adjust choice of rust nonterminals (and their names) in macro patterns

[jesse99]

Judging from the tutorial it seems that macro fragment matching is done with names coming directly from the grammar. It doesn't seem wise to tie names in user code to the entire grammar that a particular snapshot of a particular rust compiler uses internally. Perhaps it would be better to figure out a subset of non-terminals that are sensible to use in macros and that compilers are free to map however they like internally.

This would also make it easy to provide user-facing names that aren't as aggressively abbreviated as the internal names, e.g. type(?) instead of ty.

[graydon]

No, they're not taken directly from the grammar. It matches a subset of production names, chosen by paul more or less at random based on macros he expects to see people write. This could do with further cleanup but I expect expr and stmt at least are going to be more pleasant, when writing repeatedly, than expression or statement.

[jesse99]

If it's a fixed subset then they should all be documented in the reference manual (and any, like ty, that aren't obvious need to be explained).

[paulstansifer]

It haven't shown up in the release yet, but the manual now says:

"In the matcher, $ name : designator matches the nonterminal in the
Rust syntax named by designator. Valid designators are item, block,
stmt, pat, expr, ty, ident, path, tt, matchers. The last two
are the right-hand side and the left-hand side respectively of the => in
macro rules."

[graydon]

non-critical for 0.6, de-milestoning

[graydon]

nominating for backwards compatible

(at least deciding this)

[graydon]

Closing as WONTFIX, team thinks this set of names is ok as-is.