diff --git a/src/applications/project/remarkup/ProjectRemarkupRule.php b/src/applications/project/remarkup/ProjectRemarkupRule.php index 2f93a88e95..ca11a7ea8c 100644 --- a/src/applications/project/remarkup/ProjectRemarkupRule.php +++ b/src/applications/project/remarkup/ProjectRemarkupRule.php @@ -1,62 +1,62 @@ getEngine()->isTextMode()) { return '#'.$id; } return $handle->renderTag(); } protected function getObjectIDPattern() { // NOTE: This explicitly does not match strings which contain only // digits, because digit strings like "#123" are used to reference tasks at // Facebook and are somewhat conventional in general. // The latter half of this rule matches monograms with internal periods, // like `#domain.com`, but does not match monograms with terminal periods, // because they're probably just puncutation. // Broadly, this will not match every possible project monogram, and we // accept some false negatives -- like `#1` or `#dot.` -- in order to avoid // a bunch of false positives on general use of the `#` character. // In other contexts, the PhabricatorProjectProjectPHIDType pattern is // controlling and these names should parse correctly. - return '[^\s.\d!,:;{}#]+(?:[^\s!,:;{}#][^\s.!,:;{}#]+)*'; + return '[^\s.\d!,:;{}#\(\)]+(?:[^\s!,:;{}#\(\)][^\s.!,:;{}#\(\)]+)*'; } protected function loadObjects(array $ids) { $viewer = $this->getEngine()->getConfig('viewer'); // Put the "#" back on the front of these IDs. $names = array(); foreach ($ids as $id) { $names[] = '#'.$id; } // Issue a query by object name. $query = id(new PhabricatorObjectQuery()) ->setViewer($viewer) ->withNames($names); $query->execute(); $projects = $query->getNamedResults(); // Slice the "#" off again. $result = array(); foreach ($projects as $name => $project) { $result[substr($name, 1)] = $project; } return $result; } } diff --git a/src/applications/project/remarkup/__tests__/ProjectRemarkupRuleTestCase.php b/src/applications/project/remarkup/__tests__/ProjectRemarkupRuleTestCase.php index 4e72514b26..4efc98a5b8 100644 --- a/src/applications/project/remarkup/__tests__/ProjectRemarkupRuleTestCase.php +++ b/src/applications/project/remarkup/__tests__/ProjectRemarkupRuleTestCase.php @@ -1,57 +1,82 @@ array( 'embed' => array(), 'ref' => array( array( 'offset' => 8, 'id' => 'ducks', ), ), ), 'We should make a post on #blog.example.com tomorrow.' => array( 'embed' => array(), 'ref' => array( array( 'offset' => 26, 'id' => 'blog.example.com', ), ), ), 'We should make a post on #blog.example.com.' => array( 'embed' => array(), 'ref' => array( array( 'offset' => 26, 'id' => 'blog.example.com', ), ), ), '#123' => array( 'embed' => array(), 'ref' => array(), ), '#security#123' => array( 'embed' => array(), 'ref' => array( array( 'offset' => 1, 'id' => 'security', 'tail' => '123', ), ), ), + + // Don't match a terminal parenthesis. This fixes these constructs in + // natural language. + 'There is some documentation (see #guides).' => array( + 'embed' => array(), + 'ref' => array( + array( + 'offset' => 34, + 'id' => 'guides', + ), + ), + ), + + // Don't match internal parentheses either. This makes the terminal + // parenthesis behavior less arbitrary (otherwise, we match open + // parentheses but not closing parentheses, which is surprising). + '#a(b)c' => array( + 'embed' => array(), + 'ref' => array( + array( + 'offset' => 1, + 'id' => 'a', + ), + ), + ), ); foreach ($cases as $input => $expect) { $rule = new ProjectRemarkupRule(); $matches = $rule->extractReferences($input); $this->assertEqual($expect, $matches, $input); } } }