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-
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
INFIX variants look universally bad, IMO, mostly because they break the path into separate parts.
a _ b::c
a crate b::c
a:\b::c // Windows drive letters!
OPEN_DELIM a CLOSE_DELIM variants look better.
// but not <a>::b::c, it's already taken
PREFIX a::b::c (1) keywords.
From all keywords only
in looks somehow relevant.
// Less noise
// More noise
PREFIX a::b::c (2) sigils, look kinda random.
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,
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.