Hi Andrew,
thanks for your thoughts. One of the main problems here that I didn't put
on my email is that you loose the opportunity to ask if the components is a
"StyledUIBase", since we are "bifurcating" the code when implement in Group
and DataContainerBase and others the interface IClassSelectorLitsSupport.
You can ask for "IClassSelectorLitsSupport", but you probably would want
just ask for StyledUIBase.
El mar., 30 abr. 2019 a las 8:14, Frost, Andrew (<andrew.fr...@harman.com>)
escribió:
> Hi Carlos
>
> Isn't this similar to the challenge that Adobe had with EventDispatcher?
> which I think is why there are two possibilities:
> a) derive from EventDispatcher
> b) implement IEventDispatcher
>
> In the second case, you have to create an instance of EventDispatcher
> within your class, and then you have to implement the IEventDispatcher
> functions with calls to the EventDispatcher object. So it's not
> zero-effort, but it's not too bad (and probably the best option you can get
> without multiple inheritance..). Yes there's repeated code, but it's
> limited to one line per function so is relatively easy to copy/paste.
>
This solution could save code duplication if method bodies are long. In
this case method bodies are similar, so I'm afraid will not apply here to
make a significant difference
>
> Not 100% sure if the same approach can be applied in the scenario you're
> talking about, but that might help perhaps..? So the functionality you want
> to repeat would go into a class ("ClassSelectorListSupportImpl"?) and you
> create one of these in each of the classes that derive from
> UIBase/Group/etc and implement IClassSelectorLitsSupport.
>
Yes, this could be similar to before and more elegant, but the problem is
the same I put before.
>
> Some alternatives perhaps:
> - I'm also wondering whether you could create a helper function that
> actually adjusted the class prototype and programmatically added the
> necessary functions to your classes... ?
> - the repeated functionality could be put into other .as files and just
> 'included' in the classes? I know, this is a major hack, but I've seen it
> done :-(
>
>
hehe, prefer not to go that way ;-)
What I want to expose here is that beads are a good composition way and
love it. But we still has a problem in cases like this.
Thinking on this and seeing how the overall picture is now, I think the
actual solution is not helping PAYG at all. Maybe we should see if we can
add IClassSelectorLitsSupport to UIBase and remove
IClassSelectorLitsSupport implementations (StyledUIBase and the rest).
We'll be saving code, and Basic components will have access to that API
that is tiny and seems finaly make sense to be in UIBase.
At least I see other frameworks out there have it's own "addClass",
"removeClass", etc... as part of it's core, since nowadays the use of CSS
selectors is crucial and people will needed always (and that's the key for
PAYG).
Thoughts?
>
> Hope it helps..
>
> thanks
>
> Andrew
>
>
>
> -----Original Message-----
> From: Carlos Rovira [mailto:carlosrov...@apache.org]
> Sent: 29 April 2019 18:04
> To: dev@royale.apache.org
> Subject: [EXTERNAL] extending IUIBase with interfaces (was: Re: Version
> property (was: Let's bump Royale version to 1.0))
>
> Hi Alex,
>
> initial problem is: I want to add to all jewel components (that use to
> have UIBase as common class) a new interface and implement the methods. In
> concrete interface is: IClassSelectorListSupport
>
> So the expected result should be to make all Jewel components implement
> IClassSelectorListSupport, but to avoid repeat code in all classes we need
> a root for all the components.
>
> Perfect point would be UIBase, but we can't add at that point for PAYG
> reasons.
>
> Current solution in Jewel is to create StyledUIBase that is
>
> public class StyledUIBase extends UIBase implements
> IClassSelectorListSupport
>
> but this makes other classes like jewel Group that extend html Group have
> UIBase in the hierarchical chain instead StyledUIBase, so we need to make
> Group as this:
>
> public class Group extends org.apache.royale.html.Group implements
> IClassSelectorListSupport
>
> same for
>
> public class DataContainerBase extends
> org.apache.royale.core.DataContainerBase implements
> IClassSelectorListSupport
>
> This make we have 3 classes that are duplicating the same exact code
> implementing in the same maner addClass, removeClass, etc..
>
> what can we do for this cases?
>
> Hope is more clear now,
>
> thanks
>
>
>
> El lun., 29 abr. 2019 a las 17:41, Alex Harui (<aha...@adobe.com.invalid>)
> escribió:
>
> >
> > @Carlos, I couldn't understand your scenario. Feel free to start a
> > separate thread on it. The key aspect, as a guess, is trying to avoid
> > code that assumes that an instance is a subclass of a particular base
> > class instead of an implementation of an interface.
> >
> > HTH,
> > -Alex
> >
> > On 4/29/19, 7:29 AM, "Carlos Rovira" <carlosrov...@apache.org> wrote:
> >
> > @Mark, if you use Maven, then artifacts in the poms are all what
> > you need,
> > so maven takes care of pulling all right dependencies you need.
> >
> > @Alex, about extending IUIBase, my own experience is that beads
> > are a very
> > good way to extend/compose more code you need, but extending core
> > interfaces like the one you said is not the case. Let me put you an
> > example: For Jewel I use "StyledUIBase" that extends "UIBase" to add
> > IClassSelectorListSupport, the css CRUD API (addClass,
> > removeClass,
> > etc..):
> >
> > public class StyledUIBase extends UIBase implements
> > IClassSelectorListSupport
> >
> > This makes that some components are easy to extend but others need
> > to be
> > recreate. There are lots of cases in Jewel where I need to
> >
> > If you search in Jewel for implements IClassSelectorListSupport
> > you'll find Jewel Group, Jewel DataContainerBase and Jewel Table,
> are
> > implementing that class while, so they are not StyledUIBase in
> > it's core
> > are standard UIBase and at its level we implement the interface and
> add
> > again all methods repeating the code. So we are duplicating that
> > code all
> > that times.
> >
> > I think this is one of the few things I don't like in Jewel, If
> > you know
> > some way to do this in a better way, I can change it
> >
> > thanks
> >
> > --
> Carlos Rovira
>
> https://clicktime.symantec.com/37RX7Zfo3fgsrnBKZDX18bM7Vc?u=http%3A%2F%2Fabout.me%2Fcarlosrovira
>
--
Carlos Rovira
http://about.me/carlosrovira
