Awesome, I like #1 too. I was driving to work this morning and thinking about it: #2 actually encourages bad behavior, because it'll seem it's OK to fiddle with the elements between calling bind and attaching, and it's really not. We _could_ make an effort to make it work, but it's much better to make the flow clearer this way: if you're using lazy widgets, your elements have to be lazy too.
On Wed, Aug 17, 2011 at 9:41 AM, Ray Ryan <[email protected]> wrote: > I like #1 too. I think we should try to narrow the visibility of > PotentialElement as much as we can. > > So #1 means two things , right? > > • Widgets are seated in their @UiFields immediately > • In an IsRenderable owner, Element and subclasses are only available via > LazyDomElement, and @UiField Element is a compile time error > > I've tweaked the test a bit (will update soon), and I'm happy to report > that composites around non-IsRenderables work as expected, with element > fields filled immediately. Given that I don't think we need to delay the > switch to using lazy widget builder by default. > > > On Wed Aug 17 06:14:52 GMT-700 2011, Hermes Freitas wrote: > >> WidgetInterpreter and WidgetPlaceholderInterpreter shouldn't output >> LazyDomElement. Rafa, do you remember why? I don't think this aggregates any >> performance gain for us, am I missing something? >> >> And I vote for #1 >> >> On Tue, Aug 16, 2011 at 10:10 PM, Rafael Castro >> <[email protected]<http://www.google.com/url?sa=D&q=mailto%3Ardcastro%40google.com> >> > wrote: >> >> +hermes >> >> Good point, this is really tricky. The problem here is that we don't >> actually have the DOM element until the widget is attached. I see 2 options: >> 1-) We force the UiField to be a LazyDomElement, so this is explicit. >> 2-) We use PotentialElement with a resolver that throws an Exception >> (i.e., it's only really resolved when it's attached). >> >> what do you think? >> >> ps.: really nice tests, thanks for putting them together! >> >> On Tue, Aug 16, 2011 at 5:13 PM, >> <[email protected]<http://www.google.com/url?sa=D&q=mailto%3Arjrjr%40google.com> >> > wrote: >> >> On 2011/08/17 00:12:24, rjrjr wrote: >> >> Ready for review. >> >> Rafa, this turned up one issue that concerns me: most @UiField fields >> are not filled in until the widget is attached to the dom, but we're not >> consistent about it. See the big comment in testDeep. >> >> >> http://gwt-code-reviews.**appspot.com/1527804/<http://www.google.com/url?sa=D&q=http%3A%2F%2Fgwt-code-reviews.appspot.com%2F1527804%2F> >> >> >> >> >> >> -- >> --Hermes Freitas >> > -- http://groups.google.com/group/Google-Web-Toolkit-Contributors
