Flatten arrays for FTS index

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

@@ -1,4 +1,4 @@
-use pile_config::{ConfigToml, DatasetFts, Label};
+use pile_config::{ConfigToml, DatasetFts, Label, objectpath::ObjectPath};
 use pile_value::{
 	extract::traits::ExtractState,
 	value::{Item, PileValue},
@@ -10,7 +10,7 @@ use tantivy::{
 	query::QueryParser,
 	schema::{self, Schema, Value as TantivyValue},
 };
-use tracing::{debug, trace, warn};
+use tracing::warn;
 
 #[derive(Debug, Clone)]
 pub struct FtsLookupResult {
@@ -79,18 +79,12 @@ impl DbFtsIndex {
 
 		let mut empty = true;
 		for name in self.fts_cfg().fields.keys() {
-			let x = self.get_field(state, &item, name).await?;
+			let vals = self.get_field(state, &item, name).await?;
+			let field = self.schema.get_field(name)?;
 
-			let val = match x {
-				Some(x) => x,
-				None => continue,
-			};
-
-			empty = false;
-
-			let field = self.schema.get_field(name);
-			if let Ok(field) = field {
-				doc.add_text(field, val);
+			for v in vals {
+				empty = false;
+				doc.add_text(field, v);
 			}
 		}
 
@@ -106,110 +100,29 @@ impl DbFtsIndex {
 		state: &ExtractState,
 		extractor: &PileValue,
 		field_name: &Label,
-	) -> Result<Option<String>, std::io::Error> {
+	) -> Result<Vec<String>, std::io::Error> {
 		let field = match self.cfg.schema.get(field_name) {
 			Some(x) => x,
 			None => {
 				warn!("Unknown field {field_name:?}");
-				return Ok(None);
+				return Ok(Vec::new());
 			}
 		};
 
 		// Try paths in order, using the first value we find
-		'outer: for path in field.path.as_slice() {
+		for path in field.path.as_slice() {
 			let val = match extractor.query(state, path).await? {
 				Some(x) => x,
-				None => return Ok(None),
+				None => continue,
 			};
 
-			let mut val = match val {
-				PileValue::Null => {
-					trace!(
-						message = "Skipping field, is null",
-						field = field_name.to_string(),
-						?path,
-						// value = ?val
-					);
-					continue;
-				}
-				x => x.clone(),
-			};
-
-			loop {
-				val = match val {
-					PileValue::String(x) => return Ok(Some(x.to_string())),
-					PileValue::U64(x) => return Ok(Some(x.to_string())),
-					PileValue::I64(x) => return Ok(Some(x.to_string())),
-
-					PileValue::Array(x) => {
-						if x.len() == 1 {
-							x[0].clone()
-						} else if x.len() > 1 {
-							debug!(
-								message = "Skipping field, is array with more than one element",
-								field = field_name.to_string(),
-								?path,
-							);
-							continue 'outer;
-						} else {
-							debug!(
-								message = "Skipping field, is empty array",
-								field = field_name.to_string(),
-								?path,
-							);
-							continue 'outer;
-						}
-					}
-
-					PileValue::Null => {
-						trace!(
-							message = "Skipping field, is null",
-							field = field_name.to_string(),
-							?path,
-						);
-						continue 'outer;
-					}
-
-					PileValue::ObjectExtractor(_) => {
-						trace!(
-							message = "Skipping field, is object",
-							field = field_name.to_string(),
-							?path,
-						);
-						continue 'outer;
-					}
-
-					PileValue::Item(_) => {
-						trace!(
-							message = "Skipping field, is item",
-							field = field_name.to_string(),
-							?path,
-						);
-						continue 'outer;
-					}
-
-					PileValue::ListExtractor(_) => {
-						trace!(
-							message = "Skipping field, is ListExtractor",
-							field = field_name.to_string(),
-							?path,
-						);
-						continue 'outer;
-					}
-
-					PileValue::Blob { .. } => {
-						trace!(
-							message = "Skipping field, is blob",
-							field = field_name.to_string(),
-							?path,
-						);
-						continue 'outer;
-					}
-				}
+			let val = val_to_string(state, &val, path, field_name).await?;
+			if !val.is_empty() {
+				return Ok(val);
 			}
 		}
 
-		return Ok(None);
+		return Ok(Vec::new());
 	}
 
 	/// Run the given query on this table's ftx index.
@@ -298,3 +211,42 @@ impl DbFtsIndex {
 		return Ok(out);
 	}
 }
+
+async fn val_to_string(
+	state: &ExtractState,
+	val: &PileValue,
+	path: &ObjectPath,
+	field_name: &str,
+) -> Result<Vec<String>, std::io::Error> {
+	match val {
+		PileValue::String(x) => return Ok(vec![x.to_string()]),
+		PileValue::U64(x) => return Ok(vec![x.to_string()]),
+		PileValue::I64(x) => return Ok(vec![x.to_string()]),
+
+		PileValue::Array(x) => {
+			let mut out = Vec::new();
+			for x in x.iter() {
+				out.extend(Box::pin(val_to_string(state, x, path, field_name)).await?);
+			}
+			return Ok(out);
+		}
+
+		#[expect(clippy::unwrap_used)]
+		PileValue::ListExtractor(x) => {
+			let mut out = Vec::new();
+			let len = x.len(state).await?;
+			for i in 0..len {
+				let v = x.get(state, i).await?;
+				out.extend(Box::pin(val_to_string(state, &v.unwrap(), path, field_name)).await?);
+			}
+			return Ok(out);
+		}
+
+		PileValue::Null => {}
+		PileValue::ObjectExtractor(_) => {}
+		PileValue::Item(_) => {}
+		PileValue::Blob { .. } => {}
+	}
+
+	return Ok(Vec::new());
+}

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

@@ -58,13 +58,15 @@ impl Id3Extractor {
 
 		let mut output: HashMap<Label, Vec<PileValue>> = HashMap::new();
 		for frame in tag.frames() {
-			if let Some(text) = frame.content().text() {
+			if let Some(texts) = frame.content().text_values() {
 				let name = frame_id_to_field(frame.id());
 				if let Some(key) = Label::new(name) {
-					output
-						.entry(key)
-						.or_default()
-						.push(PileValue::String(Arc::new(text.into())));
+					for text in texts {
+						output
+							.entry(key.clone())
+							.or_default()
+							.push(PileValue::String(Arc::new(text.into())));
+					}
 				}
 			}
 		}

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

@@ -0,0 +1,87 @@
+use pile_config::Label;
+use std::{
+	collections::HashMap,
+	sync::{Arc, OnceLock},
+};
+
+use crate::{
+	extract::traits::{ExtractState, ObjectExtractor},
+	value::{AsyncReader, Item, PileValue},
+};
+
+fn json_to_pile(value: serde_json::Value) -> PileValue {
+	match value {
+		serde_json::Value::Null => PileValue::Null,
+		serde_json::Value::Bool(b) => PileValue::String(Arc::new(b.to_string().into())),
+		serde_json::Value::Number(n) => PileValue::String(Arc::new(n.to_string().into())),
+		serde_json::Value::String(s) => PileValue::String(Arc::new(s.into())),
+		serde_json::Value::Array(a) => {
+			PileValue::Array(Arc::new(a.into_iter().map(json_to_pile).collect()))
+		}
+		serde_json::Value::Object(_) => PileValue::Null,
+	}
+}
+
+pub struct JsonExtractor {
+	item: Item,
+	output: OnceLock<HashMap<Label, PileValue>>,
+}
+
+impl JsonExtractor {
+	pub fn new(item: &Item) -> Self {
+		Self {
+			item: item.clone(),
+			output: OnceLock::new(),
+		}
+	}
+
+	async fn get_inner(&self) -> Result<&HashMap<Label, PileValue>, std::io::Error> {
+		if let Some(x) = self.output.get() {
+			return Ok(x);
+		}
+
+		let mut reader = self.item.read().await?;
+		let bytes = reader.read_to_end().await?;
+		let json: serde_json::Value = match serde_json::from_slice(&bytes) {
+			Ok(x) => x,
+			Err(_) => return Ok(self.output.get_or_init(HashMap::new)),
+		};
+
+		let output: HashMap<Label, PileValue> = match json {
+			serde_json::Value::Object(map) => map
+				.into_iter()
+				.filter_map(|(k, v)| Label::new(&k).map(|label| (label, json_to_pile(v))))
+				.collect(),
+			_ => HashMap::new(),
+		};
+
+		return Ok(self.output.get_or_init(|| output));
+	}
+}
+
+#[async_trait::async_trait]
+impl ObjectExtractor for JsonExtractor {
+	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().type_() != mime::APPLICATION
+				&& self.item.mime().type_() != mime::TEXT)
+		{
+			return Ok(None);
+		}
+
+		Ok(self.get_inner().await?.get(name).cloned())
+	}
+
+	async fn fields(&self) -> Result<Vec<Label>, std::io::Error> {
+		Ok(self.get_inner().await?.keys().cloned().collect())
+	}
+}

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

@@ -18,6 +18,9 @@ pub use exif::*;
 mod pdf;
 pub use pdf::*;
 
+mod json;
+pub use json::*;
+
 mod toml;
 use pile_config::Label;
 pub use toml::*;
@@ -66,6 +69,10 @@ impl ItemExtractor {
 					Label::new("pdf").unwrap(),
 					PileValue::ObjectExtractor(Arc::new(PdfExtractor::new(item))),
 				),
+				(
+					Label::new("json").unwrap(),
+					PileValue::ObjectExtractor(Arc::new(JsonExtractor::new(item))),
+				),
 				(
 					Label::new("toml").unwrap(),
 					PileValue::ObjectExtractor(Arc::new(TomlExtractor::new(item))),
@@ -101,6 +108,7 @@ impl ObjectExtractor for ItemExtractor {
 			Label::new("epub").unwrap(),
 			Label::new("exif").unwrap(),
 			Label::new("pdf").unwrap(),
+			Label::new("json").unwrap(),
 			Label::new("sidecar").unwrap(),
 		]);
 	}

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

@@ -47,14 +47,20 @@ impl DataSource for Arc<DirDataSource> {
 			source: Arc::clone(self),
 			mime: mime_guess::from_path(&key).first_or_octet_stream(),
 			path: key.clone(),
-			sidecar: self.sidecars.then(|| {
-				Box::new(Item::File {
-					source: Arc::clone(self),
-					mime: mime_guess::from_path(key.with_extension("toml")).first_or_octet_stream(),
-					path: key.with_extension("toml"),
-					sidecar: None,
+			sidecar: self
+				.sidecars
+				.then(|| {
+					let sidecar_path = key.with_extension("toml");
+					sidecar_path.is_file().then(|| {
+						Box::new(Item::File {
+							source: Arc::clone(self),
+							mime: mime_guess::from_path(&sidecar_path).first_or_octet_stream(),
+							path: sidecar_path,
+							sidecar: None,
+						})
+					})
 				})
-			}),
+				.flatten(),
 		}));
 	}
 
@@ -91,15 +97,21 @@ impl DataSource for Arc<DirDataSource> {
 						mime: mime_guess::from_path(&path).first_or_octet_stream(),
 						path: path.clone(),
 
-						sidecar: source.sidecars.then(|| {
-							Box::new(Item::File {
-								source: Arc::clone(&source),
-								mime: mime_guess::from_path(path.with_extension("toml"))
-									.first_or_octet_stream(),
-								path: path.with_extension("toml"),
-								sidecar: None,
+						sidecar: source
+							.sidecars
+							.then(|| {
+								let sidecar_path = path.with_extension("toml");
+								sidecar_path.is_file().then(|| {
+									Box::new(Item::File {
+										source: Arc::clone(&source),
+										mime: mime_guess::from_path(&sidecar_path)
+											.first_or_octet_stream(),
+										path: sidecar_path,
+										sidecar: None,
+									})
+								})
 							})
-						}),
+							.flatten(),
 					},
 				};
 

crates/pile/src/command/annotate.rs

@@ -77,18 +77,16 @@ impl CliCmd for AnnotateCommand {
 						};
 
 						let item = PileValue::Item(item.clone());
-						let Some(value) = index
-							.get_field(&state, &item, &field)
-							.await
-							.with_context(|| {
-								format!("while extracting field from {}", path.display())
-							})?
-						else {
-							continue;
-						};
+						let vals =
+							index
+								.get_field(&state, &item, &field)
+								.await
+								.with_context(|| {
+									format!("while extracting field from {}", path.display())
+								})?;
 
 						// TODO: implement sidecar writing
-						let _ = (&dest_path, &value);
+						let _ = (&dest_path, &vals);
 						todo!("write_sidecar not yet implemented");
 
 						#[expect(unreachable_code)]