Styles and Classes on Widgets about widget-core

[novemberborn]

I wonder if it should be possible to provide state that isn't actually stored anywhere?

Turns out it is, yay!

[kitsonk]

Refs #45 in that what we need to do here will essentially effect that as well as #35 to a minor degree.

[novemberborn]

Following up from dojo/app#20 (via dojo/widgets#30), should it be possible to specify classes at instantiation time without specifying them in the state?

I suppose these too could be considered "static" and added after getting the widget classes. Then any dynamic stuff needs to come from the state.

[agubler]

classes and styles have been removed as options for widgets and the only way to specify classes outside of statically in the widget is via state, so in answer I don't think so.

[novemberborn]

@agubler that makes dojo/app#20 really hard to implement. I wonder if it should be possible to provide state that isn't actually stored anywhere? Otherwise App will have to create IDs for all widgets it instantiates, and might just have to require a default widget store.

I'm actually bringing this up since I'm looking to simplify the dataviz behavior around this kind of config. Sounds like the idea is that you can specify defaults on the class level, and allow customization via state. So I'll try to make that work.

[kitsonk]

Turns out it is, yay!

The only caveat/edge case that I don't think we have handled is that if .observeState() after creation, the "local" state will be patched with the observed state, which might result in some surprises. We could change it that the first thing it does when observing state is that it takes any local state and patches that before it starts observing. I suspect that is all a bit of an edge case though.

[novemberborn]

The only caveat/edge case that I don't think we have handled is that if .observeState() after creation, the "local" state will be patched with the observed state, which might result in some surprises. We could change it that the first thing it does when observing state is that it takes any local state and patches that before it starts observing. I suspect that is all a bit of an edge case though.

Would be good to avoid surprises like this.