Maven Central fingerprint kind (#6)

CHANGELOG.md

@@ -1,4 +1,10 @@
 
+# v2.1.0
+
+- Adds new `Kind::JarMavenCentralV1` fingerprint.
+- Attempts to improve performance of file-based fingerprinting by paralellizing across threads.
+  - This was done now that we have several kinds of fingerprints, and we'll probably just keep adding more.
+
 # v2.0.0
 
 Refactored to the new Sparkle-based view of fingerprints.

Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "fingerprint"
-version = "2.0.0"
+version = "2.1.0"
 edition = "2021"
 
 [features]
@@ -42,6 +42,7 @@ strum = { version = "0.26.2", features = ["derive"] }
 alphanumeric-sort = "1.5.3"
 tap = "1.0.1"
 tracing = "0.1.40"
+sha1 = "0.10.6"
 
 [dev-dependencies]
 pretty_assertions = "1.4.0"

src/fingerprint/jar.rs

@@ -1,5 +1,6 @@
 use std::io::{BufRead, Seek};
 
+use sha1::Sha1;
 use sha2::{Digest, Sha256};
 use tap::Pipe;
 use tracing::warn;
@@ -22,6 +23,15 @@ pub fn raw(stream: impl BufRead + Seek) -> Result<Option<Fingerprint>, Error> {
     }
 }
 
+/// Fingerprint the java archive the same way as Maven Central.
+#[tracing::instrument(level = tracing::Level::DEBUG, skip_all, ret)]
+pub fn maven_central(mut stream: impl BufRead + Seek) -> Result<Option<Fingerprint>, Error> {
+    let mut hasher = Sha1::new();
+    std::io::copy(&mut stream, &mut hasher)?;
+    let content = Content::from_digest(hasher);
+    Ok(Some(Fingerprint::new(Kind::JarMavenCentralV1, content)))
+}
+
 /// Fingerprint class files inside a java archive (a JAR).
 #[tracing::instrument(level = tracing::Level::DEBUG, skip_all, ret)]
 pub fn class(stream: impl BufRead + Seek) -> Result<Option<Fingerprint>, Error> {

src/lib.rs

@@ -9,12 +9,14 @@ use std::{
     fs::File,
     io::{BufRead, BufReader, Cursor, Seek},
     path::Path,
+    thread::ScopedJoinHandle,
 };
 
 use getset::Getters;
 use serde::{Deserialize, Serialize};
 use sha2::{Digest, Sha256};
 use strum::{AsRefStr, Display, EnumIter, IntoEnumIterator, VariantNames};
+use tap::Pipe;
 use thiserror::Error;
 
 mod fingerprint;
@@ -123,6 +125,15 @@ pub enum Kind {
     #[strum(serialize = "v1.raw.jar")]
     JarRawV1,
 
+    /// Represents a fingerprint derived by hashing the raw contents of a JAR file in the same manner
+    /// as Maven Central. The idea is that such fingerprints can then be looked up via the
+    /// Maven Central REST API as a fallback to our own indexing.
+    ///
+    /// Specifically:
+    /// - The content of the JAR file is hashed as-is using the sha1 algorithm.
+    #[strum(serialize = "v1.mavencentral.jar")]
+    JarMavenCentralV1,
+
     /// Represents a fingerprint derived by hashing the raw contents of a JAR file with the SHA256 algorithm
     /// in a platform-independent manner.
     ///
@@ -162,7 +173,7 @@ impl<'de> Deserialize<'de> for Kind {
 }
 
 /// An array of bytes representing a fingerprint's content.
-#[derive(Clone, Eq, PartialEq, Hash, Default)]
+#[derive(Clone, Eq, PartialEq, Ord, PartialOrd, Hash, Default)]
 pub struct Content(Vec<u8>);
 
 impl Content {
@@ -323,6 +334,7 @@ impl Fingerprint {
             Kind::CommentStrippedSha256 => fingerprint::text::comment_stripped(stream),
             Kind::JarRawV1 => fingerprint::jar::raw(stream),
             Kind::JarClassV1 => fingerprint::jar::class(stream),
+            Kind::JarMavenCentralV1 => fingerprint::jar::maven_central(stream),
         }
     }
 
@@ -390,7 +402,11 @@ impl<K: Into<Kind>, C: Into<Content>> From<(K, C)> for Fingerprint {
 pub struct Combined(HashMap<Kind, Content>);
 
 impl Combined {
-    /// Fingerprint the provided stream (typically a file handle) with all fingerprint [`Kind`]s.
+    /// Fingerprint the provided stream with all fingerprint [`Kind`]s.
+    ///
+    /// Note: this forces fingerprinting to be performed serially
+    /// since the stream has to be seeked backwards for each fingerprinter;
+    /// if this is not desired consider [`Combined::from_file`] or [`Combined::from_buffer`].
     #[tracing::instrument(level = tracing::Level::DEBUG, skip_all, ret)]
     pub fn from_stream(mut stream: impl BufRead + Seek) -> Result<Self, Error> {
         let mut fingerprints = Vec::new();
@@ -407,10 +423,22 @@ impl Combined {
     }
 
     /// Fingerprint the provided file with all fingerprint [`Kind`]s.
+    ///
+    /// Note: this opens the file multiple times, once for each kind of fingerprint,
+    /// then runs each fingerprinter in its own thread.
+    /// If this is not desired consider [`Combined::from_stream`] or [`Combined::from_buffer`].
     #[tracing::instrument(level = tracing::Level::DEBUG, ret)]
     pub fn from_file(path: &Path) -> Result<Self, Error> {
-        let mut file = BufReader::new(File::open(path)?);
-        Self::from_stream(&mut file)
+        std::thread::scope(|scope| {
+            let handles = Kind::iter()
+                .map(|kind| scope.spawn(move || Fingerprint::from_file(kind, path)))
+                .collect::<Vec<_>>();
+
+            match collapse_handles(handles) {
+                Ok(fps) => fps.into_iter().flatten().pipe(Combined::from).pipe(Ok),
+                Err(err) => Err(err),
+            }
+        })
     }
 
     /// Fingerprint the provided buffer with all fingerprint [`Kind`]s.
@@ -425,8 +453,13 @@ impl Combined {
     /// of errors in the future it isn't a breaking change.
     #[tracing::instrument(level = tracing::Level::DEBUG, fields(buf = %buf.as_ref().len()), ret)]
     pub fn from_buffer(buf: impl AsRef<[u8]>) -> Result<Self, Error> {
-        let mut content = Cursor::new(buf);
-        Self::from_stream(&mut content)
+        Kind::iter()
+            .map(|kind| Fingerprint::from_buffer(kind, buf.as_ref()))
+            .collect::<Result<Vec<_>, _>>()?
+            .into_iter()
+            .flatten()
+            .pipe(Combined::from)
+            .pipe(Ok)
     }
 
     /// Create a new instance from a single fingerprint.
@@ -462,3 +495,17 @@ impl<I: IntoIterator<Item = F>, F: Into<Fingerprint>> From<I> for Combined {
         )
     }
 }
+
+fn collapse_handles<T, E>(handles: Vec<ScopedJoinHandle<'_, Result<T, E>>>) -> Result<Vec<T>, E> {
+    let mut collected = Vec::new();
+    for handle in handles {
+        match handle.join() {
+            Err(err) => std::panic::resume_unwind(err),
+            Ok(operation) => match operation {
+                Ok(inner) => collected.push(inner),
+                Err(err) => return Err(err),
+            },
+        }
+    }
+    Ok(collected)
+}

tests/it/code_vsi.rs

@@ -1,7 +1,5 @@
 //! Tests for plain code files using legacy VSI fingerprints.
 
-use std::io::Cursor;
-
 use pretty_assertions::assert_eq;
 
 use fingerprint::*;
@@ -11,7 +9,7 @@ use fingerprint::*;
 ///
 /// ```ignore
 /// let content = b"hello world";
-/// let combined = Combined::from_stream(&mut Cursor::new(content)).expect("fingerprint");
+/// let combined = Combined::from_buffer(&content).expect("fingerprint");
 /// assert_fingerprint_eq!(Kind::RawSha256, content, combined);
 /// assert_fingerprint_eq!(Kind::CommentStrippedSha256, content, combined);
 /// ```
@@ -69,15 +67,15 @@ fn combined_getters() {
 #[test]
 fn fingerprints_binary_file() {
     let content = vec![1, 2, 3, 0, 1, 2, 3];
-    let combined = Combined::from_stream(&mut Cursor::new(content.clone())).expect("fingerprint");
+    let combined = Combined::from_buffer(&content).expect("fingerprint");
     assert_fingerprint_eq!(Kind::RawSha256, &content, combined);
     assert_fingerprint_eq!(Kind::CommentStrippedSha256, None, combined);
 }
 
 #[test]
 fn fingerprints_text_file() {
     let content = b"hello world";
-    let combined = Combined::from_stream(&mut Cursor::new(content)).expect("fingerprint");
+    let combined = Combined::from_buffer(content).expect("fingerprint");
     assert_fingerprint_eq!(Kind::RawSha256, content, combined);
     assert_fingerprint_eq!(Kind::CommentStrippedSha256, content, combined);
 }
@@ -88,7 +86,7 @@ fn fingerprints_text_file_stripping_cr() {
     let content_cs = b"hello world\nanother line\na final line";
     let without_cr = b"hello world\nanother line\na final line\n";
 
-    let combined = Combined::from_stream(&mut Cursor::new(content)).expect("fingerprint");
+    let combined = Combined::from_buffer(content).expect("fingerprint");
     assert_fingerprint_eq!(Kind::RawSha256, without_cr, combined);
     assert_fingerprint_eq!(Kind::CommentStrippedSha256, content_cs, combined);
 }
@@ -97,15 +95,15 @@ fn fingerprints_text_file_stripping_cr() {
 fn fingerprints_binary_file_appearing_as_text() {
     // Sourced from `[email protected]:chromium/chromium.git` at `tools/origin_trials/eftest.key` on commit 49249345609d505c8bb8b0b5a42ff4b68b9e6d41.
     let content = include_bytes!("../../testdata/eftest.key");
-    let combined = Combined::from_stream(&mut Cursor::new(content)).expect("fingerprint");
+    let combined = Combined::from_buffer(content).expect("fingerprint");
     assert_fingerprint_eq!(Kind::RawSha256, content, combined);
     assert_fingerprint_eq!(Kind::CommentStrippedSha256, None, combined);
 }
 
 #[test]
 fn comment_stripped_does_not_fingerprint_binary_file() {
     let content = vec![1, 2, 3, 0, 1, 2, 3];
-    let combined = Combined::from_stream(&mut Cursor::new(content)).expect("fingerprint");
+    let combined = Combined::from_buffer(content).expect("fingerprint");
     assert_fingerprint_eq!(Kind::CommentStrippedSha256, None, combined);
 }
 
@@ -123,7 +121,7 @@ int main() {
 }
 "#;
 
-    let combined = Combined::from_stream(&mut Cursor::new(content)).expect("fingerprint");
+    let combined = Combined::from_buffer(content).expect("fingerprint");
     let expected = Content::new(
         hex::decode("44fc8f68ab633c7ca0240a66e4ff038c0f2412fe69d14b6f052556edaa1b9160")
             .expect("decode hex literal"),