release v0.3.7 🥳

`-j`/`--jobs` can be used to manually set the number of threads to use for scanning files. additionally, fif won't bother with multi-threaded scanning when there are less than 32 files to scan.

.idea/runConfigurations/Run.xml

@@ -6,6 +6,7 @@
     <option name="requiredFeatures" value="true" />
     <option name="allFeatures" value="false" />
     <option name="emulateTerminal" value="false" />
+    <option name="withSudo" value="false" />
     <option name="backtrace" value="SHORT" />
     <envs>
       <env name="RUST_LOG" value="debug" />

.idea/vcs.xml

@@ -2,5 +2,6 @@
 <project version="4">
   <component name="VcsDirectoryMappings">
     <mapping directory="$PROJECT_DIR$" vcs="Git" />
+    <mapping directory="$PROJECT_DIR$/src/walkdir" vcs="Git" />
   </component>
 </project>

CHANGELOG.md

@@ -5,9 +5,15 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
 ## Unreleased
+
+## v0.3.7 - 2021-09-25
 ### Added
+- `-j`/`--jobs` flag for specifying the number of threads fif should use for scanning files
 - AIFF (Audio Interchange File Format, a PCM audio format like WAV) detection to [`infer`] backend
 - `--version` output now includes the (short) hash of the git commit fif was built from
+### Changed
+- fif will no longer use multithreading when scanning less than 32 files - the overhead of spawning threads isn't really
+  worth it
 ### Other
 - Refactoring - split fif into `main.rs` and `lib.rs`, moved file-related functionality (directory scanning, etc.) into
 files module, removed string module, etc.

Cargo.lock

@@ -1,5 +1,7 @@
 # This file is automatically @generated by Cargo.
 # It is not intended for manual editing.
+version = 3
+
 [[package]]
 name = "arrayvec"
 version = "0.5.2"
@@ -173,7 +175,7 @@ checksum = "de853764b47027c2e862a995c34978ffa63c1501f2e15f987ba11bd4f9bba193"
 
 [[package]]
 name = "fif"
-version = "0.3.6"
+version = "0.3.7"
 dependencies = [
  "bitflags",
  "cfg-if",
@@ -186,6 +188,7 @@ dependencies = [
  "log",
  "mime",
  "new_mime_guess",
+ "num_cpus",
  "once_cell",
  "rand",
  "rayon",

Cargo.toml

@@ -1,7 +1,7 @@
 [package]
 name = "fif"
 description = "A command-line tool for detecting and optionally correcting files with incorrect extensions."
-version = "0.3.6"
+version = "0.3.7"
 authors = ["Lynnesbian <lynne@bune.city>"]
 edition = "2018"
 license = "GPL-3.0-or-later"
@@ -17,7 +17,7 @@ maintenance = { status = "experimental" }
 
 [features]
 default = ["multi-threaded", "json"]
-multi-threaded = ["rayon"]
+multi-threaded = ["rayon", "num_cpus"]
 infer-backend = ["infer"]
 xdg-mime-backend = ["xdg-mime"]
 json = ["serde", "serde_json"]
@@ -36,6 +36,7 @@ itertools = "0.10.0"
 serde = { version = "1.0", features = ["derive"], optional = true }
 serde_json = { version = "1.0", optional = true }
 bitflags = "~1.2.1" # 1.3+ requires Rust >= 1.46
+num_cpus = { version = "1.13.0", optional = true }
 
 [target.'cfg(not(unix))'.dependencies]
 xdg-mime = { version = "0.3.3", optional = true }

src/files.rs

@@ -136,25 +136,42 @@ pub fn scan_file(entry: &DirEntry, canonical_paths: bool) -> Result<Findings, Sc
 }
 
 /// Takes a slice of [`DirEntry`]s and calls [`scan_file`] on each one, returning the results in a vector.
-pub fn scan_from_walkdir(entries: &[DirEntry], canonical_paths: bool) -> Vec<Result<Findings, ScanError>> {
+pub fn scan_from_walkdir(
+	entries: &[DirEntry],
+	canonical_paths: bool,
+	use_threads: bool,
+) -> Vec<Result<Findings, ScanError>> {
 	cfg_if! {
 		if #[cfg(feature = "multi-threaded")] {
 			use rayon::prelude::*;
+			const CHUNKS: usize = 32;
 
-			// split the entries into chunks of 32, and iterate over each chunk of entries in a separate thread
-			entries
-				.par_chunks(32)
-				.flat_map(|chunk| {
-					chunk
-						.iter() // iter over the chunk, which is a slice of DirEntry structs
-						.map(|entry| scan_file(entry, canonical_paths))
-						.collect::<Vec<_>>()
-				})
-				.collect()
+			if use_threads && entries.len() > CHUNKS {
+				// split the entries into chunks of 32, and iterate over each chunk of entries in a separate thread
+				return entries
+					.par_chunks(CHUNKS)
+					.flat_map_iter(|chunk| {
+						chunk
+							.iter() // iter over the chunk, which is a slice of DirEntry structs
+							.map(|entry| scan_file(entry, canonical_paths))
+							.collect::<Vec<_>>() // TODO: is there a way to avoid having to collect here?
+					})
+					.collect()
+			}
 		} else {
-			entries.iter().map(|entry: &DirEntry| scan_file(entry, canonical_paths)).collect()
+			// should always be false when multi-threading is disabled at compile time
+			assert!(!use_threads)
 		}
 	}
+
+	// if we end up here, either
+	// - there were less than CHUNKS files to scan, or
+	// - the user specified that only one thread should be used, by specifying `-j 1`
+	// - fif was compiled without the `multi-threading` feature
+	entries
+		.iter()
+		.map(|entry: &DirEntry| scan_file(entry, canonical_paths))
+		.collect()
 }
 
 /// Scans a given directory with [`WalkDir`], filters with [`wanted_file`], checks for errors, and returns a vector of

src/formats.rs

@@ -7,7 +7,7 @@ use std::os::unix::ffi::OsStrExt;
 use std::path::Path;
 
 use cfg_if::cfg_if;
-use itertools::{Either, Itertools};
+use itertools::Itertools;
 use snailquote::escape;
 
 use crate::findings::ScanError;
@@ -338,6 +338,8 @@ pub struct Json;
 #[cfg(feature = "json")]
 impl Format for Json {
 	fn write_all<W: Write>(&self, f: &mut W, entries: &Entries) -> io::Result<()> {
+		use itertools::Either;
+
 		#[derive(serde::Serialize)]
 		struct SerdeEntries<'a> {
 			errors: &'a Vec<&'a ScanError<'a>>,

src/main.rs

@@ -20,6 +20,7 @@
 use std::io::{stdout, BufWriter, Write};
 use std::process::exit;
 
+use cfg_if::cfg_if;
 use clap::Clap;
 use fif::files::{scan_directory, scan_from_walkdir};
 use fif::formats::Format;
@@ -87,7 +88,28 @@ fn main() {
 
 	trace!("Found {} items to check", entries.len());
 
-	let results: Vec<_> = scan_from_walkdir(&entries, args.canonical_paths)
+	cfg_if! {
+		if #[cfg(feature = "multi-threaded")] {
+			let use_threads = args.jobs != 1;
+
+			if use_threads {
+				// 0 is a special case - it should be understood to mean "all available host CPUs"
+				let jobs = if args.jobs == 0 { num_cpus::get() } else { args.jobs };
+
+				// set up the global thread pool with the requested number of threads
+				rayon::ThreadPoolBuilder::new().num_threads(jobs).build_global().unwrap();
+				trace!("Multithreading enabled, using {} threads", jobs);
+			} else {
+				trace!("Multithreading disabled at runtime");
+			}
+
+		} else { // `multi-threading` feature disabled
+			let use_threads = false;
+			trace!("Multithreading disabled at compile time");
+		}
+	}
+
+	let results: Vec<_> = scan_from_walkdir(&entries, args.canonical_paths, use_threads)
 		.into_iter()
 		.filter(
 			|result| result.is_err() || !result.as_ref().unwrap().valid,
@@ -101,12 +123,16 @@ fn main() {
 	for result in &results {
 		match result {
 			Ok(r) => {
-				debug!(
-					"{:?} is of type {}, should have extension \"{}\"",
-					r.file,
-					r.mime,
-					r.recommended_extension().unwrap_or_else(|| "???".into())
-				);
+				// check to see if debug logging is enabled before invoking debug! macro
+				// https://github.com/rust-lang/log/pull/394#issuecomment-630490343
+				if log::max_level() >= log::Level::Debug {
+					debug!(
+						"{:?} is of type {}, should have extension \"{}\"",
+						r.file,
+						r.mime,
+						r.recommended_extension().unwrap_or_else(|| "???".into())
+					);
+				}
 			}
 			Err(f) => warn!("{}", f),
 		}

src/parameters.rs

@@ -133,6 +133,14 @@ pub struct Parameters {
 	/// For example, with this option, fif will not rename "image.unknown" to "image.jpg"
 	#[clap(short = 'I', long)]
 	pub ignore_unknown_exts: bool,
+
+	#[cfg(feature = "multi-threaded")]
+	/// Number of jobs (threads) to use when scanning results.
+	/// The default behaviour is to use one thread per CPU thread. This behaviour can be manually requested by setting
+	/// `-j 0`. Using `-j 1` will disable multi-threading behaviour, as if you had compiled fif with the multi-threading
+	/// feature disabled. Setting more jobs than you have CPU threads is not recommended.
+	#[clap(short = 'j', long, default_value = "0")]
+	pub jobs: usize,
 }
 
 fn lowercase_exts(exts: &str) -> Result<(), String> {

src/tests/mod.rs

@@ -114,8 +114,10 @@ fn simple_directory() {
 	// there should be one file missing: "ignore.fake_ext"
 	assert_eq!(entries.len(), files.len() - 1);
 
-	let results = scan_from_walkdir(&entries, false);
-	let canonical_results = scan_from_walkdir(&entries, true);
+	let use_threads = cfg!(feature = "multi-threaded");
+
+	let results = scan_from_walkdir(&entries, false, use_threads);
+	let canonical_results = scan_from_walkdir(&entries, true, use_threads);
 	assert_eq!(results.len(), canonical_results.len());
 
 	for (result, canonical_result) in results.iter().zip(canonical_results.iter()) {
@@ -291,6 +293,8 @@ fn rejects_bad_args() {
 		vec!["fif", "-X", "pebis"],
 		// `-e` with nothing but commas:
 		vec!["fif", "-e", ",,,,,"],
+		// `-j` with a negative value:
+		vec!["fif", "-j", "-1"],
 	];
 
 	for test in &tests {