Basic package management

1 files changed, 79 insertions, 100 deletions

diff --git a/src/syntax/source.rs b/src/syntax/source.rs
index 277271db..6eb6fd5d 100644
--- a/src/syntax/source.rs
+++ b/src/syntax/source.rs
@@ -3,105 +3,107 @@
 use std::fmt::{self, Debug, Formatter};
 use std::hash::{Hash, Hasher};
 use std::ops::Range;
-use std::path::{Path, PathBuf};
+use std::sync::Arc;
 
 use comemo::Prehashed;
-use unscanny::Scanner;
 
 use super::ast::Markup;
 use super::reparser::reparse;
 use super::{is_newline, parse, LinkedNode, Span, SyntaxNode};
 use crate::diag::SourceResult;
-use crate::util::{PathExt, StrExt};
+use crate::file::FileId;
+use crate::util::StrExt;
 
 /// A source file.
 ///
 /// All line and column indices start at zero, just like byte indices. Only for
 /// user-facing display, you should add 1 to them.
+///
+/// Values of this type are cheap to clone and hash.
 #[derive(Clone)]
-pub struct Source {
-    id: SourceId,
-    path: PathBuf,
-    lines: Vec<Line>,
+pub struct Source(Arc<Repr>);
+
+/// The internal representation.
+#[derive(Clone)]
+struct Repr {
+    id: FileId,
     text: Prehashed<String>,
     root: Prehashed<SyntaxNode>,
+    lines: Vec<Line>,
 }
 
 impl Source {
     /// Create a new source file.
+    ///
+    /// The path must be canonical, so that the same source file has the same
+    /// id even if accessed through different paths.
     #[tracing::instrument(skip_all)]
-    pub fn new(id: SourceId, path: &Path, text: String) -> Self {
+    pub fn new(id: FileId, text: String) -> Self {
         let mut root = parse(&text);
         root.numberize(id, Span::FULL).unwrap();
-        Self {
+        Self(Arc::new(Repr {
             id,
-            path: path.normalize(),
             lines: lines(&text),
             text: Prehashed::new(text),
             root: Prehashed::new(root),
-        }
+        }))
     }
 
     /// Create a source file without a real id and path, usually for testing.
     pub fn detached(text: impl Into<String>) -> Self {
-        Self::new(SourceId::detached(), Path::new(""), text.into())
+        Self::new(FileId::detached(), text.into())
     }
 
     /// Create a source file with the same synthetic span for all nodes.
     pub fn synthesized(text: String, span: Span) -> Self {
         let mut root = parse(&text);
         root.synthesize(span);
-        Self {
-            id: SourceId::detached(),
-            path: PathBuf::new(),
+        Self(Arc::new(Repr {
+            id: FileId::detached(),
             lines: lines(&text),
             text: Prehashed::new(text),
             root: Prehashed::new(root),
-        }
+        }))
     }
 
     /// The root node of the file's untyped syntax tree.
     pub fn root(&self) -> &SyntaxNode {
-        &self.root
+        &self.0.root
     }
 
     /// The root node of the file's typed abstract syntax tree.
     pub fn ast(&self) -> SourceResult<Markup> {
-        let errors = self.root.errors();
+        let errors = self.root().errors();
         if errors.is_empty() {
-            Ok(self.root.cast().expect("root node must be markup"))
+            Ok(self.root().cast().expect("root node must be markup"))
         } else {
             Err(Box::new(errors))
         }
     }
 
     /// The id of the source file.
-    pub fn id(&self) -> SourceId {
-        self.id
-    }
-
-    /// The normalized path to the source file.
-    pub fn path(&self) -> &Path {
-        &self.path
+    pub fn id(&self) -> FileId {
+        self.0.id
     }
 
     /// The whole source as a string slice.
     pub fn text(&self) -> &str {
-        &self.text
+        &self.0.text
     }
 
     /// Slice out the part of the source code enclosed by the range.
     pub fn get(&self, range: Range<usize>) -> Option<&str> {
-        self.text.get(range)
+        self.text().get(range)
     }
 
     /// Fully replace the source text.
     pub fn replace(&mut self, text: String) {
-        self.text = Prehashed::new(text);
-        self.lines = lines(&self.text);
-        let mut root = parse(&self.text);
-        root.numberize(self.id, Span::FULL).unwrap();
-        self.root = Prehashed::new(root);
+        let inner = Arc::make_mut(&mut self.0);
+        inner.text = Prehashed::new(text);
+        inner.lines = lines(&inner.text);
+        let mut root = parse(&inner.text);
+        root.numberize(inner.id, Span::FULL).unwrap();
+        inner.root = Prehashed::new(root);
     }
 
     /// Edit the source file by replacing the given range.
@@ -112,72 +114,70 @@ impl Source {
     #[track_caller]
     pub fn edit(&mut self, replace: Range<usize>, with: &str) -> Range<usize> {
         let start_byte = replace.start;
-        let start_utf16 = self.byte_to_utf16(replace.start).unwrap();
-        self.text.update(|text| text.replace_range(replace.clone(), with));
+        let start_utf16 = self.byte_to_utf16(start_byte).unwrap();
+        let line = self.byte_to_line(start_byte).unwrap();
+
+        let inner = Arc::make_mut(&mut self.0);
+
+        // Update the text itself.
+        inner.text.update(|text| text.replace_range(replace.clone(), with));
 
         // Remove invalidated line starts.
-        let line = self.byte_to_line(start_byte).unwrap();
-        self.lines.truncate(line + 1);
+        inner.lines.truncate(line + 1);
 
         // Handle adjoining of \r and \n.
-        if self.text[..start_byte].ends_with('\r') && with.starts_with('\n') {
-            self.lines.pop();
+        if inner.text[..start_byte].ends_with('\r') && with.starts_with('\n') {
+            inner.lines.pop();
         }
 
         // Recalculate the line starts after the edit.
-        self.lines
-            .extend(lines_from(start_byte, start_utf16, &self.text[start_byte..]));
+        inner.lines.extend(lines_from(
+            start_byte,
+            start_utf16,
+            &inner.text[start_byte..],
+        ));
 
         // Incrementally reparse the replaced range.
-        self.root
-            .update(|root| reparse(root, &self.text, replace, with.len()))
+        inner
+            .root
+            .update(|root| reparse(root, &inner.text, replace, with.len()))
     }
 
     /// Get the length of the file in UTF-8 encoded bytes.
     pub fn len_bytes(&self) -> usize {
-        self.text.len()
+        self.text().len()
     }
 
     /// Get the length of the file in UTF-16 code units.
     pub fn len_utf16(&self) -> usize {
-        let last = self.lines.last().unwrap();
-        last.utf16_idx + self.text[last.byte_idx..].len_utf16()
+        let last = self.0.lines.last().unwrap();
+        last.utf16_idx + self.0.text[last.byte_idx..].len_utf16()
     }
 
     /// Get the length of the file in lines.
     pub fn len_lines(&self) -> usize {
-        self.lines.len()
+        self.0.lines.len()
     }
 
     /// Find the node with the given span.
     ///
     /// Returns `None` if the span does not point into this source file.
     pub fn find(&self, span: Span) -> Option<LinkedNode<'_>> {
-        LinkedNode::new(&self.root).find(span)
-    }
-
-    /// Map a span that points into this source file to a byte range.
-    ///
-    /// Panics if the span does not point into this source file.
-    #[track_caller]
-    pub fn range(&self, span: Span) -> Range<usize> {
-        self.find(span)
-            .expect("span does not point into this source file")
-            .range()
+        LinkedNode::new(self.root()).find(span)
     }
 
     /// Return the index of the UTF-16 code unit at the byte index.
     pub fn byte_to_utf16(&self, byte_idx: usize) -> Option<usize> {
         let line_idx = self.byte_to_line(byte_idx)?;
-        let line = self.lines.get(line_idx)?;
-        let head = self.text.get(line.byte_idx..byte_idx)?;
+        let line = self.0.lines.get(line_idx)?;
+        let head = self.0.text.get(line.byte_idx..byte_idx)?;
         Some(line.utf16_idx + head.len_utf16())
     }
 
     /// Return the index of the line that contains the given byte index.
     pub fn byte_to_line(&self, byte_idx: usize) -> Option<usize> {
-        (byte_idx <= self.text.len()).then(|| {
-            match self.lines.binary_search_by_key(&byte_idx, |line| line.byte_idx) {
+        (byte_idx <= self.0.text.len()).then(|| {
+            match self.0.lines.binary_search_by_key(&byte_idx, |line| line.byte_idx) {
                 Ok(i) => i,
                 Err(i) => i - 1,
             }
@@ -197,33 +197,33 @@ impl Source {
 
     /// Return the byte index at the UTF-16 code unit.
     pub fn utf16_to_byte(&self, utf16_idx: usize) -> Option<usize> {
-        let line = self.lines.get(
-            match self.lines.binary_search_by_key(&utf16_idx, |line| line.utf16_idx) {
+        let line = self.0.lines.get(
+            match self.0.lines.binary_search_by_key(&utf16_idx, |line| line.utf16_idx) {
                 Ok(i) => i,
                 Err(i) => i - 1,
             },
         )?;
 
         let mut k = line.utf16_idx;
-        for (i, c) in self.text[line.byte_idx..].char_indices() {
+        for (i, c) in self.0.text[line.byte_idx..].char_indices() {
             if k >= utf16_idx {
                 return Some(line.byte_idx + i);
             }
             k += c.len_utf16();
         }
 
-        (k == utf16_idx).then_some(self.text.len())
+        (k == utf16_idx).then_some(self.0.text.len())
     }
 
     /// Return the byte position at which the given line starts.
     pub fn line_to_byte(&self, line_idx: usize) -> Option<usize> {
-        self.lines.get(line_idx).map(|line| line.byte_idx)
+        self.0.lines.get(line_idx).map(|line| line.byte_idx)
     }
 
     /// Return the range which encloses the given line.
     pub fn line_to_range(&self, line_idx: usize) -> Option<Range<usize>> {
         let start = self.line_to_byte(line_idx)?;
-        let end = self.line_to_byte(line_idx + 1).unwrap_or(self.text.len());
+        let end = self.line_to_byte(line_idx + 1).unwrap_or(self.0.text.len());
         Some(start..end)
     }
 
@@ -248,42 +248,21 @@ impl Source {
 
 impl Debug for Source {
     fn fmt(&self, f: &mut Formatter) -> fmt::Result {
-        write!(f, "Source({})", self.path.display())
+        write!(f, "Source({})", self.id().path().display())
     }
 }
 
 impl Hash for Source {
     fn hash<H: Hasher>(&self, state: &mut H) {
-        self.id.hash(state);
-        self.path.hash(state);
-        self.text.hash(state);
-        self.root.hash(state);
+        self.0.id.hash(state);
+        self.0.text.hash(state);
+        self.0.root.hash(state);
     }
 }
 
-/// A unique identifier for a loaded source file.
-#[derive(Debug, Copy, Clone, Eq, PartialEq, Hash)]
-pub struct SourceId(u16);
-
-impl SourceId {
-    /// Create a new source id for a file that is not part of the world.
-    pub const fn detached() -> Self {
-        Self(u16::MAX)
-    }
-
-    /// Whether the source id is the detached.
-    pub const fn is_detached(self) -> bool {
-        self.0 == Self::detached().0
-    }
-
-    /// Create a source id from a number.
-    pub const fn from_u16(v: u16) -> Self {
-        Self(v)
-    }
-
-    /// Extract the underlying number.
-    pub const fn as_u16(self) -> u16 {
-        self.0
+impl AsRef<str> for Source {
+    fn as_ref(&self) -> &str {
+        self.text()
     }
 }
 
@@ -309,7 +288,7 @@ fn lines_from(
     utf16_offset: usize,
     text: &str,
 ) -> impl Iterator<Item = Line> + '_ {
-    let mut s = Scanner::new(text);
+    let mut s = unscanny::Scanner::new(text);
     let mut utf16_idx = utf16_offset;
 
     std::iter::from_fn(move || {
@@ -340,7 +319,7 @@ mod tests {
     fn test_source_file_new() {
         let source = Source::detached(TEST);
         assert_eq!(
-            source.lines,
+            source.0.lines,
             [
                 Line { byte_idx: 0, utf16_idx: 0 },
                 Line { byte_idx: 7, utf16_idx: 6 },
@@ -421,8 +400,8 @@ mod tests {
             let mut source = Source::detached(prev);
             let result = Source::detached(after);
             source.edit(range, with);
-            assert_eq!(source.text, result.text);
-            assert_eq!(source.lines, result.lines);
+            assert_eq!(source.text(), result.text());
+            assert_eq!(source.0.lines, result.0.lines);
         }
 
         // Test inserting at the beginning.