Add pdf cover extraction

crates/pile-dataset/Cargo.toml

@@ -25,6 +25,8 @@ blake3 = { workspace = true }
 epub = { workspace = true }
 kamadak-exif = { workspace = true }
 pdf = { workspace = true }
+pdfium-render = { workspace = true, optional = true }
+image = { workspace = true, optional = true }
 id3 = { workspace = true }
 tokio = { workspace = true }
 tokio-stream = { workspace = true }
@@ -40,4 +42,5 @@ utoipa-swagger-ui = { workspace = true, optional = true }
 
 [features]
 default = []
+pdfium = ["dep:pdfium-render", "dep:image"]
 axum = ["dep:axum", "dep:utoipa", "dep:utoipa-swagger-ui"]

crates/pile-dataset/build.rs

@@ -0,0 +1,54 @@
+use std::env;
+use std::path::PathBuf;
+
+const PDFIUM_URL: &str = "https://github.com/bblanchon/pdfium-binaries/releases/download/chromium%2F7725/pdfium-linux-x64.tgz";
+
+fn main() {
+	println!("cargo:rerun-if-changed=build.rs");
+
+	if env::var("CARGO_FEATURE_PDFIUM").is_err() {
+		return;
+	}
+
+	let out_dir = PathBuf::from(env::var("OUT_DIR").unwrap());
+
+	// OUT_DIR is target/<profile>/build/<pkg>-<hash>/out
+	// Go up 3 levels to reach target/<profile>/
+	let profile_dir = out_dir
+		.ancestors()
+		.nth(3)
+		.expect("unexpected OUT_DIR structure")
+		.to_path_buf();
+
+	let lib_path = profile_dir.join("libpdfium.so");
+
+	if !lib_path.exists() {
+		let tgz_path = out_dir.join("pdfium.tgz");
+
+		eprintln!("cargo:warning=Downloading PDFium from {PDFIUM_URL}");
+
+		let status = std::process::Command::new("curl")
+			.args(["-L", "--fail", "-o", tgz_path.to_str().unwrap(), PDFIUM_URL])
+			.status()
+			.expect("failed to run curl");
+		assert!(status.success(), "curl failed to download PDFium");
+
+		let status = std::process::Command::new("tar")
+			.args([
+				"-xzf",
+				tgz_path.to_str().unwrap(),
+				"-C",
+				out_dir.to_str().unwrap(),
+			])
+			.status()
+			.expect("failed to run tar");
+		assert!(status.success(), "tar failed to extract PDFium");
+
+		std::fs::copy(out_dir.join("lib").join("libpdfium.so"), &lib_path)
+			.expect("failed to copy libpdfium.so");
+	}
+
+	println!("cargo:rustc-link-search=native={}", profile_dir.display());
+	println!("cargo:rustc-link-lib=dylib=pdfium");
+	println!("cargo:rustc-link-arg=-Wl,-rpath,$ORIGIN");
+}

crates/pile-dataset/src/extract/pdf/mod.rs

@@ -1,6 +1,11 @@
 use pile_config::Label;
 use std::{collections::HashMap, sync::Arc};
 
+#[cfg(feature = "pdfium")]
+mod pdf_cover;
+#[cfg(feature = "pdfium")]
+pub use pdf_cover::*;
+
 mod pdf_meta;
 pub use pdf_meta::*;
 
@@ -19,20 +24,24 @@ pub struct PdfExtractor<'a> {
 impl<'a> PdfExtractor<'a> {
 	#[expect(clippy::unwrap_used)]
 	pub fn new(item: &'a Item) -> Self {
-		let inner = MapExtractor {
-			inner: HashMap::from([
-				(
-					Label::new("text").unwrap(),
-					PileValue::Extractor(Arc::new(PdfTextExtractor::new(item))),
-				),
-				(
-					Label::new("meta").unwrap(),
-					PileValue::Extractor(Arc::new(PdfMetaExtractor::new(item))),
-				),
-			]),
-		};
+		let mut inner_map = HashMap::new();
+		inner_map.insert(
+			Label::new("text").unwrap(),
+			PileValue::Extractor(Arc::new(PdfTextExtractor::new(item))),
+		);
+		inner_map.insert(
+			Label::new("meta").unwrap(),
+			PileValue::Extractor(Arc::new(PdfMetaExtractor::new(item))),
+		);
+		#[cfg(feature = "pdfium")]
+		inner_map.insert(
+			Label::new("cover").unwrap(),
+			PileValue::Extractor(Arc::new(PdfCoverExtractor::new(item))),
+		);
 
-		Self { inner }
+		Self {
+			inner: MapExtractor { inner: inner_map },
+		}
 	}
 }
 
@@ -50,6 +59,15 @@ impl Extractor for PdfExtractor<'_> {
 			};
 		}
 
+		#[cfg(feature = "pdfium")]
+		#[expect(clippy::unwrap_used)]
+		if name.as_str() == "cover" {
+			match self.inner.inner.get(name).unwrap() {
+				PileValue::Extractor(x) => return x.field(name).await,
+				_ => unreachable!(),
+			};
+		}
+
 		self.inner.field(name).await
 	}
 
@@ -58,6 +76,8 @@ impl Extractor for PdfExtractor<'_> {
 		Ok(vec![
 			Label::new("text").unwrap(),
 			Label::new("meta").unwrap(),
+			#[cfg(feature = "pdfium")]
+			Label::new("cover").unwrap(),
 		])
 	}
 }

crates/pile-dataset/src/extract/pdf/pdf_cover.rs

@@ -0,0 +1,98 @@
+use image::ImageFormat;
+use pdfium_render::prelude::*;
+use pile_config::Label;
+use std::{
+	collections::HashMap,
+	io::{BufReader, Cursor},
+	sync::{Arc, OnceLock},
+};
+use tracing::trace;
+
+use crate::{Item, PileValue, SyncReadBridge, extract::Extractor};
+
+pub struct PdfCoverExtractor<'a> {
+	item: &'a Item,
+	output: OnceLock<HashMap<Label, PileValue<'a>>>,
+}
+
+impl<'a> PdfCoverExtractor<'a> {
+	pub fn new(item: &'a Item) -> Self {
+		Self {
+			item,
+			output: OnceLock::new(),
+		}
+	}
+
+	async fn get_inner(&self) -> Result<&HashMap<Label, PileValue<'a>>, std::io::Error> {
+		if let Some(x) = self.output.get() {
+			return Ok(x);
+		}
+
+		let reader = SyncReadBridge::new_current(self.item.read().await?);
+		let cover = tokio::task::spawn_blocking(move || {
+			let mut bytes = Vec::new();
+			std::io::Read::read_to_end(&mut BufReader::new(reader), &mut bytes)?;
+
+			let pdfium = Pdfium::default();
+
+			let document = pdfium
+				.load_pdf_from_byte_slice(&bytes, None)
+				.map_err(|e| std::io::Error::new(std::io::ErrorKind::InvalidData, e.to_string()))?;
+
+			let render_config = PdfRenderConfig::new().set_target_width(1024);
+
+			let page = document
+				.pages()
+				.get(0)
+				.map_err(|e| std::io::Error::new(std::io::ErrorKind::InvalidData, e.to_string()))?;
+
+			let image = page
+				.render_with_config(&render_config)
+				.map_err(|e| std::io::Error::new(std::io::ErrorKind::InvalidData, e.to_string()))?
+				.as_image();
+
+			let mut png_bytes = Vec::new();
+			image
+				.write_to(&mut Cursor::new(&mut png_bytes), ImageFormat::Png)
+				.map_err(|e| std::io::Error::other(e.to_string()))?;
+
+			Ok::<_, std::io::Error>(png_bytes)
+		})
+		.await
+		.map_err(std::io::Error::other)?;
+
+		let output = match cover {
+			Ok(data) => {
+				#[expect(clippy::unwrap_used)]
+				let label = Label::new("cover").unwrap();
+				HashMap::from([(
+					label,
+					PileValue::Blob {
+						mime: mime::IMAGE_PNG,
+						bytes: Arc::new(data),
+					},
+				)])
+			}
+			Err(error) => {
+				trace!(message = "Could not render pdf cover", ?error, key = ?self.item.key());
+				HashMap::new()
+			}
+		};
+
+		return Ok(self.output.get_or_init(|| output));
+	}
+}
+
+#[async_trait::async_trait]
+impl Extractor for PdfCoverExtractor<'_> {
+	async fn field<'a>(
+		&'a self,
+		name: &Label,
+	) -> Result<Option<&'a PileValue<'a>>, std::io::Error> {
+		Ok(self.get_inner().await?.get(name))
+	}
+
+	async fn fields(&self) -> Result<Vec<Label>, std::io::Error> {
+		Ok(self.get_inner().await?.keys().cloned().collect())
+	}
+}

crates/pile/Cargo.toml

@@ -9,7 +9,7 @@ workspace = true
 
 [dependencies]
 pile-toolbox = { workspace = true }
-pile-dataset = { workspace = true, features = ["axum"] }
+pile-dataset = { workspace = true, features = ["axum", "pdfium"] }
 pile-config = { workspace = true }
 
 tracing = { workspace = true }