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 exceptions are allowed as there's known bugs in rustdoc, but this
15 //! should catch the majority of "broken link" cases.
17 use std::collections::hash_map::Entry;
18 use std::collections::{HashMap, HashSet};
21 use std::path::{Component, Path, PathBuf};
24 use crate::Redirect::*;
30 Err(e) => panic!("{} failed with {:?}", stringify!($e), e),
36 let docs = env::args_os().nth(1).unwrap();
37 let docs = env::current_dir().unwrap().join(docs);
38 let mut errors = false;
39 walk(&mut HashMap::new(), &docs, &docs, &mut errors);
41 panic!("found some broken links");
47 IOError(std::io::Error),
48 BrokenRedirect(PathBuf, std::io::Error),
62 type Cache = HashMap<PathBuf, FileEntry>;
64 fn small_url_encode(s: &str) -> String {
80 fn parse_ids(&mut self, file: &Path, contents: &str, errors: &mut bool) {
81 if self.ids.is_empty() {
82 with_attrs_in_source(contents, " id", |fragment, i, _| {
83 let frag = fragment.trim_start_matches("#").to_owned();
84 let encoded = small_url_encode(&frag);
85 if !self.ids.insert(frag) {
87 println!("{}:{}: id is not unique: `{}`", file.display(), i, fragment);
89 // Just in case, we also add the encoded id.
90 self.ids.insert(encoded);
96 fn walk(cache: &mut Cache, root: &Path, dir: &Path, errors: &mut bool) {
97 for entry in t!(dir.read_dir()).map(|e| t!(e)) {
98 let path = entry.path();
99 let kind = t!(entry.file_type());
101 walk(cache, root, &path, errors);
103 let pretty_path = check(cache, root, &path, errors);
104 if let Some(pretty_path) = pretty_path {
105 let entry = cache.get_mut(&pretty_path).unwrap();
106 // we don't need the source anymore,
107 // so drop to reduce memory-usage
108 entry.source = Rc::new(String::new());
114 fn check(cache: &mut Cache, root: &Path, file: &Path, errors: &mut bool) -> Option<PathBuf> {
115 // Ignore non-HTML files.
116 if file.extension().and_then(|s| s.to_str()) != Some("html") {
120 // Unfortunately we're not 100% full of valid links today to we need a few
121 // exceptions to get this past `make check` today.
123 if file.ends_with("std/io/struct.IoSlice.html")
129 if file.ends_with("alloc/collections/btree_map/struct.BTreeMap.html")
130 || file.ends_with("alloc/collections/btree_set/struct.BTreeSet.html")
131 || file.ends_with("std/collections/btree_map/struct.BTreeMap.html")
132 || file.ends_with("std/collections/btree_set/struct.BTreeSet.html")
133 || file.ends_with("std/collections/hash_set/struct.HashSet.html")
138 let res = load_file(cache, root, file, SkipRedirect);
139 let (pretty_file, contents) = match res {
141 Err(_) => return None,
144 cache.get_mut(&pretty_file).unwrap().parse_ids(&pretty_file, &contents, errors);
147 // Search for anything that's the regex 'href[ ]*=[ ]*".*?"'
148 with_attrs_in_source(&contents, " href", |url, i, base| {
149 // Ignore external URLs
150 if url.starts_with("http:")
151 || url.starts_with("https:")
152 || url.starts_with("javascript:")
153 || url.starts_with("ftp:")
154 || url.starts_with("irc:")
155 || url.starts_with("data:")
159 let mut parts = url.splitn(2, "#");
160 let url = parts.next().unwrap();
161 let fragment = parts.next();
162 let mut parts = url.splitn(2, "?");
163 let url = parts.next().unwrap();
165 // Once we've plucked out the URL, parse it using our base url and
166 // then try to extract a file path.
167 let mut path = file.to_path_buf();
168 if !base.is_empty() || !url.is_empty() {
170 for part in Path::new(base).join(url).components() {
172 Component::Prefix(_) | Component::RootDir => {
173 // Avoid absolute paths as they make the docs not
174 // relocatable by making assumptions on where the docs
175 // are hosted relative to the site root.
178 "{}:{}: absolute path - {}",
179 pretty_file.display(),
181 Path::new(base).join(url).display()
185 Component::CurDir => {}
186 Component::ParentDir => {
189 Component::Normal(s) => {
196 // Alright, if we've found a file name then this file had better
197 // exist! If it doesn't then we register and print an error.
200 // Links to directories show as directory listings when viewing
201 // the docs offline so it's best to avoid them.
203 let pretty_path = path.strip_prefix(root).unwrap_or(&path);
205 "{}:{}: directory link - {}",
206 pretty_file.display(),
208 pretty_path.display()
212 if let Some(extension) = path.extension() {
213 // Ignore none HTML files.
214 if extension != "html" {
218 let res = load_file(cache, root, &path, FromRedirect(false));
219 let (pretty_path, contents) = match res {
221 Err(LoadError::IOError(err)) => {
222 panic!("error loading {}: {}", path.display(), err);
224 Err(LoadError::BrokenRedirect(target, _)) => {
227 "{}:{}: broken redirect to {}",
228 pretty_file.display(),
234 Err(LoadError::IsRedirect) => unreachable!(),
237 if let Some(ref fragment) = fragment {
238 // Fragments like `#1-6` are most likely line numbers to be
239 // interpreted by javascript, so we're ignoring these
240 if fragment.splitn(2, '-').all(|f| f.chars().all(|c| c.is_numeric())) {
244 // These appear to be broken in mdbook right now?
245 if fragment.starts_with("-") {
249 let entry = &mut cache.get_mut(&pretty_path).unwrap();
250 entry.parse_ids(&pretty_path, &contents, errors);
252 if !entry.ids.contains(*fragment) {
254 print!("{}:{}: broken link fragment ", pretty_file.display(), i + 1);
255 println!("`#{}` pointing to `{}`", fragment, pretty_path.display());
260 print!("{}:{}: broken link - ", pretty_file.display(), i + 1);
261 let pretty_path = path.strip_prefix(root).unwrap_or(&path);
262 println!("{}", pretty_path.display());
273 ) -> Result<(PathBuf, Rc<String>), LoadError> {
274 let pretty_file = PathBuf::from(file.strip_prefix(root).unwrap_or(&file));
276 let (maybe_redirect, contents) = match cache.entry(pretty_file.clone()) {
277 Entry::Occupied(entry) => (None, entry.get().source.clone()),
278 Entry::Vacant(entry) => {
279 let contents = match fs::read_to_string(file) {
282 return Err(if let FromRedirect(true) = redirect {
283 LoadError::BrokenRedirect(file.to_path_buf(), err)
285 LoadError::IOError(err)
290 let maybe = maybe_redirect(&contents);
292 if let SkipRedirect = redirect {
293 return Err(LoadError::IsRedirect);
296 entry.insert(FileEntry { source: contents.clone(), ids: HashSet::new() });
301 match maybe_redirect.map(|url| file.parent().unwrap().join(url)) {
302 Some(redirect_file) => load_file(cache, root, &redirect_file, FromRedirect(true)),
303 None => Ok((pretty_file, contents)),
307 fn maybe_redirect(source: &str) -> Option<String> {
308 const REDIRECT: &'static str = "<p>Redirecting to <a href=";
310 let mut lines = source.lines();
311 let redirect_line = lines.nth(6)?;
313 redirect_line.find(REDIRECT).map(|i| {
314 let rest = &redirect_line[(i + REDIRECT.len() + 1)..];
315 let pos_quote = rest.find('"').unwrap();
316 rest[..pos_quote].to_owned()
320 fn with_attrs_in_source<F: FnMut(&str, usize, &str)>(contents: &str, attr: &str, mut f: F) {
322 for (i, mut line) in contents.lines().enumerate() {
323 while let Some(j) = line.find(attr) {
324 let rest = &line[j + attr.len()..];
325 // The base tag should always be the first link in the document so
326 // we can get away with using one pass.
327 let is_base = line[..j].ends_with("<base");
329 let pos_equals = match rest.find("=") {
333 if rest[..pos_equals].trim_start_matches(" ") != "" {
337 let rest = &rest[pos_equals + 1..];
339 let pos_quote = match rest.find(&['"', '\''][..]) {
343 let quote_delim = rest.as_bytes()[pos_quote] as char;
345 if rest[..pos_quote].trim_start_matches(" ") != "" {
348 let rest = &rest[pos_quote + 1..];
349 let url = match rest.find(quote_delim) {
350 Some(i) => &rest[..i],