Add pdf extractors

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())
+	}
+}