Filter by mime

crates/pile-dataset/src/dataset.rs

@@ -2,6 +2,7 @@ use chrono::{DateTime, Utc};
 use pile_config::{ConfigToml, Label, Source, objectpath::ObjectPath};
 use pile_toolbox::cancelabletask::{CancelFlag, CancelableTaskError};
 use pile_value::{
+	extract::traits::ExtractState,
 	source::{DataSource, DirDataSource, S3DataSource, misc::path_ts_earliest},
 	value::{Item, PileValue},
 };
@@ -187,6 +188,7 @@ impl Datasets {
 	/// Returns `None` if the item or field is not found.
 	pub async fn get_field(
 		&self,
+		state: &ExtractState,
 		source: &Label,
 		key: &str,
 		path: &ObjectPath,
@@ -196,11 +198,11 @@ impl Datasets {
 		};
 
 		let item = PileValue::Item(item);
-		let Some(value) = item.query(path).await? else {
+		let Some(value) = item.query(state, path).await? else {
 			return Ok(None);
 		};
 
-		Ok(Some(value.to_json().await?))
+		Ok(Some(value.to_json(state).await?))
 	}
 
 	//
@@ -210,6 +212,7 @@ impl Datasets {
 	/// Refresh this dataset's fts index.
 	pub async fn fts_refresh(
 		&self,
+		state: &ExtractState,
 		_threads: usize,
 		flag: Option<CancelFlag>,
 	) -> Result<(), CancelableTaskError<DatasetError>> {
@@ -265,9 +268,10 @@ impl Datasets {
 
 				let item = item_result.map_err(DatasetError::from)?;
 				let db = Arc::clone(&db_index);
+				let state = state.clone();
 				join_set.spawn(async move {
 					let key = item.key();
-					let result = db.entry_to_document(&item).await;
+					let result = db.entry_to_document(&state, &item).await;
 					(key, result)
 				});
 

crates/pile-dataset/src/index/index_fts.rs

@@ -1,5 +1,8 @@
 use pile_config::{ConfigToml, DatasetFts, Label};
-use pile_value::value::{Item, PileValue};
+use pile_value::{
+	extract::traits::ExtractState,
+	value::{Item, PileValue},
+};
 use std::{path::PathBuf, sync::LazyLock};
 use tantivy::{
 	DocAddress, Index, ReloadPolicy, TantivyDocument, TantivyError,
@@ -63,6 +66,7 @@ impl DbFtsIndex {
 	/// Turn an entry into a tantivy document
 	pub async fn entry_to_document(
 		&self,
+		state: &ExtractState,
 		item: &Item,
 	) -> Result<Option<TantivyDocument>, TantivyError> {
 		let mut doc = TantivyDocument::default();
@@ -75,7 +79,7 @@ impl DbFtsIndex {
 
 		let mut empty = true;
 		for name in self.fts_cfg().fields.keys() {
-			let x = self.get_field(&item, name).await?;
+			let x = self.get_field(state, &item, name).await?;
 
 			let val = match x {
 				Some(x) => x,
@@ -99,6 +103,7 @@ impl DbFtsIndex {
 
 	pub async fn get_field(
 		&self,
+		state: &ExtractState,
 		extractor: &PileValue,
 		field_name: &Label,
 	) -> Result<Option<String>, std::io::Error> {
@@ -112,7 +117,7 @@ impl DbFtsIndex {
 
 		// Try paths in order, using the first value we find
 		'outer: for path in field.path.as_slice() {
-			let val = match extractor.query(path).await? {
+			let val = match extractor.query(state, path).await? {
 				Some(x) => x,
 				None => return Ok(None),
 			};

crates/pile-dataset/src/serve/field.rs

@@ -5,7 +5,7 @@ use axum::{
 	response::{IntoResponse, Response},
 };
 use pile_config::{Label, objectpath::ObjectPath};
-use pile_value::value::PileValue;
+use pile_value::{extract::traits::ExtractState, value::PileValue};
 use serde::Deserialize;
 use std::{sync::Arc, time::Instant};
 use tracing::debug;
@@ -62,8 +62,10 @@ pub async fn get_field(
 		return StatusCode::NOT_FOUND.into_response();
 	};
 
+	let state = ExtractState { ignore_mime: false };
+
 	let item = PileValue::Item(item);
-	let value = match item.query(&path).await {
+	let value = match item.query(&state, &path).await {
 		Ok(Some(v)) => v,
 		Ok(None) => return StatusCode::NOT_FOUND.into_response(),
 		Err(e) => return (StatusCode::INTERNAL_SERVER_ERROR, format!("{e:?}")).into_response(),
@@ -90,7 +92,7 @@ pub async fn get_field(
 			bytes.as_ref().clone(),
 		)
 			.into_response(),
-		_ => match value.to_json().await {
+		_ => match value.to_json(&state).await {
 			Ok(json) => (StatusCode::OK, Json(json)).into_response(),
 			Err(e) => (StatusCode::INTERNAL_SERVER_ERROR, format!("{e:?}")).into_response(),
 		},

crates/pile-value/src/extract/item/epub/epub_meta.rs

@@ -7,7 +7,7 @@ use std::{
 use tracing::trace;
 
 use crate::{
-	extract::traits::ObjectExtractor,
+	extract::traits::{ExtractState, ObjectExtractor},
 	value::{Item, PileValue, SyncReadBridge},
 };
 
@@ -82,6 +82,7 @@ impl EpubMetaExtractor {
 impl ObjectExtractor for EpubMetaExtractor {
 	async fn field(
 		&self,
+		state: &ExtractState,
 		name: &Label,
 		args: Option<&str>,
 	) -> Result<Option<PileValue>, std::io::Error> {
@@ -89,6 +90,10 @@ impl ObjectExtractor for EpubMetaExtractor {
 			return Ok(None);
 		}
 
+		if !state.ignore_mime && self.item.mime().essence_str() != "application/epub+zip" {
+			return Ok(None);
+		}
+
 		Ok(self.get_inner().await?.get(name).cloned())
 	}
 

crates/pile-value/src/extract/item/epub/epub_text.rs

@@ -7,7 +7,7 @@ use std::{
 use tracing::trace;
 
 use crate::{
-	extract::traits::ObjectExtractor,
+	extract::traits::{ExtractState, ObjectExtractor},
 	value::{Item, PileValue, SyncReadBridge},
 };
 
@@ -92,6 +92,7 @@ fn strip_html(html: &str) -> String {
 impl ObjectExtractor for EpubTextExtractor {
 	async fn field(
 		&self,
+		state: &ExtractState,
 		name: &Label,
 		args: Option<&str>,
 	) -> Result<Option<PileValue>, std::io::Error> {
@@ -99,6 +100,10 @@ impl ObjectExtractor for EpubTextExtractor {
 			return Ok(None);
 		}
 
+		if !state.ignore_mime && self.item.mime().essence_str() != "application/epub+zip" {
+			return Ok(None);
+		}
+
 		Ok(self.get_inner().await?.get(name).cloned())
 	}
 

crates/pile-value/src/extract/item/epub/mod.rs

@@ -8,7 +8,7 @@ mod epub_text;
 pub use epub_text::*;
 
 use crate::{
-	extract::traits::ObjectExtractor,
+	extract::traits::{ExtractState, ObjectExtractor},
 	value::{Item, PileValue},
 };
 
@@ -30,13 +30,14 @@ impl EpubExtractor {
 impl ObjectExtractor for EpubExtractor {
 	async fn field(
 		&self,
+		state: &ExtractState,
 		name: &pile_config::Label,
 		args: Option<&str>,
 	) -> Result<Option<PileValue>, std::io::Error> {
 		match (name.as_str(), args) {
 			("text", args) => Ok(Some(
 				self.text
-					.field(name, args)
+					.field(state, name, args)
 					.await
 					.map(|x| x.unwrap_or(PileValue::Null))?,
 			)),

crates/pile-value/src/extract/item/exif.rs

@@ -7,7 +7,7 @@ use std::{
 use tracing::trace;
 
 use crate::{
-	extract::traits::ObjectExtractor,
+	extract::traits::{ExtractState, ObjectExtractor},
 	value::{Item, PileValue, SyncReadBridge},
 };
 
@@ -87,6 +87,7 @@ fn tag_to_label(tag: &str) -> Option<Label> {
 impl ObjectExtractor for ExifExtractor {
 	async fn field(
 		&self,
+		state: &ExtractState,
 		name: &Label,
 		args: Option<&str>,
 	) -> Result<Option<PileValue>, std::io::Error> {
@@ -100,6 +101,10 @@ impl ObjectExtractor for ExifExtractor {
 			return Ok(None);
 		}
 
+		if !state.ignore_mime && self.item.mime().type_() != mime::IMAGE {
+			return Ok(None);
+		}
+
 		Ok(self.get_inner().await?.get(name).cloned())
 	}
 

crates/pile-value/src/extract/item/flac.rs

@@ -9,7 +9,7 @@ use std::{
 use tracing::trace;
 
 use crate::{
-	extract::traits::{ListExtractor, ObjectExtractor},
+	extract::traits::{ExtractState, ListExtractor, ObjectExtractor},
 	value::{Item, PileValue, SyncReadBridge},
 };
 
@@ -46,16 +46,32 @@ impl FlacImagesExtractor {
 
 		return Ok(count);
 	}
+
+	fn mime_ok(&self, state: &ExtractState) -> bool {
+		if state.ignore_mime {
+			return true;
+		}
+		let essence = self.item.mime().essence_str();
+		essence == "audio/flac" || essence == "audio/x-flac"
+	}
 }
 
 #[async_trait::async_trait]
 impl ListExtractor for FlacImagesExtractor {
-	async fn get<'a>(&'a self, mut idx: usize) -> Result<Option<PileValue>, std::io::Error> {
+	async fn get(
+		&self,
+		state: &ExtractState,
+		mut idx: usize,
+	) -> Result<Option<PileValue>, std::io::Error> {
 		trace!(
 			key = self.item.key().as_str(),
 			"Getting index {idx} from FlacImagesExtractor",
 		);
 
+		if !self.mime_ok(state) {
+			return Ok(None);
+		}
+
 		let key = self.item.key();
 		let reader = SyncReadBridge::new_current(self.item.read().await?);
 		let image = tokio::task::spawn_blocking(move || {
@@ -98,7 +114,11 @@ impl ListExtractor for FlacImagesExtractor {
 		}))
 	}
 
-	async fn len(&self) -> Result<usize, std::io::Error> {
+	async fn len(&self, state: &ExtractState) -> Result<usize, std::io::Error> {
+		if !self.mime_ok(state) {
+			return Ok(0);
+		}
+
 		if let Some(x) = self.cached_count.get() {
 			return Ok(*x);
 		}
@@ -178,12 +198,21 @@ impl FlacExtractor {
 
 		return Ok(self.output.get_or_init(|| output));
 	}
+
+	fn mime_ok(&self, state: &ExtractState) -> bool {
+		if state.ignore_mime {
+			return true;
+		}
+		let essence = self.item.mime().essence_str();
+		essence == "audio/flac" || essence == "audio/x-flac"
+	}
 }
 
 #[async_trait::async_trait]
 impl ObjectExtractor for FlacExtractor {
 	async fn field(
 		&self,
+		state: &ExtractState,
 		name: &Label,
 		args: Option<&str>,
 	) -> Result<Option<PileValue>, std::io::Error> {
@@ -191,6 +220,10 @@ impl ObjectExtractor for FlacExtractor {
 			return Ok(None);
 		}
 
+		if !self.mime_ok(state) {
+			return Ok(None);
+		}
+
 		if name.as_str() == "images" {
 			return Ok(Some(self.images.clone()));
 		}

crates/pile-value/src/extract/item/fs.rs

@@ -1,5 +1,5 @@
 use crate::{
-	extract::traits::ObjectExtractor,
+	extract::traits::{ExtractState, ObjectExtractor},
 	value::{Item, PileValue},
 };
 use pile_config::Label;
@@ -88,6 +88,7 @@ impl FsExtractor {
 impl ObjectExtractor for FsExtractor {
 	async fn field(
 		&self,
+		_state: &ExtractState,
 		name: &Label,
 		args: Option<&str>,
 	) -> Result<Option<PileValue>, std::io::Error> {

crates/pile-value/src/extract/item/id3.rs

@@ -9,7 +9,7 @@ use std::{
 use tracing::trace;
 
 use crate::{
-	extract::traits::ObjectExtractor,
+	extract::traits::{ExtractState, ObjectExtractor},
 	value::{Item, PileValue, SyncReadBridge},
 };
 
@@ -125,6 +125,7 @@ fn frame_id_to_field(id: &str) -> Cow<'static, str> {
 impl ObjectExtractor for Id3Extractor {
 	async fn field(
 		&self,
+		state: &ExtractState,
 		name: &Label,
 		args: Option<&str>,
 	) -> Result<Option<PileValue>, std::io::Error> {
@@ -132,6 +133,10 @@ impl ObjectExtractor for Id3Extractor {
 			return Ok(None);
 		}
 
+		if !state.ignore_mime && self.item.mime().essence_str() != "audio/mpeg" {
+			return Ok(None);
+		}
+
 		Ok(self.get_inner().await?.get(name).cloned())
 	}
 

crates/pile-value/src/extract/item/mod.rs

@@ -26,7 +26,10 @@ mod sidecar;
 pub use sidecar::*;
 
 use crate::{
-	extract::{misc::MapExtractor, traits::ObjectExtractor},
+	extract::{
+		misc::MapExtractor,
+		traits::{ExtractState, ObjectExtractor},
+	},
 	value::{Item, PileValue},
 };
 
@@ -82,10 +85,11 @@ impl ItemExtractor {
 impl ObjectExtractor for ItemExtractor {
 	async fn field(
 		&self,
+		state: &ExtractState,
 		name: &pile_config::Label,
 		args: Option<&str>,
 	) -> Result<Option<PileValue>, std::io::Error> {
-		self.inner.field(name, args).await
+		self.inner.field(state, name, args).await
 	}
 
 	#[expect(clippy::unwrap_used)]

crates/pile-value/src/extract/item/pdf/mod.rs

@@ -14,7 +14,7 @@ mod pdf_text;
 pub use pdf_text::*;
 
 use crate::{
-	extract::traits::ObjectExtractor,
+	extract::traits::{ExtractState, ObjectExtractor},
 	value::{Item, PileValue},
 };
 
@@ -40,6 +40,7 @@ impl PdfExtractor {
 impl ObjectExtractor for PdfExtractor {
 	async fn field(
 		&self,
+		state: &ExtractState,
 		name: &pile_config::Label,
 		args: Option<&str>,
 	) -> Result<Option<PileValue>, std::io::Error> {
@@ -50,7 +51,7 @@ impl ObjectExtractor for PdfExtractor {
 		);
 
 		match (name.as_str(), args) {
-			("text", args) => self.text.field(name, args).await,
+			("text", args) => self.text.field(state, name, args).await,
 			("meta", None) => Ok(Some(PileValue::ObjectExtractor(self.meta.clone()))),
 			#[cfg(feature = "pdfium")]
 			("pages", None) => Ok(Some(PileValue::ListExtractor(self.pages.clone()))),

crates/pile-value/src/extract/item/pdf/pdf_meta.rs

@@ -9,7 +9,7 @@ use std::{
 use tracing::trace;
 
 use crate::{
-	extract::traits::ObjectExtractor,
+	extract::traits::{ExtractState, ObjectExtractor},
 	value::{Item, PileValue, SyncReadBridge},
 };
 
@@ -124,12 +124,18 @@ fn format_date(d: &Date) -> String {
 impl ObjectExtractor for PdfMetaExtractor {
 	async fn field(
 		&self,
+		state: &ExtractState,
 		name: &Label,
 		args: Option<&str>,
 	) -> Result<Option<PileValue>, std::io::Error> {
 		if args.is_some() {
 			return Ok(None);
 		}
+
+		if !state.ignore_mime && self.item.mime().essence_str() != "application/pdf" {
+			return Ok(None);
+		}
+
 		Ok(self.get_inner().await?.get(name).cloned())
 	}
 

crates/pile-value/src/extract/item/pdf/pdf_pages.rs

@@ -7,7 +7,7 @@ use std::{
 use tracing::trace;
 
 use crate::{
-	extract::traits::ListExtractor,
+	extract::traits::{ExtractState, ListExtractor},
 	value::{Item, PileValue, SyncReadBridge},
 };
 
@@ -34,12 +34,20 @@ impl PdfPagesExtractor {
 
 #[async_trait::async_trait]
 impl ListExtractor for PdfPagesExtractor {
-	async fn get(&self, idx: usize) -> Result<Option<PileValue>, std::io::Error> {
+	async fn get(
+		&self,
+		state: &ExtractState,
+		idx: usize,
+	) -> Result<Option<PileValue>, std::io::Error> {
 		trace!(
 			key = self.item.key().as_str(),
 			"Getting index {idx} from PdfPagesExtractor",
 		);
 
+		if !state.ignore_mime && self.item.mime().essence_str() != "application/pdf" {
+			return Ok(None);
+		}
+
 		let bytes = self.get_bytes().await?;
 		let png = tokio::task::spawn_blocking(move || {
 			let pdfium = Pdfium::default();
@@ -81,7 +89,11 @@ impl ListExtractor for PdfPagesExtractor {
 		Ok(Some(value))
 	}
 
-	async fn len(&self) -> Result<usize, std::io::Error> {
+	async fn len(&self, state: &ExtractState) -> Result<usize, std::io::Error> {
+		if !state.ignore_mime && self.item.mime().essence_str() != "application/pdf" {
+			return Ok(0);
+		}
+
 		let bytes = self.get_bytes().await?;
 		let count = tokio::task::spawn_blocking(move || {
 			let pdfium = Pdfium::default();
@@ -103,10 +115,10 @@ impl ListExtractor for PdfPagesExtractor {
 
 	// Override, extracting all pages is very slow,
 	// and we can't display binary in json anyway
-	async fn to_json(&self) -> Result<serde_json::Value, std::io::Error> {
+	async fn to_json(&self, state: &ExtractState) -> Result<serde_json::Value, std::io::Error> {
 		Ok(serde_json::Value::String(format!(
 			"<PdfPages ({} pages)>",
-			self.len().await?
+			self.len(state).await?
 		)))
 	}
 }

crates/pile-value/src/extract/item/pdf/pdf_text.rs

@@ -9,7 +9,7 @@ use std::{
 use tracing::trace;
 
 use crate::{
-	extract::traits::ObjectExtractor,
+	extract::traits::{ExtractState, ObjectExtractor},
 	value::{Item, PileValue, SyncReadBridge},
 };
 
@@ -104,6 +104,7 @@ impl PdfTextExtractor {
 impl ObjectExtractor for PdfTextExtractor {
 	async fn field(
 		&self,
+		state: &ExtractState,
 		name: &Label,
 		args: Option<&str>,
 	) -> Result<Option<PileValue>, std::io::Error> {
@@ -111,6 +112,10 @@ impl ObjectExtractor for PdfTextExtractor {
 			return Ok(None);
 		}
 
+		if !state.ignore_mime && self.item.mime().essence_str() != "application/pdf" {
+			return Ok(None);
+		}
+
 		Ok(self.get_inner().await?.get(name).cloned())
 	}
 

crates/pile-value/src/extract/item/sidecar.rs

@@ -4,7 +4,7 @@ use tracing::trace;
 
 use super::TomlExtractor;
 use crate::{
-	extract::traits::ObjectExtractor,
+	extract::traits::{ExtractState, ObjectExtractor},
 	value::{Item, PileValue},
 };
 
@@ -26,6 +26,7 @@ impl SidecarExtractor {
 impl ObjectExtractor for SidecarExtractor {
 	async fn field(
 		&self,
+		state: &ExtractState,
 		name: &Label,
 		args: Option<&str>,
 	) -> Result<Option<PileValue>, std::io::Error> {
@@ -39,7 +40,7 @@ impl ObjectExtractor for SidecarExtractor {
 			.output
 			.get_or_init(|| self.item.sidecar().map(TomlExtractor::new))
 		{
-			Some(x) => Ok(x.field(name, args).await?),
+			Some(x) => Ok(x.field(state, name, args).await?),
 			None => Ok(Some(PileValue::Null)),
 		}
 	}

crates/pile-value/src/extract/item/toml.rs

@@ -5,7 +5,7 @@ use std::{
 };
 
 use crate::{
-	extract::traits::ObjectExtractor,
+	extract::traits::{ExtractState, ObjectExtractor},
 	value::{AsyncReader, Item, PileValue},
 };
 
@@ -64,6 +64,7 @@ impl TomlExtractor {
 impl ObjectExtractor for TomlExtractor {
 	async fn field(
 		&self,
+		state: &ExtractState,
 		name: &Label,
 		args: Option<&str>,
 	) -> Result<Option<PileValue>, std::io::Error> {
@@ -71,6 +72,10 @@ impl ObjectExtractor for TomlExtractor {
 			return Ok(None);
 		}
 
+		if !state.ignore_mime && self.item.mime().type_() != mime::TEXT {
+			return Ok(None);
+		}
+
 		Ok(self.get_inner().await?.get(name).cloned())
 	}
 

crates/pile-value/src/extract/misc/list.rs

@@ -1,6 +1,9 @@
 use std::sync::Arc;
 
-use crate::{extract::traits::ListExtractor, value::PileValue};
+use crate::{
+	extract::traits::{ExtractState, ListExtractor},
+	value::PileValue,
+};
 
 pub struct ArrayExtractor {
 	inner: Arc<Vec<PileValue>>,
@@ -14,11 +17,15 @@ impl ArrayExtractor {
 
 #[async_trait::async_trait]
 impl ListExtractor for ArrayExtractor {
-	async fn get(&self, idx: usize) -> Result<Option<PileValue>, std::io::Error> {
+	async fn get(
+		&self,
+		_state: &ExtractState,
+		idx: usize,
+	) -> Result<Option<PileValue>, std::io::Error> {
 		Ok(self.inner.get(idx).cloned())
 	}
 
-	async fn len(&self) -> Result<usize, std::io::Error> {
+	async fn len(&self, _state: &ExtractState) -> Result<usize, std::io::Error> {
 		Ok(self.inner.len())
 	}
 }

crates/pile-value/src/extract/misc/map.rs

@@ -1,7 +1,10 @@
 use pile_config::Label;
 use std::collections::HashMap;
 
-use crate::{extract::traits::ObjectExtractor, value::PileValue};
+use crate::{
+	extract::traits::{ExtractState, ObjectExtractor},
+	value::PileValue,
+};
 
 #[derive(Default)]
 pub struct MapExtractor {
@@ -12,6 +15,7 @@ pub struct MapExtractor {
 impl ObjectExtractor for MapExtractor {
 	async fn field(
 		&self,
+		_state: &ExtractState,
 		name: &Label,
 		args: Option<&str>,
 	) -> Result<Option<PileValue>, std::io::Error> {

crates/pile-value/src/extract/misc/vec.rs

@@ -1,4 +1,7 @@
-use crate::{extract::traits::ListExtractor, value::PileValue};
+use crate::{
+	extract::traits::{ExtractState, ListExtractor},
+	value::PileValue,
+};
 
 #[derive(Default)]
 pub struct VecExtractor {
@@ -7,11 +10,15 @@ pub struct VecExtractor {
 
 #[async_trait::async_trait]
 impl ListExtractor for VecExtractor {
-	async fn get(&self, idx: usize) -> Result<Option<PileValue>, std::io::Error> {
+	async fn get(
+		&self,
+		_state: &ExtractState,
+		idx: usize,
+	) -> Result<Option<PileValue>, std::io::Error> {
 		Ok(self.inner.get(idx).cloned())
 	}
 
-	async fn len(&self) -> Result<usize, std::io::Error> {
+	async fn len(&self, _state: &ExtractState) -> Result<usize, std::io::Error> {
 		Ok(self.inner.len())
 	}
 }

crates/pile-value/src/extract/string.rs

@@ -2,7 +2,10 @@ use pile_config::Label;
 use smartstring::{LazyCompact, SmartString};
 use std::sync::Arc;
 
-use crate::{extract::traits::ObjectExtractor, value::PileValue};
+use crate::{
+	extract::traits::{ExtractState, ObjectExtractor},
+	value::PileValue,
+};
 
 pub struct StringExtractor {
 	item: Arc<SmartString<LazyCompact>>,
@@ -18,6 +21,7 @@ impl StringExtractor {
 impl ObjectExtractor for StringExtractor {
 	async fn field(
 		&self,
+		_state: &ExtractState,
 		name: &Label,
 		args: Option<&str>,
 	) -> Result<Option<PileValue>, std::io::Error> {
@@ -89,7 +93,10 @@ mod tests {
 
 	#[expect(clippy::unwrap_used)]
 	async fn field(ext: &StringExtractor, name: &str, args: Option<&str>) -> Option<PileValue> {
-		ext.field(&Label::new(name).unwrap(), args).await.unwrap()
+		let state = ExtractState { ignore_mime: false };
+		ext.field(&state, &Label::new(name).unwrap(), args)
+			.await
+			.unwrap()
 	}
 
 	fn string(v: Option<PileValue>) -> Option<String> {

crates/pile-value/src/extract/traits.rs

@@ -1,3 +1,13 @@
+#[derive(Debug, Clone)]
+pub struct ExtractState {
+	/// If true, extract all fields from all items.
+	/// Do not pre-filter using mime type.
+	///
+	/// This may detect additional fields, but
+	/// makes extraction take much longer
+	pub ignore_mime: bool,
+}
+
 /// An attachment that extracts metadata from an [Item].
 ///
 /// Metadata is exposed as an immutable map of {label: value},
@@ -15,6 +25,7 @@ pub trait ObjectExtractor: Send + Sync {
 	/// this fn should return `Ok(Some(None))`.
 	async fn field(
 		&self,
+		state: &ExtractState,
 		name: &pile_config::Label,
 		args: Option<&str>,
 	) -> Result<Option<crate::value::PileValue>, std::io::Error>;
@@ -25,15 +36,15 @@ pub trait ObjectExtractor: Send + Sync {
 	async fn fields(&self) -> Result<Vec<pile_config::Label>, std::io::Error>;
 
 	/// Convert this to a JSON value.
-	async fn to_json(&self) -> Result<serde_json::Value, std::io::Error> {
+	async fn to_json(&self, state: &ExtractState) -> Result<serde_json::Value, std::io::Error> {
 		let keys = self.fields().await?;
 		let mut map = serde_json::Map::new();
 		for k in &keys {
-			let v = match self.field(k, None).await? {
+			let v = match self.field(state, k, None).await? {
 				Some(x) => x,
 				None => continue,
 			};
-			map.insert(k.to_string(), Box::pin(v.to_json()).await?);
+			map.insert(k.to_string(), Box::pin(v.to_json(state)).await?);
 		}
 
 		Ok(serde_json::Value::Object(map))
@@ -49,25 +60,25 @@ pub trait ListExtractor: Send + Sync {
 	/// Indices start at zero, and must be consecutive.
 	/// - returns `None` if `idx` is out of range
 	/// - returns `Some(Null)` if `None` is at `idx`
-	async fn get(&self, idx: usize) -> Result<Option<crate::value::PileValue>, std::io::Error>;
+	async fn get(
+		&self,
+		state: &ExtractState,
+		idx: usize,
+	) -> Result<Option<crate::value::PileValue>, std::io::Error>;
 
-	async fn len(&self) -> Result<usize, std::io::Error>;
+	async fn len(&self, state: &ExtractState) -> Result<usize, std::io::Error>;
-
-	async fn is_empty(&self) -> Result<bool, std::io::Error> {
-		Ok(self.len().await? == 0)
-	}
 
 	/// Convert this list to a JSON value.
-	async fn to_json(&self) -> Result<serde_json::Value, std::io::Error> {
+	async fn to_json(&self, state: &ExtractState) -> Result<serde_json::Value, std::io::Error> {
-		let len = self.len().await?;
+		let len = self.len(state).await?;
 		let mut list = Vec::with_capacity(len);
 		for i in 0..len {
 			#[expect(clippy::expect_used)]
 			let v = self
-				.get(i)
+				.get(state, i)
 				.await?
 				.expect("value must be present according to length");
-			list.push(Box::pin(v.to_json()).await?);
+			list.push(Box::pin(v.to_json(state)).await?);
 		}
 
 		Ok(serde_json::Value::Array(list))

crates/pile-value/src/value/value.rs

@@ -9,7 +9,7 @@ use crate::{
 		item::ItemExtractor,
 		misc::{ArrayExtractor, MapExtractor, VecExtractor},
 		string::StringExtractor,
-		traits::{ListExtractor, ObjectExtractor},
+		traits::{ExtractState, ListExtractor, ObjectExtractor},
 	},
 	value::Item,
 };
@@ -91,7 +91,11 @@ impl PileValue {
 		}
 	}
 
-	pub async fn query(&self, query: &ObjectPath) -> Result<Option<Self>, std::io::Error> {
+	pub async fn query(
+		&self,
+		state: &ExtractState,
+		query: &ObjectPath,
+	) -> Result<Option<Self>, std::io::Error> {
 		let mut out: Option<PileValue> = Some(self.clone());
 
 		for s in &query.segments {
@@ -106,7 +110,7 @@ impl PileValue {
 						}
 					};
 
-					out = e.field(name, args.as_deref()).await?;
+					out = e.field(state, name, args.as_deref()).await?;
 				}
 
 				PathSegment::Index(idx) => {
@@ -121,7 +125,7 @@ impl PileValue {
 					let idx = if *idx >= 0 {
 						usize::try_from(*idx).ok()
 					} else {
-						usize::try_from(e.len().await? as i64 - idx).ok()
+						usize::try_from(e.len(state).await? as i64 - idx).ok()
 					};
 
 					let idx = match idx {
@@ -132,7 +136,7 @@ impl PileValue {
 						}
 					};
 
-					out = e.get(idx).await?;
+					out = e.get(state, idx).await?;
 				}
 			}
 		}
@@ -147,7 +151,10 @@ impl PileValue {
 	/// - `ObjectExtractor` is recursed into; returns `Some(Object(map))` with
 	///   only the fields that had data, or `None` if all fields were absent.
 	/// - `Array` / `ListExtractor` are treated as opaque leaf values (not descended into).
-	pub async fn count_fields(&self) -> Result<Option<Value>, std::io::Error> {
+	pub async fn count_fields(
+		&self,
+		state: &ExtractState,
+	) -> Result<Option<Value>, std::io::Error> {
 		Ok(match self {
 			Self::Null => None,
 
@@ -156,18 +163,18 @@ impl PileValue {
 			}
 
 			Self::Array(x) => (!x.is_empty()).then(|| Value::Number(1u64.into())),
-			Self::ListExtractor(x) => (!x.is_empty().await?).then(|| Value::Number(1u64.into())),
+			Self::ListExtractor(x) => (x.len(state).await? > 0).then(|| Value::Number(1u64.into())),
 
 			Self::ObjectExtractor(_) | Self::Item(_) => {
 				let e = self.object_extractor();
 				let keys = e.fields().await?;
 				let mut map = Map::new();
 				for k in &keys {
-					let v = match e.field(k, None).await? {
+					let v = match e.field(state, k, None).await? {
 						Some(x) => x,
 						None => continue,
 					};
-					if let Some(counted) = Box::pin(v.count_fields()).await? {
+					if let Some(counted) = Box::pin(v.count_fields(state)).await? {
 						map.insert(k.to_string(), counted);
 					}
 				}
@@ -187,7 +194,7 @@ impl PileValue {
 		}
 	}
 
-	pub async fn to_json(&self) -> Result<Value, std::io::Error> {
+	pub async fn to_json(&self, state: &ExtractState) -> Result<Value, std::io::Error> {
 		Ok(match self {
 			Self::Null => Value::Null,
 			Self::U64(x) => Value::Number((*x).into()),
@@ -201,12 +208,12 @@ impl PileValue {
 
 			Self::Array(_) | Self::ListExtractor(_) => {
 				let e = self.list_extractor();
-				return e.to_json().await;
+				return e.to_json(state).await;
 			}
 
 			Self::ObjectExtractor(_) | Self::Item(_) => {
 				let e = self.object_extractor();
-				return e.to_json().await;
+				return e.to_json(state).await;
 			}
 		})
 	}

crates/pile/src/command/annotate.rs

@@ -4,6 +4,7 @@ use pile_config::{Label, Source};
 use pile_dataset::{Datasets, index::DbFtsIndex};
 use pile_toolbox::cancelabletask::{CancelFlag, CancelableTaskError};
 use pile_value::{
+	extract::traits::ExtractState,
 	source::{DataSource, DirDataSource},
 	value::{Item, PileValue},
 };
@@ -46,6 +47,7 @@ impl CliCmd for AnnotateCommand {
 			.ok_or_else(|| anyhow::anyhow!("invalid field name {:?}", self.field))?;
 		let dest_path = Self::parse_dest(&self.dest)?;
 
+		let state = ExtractState { ignore_mime: false };
 		let ds = Datasets::open(&self.config)
 			.with_context(|| format!("while opening dataset for {}", self.config.display()))?;
 
@@ -75,8 +77,10 @@ impl CliCmd for AnnotateCommand {
 						};
 
 						let item = PileValue::Item(item.clone());
-						let Some(value) =
+						let Some(value) = index
-							index.get_field(&item, &field).await.with_context(|| {
+							.get_field(&state, &item, &field)
+							.await
+							.with_context(|| {
 								format!("while extracting field from {}", path.display())
 							})?
 						else {

crates/pile/src/command/fields.rs

@@ -2,7 +2,7 @@ use anyhow::{Context, Result};
 use clap::Args;
 use pile_dataset::Datasets;
 use pile_toolbox::cancelabletask::{CancelFlag, CancelableTaskError};
-use pile_value::value::PileValue;
+use pile_value::{extract::traits::ExtractState, value::PileValue};
 use serde_json::{Map, Value};
 use std::{path::PathBuf, time::Instant};
 use tokio::task::JoinSet;
@@ -61,6 +61,7 @@ impl CliCmd for FieldsCommand {
 		let mut total_counts: Map<String, Value> = Map::new();
 		let mut total_items = 0u64;
 		let jobs = self.jobs.max(1);
+		let state = ExtractState { ignore_mime: false };
 
 		for (name, dataset) in ds.sources.iter().filter(|(name, _)| {
 			self.source.is_empty() || self.source.iter().any(|s| s == name.as_str())
@@ -93,9 +94,10 @@ impl CliCmd for FieldsCommand {
 						let item =
 							item_result.with_context(|| format!("while reading source {name}"))?;
 						let name = name.clone();
+						let state = state.clone();
 						join_set.spawn(async move {
 							let item = PileValue::Item(item);
-							let result = item.count_fields().await.with_context(|| {
+							let result = item.count_fields(&state).await.with_context(|| {
 								format!("while counting fields in source {name}")
 							})?;
 							Ok(result.and_then(|v| {

crates/pile/src/command/index.rs

@@ -2,6 +2,7 @@ use anyhow::{Context, Result};
 use clap::Args;
 use pile_dataset::Datasets;
 use pile_toolbox::cancelabletask::{CancelFlag, CancelableTaskError};
+use pile_value::extract::traits::ExtractState;
 use std::{fmt::Debug, path::PathBuf};
 
 use crate::{CliCmd, GlobalContext};
@@ -26,7 +27,10 @@ impl CliCmd for IndexCommand {
 		let ds = Datasets::open(&self.config)
 			.with_context(|| format!("while opening dataset for {}", self.config.display()))?;
 
-		ds.fts_refresh(self.jobs, Some(flag)).await.map_err(|x| {
+		let state = ExtractState { ignore_mime: false };
+		ds.fts_refresh(&state, self.jobs, Some(flag))
+			.await
+			.map_err(|x| {
 				x.map_err(|x| {
 					anyhow::Error::from(x).context(format!(
 						"while refreshing fts for {}",

crates/pile/src/command/list.rs

@@ -3,7 +3,7 @@ use clap::Args;
 use pile_config::objectpath::ObjectPath;
 use pile_dataset::Datasets;
 use pile_toolbox::cancelabletask::{CancelFlag, CancelableTaskError};
-use pile_value::value::PileValue;
+use pile_value::{extract::traits::ExtractState, value::PileValue};
 use std::{path::PathBuf, str::FromStr, sync::Arc};
 use tokio::task::JoinSet;
 use tokio_stream::StreamExt;
@@ -48,6 +48,7 @@ impl CliCmd for ListCommand {
 			.with_context(|| format!("while opening dataset for {}", self.config.display()))?;
 
 		let jobs = self.jobs.max(1);
+		let state = ExtractState { ignore_mime: false };
 
 		for (name, dataset) in ds.sources.iter().filter(|(name, _)| {
 			self.source.is_empty() || self.source.iter().any(|s| s == name.as_str())
@@ -78,10 +79,11 @@ impl CliCmd for ListCommand {
 						let key = item.key().to_string();
 						let path = path.clone();
 						let invert = self.invert;
+						let state = state.clone();
 
 						join_set.spawn(async move {
 							let item = PileValue::Item(item);
-							let value = item.query(&path).await?;
+							let value = item.query(&state, &path).await?;
 
 							let is_present =
 								matches!(value, Some(v) if !matches!(v, PileValue::Null));

crates/pile/src/command/lookup.rs

@@ -2,6 +2,7 @@ use anyhow::{Context, Result};
 use clap::Args;
 use pile_dataset::Datasets;
 use pile_toolbox::cancelabletask::{CancelFlag, CancelableTaskError};
+use pile_value::extract::traits::ExtractState;
 use std::{fmt::Debug, path::PathBuf};
 use tracing::info;
 
@@ -42,9 +43,13 @@ impl CliCmd for LookupCommand {
 		let ds = Datasets::open(&self.config)
 			.with_context(|| format!("while opening dataset for {}", self.config.display()))?;
 
+		let state = ExtractState { ignore_mime: false };
+
 		if self.refresh && ds.needs_fts().await.context("while checking dataset fts")? {
 			info!("FTS index is missing or out-of-date, regenerating");
-			ds.fts_refresh(self.jobs, Some(flag)).await.map_err(|x| {
+			ds.fts_refresh(&state, self.jobs, Some(flag))
+				.await
+				.map_err(|x| {
 					x.map_err(|x| {
 						anyhow::Error::from(x).context(format!(
 							"while refreshing fts for {}",

crates/pile/src/command/probe.rs

@@ -3,7 +3,7 @@ use clap::Args;
 use pile_config::{Label, objectpath::ObjectPath};
 use pile_dataset::Datasets;
 use pile_toolbox::cancelabletask::{CancelFlag, CancelableTaskError};
-use pile_value::value::PileValue;
+use pile_value::{extract::traits::ExtractState, value::PileValue};
 use std::path::PathBuf;
 
 use crate::{CliCmd, GlobalContext};
@@ -39,12 +39,14 @@ impl CliCmd for ProbeCommand {
 		let ds = Datasets::open(&self.config)
 			.with_context(|| format!("while opening dataset for {}", self.config.display()))?;
 
+		let state = ExtractState { ignore_mime: false };
+
 		let json = if let Some(path_str) = self.path {
 			let path: ObjectPath = path_str
 				.parse()
 				.with_context(|| format!("invalid path {path_str:?}"))?;
 
-			ds.get_field(&source, &self.key, &path)
+			ds.get_field(&state, &source, &self.key, &path)
 				.await
 				.with_context(|| format!("while extracting {}", self.key))?
 				.ok_or_else(|| {
@@ -56,7 +58,7 @@ impl CliCmd for ProbeCommand {
 			})?;
 
 			let item = PileValue::Item(item);
-			item.to_json()
+			item.to_json(&state)
 				.await
 				.with_context(|| format!("while extracting {}", self.key))?
 		};

crates/pile/src/command/serve.rs

@@ -2,6 +2,7 @@ use anyhow::{Context, Result};
 use clap::Args;
 use pile_dataset::Datasets;
 use pile_toolbox::cancelabletask::{CancelFlag, CancelableTaskError};
+use pile_value::extract::traits::ExtractState;
 use std::{fmt::Debug, path::PathBuf, sync::Arc};
 use tracing::{error, info};
 
@@ -35,9 +36,11 @@ impl CliCmd for ServeCommand {
 		let ds = Datasets::open(&self.config)
 			.with_context(|| format!("while opening dataset for {}", self.config.display()))?;
 
+		let state = ExtractState { ignore_mime: false };
+
 		if self.refresh && ds.needs_fts().await.context("while checking dataset fts")? {
 			info!("FTS index is missing or out-of-date, regenerating");
-			ds.fts_refresh(self.jobs, Some(flag.clone()))
+			ds.fts_refresh(&state, self.jobs, Some(flag.clone()))
 				.await
 				.map_err(|x| {
 					x.map_err(|x| {