Add some useful interfaces to the core and std libs

[marijnh]

To find out whether they work well, we should try to actually use them. to_str and seq are some obvious candidates.

If #1438 is resolved, we can provide a core::impl::common module that exports impls for common interfaces of built-in types, so you'd only need a single import impl::common::* to get access to [1].len() and 10.to_str().

[kud1ing]

Would it be possible to do this re-exporting in libcore/core.rs?

[marijnh]

I think I'd rather not put a bunch of non-optional interfaces in every scope.

[kud1ing]

Fair enough. I don't find it easy to map Haskell concepts to Rust all the time.

May i suggest that we consider that the interfaces are adjectives where possible?

[marijnh]

Adjectives? Don't nouns (seq) or verbs (show) make more sense? Can you give some examples?

[kud1ing]

Haskell is using adjectives for newer type classes, e.g. "IsString", "Foldable", "Traversable", "Typeable", "Storable"

I think TypeState predicates in Rust are mostly adjectives like "is_not_empty" etc.
Type classes are a similar requirement, but on the type.

My suggestion would be:

• "to_str" => "is_showable"
• "seq" => "is_iterable"

It's certainly a matter of taste:

fn foo<T: is_showable>(v: T) -> str
fn any<T: is_iterable>(v: T, f: block(U) -> bool) -> bool

which to me would be as nice as

 fn foo<T>(v: [T]) : is_not_empty(v) -> T

Why do nouns make more sense in your opinion?

[boggle]

I'd reserve adjectives for predicates and find some other naming scheme for interfaces. In any case, please do not use "is_", we just agreed to use that for predicates. I like marijns suggestion to try and use verbs for interfaces ("t: shows, t: iterates")

[graydon]

I agree interfaces are at least as likely verb-y as noun-y. But the -able and even -s suffixes seem noisy. The syntactic context makes it quite clear we're talking about an interface.

fn foo<T: fold iterate write>(...) 

looks ok to me.

[kud1ing]

I'm fine with verbs. I've update the yet sketchy Core styleguide

[kud1ing]

Was Stop passing spans to middle::trans functions that don't need them really meant to close Add some useful interfaces to the core and std libs?

[marijnh]

No! I got the bug number wrong.