Add exif extractor

crates/pile-dataset/Cargo.toml

@@ -23,6 +23,7 @@ thiserror = { workspace = true }
 smartstring = { workspace = true }
 blake3 = { workspace = true }
 epub = { workspace = true }
+kamadak-exif = { workspace = true }
 pdf = { workspace = true }
 id3 = { workspace = true }
 tokio = { workspace = true }

crates/pile-dataset/src/extract/exif.rs

@@ -0,0 +1,74 @@
+use pile_config::Label;
+use std::{collections::HashMap, io::Cursor, sync::OnceLock};
+use tracing::debug;
+
+use crate::{Item, PileValue, extract::Extractor};
+
+pub struct ExifExtractor<'a> {
+	item: &'a Item,
+	output: OnceLock<HashMap<Label, PileValue<'a>>>,
+}
+
+impl<'a> ExifExtractor<'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 bytes = self.item.read().await?.read_to_end().await?;
+		let mut cursor = Cursor::new(bytes);
+
+		let exif = match exif::Reader::new().read_from_container(&mut cursor) {
+			Ok(x) => x,
+			Err(error) => {
+				debug!(message = "Could not process exif", ?error, key = ?self.item.key());
+				return Ok(self.output.get_or_init(HashMap::new));
+			}
+		};
+
+		let mut output: HashMap<Label, PileValue<'a>> = HashMap::new();
+
+		for field in exif.fields() {
+			let Some(label) = tag_to_label(&field.tag) else {
+				continue;
+			};
+			// First occurrence wins (PRIMARY IFD comes before THUMBNAIL)
+			output.entry(label).or_insert_with(|| {
+				PileValue::String(field.display_value().with_unit(&exif).to_string().into())
+			});
+		}
+
+		return Ok(self.output.get_or_init(|| output));
+	}
+}
+
+fn tag_to_label(tag: &exif::Tag) -> Option<Label> {
+	let sanitized: String = tag
+		.to_string()
+		.chars()
+		.map(|c| if c == ' ' { '_' } else { c })
+		.filter(|c| Label::VALID_CHARS.contains(*c))
+		.collect();
+	Label::new(sanitized)
+}
+
+#[async_trait::async_trait]
+impl Extractor for ExifExtractor<'_> {
+	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-dataset/src/extract/mod.rs

@@ -13,6 +13,9 @@ pub use fs::*;
 mod epub;
 pub use epub::*;
 
+mod exif;
+pub use exif::*;
+
 mod pdf;
 pub use pdf::*;
 
@@ -76,6 +79,10 @@ impl<'a> MetaExtractor<'a> {
 					Label::new("epub").unwrap(),
 					crate::PileValue::Extractor(Arc::new(EpubExtractor::new(item))),
 				),
+				(
+					Label::new("exif").unwrap(),
+					crate::PileValue::Extractor(Arc::new(ExifExtractor::new(item))),
+				),
 				(
 					Label::new("pdf").unwrap(),
 					crate::PileValue::Extractor(Arc::new(PdfExtractor::new(item))),
@@ -111,6 +118,7 @@ impl Extractor for MetaExtractor<'_> {
 			Label::new("id3").unwrap(),
 			Label::new("fs").unwrap(),
 			Label::new("epub").unwrap(),
+			Label::new("exif").unwrap(),
 			Label::new("pdf").unwrap(),
 			Label::new("sidecar").unwrap(),
 		]);