished");	4✔
157	let sorted = arc_list_graph::ArcListGraph::new_labeled(graph.num_nodes(), edges);	20✔
158
159	drop(dirs);	8✔
160	Ok(Left(sorted))	4✔
161	}

vigna / webgraph-rs / 19278972970

Source File Press 'n' to go to next uncovered line, 'b' for previous

Source File
Press 'n' to go to next uncovered line, 'b' for previous