11837625908

Committed 12 Nov 2024 07:33PM UTC coverage: 53.594% (-0.04%) from 53.631%

Build # 11837625908

Build Type

push

github

Committed by

vigna

Commit Message

No more Borrow

Run Details

5 of 13 new or added lines in 6 files covered. (38.46%)

2 existing lines in 2 files now uncovered.

2371 of 4424 relevant lines covered (53.59%)

23193450.13 hits per line

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

13.74

/src/cli/transform/simplify.rs

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

use crate::cli::*;
use crate::graphs::union_graph::UnionGraph;
use crate::prelude::*;
use anyhow::Result;
use clap::{ArgMatches, Args, Command, FromArgMatches};
use dsi_bitstream::prelude::*;
use mmap_rs::MmapFlags;
use std::path::PathBuf;
use tempfile::Builder;

pub const COMMAND_NAME: &str = "simplify";

#[derive(Args, Debug)]
#[command(about = "Makes a BvGraph simple (undirected and loopless) by adding missing arcs and removing loops, optionally applying a permutation.", long_about = None)]
pub struct CliArgs {
    /// The basename of the graph.
    pub src: PathBuf,
    /// The basename of the simplified graph.
    pub dst: PathBuf,

    #[arg(long)]
    /// The basename of a pre-computed transposed version of the source graph, which
    /// will be use to speed up the simplification.
    pub transposed: Option<PathBuf>,

    #[clap(flatten)]
    pub num_threads: NumThreadsArg,

    #[clap(flatten)]
    pub batch_size: BatchSizeArg,

    #[clap(flatten)]
    pub ca: CompressArgs,

    #[arg(long)]
    /// The path to an optional permutation in binary big-endian format to apply to the graph.
    pub permutation: Option<PathBuf>,
}

pub fn cli(command: Command) -> Command {
    command.subcommand(CliArgs::augment_args(Command::new(COMMAND_NAME)).display_order(0))
}

pub fn main(submatches: &ArgMatches) -> Result<()> {
    let args = CliArgs::from_arg_matches(submatches)?;

    create_parent_dir(&args.dst)?;

    match get_endianness(&args.src)?.as_str() {
        #[cfg(any(
            feature = "be_bins",
            not(any(feature = "be_bins", feature = "le_bins"))
        ))]
        BE::NAME => simplify::<BE>(args),
        #[cfg(any(
            feature = "le_bins",
            not(any(feature = "be_bins", feature = "le_bins"))
        ))]
        LE::NAME => simplify::<LE>(args),
        e => panic!("Unknown endianness: {}", e),
    }
}

fn no_ef_warn(basepath: impl AsRef<std::path::Path>) {
    log::warn!("The .ef file was not found so the simplification will proceed sequentially. This may be slow. To speed it up, you can use `webgraph build ef {}` which would allow us create batches in parallel", basepath.as_ref().display());
}

pub fn simplify<E: Endianness + Send + Sync + 'static>(args: CliArgs) -> Result<()>
where
    for<'a> BufBitReader<E, MemWordReader<u32, &'a [u32]>>: CodeRead<E> + BitSeek,
{
    // TODO!: speed it up by using random access graph if possible
    let thread_pool = crate::cli::get_thread_pool(args.num_threads.num_threads);

    let target_endianness = args.ca.endianness.clone().unwrap_or_else(|| E::NAME.into());

    let dir = Builder::new().prefix("transform_simplify_").tempdir()?;

    match (args.permutation, args.transposed) {
        // load the transposed graph and use it to directly compress the graph
        // without doing any sorting
        (None, Some(t_path)) => {
            log::info!("Transposed graph provided, using it to simplify the graph");

            let has_ef_graph =
                std::fs::metadata(args.src.with_extension(".ef")).is_ok_and(|x| x.is_file());
            let has_ef_t_graph =
                std::fs::metadata(t_path.with_extension(".ef")).is_ok_and(|x| x.is_file());

            match (has_ef_graph, has_ef_t_graph) {
                (true, true) => {
                    log::info!("Both .ef files found, using simplify split");

                    let graph =
                        crate::graphs::bvgraph::random_access::BvGraph::with_basename(&args.src)
                            .endianness::<E>()
                            .load()?;
                    let num_nodes = graph.num_nodes();
                    let graph_t =
                        crate::graphs::bvgraph::random_access::BvGraph::with_basename(&t_path)
                            .endianness::<E>()
                            .load()?;

                    if graph_t.num_nodes() != num_nodes {
                        anyhow::bail!("The number of nodes in the graph and its transpose do not match! {} != {}", num_nodes, graph_t.num_nodes());
                    }

                    let sorted = NoSelfLoopsGraph(UnionGraph(graph, graph_t));

                    BvComp::parallel_endianness(
                        &args.dst,
                        &sorted,
                        num_nodes,
                        args.ca.into(),
                        &thread_pool,
                        dir,
                        &target_endianness,
                    )?;

                    return Ok(());
                }
                (true, false) => {
                    no_ef_warn(&args.src);
                }
                (false, true) => {
                    no_ef_warn(&t_path);
                }
                (false, false) => {
                    no_ef_warn(&args.src);
                    no_ef_warn(&t_path);
                }
            }

            let seq_graph =
                crate::graphs::bvgraph::sequential::BvGraphSeq::with_basename(&args.src)
                    .endianness::<E>()
                    .load()?;
            let num_nodes = seq_graph.num_nodes();
            let seq_graph_t =
                crate::graphs::bvgraph::sequential::BvGraphSeq::with_basename(&t_path)
                    .endianness::<E>()
                    .load()?;

            if seq_graph_t.num_nodes() != num_nodes {
                anyhow::bail!(
                    "The number of nodes in the graph and its transpose do not match! {} != {}",
                    num_nodes,
                    seq_graph_t.num_nodes()
                );
            }

            let sorted = NoSelfLoopsGraph(UnionGraph(seq_graph, seq_graph_t));

            BvComp::parallel_endianness(
                &args.dst,
                &sorted,
                num_nodes,
                args.ca.into(),
                &thread_pool,
                dir,
                &target_endianness,
            )?;
        }
        // apply the permutation, don't care if the transposed graph is already computed
        // as we cannot really exploit it
        (Some(perm_path), None | Some(_)) => {
            log::info!("Permutation provided, applying it to the graph");

            let perm = JavaPermutation::mmap(perm_path, MmapFlags::RANDOM_ACCESS)?;

            // if the .ef file exists, we can use the simplify split
            if std::fs::metadata(args.src.with_extension(".ef")).is_ok_and(|x| x.is_file()) {
                log::info!(".ef file found, using simplify split");
                let graph =
                    crate::graphs::bvgraph::random_access::BvGraph::with_basename(&args.src)
                        .endianness::<E>()
                        .load()?;

                let perm_graph = PermutedGraph {
                    graph: &graph,
                    perm: &perm,
                };

                let sorted = crate::transform::simplify_split(
                    &perm_graph,
                    args.batch_size.batch_size,
                    &thread_pool,
                )?;

                BvComp::parallel_endianness(
                    &args.dst,
                    &sorted,
                    graph.num_nodes(),
                    args.ca.into(),
                    &thread_pool,
                    dir,
                    &target_endianness,
                )?;

                return Ok(());
            }

            no_ef_warn(&args.src);

            let seq_graph =
                crate::graphs::bvgraph::sequential::BvGraphSeq::with_basename(&args.src)
                    .endianness::<E>()
                    .load()?;

            let perm_graph = PermutedGraph {
                graph: &seq_graph,
                perm: &perm,
            };

            // simplify the graph
            let sorted =
                crate::transform::simplify(&perm_graph, args.batch_size.batch_size).unwrap();

            BvComp::parallel_endianness(
                &args.dst,
                &sorted,
                sorted.num_nodes(),
                args.ca.into(),
                &thread_pool,
                dir,
                &target_endianness,
            )?;
        }
        // just compute the transpose on the fly
        (None, None) => {
            log::info!(
                "No permutation or transposed graph provided, computing the transpose on the fly"
            );
            // if the .ef file exists, we can use the simplify split
            if std::fs::metadata(args.src.with_extension(".ef")).is_ok_and(|x| x.is_file()) {
                log::info!(".ef file found, using simplify split");

                let graph =
                    crate::graphs::bvgraph::random_access::BvGraph::with_basename(&args.src)
                        .endianness::<E>()
                        .load()?;

                let sorted = crate::transform::simplify_split(
                    &graph,
                    args.batch_size.batch_size,
                    &thread_pool,
                )?;

                BvComp::parallel_endianness(
                    &args.dst,
                    &sorted,
                    graph.num_nodes(),
                    args.ca.into(),
                    &thread_pool,
                    dir,
                    &target_endianness,
                )?;

                return Ok(());
            }

            no_ef_warn(&args.src);

            let seq_graph =
                crate::graphs::bvgraph::sequential::BvGraphSeq::with_basename(&args.src)
                    .endianness::<E>()
                    .load()?;

            // transpose the graph
            let sorted =
                crate::transform::simplify(&seq_graph, args.batch_size.batch_size).unwrap();

            BvComp::parallel_endianness(
                &args.dst,
                &sorted,
                seq_graph.num_nodes(),
                args.ca.into(),
                &thread_pool,
                dir,
                &target_endianness,
            )?;
        }
    }

    Ok(())
}

1	/*
2	* SPDX-FileCopyrightText: 2023 Inria
3	* SPDX-FileCopyrightText: 2023 Tommaso Fontana
4	*
5	* SPDX-License-Identifier: Apache-2.0 OR LGPL-2.1-or-later
6	*/
7
8	use crate::cli::*;
9	use crate::graphs::union_graph::UnionGraph;
10	use crate::prelude::*;
11	use anyhow::Result;
12	use clap::{ArgMatches, Args, Command, FromArgMatches};
13	use dsi_bitstream::prelude::*;
14	use mmap_rs::MmapFlags;
15	use std::path::PathBuf;
16	use tempfile::Builder;
17
18	pub const COMMAND_NAME: &str = "simplify";
19
20	#[derive(Args, Debug)]
21	#[command(about = "Makes a BvGraph simple (undirected and loopless) by adding missing arcs and removing loops, optionally applying a permutation.", long_about = None)]
22	pub struct CliArgs {
23	/// The basename of the graph.
24	pub src: PathBuf,
25	/// The basename of the simplified graph.
26	pub dst: PathBuf,
27
28	#[arg(long)]
29	/// The basename of a pre-computed transposed version of the source graph, which
30	/// will be use to speed up the simplification.
31	pub transposed: Option<PathBuf>,
32
33	#[clap(flatten)]
34	pub num_threads: NumThreadsArg,
35
36	#[clap(flatten)]
37	pub batch_size: BatchSizeArg,
38
39	#[clap(flatten)]
40	pub ca: CompressArgs,
41
42	#[arg(long)]
43	/// The path to an optional permutation in binary big-endian format to apply to the graph.
44	pub permutation: Option<PathBuf>,
45	}
46
47	pub fn cli(command: Command) -> Command {	18✔
48	command.subcommand(CliArgs::augment_args(Command::new(COMMAND_NAME)).display_order(0))	18✔
49	}
50
51	pub fn main(submatches: &ArgMatches) -> Result<()> {	2✔
52	let args = CliArgs::from_arg_matches(submatches)?;	4✔
53
54	create_parent_dir(&args.dst)?;	×
55
56	match get_endianness(&args.src)?.as_str() {	4✔
57	#[cfg(any(
58	feature = "be_bins",
59	not(any(feature = "be_bins", feature = "le_bins"))
60	))]
61	BE::NAME => simplify::<BE>(args),	4✔
62	#[cfg(any(
63	feature = "le_bins",
64	not(any(feature = "be_bins", feature = "le_bins"))
65	))]
66	LE::NAME => simplify::<LE>(args),	×
67	e => panic!("Unknown endianness: {}", e),	×
68	}
69	}
70
71	fn no_ef_warn(basepath: impl AsRef<std::path::Path>) {	2✔
72	log::warn!("The .ef file was not found so the simplification will proceed sequentially. This may be slow. To speed it up, you can use `webgraph build ef {}` which would allow us create batches in parallel", basepath.as_ref().display());	4✔
73	}
74
75	pub fn simplify<E: Endianness + Send + Sync + 'static>(args: CliArgs) -> Result<()>	2✔
76	where
77	for<'a> BufBitReader<E, MemWordReader<u32, &'a [u32]>>: CodeRead<E> + BitSeek,
78	{
79	// TODO!: speed it up by using random access graph if possible
80	let thread_pool = crate::cli::get_thread_pool(args.num_threads.num_threads);	2✔
81
82	let target_endianness = args.ca.endianness.clone().unwrap_or_else(\|\| E::NAME.into());	6✔
83
84	let dir = Builder::new().prefix("transform_simplify_").tempdir()?;	4✔
85
86	match (args.permutation, args.transposed) {	×
87	// load the transposed graph and use it to directly compress the graph
88	// without doing any sorting
89	(None, Some(t_path)) => {	×
90	log::info!("Transposed graph provided, using it to simplify the graph");	×
91
92	let has_ef_graph =	×
93	std::fs::metadata(args.src.with_extension(".ef")).is_ok_and(\|x\| x.is_file());	×
94	let has_ef_t_graph =	×
95	std::fs::metadata(t_path.with_extension(".ef")).is_ok_and(\|x\| x.is_file());	×
96
97	match (has_ef_graph, has_ef_t_graph) {	×
98	(true, true) => {	×
99	log::info!("Both .ef files found, using simplify split");	×
100
101	let graph =	×
102	crate::graphs::bvgraph::random_access::BvGraph::with_basename(&args.src)	×
103	.endianness::<E>()
104	.load()?;
105	let num_nodes = graph.num_nodes();	×
106	let graph_t =	×
107	crate::graphs::bvgraph::random_access::BvGraph::with_basename(&t_path)	×
108	.endianness::<E>()
109	.load()?;
110
111	if graph_t.num_nodes() != num_nodes {	×
112	anyhow::bail!("The number of nodes in the graph and its transpose do not match! {} != {}", num_nodes, graph_t.num_nodes());	×
113	}
114
115	let sorted = NoSelfLoopsGraph(UnionGraph(graph, graph_t));	×
116
117	BvComp::parallel_endianness(
118	&args.dst,	×
119	&sorted,	×
120	num_nodes,	×
121	args.ca.into(),	×
122	&thread_pool,	×
123	dir,	×
124	&target_endianness,	×
125	)?;
126
127	return Ok(());	×
128	}
129	(true, false) => {	×
130	no_ef_warn(&args.src);	×
131	}
132	(false, true) => {	×
133	no_ef_warn(&t_path);	×
134	}
135	(false, false) => {	×
136	no_ef_warn(&args.src);	×
137	no_ef_warn(&t_path);	×
138	}
139	}
140
141	let seq_graph =	×
142	crate::graphs::bvgraph::sequential::BvGraphSeq::with_basename(&args.src)	×
143	.endianness::<E>()
144	.load()?;
145	let num_nodes = seq_graph.num_nodes();	×
146	let seq_graph_t =	×
147	crate::graphs::bvgraph::sequential::BvGraphSeq::with_basename(&t_path)	×
148	.endianness::<E>()
149	.load()?;
150
151	if seq_graph_t.num_nodes() != num_nodes {	×
152	anyhow::bail!(	×
153	"The number of nodes in the graph and its transpose do not match! {} != {}",	×
154	num_nodes,	×
155	seq_graph_t.num_nodes()	×
156	);
157	}
158
159	let sorted = NoSelfLoopsGraph(UnionGraph(seq_graph, seq_graph_t));	×
160
161	BvComp::parallel_endianness(
162	&args.dst,	×
163	&sorted,	×
164	num_nodes,	×
165	args.ca.into(),	×
166	&thread_pool,	×
167	dir,	×
168	&target_endianness,	×
169	)?;
170	}
171	// apply the permutation, don't care if the transposed graph is already computed
172	// as we cannot really exploit it
173	(Some(perm_path), None \| Some(_)) => {	2✔
174	log::info!("Permutation provided, applying it to the graph");	4✔
175
176	let perm = JavaPermutation::mmap(perm_path, MmapFlags::RANDOM_ACCESS)?;	4✔
177
178	// if the .ef file exists, we can use the simplify split
179	if std::fs::metadata(args.src.with_extension(".ef")).is_ok_and(\|x\| x.is_file()) {	×
180	log::info!(".ef file found, using simplify split");	×
181	let graph =	×
182	crate::graphs::bvgraph::random_access::BvGraph::with_basename(&args.src)	×
183	.endianness::<E>()
184	.load()?;
185
186	let perm_graph = PermutedGraph {
187	graph: &graph,	×
188	perm: &perm,	×
189	};
190
191	let sorted = crate::transform::simplify_split(
192	&perm_graph,	×
193	args.batch_size.batch_size,	×
194	&thread_pool,	×
195	)?;
196
197	BvComp::parallel_endianness(
198	&args.dst,	×
199	&sorted,	×
200	graph.num_nodes(),	×
201	args.ca.into(),	×
NEW 202	&thread_pool,	×
203	dir,	×
204	&target_endianness,	×
205	)?;
206
207	return Ok(());	×
208	}
209
210	no_ef_warn(&args.src);	2✔
211
212	let seq_graph =	2✔
213	crate::graphs::bvgraph::sequential::BvGraphSeq::with_basename(&args.src)	×
214	.endianness::<E>()
215	.load()?;
216
217	let perm_graph = PermutedGraph {
218	graph: &seq_graph,	×
219	perm: &perm,	×
220	};
221
222	// simplify the graph
223	let sorted =	×
224	crate::transform::simplify(&perm_graph, args.batch_size.batch_size).unwrap();	×
225
226	BvComp::parallel_endianness(
227	&args.dst,	×
228	&sorted,	×
229	sorted.num_nodes(),	×
230	args.ca.into(),	×
231	&thread_pool,	×
232	dir,	×
233	&target_endianness,	×
234	)?;
235	}
236	// just compute the transpose on the fly
237	(None, None) => {	×
238	log::info!(	×
239	"No permutation or transposed graph provided, computing the transpose on the fly"	×
240	);
241	// if the .ef file exists, we can use the simplify split
242	if std::fs::metadata(args.src.with_extension(".ef")).is_ok_and(\|x\| x.is_file()) {	×
243	log::info!(".ef file found, using simplify split");	×
244
245	let graph =	×
246	crate::graphs::bvgraph::random_access::BvGraph::with_basename(&args.src)	×
247	.endianness::<E>()
248	.load()?;
249
250	let sorted = crate::transform::simplify_split(
251	&graph,	×
252	args.batch_size.batch_size,	×
253	&thread_pool,	×
254	)?;
255
256	BvComp::parallel_endianness(
257	&args.dst,	×
258	&sorted,	×
259	graph.num_nodes(),	×
260	args.ca.into(),	×
NEW 261	&thread_pool,	×
262	dir,	×
263	&target_endianness,	×
264	)?;
265
266	return Ok(());	×
267	}
268
269	no_ef_warn(&args.src);	×
270
271	let seq_graph =	×
272	crate::graphs::bvgraph::sequential::BvGraphSeq::with_basename(&args.src)	×
273	.endianness::<E>()
274	.load()?;
275
276	// transpose the graph
277	let sorted =	×
278	crate::transform::simplify(&seq_graph, args.batch_size.batch_size).unwrap();	×
279
280	BvComp::parallel_endianness(
281	&args.dst,	×
282	&sorted,	×
283	seq_graph.num_nodes(),	×
284	args.ca.into(),	×
NEW 285	&thread_pool,	×
286	dir,	×
287	&target_endianness,	×
288	)?;
289	}
290	}
291
292	Ok(())	2✔
293	}

vigna / webgraph-rs / 11837625908

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