Add epub extractor

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

@@ -0,0 +1,63 @@
+use pile_config::Label;
+use std::{collections::HashMap, sync::Arc};
+
+mod epub_meta;
+pub use epub_meta::*;
+
+mod epub_text;
+pub use epub_text::*;
+
+use crate::{
+	Item, PileValue,
+	extract::{Extractor, MapExtractor},
+};
+
+pub struct EpubExtractor<'a> {
+	inner: MapExtractor<'a>,
+}
+
+impl<'a> EpubExtractor<'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(EpubTextExtractor::new(item))),
+				),
+				(
+					Label::new("meta").unwrap(),
+					PileValue::Extractor(Arc::new(EpubMetaExtractor::new(item))),
+				),
+			]),
+		};
+
+		Self { inner }
+	}
+}
+
+#[async_trait::async_trait]
+impl Extractor for EpubExtractor<'_> {
+	async fn field<'a>(
+		&'a self,
+		name: &pile_config::Label,
+	) -> Result<Option<&'a PileValue<'a>>, std::io::Error> {
+		#[expect(clippy::unwrap_used)]
+		if name.as_str() == "text" {
+			match self.inner.inner.get(name).unwrap() {
+				PileValue::Extractor(x) => return x.field(name).await,
+				_ => unreachable!(),
+			};
+		}
+
+		self.inner.field(name).await
+	}
+
+	#[expect(clippy::unwrap_used)]
+	async fn fields(&self) -> Result<Vec<Label>, std::io::Error> {
+		Ok(vec![
+			Label::new("text").unwrap(),
+			Label::new("meta").unwrap(),
+		])
+	}
+}