Consider making pointer boxing a higher priority

[Kimundi]

Right now, ~[1,2,3].to_str() is the same as ~([1,2,3].to_str()), which is confusing because in the case of of strings and vectors, the sigil is basically part of the literal.

If pointer boxing gets at least a higher priority than method application, ~[1,2,3].to_str() would mean (~[1,2,3]).to_str(), and you'd have to add explicit delimiters to box the whole method chain, like so: ~([1,2,3].to_str())

Which, I think, would be more intuitive: A pointer sigil puts the thing right next to it in a box, if you want to box the result of an complex operation, but it in brackets.

[catamorphism]

Seems non-critical for 0.7. Nominating for milestone #1.

[graydon]

rejected as wontfix; it's not great but unary operators are low-precedence by design

[nikomatsakis]

I prefer to keep all unary operators with the same precedence, and I prefer to use C's precedence rules, so that &a.b.c continues to work.