Remember anonymous comment posters. (UTF-8 bug)

Here is my first go at this patch.

The css is in system.css

When caching is disabled the session stores the anonymous users info.

When caching is enabled jQuery and cookies do the remembering. I split up the javascript. The cookie code is the jQuery cookie plugin and in is /misc/jquery.cookie.js. This can be used by other modules and is the .cookie code described at visualjquery.com.

The rest of the javascript is in /modules/comment/comment.js.

Simplification = good

@todo: Remove checkbox to remember. Just do it.
@todo: Remove the on/off setting.

I thought about requesting the information with jQuery from drupal but figured it wasn't worth the performance hit.

Time to do some digging into how wordpress does this.

Here is another shot that is much more simplified. The jquery.cookie.js in /misc is compressed with packer. All of the configuration settings are gone and it just works when the fields are there. The comment.js file has been greatly simplified.

Lets try that attachment again

added "// $Id: $" to the comment.js file

The compressed function is jQuery.cookie. It's the jQuery cookie plugin from jquery.com. you can read about it's usage at http://www.visualjquery.com in the plugins section. The function reads:

jQuery.cookie = function(name, value, options) {
  if (typeof value != 'undefined') { // name and value given, set cookie
    options = options || {};
    var expires = '';
    if (options.expires && (typeof options.expires == 'number' || options.expires.toGMTString)) {
      var date;
      if (typeof options.expires == 'number') {
        date = new Date();
        date.setTime(date.getTime() + (options.expires * 24 * 60 * 60 * 1000));
      } else {
        date = options.expires;
      }
      expires = '; expires=' + date.toGMTString(); // use expires attribute, max-age is not supported by IE
    }
    var path = options.path ? '; path=' + options.path : '';
    var domain = options.domain ? '; domain=' + options.domain : '';
    var secure = options.secure ? '; secure' : '';
    document.cookie = [name, '=', encodeURIComponent(value), expires, path, domain, secure].join('');
  } else { // only name given, get cookie
    var cookieValue = null;
    if (document.cookie && document.cookie != '') {
      var cookies = document.cookie.split(';');
      for (var i = 0; i < cookies.length; i++) {
        var cookie = jQuery.trim(cookies[i]);
        // Does this cookie string begin with the name we want?
        if (cookie.substring(0, name.length + 1) == (name + '=')) {
          cookieValue = decodeURIComponent(cookie.substring(name.length + 1));
          break;
        }
      }
    }
    return cookieValue;
  }
};

I figured it is better to use the jQuery plugin than to roll my own. And, to separate it out in case anyone else wants to use it.

I compared the jQuery.cookie code to some roll your own cookie reading and writing javascript in the contrib modules. Going with custom cookie javascript can save about 300 bytes of javascript.

This patch sets the cookie in the comment_form_submit and reads it with javascript. The jquery.cookie plugin is not used here and a custom function is used to read the cookie. It's less javascript.

The only question I have is how to word the watchdog event when a cookie isn't sent from php.

This or the solution in #7 both work. They just solve it different ways.

fixed the concatenation and spacing issues. Changed setcookie to setrawcookie with rawurlencode so the escape characters would be the same between php and javascript. Updated javascript to returnValue.

Moved setting the cookie from _submit to _validate to work with preview. When it saves it in _submit, you have saved values, you change a value and hit preview, the value would revert.

Still looking in to how to do it with one cookie.

This patch moves everything into one cookie and is post FAPI3.

I'm not sure what I was thinking with that warning message. I removed it in this patch.

I moved everything to one cookie due to the suggestion by ChrisKennedy in #13. It seemed to make sense. There is a 20 cookie limit per domain per rfc2109 which is what most browsers including IE and Firefox go by.

If we go with 3 cookies that would put drupal core at 4 cookie (with the phpsession). This way there are just 2 for drupal core. One for the session and one for the anonymous comment information. This leaves 18 open cookies rather than 16 for other modules or apps to use.

3 cookies it is :-D

Back to 3 cookies. This is similar to #14.

I, also, found that if the cookie doesn't exist we shouldn't be overwriting the value of the form field with '' which is what the cookie variable in the javascript is initialized with. So, if the cookie is empty no changes are made to that part of the form. This may come in handy if someone tries to theme the form to add default values.

@dries - The reason for the setrawcookie and rawurlencode has to do with the difference between php and javascript handling spaces, which drupal allows in the username. Some details in the php handbook at http://us2.php.net/manual/en/function.setcookie.php#62848.

Basically, setcookie uses urlencode to encode the cookie. This function is not RFC 1738 compliant and encodes spaces as a + which javascript does does not turn back into a space. According to the php manual on urlencode, this functionality is used for historical reasons. rawurlencode is RFC 1738 compliant and uses the same encoding as javascript.

@Senpai: I'm not sure that the 3 cookies will really be much of an issue compared to one. The trade off is about 10 more lines of javascript for that handful of users while wordpress and others are already using 3 cookies for this which has set a precedence.

Dries wanted the 3 and I'm really OK with that. If you want to submit an issue I'll subscribe to it and roll a patch if the powers at be deem it worthy.

oops. Looks like unescape() is deprecated. decodeURIComponent() seems to be the utf8 compliant solution here.

I haven't tested it. I'll put that on my to do list for the weekend but everything I've read says this patch has it right.

Nice catch.

erdemkose and webchick, nice catches. That'll teach me to read the manual and to not only develop in php5. :-)

This patch fixes the uft bug of eremkose and worked for me on php4.