map on homogenous tuples

[bend-n]

Proposal

Problem statement

It is impossible to map homogenous tuples as if they were arrays.

Motivating examples or use cases

Often times you see this kind of code: (f(x.0), f(x.1)) (see search).

Solution sketch

impl<T> (T, T…) {
    pub fn map<F, U>(self, f: F) -> (U, U…)
    where
        F: FnMut(T) -> U;
}

Alternatives

One may:

• simply use an array
• <[_; _]>::from(x).map(f).into()

Links and related work

rust-lang/rust#160640

What happens now?

This issue contains an API change proposal (or ACP) and is part of the libs-api team feature lifecycle. Once this issue is filed, the libs-api team will review open proposals as capability becomes available. Current response times do not have a clear estimate, but may be up to several months.

Possible responses

The libs team may respond in various different ways. First, the team will consider the problem (this doesn't require any concrete solution or alternatives to have been proposed):

• We think this problem seems worth solving, and the standard library might be the right place to solve it.
• We think that this probably doesn't belong in the standard library.

Second, if there's a concrete solution:

• We think this specific solution looks roughly right, approved, you or someone else should implement this. (Further review will still happen on the subsequent implementation PR.)
• We're not sure this is the right solution, and the alternatives or other materials don't give us enough information to be sure about that. Here are some questions we have that aren't answered, or rough ideas about alternatives we'd want to see discussed.

[kennytm]

note that some of the search results aren't homogeneous mappings:

// from rusqlite:

    fn next(&mut self) -> Option<(IndexConstraint<'a>, IndexConstraintUsage<'a>)> {
        self.iter
            .next()
            .map(|raw| (IndexConstraint(raw.0), IndexConstraintUsage(raw.1)))
//                                                             ^~~~~ note the returned types are different
    }

// from ssb-handshake:

    let (eph_pk, eph_sk) = (ClientEphPublicKey(eph_kp.0), ClientEphSecretKey(eph_kp.1));
//                                   ^~~~~~                        ^~~~~~
    let server_pk = ServerPublicKey(*server_pk);

// from BenoitZugmeyer/RustyAdventOfCode:

                Some((keypad1_char(state.0), keypad2_char(state.1)))
//                          ^                      ^

[scottmcm]

I think that "go through an array if you want this" is probably the best answer. (And maybe just use arrays instead of tuples in the first place -- like using array_chunks instead of Itertools::tuples.)

We should save map on tuples for when we get generic closures, IMHO.

EDIT later to expand: For example, my_tuple.map(Some) should "obviously" do (T, U) -> (Option<T>, Option<U>), so we shouldn't use map for anything that would conflict with doing that later.

[scottmcm]

I suppose another alternative would be to add inherent methods to make the array ↔ tuple conversions easier to type.

I agree that <[_; _]>::from(x).map(f) is kinda ugly, but x.into_array().map(f).into_tuple() might be fine?

[bend-n]

I think that would be a good idea, separate from map on homogenous tuples. I think it doesn't quite replace the want for this, but it does ease the pain.

[scottmcm]

I think it'd also help to have not just a link to a search result but an analysis of those matches.

What are people using tuples for like this? How new is the code? One possible answer here is "well yes, before array patterns existed of course they used tuples, but now they should just use arrays" and to do nothing.

[bend-n]

Most of these cases are not because people need to use a tuple, but rather because some other method (not necessarily their own) that they use produces or requires a tuple, I've found from looking through a few of them. People seem to generally dislike doing fn x(y: [u32; 2]) and prefer y: (u32, u32), which causes this.

[ChayimFriedman2]

So, should we document somewhere visible that you should prefer [T; N] over (T, T, ...)?

[bend-n]

There's no tuple.0 notation for arrays though, people have to use tuple[0], which they tend to be a little averse to.

[cuviper]

Even if we don't prescribe a preference, we can point out such advantages of arrays.

[kpreid]

There's no tuple.0 notation for arrays though, people have to use tuple[0], which they tend to be a little averse to.

I think it would be good to add that notation, not merely for more concise/“pretty”/“clean” code, but to make it syntactically obvious that this index cannot overflow at run time.

[bend-n]

That's a cool idea, would have to be a whole-ass RFC though, right?