Add pdf extractors

crates/pile-dataset/Cargo.toml

@@ -24,3 +24,4 @@ rayon = { workspace = true }
 smartstring = { workspace = true }
 blake3 = { workspace = true }
 toml_edit = { workspace = true }
+pdf = { workspace = true }

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

@@ -22,6 +22,11 @@ impl<'a> FlacExtractor<'a> {
 			return Ok(x);
 		}
 
+		// If this isn't a flac file, ignore it.
+		if self.item.path.extension().map(|x| x.to_str()).flatten() != Some("flac") {
+			return Ok(self.output.get_or_init(|| HashMap::new()));
+		}
+
 		let file = File::open(&self.item.path)?;
 		let reader = FlacReader::new(BufReader::new(file));
 

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

@@ -4,7 +4,7 @@ use std::collections::HashMap;
 use crate::{Item, PileValue, extract::Extractor};
 
 pub struct MapExtractor<'a, I: Item> {
-	pub(super) inner: HashMap<Label, PileValue<'a, I>>,
+	pub(crate) inner: HashMap<Label, PileValue<'a, I>>,
 }
 
 impl<I: Item> Extractor<I> for MapExtractor<'_, I> {

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

@@ -7,6 +7,9 @@ pub use flac::*;
 mod fs;
 pub use fs::*;
 
+mod pdf;
+pub use pdf::*;
+
 mod sidecar;
 pub use sidecar::*;
 
@@ -49,6 +52,10 @@ impl<'a> MetaExtractor<'a, crate::FileItem> {
 					Label::new("fs").unwrap(),
 					crate::PileValue::Extractor(Rc::new(FsExtractor::new(item))),
 				),
+				(
+					Label::new("pdf").unwrap(),
+					crate::PileValue::Extractor(Rc::new(PdfExtractor::new(item))),
+				),
 				(
 					Label::new("sidecar").unwrap(),
 					crate::PileValue::Extractor(Rc::new(SidecarExtractor::new(item))),
@@ -73,6 +80,7 @@ impl Extractor<crate::FileItem> for MetaExtractor<'_, crate::FileItem> {
 		return Ok(vec![
 			Label::new("flac").unwrap(),
 			Label::new("fs").unwrap(),
+			Label::new("pdf").unwrap(),
 			Label::new("sidecar").unwrap(),
 		]);
 	}

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

@@ -0,0 +1,61 @@
+use pile_config::Label;
+use std::{collections::HashMap, rc::Rc};
+
+mod pdf_meta;
+pub use pdf_meta::*;
+
+mod pdf_text;
+pub use pdf_text::*;
+
+use crate::{
+	FileItem, PileValue,
+	extract::{Extractor, MapExtractor},
+};
+
+pub struct PdfExtractor<'a> {
+	inner: MapExtractor<'a, FileItem>,
+}
+
+impl<'a> PdfExtractor<'a> {
+	#[expect(clippy::unwrap_used)]
+	pub fn new(item: &'a FileItem) -> Self {
+		let inner = MapExtractor {
+			inner: HashMap::from([
+				(
+					Label::new("text").unwrap(),
+					PileValue::Extractor(Rc::new(PdfTextExtractor::new(item))),
+				),
+				(
+					Label::new("meta").unwrap(),
+					PileValue::Extractor(Rc::new(PdfMetaExtractor::new(item))),
+				),
+			]),
+		};
+
+		Self { inner }
+	}
+}
+
+impl Extractor<FileItem> for PdfExtractor<'_> {
+	fn field<'a>(
+		&'a self,
+		name: &pile_config::Label,
+	) -> Result<Option<&'a PileValue<'a, FileItem>>, std::io::Error> {
+		if name.as_str() == "text" {
+			match self.inner.inner.get(name).unwrap() {
+				PileValue::Extractor(x) => return x.field(name),
+				_ => unreachable!(),
+			};
+		}
+
+		self.inner.field(name)
+	}
+
+	#[expect(clippy::unwrap_used)]
+	fn fields(&self) -> Result<Vec<Label>, std::io::Error> {
+		Ok(vec![
+			Label::new("text").unwrap(),
+			Label::new("meta").unwrap(),
+		])
+	}
+}

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

@@ -0,0 +1,98 @@
+use pdf::file::FileOptions;
+use pdf::primitive::{Date, TimeRel};
+use pile_config::Label;
+use std::{collections::HashMap, sync::OnceLock};
+
+use crate::{FileItem, PileValue, extract::Extractor};
+
+pub struct PdfMetaExtractor<'a> {
+	item: &'a FileItem,
+	output: OnceLock<HashMap<Label, PileValue<'a, FileItem>>>,
+}
+
+impl<'a> PdfMetaExtractor<'a> {
+	pub fn new(item: &'a FileItem) -> Self {
+		Self {
+			item,
+			output: OnceLock::new(),
+		}
+	}
+
+	fn get_inner(&self) -> Result<&HashMap<Label, PileValue<'a, FileItem>>, std::io::Error> {
+		if let Some(x) = self.output.get() {
+			return Ok(x);
+		}
+
+		let file = FileOptions::cached()
+			.open(&self.item.path)
+			.map_err(|e| std::io::Error::new(std::io::ErrorKind::InvalidData, e.to_string()))?;
+
+		let mut output: HashMap<Label, PileValue<'a, FileItem>> = HashMap::new();
+
+		if let Some(info) = &file.trailer.info_dict {
+			let fields: &[(&str, Option<&_>)] = &[
+				("title", info.title.as_ref()),
+				("author", info.author.as_ref()),
+				("subject", info.subject.as_ref()),
+				("keywords", info.keywords.as_ref()),
+				("creator", info.creator.as_ref()),
+				("producer", info.producer.as_ref()),
+			];
+
+			#[expect(clippy::unwrap_used)]
+			for (key, val) in fields {
+				let label = Label::new(*key).unwrap();
+				let value = match val {
+					Some(s) => PileValue::String(s.to_string_lossy().into()),
+					None => PileValue::Null,
+				};
+				output.insert(label, value);
+			}
+
+			#[expect(clippy::unwrap_used)]
+			{
+				output.insert(
+					Label::new("creation_date").unwrap(),
+					info.creation_date
+						.as_ref()
+						.map(|d| PileValue::String(format_date(d).into()))
+						.unwrap_or(PileValue::Null),
+				);
+				output.insert(
+					Label::new("mod_date").unwrap(),
+					info.mod_date
+						.as_ref()
+						.map(|d| PileValue::String(format_date(d).into()))
+						.unwrap_or(PileValue::Null),
+				);
+			}
+		}
+
+		return Ok(self.output.get_or_init(|| output));
+	}
+}
+
+fn format_date(d: &Date) -> String {
+	let tz = match d.rel {
+		TimeRel::Universal => "Z".to_owned(),
+		TimeRel::Later => format!("+{:02}:{:02}", d.tz_hour, d.tz_minute),
+		TimeRel::Earlier => format!("-{:02}:{:02}", d.tz_hour, d.tz_minute),
+	};
+	format!(
+		"{:04}-{:02}-{:02}T{:02}:{:02}:{:02}{}",
+		d.year, d.month, d.day, d.hour, d.minute, d.second, tz
+	)
+}
+
+impl Extractor<FileItem> for PdfMetaExtractor<'_> {
+	fn field<'a>(
+		&'a self,
+		name: &Label,
+	) -> Result<Option<&'a PileValue<'a, FileItem>>, std::io::Error> {
+		Ok(self.get_inner()?.get(name))
+	}
+
+	fn fields(&self) -> Result<Vec<Label>, std::io::Error> {
+		Ok(self.get_inner()?.keys().cloned().collect())
+	}
+}

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

@@ -0,0 +1,79 @@
+use pdf::content::{Op, TextDrawAdjusted};
+use pdf::file::FileOptions;
+use pile_config::Label;
+use std::{collections::HashMap, sync::OnceLock};
+
+use crate::{FileItem, PileValue, extract::Extractor};
+
+pub struct PdfTextExtractor<'a> {
+	item: &'a FileItem,
+	output: OnceLock<HashMap<Label, PileValue<'a, FileItem>>>,
+}
+
+impl<'a> PdfTextExtractor<'a> {
+	pub fn new(item: &'a FileItem) -> Self {
+		Self {
+			item,
+			output: OnceLock::new(),
+		}
+	}
+
+	fn get_inner(&self) -> Result<&HashMap<Label, PileValue<'a, FileItem>>, std::io::Error> {
+		if let Some(x) = self.output.get() {
+			return Ok(x);
+		}
+
+		let file = FileOptions::cached()
+			.open(&self.item.path)
+			.map_err(|e| std::io::Error::new(std::io::ErrorKind::InvalidData, e.to_string()))?;
+
+		let mut text_parts: Vec<String> = Vec::new();
+
+		for page in file.pages() {
+			let page = page
+				.map_err(|e| std::io::Error::new(std::io::ErrorKind::InvalidData, e.to_string()))?;
+
+			if let Some(content) = &page.contents {
+				let ops = content.operations(&file.resolver()).map_err(|e| {
+					std::io::Error::new(std::io::ErrorKind::InvalidData, e.to_string())
+				})?;
+
+				for op in ops {
+					match op {
+						Op::TextDraw { text } => {
+							text_parts.push(text.to_string_lossy());
+						}
+						Op::TextDrawAdjusted { array } => {
+							for item in array {
+								if let TextDrawAdjusted::Text(text) = item {
+									text_parts.push(text.to_string_lossy());
+								}
+							}
+						}
+						_ => {}
+					}
+				}
+			}
+		}
+
+		let text = text_parts.join(" ");
+
+		#[expect(clippy::unwrap_used)]
+		let output = HashMap::from([(Label::new("text").unwrap(), PileValue::String(text.into()))]);
+
+		return Ok(self.output.get_or_init(|| output));
+	}
+}
+
+impl Extractor<FileItem> for PdfTextExtractor<'_> {
+	fn field<'a>(
+		&'a self,
+		name: &Label,
+	) -> Result<Option<&'a PileValue<'a, FileItem>>, std::io::Error> {
+		Ok(self.get_inner()?.get(name))
+	}
+
+	fn fields(&self) -> Result<Vec<Label>, std::io::Error> {
+		Ok(self.get_inner()?.keys().cloned().collect())
+	}
+}