1 //! Script to check the validity of `href` links in our HTML documentation.
3 //! In the past we've been quite error prone to writing in broken links as most
4 //! of them are manually rather than automatically added. As files move over
5 //! time or apis change old links become stale or broken. The purpose of this
6 //! script is to check all relative links in our documentation to make sure they
7 //! actually point to a valid place.
9 //! Currently this doesn't actually do any HTML parsing or anything fancy like
10 //! that, it just has a simple "regex" to search for `href` and `id` tags.
11 //! These values are then translated to file URLs if possible and then the
12 //! destination is asserted to exist.
14 //! A few whitelisted exceptions are allowed as there's known bugs in rustdoc,
15 //! but this should catch the majority of "broken link" cases.
19 use std::collections::hash_map::Entry;
20 use std::collections::{HashMap, HashSet};
23 use std::path::{Component, Path, PathBuf};
26 use crate::Redirect::*;
32 Err(e) => panic!("{} failed with {:?}", stringify!($e), e),
38 let docs = env::args_os().nth(1).unwrap();
39 let docs = env::current_dir().unwrap().join(docs);
40 let mut errors = false;
41 walk(&mut HashMap::new(), &docs, &docs, &mut errors);
43 panic!("found some broken links");
49 IOError(std::io::Error),
50 BrokenRedirect(PathBuf, std::io::Error),
64 type Cache = HashMap<PathBuf, FileEntry>;
66 fn small_url_encode(s: &str) -> String {
82 fn parse_ids(&mut self, file: &Path, contents: &str, errors: &mut bool) {
83 if self.ids.is_empty() {
84 with_attrs_in_source(contents, " id", |fragment, i, _| {
85 let frag = fragment.trim_start_matches("#").to_owned();
86 let encoded = small_url_encode(&frag);
87 if !self.ids.insert(frag) {
89 println!("{}:{}: id is not unique: `{}`", file.display(), i, fragment);
91 // Just in case, we also add the encoded id.
92 self.ids.insert(encoded);
98 fn walk(cache: &mut Cache, root: &Path, dir: &Path, errors: &mut bool) {
99 for entry in t!(dir.read_dir()).map(|e| t!(e)) {
100 let path = entry.path();
101 let kind = t!(entry.file_type());
103 walk(cache, root, &path, errors);
105 let pretty_path = check(cache, root, &path, errors);
106 if let Some(pretty_path) = pretty_path {
107 let entry = cache.get_mut(&pretty_path).unwrap();
108 // we don't need the source anymore,
109 // so drop to reduce memory-usage
110 entry.source = Rc::new(String::new());
116 fn check(cache: &mut Cache, root: &Path, file: &Path, errors: &mut bool) -> Option<PathBuf> {
117 // Ignore none HTML files.
118 if file.extension().and_then(|s| s.to_str()) != Some("html") {
122 // Unfortunately we're not 100% full of valid links today to we need a few
123 // whitelists to get this past `make check` today.
125 if file.ends_with("std/string/struct.String.html")
126 || file.ends_with("interpret/struct.ImmTy.html")
127 || file.ends_with("ast/struct.ThinVec.html")
128 || file.ends_with("util/struct.ThinVec.html")
129 || file.ends_with("layout/struct.TyLayout.html")
130 || file.ends_with("humantime/struct.Timestamp.html")
131 || file.ends_with("log/index.html")
132 || file.ends_with("ty/struct.Slice.html")
133 || file.ends_with("ty/enum.Attributes.html")
134 || file.ends_with("ty/struct.SymbolName.html")
135 || file.ends_with("io/struct.IoSlice.html")
136 || file.ends_with("io/struct.IoSliceMut.html")
141 if file.ends_with("string/struct.String.html") {
145 if file.ends_with("btree_set/struct.BTreeSet.html")
146 || file.ends_with("struct.BTreeSet.html")
147 || file.ends_with("btree_map/struct.BTreeMap.html")
148 || file.ends_with("hash_map/struct.HashMap.html")
149 || file.ends_with("hash_set/struct.HashSet.html")
150 || file.ends_with("sync/struct.Lrc.html")
151 || file.ends_with("sync/struct.RwLock.html")
156 let res = load_file(cache, root, file, SkipRedirect);
157 let (pretty_file, contents) = match res {
159 Err(_) => return None,
162 cache.get_mut(&pretty_file).unwrap().parse_ids(&pretty_file, &contents, errors);
165 // Search for anything that's the regex 'href[ ]*=[ ]*".*?"'
166 with_attrs_in_source(&contents, " href", |url, i, base| {
167 // Ignore external URLs
168 if url.starts_with("http:")
169 || url.starts_with("https:")
170 || url.starts_with("javascript:")
171 || url.starts_with("ftp:")
172 || url.starts_with("irc:")
173 || url.starts_with("data:")
177 let mut parts = url.splitn(2, "#");
178 let url = parts.next().unwrap();
179 let fragment = parts.next();
180 let mut parts = url.splitn(2, "?");
181 let url = parts.next().unwrap();
183 // Once we've plucked out the URL, parse it using our base url and
184 // then try to extract a file path.
185 let mut path = file.to_path_buf();
186 if !base.is_empty() || !url.is_empty() {
188 for part in Path::new(base).join(url).components() {
190 Component::Prefix(_) | Component::RootDir => {
191 // Avoid absolute paths as they make the docs not
192 // relocatable by making assumptions on where the docs
193 // are hosted relative to the site root.
196 "{}:{}: absolute path - {}",
197 pretty_file.display(),
199 Path::new(base).join(url).display()
203 Component::CurDir => {}
204 Component::ParentDir => {
207 Component::Normal(s) => {
214 // Alright, if we've found a file name then this file had better
215 // exist! If it doesn't then we register and print an error.
218 // Links to directories show as directory listings when viewing
219 // the docs offline so it's best to avoid them.
221 let pretty_path = path.strip_prefix(root).unwrap_or(&path);
223 "{}:{}: directory link - {}",
224 pretty_file.display(),
226 pretty_path.display()
230 if let Some(extension) = path.extension() {
231 // Ignore none HTML files.
232 if extension != "html" {
236 let res = load_file(cache, root, &path, FromRedirect(false));
237 let (pretty_path, contents) = match res {
239 Err(LoadError::IOError(err)) => {
240 panic!("error loading {}: {}", path.display(), err);
242 Err(LoadError::BrokenRedirect(target, _)) => {
245 "{}:{}: broken redirect to {}",
246 pretty_file.display(),
252 Err(LoadError::IsRedirect) => unreachable!(),
255 if let Some(ref fragment) = fragment {
256 // Fragments like `#1-6` are most likely line numbers to be
257 // interpreted by javascript, so we're ignoring these
258 if fragment.splitn(2, '-').all(|f| f.chars().all(|c| c.is_numeric())) {
262 // These appear to be broken in mdbook right now?
263 if fragment.starts_with("-") {
267 let entry = &mut cache.get_mut(&pretty_path).unwrap();
268 entry.parse_ids(&pretty_path, &contents, errors);
270 if !entry.ids.contains(*fragment) {
272 print!("{}:{}: broken link fragment ", pretty_file.display(), i + 1);
273 println!("`#{}` pointing to `{}`", fragment, pretty_path.display());
278 print!("{}:{}: broken link - ", pretty_file.display(), i + 1);
279 let pretty_path = path.strip_prefix(root).unwrap_or(&path);
280 println!("{}", pretty_path.display());
291 ) -> Result<(PathBuf, Rc<String>), LoadError> {
292 let pretty_file = PathBuf::from(file.strip_prefix(root).unwrap_or(&file));
294 let (maybe_redirect, contents) = match cache.entry(pretty_file.clone()) {
295 Entry::Occupied(entry) => (None, entry.get().source.clone()),
296 Entry::Vacant(entry) => {
297 let contents = match fs::read_to_string(file) {
300 return Err(if let FromRedirect(true) = redirect {
301 LoadError::BrokenRedirect(file.to_path_buf(), err)
303 LoadError::IOError(err)
308 let maybe = maybe_redirect(&contents);
310 if let SkipRedirect = redirect {
311 return Err(LoadError::IsRedirect);
314 entry.insert(FileEntry { source: contents.clone(), ids: HashSet::new() });
319 match maybe_redirect.map(|url| file.parent().unwrap().join(url)) {
320 Some(redirect_file) => load_file(cache, root, &redirect_file, FromRedirect(true)),
321 None => Ok((pretty_file, contents)),
325 fn maybe_redirect(source: &str) -> Option<String> {
326 const REDIRECT: &'static str = "<p>Redirecting to <a href=";
328 let mut lines = source.lines();
329 let redirect_line = lines.nth(6)?;
331 redirect_line.find(REDIRECT).map(|i| {
332 let rest = &redirect_line[(i + REDIRECT.len() + 1)..];
333 let pos_quote = rest.find('"').unwrap();
334 rest[..pos_quote].to_owned()
338 fn with_attrs_in_source<F: FnMut(&str, usize, &str)>(contents: &str, attr: &str, mut f: F) {
340 for (i, mut line) in contents.lines().enumerate() {
341 while let Some(j) = line.find(attr) {
342 let rest = &line[j + attr.len()..];
343 // The base tag should always be the first link in the document so
344 // we can get away with using one pass.
345 let is_base = line[..j].ends_with("<base");
347 let pos_equals = match rest.find("=") {
351 if rest[..pos_equals].trim_start_matches(" ") != "" {
355 let rest = &rest[pos_equals + 1..];
357 let pos_quote = match rest.find(&['"', '\''][..]) {
361 let quote_delim = rest.as_bytes()[pos_quote] as char;
363 if rest[..pos_quote].trim_start_matches(" ") != "" {
366 let rest = &rest[pos_quote + 1..];
367 let url = match rest.find(quote_delim) {
368 Some(i) => &rest[..i],