src/tools/linkchecker/main.rs

   1 // Copyright 2016 The Rust Project Developers. See the COPYRIGHT
   2 // file at the top-level directory of this distribution and at
   3 // http://rust-lang.org/COPYRIGHT.
   4 //
   5 // Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
   6 // http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
   7 // <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
   8 // option. This file may not be copied, modified, or distributed
   9 // except according to those terms.
  10
  11 //! Script to check the validity of `href` links in our HTML documentation.
  12 //!
  13 //! In the past we've been quite error prone to writing in broken links as most
  14 //! of them are manually rather than automatically added. As files move over
  15 //! time or apis change old links become stale or broken. The purpose of this
  16 //! script is to check all relative links in our documentation to make sure they
  17 //! actually point to a valid place.
  18 //!
  19 //! Currently this doesn't actually do any HTML parsing or anything fancy like
  20 //! that, it just has a simple "regex" to search for `href` and `id` tags.
  21 //! These values are then translated to file URLs if possible and then the
  22 //! destination is asserted to exist.
  23 //!
  24 //! A few whitelisted exceptions are allowed as there's known bugs in rustdoc,
  25 //! but this should catch the majority of "broken link" cases.
  26
  27 use std::env;
  28 use std::fs::File;
  29 use std::io::prelude::*;
  30 use std::path::{Path, PathBuf, Component};
  31 use std::collections::{HashMap, HashSet};
  32 use std::collections::hash_map::Entry;
  33
  34 use Redirect::*;
  35
  36 macro_rules! t {
  37     ($e:expr) => (match $e {
  38         Ok(e) => e,
  39         Err(e) => panic!("{} failed with {:?}", stringify!($e), e),
  40     })
  41 }
  42
  43 fn main() {
  44     let docs = env::args_os().nth(1).unwrap();
  45     let docs = env::current_dir().unwrap().join(docs);
  46     let mut errors = false;
  47     walk(&mut HashMap::new(), &docs, &docs, &mut errors);
  48     if errors {
  49         panic!("found some broken links");
  50     }
  51 }
  52
  53 #[derive(Debug)]
  54 pub enum LoadError {
  55     IOError(std::io::Error),
  56     BrokenRedirect(PathBuf, std::io::Error),
  57     IsRedirect,
  58 }
  59
  60 enum Redirect {
  61     SkipRedirect,
  62     FromRedirect(bool),
  63 }
  64
  65 struct FileEntry {
  66     source: String,
  67     ids: HashSet<String>,
  68 }
  69
  70 type Cache = HashMap<PathBuf, FileEntry>;
  71
  72 fn small_url_encode(s: &str) -> String {
  73     s.replace("<", "%3C")
  74      .replace(">", "%3E")
  75      .replace(" ", "%20")
  76      .replace("?", "%3F")
  77      .replace("'", "%27")
  78      .replace("&", "%26")
  79      .replace(",", "%2C")
  80      .replace(":", "%3A")
  81      .replace(";", "%3B")
  82      .replace("[", "%5B")
  83      .replace("]", "%5D")
  84      .replace("\"", "%22")
  85 }
  86
  87 impl FileEntry {
  88     fn parse_ids(&mut self, file: &Path, contents: &str, errors: &mut bool) {
  89         if self.ids.is_empty() {
  90             with_attrs_in_source(contents, " id", |fragment, i, _| {
  91                 let frag = fragment.trim_left_matches("#").to_owned();
  92                 let encoded = small_url_encode(&frag);
  93                 if !self.ids.insert(frag) {
  94                     *errors = true;
  95                     println!("{}:{}: id is not unique: `{}`", file.display(), i, fragment);
  96                 }
  97                 // Just in case, we also add the encoded id.
  98                 self.ids.insert(encoded);
  99             });
 100         }
 101     }
 102 }
 103
 104 fn walk(cache: &mut Cache, root: &Path, dir: &Path, errors: &mut bool) {
 105     for entry in t!(dir.read_dir()).map(|e| t!(e)) {
 106         let path = entry.path();
 107         let kind = t!(entry.file_type());
 108         if kind.is_dir() {
 109             walk(cache, root, &path, errors);
 110         } else {
 111             let pretty_path = check(cache, root, &path, errors);
 112             if let Some(pretty_path) = pretty_path {
 113                 let entry = cache.get_mut(&pretty_path).unwrap();
 114                 // we don't need the source anymore,
 115                 // so drop to reduce memory-usage
 116                 entry.source = String::new();
 117             }
 118         }
 119     }
 120 }
 121
 122 fn check(cache: &mut Cache,
 123          root: &Path,
 124          file: &Path,
 125          errors: &mut bool)
 126          -> Option<PathBuf> {
 127     // Ignore none HTML files.
 128     if file.extension().and_then(|s| s.to_str()) != Some("html") {
 129         return None;
 130     }
 131
 132     // Unfortunately we're not 100% full of valid links today to we need a few
 133     // whitelists to get this past `make check` today.
 134     // FIXME(#32129)
 135     if file.ends_with("std/string/struct.String.html") ||
 136        file.ends_with("interpret/struct.ValTy.html") ||
 137        file.ends_with("symbol/struct.InternedString.html") ||
 138        file.ends_with("ast/struct.ThinVec.html") ||
 139        file.ends_with("util/struct.ThinVec.html") ||
 140        file.ends_with("util/struct.RcSlice.html") ||
 141        file.ends_with("layout/struct.TyLayout.html") ||
 142        file.ends_with("humantime/struct.Timestamp.html") ||
 143        file.ends_with("log/index.html") ||
 144        file.ends_with("ty/struct.Slice.html") ||
 145        file.ends_with("ty/enum.Attributes.html") ||
 146        file.ends_with("ty/struct.SymbolName.html") {
 147         return None;
 148     }
 149     // FIXME(#32553)
 150     if file.ends_with("string/struct.String.html") {
 151         return None;
 152     }
 153     // FIXME(#32130)
 154     if file.ends_with("btree_set/struct.BTreeSet.html") ||
 155        file.ends_with("struct.BTreeSet.html") ||
 156        file.ends_with("btree_map/struct.BTreeMap.html") ||
 157        file.ends_with("hash_map/struct.HashMap.html") ||
 158        file.ends_with("hash_set/struct.HashSet.html") ||
 159        file.ends_with("sync/struct.Lrc.html") ||
 160        file.ends_with("sync/struct.RwLock.html") {
 161         return None;
 162     }
 163
 164     let res = load_file(cache, root, file, SkipRedirect);
 165     let (pretty_file, contents) = match res {
 166         Ok(res) => res,
 167         Err(_) => return None,
 168     };
 169     {
 170         cache.get_mut(&pretty_file)
 171              .unwrap()
 172              .parse_ids(&pretty_file, &contents, errors);
 173     }
 174
 175     // Search for anything that's the regex 'href[ ]*=[ ]*".*?"'
 176     with_attrs_in_source(&contents, " href", |url, i, base| {
 177         // Ignore external URLs
 178         if url.starts_with("http:") || url.starts_with("https:") ||
 179            url.starts_with("javascript:") || url.starts_with("ftp:") ||
 180            url.starts_with("irc:") || url.starts_with("data:") {
 181             return;
 182         }
 183         let mut parts = url.splitn(2, "#");
 184         let url = parts.next().unwrap();
 185         let fragment = parts.next();
 186         let mut parts = url.splitn(2, "?");
 187         let url = parts.next().unwrap();
 188
 189         // Once we've plucked out the URL, parse it using our base url and
 190         // then try to extract a file path.
 191         let mut path = file.to_path_buf();
 192         if !base.is_empty() || !url.is_empty() {
 193             path.pop();
 194             for part in Path::new(base).join(url).components() {
 195                 match part {
 196                     Component::Prefix(_) |
 197                     Component::RootDir => {
 198                         // Avoid absolute paths as they make the docs not
 199                         // relocatable by making assumptions on where the docs
 200                         // are hosted relative to the site root.
 201                         *errors = true;
 202                         println!("{}:{}: absolute path - {}",
 203                                  pretty_file.display(),
 204                                  i + 1,
 205                                  Path::new(base).join(url).display());
 206                         return;
 207                     }
 208                     Component::CurDir => {}
 209                     Component::ParentDir => { path.pop(); }
 210                     Component::Normal(s) => { path.push(s); }
 211                 }
 212             }
 213         }
 214
 215         // Alright, if we've found a file name then this file had better
 216         // exist! If it doesn't then we register and print an error.
 217         if path.exists() {
 218             if path.is_dir() {
 219                 // Links to directories show as directory listings when viewing
 220                 // the docs offline so it's best to avoid them.
 221                 *errors = true;
 222                 let pretty_path = path.strip_prefix(root).unwrap_or(&path);
 223                 println!("{}:{}: directory link - {}",
 224                          pretty_file.display(),
 225                          i + 1,
 226                          pretty_path.display());
 227                 return;
 228             }
 229             if let Some(extension) = path.extension() {
 230                 // Ignore none HTML files.
 231                 if extension != "html" {
 232                     return;
 233                 }
 234             }
 235             let res = load_file(cache, root, &path, FromRedirect(false));
 236             let (pretty_path, contents) = match res {
 237                 Ok(res) => res,
 238                 Err(LoadError::IOError(err)) => {
 239                     panic!("error loading {}: {}", path.display(), err);
 240                 }
 241                 Err(LoadError::BrokenRedirect(target, _)) => {
 242                     *errors = true;
 243                     println!("{}:{}: broken redirect to {}",
 244                              pretty_file.display(),
 245                              i + 1,
 246                              target.display());
 247                     return;
 248                 }
 249                 Err(LoadError::IsRedirect) => unreachable!(),
 250             };
 251
 252             if let Some(ref fragment) = fragment {
 253                 // Fragments like `#1-6` are most likely line numbers to be
 254                 // interpreted by javascript, so we're ignoring these
 255                 if fragment.splitn(2, '-')
 256                            .all(|f| f.chars().all(|c| c.is_numeric())) {
 257                     return;
 258                 }
 259
 260                 // These appear to be broken in mdbook right now?
 261                 if fragment.starts_with("-") {
 262                     return;
 263                 }
 264
 265                 let entry = &mut cache.get_mut(&pretty_path).unwrap();
 266                 entry.parse_ids(&pretty_path, &contents, errors);
 267
 268                 if !entry.ids.contains(*fragment) {
 269                     *errors = true;
 270                     print!("{}:{}: broken link fragment ",
 271                            pretty_file.display(),
 272                            i + 1);
 273                     println!("`#{}` pointing to `{}`", fragment, pretty_path.display());
 274                 };
 275             }
 276         } else {
 277             *errors = true;
 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());
 281         }
 282     });
 283     Some(pretty_file)
 284 }
 285
 286 fn load_file(cache: &mut Cache,
 287              root: &Path,
 288              file: &Path,
 289              redirect: Redirect)
 290              -> Result<(PathBuf, String), LoadError> {
 291     let mut contents = String::new();
 292     let pretty_file = PathBuf::from(file.strip_prefix(root).unwrap_or(&file));
 293
 294     let maybe_redirect = match cache.entry(pretty_file.clone()) {
 295         Entry::Occupied(entry) => {
 296             contents = entry.get().source.clone();
 297             None
 298         }
 299         Entry::Vacant(entry) => {
 300             let mut fp = File::open(file).map_err(|err| {
 301                 if let FromRedirect(true) = redirect {
 302                     LoadError::BrokenRedirect(file.to_path_buf(), err)
 303                 } else {
 304                     LoadError::IOError(err)
 305                 }
 306             })?;
 307             fp.read_to_string(&mut contents).map_err(|err| LoadError::IOError(err))?;
 308
 309             let maybe = maybe_redirect(&contents);
 310             if maybe.is_some() {
 311                 if let SkipRedirect = redirect {
 312                     return Err(LoadError::IsRedirect);
 313                 }
 314             } else {
 315                 entry.insert(FileEntry {
 316                     source: contents.clone(),
 317                     ids: HashSet::new(),
 318                 });
 319             }
 320             maybe
 321         }
 322     };
 323     match maybe_redirect.map(|url| file.parent().unwrap().join(url)) {
 324         Some(redirect_file) => {
 325             load_file(cache, root, &redirect_file, FromRedirect(true))
 326         }
 327         None => Ok((pretty_file, contents)),
 328     }
 329 }
 330
 331 fn maybe_redirect(source: &str) -> Option<String> {
 332     const REDIRECT: &'static str = "<p>Redirecting to <a href=";
 333
 334     let mut lines = source.lines();
 335     let redirect_line = match lines.nth(6) {
 336         Some(l) => l,
 337         None => return None,
 338     };
 339
 340     redirect_line.find(REDIRECT).map(|i| {
 341         let rest = &redirect_line[(i + REDIRECT.len() + 1)..];
 342         let pos_quote = rest.find('"').unwrap();
 343         rest[..pos_quote].to_owned()
 344     })
 345 }
 346
 347 fn with_attrs_in_source<F: FnMut(&str, usize, &str)>(contents: &str, attr: &str, mut f: F) {
 348     let mut base = "";
 349     for (i, mut line) in contents.lines().enumerate() {
 350         while let Some(j) = line.find(attr) {
 351             let rest = &line[j + attr.len()..];
 352             // The base tag should always be the first link in the document so
 353             // we can get away with using one pass.
 354             let is_base = line[..j].ends_with("<base");
 355             line = rest;
 356             let pos_equals = match rest.find("=") {
 357                 Some(i) => i,
 358                 None => continue,
 359             };
 360             if rest[..pos_equals].trim_left_matches(" ") != "" {
 361                 continue;
 362             }
 363
 364             let rest = &rest[pos_equals + 1..];
 365
 366             let pos_quote = match rest.find(&['"', '\''][..]) {
 367                 Some(i) => i,
 368                 None => continue,
 369             };
 370             let quote_delim = rest.as_bytes()[pos_quote] as char;
 371
 372             if rest[..pos_quote].trim_left_matches(" ") != "" {
 373                 continue;
 374             }
 375             let rest = &rest[pos_quote + 1..];
 376             let url = match rest.find(quote_delim) {
 377                 Some(i) => &rest[..i],
 378                 None => continue,
 379             };
 380             if is_base {
 381                 base = url;
 382                 continue;
 383             }
 384             f(url, i, base)
 385         }
 386     }
 387 }