Add to definition of `P<T>` type to Appendix D: Code Index

[alexreg]

And why it's needed.

[alexreg]

@eddyb's definition of it is an "immutable Box<T>". I suppose it's needed just due to the nature of tree data structures?

[eddyb]

It's not "needed", not to a large extent. The actual code has (potentially-outdated) documentation.

[alexreg]

@eddyb Well, heap allocation is surely needed, otherwise we'd have to deal with an immutable tree, and that would be very expensive during construction and later modification.

[eddyb]

I mean as a separate concept from Box or anything else.

[mark-i-m]

Sorry, I'm not sure I understand. We use P when we need an owned immutable pointer. We can't use &T because that is not owned and we can't use Box because it is not immutable. Is that correct?

[alexreg]

@eddyb Right, fair.

@mark-i-m More or less, yes... @eddyb is saying that P<> doesn't really provide anything new over Box<> though – it's just a new-type with a immutability guarantee.