Merge #16

16: Proof-of-concept SyntaxKind as enum r=matklad a=CAD97

This was the one bit about the original RFC I was most confused about. Why isn't `SyntaxKind` a normal `enum`?

If it's to signify that it's non-exhaustive [`#[non_exhaustive]`](https://github.com/rust-lang/rust/issues/44109) should be used. (Or `#[doc(hidden)] __Nonexhaustive` on stable.)

If it's so that more variants can be added externally, why? There's no need for that, that I can foresee.

If it's to expose the `SyntaxKind` type but not any of its instances, why? This is the only actual benefit I can see of this pseudo-enum style.

This diff is meant to be as non-invasive as possible, and as such reexports all symbols as they existed prior to this. It's diffed on top of the assumed-good-to-merge #15 to avoid the conflict between them.

Diff without #15: <https://github.com/matklad/libsyntax2/pull/16/commits/4244948c6b1a62bd0e1ef276d1f0cc22c36f6f66>
Just `src/syntax_kinds.rs`: <https://github.com/matklad/libsyntax2/pull/16/commits/4244948c6b1a62bd0e1ef276d1f0cc22c36f6f66#diff-8f0d69eb4fe0148851505f787b6fd3bb>

2 files changed, 1 insertions, 3 deletions

diff --git a/src/parser/event_parser/grammar/mod.rs b/src/parser/event_parser/grammar/mod.rs
index c6ab1fbe2..82f8b7f3e 100644
--- a/src/parser/event_parser/grammar/mod.rs
+++ b/src/parser/event_parser/grammar/mod.rs
@@ -1,6 +1,5 @@
 use super::parser::{Parser, TokenSet};
 use SyntaxKind;
-use tree::EOF;
 use syntax_kinds::*;
 
 mod items;
diff --git a/src/parser/event_parser/parser.rs b/src/parser/event_parser/parser.rs
index 5ba3071cb..573e3d73a 100644
--- a/src/parser/event_parser/parser.rs
+++ b/src/parser/event_parser/parser.rs
@@ -1,8 +1,7 @@
 use {SyntaxKind, TextUnit, Token};
 use super::Event;
 use super::super::is_insignificant;
-use syntax_kinds::{ERROR, L_CURLY, R_CURLY};
-use tree::{EOF, TOMBSTONE};
+use SyntaxKind::{EOF, ERROR, L_CURLY, R_CURLY, TOMBSTONE};
 
 pub(crate) struct Marker {
     pos: u32,