[udig-devel] Re: IGeoResource / IGeoResource clarification / discussion

[Jody Garnett <jgarnett@xxxxxxxxxxxxxxx>]

Seems we are on the same track,

Comments in line:

> > These IGeoResource methods are considered ui safe:
> > - getIdentifier()
> > - getService()
> > - equals() & hashCode()
> > - getStatus() & getStatusMessage()
> >    
>
> mmm, this isn't always posible/advisable. It is entirely posible to have
> a resource without an associated Service until it is required to be
> resolved. That said ... a placeholder could be created and returned.
>  
I need do need a parent for user interface land, suggestions? You could 
return null and use events when you have a parent (nasty for the poor 
user interface).
Also see the comment about IResource and Phantom, we could have a Type 
of Phantom and the ui could know not to look for a parent...

> > This IGeoResourceInfo method:
> > - getType()
> >
> > And these are considered blocking:
> > - getInfo()
> Can force a load when the info is loaded ... most cases would do this
> anyways.
>  
That would be the "blocking" part :-)

> > Recomendation/Discussion:
> > - make getInfo() throw an IOException (so it is *really* obvious I/O is 
> >
> >    
> happeningSure, this would change the current assumption though that the info is
> loaded at this time.
>  
Not sure I understand - I would think an IOException would enforce the 
idea that Info is loaded. Although I guess an IO problem
should result in a phantom with enough information in getStatus to 
communicate the error.

> > - move getType() to IGeoResource, we will need to remove the constant 
> > WRITE as it the only ability we would be "know" simply from the identifier.
> >    
> The move is fine, but unsure as to the WRITE constant.
>  
Don't we have to check the service to figure out if something is writable?

> > Consider/Discussion:
> > You may also want to look at the IResource link bellow:
> > - it has a getWorkspace() indicating that a getCatalog() method may be 
> > useful (allow client code the ability to know where to listen for event 
> > notification).
> >    
> Yes and No, I currently magically move the IService around ...
>  
Still as client I *need* to know where to get my notification from, 
besides I think the replace event can be used to communicate the Catalog 
Move.
You are replacing the IGeoResource with another in a different catalog 
after all.

> > - getName() defined to be last section of the Path, we can define 
> > getName() to be the name of the getIdentifierURL()
> >    
> in most cases it's the reference of the url.
>  
Right now I have the following code in the LabelItemProvider:

> IGeoResource resource = layer.resource();
>         
>         label = resource.getIdentifier().getPath();
>         if( label != null || label.length() == 0 ){
>             int split = label.indexOf('/');
>             if( split != -1 ){
>                 label = label.substring( label.lastIndexOf('/') );
>             }
>             return label;
>         }

Alternate proposal: Make a Services utility class with:
- getName( IService )
- getDefaultIcon( IService )
- getName( IGeoResource )
- getDefaultIcon( IGeoResource )

This would prevent everyone from making up their own recipe for 
identification to label (when title is not available of course).

> > - contains the idea of Phantom resources (ie pending), a similar idea 
> >    
It may be nice to have the Phantom type, just so we can indicate that in 
the user interface. See above comment about
replace event for moving a Phantom into a the local Catalog for actual 
data access..

> > may be useful for search results that we are not connected to yet.
> > - has both session properties, and persistent properties (w/ 2k limit)
> >    
> Dont like this ... the represent the service/resource ... don't want
> then to hold people's state too.
>  
Not sure I like it either, but I suspect people will ask for it. We can 
wait until they do. Most of them will be content to play with
Layer (which already has a properties Blackboard for collaboration)..