Duplicate query strings such as array parameters are removed

The problem stems from the function _link_parse_str($query), called by _link_parse_url($url).

function _link_parse_str($query) {
  $query_array = array();

  $pairs = explode('&', $query);
  foreach ($pairs as $pair) {
    $name_value = explode('=', $pair, 2);
    $name = urldecode($name_value[0]);
    $value = isset($name_value[1]) ? urldecode($name_value[1]) : NULL;
    $query_array[$name] = $value;
  
  return $query_array;
}

It's splitting the query string up into components & storing it in an array for use by _link_parse_url, which creates URL parts (path, query string, fragment) for use in _link_sanitize(), which runs those parts through the Drupal function url() to create & display the URL.

The problem arises from the fact that _link_parse_str is creating an array whose keys are the query parameter names. In the case here, the query parameter names are identical, so the array keys are identical. In that case PHP only keeps the last array key, hence the cut off query string when it's all done.

Since Drupal's url() (https://api.drupal.org/api/drupal/includes!common.inc/function/url/7) can accept a query string as part of the 'path' parameter, why not just run the whole input URL into the 'path' parameter rather than splitting off the query string, making it an array, & passing it in as the 'query' parameter? What are the advantages of using _link_parse_str($query) & _link_parse_url($url) to get the query part as an array & passing it into url() as such rather than just plugging the whole URL into the 'path' parameter of url()?

I'm attaching a patch which solves this problem by taking the approach outlined above. Not sure if it has some liability since I'm not clear on why it's being done the way it is currently--I could be missing something here.

By the way, it is possible to cause _link_parse_str to create a multidimensional array which will create a URL which preserves both identical query parameters, but then those parameters are suffixed with the integer key/index of the query string values, e.g.: page?content_type[0]=foo&content_type[1]=bar. Simply change $query_array[$name] = $value; to $query_array[$name][] = $value; (or a more complicated version of that with if/then checks a la https://php.net/manual/en/reserved.variables.get.php#92439 ).

PS: The URL module has this same problem; while it will display a URL such as page?content_type=foo&content_type=bar correctly, the HTML link it creates (href) actually is truncated exactly as discribed in this problem. It's using drupal_parse_url() to split up the URL parts & passing them into url just like this module.

The patch I uploaded has been working for me for two years. Maybe somebody can test it or one of the re-rolls & mark it RTBC, & perhaps it can get into the module eventually?