25686 2014-05-13 17:29:36 +0000 alert() needs two overloads so alert(undefined) can alert "undefined" 2014-05-16 16:54:15 +0000 1 1 1 Unclassified WHATWG HTML unspecified Other other RESOLVED FIXED http://www.whatwg.org/specs/web-apps/current-work/#the-window-object P3 normal Unsorted 1 contributor ian bzbarsky d ian mike contributor oldest_to_newest 105871 0 contributor 2014-05-13 17:29:36 +0000 Specification: http://www.whatwg.org/specs/web-apps/current-work/multipage/browsers.html Multipage: http://www.whatwg.org/C#the-window-object Complete: http://www.whatwg.org/c#the-window-object Referrer: http://www.whatwg.org/specs/web-apps/current-work/multipage/ Comment: alert() needs two overloads so alert(undefined) can alert "undefined" Posted from: 63.245.221.34 by bzbarsky@mit.edu User agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.9; rv:32.0) Gecko/20100101 Firefox/32.0 105872 1 bzbarsky 2014-05-13 17:34:44 +0000 People are asking for alert(undefined) to show "undefined". Current behavior in browsers: Chrome, Firefox 27 and earlier, IE11: alert() shows "" alert(undefined) shows "undefined" Firefox 28 and later, current spec: alert() shows "" alert(undefined) shows "" Safari 7: alert() shows "undefined" alert(undefined) shows "undefined" To get the Chrome/IE/old-Firefox behavior, the IDL needs to be: void alert(); void alert(DOMString message); with prose that says that the first overload uses "" as the message. To get the Safari behavior, the IDL can be: void alert(optional DOMString message = "undefined"); For now I'm switching Firefox to our old behavior over in https://bugzilla.mozilla.org/show_bug.cgi?id=999315 105873 2 bzbarsky 2014-05-13 17:35:38 +0000 But note that the Safari behavior here is more JavaScripty in principle. 105874 3 d 2014-05-13 17:39:50 +0000 Just to chime in: Both the Firefox 28+ and Safari 7+ behaviors are reasonably JavaScriptey. Firefox 28+ behaves as if it were function alert(message = "") { message = String(message); // standard type coercion here } and Safari 7+ behaves as if it were function alert(message) { message = String(message); } If we can get away with either of those two that would be nice. Sounds like the Firefox 28+ behavior is a no-go though. 105875 4 bzbarsky 2014-05-13 17:57:31 +0000 Right, to be clear I was comparing the Safari behavior to the Chrome/IE/old-Firefox behavior. The new-Firefox behavior is indeed reasonably JavaScripty, but seems to not be what people seem to want from this API. 105876 5 bzbarsky 2014-05-13 17:57:56 +0000 And I personally would be OK with trying to switch Gecko to the Safari behavior. 105879 6 ian 2014-05-13 18:00:32 +0000 If I define: void alert(); void alert(DOMString message); ...and then you pass the undefined value explicitly, it picks the DOMString overload? It doesn't pretend the argument was omitted? 105886 7 bzbarsky 2014-05-13 18:28:09 +0000 I believe that is the current state of WebIDL, yes, and some other APIs rely on that.... Of course if we do the Safari behavior it's a nonissue. 106095 8 ian 2014-05-15 21:02:06 +0000 Yeah, looks like Web IDL doesn't do anything to treat explicit "undefined"s as missing arguments. Ok. Do you want this for prompt() or confirm() too, or just alert()? 106174 9 bzbarsky 2014-05-16 16:18:13 +0000 It's only come up as an issue for alert(), since people use that to examine the value. For prompt() and confirm() both "" and "undefined" are pretty much pointless given how the string is supposed to be used, so I don't have an opinion as to which we should do. 106175 10 ian 2014-05-16 16:53:37 +0000 Ok. Thanks! 106176 11 contributor 2014-05-16 16:54:04 +0000 Checked in as WHATWG revision r8638. Check-in comment: for compat, apparently alert() should alert '' and alert(undefined) should alert 'undefined' http://html5.org/tools/web-apps-tracker?from=8637&to=8638 106177 12 ian 2014-05-16 16:54:15 +0000 (I used the more widely implemented behaviour; I don't mind switching to another behaviour but we should get more vendors on board if we want to do that.)