rustdoc.rs

  1mod item;
  2mod to_markdown;
  3
  4pub use item::*;
  5pub use to_markdown::convert_rustdoc_to_markdown;
  6
  7use std::path::PathBuf;
  8use std::sync::Arc;
  9
 10use anyhow::{bail, Context, Result};
 11use async_trait::async_trait;
 12use collections::{HashSet, VecDeque};
 13use fs::Fs;
 14use futures::AsyncReadExt;
 15use http::{AsyncBody, HttpClient, HttpClientWithUrl};
 16
 17use crate::{IndexedDocsDatabase, IndexedDocsProvider, PackageName, ProviderId};
 18
 19#[derive(Debug, Clone, Copy)]
 20pub enum RustdocSource {
 21    /// The docs were sourced from Zed's rustdoc index.
 22    Index,
 23    /// The docs were sourced from local `cargo doc` output.
 24    Local,
 25    /// The docs were sourced from `docs.rs`.
 26    DocsDotRs,
 27}
 28
 29#[derive(Debug)]
 30struct RustdocItemWithHistory {
 31    pub item: RustdocItem,
 32    #[cfg(debug_assertions)]
 33    pub history: Vec<String>,
 34}
 35
 36#[async_trait]
 37pub trait RustdocProvider {
 38    async fn fetch_page(
 39        &self,
 40        package: &PackageName,
 41        item: Option<&RustdocItem>,
 42    ) -> Result<Option<String>>;
 43}
 44
 45pub struct RustdocIndexer {
 46    provider: Box<dyn RustdocProvider + Send + Sync + 'static>,
 47}
 48
 49impl RustdocIndexer {
 50    pub fn new(provider: Box<dyn RustdocProvider + Send + Sync + 'static>) -> Self {
 51        Self { provider }
 52    }
 53}
 54
 55#[async_trait]
 56impl IndexedDocsProvider for RustdocIndexer {
 57    fn id(&self) -> ProviderId {
 58        ProviderId::rustdoc()
 59    }
 60
 61    fn database_path(&self) -> PathBuf {
 62        paths::support_dir().join("docs/rust/rustdoc-db.1.mdb")
 63    }
 64
 65    async fn index(&self, package: PackageName, database: Arc<IndexedDocsDatabase>) -> Result<()> {
 66        let Some(package_root_content) = self.provider.fetch_page(&package, None).await? else {
 67            return Ok(());
 68        };
 69
 70        let (crate_root_markdown, items) =
 71            convert_rustdoc_to_markdown(package_root_content.as_bytes())?;
 72
 73        database
 74            .insert(package.to_string(), crate_root_markdown)
 75            .await?;
 76
 77        let mut seen_items = HashSet::from_iter(items.clone());
 78        let mut items_to_visit: VecDeque<RustdocItemWithHistory> =
 79            VecDeque::from_iter(items.into_iter().map(|item| RustdocItemWithHistory {
 80                item,
 81                #[cfg(debug_assertions)]
 82                history: Vec::new(),
 83            }));
 84
 85        while let Some(item_with_history) = items_to_visit.pop_front() {
 86            let item = &item_with_history.item;
 87
 88            let Some(result) = self
 89                .provider
 90                .fetch_page(&package, Some(&item))
 91                .await
 92                .with_context(|| {
 93                    #[cfg(debug_assertions)]
 94                    {
 95                        format!(
 96                            "failed to fetch {item:?}: {history:?}",
 97                            history = item_with_history.history
 98                        )
 99                    }
100
101                    #[cfg(not(debug_assertions))]
102                    {
103                        format!("failed to fetch {item:?}")
104                    }
105                })?
106            else {
107                continue;
108            };
109
110            let (markdown, referenced_items) = convert_rustdoc_to_markdown(result.as_bytes())?;
111
112            database
113                .insert(format!("{package}::{}", item.display()), markdown)
114                .await?;
115
116            let parent_item = item;
117            for mut item in referenced_items {
118                if seen_items.contains(&item) {
119                    continue;
120                }
121
122                seen_items.insert(item.clone());
123
124                item.path.extend(parent_item.path.clone());
125                match parent_item.kind {
126                    RustdocItemKind::Mod => {
127                        item.path.push(parent_item.name.clone());
128                    }
129                    _ => {}
130                }
131
132                items_to_visit.push_back(RustdocItemWithHistory {
133                    #[cfg(debug_assertions)]
134                    history: {
135                        let mut history = item_with_history.history.clone();
136                        history.push(item.url_path());
137                        history
138                    },
139                    item,
140                });
141            }
142        }
143
144        Ok(())
145    }
146}
147
148pub struct LocalProvider {
149    fs: Arc<dyn Fs>,
150    cargo_workspace_root: PathBuf,
151}
152
153impl LocalProvider {
154    pub fn new(fs: Arc<dyn Fs>, cargo_workspace_root: PathBuf) -> Self {
155        Self {
156            fs,
157            cargo_workspace_root,
158        }
159    }
160}
161
162#[async_trait]
163impl RustdocProvider for LocalProvider {
164    async fn fetch_page(
165        &self,
166        crate_name: &PackageName,
167        item: Option<&RustdocItem>,
168    ) -> Result<Option<String>> {
169        let mut local_cargo_doc_path = self.cargo_workspace_root.join("target/doc");
170        local_cargo_doc_path.push(crate_name.as_ref());
171        if let Some(item) = item {
172            local_cargo_doc_path.push(item.url_path());
173        } else {
174            local_cargo_doc_path.push("index.html");
175        }
176
177        let Ok(contents) = self.fs.load(&local_cargo_doc_path).await else {
178            return Ok(None);
179        };
180
181        Ok(Some(contents))
182    }
183}
184
185pub struct DocsDotRsProvider {
186    http_client: Arc<HttpClientWithUrl>,
187}
188
189impl DocsDotRsProvider {
190    pub fn new(http_client: Arc<HttpClientWithUrl>) -> Self {
191        Self { http_client }
192    }
193}
194
195#[async_trait]
196impl RustdocProvider for DocsDotRsProvider {
197    async fn fetch_page(
198        &self,
199        crate_name: &PackageName,
200        item: Option<&RustdocItem>,
201    ) -> Result<Option<String>> {
202        let version = "latest";
203        let path = format!(
204            "{crate_name}/{version}/{crate_name}{item_path}",
205            item_path = item
206                .map(|item| format!("/{}", item.url_path()))
207                .unwrap_or_default()
208        );
209
210        let mut response = self
211            .http_client
212            .get(
213                &format!("https://docs.rs/{path}"),
214                AsyncBody::default(),
215                true,
216            )
217            .await?;
218
219        let mut body = Vec::new();
220        response
221            .body_mut()
222            .read_to_end(&mut body)
223            .await
224            .context("error reading docs.rs response body")?;
225
226        if response.status().is_client_error() {
227            let text = String::from_utf8_lossy(body.as_slice());
228            bail!(
229                "status error {}, response: {text:?}",
230                response.status().as_u16()
231            );
232        }
233
234        Ok(Some(String::from_utf8(body)?))
235    }
236}