use std::mem; use rustc_data_structures::fx::{FxHashMap, FxHashSet}; use rustc_hir::def_id::{CrateNum, DefId}; use rustc_middle::ty::{self, TyCtxt}; use rustc_span::Symbol; use crate::clean::{self, types::ExternalLocation, ExternalCrate, ItemId, PrimitiveType}; use crate::core::DocContext; use crate::fold::DocFolder; use crate::formats::item_type::ItemType; use crate::formats::Impl; use crate::html::format::join_with_double_colon; use crate::html::markdown::short_markdown_summary; use crate::html::render::search_index::get_function_type_for_search; use crate::html::render::IndexItem; use crate::visit_lib::RustdocEffectiveVisibilities; /// This cache is used to store information about the [`clean::Crate`] being /// rendered in order to provide more useful documentation. This contains /// information like all implementors of a trait, all traits a type implements, /// documentation for all known traits, etc. /// /// This structure purposefully does not implement `Clone` because it's intended /// to be a fairly large and expensive structure to clone. Instead this adheres /// to `Send` so it may be stored in an `Arc` instance and shared among the various /// rendering threads. #[derive(Default)] pub(crate) struct Cache { /// Maps a type ID to all known implementations for that type. This is only /// recognized for intra-crate [`clean::Type::Path`]s, and is used to print /// out extra documentation on the page of an enum/struct. /// /// The values of the map are a list of implementations and documentation /// found on that implementation. pub(crate) impls: FxHashMap>, /// Maintains a mapping of local crate `DefId`s to the fully qualified name /// and "short type description" of that node. This is used when generating /// URLs when a type is being linked to. External paths are not located in /// this map because the `External` type itself has all the information /// necessary. pub(crate) paths: FxHashMap, ItemType)>, /// Similar to `paths`, but only holds external paths. This is only used for /// generating explicit hyperlinks to other crates. pub(crate) external_paths: FxHashMap, ItemType)>, /// Maps local `DefId`s of exported types to fully qualified paths. /// Unlike 'paths', this mapping ignores any renames that occur /// due to 'use' statements. /// /// This map is used when writing out the special 'implementors' /// javascript file. By using the exact path that the type /// is declared with, we ensure that each path will be identical /// to the path used if the corresponding type is inlined. By /// doing this, we can detect duplicate impls on a trait page, and only display /// the impl for the inlined type. pub(crate) exact_paths: FxHashMap>, /// This map contains information about all known traits of this crate. /// Implementations of a crate should inherit the documentation of the /// parent trait if no extra documentation is specified, and default methods /// should show up in documentation about trait implementations. pub(crate) traits: FxHashMap, /// When rendering traits, it's often useful to be able to list all /// implementors of the trait, and this mapping is exactly, that: a mapping /// of trait ids to the list of known implementors of the trait pub(crate) implementors: FxHashMap>, /// Cache of where external crate documentation can be found. pub(crate) extern_locations: FxHashMap, /// Cache of where documentation for primitives can be found. pub(crate) primitive_locations: FxHashMap, // Note that external items for which `doc(hidden)` applies to are shown as // non-reachable while local items aren't. This is because we're reusing // the effective visibilities from the privacy check pass. pub(crate) effective_visibilities: RustdocEffectiveVisibilities, /// The version of the crate being documented, if given from the `--crate-version` flag. pub(crate) crate_version: Option, /// Whether to document private items. /// This is stored in `Cache` so it doesn't need to be passed through all rustdoc functions. pub(crate) document_private: bool, /// Crates marked with [`#[doc(masked)]`][doc_masked]. /// /// [doc_masked]: https://doc.rust-lang.org/nightly/unstable-book/language-features/doc-masked.html pub(crate) masked_crates: FxHashSet, // Private fields only used when initially crawling a crate to build a cache stack: Vec, parent_stack: Vec, stripped_mod: bool, pub(crate) search_index: Vec, // In rare case where a structure is defined in one module but implemented // in another, if the implementing module is parsed before defining module, // then the fully qualified name of the structure isn't presented in `paths` // yet when its implementation methods are being indexed. Caches such methods // and their parent id here and indexes them at the end of crate parsing. pub(crate) orphan_impl_items: Vec, // Similarly to `orphan_impl_items`, sometimes trait impls are picked up // even though the trait itself is not exported. This can happen if a trait // was defined in function/expression scope, since the impl will be picked // up by `collect-trait-impls` but the trait won't be scraped out in the HIR // crawl. In order to prevent crashes when looking for notable traits or // when gathering trait documentation on a type, hold impls here while // folding and add them to the cache later on if we find the trait. orphan_trait_impls: Vec<(DefId, FxHashSet, Impl)>, /// All intra-doc links resolved so far. /// /// Links are indexed by the DefId of the item they document. pub(crate) intra_doc_links: FxHashMap>, /// Cfg that have been hidden via #![doc(cfg_hide(...))] pub(crate) hidden_cfg: FxHashSet, } /// This struct is used to wrap the `cache` and `tcx` in order to run `DocFolder`. struct CacheBuilder<'a, 'tcx> { cache: &'a mut Cache, /// This field is used to prevent duplicated impl blocks. impl_ids: FxHashMap>, tcx: TyCtxt<'tcx>, } impl Cache { pub(crate) fn new(document_private: bool) -> Self { Cache { document_private, ..Cache::default() } } /// Populates the `Cache` with more data. The returned `Crate` will be missing some data that was /// in `krate` due to the data being moved into the `Cache`. pub(crate) fn populate(cx: &mut DocContext<'_>, mut krate: clean::Crate) -> clean::Crate { let tcx = cx.tcx; // Crawl the crate to build various caches used for the output debug!(?cx.cache.crate_version); cx.cache.traits = krate.external_traits.take(); // Cache where all our extern crates are located // FIXME: this part is specific to HTML so it'd be nice to remove it from the common code for &crate_num in cx.tcx.crates(()) { let e = ExternalCrate { crate_num }; let name = e.name(tcx); let render_options = &cx.render_options; let extern_url = render_options.extern_html_root_urls.get(name.as_str()).map(|u| &**u); let extern_url_takes_precedence = render_options.extern_html_root_takes_precedence; let dst = &render_options.output; let location = e.location(extern_url, extern_url_takes_precedence, dst, tcx); cx.cache.extern_locations.insert(e.crate_num, location); cx.cache.external_paths.insert(e.def_id(), (vec![name], ItemType::Module)); } // FIXME: avoid this clone (requires implementing Default manually) cx.cache.primitive_locations = PrimitiveType::primitive_locations(tcx).clone(); for (prim, &def_id) in &cx.cache.primitive_locations { let crate_name = tcx.crate_name(def_id.krate); // Recall that we only allow primitive modules to be at the root-level of the crate. // If that restriction is ever lifted, this will have to include the relative paths instead. cx.cache .external_paths .insert(def_id, (vec![crate_name, prim.as_sym()], ItemType::Primitive)); } let (krate, mut impl_ids) = { let mut cache_builder = CacheBuilder { tcx, cache: &mut cx.cache, impl_ids: FxHashMap::default() }; krate = cache_builder.fold_crate(krate); (krate, cache_builder.impl_ids) }; for (trait_did, dids, impl_) in cx.cache.orphan_trait_impls.drain(..) { if cx.cache.traits.contains_key(&trait_did) { for did in dids { if impl_ids.entry(did).or_default().insert(impl_.def_id()) { cx.cache.impls.entry(did).or_default().push(impl_.clone()); } } } } krate } } impl<'a, 'tcx> DocFolder for CacheBuilder<'a, 'tcx> { fn fold_item(&mut self, item: clean::Item) -> Option { if item.item_id.is_local() { debug!("folding {} \"{:?}\", id {:?}", item.type_(), item.name, item.item_id); } // If this is a stripped module, // we don't want it or its children in the search index. let orig_stripped_mod = match *item.kind { clean::StrippedItem(box clean::ModuleItem(..)) => { mem::replace(&mut self.cache.stripped_mod, true) } _ => self.cache.stripped_mod, }; // If the impl is from a masked crate or references something from a // masked crate then remove it completely. if let clean::ImplItem(ref i) = *item.kind { if self.cache.masked_crates.contains(&item.item_id.krate()) || i.trait_ .as_ref() .map_or(false, |t| self.cache.masked_crates.contains(&t.def_id().krate)) || i.for_ .def_id(self.cache) .map_or(false, |d| self.cache.masked_crates.contains(&d.krate)) { return None; } } // Propagate a trait method's documentation to all implementors of the // trait. if let clean::TraitItem(ref t) = *item.kind { self.cache.traits.entry(item.item_id.expect_def_id()).or_insert_with(|| (**t).clone()); } // Collect all the implementors of traits. if let clean::ImplItem(ref i) = *item.kind { if let Some(trait_) = &i.trait_ { if !i.kind.is_blanket() { self.cache .implementors .entry(trait_.def_id()) .or_default() .push(Impl { impl_item: item.clone() }); } } } // Index this method for searching later on. if let Some(s) = item.name.or_else(|| { if item.is_stripped() { None } else if let clean::ImportItem(ref i) = *item.kind && let clean::ImportKind::Simple(s) = i.kind { Some(s) } else { None } }) { let (parent, is_inherent_impl_item) = match *item.kind { clean::StrippedItem(..) => ((None, None), false), clean::AssocConstItem(..) | clean::AssocTypeItem(..) if self .cache .parent_stack .last() .map_or(false, |parent| parent.is_trait_impl()) => { // skip associated items in trait impls ((None, None), false) } clean::TyMethodItem(..) | clean::TyAssocConstItem(..) | clean::TyAssocTypeItem(..) | clean::StructFieldItem(..) | clean::VariantItem(..) => ( ( Some( self.cache .parent_stack .last() .expect("parent_stack is empty") .item_id() .expect_def_id(), ), Some(&self.cache.stack[..self.cache.stack.len() - 1]), ), false, ), clean::MethodItem(..) | clean::AssocConstItem(..) | clean::AssocTypeItem(..) => { if self.cache.parent_stack.is_empty() { ((None, None), false) } else { let last = self.cache.parent_stack.last().expect("parent_stack is empty 2"); let did = match &*last { ParentStackItem::Impl { for_, .. } => for_.def_id(&self.cache), ParentStackItem::Type(item_id) => item_id.as_def_id(), }; let path = match did.and_then(|did| self.cache.paths.get(&did)) { // The current stack not necessarily has correlation // for where the type was defined. On the other // hand, `paths` always has the right // information if present. Some((fqp, _)) => Some(&fqp[..fqp.len() - 1]), None => None, }; ((did, path), true) } } _ => ((None, Some(&*self.cache.stack)), false), }; match parent { (parent, Some(path)) if is_inherent_impl_item || !self.cache.stripped_mod => { debug_assert!(!item.is_stripped()); // A crate has a module at its root, containing all items, // which should not be indexed. The crate-item itself is // inserted later on when serializing the search-index. if item.item_id.as_def_id().map_or(false, |idx| !idx.is_crate_root()) { let desc = item.doc_value().map_or_else(String::new, |x| { short_markdown_summary(x.as_str(), &item.link_names(self.cache)) }); let ty = item.type_(); if ty != ItemType::StructField || u16::from_str_radix(s.as_str(), 10).is_err() { // In case this is a field from a tuple struct, we don't add it into // the search index because its name is something like "0", which is // not useful for rustdoc search. self.cache.search_index.push(IndexItem { ty, name: s, path: join_with_double_colon(path), desc, parent, parent_idx: None, search_type: get_function_type_for_search( &item, self.tcx, clean_impl_generics(self.cache.parent_stack.last()).as_ref(), self.cache, ), aliases: item.attrs.get_doc_aliases(), }); } } } (Some(parent), None) if is_inherent_impl_item => { // We have a parent, but we don't know where they're // defined yet. Wait for later to index this item. let impl_generics = clean_impl_generics(self.cache.parent_stack.last()); self.cache.orphan_impl_items.push(OrphanImplItem { parent, item: item.clone(), impl_generics, }); } _ => {} } } // Keep track of the fully qualified path for this item. let pushed = match item.name { Some(n) if !n.is_empty() => { self.cache.stack.push(n); true } _ => false, }; match *item.kind { clean::StructItem(..) | clean::EnumItem(..) | clean::TypedefItem(..) | clean::TraitItem(..) | clean::TraitAliasItem(..) | clean::FunctionItem(..) | clean::ModuleItem(..) | clean::ForeignFunctionItem(..) | clean::ForeignStaticItem(..) | clean::ConstantItem(..) | clean::StaticItem(..) | clean::UnionItem(..) | clean::ForeignTypeItem | clean::MacroItem(..) | clean::ProcMacroItem(..) | clean::VariantItem(..) => { if !self.cache.stripped_mod { // Re-exported items mean that the same id can show up twice // in the rustdoc ast that we're looking at. We know, // however, that a re-exported item doesn't show up in the // `public_items` map, so we can skip inserting into the // paths map if there was already an entry present and we're // not a public item. if !self.cache.paths.contains_key(&item.item_id.expect_def_id()) || self .cache .effective_visibilities .is_directly_public(self.tcx, item.item_id.expect_def_id()) { self.cache.paths.insert( item.item_id.expect_def_id(), (self.cache.stack.clone(), item.type_()), ); } } } clean::PrimitiveItem(..) => { self.cache .paths .insert(item.item_id.expect_def_id(), (self.cache.stack.clone(), item.type_())); } clean::ExternCrateItem { .. } | clean::ImportItem(..) | clean::OpaqueTyItem(..) | clean::ImplItem(..) | clean::TyMethodItem(..) | clean::MethodItem(..) | clean::StructFieldItem(..) | clean::TyAssocConstItem(..) | clean::AssocConstItem(..) | clean::TyAssocTypeItem(..) | clean::AssocTypeItem(..) | clean::StrippedItem(..) | clean::KeywordItem => { // FIXME: Do these need handling? // The person writing this comment doesn't know. // So would rather leave them to an expert, // as at least the list is better than `_ => {}`. } } // Maintain the parent stack. let (item, parent_pushed) = match *item.kind { clean::TraitItem(..) | clean::EnumItem(..) | clean::ForeignTypeItem | clean::StructItem(..) | clean::UnionItem(..) | clean::VariantItem(..) | clean::ImplItem(..) => { self.cache.parent_stack.push(ParentStackItem::new(&item)); (self.fold_item_recur(item), true) } _ => (self.fold_item_recur(item), false), }; // Once we've recursively found all the generics, hoard off all the // implementations elsewhere. let ret = if let clean::Item { kind: box clean::ImplItem(ref i), .. } = item { // Figure out the id of this impl. This may map to a // primitive rather than always to a struct/enum. // Note: matching twice to restrict the lifetime of the `i` borrow. let mut dids = FxHashSet::default(); match i.for_ { clean::Type::Path { ref path } | clean::BorrowedRef { type_: box clean::Type::Path { ref path }, .. } => { dids.insert(path.def_id()); if let Some(generics) = path.generics() && let ty::Adt(adt, _) = self.tcx.type_of(path.def_id()).kind() && adt.is_fundamental() { for ty in generics { if let Some(did) = ty.def_id(self.cache) { dids.insert(did); } } } } clean::DynTrait(ref bounds, _) | clean::BorrowedRef { type_: box clean::DynTrait(ref bounds, _), .. } => { dids.insert(bounds[0].trait_.def_id()); } ref t => { let did = t .primitive_type() .and_then(|t| self.cache.primitive_locations.get(&t).cloned()); if let Some(did) = did { dids.insert(did); } } } if let Some(generics) = i.trait_.as_ref().and_then(|t| t.generics()) { for bound in generics { if let Some(did) = bound.def_id(self.cache) { dids.insert(did); } } } let impl_item = Impl { impl_item: item }; if impl_item.trait_did().map_or(true, |d| self.cache.traits.contains_key(&d)) { for did in dids { if self.impl_ids.entry(did).or_default().insert(impl_item.def_id()) { self.cache .impls .entry(did) .or_insert_with(Vec::new) .push(impl_item.clone()); } } } else { let trait_did = impl_item.trait_did().expect("no trait did"); self.cache.orphan_trait_impls.push((trait_did, dids, impl_item)); } None } else { Some(item) }; if pushed { self.cache.stack.pop().expect("stack already empty"); } if parent_pushed { self.cache.parent_stack.pop().expect("parent stack already empty"); } self.cache.stripped_mod = orig_stripped_mod; ret } } pub(crate) struct OrphanImplItem { pub(crate) parent: DefId, pub(crate) item: clean::Item, pub(crate) impl_generics: Option<(clean::Type, clean::Generics)>, } /// Information about trait and type parents is tracked while traversing the item tree to build /// the cache. /// /// We don't just store `Item` in there, because `Item` contains the list of children being /// traversed and it would be wasteful to clone all that. We also need the item id, so just /// storing `ItemKind` won't work, either. enum ParentStackItem { Impl { for_: clean::Type, trait_: Option, generics: clean::Generics, kind: clean::ImplKind, item_id: ItemId, }, Type(ItemId), } impl ParentStackItem { fn new(item: &clean::Item) -> Self { match &*item.kind { clean::ItemKind::ImplItem(box clean::Impl { for_, trait_, generics, kind, .. }) => { ParentStackItem::Impl { for_: for_.clone(), trait_: trait_.clone(), generics: generics.clone(), kind: kind.clone(), item_id: item.item_id, } } _ => ParentStackItem::Type(item.item_id), } } fn is_trait_impl(&self) -> bool { matches!(self, ParentStackItem::Impl { trait_: Some(..), .. }) } fn item_id(&self) -> ItemId { match self { ParentStackItem::Impl { item_id, .. } => *item_id, ParentStackItem::Type(item_id) => *item_id, } } } fn clean_impl_generics(item: Option<&ParentStackItem>) -> Option<(clean::Type, clean::Generics)> { if let Some(ParentStackItem::Impl { for_, generics, kind: clean::ImplKind::Normal, .. }) = item { Some((for_.clone(), generics.clone())) } else { None } }