Fix nan/inf output for very fast commands

closes #319

CHANGELOG.md

@@ -24,6 +24,8 @@
 ## Bugfixes
 
 - Fix a bug in the outlier detection which would only detect "slow outliers" but not the fast ones (runs that are much faster than the rest of the benchmarking runs), see #329
+- Better error messages for very fast commands that would lead to inf/nan results in the relative
+  speed comparison, see #319
 - Keep output colorized when the output is not interactive and `--style=full` or `--style=color` is used.
 
 ## Other

src/hyperfine/export/markdown.rs

@@ -5,7 +5,7 @@ use crate::hyperfine::internal::{compute_relative_speed, BenchmarkResultWithRela
 use crate::hyperfine::types::BenchmarkResult;
 use crate::hyperfine::units::Unit;
 
-use std::io::Result;
+use std::io::{Error, ErrorKind, Result};
 
 #[derive(Default)]
 pub struct MarkdownExporter {}
@@ -23,15 +23,20 @@ impl Exporter for MarkdownExporter {
             Unit::Second
         };
 
-        let annotated_results = compute_relative_speed(results);
+        if let Some(annotated_results) = compute_relative_speed(results) {
+            let mut destination = start_table(unit);
 
-        let mut destination = start_table(unit);
+            for result in annotated_results {
+                add_table_row(&mut destination, &result, unit);
+            }
 
-        for result in annotated_results {
-            add_table_row(&mut destination, &result, unit);
+            Ok(destination)
+        } else {
+            Err(Error::new(
+                ErrorKind::Other,
+                "Relative speed comparison is not available for Markdown export.",
+            ))
         }
-
-        Ok(destination)
     }
 }
 

src/hyperfine/internal.rs

@@ -52,6 +52,7 @@ pub fn min(vals: &[f64]) -> f64 {
         .unwrap()
 }
 
+#[derive(Debug)]
 pub struct BenchmarkResultWithRelativeSpeed<'a> {
     pub result: &'a BenchmarkResult,
     pub relative_speed: Scalar,
@@ -65,53 +66,70 @@ fn compare_mean_time(l: &BenchmarkResult, r: &BenchmarkResult) -> Ordering {
 
 pub fn compute_relative_speed<'a>(
     results: &'a [BenchmarkResult],
-) -> Vec<BenchmarkResultWithRelativeSpeed<'a>> {
+) -> Option<Vec<BenchmarkResultWithRelativeSpeed<'a>>> {
     let fastest: &BenchmarkResult = results
         .iter()
         .min_by(|&l, &r| compare_mean_time(l, r))
         .expect("at least one benchmark result");
 
-    results
-        .iter()
-        .map(|result| {
-            let ratio = result.mean / fastest.mean;
+    if fastest.mean == 0.0 {
+        return None;
+    }
 
-            // https://en.wikipedia.org/wiki/Propagation_of_uncertainty#Example_formulas
-            // Covariance asssumed to be 0, i.e. variables are assumed to be independent
-            let ratio_stddev = ratio
-                * ((result.stddev / result.mean).powi(2) + (fastest.stddev / fastest.mean).powi(2))
+    Some(
+        results
+            .iter()
+            .map(|result| {
+                let ratio = result.mean / fastest.mean;
+
+                // https://en.wikipedia.org/wiki/Propagation_of_uncertainty#Example_formulas
+                // Covariance asssumed to be 0, i.e. variables are assumed to be independent
+                let ratio_stddev = ratio
+                    * ((result.stddev / result.mean).powi(2)
+                        + (fastest.stddev / fastest.mean).powi(2))
                     .sqrt();
 
-            BenchmarkResultWithRelativeSpeed {
-                result,
-                relative_speed: ratio,
-                relative_speed_stddev: ratio_stddev,
-                is_fastest: result == fastest,
-            }
-        })
-        .collect()
+                BenchmarkResultWithRelativeSpeed {
+                    result,
+                    relative_speed: ratio,
+                    relative_speed_stddev: ratio_stddev,
+                    is_fastest: result == fastest,
+                }
+            })
+            .collect(),
+    )
 }
 
 pub fn write_benchmark_comparison(results: &[BenchmarkResult]) {
     if results.len() < 2 {
         return;
     }
 
-    let mut annotated_results = compute_relative_speed(&results);
-    annotated_results.sort_by(|l, r| compare_mean_time(l.result, r.result));
-
-    let fastest = &annotated_results[0];
-    let others = &annotated_results[1..];
-
-    println!("{}", "Summary".bold());
-    println!("  '{}' ran", fastest.result.command.cyan());
-
-    for item in others {
-        println!(
-            "{} ± {} times faster than '{}'",
-            format!("{:8.2}", item.relative_speed).bold().green(),
-            format!("{:.2}", item.relative_speed_stddev).green(),
-            &item.result.command.magenta()
+    if let Some(mut annotated_results) = compute_relative_speed(&results) {
+        annotated_results.sort_by(|l, r| compare_mean_time(l.result, r.result));
+
+        let fastest = &annotated_results[0];
+        let others = &annotated_results[1..];
+
+        println!("{}", "Summary".bold());
+        println!("  '{}' ran", fastest.result.command.cyan());
+
+        for item in others {
+            println!(
+                "{} ± {} times faster than '{}'",
+                format!("{:8.2}", item.relative_speed).bold().green(),
+                format!("{:.2}", item.relative_speed_stddev).green(),
+                &item.result.command.magenta()
+            );
+        }
+    } else {
+        eprintln!(
+            "{}: The benchmark comparison could not be computed as some benchmark times are zero. \
+             This could be caused by background interference during the initial calibration phase \
+             of hyperfine, in combination with very fast commands (faster than a few milliseconds). \
+             Try to re-run the benchmark on a quiet system. If it does not help, you command is \
+             most likely too fast to be accurately benchmarked by hyperfine.",
+             "Note".bold().red()
         );
     }
 }
@@ -125,12 +143,11 @@ fn test_max() {
     assert_eq!(1.0, max(&[-1.0, 1.0, 0.0]));
 }
 
-#[test]
-fn test_compute_relative_speed() {
-    use approx::assert_relative_eq;
+#[cfg(test)]
+fn create_result(name: &str, mean: Scalar) -> BenchmarkResult {
     use std::collections::BTreeMap;
 
-    let create_result = |name: &str, mean| BenchmarkResult {
+    BenchmarkResult {
         command: name.into(),
         mean,
         stddev: 1.0,
@@ -141,21 +158,35 @@ fn test_compute_relative_speed() {
         max: mean,
         times: None,
         parameters: BTreeMap::new(),
-    };
+    }
+}
+
+#[test]
+fn test_compute_relative_speed() {
+    use approx::assert_relative_eq;
 
     let results = vec![
         create_result("cmd1", 3.0),
         create_result("cmd2", 2.0),
         create_result("cmd3", 5.0),
     ];
 
-    let annotated_results = compute_relative_speed(&results);
+    let annotated_results = compute_relative_speed(&results).unwrap();
 
     assert_relative_eq!(1.5, annotated_results[0].relative_speed);
     assert_relative_eq!(1.0, annotated_results[1].relative_speed);
     assert_relative_eq!(2.5, annotated_results[2].relative_speed);
 }
 
+#[test]
+fn test_compute_relative_speed_for_zero_times() {
+    let results = vec![create_result("cmd1", 1.0), create_result("cmd2", 0.0)];
+
+    let annotated_results = compute_relative_speed(&results);
+
+    assert!(annotated_results.is_none());
+}
+
 pub fn tokenize<'a>(values: &'a str) -> Vec<String> {
     let mut tokens = vec![];
     let mut buf = String::new();

src/main.rs

@@ -16,7 +16,7 @@ use crate::hyperfine::export::{ExportManager, ExportType};
 use crate::hyperfine::internal::{tokenize, write_benchmark_comparison};
 use crate::hyperfine::parameter_range::get_parameterized_commands;
 use crate::hyperfine::types::{
-    BenchmarkResult, CmdFailureAction, Command, HyperfineOptions, OutputStyleOption, ParameterValue,
+    CmdFailureAction, Command, HyperfineOptions, OutputStyleOption, ParameterValue,
 };
 use crate::hyperfine::units::Unit;
 
@@ -27,7 +27,11 @@ pub fn error(message: &str) -> ! {
 }
 
 /// Runs the benchmark for the given commands
-fn run(commands: &[Command<'_>], options: &HyperfineOptions) -> io::Result<Vec<BenchmarkResult>> {
+fn run(
+    commands: &[Command<'_>],
+    options: &HyperfineOptions,
+    export_manager: &ExportManager,
+) -> io::Result<()> {
     let shell_spawning_time =
         mean_shell_spawning_time(&options.shell, options.output_style, options.show_output)?;
 
@@ -47,7 +51,21 @@ fn run(commands: &[Command<'_>], options: &HyperfineOptions) -> io::Result<Vec<B
         timing_results.push(run_benchmark(num, cmd, shell_spawning_time, options)?);
     }
 
-    Ok(timing_results)
+    // Print relative speed comparison
+    if options.output_style != OutputStyleOption::Disabled {
+        write_benchmark_comparison(&timing_results);
+    }
+
+    // Export results
+    let ans = export_manager.write_results(timing_results, options.time_unit);
+    if let Err(e) = ans {
+        error(&format!(
+            "The following error occurred while exporting: {}",
+            e
+        ));
+    }
+
+    Ok(())
 }
 
 fn main() {
@@ -57,26 +75,12 @@ fn main() {
     let commands = build_commands(&matches);
 
     let res = match options {
-        Ok(ref opts) => run(&commands, &opts),
+        Ok(ref opts) => run(&commands, &opts, &export_manager),
         Err(ref e) => error(&e.to_string()),
     };
 
     match res {
-        Ok(timing_results) => {
-            let options = options.unwrap();
-
-            if options.output_style != OutputStyleOption::Disabled {
-                write_benchmark_comparison(&timing_results);
-            }
-
-            let ans = export_manager.write_results(timing_results, options.time_unit);
-            if let Err(e) = ans {
-                error(&format!(
-                    "The following error occurred while exporting: {}",
-                    e
-                ));
-            }
-        }
+        Ok(_) => {}
         Err(e) => error(&e.to_string()),
     }
 }