blob: 7ae1e5f82e56330f1f5dcb827e74c612ebae1db0 [file] [log] [blame]
//! This is the actual "grammar" of the Rust language.
//!
//! Each function in this module and its children corresponds
//! to a production of the formal grammar. Submodules roughly
//! correspond to different *areas* of the grammar. By convention,
//! each submodule starts with `use super::*` import and exports
//! "public" productions via `pub(super)`.
//!
//! See docs for [`Parser`](super::parser::Parser) to learn about API,
//! available to the grammar, and see docs for [`Event`](super::event::Event)
//! to learn how this actually manages to produce parse trees.
//!
//! Code in this module also contains inline tests, which start with
//! `// test name-of-the-test` comment and look like this:
//!
//! ```text
//! // test function_with_zero_parameters
//! // fn foo() {}
//! ```
//!
//! After adding a new inline-test, run `cargo test -p xtask` to
//! extract it as a standalone text-fixture into
//! `crates/syntax/test_data/parser/`, and run `cargo test` once to
//! create the "gold" value.
//!
//! Coding convention: rules like `where_clause` always produce either a
//! node or an error, rules like `opt_where_clause` may produce nothing.
//! Non-opt rules typically start with `assert!(p.at(FIRST_TOKEN))`, the
//! caller is responsible for branching on the first token.
mod attributes;
mod expressions;
mod generic_args;
mod generic_params;
mod items;
mod params;
mod paths;
mod patterns;
mod types;
use crate::{
parser::{CompletedMarker, Marker, Parser},
SyntaxKind::{self, *},
TokenSet, T,
};
pub(crate) mod entry {
use super::*;
pub(crate) mod prefix {
use super::*;
pub(crate) fn vis(p: &mut Parser<'_>) {
opt_visibility(p, false);
}
pub(crate) fn block(p: &mut Parser<'_>) {
expressions::block_expr(p);
}
pub(crate) fn stmt(p: &mut Parser<'_>) {
expressions::stmt(p, expressions::Semicolon::Forbidden);
}
pub(crate) fn pat(p: &mut Parser<'_>) {
patterns::pattern_single(p);
}
pub(crate) fn pat_top(p: &mut Parser<'_>) {
patterns::pattern_top(p);
}
pub(crate) fn ty(p: &mut Parser<'_>) {
types::type_(p);
}
pub(crate) fn expr(p: &mut Parser<'_>) {
expressions::expr(p);
}
pub(crate) fn path(p: &mut Parser<'_>) {
paths::type_path(p);
}
pub(crate) fn item(p: &mut Parser<'_>) {
items::item_or_macro(p, true);
}
// Parse a meta item , which excluded [], e.g : #[ MetaItem ]
pub(crate) fn meta_item(p: &mut Parser<'_>) {
attributes::meta(p);
}
}
pub(crate) mod top {
use super::*;
pub(crate) fn source_file(p: &mut Parser<'_>) {
let m = p.start();
p.eat(SHEBANG);
items::mod_contents(p, false);
m.complete(p, SOURCE_FILE);
}
pub(crate) fn macro_stmts(p: &mut Parser<'_>) {
let m = p.start();
while !p.at(EOF) {
expressions::stmt(p, expressions::Semicolon::Optional);
}
m.complete(p, MACRO_STMTS);
}
pub(crate) fn macro_items(p: &mut Parser<'_>) {
let m = p.start();
items::mod_contents(p, false);
m.complete(p, MACRO_ITEMS);
}
pub(crate) fn pattern(p: &mut Parser<'_>) {
let m = p.start();
patterns::pattern_top(p);
if p.at(EOF) {
m.abandon(p);
return;
}
while !p.at(EOF) {
p.bump_any();
}
m.complete(p, ERROR);
}
pub(crate) fn type_(p: &mut Parser<'_>) {
let m = p.start();
types::type_(p);
if p.at(EOF) {
m.abandon(p);
return;
}
while !p.at(EOF) {
p.bump_any();
}
m.complete(p, ERROR);
}
pub(crate) fn expr(p: &mut Parser<'_>) {
let m = p.start();
expressions::expr(p);
if p.at(EOF) {
m.abandon(p);
return;
}
while !p.at(EOF) {
p.bump_any();
}
m.complete(p, ERROR);
}
pub(crate) fn meta_item(p: &mut Parser<'_>) {
let m = p.start();
attributes::meta(p);
if p.at(EOF) {
m.abandon(p);
return;
}
while !p.at(EOF) {
p.bump_any();
}
m.complete(p, ERROR);
}
}
}
pub(crate) fn reparser(
node: SyntaxKind,
first_child: Option<SyntaxKind>,
parent: Option<SyntaxKind>,
) -> Option<fn(&mut Parser<'_>)> {
let res = match node {
BLOCK_EXPR => expressions::block_expr,
RECORD_FIELD_LIST => items::record_field_list,
RECORD_EXPR_FIELD_LIST => items::record_expr_field_list,
VARIANT_LIST => items::variant_list,
MATCH_ARM_LIST => items::match_arm_list,
USE_TREE_LIST => items::use_tree_list,
EXTERN_ITEM_LIST => items::extern_item_list,
TOKEN_TREE if first_child? == T!['{'] => items::token_tree,
ASSOC_ITEM_LIST => match parent? {
IMPL | TRAIT => items::assoc_item_list,
_ => return None,
},
ITEM_LIST => items::item_list,
_ => return None,
};
Some(res)
}
#[derive(Clone, Copy, PartialEq, Eq)]
enum BlockLike {
Block,
NotBlock,
}
impl BlockLike {
fn is_block(self) -> bool {
self == BlockLike::Block
}
fn is_blocklike(kind: SyntaxKind) -> bool {
matches!(kind, BLOCK_EXPR | IF_EXPR | WHILE_EXPR | FOR_EXPR | LOOP_EXPR | MATCH_EXPR)
}
}
const VISIBILITY_FIRST: TokenSet = TokenSet::new(&[T![pub]]);
fn opt_visibility(p: &mut Parser<'_>, in_tuple_field: bool) -> bool {
if !p.at(T![pub]) {
return false;
}
let m = p.start();
p.bump(T![pub]);
if p.at(T!['(']) {
match p.nth(1) {
// test crate_visibility
// pub(crate) struct S;
// pub(self) struct S;
// pub(super) struct S;
// test_err crate_visibility_empty_recover
// pub() struct S;
// test pub_parens_typepath
// struct B(pub (super::A));
// struct B(pub (crate::A,));
T![crate] | T![self] | T![super] | T![ident] | T![')'] if p.nth(2) != T![:] => {
// If we are in a tuple struct, then the parens following `pub`
// might be an tuple field, not part of the visibility. So in that
// case we don't want to consume an identifier.
// test pub_tuple_field
// struct MyStruct(pub (u32, u32));
// struct MyStruct(pub (u32));
// struct MyStruct(pub ());
if !(in_tuple_field && matches!(p.nth(1), T![ident] | T![')'])) {
p.bump(T!['(']);
paths::use_path(p);
p.expect(T![')']);
}
}
// test crate_visibility_in
// pub(in super::A) struct S;
// pub(in crate) struct S;
T![in] => {
p.bump(T!['(']);
p.bump(T![in]);
paths::use_path(p);
p.expect(T![')']);
}
_ => {}
}
}
m.complete(p, VISIBILITY);
true
}
fn opt_rename(p: &mut Parser<'_>) {
if p.at(T![as]) {
let m = p.start();
p.bump(T![as]);
if !p.eat(T![_]) {
name(p);
}
m.complete(p, RENAME);
}
}
fn abi(p: &mut Parser<'_>) {
assert!(p.at(T![extern]));
let abi = p.start();
p.bump(T![extern]);
p.eat(STRING);
abi.complete(p, ABI);
}
fn opt_ret_type(p: &mut Parser<'_>) -> bool {
if p.at(T![->]) {
let m = p.start();
p.bump(T![->]);
types::type_no_bounds(p);
m.complete(p, RET_TYPE);
true
} else {
false
}
}
fn name_r(p: &mut Parser<'_>, recovery: TokenSet) {
if p.at(IDENT) {
let m = p.start();
p.bump(IDENT);
m.complete(p, NAME);
} else {
p.err_recover("expected a name", recovery);
}
}
fn name(p: &mut Parser<'_>) {
name_r(p, TokenSet::EMPTY);
}
fn name_ref(p: &mut Parser<'_>) {
if p.at(IDENT) {
let m = p.start();
p.bump(IDENT);
m.complete(p, NAME_REF);
} else {
p.err_and_bump("expected identifier");
}
}
fn name_ref_or_index(p: &mut Parser<'_>) {
assert!(p.at(IDENT) || p.at(INT_NUMBER));
let m = p.start();
p.bump_any();
m.complete(p, NAME_REF);
}
fn lifetime(p: &mut Parser<'_>) {
assert!(p.at(LIFETIME_IDENT));
let m = p.start();
p.bump(LIFETIME_IDENT);
m.complete(p, LIFETIME);
}
fn error_block(p: &mut Parser<'_>, message: &str) {
assert!(p.at(T!['{']));
let m = p.start();
p.error(message);
p.bump(T!['{']);
expressions::expr_block_contents(p);
p.eat(T!['}']);
m.complete(p, ERROR);
}
// test_err top_level_let
// let ref foo: fn() = 1 + 3;
fn error_let_stmt(p: &mut Parser<'_>, message: &str) {
assert!(p.at(T![let]));
let m = p.start();
p.error(message);
expressions::let_stmt(p, expressions::Semicolon::Optional);
m.complete(p, ERROR);
}
/// The `parser` passed this is required to at least consume one token if it returns `true`.
/// If the `parser` returns false, parsing will stop.
fn delimited(
p: &mut Parser<'_>,
bra: SyntaxKind,
ket: SyntaxKind,
delim: SyntaxKind,
unexpected_delim_message: impl Fn() -> String,
first_set: TokenSet,
mut parser: impl FnMut(&mut Parser<'_>) -> bool,
) {
p.bump(bra);
while !p.at(ket) && !p.at(EOF) {
if p.at(delim) {
// Recover if an argument is missing and only got a delimiter,
// e.g. `(a, , b)`.
// Wrap the erroneous delimiter in an error node so that fixup logic gets rid of it.
// FIXME: Ideally this should be handled in fixup in a structured way, but our list
// nodes currently have no concept of a missing node between two delimiters.
// So doing it this way is easier.
let m = p.start();
p.error(unexpected_delim_message());
p.bump(delim);
m.complete(p, ERROR);
continue;
}
if !parser(p) {
break;
}
if !p.eat(delim) {
if p.at_ts(first_set) {
p.error(format!("expected {delim:?}"));
} else {
break;
}
}
}
p.expect(ket);
}