Triple primary key 2

[maurolacy]

Follow-up of #210.

Fixes triple key Prefix type. Add Map triple keys tests.

[maurolacy]

I ran into problems when trying to use (T, U):

error[E0277]: the trait bound `(T, U): std::convert::AsRef<keys::PkOwned>` is not satisfied
  --> packages/storage-plus/src/keys.rs:83:19
   |
8  | pub trait PrimaryKey<'a>: Clone {
   |           ---------- required by a bound in this
9  |     type Prefix: Prefixer<'a>;
   |                  ------------ required by this bound in `keys::PrimaryKey`
...
83 |     type Prefix = (T, U);
   |                   ^^^^^^ the trait `std::convert::AsRef<keys::PkOwned>` is not implemented for `(T, U)`
   |
   = note: required because of the requirements on the impl of `keys::Prefixer<'a>` for `(T, U)`

Tried to implement AsRef<PkOwned> explicitly for (T, U) but started to get collissions with other implementations... notably, of Prefixer... not sure how to do it properly.
So, I "cheated" and fallback to using Pk2. If you know how to solve the previous error, I'll be glad to generalize / fix this.

[ethanfrey]

Please test and update for tripple keys that don't only use &[u8]

[ethanfrey]

Good test. This is what I want to see for the full-stack usage.

Can you update the TRIPLE to (&[u8], Int32Key, Int64Key) and verify this works just as well?

[maurolacy]

Sure.

[ethanfrey]

I fixed the prefix implemention in my last commit.
You should be able to use it in the new code.

[ethanfrey]

I need to reflect why we cannot satisfy Prefix automatically for anything that implements PrimaryKey - I assume this is a subset of the work.

[maurolacy]

I need to reflect why we cannot satisfy Prefix automatically for anything that implements PrimaryKey - I assume this is a subset of the work.

I guess the idea is to be able to range() over the last part of a given key; that is, the one that doesn't implement Prefixer.

That's OK IMO. What I don't like is that currently prefix() only works for the "full" prefix, i. e. in case of triple keys, we cannot prefix only on the first element of the triple.
What I'm now thinking is that maybe we can use (abuse?) notation like (T, (U, V)), for definining a key that is internally "flat", but can be prefixed over its first element.

[ethanfrey]

So, I "cheated" and fallback to using Pk2. If you know how to solve the previous error, I'll be glad to generalize / fix this.

Yes, this is because the Prefixer implementation for Pk2 is broken (it only works with default type &[u8] not any T, U). Time to fix that here.

[ethanfrey]

Looking better. Thank you for finding these issues

[maurolacy]

Looking better. Thank you for finding these issues

Thank you for the review, and for the fixes.