Poll: Which other-crate-relative-path syntax do you prefer?

[elahn]

@petrochenkov

I’ve looked through the token list and selected tokens that can be more or less usable for other-crate-relative paths without ambiguities (including “inline” non-use paths).

Most of tokens were thrown away immediately because they can be used as binary/unary operators and they can be used freely with paths in expressions. I tried to not introduce new tokens for this, but in principle ~ can be used regardless of syntax details because it’s unused at the moment, or the backslash token (\) can be added to the language (can’t say I’m a fan though).

There are roughly three groups of possible syntaxes:

a INFIX b::c
OPEN_DELIM a CLOSE_DELIM b::c
PREFIX a::b::c

INFIX variants look universally bad, IMO, mostly because they break the path into separate parts.

a _ b::c
a::_::b::c
a crate b::c
a:\b::c // Windows drive letters!
a#b::c
a#::b::c
// etc

OPEN_DELIM a CLOSE_DELIM variants look better.

[a]::b::c
(a)::b::c
{a}::b::c
// but not <a>::b::c, it's already taken

PREFIX a::b::c (1) keywords. From all keywords only crate, extern and in looks somehow relevant.

// Less noise
extern a::b::c
crate a::b::c
in a::b::c
// More noise
extern::a::b::c
crate::a::b::c
in::a::b::c

PREFIX a::b::c (2) sigils, look kinda random.

_ a::b::c
_::a::b::c
#a::b::c
#::a::b::c
~a::b::c
~::a::b::c

Aaaand… ta-da!

@a::b::c

is not actually ambiguous!11

@ can currently be used only in patterns like this ref? mut? IDENT @ PATTERN and IDENT PATH is never a valid token sequence, so ident @a::b::c is unambiguously a pattern (ident @ a::b::c == IDENT @ PATTERN) and not ident @a::b::c == IDENT PATH, because the latter is not a valid syntax.

@petrochenkov

Basically, the three most viable syntaxes were already discussed in various threads:

// The least noisy syntax, and well fitting because `@` has "location" semantics.
@a::b::c

// Slightly more noisy syntax, and memorable because `[ ### ]` is a *crate*.
[a]::b::c

// The "no sigils" variant, good if `use extern a;` will be used roughly like
// `extern crate a;` is used now, with "inline" uses being rare. Probably too
// wordy if used more often (e.g. in every `use` import).
extern a::b::c

Note: this poll includes “inline” non-use paths

• @a::b::c
• [a]::b::c
• extern a::b::c

[phaylon]

I checked extern a::b::c, though I’d certainly prefer extern::a::b::c because of more regularity across the full path. It’s an extension of self:: and super:: to me.

So I vote for the non-listed extern::cratename::b::c.

[phaylon]

I should also note, while @foo::bar is not ambiguous, it would make foo@@bar::Baz a valid pattern.

[elahn]

Sorry, I can’t add it to the poll as > 5 mins have passed, but a moderator can…Until then, if anyone wants to vote for extern::a::b::c, please like @phaylon’s comment

[ahmedcharles]

There isn’t an option for “don’t do this, please”.