Fix resolving absolute paths

The previous solution didn't resolve to absolute paths
and rather removed things like `.` and `..`.

Cargo.lock

@@ -1394,6 +1394,7 @@ dependencies = [
  "log",
  "markup5ever_rcdom",
  "openssl-sys",
+ "path-clean",
  "pretty_assertions",
  "pulldown-cmark",
  "regex",
@@ -1718,6 +1719,12 @@ dependencies = [
  "winapi",
 ]
 
+[[package]]
+name = "path-clean"
+version = "0.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ecba01bf2678719532c5e3059e0b5f0811273d94b397088b82e3bd0a78c78fdd"
+
 [[package]]
 name = "pem"
 version = "0.8.3"

README.md

@@ -148,11 +148,15 @@ lychee ~/projects/*/README.md
 
 # check links in local files (lychee supports advanced globbing and ~ expansion):
 lychee "~/projects/big_project/**/README.*"
+
 # ignore case when globbing and check result for each link:
 lychee --glob-ignore-case --verbose "~/projects/**/[r]eadme.*"
 
 # check links from epub file (requires atool: https://www.nongnu.org/atool)
 acat -F zip {file.epub} "*.xhtml" "*.html" | lychee -
+
+# check links in directory; block network requests
+lychee --offline path/to/directory
 ```
 
 ### GitHub token

lychee-lib/Cargo.toml

@@ -41,6 +41,7 @@ tokio = { version = "1.6.0", features = ["full"] }
 typed-builder = "0.9.1"
 url = { version = "2.2.2", features = ["serde"] }
 log = "0.4.14"
+path-clean = "0.1.0"
 
 [dev-dependencies]
 doc-comment = "0.3.3"

lychee-lib/src/extract.rs

@@ -39,7 +39,7 @@ pub(crate) fn extract_links(
                 // Silently ignore anchor links for now
                 continue;
             }
-            let uri = create_uri(root, base, &link)?;
+            let uri = create_uri_from_path(root, base, &link)?;
             Request::new(Uri { inner: uri }, input_content.input.clone())
         } else {
             info!("Handling of {} not implemented yet", &link);
@@ -122,7 +122,7 @@ fn extract_links_from_plaintext(input: &str) -> Vec<String> {
         .collect()
 }
 
-fn create_uri(root: &PathBuf, base: &Option<Base>, link: &str) -> Result<Url> {
+fn create_uri_from_path(root: &PathBuf, base: &Option<Base>, link: &str) -> Result<Url> {
     let link = url::remove_get_params(&link);
     let path = path::resolve(root, &PathBuf::from(&link), base)?;
     Ok(Url::from_file_path(&path).map_err(|_e| ErrorKind::InvalidPath(path))?)

lychee-lib/src/helpers/path.rs

@@ -1,45 +1,25 @@
 use crate::{Base, ErrorKind, Result};
-use std::path::{Component, Path, PathBuf};
+use path_clean::PathClean;
+use std::env;
+use std::path::{Path, PathBuf};
 
 // Returns the base if it is a valid `PathBuf`
 fn get_base_dir(base: &Option<Base>) -> Option<PathBuf> {
     base.as_ref().and_then(Base::dir)
 }
 
-/// Normalize a path, removing things like `.` and `..`.
-///
-/// CAUTION: This does not resolve symlinks (unlike
-/// [`std::fs::canonicalize`]). This may cause incorrect or surprising
-/// behavior at times. This should be used carefully. Unfortunately,
-/// [`std::fs::canonicalize`] can be hard to use correctly, since it can often
-/// fail, or on Windows returns annoying device paths. This is a problem Cargo
-/// needs to improve on.
-///
-/// Taken from [`cargo`](https://github.com/rust-lang/cargo/blob/fede83ccf973457de319ba6fa0e36ead454d2e20/src/cargo/util/paths.rs#L61)
-pub(crate) fn normalize(path: &Path) -> PathBuf {
-    let mut components = path.components().peekable();
+// https://stackoverflow.com/a/54817755/270334
+pub(crate) fn absolute_path(path: impl AsRef<Path>) -> Result<PathBuf> {
+    let path = path.as_ref();
 
-    let mut ret = components.peek().copied().map_or_else(PathBuf::new, |c| {
-        components.next();
-        PathBuf::from(c.as_os_str())
-    });
-
-    for component in components {
-        match component {
-            Component::Prefix(..) => unreachable!(),
-            Component::RootDir => {
-                ret.push(component.as_os_str());
-            }
-            Component::CurDir => {}
-            Component::ParentDir => {
-                ret.pop();
-            }
-            Component::Normal(c) => {
-                ret.push(c);
-            }
-        }
+    let absolute_path = if path.is_absolute() {
+        path.to_path_buf()
+    } else {
+        env::current_dir()?.join(path)
     }
-    ret
+    .clean();
+
+    Ok(absolute_path)
 }
 
 // Get the parent directory of a given `Path`.
@@ -59,7 +39,7 @@ pub(crate) fn resolve(src: &Path, dst: &Path, base: &Option<Base>) -> Result<Pat
         // Find `dst` in the parent directory of `src`
         if let Some(parent) = src.parent() {
             let rel_path = parent.join(dst.to_path_buf());
-            return Ok(normalize(&rel_path));
+            return Ok(absolute_path(&rel_path)?);
         }
     }
     if dst.is_absolute() {
@@ -73,7 +53,7 @@ pub(crate) fn resolve(src: &Path, dst: &Path, base: &Option<Base>) -> Result<Pat
             )
         })?;
         let abs_path = join(dirname(&base), dst);
-        return Ok(normalize(&abs_path));
+        return Ok(absolute_path(&abs_path)?);
     }
     Err(ErrorKind::FileNotFound(dst.to_path_buf()))
 }