URL array variables (?var[]=value) results not properly cached

Quick fix is to turn off "Only allow ASCII characters in path". I'll be looking for a proper fix soon.

above comment is false. Root of the issue has to do with multi dimensional arrays in the $GET php variable. Here is the problem code

  foreach ($_GET as $key => $val) {
    if ($key != 'q' && $key != 'destination') {
      if (!empty($val)) {
        $GLOBALS['_boost_query'] .= (($GLOBALS['_boost_query'] == BOOST_CHAR) ? '' : '&') . $key . '=' . $val;
      }

$GET structure for specials?page=1&field_type_value[0]=3 & specials?page=1&field_type_value[]=3

	Array

      [q] => specials
      [page] => 1
      [field_type_value] => Array

            [0] => 3

I'll be searching for code that reverses this; reproduce the URL given the $GET variable. I would rather not use $_SERVER['REQUEST_URI']; but I might have to. Looks like I need to have 2 sets of URLs as well, [0] & []

http://php.net/http-build-query PHP 5 Function with it's PHP 4 backwards compatibility function

// php.net/http-build-query#90438
if (!function_exists('http_build_query')) {
  function http_build_query($data, $prefix='', $sep='', $key='') {
    $ret = array();
    foreach ((array)$data as $k => $v) {
      if (is_int($k) && $prefix != null) {
        $k = urlencode($prefix . $k);
      }
      if ((!empty($key)) || ($key === 0)) {
        $k = $key.'['.urlencode($k).']';
      }
      if (is_array($v) || is_object($v)) {
        array_push($ret, http_build_query($v, '', $sep, $k));
      }
      else {
        array_push($ret, $k.'='.urlencode($v));
      }
    }
    if (empty($sep)) {
      $sep = ini_get('arg_separator.output');
    }
    return implode($sep, $ret);
  }
}

From this I need to do a http://php.net/urldecode on the string; and then deal with [] and [0]; [] [] and [0] [1]

Opened up an issue in views for the duplicate URL issue
#637900: More consistent looking URL's when submitting exposed filter

[] are valid character names so why would I replace them with _? That would create a huge headache in htaccess. Long story short, I need to cache these 2 different files, one with [] the other with [0]; which brings the total number of files created with one action up to a potential 8 files... kinda nuts but necessary. This will take some thinking on how to do this correctly. I've started to implement the framework for this since boost can create up to 4 files for each cache action already, using this function

/**
 * Returns all possible filenames given the input and current settings
 *
 * @param $filename
 *   Name of file
 * @param $base_dir
 *   Value from base_dir column in database
 * @return
 *   returns a 2 dimensional array
 *    1st dimension key is either gzip or normal
 *    2nd dimension contains all the filenames
 */
function boost_get_all_filenames($filename, $base_dir = NULL) {
  $names = array();
  $filenames = array();
  $names[] = $filename;
  $paths = explode('/', $filename);
  $end = array_pop($paths);
  $base_dir = is_null($base_dir) ? BOOST_FILE_PATH : $base_dir;

  // Generate urlencoded filename, if name contains decoded characters
  $chars = '*:<>|?#,';
  if (preg_match("/[" . $chars . "]/", $end)) {
    $end = urlencode($end);
    // '=' sign should not be encoded
    $end = str_replace('%3D', '=', $end);
    // '&' should not be encoded
    $end = str_replace('%26', '&', $end);
    // '#' should not be encoded
    $end = str_replace('%23', '#', $end);
    // '?' is not an issue because the first ? is converted to a _ with the default settings.
    // If not using the default then this could be an issue. Will deal with it when that happens.
    $paths[] = $end;
    $names[] = implode('/', $paths);
  }

  // Generate gzip filenames
  foreach ($names as $name) {
    $filenames['normal'][] = $name;
    if (BOOST_GZIP) {
      // Replace the correct dir with the gzip version for the given base dir.
      $gzip_base_path = implode('/', array_filter(explode('/', str_replace(BOOST_ROOT_CACHE_DIR . '/' . BOOST_NORMAL_DIR, BOOST_ROOT_CACHE_DIR . '/' . BOOST_GZIP_DIR . '/', $base_dir))));
      $filenames['gzip'][] = str_replace($base_dir, $gzip_base_path, $name) . BOOST_GZIP_EXTENSION;
    }
  }
  return $filenames;
}

I should always have the number in the brackets [0] and then use the boost_get_all_filenames() function to strip all numbers out of []. This means I can not use $_SERVER['REQUEST_URI'] and must use something like http_build_query()

Plan:
Copy $GET to $query
remove q & destination from $query
run $query through http_build_query()
run that through urldecode()
this is the filename I save in the database
boost_get_all_filenames() will remove all [n] and replace with [] as a variant of the filename - will probably have to use regex to do this.

http_build_query() explodes the array for me, using built in php functions rather then my own.

committed

@kees
I already took care of that, each page that contains a [0] will also get a page that has [], at the same time. One action will generate 2 files. As merlinofchaos stated, it's a browser issue.

Is there a reason why saving the link twice would cause any harm?

if "Asynchronous Operation: output HTML, close connection, then store static file" is enabled then the performance impact should be minimal. My guess is that the worse case, is an extra 20ms added due to this extra file being written; something I'm not too worried about since its a one time occurrence; a cache miss is generally more expensive then writing an extra file.