diff --git a/src/markup/engine/remarkup/markuprule/PhutilRemarkupDocumentLinkRule.php b/src/markup/engine/remarkup/markuprule/PhutilRemarkupDocumentLinkRule.php
index 27c7290..a6effa0 100644
--- a/src/markup/engine/remarkup/markuprule/PhutilRemarkupDocumentLinkRule.php
+++ b/src/markup/engine/remarkup/markuprule/PhutilRemarkupDocumentLinkRule.php
@@ -1,175 +1,175 @@
 <?php
 
 final class PhutilRemarkupDocumentLinkRule extends PhutilRemarkupRule {
 
   public function getPriority() {
     return 150.0;
   }
 
   public function apply($text) {
     // Handle mediawiki-style links: [[ href | name ]]
     $text = preg_replace_callback(
       '@\B\\[\\[([^|\\]]+)(?:\\|([^\\]]+))?\\]\\]\B@U',
       array($this, 'markupDocumentLink'),
       $text);
 
     // Handle markdown-style links: [name](href)
     $text = preg_replace_callback(
       '@'.
         '\B'.
         '\\[([^\\]]+)\\]'.
         '\\('.
           '(\s*'.
             // See T12343. This is making some kind of effort to implement
             // parenthesis balancing rules. It won't get nested parentheses
             // right, but should do OK for Wikipedia pages, which seem to be
             // the most important use case.
 
             // Match zero or more non-parenthesis, non-space characters.
             '[^\s()]*'.
             // Match zero or more sequences of "(...)", where two balanced
             // parentheses enclose zero or more normal characters. If we
             // match some, optionally match more stuff at the end.
             '(?:(?:\\([^ ()]*\\))+[^\s()]*)?'.
           '\s*)'.
         '\\)'.
       '\B'.
       '@U',
       array($this, 'markupAlternateLink'),
       $text);
 
     return $text;
   }
 
   protected function renderHyperlink($link, $name) {
     $engine = $this->getEngine();
 
     $is_anchor = false;
     if (strncmp($link, '/', 1) == 0) {
       $base = $engine->getConfig('uri.base');
       $base = rtrim($base, '/');
       $link = $base.$link;
     } else if (strncmp($link, '#', 1) == 0) {
       $here = $engine->getConfig('uri.here');
       $link = $here.$link;
 
       $is_anchor = true;
     }
 
     if ($engine->isTextMode()) {
       // If present, strip off "mailto:" or "tel:".
       $link = preg_replace('/^(?:mailto|tel):/', '', $link);
 
       if (!strlen($name)) {
         return $link;
       }
 
       return $name.' <'.$link.'>';
     }
 
     if (!strlen($name)) {
       $name = $link;
       $name = preg_replace('/^(?:mailto|tel):/', '', $name);
     }
 
     if ($engine->getState('toc')) {
       return $name;
     }
 
     $same_window = $engine->getConfig('uri.same-window', false);
     if ($same_window) {
       $target = null;
     } else {
       $target = '_blank';
     }
 
     // For anchors on the same page, always stay here.
     if ($is_anchor) {
       $target = null;
     }
 
     return phutil_tag(
       'a',
       array(
         'href' => $link,
         'class' => 'remarkup-link',
         'target' => $target,
         'rel' => 'noreferrer',
       ),
       $name);
   }
 
   public function markupAlternateLink(array $matches) {
     $uri = trim($matches[2]);
 
+    if (!strlen($uri)) {
+      return $matches[0];
+    }
+
     // NOTE: We apply some special rules to avoid false positives here. The
     // major concern is that we do not want to convert `x[0][1](y)` in a
     // discussion about C source code into a link. To this end, we:
     //
     //   - Don't match at word boundaries;
     //   - require the URI to contain a "/" character or "@" character; and
     //   - reject URIs which being with a quote character.
 
     if ($uri[0] == '"' || $uri[0] == "'" || $uri[0] == '`') {
       return $matches[0];
     }
 
-    if (!strlen($uri[0])) {
-      return $matches[0];
-    }
-
     if (strpos($uri, '/') === false &&
         strpos($uri, '@') === false &&
         strncmp($uri, 'tel:', 4)) {
       return $matches[0];
     }
 
     return $this->markupDocumentLink(
       array(
         $matches[0],
         $matches[2],
         $matches[1],
       ));
   }
 
   public function markupDocumentLink(array $matches) {
     $uri = trim($matches[1]);
     $name = trim(idx($matches, 2));
 
     // If whatever is being linked to begins with "/" or "#", or has "://",
     // or is "mailto:" or "tel:", treat it as a URI instead of a wiki page.
     $is_uri = preg_match('@(^/)|(://)|(^#)|(^(?:mailto|tel):)@', $uri);
 
     if ($is_uri && strncmp('/', $uri, 1) && strncmp('#', $uri, 1)) {
       $protocols = $this->getEngine()->getConfig(
         'uri.allowed-protocols',
         array());
 
       try {
         $protocol = id(new PhutilURI($uri))->getProtocol();
         if (!idx($protocols, $protocol)) {
           // Don't treat this as a URI if it's not an allowed protocol.
           $is_uri = false;
         }
       } catch (Exception $ex) {
         // We can end up here if we try to parse an ambiguous URI, see
         // T12796.
         $is_uri = false;
       }
     }
 
     // As a special case, skip "[[ / ]]" so that Phriction picks it up as a
     // link to the Phriction root. It is more useful to be able to use this
     // syntax to link to the root document than the home page of the install.
     if ($uri == '/') {
       $is_uri = false;
     }
 
     if (!$is_uri) {
       return $matches[0];
     }
 
     return $this->getEngine()->storeText($this->renderHyperlink($uri, $name));
   }
 
 }