refactoring: remove dependency on SimpleFiles, make tree parsing multithreaded

content/music/reviews.tree

@@ -6,7 +6,6 @@ this page used to host all my music reviews, but those have since been moved to
 + I haven't had time to review the following albums, but I'm including them for permalinking purposes.
 
     % id = "music/album/aphex-twin/drukqs"
-      content.link = "music/reviews/aphex-twin/drukqs"
     + ### :page: album: Aphex Twin - drukQs
 
         % id = "01H9JB094GTG1TJ029CQ4PNMS0"

crates/treehouse/src/cli/fix.rs

@@ -108,7 +108,8 @@ pub fn fix_file(
     diagnostics: &mut Vec<Diagnostic<FileId>>,
     file_id: FileId,
 ) -> Result<String, parse::ErrorsEmitted> {
-    parse_tree_with_diagnostics(treehouse, file_id)
+    let source = treehouse.source(file_id).input();
+    parse_tree_with_diagnostics(file_id, source)
         .map(|roots| {
             let mut source = treehouse.source(file_id).input().to_owned();
             let mut state = State::default();
@@ -146,7 +147,7 @@ pub fn fix_file_cli(fix_args: FixArgs, root: &dyn Dir) -> anyhow::Result<Edit> {
 
     let mut treehouse = Treehouse::new();
     let mut diagnostics = vec![];
-    let file_id = treehouse.add_file(fix_args.file.as_str().to_owned(), Source::Other(file));
+    let file_id = treehouse.add_file(fix_args.file.clone(), Source::Other(file));
     let edit_path = root.edit_path(&fix_args.file).ok_or_else(|| {
         anyhow!(
             "{} is not an editable file (perhaps it is not in a persistent path?)",
@@ -178,7 +179,7 @@ pub fn fix_file_cli(fix_args: FixArgs, root: &dyn Dir) -> anyhow::Result<Edit> {
                 Edit::NoOp
             }
         } else {
-            report_diagnostics(&treehouse.files, &diagnostics)?;
+            report_diagnostics(&treehouse, &diagnostics)?;
             Edit::NoOp
         },
     )
@@ -196,7 +197,7 @@ pub fn fix_all_cli(fix_all_args: FixAllArgs, dir: &dyn Dir) -> anyhow::Result<Ed
 
             let mut treehouse = Treehouse::new();
             let mut diagnostics = vec![];
-            let file_id = treehouse.add_file(path.as_str().to_string(), Source::Other(content));
+            let file_id = treehouse.add_file(path.to_owned(), Source::Other(content));
             let edit_path = dir.edit_path(path).context("path is not editable")?;
 
             if let Ok(fixed) = fix_file(&mut treehouse, &mut diagnostics, file_id) {
@@ -204,7 +205,7 @@ pub fn fix_all_cli(fix_all_args: FixAllArgs, dir: &dyn Dir) -> anyhow::Result<Ed
                     return Ok(Edit::Write(edit_path, fixed));
                 }
             } else {
-                report_diagnostics(&treehouse.files, &diagnostics)?;
+                report_diagnostics(&treehouse, &diagnostics)?;
             }
         }
 

crates/treehouse/src/cli/serve.rs

@@ -144,13 +144,12 @@ async fn branch(RawQuery(named_id): RawQuery, State(state): State<Arc<Server>>)
             });
         if let Some(branch_id) = branch_id {
             let branch = state.sources.treehouse.tree.branch(branch_id);
-            if let Source::Tree {
+            if let Source::Tree { input, tree_path } =
-                input, target_path, ..
+                state.sources.treehouse.source(branch.file_id)
-            } = state.sources.treehouse.source(branch.file_id)
             {
                 if let Some(content) = state
                     .target
-                    .content(target_path)
+                    .content(tree_path)
                     .await
                     .and_then(|s| String::from_utf8(s).ok())
                 {
@@ -172,7 +171,7 @@ async fn branch(RawQuery(named_id): RawQuery, State(state): State<Arc<Server>>)
                 } else {
                     return (
                         StatusCode::INTERNAL_SERVER_ERROR,
-                        format!("500 Internal Server Error: branch metadata points to entry {target_path} which does not have readable content")
+                        format!("500 Internal Server Error: branch metadata points to entry {tree_path} which does not have readable content")
                     )
                         .into_response();
                 }

crates/treehouse/src/cli/wc.rs

@@ -47,16 +47,16 @@ pub fn wc_cli(content_dir: &dyn Dir, mut wc_args: WcArgs) -> anyhow::Result<()>
             .content(path)
             .and_then(|b| String::from_utf8(b).ok())
         {
-            let file_id = treehouse.add_file(path.to_string(), Source::Other(content));
+            let file_id = treehouse.add_file(path.clone(), Source::Other(content.clone()));
-            match parse_tree_with_diagnostics(&mut treehouse, file_id) {
+            match parse_tree_with_diagnostics(file_id, &content) {
                 Ok(parsed) => {
                     let source = treehouse.source(file_id);
                     let word_count = wc_roots(source.input(), &parsed);
-                    println!("{word_count:>8} {}", treehouse.filename(file_id));
+                    println!("{word_count:>8} {}", treehouse.path(file_id));
                     total += word_count;
                 }
                 Err(diagnostics) => {
-                    report_diagnostics(&treehouse.files, &diagnostics)?;
+                    report_diagnostics(&treehouse, &diagnostics)?;
                 }
             }
         }

crates/treehouse/src/config.rs

@@ -1,6 +1,7 @@
 use std::{collections::HashMap, ops::ControlFlow};
 
 use anyhow::{anyhow, Context};
+use rayon::iter::{IntoParallelRefIterator, ParallelIterator};
 use serde::{Deserialize, Serialize};
 use tracing::{error, info_span, instrument};
 
@@ -10,7 +11,7 @@ use crate::{
         Syntax,
     },
     import_map::ImportRoot,
-    vfs::{self, Dir, ImageSize, VPath, VPathBuf},
+    vfs::{self, Dir, DynDir, ImageSize, VPath, VPathBuf},
 };
 
 #[derive(Debug, Clone, Deserialize, Serialize)]
@@ -159,9 +160,18 @@ impl Config {
 
     /// Loads all syntax definition files.
     #[instrument(name = "Config::load_syntaxes", skip(self))]
-    pub fn load_syntaxes(&mut self, dir: &dyn Dir) -> anyhow::Result<()> {
+    pub fn load_syntaxes(&mut self, dir: DynDir) -> anyhow::Result<()> {
-        vfs::walk_dir_rec(dir, VPath::ROOT, &mut |path| {
+        let mut paths = vec![];
+        vfs::walk_dir_rec(&dir, VPath::ROOT, &mut |path| {
             if path.extension() == Some("json") {
+                paths.push(path.to_owned());
+            }
+            ControlFlow::Continue(())
+        });
+
+        let syntaxes: Vec<_> = paths
+            .par_iter()
+            .flat_map(|path| {
                 let name = path
                     .file_stem()
                     .expect("syntax file name should have a stem due to the .json extension");
@@ -180,14 +190,19 @@ impl Config {
                     Ok(syntax) => {
                         let _span = info_span!("Config::load_syntaxes::compile", ?name).entered();
                         let compiled = compile_syntax(&syntax);
-                        self.syntaxes.insert(name.to_owned(), compiled);
+                        Some((name.to_owned(), compiled))
+                    }
+                    Err(err) => {
+                        error!("error while loading syntax file `{path}`: {err}");
+                        None
                     }
-                    Err(err) => error!("error while loading syntax file `{path}`: {err}"),
                 }
-            }
+            })
+            .collect();
 
-            ControlFlow::Continue(())
+        for (name, compiled) in syntaxes {
-        });
+            self.syntaxes.insert(name, compiled);
+        }
 
         Ok(())
     }

crates/treehouse/src/generate.rs

@@ -4,10 +4,10 @@ mod include_static_helper;
 use std::{collections::HashMap, fmt, ops::ControlFlow, sync::Arc};
 
 use anyhow::{anyhow, ensure, Context};
-use codespan_reporting::diagnostic::Diagnostic;
 use dir_helper::DirHelper;
 use handlebars::{handlebars_helper, Handlebars};
 use include_static_helper::IncludeStaticHelper;
+use rayon::iter::{IndexedParallelIterator, IntoParallelIterator, ParallelIterator};
 use serde::Serialize;
 use tracing::{error, info_span, instrument};
 
@@ -18,7 +18,7 @@ use crate::{
     html::{breadcrumbs::breadcrumbs_to_html, navmap::NavigationMap, tree::branches_to_html},
     import_map::ImportMap,
     parse::parse_tree_with_diagnostics,
-    state::{report_diagnostics, Source},
+    state::{report_diagnostics, FileId, Source},
     tree::SemaRoots,
     vfs::{
         self, Cd, ContentCache, Dir, DirEntry, DynDir, EditPath, ImageSize, MemDir, Overlay,
@@ -26,13 +26,7 @@ use crate::{
     },
 };
 
-use crate::state::{FileId, Treehouse};
+use crate::state::Treehouse;
-
-#[derive(Debug, Clone)]
-pub struct ParsedTree {
-    root_key: String,
-    file_id: FileId,
-}
 
 #[derive(Serialize)]
 struct Page {
@@ -96,84 +90,81 @@ fn load_templates(handlebars: &mut Handlebars, dir: &dyn Dir) {
     });
 }
 
-#[instrument(skip(treehouse, config, source, target_path, tree_path))]
-fn parse_tree(
-    treehouse: &mut Treehouse,
-    config: &Config,
-    source: String,
-    source_path: VPathBuf,
-    target_path: VPathBuf,
-    tree_path: String,
-) -> anyhow::Result<(Option<ParsedTree>, Vec<Diagnostic<FileId>>)> {
-    let file_id = treehouse.add_file(
-        source_path.as_str().to_owned(),
-        Source::Tree {
-            input: source,
-            target_path: target_path.clone(),
-            tree_path: tree_path.clone(),
-        },
-    );
-
-    match parse_tree_with_diagnostics(treehouse, file_id) {
-        Ok(roots) => {
-            let mut diagnostics = vec![];
-            let roots = SemaRoots::from_roots(treehouse, &mut diagnostics, config, file_id, roots);
-
-            let root_key = tree_path.clone();
-            treehouse.roots.insert(root_key.clone(), roots);
-
-            Ok((Some(ParsedTree { root_key, file_id }), diagnostics))
-        }
-        Err(diagnostics) => Ok((None, diagnostics)),
-    }
-}
-
 #[instrument(skip(config, dirs))]
-fn parse_trees(
+fn load_trees(config: &Config, dirs: &Dirs) -> anyhow::Result<Treehouse> {
-    config: &Config,
-    dirs: &Dirs,
-) -> anyhow::Result<(Treehouse, HashMap<VPathBuf, ParsedTree>)> {
     let mut treehouse = Treehouse::new();
     let mut diagnostics = vec![];
     let mut parsed_trees = HashMap::new();
 
+    let mut paths = vec![];
+
     vfs::walk_dir_rec(&*dirs.content, VPath::ROOT, &mut |path| {
         if path.extension() == Some("tree") {
-            if let Some(source) = dirs
+            paths.push(path.to_owned());
-                .content
-                .content(path)
-                .and_then(|b| String::from_utf8(b).ok())
-            {
-                let tree_path = path.with_extension("");
-                let target_path = path.with_extension("html");
-
-                match parse_tree(
-                    &mut treehouse,
-                    config,
-                    source,
-                    path.to_owned(),
-                    target_path,
-                    tree_path.as_str().to_owned(),
-                ) {
-                    Ok((parsed_tree, mut parse_diagnostics)) => {
-                        diagnostics.append(&mut parse_diagnostics);
-                        if let Some(parsed_tree) = parsed_tree {
-                            parsed_trees.insert(tree_path, parsed_tree);
-                        }
-                    }
-                    Err(err) => {
-                        error!("failed to parse tree {path}: {err:?}")
-                    }
-                }
-            }
         }
-
         ControlFlow::Continue(())
     });
 
-    report_diagnostics(&treehouse.files, &diagnostics)?;
+    // NOTE: Sources are filled in later; they can be left out until a call to report_diagnostics.
+    let file_ids: Vec<_> = paths
+        .iter()
+        .map(|path| treehouse.add_file(path.clone(), Source::Other(String::new())))
+        .collect();
 
-    Ok((treehouse, parsed_trees))
+    let parse_results: Vec<_> = {
+        let _span = info_span!("load_trees::parse").entered();
+        paths
+            .into_par_iter()
+            .zip(&file_ids)
+            .flat_map(|(path, &file_id)| {
+                dirs.content
+                    .content(&path)
+                    .and_then(|b| String::from_utf8(b).ok())
+                    .map(|input| {
+                        let parse_result = parse_tree_with_diagnostics(file_id, &input);
+                        (path, file_id, input, parse_result)
+                    })
+            })
+            .collect()
+    };
+
+    for (path, file_id, input, _) in &parse_results {
+        let tree_path = path.with_extension("");
+        treehouse
+            .files_by_tree_path
+            .insert(tree_path.clone(), *file_id);
+        treehouse.set_source(
+            *file_id,
+            Source::Tree {
+                input: input.clone(),
+                tree_path,
+            },
+        );
+    }
+
+    {
+        let _span = info_span!("load_trees::sema").entered();
+        for (path, file_id, _, result) in parse_results {
+            match result {
+                Ok(roots) => {
+                    let roots = SemaRoots::from_roots(
+                        &mut treehouse,
+                        &mut diagnostics,
+                        config,
+                        file_id,
+                        roots,
+                    );
+                    treehouse.roots.insert(file_id, roots);
+                    parsed_trees.insert(path, file_id);
+                }
+                Err(mut parse_diagnostics) => diagnostics.append(&mut parse_diagnostics),
+            }
+        }
+    }
+
+    report_diagnostics(&treehouse, &diagnostics)?;
+
+    Ok(treehouse)
 }
 
 #[instrument(skip(sources, handlebars))]
@@ -205,23 +196,19 @@ fn generate_simple_template_or_error(
     }
 }
 
-#[instrument(skip(sources, dirs, handlebars, parsed_tree), fields(root_key = parsed_tree.root_key))]
+#[instrument(skip(sources, dirs, handlebars))]
 fn generate_tree(
     sources: &Sources,
     dirs: &Dirs,
     handlebars: &Handlebars,
-    parsed_tree: &ParsedTree,
+    file_id: FileId,
 ) -> anyhow::Result<String> {
-    let breadcrumbs = breadcrumbs_to_html(
+    let breadcrumbs = breadcrumbs_to_html(&sources.config, &sources.navigation_map, file_id);
-        &sources.config,
-        &sources.navigation_map,
-        &parsed_tree.root_key,
-    );
 
     let roots = sources
         .treehouse
         .roots
-        .get(&parsed_tree.root_key)
+        .get(&file_id)
         .expect("tree should have been added to the treehouse");
 
     let tree = {
@@ -232,7 +219,7 @@ fn generate_tree(
             &sources.treehouse,
             &sources.config,
             dirs,
-            parsed_tree.file_id,
+            file_id,
             &roots.branches,
         );
         tree
@@ -261,10 +248,7 @@ fn generate_tree(
             scripts: roots.attributes.scripts.clone(),
             styles: roots.attributes.styles.clone(),
             breadcrumbs,
-            tree_path: sources
+            tree_path: sources.treehouse.tree_path(file_id).map(|s| s.to_string()),
-                .treehouse
-                .tree_path(parsed_tree.file_id)
-                .map(|s| s.to_owned()),
             tree,
         },
     };
@@ -289,9 +273,9 @@ fn generate_tree_or_error(
     sources: &Sources,
     dirs: &Dirs,
     handlebars: &Handlebars,
-    parsed_tree: &ParsedTree,
+    file_id: FileId,
 ) -> String {
-    match generate_tree(sources, dirs, handlebars, parsed_tree) {
+    match generate_tree(sources, dirs, handlebars, file_id) {
         Ok(html) => html,
         Err(error) => format!("error: {error:?}"),
     }
@@ -300,7 +284,6 @@ fn generate_tree_or_error(
 pub struct Sources {
     pub config: Config,
     pub treehouse: Treehouse,
-    pub parsed_trees: HashMap<VPathBuf, ParsedTree>,
     pub navigation_map: NavigationMap,
     pub import_map: ImportMap,
 }
@@ -320,12 +303,15 @@ impl Sources {
             config.site = std::env::var("TREEHOUSE_SITE").unwrap_or(config.site);
             config.autopopulate_emoji(&*dirs.emoji)?;
             config.autopopulate_pics(&*dirs.pic)?;
-            config.load_syntaxes(&*dirs.syntax)?;
+            config.load_syntaxes(dirs.syntax.clone())?;
             config
         };
 
-        let (treehouse, parsed_trees) = parse_trees(&config, dirs)?;
+        let treehouse = load_trees(&config, dirs)?;
-        let navigation_map = NavigationMap::build(&treehouse, "index");
+        let navigation_map = NavigationMap::build(
+            &treehouse,
+            treehouse.files_by_tree_path[VPath::new("index")],
+        );
         let import_map = ImportMap::generate(
             &config.site,
             &Cd::new(dirs.static_.clone(), VPathBuf::new("js")),
@@ -335,7 +321,6 @@ impl Sources {
         Ok(Sources {
             config,
             treehouse,
-            parsed_trees,
             navigation_map,
             import_map,
         })
@@ -429,11 +414,11 @@ impl Dir for TreehouseDir {
         };
 
         self.sources
-            .parsed_trees
+            .treehouse
+            .files_by_tree_path
             .get(path)
-            .map(|parsed_tree| {
+            .map(|&file_id| {
-                generate_tree_or_error(&self.sources, &self.dirs, &self.handlebars, parsed_tree)
+                generate_tree_or_error(&self.sources, &self.dirs, &self.handlebars, file_id).into()
-                    .into()
             })
             .or_else(|| {
                 if path.file_name().is_some_and(|s| !s.starts_with('_')) {
@@ -525,7 +510,7 @@ pub fn target(dirs: Arc<Dirs>, sources: Arc<Sources>) -> DynDir {
         Cd::new(dirs.static_.clone(), VPathBuf::new("robots.txt")).to_dyn(),
     );
 
-    let dir_index = DirIndex::new(sources.parsed_trees.keys().map(|x| &**x));
+    let dir_index = DirIndex::new(sources.treehouse.files_by_tree_path.keys().map(|x| &**x));
     let tree_view = TreehouseDir::new(dirs, sources, dir_index);
 
     let tree_view = ContentCache::new(tree_view);

crates/treehouse/src/html/breadcrumbs.rs

@@ -2,7 +2,7 @@ use std::{borrow::Cow, fmt::Write};
 
 use tracing::instrument;
 
-use crate::config::Config;
+use crate::{config::Config, state::FileId, vfs::VPath};
 
 use super::{navmap::NavigationMap, EscapeAttribute};
 
@@ -10,26 +10,31 @@ use super::{navmap::NavigationMap, EscapeAttribute};
 pub fn breadcrumbs_to_html(
     config: &Config,
     navigation_map: &NavigationMap,
-    tree_path: &str,
+    file_id: FileId,
 ) -> String {
     let mut s = String::new();
 
-    if let Some(path) = navigation_map.paths.get(tree_path) {
+    if let Some(path) = navigation_map.paths.get(&file_id) {
         for (i, element) in path.iter().enumerate() {
             // Skip the index because it's implied by the logo on the left.
-            if element != "index" {
+            if &**element != VPath::new_const("index") {
                 s.push_str("<li class=\"breadcrumb\">");
                 {
                     let short_element = path
                         .get(i - 1)
                         .map(|p| format!("{p}/"))
-                        .and_then(|prefix| element.strip_prefix(prefix.as_str()).map(Cow::Borrowed))
+                        .and_then(|prefix| {
+                            element
+                                .as_str()
+                                .strip_prefix(prefix.as_str())
+                                .map(Cow::Borrowed)
+                        })
                         .unwrap_or_else(|| Cow::Owned(format!("/{element}")));
                     write!(
                         s,
                         "<a href=\"{site}/{element}\">{short_element}</a>",
                         site = EscapeAttribute(&config.site),
-                        element = EscapeAttribute(element)
+                        element = EscapeAttribute(element.as_str())
                     )
                     .unwrap();
                 }

crates/treehouse/src/html/navmap.rs

@@ -3,20 +3,23 @@ use std::collections::HashMap;
 use tracing::instrument;
 
 use crate::{
-    state::Treehouse,
+    state::{FileId, Treehouse},
     tree::{attributes::Content, SemaBranchId},
+    vfs::VPathBuf,
 };
 
 #[derive(Debug, Clone, Default)]
 struct NavigationMapBuilder {
-    stack: Vec<String>,
+    stack: Vec<VPathBuf>,
     navigation_map: NavigationMap,
 }
 
 impl NavigationMapBuilder {
-    fn enter_tree(&mut self, tree: String) {
+    fn enter_tree(&mut self, file_id: FileId, tree_path: VPathBuf) {
-        self.stack.push(tree.clone());
+        self.stack.push(tree_path.clone());
-        self.navigation_map.paths.insert(tree, self.stack.clone());
+        self.navigation_map
+            .paths
+            .insert(file_id, self.stack.clone());
     }
 
     fn exit_tree(&mut self) {
@@ -31,12 +34,12 @@ impl NavigationMapBuilder {
 #[derive(Debug, Clone, Default)]
 pub struct NavigationMap {
     /// Tells you which pages need to be opened to get to the key.
-    pub paths: HashMap<String, Vec<String>>,
+    pub paths: HashMap<FileId, Vec<VPathBuf>>,
 }
 
 impl NavigationMap {
     #[instrument(name = "NavigationMap::build", skip(treehouse))]
-    pub fn build(treehouse: &Treehouse, root_tree_path: &str) -> Self {
+    pub fn build(treehouse: &Treehouse, root_file_id: FileId) -> Self {
         let mut builder = NavigationMapBuilder::default();
 
         fn rec_branch(
@@ -45,8 +48,8 @@ impl NavigationMap {
             branch_id: SemaBranchId,
         ) {
             let branch = treehouse.tree.branch(branch_id);
-            if let Content::Link(linked) = &branch.attributes.content {
+            if let Content::ResolvedLink(linked) = &branch.attributes.content {
-                rec_tree(treehouse, builder, linked);
+                rec_tree(treehouse, builder, *linked);
             } else {
                 for &child_id in &branch.children {
                     rec_branch(treehouse, builder, child_id);
@@ -54,12 +57,18 @@ impl NavigationMap {
             }
         }
 
-        fn rec_tree(treehouse: &Treehouse, builder: &mut NavigationMapBuilder, tree_path: &str) {
+        fn rec_tree(treehouse: &Treehouse, builder: &mut NavigationMapBuilder, file_id: FileId) {
-            if let Some(roots) = treehouse.roots.get(tree_path) {
+            if let Some(roots) = treehouse.roots.get(&file_id) {
                 // Pages can link to each other causing infinite recursion, so we need to handle that
                 // case by skipping pages that already have been analyzed.
-                if !builder.navigation_map.paths.contains_key(tree_path) {
+                if !builder.navigation_map.paths.contains_key(&file_id) {
-                    builder.enter_tree(tree_path.to_owned());
+                    builder.enter_tree(
+                        file_id,
+                        treehouse
+                            .tree_path(file_id)
+                            .expect("tree files may only link to other tree files")
+                            .to_owned(),
+                    );
                     for &branch_id in &roots.branches {
                         rec_branch(treehouse, builder, branch_id);
                     }
@@ -68,7 +77,7 @@ impl NavigationMap {
             }
         }
 
-        rec_tree(treehouse, &mut builder, root_tree_path);
+        rec_tree(treehouse, &mut builder, root_file_id);
 
         builder.finish()
     }

crates/treehouse/src/html/tree.rs

@@ -30,8 +30,8 @@ pub fn branch_to_html(
         return;
     }
 
-    let has_children =
+    let has_children = !branch.children.is_empty()
-        !branch.children.is_empty() || matches!(branch.attributes.content, Content::Link(_));
+        || matches!(branch.attributes.content, Content::ResolvedLink(_));
 
     let class = if has_children { "branch" } else { "leaf" };
     let mut class = String::from(class);
@@ -44,7 +44,7 @@ pub fn branch_to_html(
         class.push_str(" draft");
     }
 
-    let component = if let Content::Link(_) = branch.attributes.content {
+    let component = if let Content::ResolvedLink(_) = branch.attributes.content {
         "b-linked"
     } else {
         "b"
@@ -55,8 +55,9 @@ pub fn branch_to_html(
         Cow::Borrowed(component)
     };
 
-    let linked_branch = if let Content::Link(link) = &branch.attributes.content {
+    let linked_branch = if let Content::ResolvedLink(file_id) = &branch.attributes.content {
-        format!(" data-th-link=\"{}\"", EscapeHtml(link))
+        let path = treehouse.tree_path(*file_id).expect(".tree file expected");
+        format!(" data-th-link=\"{}\"", EscapeHtml(path.as_str()))
     } else {
         String::new()
     };
@@ -126,7 +127,7 @@ pub fn branch_to_html(
             page_id: treehouse
                 .tree_path(file_id)
                 .expect(".tree file expected")
-                .to_owned(),
+                .to_string(),
 
             config,
             dirs,
@@ -137,13 +138,14 @@ pub fn branch_to_html(
         .render(&events, s);
 
         let branch = treehouse.tree.branch(branch_id);
-        if let Content::Link(link) = &branch.attributes.content {
+        if let Content::ResolvedLink(file_id) = &branch.attributes.content {
+            let path = treehouse.tree_path(*file_id).expect(".tree file expected");
             write!(
                 s,
                 "<noscript><a class=\"navigate icon-go\" href=\"{}/{}\">Go to linked tree: <code>{}</code></a></noscript>",
                 EscapeAttribute(&config.site),
-                EscapeAttribute(link),
+                EscapeAttribute(path.as_str()),
-                EscapeHtml(link),
+                EscapeHtml(path.as_str()),
             )
             .unwrap();
         }
@@ -151,12 +153,13 @@ pub fn branch_to_html(
 
         s.push_str("<th-bb>");
         {
-            if let Content::Link(link) = &branch.attributes.content {
+            if let Content::ResolvedLink(file_id) = &branch.attributes.content {
+                let path = treehouse.tree_path(*file_id).expect(".tree file expected");
                 write!(
                     s,
                     "<a class=\"icon icon-go\" href=\"{}/{}\" title=\"linked tree\"></a>",
                     EscapeAttribute(&config.site),
-                    EscapeAttribute(link),
+                    EscapeAttribute(path.as_str()),
                 )
                 .unwrap();
             } else {

crates/treehouse/src/parse.rs

@@ -8,12 +8,11 @@ use crate::state::{toml_error_to_diagnostic, FileId, TomlError, Treehouse};
 
 pub struct ErrorsEmitted;
 
-#[instrument(skip(treehouse))]
+#[instrument(skip(input))]
 pub fn parse_tree_with_diagnostics(
-    treehouse: &mut Treehouse,
     file_id: FileId,
+    input: &str,
 ) -> Result<Roots, Vec<Diagnostic<FileId>>> {
-    let input = &treehouse.source(file_id).input();
     Roots::parse(&mut treehouse_format::pull::Parser { input, position: 0 }).map_err(|error| {
         vec![Diagnostic {
             severity: Severity::Error,

crates/treehouse/src/state.rs

@@ -3,23 +3,19 @@ use std::{collections::HashMap, ops::Range};
 use anyhow::Context;
 use codespan_reporting::{
     diagnostic::{Diagnostic, Label, LabelStyle, Severity},
-    files::SimpleFiles,
     term::termcolor::{ColorChoice, StandardStream},
 };
+use tracing::instrument;
 use ulid::Ulid;
 
 use crate::{
     tree::{SemaBranchId, SemaRoots, SemaTree},
-    vfs::VPathBuf,
+    vfs::{VPath, VPathBuf},
 };
 
 #[derive(Debug, Clone)]
 pub enum Source {
-    Tree {
+    Tree { input: String, tree_path: VPathBuf },
-        input: String,
-        tree_path: String,
-        target_path: VPathBuf,
-    },
     Other(String),
 }
 
@@ -38,26 +34,54 @@ impl AsRef<str> for Source {
     }
 }
 
-pub type Files = SimpleFiles<String, Source>;
+#[derive(Debug, Clone)]
-pub type FileId = <Files as codespan_reporting::files::Files<'static>>::FileId;
+pub struct File {
+    pub path: VPathBuf,
+    pub source: Source,
+    pub line_starts: Vec<usize>,
+}
+
+impl File {
+    fn line_start(&self, line_index: usize) -> Result<usize, codespan_reporting::files::Error> {
+        use std::cmp::Ordering;
+
+        match line_index.cmp(&self.line_starts.len()) {
+            Ordering::Less => Ok(self
+                .line_starts
+                .get(line_index)
+                .cloned()
+                .expect("failed despite previous check")),
+            Ordering::Equal => Ok(self.source.as_ref().len()),
+            Ordering::Greater => Err(codespan_reporting::files::Error::LineTooLarge {
+                given: line_index,
+                max: self.line_starts.len() - 1,
+            }),
+        }
+    }
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Hash)]
+pub struct FileId(usize);
 
 /// Treehouse compilation context.
 pub struct Treehouse {
-    pub files: Files,
+    pub files: Vec<File>,
+    pub files_by_tree_path: HashMap<VPathBuf, FileId>,
 
     pub tree: SemaTree,
     pub branches_by_named_id: HashMap<String, SemaBranchId>,
-    pub roots: HashMap<String, SemaRoots>,
+    pub roots: HashMap<FileId, SemaRoots>,
 
     pub branch_redirects: HashMap<String, SemaBranchId>,
 
-    missingno_generator: ulid::Generator,
+    pub missingno_generator: ulid::Generator,
 }
 
 impl Treehouse {
     pub fn new() -> Self {
         Self {
-            files: Files::new(),
+            files: vec![],
+            files_by_tree_path: HashMap::new(),
 
             tree: SemaTree::default(),
             branches_by_named_id: HashMap::new(),
@@ -69,27 +93,34 @@ impl Treehouse {
         }
     }
 
-    pub fn add_file(&mut self, filename: String, source: Source) -> FileId {
+    pub fn add_file(&mut self, path: VPathBuf, source: Source) -> FileId {
-        self.files.add(filename, source)
+        let id = FileId(self.files.len());
+        self.files.push(File {
+            line_starts: codespan_reporting::files::line_starts(source.input()).collect(),
+
+            path,
+            source,
+        });
+        id
+    }
+
+    /// Get the name of a file, assuming it was previously registered.
+    pub fn path(&self, file_id: FileId) -> &VPath {
+        &self.files[file_id.0].path
     }
 
     /// Get the source code of a file, assuming it was previously registered.
     pub fn source(&self, file_id: FileId) -> &Source {
-        self.files
+        &self.files[file_id.0].source
-            .get(file_id)
-            .expect("file should have been registered previously")
-            .source()
     }
 
-    /// Get the name of a file, assuming it was previously registered.
+    pub fn set_source(&mut self, file_id: FileId, source: Source) {
-    pub fn filename(&self, file_id: FileId) -> &str {
+        self.files[file_id.0].line_starts =
-        self.files
+            codespan_reporting::files::line_starts(source.input()).collect();
-            .get(file_id)
+        self.files[file_id.0].source = source;
-            .expect("file should have been registered previously")
-            .name()
     }
 
-    pub fn tree_path(&self, file_id: FileId) -> Option<&str> {
+    pub fn tree_path(&self, file_id: FileId) -> Option<&VPath> {
         match self.source(file_id) {
             Source::Tree { tree_path, .. } => Some(tree_path),
             Source::Other(_) => None,
@@ -109,6 +140,49 @@ impl Default for Treehouse {
     }
 }
 
+impl<'a> codespan_reporting::files::Files<'a> for Treehouse {
+    type FileId = FileId;
+
+    type Name = &'a VPath;
+
+    type Source = &'a str;
+
+    fn name(&'a self, id: Self::FileId) -> Result<Self::Name, codespan_reporting::files::Error> {
+        Ok(self.path(id))
+    }
+
+    fn source(
+        &'a self,
+        id: Self::FileId,
+    ) -> Result<Self::Source, codespan_reporting::files::Error> {
+        Ok(self.source(id).input())
+    }
+
+    fn line_index(
+        &'a self,
+        id: Self::FileId,
+        byte_index: usize,
+    ) -> Result<usize, codespan_reporting::files::Error> {
+        let file = &self.files[id.0];
+        Ok(file
+            .line_starts
+            .binary_search(&byte_index)
+            .unwrap_or_else(|next_line| next_line - 1))
+    }
+
+    fn line_range(
+        &'a self,
+        id: Self::FileId,
+        line_index: usize,
+    ) -> Result<Range<usize>, codespan_reporting::files::Error> {
+        let file = &self.files[id.0];
+        let line_start = file.line_start(line_index)?;
+        let next_line_start = file.line_start(line_index + 1)?;
+
+        Ok(line_start..next_line_start)
+    }
+}
+
 pub struct TomlError {
     pub message: String,
     pub span: Option<Range<usize>>,
@@ -135,7 +209,11 @@ pub fn toml_error_to_diagnostic(error: TomlError) -> Diagnostic<FileId> {
     }
 }
 
-pub fn report_diagnostics(files: &Files, diagnostics: &[Diagnostic<FileId>]) -> anyhow::Result<()> {
+#[instrument(skip(files, diagnostics))]
+pub fn report_diagnostics(
+    files: &Treehouse,
+    diagnostics: &[Diagnostic<FileId>],
+) -> anyhow::Result<()> {
     let writer = StandardStream::stderr(ColorChoice::Auto);
     let config = codespan_reporting::term::Config::default();
     for diagnostic in diagnostics {

crates/treehouse/src/tree.rs

@@ -96,7 +96,7 @@ impl SemaRoots {
 
         if successfully_parsed && attributes.title.is_empty() {
             attributes.title = match treehouse.source(file_id) {
-                Source::Tree { tree_path, .. } => tree_path.clone(),
+                Source::Tree { tree_path, .. } => tree_path.to_string(),
                 _ => panic!("parse_attributes called for a non-.tree file"),
             }
         }
@@ -310,7 +310,7 @@ impl SemaBranch {
                             "note: a generated id `{}` will be used, but this id is unstable and will not persist across generations",
                             attributes.id
                         ),
-                        format!("help: run `treehouse fix {}` to add missing ids to branches", treehouse.filename(file_id)),
+                        format!("help: run `treehouse fix {}` to add missing ids to branches", treehouse.path(file_id)),
                     ],
                 });
             }
@@ -334,6 +334,26 @@ impl SemaBranch {
                     });
                 }
             }
+
+            // Resolve content.links.
+            if let Content::Link(tree_path) = &attributes.content {
+                if let Some(file_id) = treehouse.files_by_tree_path.get(tree_path) {
+                    attributes.content = Content::ResolvedLink(*file_id);
+                } else {
+                    diagnostics.push(Diagnostic {
+                        severity: Severity::Error,
+                        code: Some("attr".into()),
+                        message: format!("linked tree `{tree_path}` does not exist"),
+                        labels: vec![Label {
+                            style: LabelStyle::Primary,
+                            file_id,
+                            range: attribute_warning_span.clone(),
+                            message: "".into(),
+                        }],
+                        notes: vec![],
+                    })
+                }
+            }
         }
         attributes
     }

crates/treehouse/src/tree/attributes.rs

@@ -2,6 +2,8 @@ use std::collections::HashMap;
 
 use serde::{Deserialize, Serialize};
 
+use crate::{state::FileId, vfs::VPathBuf};
+
 /// Top-level `%%` root attributes.
 #[derive(Debug, Default, Clone, PartialEq, Eq, Deserialize, Serialize)]
 pub struct RootAttributes {
@@ -118,7 +120,12 @@ pub enum Content {
     ///
     /// Note that `Link` branches must not contain any children. If a `Link` branch does contain
     /// children, an `attribute`-type error is raised.
-    Link(String),
+    Link(VPathBuf),
+
+    /// Valid link to another tree.
+    /// This replaces `Content::Link` during semantic analysis.
+    #[serde(skip)]
+    ResolvedLink(FileId),
 }
 
 #[derive(Debug, Default, Clone, PartialEq, Eq, Deserialize)]