Add regex extractor

Cargo.lock

@@ -2574,6 +2574,7 @@ dependencies = [
  "pdfium-render",
  "pile-config",
  "pile-flac",
+ "regex",
  "serde_json",
  "smartstring",
  "tokio",

Cargo.toml

@@ -127,6 +127,7 @@ mime = "0.3.17"
 mime_guess = "2.0.5"
 paste = "1.0.15"
 smartstring = "1.0.1"
+regex = "1"
 chrono = "0.4.43"
 parking_lot = "0.12.5"
 rayon = "1.11.0"

crates/pile-config/src/objectpath/tokenizer.rs

@@ -21,7 +21,52 @@ impl Tokenizer {
 		let mut tokens = Vec::new();
 
 		let mut window_start = None;
+		// Paren depth: while > 0, `.` / `[` / `]` / `$` are part of the ident.
+		let mut paren_depth: usize = 0;
+		// When true, the current char is escaped by a preceding `\` and is
+		// treated as a plain ident character with no special meaning.
+		let mut skip_next = false;
+
 		for (i, c) in source.char_indices() {
+			if skip_next {
+				skip_next = false;
+				// Escaped char: just extend the ident window (already opened by `\`).
+				continue;
+			}
+
+			if c == '\\' {
+				if window_start.is_none() {
+					window_start = Some(i);
+				}
+				skip_next = true;
+				continue;
+			}
+
+			if paren_depth > 0 {
+				// Inside parens: only track depth changes, everything else is ident.
+				match c {
+					'(' => {
+						if window_start.is_none() {
+							window_start = Some(i);
+						}
+						paren_depth += 1;
+					}
+					')' => {
+						if window_start.is_none() {
+							window_start = Some(i);
+						}
+						paren_depth -= 1;
+					}
+					x if x.is_ascii() => {
+						if window_start.is_none() {
+							window_start = Some(i);
+						}
+					}
+					char => return Err(PathParseError::NonAsciiChar { position: i, char }),
+				}
+				continue;
+			}
+
 			match c {
 				'$' => {
 					if let Some(s) = window_start.take() {
@@ -51,10 +96,26 @@ impl Tokenizer {
 					tokens.push((i, Token::SqbClose));
 				}
 
-				x if x.is_ascii() => match window_start {
+				'(' => {
-					None => window_start = Some(i),
+					if window_start.is_none() {
-					Some(_) => continue,
+						window_start = Some(i);
-				},
+					}
+					paren_depth += 1;
+				}
+
+				')' => {
+					if window_start.is_none() {
+						window_start = Some(i);
+					}
+					// paren_depth is 0 here — stray `)` is an ident char and
+					// parse_field will surface the error later.
+				}
+
+				x if x.is_ascii() => {
+					if window_start.is_none() {
+						window_start = Some(i);
+					}
+				}
 
 				char => return Err(PathParseError::NonAsciiChar { position: i, char }),
 			}

crates/pile-config/src/pattern/parser.rs

@@ -122,6 +122,14 @@ mod tests {
 		GroupSegment::Literal(s.into())
 	}
 
+	#[test]
+	fn regex() {
+		assert_eq!(
+			parse("{$.split(/)[-1].regex((.*).pub \\((.*)\\).pdf)[0]}").unwrap(),
+			vec![(0, path("$.split(/)[-1].regex((.*).pub \\((.*)\\).pdf)[0]"))]
+		);
+	}
+
 	#[test]
 	fn single_path() {
 		assert_eq!(parse("{$.foo}").unwrap(), vec![(0, path("$.foo"))]);

crates/pile-value/Cargo.toml

@@ -18,6 +18,7 @@ tracing = { workspace = true }
 chrono = { workspace = true }
 toml = { workspace = true }
 smartstring = { workspace = true }
+regex = { workspace = true }
 blake3 = { workspace = true }
 epub = { workspace = true }
 kamadak-exif = { workspace = true }

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

@@ -1,4 +1,5 @@
 pub mod item;
 pub mod misc;
+pub mod regex;
 pub mod string;
 pub mod traits;

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

@@ -0,0 +1,104 @@
+use std::sync::Arc;
+
+use pile_config::Label;
+use regex::Regex;
+use smartstring::{LazyCompact, SmartString};
+
+use crate::{
+	extract::traits::{ExtractState, ListExtractor, ObjectExtractor},
+	value::PileValue,
+};
+
+struct RegexData {
+	regex: Arc<Regex>,
+	/// Captured substrings indexed by group index (0 = whole match).
+	captures: Vec<Option<Arc<SmartString<LazyCompact>>>>,
+}
+
+impl RegexData {
+	fn new(regex: Arc<Regex>, input: &str) -> Option<Self> {
+		let caps = regex.captures(input)?;
+		let captures = caps
+			.iter()
+			.map(|m| m.map(|m| Arc::new(m.as_str().into())))
+			.collect();
+		Some(Self { regex, captures })
+	}
+}
+
+/// Exposes named capture groups as object fields.
+pub struct RegexExtractor(Arc<RegexData>);
+
+impl RegexExtractor {
+	/// Run `regex` against `input`. Returns `None` if there is no match.
+	pub fn new(regex: Arc<Regex>, input: &str) -> Option<Self> {
+		Some(Self(Arc::new(RegexData::new(regex, input)?)))
+	}
+}
+
+#[async_trait::async_trait]
+impl ObjectExtractor for RegexExtractor {
+	async fn field(
+		&self,
+		_state: &ExtractState,
+		name: &Label,
+		args: Option<&str>,
+	) -> Result<Option<PileValue>, std::io::Error> {
+		if args.is_some() {
+			return Ok(None);
+		}
+
+		let Some(idx) = self
+			.0
+			.regex
+			.capture_names()
+			.position(|n| n == Some(name.as_str()))
+		else {
+			return Ok(None);
+		};
+
+		Ok(Some(
+			match self.0.captures.get(idx).and_then(|v| v.as_ref()) {
+				Some(s) => PileValue::String(s.clone()),
+				None => PileValue::Null,
+			},
+		))
+	}
+
+	async fn fields(&self) -> Result<Vec<Label>, std::io::Error> {
+		#[expect(clippy::unwrap_used)]
+		Ok(self
+			.0
+			.regex
+			.capture_names()
+			.flatten()
+			.map(|n| Label::new(n).unwrap())
+			.collect())
+	}
+
+	fn as_list(&self) -> Option<Arc<dyn ListExtractor>> {
+		Some(Arc::new(RegexExtractor(self.0.clone())))
+	}
+}
+
+#[async_trait::async_trait]
+impl ListExtractor for RegexExtractor {
+	async fn get(
+		&self,
+		_state: &ExtractState,
+		idx: usize,
+	) -> Result<Option<PileValue>, std::io::Error> {
+		let raw_idx = idx + 1;
+		let Some(slot) = self.0.captures.get(raw_idx) else {
+			return Ok(None);
+		};
+		Ok(Some(match slot {
+			Some(s) => PileValue::String(s.clone()),
+			None => PileValue::Null,
+		}))
+	}
+
+	async fn len(&self, _state: &ExtractState) -> Result<usize, std::io::Error> {
+		Ok(self.0.captures.len().saturating_sub(1))
+	}
+}

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

@@ -1,9 +1,13 @@
 use pile_config::Label;
+use regex::Regex;
 use smartstring::{LazyCompact, SmartString};
 use std::sync::Arc;
 
 use crate::{
-	extract::traits::{ExtractState, ObjectExtractor},
+	extract::{
+		regex::RegexExtractor,
+		traits::{ExtractState, ObjectExtractor},
+	},
 	value::PileValue,
 };
 
@@ -67,6 +71,18 @@ impl ObjectExtractor for StringExtractor {
 					.collect(),
 			))),
 
+			("regex", Some(pattern)) => {
+				let Ok(re) = Regex::new(pattern) else {
+					return Ok(None);
+				};
+				Some(
+					match RegexExtractor::new(Arc::new(re), self.item.as_str()) {
+						Some(ext) => PileValue::ObjectExtractor(Arc::new(ext)),
+						None => PileValue::Null,
+					},
+				)
+			}
+
 			_ => None,
 		})
 	}
@@ -78,6 +94,9 @@ impl ObjectExtractor for StringExtractor {
 			Label::new("upper").unwrap(),
 			Label::new("lower").unwrap(),
 			Label::new("nonempty").unwrap(),
+			Label::new("trimprefix").unwrap(),
+			Label::new("trimsuffix").unwrap(),
+			Label::new("split").unwrap(),
 		]);
 	}
 }

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

@@ -35,6 +35,11 @@ pub trait ObjectExtractor: Send + Sync {
 	/// and [None] for all others.
 	async fn fields(&self) -> Result<Vec<pile_config::Label>, std::io::Error>;
 
+	/// Return a list view of this extractor, if supported.
+	fn as_list(&self) -> Option<std::sync::Arc<dyn ListExtractor>> {
+		None
+	}
+
 	/// Convert this to a JSON value.
 	async fn to_json(&self, state: &ExtractState) -> Result<serde_json::Value, std::io::Error> {
 		let keys = self.fields().await?;

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

@@ -86,7 +86,9 @@ impl PileValue {
 			Self::String(_) => Arc::new(VecExtractor::default()),
 			Self::Blob { .. } => Arc::new(VecExtractor::default()),
 			Self::ListExtractor(e) => e.clone(),
-			Self::ObjectExtractor(_) => Arc::new(VecExtractor::default()),
+			Self::ObjectExtractor(e) => e
+				.as_list()
+				.unwrap_or_else(|| Arc::new(VecExtractor::default())),
 			Self::Item(_) => Arc::new(VecExtractor::default()),
 		}
 	}