Change to use Pathos

pkg/docgen/lib/src/io.dart

+library io;
+/// This is a helper library to make working with io easier. 
+// TODO(janicejl): listDir, canonicalize, resolveLink, and linkExists are from  
+// pub/lib/src/io.dart. If the io.dart file becomes a package, should remove 
+// copy of the functions. 
+
+import 'dart:collection';
+import 'dart:io';
+import 'package:pathos/path.dart' as path;
+
+/// Lists the contents of [dir]. If [recursive] is `true`, lists subdirectory
+/// contents (defaults to `false`). If [includeHidden] is `true`, includes files
+/// and directories beginning with `.` (defaults to `false`).
+///
+/// The returned paths are guaranteed to begin with [dir].
+List<String> listDir(String dir, {bool recursive: false,
+    bool includeHidden: false}) {
+  List<String> doList(String dir, Set<String> listedDirectories) {
+    var contents = <String>[];
+
+    // Avoid recursive symlinks.
+    var resolvedPath = canonicalize(dir);
+    if (listedDirectories.contains(resolvedPath)) return [];
+
+    listedDirectories = new Set<String>.from(listedDirectories);
+    listedDirectories.add(resolvedPath);
+    
+    var children = <String>[];
+    for (var entity in new Directory(dir).listSync()) {
+      if (!includeHidden && path.basename(entity.path).startsWith('.')) {
+        continue;
+      }
+
+      contents.add(entity.path);
+      if (entity is Directory) {
+        // TODO(nweiz): don't manually recurse once issue 4794 is fixed.
+        // Note that once we remove the manual recursion, we'll need to
+        // explicitly filter out files in hidden directories.
+        if (recursive) {
+          children.addAll(doList(entity.path, listedDirectories));
+        }
+      }
+    }
+
+    contents.addAll(children);
+    return contents;
+  }
+
+  return doList(dir, new Set<String>());
+}
+
+/// Returns the canonical path for [pathString]. This is the normalized,
+/// absolute path, with symlinks resolved. As in [transitiveTarget], broken or
+/// recursive symlinks will not be fully resolved.
+///
+/// This doesn't require [pathString] to point to a path that exists on the
+/// filesystem; nonexistent or unreadable path entries are treated as normal
+/// directories.
+String canonicalize(String pathString) {
+  var seen = new Set<String>();
+  var components = new Queue<String>.from(
+      path.split(path.normalize(path.absolute(pathString))));
+
+  // The canonical path, built incrementally as we iterate through [components].
+  var newPath = components.removeFirst();
+
+  // Move through the components of the path, resolving each one's symlinks as
+  // necessary. A resolved component may also add new components that need to be
+  // resolved in turn.
+  while (!components.isEmpty) {
+    seen.add(path.join(newPath, path.joinAll(components)));
+    var resolvedPath = resolveLink(
+        path.join(newPath, components.removeFirst()));
+    var relative = path.relative(resolvedPath, from: newPath);
+
+    // If the resolved path of the component relative to `newPath` is just ".",
+    // that means component was a symlink pointing to its parent directory. We
+    // can safely ignore such components.
+    if (relative == '.') continue;
+
+    var relativeComponents = new Queue<String>.from(path.split(relative));
+
+    // If the resolved path is absolute relative to `newPath`, that means it's
+    // on a different drive. We need to canonicalize the entire target of that
+    // symlink again.
+    if (path.isAbsolute(relative)) {
+      // If we've already tried to canonicalize the new path, we've encountered
+      // a symlink loop. Avoid going infinite by treating the recursive symlink
+      // as the canonical path.
+      if (seen.contains(relative)) {
+        newPath = relative;
+      } else {
+        newPath = relativeComponents.removeFirst();
+        relativeComponents.addAll(components);
+        components = relativeComponents;
+      }
+      continue;
+    }
+
+    // Pop directories off `newPath` if the component links upwards in the
+    // directory hierarchy.
+    while (relativeComponents.first == '..') {
+      newPath = path.dirname(newPath);
+      relativeComponents.removeFirst();
+    }
+
+    // If there's only one component left, [resolveLink] guarantees that it's
+    // not a link (or is a broken link). We can just add it to `newPath` and
+    // continue resolving the remaining components.
+    if (relativeComponents.length == 1) {
+      newPath = path.join(newPath, relativeComponents.single);
+      continue;
+    }
+
+    // If we've already tried to canonicalize the new path, we've encountered a
+    // symlink loop. Avoid going infinite by treating the recursive symlink as
+    // the canonical path.
+    var newSubPath = path.join(newPath, path.joinAll(relativeComponents));
+    if (seen.contains(newSubPath)) {
+      newPath = newSubPath;
+      continue;
+    }
+
+    // If there are multiple new components to resolve, add them to the
+    // beginning of the queue.
+    relativeComponents.addAll(components);
+    components = relativeComponents;
+  }
+  return newPath;
+}
+
+/// Returns the transitive target of [link] (if A links to B which links to C,
+/// this will return C). If [link] is part of a symlink loop (e.g. A links to B
+/// which links back to A), this returns the path to the first repeated link (so
+/// `transitiveTarget("A")` would return `"A"` and `transitiveTarget("A")` would
+/// return `"B"`).
+///
+/// This accepts paths to non-links or broken links, and returns them as-is.
+String resolveLink(String link) {
+  var seen = new Set<String>();
+  while (linkExists(link) && !seen.contains(link)) {
+    seen.add(link);
+    link = path.normalize(path.join(
+        path.dirname(link), new Link(link).targetSync()));
+  }
+  return link;
+}
+
+/// Returns whether [link] exists on the file system. This will return `true`
+/// for any symlink, regardless of what it points at or whether it's broken.
+bool linkExists(String link) => new Link(link).existsSync();