use std::collections::{HashMap, HashSet}; use std::path::{Path, PathBuf}; use slotmap::{DefaultKey, DenseSlotMap}; use front_matter::SortBy; use config::Config; use content::{Page, Section}; use sorting::{find_siblings, sort_pages_by_date, sort_pages_by_weight}; // Like vec! but for HashSet macro_rules! set { ( $( $x:expr ),* ) => { { let mut s = HashSet::new(); $( s.insert($x); )* s } }; } /// Houses everything about pages and sections /// Think of it as a database where each page and section has an id (Key here) /// that can be used to find the actual value /// Sections and pages can then refer to other elements by those keys, which are very cheap to /// copy. /// We can assume the keys are always existing as removing a page/section deletes all references /// to that key. #[derive(Debug)] pub struct Library { /// All the pages of the site pages: DenseSlotMap, /// All the sections of the site sections: DenseSlotMap, /// A mapping path -> key for pages so we can easily get their key pub paths_to_pages: HashMap, /// A mapping path -> key for sections so we can easily get their key pub paths_to_sections: HashMap, /// Whether we need to look for translations is_multilingual: bool, } impl Library { pub fn new(cap_pages: usize, cap_sections: usize, is_multilingual: bool) -> Self { Library { pages: DenseSlotMap::with_capacity(cap_pages), sections: DenseSlotMap::with_capacity(cap_sections), paths_to_pages: HashMap::with_capacity(cap_pages), paths_to_sections: HashMap::with_capacity(cap_sections), is_multilingual, } } /// Add a section and return its Key pub fn insert_section(&mut self, section: Section) -> DefaultKey { let path = section.file.path.clone(); let key = self.sections.insert(section); self.paths_to_sections.insert(path, key); key } /// Add a page and return its Key pub fn insert_page(&mut self, page: Page) -> DefaultKey { let path = page.file.path.clone(); let key = self.pages.insert(page); self.paths_to_pages.insert(path, key); key } pub fn pages(&self) -> &DenseSlotMap { &self.pages } pub fn pages_mut(&mut self) -> &mut DenseSlotMap { &mut self.pages } pub fn pages_values(&self) -> Vec<&Page> { self.pages.values().collect::>() } pub fn sections(&self) -> &DenseSlotMap { &self.sections } pub fn sections_mut(&mut self) -> &mut DenseSlotMap { &mut self.sections } pub fn sections_values(&self) -> Vec<&Section> { self.sections.values().collect::>() } /// Find out the direct subsections of each subsection if there are some /// as well as the pages for each section pub fn populate_sections(&mut self, config: &Config) { let root_path = self.sections.values().find(|s| s.is_index()).map(|s| s.file.parent.clone()).unwrap(); // We are going to get both the ancestors and grandparents for each section in one go let mut ancestors: HashMap> = HashMap::new(); let mut subsections: HashMap> = HashMap::new(); for section in self.sections.values_mut() { // Make sure the pages of a section are empty since we can call that many times on `serve` section.pages = vec![]; section.ignored_pages = vec![]; if let Some(ref grand_parent) = section.file.grand_parent { subsections // Using the original filename to work for multi-lingual sections .entry(grand_parent.join(§ion.file.filename)) .or_insert_with(|| vec![]) .push(section.file.path.clone()); } // Index has no ancestors, no need to go through it if section.is_index() { ancestors.insert(section.file.path.clone(), vec![]); continue; } let mut path = root_path.clone(); let root_key = self.paths_to_sections[&root_path.join(§ion.file.filename)]; // Index section is the first ancestor of every single section let mut parents = vec![root_key]; for component in §ion.file.components { path = path.join(component); // Skip itself if path == section.file.parent { continue; } if let Some(section_key) = self.paths_to_sections.get(&path.join(§ion.file.filename)) { parents.push(*section_key); } } ancestors.insert(section.file.path.clone(), parents); } for (key, page) in &mut self.pages { let parent_filename = if page.lang != config.default_language { format!("_index.{}.md", page.lang) } else { "_index.md".to_string() }; let mut parent_section_path = page.file.parent.join(&parent_filename); while let Some(section_key) = self.paths_to_sections.get(&parent_section_path) { let parent_is_transparent; // We need to get a reference to a section later so keep the scope of borrowing small { let section = self.sections.get_mut(*section_key).unwrap(); section.pages.push(key); parent_is_transparent = section.meta.transparent; } page.ancestors = ancestors.get(&parent_section_path).cloned().unwrap_or_else(|| vec![]); // Don't forget to push the actual parent page.ancestors.push(*section_key); // Find the page template if one of a parent has page_template set // Stops after the first one found, keep in mind page.ancestors // is [index, ..., parent] so we need to reverse it first if page.meta.template.is_none() { for ancestor in page.ancestors.iter().rev() { let s = self.sections.get(*ancestor).unwrap(); if s.meta.page_template.is_some() { page.meta.template = s.meta.page_template.clone(); break; } } } if !parent_is_transparent { break; } // We've added `_index(.{LANG})?.md` so if we are here so we need to go up twice match parent_section_path.clone().parent().unwrap().parent() { Some(parent) => parent_section_path = parent.join(&parent_filename), None => break, } } } self.populate_translations(); self.sort_sections_pages(); let sections = self.paths_to_sections.clone(); let mut sections_weight = HashMap::new(); for (key, section) in &self.sections { sections_weight.insert(key, section.meta.weight); } for section in self.sections.values_mut() { if let Some(ref children) = subsections.get(§ion.file.path) { let mut children: Vec<_> = children.iter().map(|p| sections[p]).collect(); children.sort_by(|a, b| sections_weight[a].cmp(§ions_weight[b])); section.subsections = children; } section.ancestors = ancestors.get(§ion.file.path).cloned().unwrap_or_else(|| vec![]); } } /// Sort all sections pages according to sorting method given /// Pages that cannot be sorted are set to the section.ignored_pages instead pub fn sort_sections_pages(&mut self) { let mut updates = HashMap::new(); for (key, section) in &self.sections { let (sorted_pages, cannot_be_sorted_pages) = match section.meta.sort_by { SortBy::None => continue, SortBy::Date => { let data = section .pages .iter() .map(|k| { if let Some(page) = self.pages.get(*k) { (k, page.meta.datetime, page.permalink.as_ref()) } else { unreachable!("Sorting got an unknown page") } }) .collect(); sort_pages_by_date(data) } SortBy::Weight => { let data = section .pages .iter() .map(|k| { if let Some(page) = self.pages.get(*k) { (k, page.meta.weight, page.permalink.as_ref()) } else { unreachable!("Sorting got an unknown page") } }) .collect(); sort_pages_by_weight(data) } }; updates.insert(key, (sorted_pages, cannot_be_sorted_pages, section.meta.sort_by)); } for (key, (sorted, cannot_be_sorted, sort_by)) in updates { // Find sibling between sorted pages first let with_siblings = find_siblings(&sorted); for (k2, val1, val2) in with_siblings { if let Some(page) = self.pages.get_mut(k2) { match sort_by { SortBy::Date => { page.earlier = val2; page.later = val1; } SortBy::Weight => { page.lighter = val1; page.heavier = val2; } SortBy::None => unreachable!("Impossible to find siblings in SortBy::None"), } } else { unreachable!("Sorting got an unknown page") } } if let Some(s) = self.sections.get_mut(key) { s.pages = sorted; s.ignored_pages = cannot_be_sorted; } } } /// Finds all the translations for each section/page and set the `translations` /// field of each as needed /// A no-op for sites without multiple languages fn populate_translations(&mut self) { if !self.is_multilingual { return; } // Sections first let mut sections_translations = HashMap::new(); for (key, section) in &self.sections { sections_translations .entry(section.file.canonical.clone()) // TODO: avoid this clone .or_insert_with(Vec::new) .push(key); } for (key, section) in self.sections.iter_mut() { let translations = §ions_translations[§ion.file.canonical]; if translations.len() == 1 { section.translations = vec![]; continue; } section.translations = translations.iter().filter(|k| **k != key).cloned().collect(); } // Same thing for pages let mut pages_translations = HashMap::new(); for (key, page) in &self.pages { pages_translations .entry(page.file.canonical.clone()) // TODO: avoid this clone .or_insert_with(Vec::new) .push(key); } for (key, page) in self.pages.iter_mut() { let translations = &pages_translations[&page.file.canonical]; if translations.len() == 1 { page.translations = vec![]; continue; } page.translations = translations.iter().filter(|k| **k != key).cloned().collect(); } } /// Find all the orphan pages: pages that are in a folder without an `_index.md` pub fn get_all_orphan_pages(&self) -> Vec<&Page> { let pages_in_sections = self.sections.values().flat_map(|s| &s.pages).collect::>(); self.pages .iter() .filter(|(key, _)| !pages_in_sections.contains(&key)) .map(|(_, page)| page) .collect() } /// Find the parent section & all grandparents section that have transparent=true /// Only used in rebuild. pub fn find_parent_sections>(&self, path: P) -> Vec<&Section> { let mut parents = vec![]; let page = self.get_page(path.as_ref()).unwrap(); for ancestor in page.ancestors.iter().rev() { let section = self.get_section_by_key(*ancestor); if parents.is_empty() || section.meta.transparent { parents.push(section); } } parents } /// Only used in tests pub fn get_section_key>(&self, path: P) -> Option<&DefaultKey> { self.paths_to_sections.get(path.as_ref()) } pub fn get_section>(&self, path: P) -> Option<&Section> { self.sections.get(self.paths_to_sections.get(path.as_ref()).cloned().unwrap_or_default()) } pub fn get_section_mut>(&mut self, path: P) -> Option<&mut Section> { self.sections .get_mut(self.paths_to_sections.get(path.as_ref()).cloned().unwrap_or_default()) } pub fn get_section_by_key(&self, key: DefaultKey) -> &Section { self.sections.get(key).unwrap() } pub fn get_section_mut_by_key(&mut self, key: DefaultKey) -> &mut Section { self.sections.get_mut(key).unwrap() } pub fn get_section_path_by_key(&self, key: DefaultKey) -> &str { &self.get_section_by_key(key).file.relative } pub fn get_page>(&self, path: P) -> Option<&Page> { self.pages.get(self.paths_to_pages.get(path.as_ref()).cloned().unwrap_or_default()) } pub fn get_page_by_key(&self, key: DefaultKey) -> &Page { self.pages.get(key).unwrap() } pub fn get_page_mut_by_key(&mut self, key: DefaultKey) -> &mut Page { self.pages.get_mut(key).unwrap() } pub fn remove_section>(&mut self, path: P) -> Option
{ if let Some(k) = self.paths_to_sections.remove(path.as_ref()) { self.sections.remove(k) } else { None } } pub fn remove_page>(&mut self, path: P) -> Option { if let Some(k) = self.paths_to_pages.remove(path.as_ref()) { self.pages.remove(k) } else { None } } /// Used in rebuild, to check if we know it already pub fn contains_section>(&self, path: P) -> bool { self.paths_to_sections.contains_key(path.as_ref()) } /// Used in rebuild, to check if we know it already pub fn contains_page>(&self, path: P) -> bool { self.paths_to_pages.contains_key(path.as_ref()) } /// This will check every section/page paths + the aliases and ensure none of them /// are colliding. /// Returns (path colliding, [list of files causing that collision]) pub fn check_for_path_collisions(&self) -> Vec<(&str, Vec)> { let mut paths: HashMap<&str, HashSet> = HashMap::new(); for (key, page) in &self.pages { paths .entry(&page.path) .and_modify(|s| { s.insert(key); }) .or_insert_with(|| set!(key)); for alias in &page.meta.aliases { paths .entry(&alias) .and_modify(|s| { s.insert(key); }) .or_insert_with(|| set!(key)); } } for (key, section) in &self.sections { if !section.meta.render { continue; } paths .entry(§ion.path) .and_modify(|s| { s.insert(key); }) .or_insert_with(|| set!(key)); } let mut collisions = vec![]; for (p, keys) in paths { if keys.len() > 1 { let file_paths: Vec = keys .iter() .map(|k| { self.pages.get(*k).map(|p| p.file.relative.clone()).unwrap_or_else(|| { self.sections.get(*k).map(|s| s.file.relative.clone()).unwrap() }) }) .collect(); collisions.push((p, file_paths)); } } collisions } } #[cfg(test)] mod tests { use super::*; #[test] fn can_find_no_collisions() { let mut library = Library::new(10, 10, false); let mut page = Page::default(); page.path = "hello".to_string(); let mut page2 = Page::default(); page2.path = "hello-world".to_string(); let mut section = Section::default(); section.path = "blog".to_string(); library.insert_page(page); library.insert_page(page2); library.insert_section(section); let collisions = library.check_for_path_collisions(); assert_eq!(collisions.len(), 0); } #[test] fn can_find_collisions_between_pages() { let mut library = Library::new(10, 10, false); let mut page = Page::default(); page.path = "hello".to_string(); page.file.relative = "hello".to_string(); let mut page2 = Page::default(); page2.path = "hello".to_string(); page2.file.relative = "hello-world".to_string(); let mut section = Section::default(); section.path = "blog".to_string(); section.file.relative = "hello-world".to_string(); library.insert_page(page.clone()); library.insert_page(page2.clone()); library.insert_section(section); let collisions = library.check_for_path_collisions(); assert_eq!(collisions.len(), 1); assert_eq!(collisions[0].0, page.path); assert!(collisions[0].1.contains(&page.file.relative)); assert!(collisions[0].1.contains(&page2.file.relative)); } #[test] fn can_find_collisions_with_an_alias() { let mut library = Library::new(10, 10, false); let mut page = Page::default(); page.path = "hello".to_string(); page.file.relative = "hello".to_string(); let mut page2 = Page::default(); page2.path = "hello-world".to_string(); page2.file.relative = "hello-world".to_string(); page2.meta.aliases = vec!["hello".to_string()]; let mut section = Section::default(); section.path = "blog".to_string(); section.file.relative = "hello-world".to_string(); library.insert_page(page.clone()); library.insert_page(page2.clone()); library.insert_section(section); let collisions = library.check_for_path_collisions(); assert_eq!(collisions.len(), 1); assert_eq!(collisions[0].0, page.path); assert!(collisions[0].1.contains(&page.file.relative)); assert!(collisions[0].1.contains(&page2.file.relative)); } }