Add 'NFDataX.ensureSpine`

[martijnbastiaan]

Wanted: a new name for spinify! Feel free to comment below ensureSpine it is!

---

Consider the following example:

>>> x = undefined :: Vec 3 Char
>>> y = map (\_ -> 3) x
>>> y
<error>

Currently, this would fail, as map tries to pattern match on the constructor of vector Cons. This constructor doesn't exist however, and resolves to undefined. In this specific instance we could use lazyV to create the "spine" of Vec for us:

>>> z = map (\_ -> 3) (lazyV x)
>>> z
<3,3,3>

spinify generalizes this behavior to all data structures having an instance for NFDataX.

---

Fixes #748

[christiaanb]

Some other suggestions (I "hate" the name spinify):

• defineSpine
• forceSpice
• deepseqSpine
• lazyLeaf
• @leonschoorl says: lazy (generalization of lazyV) <- but it's not lazy for sum-types

[gergoerdi]

ensureSpine? ensureShape?

[basile-henry]

deepDefine

[rowanG077]

I like ensureSpine the best.

[masterdezign]

proto?

[martijnbastiaan]

Thanks everyone :). Seems like ensureSpine is the most popular. Leave a comment if you strongly disagree! If no one does, I'll change the name and merge this PR next week.

[martijnbastiaan]

@christiaanb This is ready for review. I don't believe it's possible to deprecate "custom" implementations of type class functions (in this case deepErrorX), is it?

[christiaanb]

Please keep the method implementations of deepErrorX; it probably saves on compile time, and maybe even run-time efficiency during simulation.

[martijnbastiaan]

My original idea was to deprecate "custom" implementations of deepErrorX, as deepErrorX ~ ensureSpine . errorX. As we discussed online, that's not really possible, so I'll revert the deletion of custom implementations.

[martijnbastiaan]

@christiaanb I've readded withFrozenStack too. I'll merge it if it passed the tests!