RFC: Derived Contexts

[probeiuscorp]

View Rendered Text

Selectors for Context are an in-demand feature, but with a slightly different interface, can be made much more universally useful (.slice as proposed here is similar).

Following from functional programming techniques, this RFC proposes adding:

• a ReadonlyContext<T> type which is a supertype of Context<T> without .Provider

and these methods to Context:

• .map, of type: <T, U>(this: ReadonlyContext<T>, fn: (data: T) => U) => ReadonlyContext<U>
• .apply, of type: <T, U>(this: ReadonlyContext<T>, fn: ReadonlyContext<(data: T) => U>) => ReadonlyContext<U> (I actually recommend one of its alternatives)

.map allows users to refine and select their Contexts, while .apply allows users to combine multiple Contexts together. This is covered in detail in Motivation.

Every call, without respect to the identity of the given argument (function / context of functions), returns a distinct Context. To be usable within renders, existing user-land techniques like WeakMap can be used to stabilize the identity of the returned Context to the identity of the given argument.

---

RFC 119 useContextSelector draws from much of the same motivation.

[probeiuscorp]

@gnoff as you proposed #119, if you had any thoughts on this approach they'd be incredibly interesting. My proposal is essentially an extension to dantman's .slice, and your comment on useContexts is very similar to my proposed sequence alternative to .with. It's very interesting that seemingly you both came up with applicative functors. I proposed this because it draws from established, well-known patterns and theory. Also because it eliminates worrying about whether the selector is stable or not.

I would love to see progress towards a solution to this problem, but I don't know if you or the community in general has moved on and is no longer interested in this direction.