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.
17 #![deny(rust_2018_idioms)]
19 use std::collections::hash_map::Entry;
20 use std::collections::{HashMap, HashSet};
23 use std::path::{Path, PathBuf, Component};
26 use crate::Redirect::*;
29 ($e:expr) => (match $e {
31 Err(e) => panic!("{} failed with {:?}", stringify!($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,
119 // Ignore none HTML files.
120 if file.extension().and_then(|s| s.to_str()) != Some("html") {
124 // Unfortunately we're not 100% full of valid links today to we need a few
125 // whitelists to get this past `make check` today.
127 if file.ends_with("std/string/struct.String.html") ||
128 file.ends_with("interpret/struct.ImmTy.html") ||
129 file.ends_with("symbol/struct.InternedString.html") ||
130 file.ends_with("ast/struct.ThinVec.html") ||
131 file.ends_with("util/struct.ThinVec.html") ||
132 file.ends_with("layout/struct.TyLayout.html") ||
133 file.ends_with("humantime/struct.Timestamp.html") ||
134 file.ends_with("log/index.html") ||
135 file.ends_with("ty/struct.Slice.html") ||
136 file.ends_with("ty/enum.Attributes.html") ||
137 file.ends_with("ty/struct.SymbolName.html") ||
138 file.ends_with("io/struct.IoSlice.html") ||
139 file.ends_with("io/struct.IoSliceMut.html") {
143 if file.ends_with("string/struct.String.html") {
147 if file.ends_with("btree_set/struct.BTreeSet.html") ||
148 file.ends_with("struct.BTreeSet.html") ||
149 file.ends_with("btree_map/struct.BTreeMap.html") ||
150 file.ends_with("hash_map/struct.HashMap.html") ||
151 file.ends_with("hash_set/struct.HashSet.html") ||
152 file.ends_with("sync/struct.Lrc.html") ||
153 file.ends_with("sync/struct.RwLock.html") {
157 let res = load_file(cache, root, file, SkipRedirect);
158 let (pretty_file, contents) = match res {
160 Err(_) => return None,
163 cache.get_mut(&pretty_file)
165 .parse_ids(&pretty_file, &contents, errors);
168 // Search for anything that's the regex 'href[ ]*=[ ]*".*?"'
169 with_attrs_in_source(&contents, " href", |url, i, base| {
170 // Ignore external URLs
171 if url.starts_with("http:") || url.starts_with("https:") ||
172 url.starts_with("javascript:") || url.starts_with("ftp:") ||
173 url.starts_with("irc:") || url.starts_with("data:") {
176 let mut parts = url.splitn(2, "#");
177 let url = parts.next().unwrap();
178 let fragment = parts.next();
179 let mut parts = url.splitn(2, "?");
180 let url = parts.next().unwrap();
182 // Once we've plucked out the URL, parse it using our base url and
183 // then try to extract a file path.
184 let mut path = file.to_path_buf();
185 if !base.is_empty() || !url.is_empty() {
187 for part in Path::new(base).join(url).components() {
189 Component::Prefix(_) |
190 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.
195 println!("{}:{}: absolute path - {}",
196 pretty_file.display(),
198 Path::new(base).join(url).display());
201 Component::CurDir => {}
202 Component::ParentDir => { path.pop(); }
203 Component::Normal(s) => { path.push(s); }
208 // Alright, if we've found a file name then this file had better
209 // exist! If it doesn't then we register and print an error.
212 // Links to directories show as directory listings when viewing
213 // the docs offline so it's best to avoid them.
215 let pretty_path = path.strip_prefix(root).unwrap_or(&path);
216 println!("{}:{}: directory link - {}",
217 pretty_file.display(),
219 pretty_path.display());
222 if let Some(extension) = path.extension() {
223 // Ignore none HTML files.
224 if extension != "html" {
228 let res = load_file(cache, root, &path, FromRedirect(false));
229 let (pretty_path, contents) = match res {
231 Err(LoadError::IOError(err)) => {
232 panic!("error loading {}: {}", path.display(), err);
234 Err(LoadError::BrokenRedirect(target, _)) => {
236 println!("{}:{}: broken redirect to {}",
237 pretty_file.display(),
242 Err(LoadError::IsRedirect) => unreachable!(),
245 if let Some(ref fragment) = fragment {
246 // Fragments like `#1-6` are most likely line numbers to be
247 // interpreted by javascript, so we're ignoring these
248 if fragment.splitn(2, '-')
249 .all(|f| f.chars().all(|c| c.is_numeric())) {
253 // These appear to be broken in mdbook right now?
254 if fragment.starts_with("-") {
258 let entry = &mut cache.get_mut(&pretty_path).unwrap();
259 entry.parse_ids(&pretty_path, &contents, errors);
261 if !entry.ids.contains(*fragment) {
263 print!("{}:{}: broken link fragment ",
264 pretty_file.display(),
266 println!("`#{}` pointing to `{}`", fragment, pretty_path.display());
271 print!("{}:{}: broken link - ", pretty_file.display(), i + 1);
272 let pretty_path = path.strip_prefix(root).unwrap_or(&path);
273 println!("{}", pretty_path.display());
279 fn load_file(cache: &mut Cache,
283 -> Result<(PathBuf, Rc<String>), LoadError> {
284 let pretty_file = PathBuf::from(file.strip_prefix(root).unwrap_or(&file));
286 let (maybe_redirect, contents) = match cache.entry(pretty_file.clone()) {
287 Entry::Occupied(entry) => {
288 (None, entry.get().source.clone())
290 Entry::Vacant(entry) => {
291 let contents = match fs::read_to_string(file) {
294 return Err(if let FromRedirect(true) = redirect {
295 LoadError::BrokenRedirect(file.to_path_buf(), err)
297 LoadError::IOError(err)
302 let maybe = maybe_redirect(&contents);
304 if let SkipRedirect = redirect {
305 return Err(LoadError::IsRedirect);
308 entry.insert(FileEntry {
309 source: contents.clone(),
316 match maybe_redirect.map(|url| file.parent().unwrap().join(url)) {
317 Some(redirect_file) => {
318 load_file(cache, root, &redirect_file, FromRedirect(true))
320 None => Ok((pretty_file, contents)),
324 fn maybe_redirect(source: &str) -> Option<String> {
325 const REDIRECT: &'static str = "<p>Redirecting to <a href=";
327 let mut lines = source.lines();
328 let redirect_line = lines.nth(6)?;
330 redirect_line.find(REDIRECT).map(|i| {
331 let rest = &redirect_line[(i + REDIRECT.len() + 1)..];
332 let pos_quote = rest.find('"').unwrap();
333 rest[..pos_quote].to_owned()
337 fn with_attrs_in_source<F: FnMut(&str, usize, &str)>(contents: &str, attr: &str, mut f: F) {
339 for (i, mut line) in contents.lines().enumerate() {
340 while let Some(j) = line.find(attr) {
341 let rest = &line[j + attr.len()..];
342 // The base tag should always be the first link in the document so
343 // we can get away with using one pass.
344 let is_base = line[..j].ends_with("<base");
346 let pos_equals = match rest.find("=") {
350 if rest[..pos_equals].trim_start_matches(" ") != "" {
354 let rest = &rest[pos_equals + 1..];
356 let pos_quote = match rest.find(&['"', '\''][..]) {
360 let quote_delim = rest.as_bytes()[pos_quote] as char;
362 if rest[..pos_quote].trim_start_matches(" ") != "" {
365 let rest = &rest[pos_quote + 1..];
366 let url = match rest.find(quote_delim) {
367 Some(i) => &rest[..i],