effective context pre-processing

[marc-portier]

Over time the various iterations of the specification (specially since the 0.4) has been pushing out some of the originally (0.4 draft) ambitions out of the spec into what I would call "context-pre-processing".

We could argue that this makes the URI-templates less usefull, but on the positive side the spec becomes more readable, and less open to intrepretation and discussion. The good news is that outside the spec we could ensure end user effectiveness still by providing smart and helpfull context-pre-processing features.

Things coming to mind:

• Introducing defaults for various fields
• Expanding lists to maps of indexed-names: so list: ['a','b','c'] becomes list: {'list.1': 'a', 'list.2': 'b', 'list.3': 'c'}
• Similar expansion for the hash-tables so the original name gets associated.
• producing json serilaized forms for deeper structures (with some way to negotiate between these two)

Maybe some builder-syntax can be used to express the final context from the original one:

var finalContext = ContextBuilder(originalContext).defaultsTo({...}).expand('list',...).jsonify('deepstruct',...).get();

We already have some built-in context-wrapper for function evaluation, maybe this concept can be extended to offer these new features.

See also my expressed opinion on w3c-uri-mailinglist