There are 2 alternatives to exposing setters. (1) use setAccessible() - not very nice, but it's pretty much what serialisation does. (2) use bytecode enhancement - not very easy in Pivot's context. I'm not sure what technique the EJB @Resource annotations use, but they don't need to expose setters. But then an EJB container has more control over the application class lifecycle than Pivot does.
But also, Pivot code could be make a lot neater by simply wrapping WTKXSerializer in a helper class like this: PivotLoader loader = .. window = loader.window(...) feedListView = loader.listView(...); final CardPane cardPane = loader.cardPane(...); final Label statusLabel = loader.label(...); Sure, loader will need quite a few helper methods, but it would make the majority of UI code neater without any extra magic being required. Regards, Noel Grandin On Sat, May 2, 2009 at 15:06, Dominique de Vito <[email protected]> wrote: > > Hi, > > I re-post here some ideas about annotations with Pivot; I have already > written about those ideas in > http://java.dzone.com/news/building-simple-rss-client - a RSS example. > > Greg Brown mentionned in his RSS example the startup() method including the > following instructions: > > WTKXSerializer wtkxSerializer = .. > window = ... > feedListView = (ListView) wtkxSerializer.getObjectByName(...); > final CardPane cardPane = (CardPane) wtkxSerializer.getObjectByName(...); > final Label statusLabel = (Label) wtkxSerializer.getObjectByName(...); > > Here, 'window' and 'feedListView' are class fields. And 'cardPane' and > 'statusLabel' are method-scoped variables. These fields and variables are > assignated explicitly while fetching the corresponding XML element's value. > > While reading these instructions, I have first thought about an implicit > binding, that is, about something like (with appropriate setter definitions) : > > private WTKXSerializer wtkxSerializer; > > @WtkxResource("rss_feed_demo.wtkx") > private Window window = null; > > @Bind("feedListView") > private ListView feedListView; > > @Bind("cardPane") > private CardPane cardPane; > > @Bind("statusLabel") > private Label statusLabel; > > Here, all these Java elements are Java fields. > > I thought, if the 'Application' class lifecycle/method chain call enables it, > annotations could be used to bind XML elements/Java fields, before the call > of the startup() method. Then, according to annotations, all the assignments > above could be done automatically, following annotation directives. > > It's a first sketch. I am not fully happy with it. > > I just feel annoyed needing to expose setter methods for all possible fields > to be assigned, while I just need them only here for initialization! So, I > don't feel fully happy with the way constructors/DI works today. > > So, nowadays, I am looking deeper into DI internal details in order to see > how to mix annotation+DI without exposing too much about the object internals > (I just feel property-DI break unfortunately OO encapsulation). > > Another interesting topic is: does Pivot need a specific binding annotation > set, or is it possible re-use the annotation set of some library (like JiBX > or Castor)? > > Any comment/idea about Pivot with annotations ? > > Thanks. > > Regards, > Dominique > http://www.jroller.com/dmdevito > > PS : for those interested, I have mentionned Pivot into 2 of my posts. > > Swing browsers - other interesting DSL Swing projects to follow > http://www.jroller.com/dmdevito/entry/swing_browsers_other_interesting_dsl > > HotJava may come back, due to existing, or coming, components, like JWebPane > http://www.jroller.com/dmdevito/entry/hotjava_may_come_back_due > > > > > >
