Bump version

Use a generic param to specialize on the truncate arg to avoid needing
to check it and/or host bits if we'll end up accepting it regardless
when truncate is enabled.

Refactor host bits check into iputils.

.github/workflows/release.yml

@@ -16,7 +16,7 @@ jobs:
           - target: x86_64-pc-windows-gnu
             archive: zip
           - target: x86_64-unknown-linux-musl
-            archive: tar.gz tar.xz tar.zst
+            archive: tar.gz
           - target: x86_64-apple-darwin
             archive: zip
     steps:

Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "rs-aggregate"
-version = "0.3.0"
+version = "0.3.1"
 authors = ["Keenan Tims <ktims@gotroot.ca>"]
 edition = "2021"
 description = "Aggregate a list of IP prefixes into their minimum equivalent representation"
@@ -8,11 +8,7 @@ readme = "README.md"
 repository = "https://github.com/ktims/rs-aggregate"
 license = "MIT"
 categories = ["network-programming"]
-exclude = [
+exclude = [".github/*", "doc/*", "test-data/*"]
-	".github/*",
-	"doc/*",
-	"test-data/*",
-]
 
 [dependencies]
 clap = { version = "4.4.6", features = ["derive"] }
@@ -25,6 +21,15 @@ assert_fs = "1.0.12"
 predicates = "3.0.1"
 rstest = "0.16.0"
 glob = "0.3.1"
+tempfile = "3.8.1"
+json = "0.12.4"
+plotters = "0.3.5"
+rand_chacha = "0.3.1"
+rand = "0.8.5"
 
 [[bin]]
 name = "rs-aggregate"
+
+[[bench]]
+name = "perf"
+harness = false

README.md

@@ -5,20 +5,44 @@ Intended to be a drop-in replacement for [aggregate6](https://github.com/job/agg
 
 Takes a list of whitespace-separated IPs or IP networks and aggregates them to their minimal representation.
 
-## Known discrepancies with `aggregate6`
+## Installation
 
-* `rs-aggregate` accepts subnet and wilcard mask formats in addition to CIDR, ie all these are valid and equivalent:
+`rs-aggregate` is built statically. CI-built binaries can be found in the GitHub
-  * `1.1.1.0/255.255.255.0`
+releases for most common platforms. Simply download the appropriate binary and
-  * `1.1.1.0/0.0.0.255`
+place it in your path.
-  * `1.1.1.0/24`
+
-* `-m/--max-prefixlen` supports different maximums for each address family as ipv4,ipv6 format
+It can also be installed via some software management tools:
+
+### FreeBSD
+```
+pkg install rs-aggregate
+```
+
+### Cargo
+```
+cargo install rs-aggregate
+```
+
+
+## Known differences from `aggregate6`
+
+* `-m/--max-prefixlen` supports different maximums for each address family as
+  ipv4,ipv6 format. A single value is also supported and has the same behaviour
+  as `aggregate6` (apply the same maximum to both address families).
+* `-v` verbose dump is not supported
+* Truncation errors (when host bits are set without the `-t` flag) are printed
+  based on the parsed address, ie. always in CIDR format, whereas `aggregate6`
+  prints errors based on the input.
 
 ## Performance
 
 Performance comparison of `rs-aggregate` vs `aggregate6`. A speedup of >100x is achieved on DFZ data.
 
-Full DFZ (1154968 total, 202729 aggregates):
+### Full DFZ (1154968 total, 202729 aggregates):
 ![dfz perf comparison](doc/perfcomp_all.png)
 
-IPv4 DFZ (968520 total, 154061 aggregates):
+### IPv4 DFZ (968520 total, 154061 aggregates):
 ![ipv4 dfz perf comparison](doc/perfcomp_v4.png)
+
+### 1024 random prefixes (startup time):
+![startup time comparison](doc/perfcomp_startup.png)

benches/perf.rs

@@ -0,0 +1,282 @@
+use ipnet::Ipv4Net;
+use json::JsonValue;
+use plotters::backend::BitMapBackend;
+use plotters::chart::ChartBuilder;
+use plotters::coord::ranged1d::{IntoSegmentedCoord, SegmentValue};
+use plotters::drawing::IntoDrawingArea;
+use plotters::element::{EmptyElement, Text};
+use plotters::series::{Histogram, PointSeries};
+use plotters::style::full_palette::GREY;
+use plotters::style::text_anchor::{HPos, Pos, VPos};
+use plotters::style::{Color, IntoFont, RGBColor, ShapeStyle, BLACK, WHITE};
+use rand::prelude::*;
+use rand_chacha::ChaChaRng;
+use std::ffi::OsStr;
+use std::io::{Read, Write};
+
+use std::process::Stdio;
+use tempfile::NamedTempFile;
+
+const BAR_COLOUR: RGBColor = RGBColor(66, 133, 244);
+
+#[derive(Clone, Debug)]
+struct TestDefinition {
+    cmd: String,
+    name: String, // including version
+}
+
+#[derive(Clone, Debug)]
+struct TestResult {
+    mean: f64,
+    stddev: f64,
+    median: f64,
+    min: f64,
+    max: f64,
+}
+
+impl From<JsonValue> for TestResult {
+    fn from(value: JsonValue) -> Self {
+        Self {
+            mean: value["mean"].as_f64().unwrap(),
+            stddev: value["stddev"].as_f64().unwrap(),
+            median: value["median"].as_f64().unwrap(),
+            min: value["min"].as_f64().unwrap(),
+            max: value["max"].as_f64().unwrap(),
+        }
+    }
+}
+
+fn make_tests(input_path: &str) -> Vec<TestDefinition> {
+    let our_version = format!("rs-aggregate {}", env!("CARGO_PKG_VERSION"));
+    let our_path = env!("CARGO_BIN_EXE_rs-aggregate");
+
+    let python_version_raw = std::process::Command::new("python3")
+        .arg("--version")
+        .stdout(Stdio::piped())
+        .spawn()
+        .expect("Unable to run python3")
+        .wait_with_output()
+        .expect("Couldn't get python3 output")
+        .stdout;
+    let python_version = String::from_utf8_lossy(&python_version_raw);
+
+    let agg6_version_raw = std::process::Command::new("python3")
+        .arg("-m")
+        .arg("aggregate6")
+        .arg("-V")
+        .stdout(Stdio::piped())
+        .spawn()
+        .expect("Unable to run aggregate6")
+        .wait_with_output()
+        .expect("Couldn't get aggregate6 output")
+        .stdout;
+    let agg6_version = String::from_utf8_lossy(&agg6_version_raw);
+
+    vec![
+        TestDefinition {
+            cmd: format!("{} {}", our_path, input_path),
+            name: our_version.into(),
+        },
+        TestDefinition {
+            cmd: format!("python3 -m aggregate6 {}", input_path),
+            name: format!("{} ({})", agg6_version.trim(), python_version.trim()),
+        },
+    ]
+}
+
+fn make_v4_tests(input_path: &str) -> Vec<TestDefinition> {
+    let mut all_tests = make_tests(input_path);
+
+    let iprange_version_raw = std::process::Command::new("iprange")
+        .arg("--version")
+        .stdout(Stdio::piped())
+        .spawn()
+        .expect("Unable to run iprange")
+        .wait_with_output()
+        .expect("Couldn't get iprange output")
+        .stdout;
+    let iprange_version = String::from_utf8_lossy(&iprange_version_raw);
+
+    all_tests.push(TestDefinition {
+        cmd: format!("iprange --optimize {}", input_path),
+        name: iprange_version.lines().nth(0).unwrap().into(),
+    });
+
+    all_tests
+}
+
+// We don't really care if aggregation will actually be possible, but we'll only
+// generate prefixes with length 8->24 so some should be possible.
+fn make_random_prefix(rng: &mut impl Rng) -> Ipv4Net {
+    let prefix_len: u8 = rng.gen_range(8..25);
+    let netaddr: u32 = rng.gen_range(0..(1 << prefix_len)) << 32 - prefix_len;
+
+    Ipv4Net::new(netaddr.into(), prefix_len).unwrap()
+}
+
+// Generate 1024 random v4 addresses as a startup time test
+fn make_startup_tests() -> (NamedTempFile, Vec<TestDefinition>) {
+    let mut rng = ChaChaRng::seed_from_u64(0); // use a repeatable rng with custom seed
+    let addresses = std::iter::repeat_with(|| make_random_prefix(&mut rng)).take(1024);
+
+    let mut outfile = NamedTempFile::new().unwrap();
+    let mut outfile_f = outfile.as_file();
+    for addr in addresses {
+        outfile_f.write_fmt(format_args!("{}\n", addr)).unwrap();
+    }
+    outfile.flush().unwrap();
+
+    let outpath = outfile.path().as_os_str().to_string_lossy().to_string();
+
+    // outfile needs to live on so destructor doesn't delete it before we run the benches
+    (outfile, make_v4_tests(outpath.as_str()))
+}
+
+fn hyperfine_harness<S>(cmd: S) -> Result<TestResult, Box<dyn std::error::Error>>
+where
+    S: AsRef<OsStr>,
+{
+    let resultfile = NamedTempFile::new().expect("Unable to create tempfile");
+
+    let mut process = std::process::Command::new("hyperfine")
+        .arg("--export-json")
+        .arg(resultfile.path())
+        .arg("--min-runs")
+        .arg("10")
+        .arg("-N")
+        .arg("--")
+        .arg(&cmd)
+        .stdout(Stdio::null())
+        .spawn()
+        .expect("unable to run command");
+    let _rc = process.wait().expect("unable to wait on process");
+
+    let mut raw_result_buf = Vec::new();
+    resultfile
+        .as_file()
+        .read_to_end(&mut raw_result_buf)
+        .expect("Can't read results");
+    resultfile.close().unwrap();
+
+    let hf_result = json::parse(&String::from_utf8_lossy(&raw_result_buf)).expect(
+        format!(
+            "Can't parse hyperfine json results from command `{}`",
+            cmd.as_ref().to_string_lossy()
+        )
+        .as_str(),
+    );
+
+    let final_result = &hf_result["results"][0];
+
+    Ok((final_result.clone()).into())
+}
+
+fn plot_results(
+    results: &Vec<(TestDefinition, TestResult)>,
+    caption: &str,
+    outfile: &str,
+) -> Result<(), Box<dyn std::error::Error>> {
+    // Second result is our baseline
+    let norm_numerator = results[1].1.mean;
+    let max_result = norm_numerator / results.iter().map(|x| x.1.mean).reduce(f64::min).unwrap();
+
+    let drawing = BitMapBackend::new(outfile, (640, 480)).into_drawing_area();
+    drawing.fill(&WHITE)?;
+    let mut chart = ChartBuilder::on(&drawing)
+        .x_label_area_size(40)
+        .y_label_area_size(40)
+        .caption(caption, ("Roboto", 24).into_font())
+        .build_cartesian_2d((0..results.len() - 1).into_segmented(), 0.0..max_result)?;
+
+    chart
+        .configure_mesh()
+        .y_desc("Speedup vs aggregate6")
+        .y_labels(5)
+        .y_label_formatter(&|x| std::fmt::format(format_args!("{:.0}", *x)))
+        .light_line_style(WHITE)
+        .bold_line_style(GREY)
+        .disable_x_mesh()
+        .x_label_style(("Roboto", 18).into_font())
+        .x_label_formatter(&|x| match x {
+            SegmentValue::Exact(val) => results[*val].0.name.clone(),
+            SegmentValue::CenterOf(val) => results[*val].0.name.clone(),
+            SegmentValue::Last => String::new(),
+        })
+        .draw()?;
+
+    chart.draw_series(
+        Histogram::vertical(&chart)
+            .style(BAR_COLOUR.filled())
+            .margin(10)
+            .data(
+                results
+                    .iter()
+                    .enumerate()
+                    .map(|(x, y)| (x, norm_numerator / y.1.mean)),
+            ),
+    )?;
+
+    chart.draw_series(PointSeries::of_element(
+        results
+            .iter()
+            .enumerate()
+            .map(|(x, y)| (SegmentValue::CenterOf(x), norm_numerator / y.1.mean)),
+        5,
+        ShapeStyle::from(&BLACK).filled(),
+        &|coord, _size, _style| {
+            let (target_y, target_colour) = if coord.1 < 25.0 {
+                (-25, BAR_COLOUR)
+            } else {
+                (25, WHITE)
+            };
+            EmptyElement::at(coord.clone())
+                + Text::new(
+                    format!("{:.1} x", coord.1),
+                    (0, target_y),
+                    ("Roboto", 18)
+                        .into_font()
+                        .color(&target_colour)
+                        .pos(Pos::new(HPos::Center, VPos::Center)),
+                )
+        },
+    ))?;
+
+    Ok(())
+}
+
+fn main() -> Result<(), Box<dyn std::error::Error>> {
+    run_and_plot(
+        make_tests("test-data/dfz_combined/input"),
+        "doc/perfcomp_all.png",
+        "IPv4 & IPv6 Full DFZ",
+    )?;
+    run_and_plot(
+        make_v4_tests("test-data/dfz_v4/input"),
+        "doc/perfcomp_v4.png",
+        "IPv4 Full DFZ",
+    )?;
+
+    // Need to hold on to tmpfile so it doesn't get deleted before we can bench
+    let (_tmpfile, tests) = make_startup_tests();
+    run_and_plot(
+        tests,
+        "doc/perfcomp_startup.png",
+        "1024 Random IPv4 Prefixes",
+    )?;
+
+    Ok(())
+}
+
+fn run_and_plot(
+    tests: Vec<TestDefinition>,
+    filename: &str,
+    caption: &str,
+) -> Result<(), Box<dyn std::error::Error>> {
+    let mut results: Vec<(TestDefinition, TestResult)> = Vec::new();
+    for test in tests {
+        println!("Running bench: {:?}", test);
+        results.push((test.clone(), hyperfine_harness(&test.cmd)?));
+    }
+    plot_results(&results, caption, filename)?;
+    Ok(())
+}

src/iputils.rs

@@ -63,7 +63,7 @@ pub struct IpOrNet(IpNet);
 #[derive(Debug, Clone)]
 pub struct NetParseError {
     #[allow(dead_code)]
-    msg: String,
+    msg: &'static str,
 }
 
 impl Display for NetParseError {
@@ -87,7 +87,7 @@ impl IpOrNet {
                 Ok(lead_ones.try_into()?)
             } else {
                 Err(Box::new(NetParseError {
-                    msg: "Invalid subnet mask".to_owned(),
+                    msg: "Invalid subnet mask",
                 }))
             }
         } else {
@@ -96,7 +96,7 @@ impl IpOrNet {
                 Ok(lead_zeros.try_into()?)
             } else {
                 Err(Box::new(NetParseError {
-                    msg: "Invalid wildcard mask".to_owned(),
+                    msg: "Invalid wildcard mask",
                 }))
             }
         }
@@ -113,7 +113,7 @@ impl IpOrNet {
                     Ok(IpNet::new(ip, IpOrNet::parse_mask(pfxlen)?)?.into())
                 } else {
                     Err(Box::new(NetParseError {
-                        msg: "Mask form is not valid for IPv6 address".to_owned(),
+                        msg: "Mask form is not valid for IPv6 address",
                     }))
                 }
             }
@@ -140,6 +140,9 @@ impl IpOrNet {
     pub fn network(&self) -> IpAddr {
         self.0.network()
     }
+    pub fn has_host_bits(&self) -> bool {
+        self.0.addr() != self.0.network()
+    }
 }
 
 impl FromStr for IpOrNet {
@@ -209,7 +212,7 @@ impl Default for PrefixlenPair {
 
 impl Display for PrefixlenPair {
     fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
-        f.write_str(format!("{},{}", self.v4, self.v6).as_str())
+        f.write_fmt(format_args!("{},{}", self.v4, self.v6))
     }
 }
 
@@ -263,12 +266,12 @@ impl PartialOrd<IpOrNet> for PrefixlenPair {
 
 #[derive(Debug)]
 pub struct ParsePrefixlenError {
-    msg: String,
+    msg: &'static str,
 }
 
 impl Display for ParsePrefixlenError {
     fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
-        f.write_str(self.msg.as_str())
+        f.write_str(self.msg)
     }
 }
 
@@ -280,25 +283,25 @@ impl FromStr for PrefixlenPair {
         match s.split_once(',') {
             Some(pair) => {
                 let v4 = u8::from_str(pair.0).or(Err(ParsePrefixlenError {
-                    msg: "Unable to parse integer".to_owned(),
+                    msg: "Unable to parse integer",
                 }))?;
                 let v6 = u8::from_str(pair.1).or(Err(ParsePrefixlenError {
-                    msg: "Unable to parse integer".to_owned(),
+                    msg: "Unable to parse integer",
                 }))?;
                 if v4 > 32 || v6 > 128 {
                     return Err(ParsePrefixlenError {
-                        msg: "Invalid prefix length".to_owned(),
+                        msg: "Invalid prefix length",
                     });
                 }
                 Ok(PrefixlenPair { v4, v6 })
             }
             None => {
                 let len = u8::from_str(s).or(Err(ParsePrefixlenError {
-                    msg: "Unable to parse integer".to_owned(),
+                    msg: "Unable to parse integer",
                 }))?;
                 if len > 128 {
                     return Err(ParsePrefixlenError {
-                        msg: "Invalid prefix length".to_owned(),
+                        msg: "Invalid prefix length",
                     });
                 }
                 Ok(PrefixlenPair { v4: len, v6: len })

src/main.rs

@@ -1,44 +1,33 @@
 extern crate ipnet;
 
-use std::{process::exit, io};
+use std::{io, process::exit};
 
 mod iputils;
 use iputils::{IpBothRange, IpOrNet, PrefixlenPair};
 
 use clio::*;
-use std::io::{Write, BufRead};
+use std::io::{BufRead, Write};
 
 use clap::Parser;
 
-const WRITER_BUFSIZE: usize = 32 * 1024;
+const WRITER_BUFSIZE: usize = 1 * 1024;
 
 #[derive(Parser)]
-#[command(author, version, about, long_about=None)]
+#[command(author, version, about)]
 struct Args {
     #[clap(value_parser, default_value = "-")]
     input: Vec<Input>,
-    #[structopt(
+    /// Maximum prefix length for prefixes read. Single value applies to IPv4 and IPv6, comma-separated [IPv4],[IPv6].
-        short,
+    #[structopt(short, long, default_value = "32,128")]
-        long,
-        default_value = "32,128",
-        help = "Maximum prefix length for prefixes read. Single value applies to IPv4 and IPv6, comma-separated [IPv4],[IPv6]."
-    )]
     max_prefixlen: PrefixlenPair,
-    #[arg(short, long, help = "truncate IP/mask to network/mask (else ignore)")]
+    /// Truncate IP/mask to network/mask (else ignore)
+    #[arg(short, long)]
     truncate: bool,
-    #[arg(
+    /// Only output IPv4 prefixes
-        id = "4",
+    #[arg(id = "4", short, conflicts_with("6"))]
-        short,
-        help = "Only output IPv4 prefixes",
-        conflicts_with("6")
-    )]
     only_v4: bool,
-    #[arg(
+    /// Only output IPv6 prefixes
-        id = "6",
+    #[arg(id = "6", short, conflicts_with("4"))]
-        short,
-        help = "Only output IPv6 prefixes",
-        conflicts_with("4")
-    )]
     only_v6: bool,
 }
 
@@ -71,16 +60,17 @@ struct App {
 }
 
 impl App {
-    fn add_prefix(&mut self, pfx: IpOrNet) {
+    fn add_prefix<const TRUNCATE: bool>(&mut self, pfx: IpOrNet) {
         // Parser accepts host bits set, so detect that case and error if not truncate mode
         // Note: aggregate6 errors in this case regardless of -4, -6 so do the same
-        if !self.args.truncate {
+        if !TRUNCATE && pfx.has_host_bits() {
-            if pfx.addr() != pfx.network() {
+            // We don't have the original string any more so our error
-                eprintln!("ERROR: '{}' is not a valid IP network, ignoring.", pfx);
+            // differs from `aggregate6` in that it prints the pfxlen as
-                return;
+            // parsed, not as in the source.
-            }
+            eprintln!("ERROR: '{}' is not a valid IP network, ignoring.", pfx);
+            return;
         }
-        // Don't bother saving if we won't display.
+
         if self.args.only_v4 && pfx.is_ipv6() {
             return;
         } else if self.args.only_v6 && pfx.is_ipv4() {
@@ -90,14 +80,14 @@ impl App {
             self.prefixes.add(pfx);
         }
     }
-    fn consume_input(&mut self, input: &mut Input) {
+    fn consume_input<const TRUNCATE: bool>(&mut self, input: &mut Input) {
         for line in input.lock().lines() {
             match line {
                 Ok(line) => {
-                    for net in line.split_whitespace() {
+                    for net in line.split_ascii_whitespace() {
                         let pnet = net.parse::<IpOrNet>();
                         match pnet {
-                            Ok(pnet) => self.add_prefix(pnet),
+                            Ok(pnet) => self.add_prefix::<TRUNCATE>(pnet),
                             Err(_e) => {
                                 eprintln!("ERROR: '{}' is not a valid IP network, ignoring.", net);
                             }
@@ -114,7 +104,10 @@ impl App {
     fn simplify_inputs(&mut self) {
         let inputs = self.args.input.to_owned();
         for mut input in inputs {
-            self.consume_input(&mut input);
+            match self.args.truncate {
+                true => self.consume_input::<true>(&mut input),
+                false => self.consume_input::<false>(&mut input),
+            }
         }
         self.prefixes.simplify();
     }

tests/cli.rs

@@ -43,14 +43,24 @@ impl Predicate<[u8]> for SortedEquals {
 
 impl PredicateReflection for SortedEquals {}
 
-// Really should normalize the data (lex sort) before comparison
+/// Compare the output with pre-prepared expected outputs. When functionality is
+/// matching, we generate expected outputs with `aggregate6`, and expect byte-for-byte
+/// output consistency, including ordering. When our functionality and `aggregate6`'s
+/// diverge, we generate expected outputs ourselves, and expect output sorted by numeric
+/// value of the address.
+///
+/// Normalization is available for future test cases.
 #[rstest]
-#[case("test-data/dfz_combined", "")] // Basic aggregation test
+#[case::dfz_combined("test-data/dfz_combined", "", false)] // Basic aggregation test
-#[case("test-data/max_pfxlen", "-m 20")] // Filter on prefix length
+#[case::max_pfxlen("test-data/max_pfxlen", "-m 20", false)] // Filter on prefix length
-#[case("test-data/max_pfxlen_split", "-m 20,32")] // Filter on prefix length (split v4/v6)
+#[case::max_pfxlen_split("test-data/max_pfxlen_split", "-m 20,32", false)] // Filter on prefix length (split v4/v6)
-#[case("test-data/v4_only", "-4")] // Filter v4 only
+#[case::v4_only("test-data/v4_only", "-4", false)] // Filter v4 only
-#[case("test-data/v6_only", "-6")] // Filter v4 only
+#[case::v6_only("test-data/v6_only", "-6", false)] // Filter v6 only
-fn dfz_test(#[case] path: &str, #[case] args: &str) -> Result<(), Box<dyn Error>> {
+fn dfz_test(
+    #[case] path: &str,
+    #[case] args: &str,
+    #[case] normalize_data: bool,
+) -> Result<(), Box<dyn Error>> {
     let mut cmd = Command::cargo_bin("rs-aggregate")?;
     let in_path = Path::new(path).join("input");
     let expect_path = Path::new(path).join("expected");
@@ -65,10 +75,17 @@ fn dfz_test(#[case] path: &str, #[case] args: &str) -> Result<(), Box<dyn Error>
         .timeout(std::time::Duration::from_secs(30))
         .assert();
 
-    assert
+    if normalize_data {
-        .success()
+        assert
-        .stdout(SortedEquals::new(&expect_data))
+            .success()
-        .stderr(predicate::str::is_empty());
+            .stdout(SortedEquals::new(&expect_data))
+            .stderr(predicate::str::is_empty());
+    } else {
+        assert
+            .success()
+            .stdout(predicate::eq(expect_data))
+            .stderr(predicate::str::is_empty());
+    }
 
     Ok(())
 }