> On 2011-08-16 17:24:01, Henry Saputra wrote: > > http://svn.apache.org/repos/asf/shindig/trunk/features/src/main/javascript/features/core.util.onload/onload.js, > > line 47 > > <https://reviews.apache.org/r/1525/diff/1/?file=32902#file32902line47> > > > > Why does the definition registerOnLoadHandler here? > > Dan Dumont wrote: > It is redefined to immediately execute any onload callbacks. > > The sequence of events is: > 1) Gadget starts loading and registers some onload. > 2) Gadget fires an async operation that takes an unknown amount of time > 3) Open Social api is ready and runOnLoadHandlers is called. > 4) Gadget's async operation is done. To be safe it uses > registerOnLoadHandler to make sure everything is all set (in case the async > operation beat the runOnLoadHandlers call) > > At this point, before the change, the call to registerOnLoadHandler would > never result in the callback being executed. > This change makes it so once runOnLoadHandlers is called, > registerOnLoadHandler executes the callback immediately. > > Henry Saputra wrote: > In step 2, the async operation is called outside of the function > registered with registerOnLoadHandler? > > So in step 4, why would the gadget's async handler call > registerOnLoadHandler? > > Dan Dumont wrote: > This is important with some of the changes in dojo 1.7 where the > bootstrap may be very small and included in the gadget page but > asynchronously load more code in. When the application is ready, it should > be able to rely on registerOnLoadHandler to provide them a means of executing > gadget code safely. > > Really, the gadget should be able to do any non-gadget-api async > operations while the page is loading. To sync everything back up, the gadget > should be able to rely on registerOnLoadHandler executing their callbacks > that rely on gadget-api operations so that we don't introduce race conditions.
+1 for this change. I think this is pretty much how all of the "ready" functions I've seen in other JS libraries work (dojo, jquery, ...) -- in any of those libraries registering code after the trigger has already fired results in immediate execution, so I think this is probably something people would expect. - Jesse ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/1525/#review1479 ----------------------------------------------------------- On 2011-08-17 15:18:59, Dan Dumont wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/1525/ > ----------------------------------------------------------- > > (Updated 2011-08-17 15:18:59) > > > Review request for shindig. > > > Summary > ------- > > If gadgets.util.registerOnLoadHandler was called after the runOnLoad > counterpart was called, the callback would sit in the array... forever alone. > > > This addresses bug SHINDIG-1578. > https://issues.apache.org/jira/browse/SHINDIG-1578 > > > Diffs > ----- > > > http://svn.apache.org/repos/asf/shindig/trunk/features/src/main/javascript/features/core.util.onload/onload.js > 1157211 > > Diff: https://reviews.apache.org/r/1525/diff > > > Testing > ------- > > > Thanks, > > Dan > >
