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/io/struct.IoSlice.html")
126 || file.ends_with("std/string/struct.String.html")
131 if file.ends_with("alloc/string/struct.String.html") {
135 if file.ends_with("alloc/collections/btree_map/struct.BTreeMap.html")
136 || file.ends_with("alloc/collections/btree_set/struct.BTreeSet.html")
137 || file.ends_with("std/collections/btree_map/struct.BTreeMap.html")
138 || file.ends_with("std/collections/btree_set/struct.BTreeSet.html")
139 || file.ends_with("std/collections/hash_map/struct.HashMap.html")
140 || file.ends_with("std/collections/hash_set/struct.HashSet.html")
145 let res = load_file(cache, root, file, SkipRedirect);
146 let (pretty_file, contents) = match res {
148 Err(_) => return None,
151 cache.get_mut(&pretty_file).unwrap().parse_ids(&pretty_file, &contents, errors);
154 // Search for anything that's the regex 'href[ ]*=[ ]*".*?"'
155 with_attrs_in_source(&contents, " href", |url, i, base| {
156 // Ignore external URLs
157 if url.starts_with("http:")
158 || url.starts_with("https:")
159 || url.starts_with("javascript:")
160 || url.starts_with("ftp:")
161 || url.starts_with("irc:")
162 || url.starts_with("data:")
166 let mut parts = url.splitn(2, "#");
167 let url = parts.next().unwrap();
168 let fragment = parts.next();
169 let mut parts = url.splitn(2, "?");
170 let url = parts.next().unwrap();
172 // Once we've plucked out the URL, parse it using our base url and
173 // then try to extract a file path.
174 let mut path = file.to_path_buf();
175 if !base.is_empty() || !url.is_empty() {
177 for part in Path::new(base).join(url).components() {
179 Component::Prefix(_) | Component::RootDir => {
180 // Avoid absolute paths as they make the docs not
181 // relocatable by making assumptions on where the docs
182 // are hosted relative to the site root.
185 "{}:{}: absolute path - {}",
186 pretty_file.display(),
188 Path::new(base).join(url).display()
192 Component::CurDir => {}
193 Component::ParentDir => {
196 Component::Normal(s) => {
203 // Alright, if we've found a file name then this file had better
204 // exist! If it doesn't then we register and print an error.
207 // Links to directories show as directory listings when viewing
208 // the docs offline so it's best to avoid them.
210 let pretty_path = path.strip_prefix(root).unwrap_or(&path);
212 "{}:{}: directory link - {}",
213 pretty_file.display(),
215 pretty_path.display()
219 if let Some(extension) = path.extension() {
220 // Ignore none HTML files.
221 if extension != "html" {
225 let res = load_file(cache, root, &path, FromRedirect(false));
226 let (pretty_path, contents) = match res {
228 Err(LoadError::IOError(err)) => {
229 panic!("error loading {}: {}", path.display(), err);
231 Err(LoadError::BrokenRedirect(target, _)) => {
234 "{}:{}: broken redirect to {}",
235 pretty_file.display(),
241 Err(LoadError::IsRedirect) => unreachable!(),
244 if let Some(ref fragment) = fragment {
245 // Fragments like `#1-6` are most likely line numbers to be
246 // interpreted by javascript, so we're ignoring these
247 if fragment.splitn(2, '-').all(|f| f.chars().all(|c| c.is_numeric())) {
251 // These appear to be broken in mdbook right now?
252 if fragment.starts_with("-") {
256 let entry = &mut cache.get_mut(&pretty_path).unwrap();
257 entry.parse_ids(&pretty_path, &contents, errors);
259 if !entry.ids.contains(*fragment) {
261 print!("{}:{}: broken link fragment ", pretty_file.display(), i + 1);
262 println!("`#{}` pointing to `{}`", fragment, pretty_path.display());
267 print!("{}:{}: broken link - ", pretty_file.display(), i + 1);
268 let pretty_path = path.strip_prefix(root).unwrap_or(&path);
269 println!("{}", pretty_path.display());
280 ) -> Result<(PathBuf, Rc<String>), LoadError> {
281 let pretty_file = PathBuf::from(file.strip_prefix(root).unwrap_or(&file));
283 let (maybe_redirect, contents) = match cache.entry(pretty_file.clone()) {
284 Entry::Occupied(entry) => (None, entry.get().source.clone()),
285 Entry::Vacant(entry) => {
286 let contents = match fs::read_to_string(file) {
289 return Err(if let FromRedirect(true) = redirect {
290 LoadError::BrokenRedirect(file.to_path_buf(), err)
292 LoadError::IOError(err)
297 let maybe = maybe_redirect(&contents);
299 if let SkipRedirect = redirect {
300 return Err(LoadError::IsRedirect);
303 entry.insert(FileEntry { source: contents.clone(), ids: HashSet::new() });
308 match maybe_redirect.map(|url| file.parent().unwrap().join(url)) {
309 Some(redirect_file) => load_file(cache, root, &redirect_file, FromRedirect(true)),
310 None => Ok((pretty_file, contents)),
314 fn maybe_redirect(source: &str) -> Option<String> {
315 const REDIRECT: &'static str = "<p>Redirecting to <a href=";
317 let mut lines = source.lines();
318 let redirect_line = lines.nth(6)?;
320 redirect_line.find(REDIRECT).map(|i| {
321 let rest = &redirect_line[(i + REDIRECT.len() + 1)..];
322 let pos_quote = rest.find('"').unwrap();
323 rest[..pos_quote].to_owned()
327 fn with_attrs_in_source<F: FnMut(&str, usize, &str)>(contents: &str, attr: &str, mut f: F) {
329 for (i, mut line) in contents.lines().enumerate() {
330 while let Some(j) = line.find(attr) {
331 let rest = &line[j + attr.len()..];
332 // The base tag should always be the first link in the document so
333 // we can get away with using one pass.
334 let is_base = line[..j].ends_with("<base");
336 let pos_equals = match rest.find("=") {
340 if rest[..pos_equals].trim_start_matches(" ") != "" {
344 let rest = &rest[pos_equals + 1..];
346 let pos_quote = match rest.find(&['"', '\''][..]) {
350 let quote_delim = rest.as_bytes()[pos_quote] as char;
352 if rest[..pos_quote].trim_start_matches(" ") != "" {
355 let rest = &rest[pos_quote + 1..];
356 let url = match rest.find(quote_delim) {
357 Some(i) => &rest[..i],