rust/crates/mbe/src/lib.rs

368 lines
11 KiB
Rust
Raw Normal View History

//! `mbe` (short for Macro By Example) crate contains code for handling
2020-08-12 09:46:20 -05:00
//! `macro_rules` macros. It uses `TokenTree` (from `tt` package) as the
//! interface, although it contains some code to bridge `SyntaxNode`s and
//! `TokenTree`s as well!
2019-01-31 12:29:04 -06:00
mod parser;
2021-01-29 06:23:38 -06:00
mod expander;
2019-01-31 12:29:04 -06:00
mod syntax_bridge;
mod tt_iter;
2019-04-07 08:42:53 -05:00
mod subtree_source;
2019-01-30 14:17:32 -06:00
2020-07-11 07:57:07 -05:00
#[cfg(test)]
mod tests;
2021-02-05 05:57:32 -06:00
#[cfg(test)]
mod benchmark;
2021-05-24 11:43:42 -05:00
mod token_map;
2021-02-05 05:57:32 -06:00
use std::fmt;
2020-12-29 12:35:21 -06:00
pub use tt::{Delimiter, DelimiterKind, Punct};
2019-01-30 14:25:02 -06:00
use crate::{
2021-02-01 14:42:37 -06:00
parser::{parse_pattern, parse_template, MetaTemplate, Op},
tt_iter::TtIter,
};
2019-03-02 13:20:26 -06:00
#[derive(Debug, PartialEq, Eq)]
2019-03-03 03:40:03 -06:00
pub enum ParseError {
UnexpectedToken(String),
2019-03-03 05:45:30 -06:00
Expected(String),
InvalidRepeat,
2021-01-08 08:42:40 -06:00
RepetitionEmptyTokenTree,
2019-03-03 03:40:03 -06:00
}
2020-05-14 04:57:37 -05:00
#[derive(Debug, PartialEq, Eq, Clone)]
2019-03-03 03:40:03 -06:00
pub enum ExpandError {
2019-03-02 13:20:26 -06:00
NoMatchingRule,
UnexpectedToken,
BindingError(String),
ConversionError,
2020-03-26 11:41:44 -05:00
ProcMacroError(tt::ExpansionError),
UnresolvedProcMacro,
Other(String),
2020-03-26 11:41:44 -05:00
}
impl From<tt::ExpansionError> for ExpandError {
fn from(it: tt::ExpansionError) -> Self {
ExpandError::ProcMacroError(it)
}
2019-03-02 13:20:26 -06:00
}
impl fmt::Display for ExpandError {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
match self {
ExpandError::NoMatchingRule => f.write_str("no rule matches input tokens"),
ExpandError::UnexpectedToken => f.write_str("unexpected token in input"),
ExpandError::BindingError(e) => f.write_str(e),
ExpandError::ConversionError => f.write_str("could not convert tokens"),
2020-11-26 12:56:38 -06:00
ExpandError::ProcMacroError(e) => e.fmt(f),
ExpandError::UnresolvedProcMacro => f.write_str("unresolved proc macro"),
ExpandError::Other(e) => f.write_str(e),
}
}
}
2021-05-24 14:47:01 -05:00
pub use crate::{
syntax_bridge::{
ast_to_token_tree, parse_exprs_with_sep, parse_to_token_tree, syntax_node_to_token_tree,
token_tree_to_syntax_node,
},
token_map::TokenMap,
};
2019-01-31 04:46:40 -06:00
2019-02-11 10:07:49 -06:00
/// This struct contains AST for a single `macro_rules` definition. What might
2019-01-31 13:14:28 -06:00
/// be very confusing is that AST has almost exactly the same shape as
/// `tt::TokenTree`, but there's a crucial difference: in macro rules, `$ident`
/// and `$()*` have special meaning (see `Var` and `Repeat` data structures)
#[derive(Clone, Debug, PartialEq, Eq)]
2019-01-31 04:40:05 -06:00
pub struct MacroRules {
rules: Vec<Rule>,
2019-11-04 09:22:18 -06:00
/// Highest id of the token we have in TokenMap
shift: Shift,
2019-01-30 14:17:32 -06:00
}
2021-01-25 15:15:47 -06:00
/// For Macro 2.0
#[derive(Clone, Debug, PartialEq, Eq)]
pub struct MacroDef {
rules: Vec<Rule>,
/// Highest id of the token we have in TokenMap
shift: Shift,
}
#[derive(Clone, Debug, PartialEq, Eq)]
struct Rule {
2020-12-29 12:35:21 -06:00
lhs: MetaTemplate,
rhs: MetaTemplate,
}
#[derive(Clone, Copy, Debug, PartialEq, Eq)]
struct Shift(u32);
impl Shift {
fn new(tt: &tt::Subtree) -> Shift {
// Note that TokenId is started from zero,
// We have to add 1 to prevent duplication.
let value = max_id(tt).map_or(0, |it| it + 1);
return Shift(value);
// Find the max token id inside a subtree
fn max_id(subtree: &tt::Subtree) -> Option<u32> {
subtree
.token_trees
.iter()
.filter_map(|tt| match tt {
tt::TokenTree::Subtree(subtree) => {
let tree_id = max_id(subtree);
match subtree.delimiter {
Some(it) if it.id != tt::TokenId::unspecified() => {
Some(tree_id.map_or(it.id.0, |t| t.max(it.id.0)))
}
_ => tree_id,
}
}
tt::TokenTree::Leaf(tt::Leaf::Ident(ident))
if ident.id != tt::TokenId::unspecified() =>
{
Some(ident.id.0)
}
_ => None,
})
.max()
}
}
/// Shift given TokenTree token id
fn shift_all(self, tt: &mut tt::Subtree) {
2021-06-07 06:59:01 -05:00
for t in &mut tt.token_trees {
match t {
tt::TokenTree::Leaf(leaf) => match leaf {
tt::Leaf::Ident(ident) => ident.id = self.shift(ident.id),
tt::Leaf::Punct(punct) => punct.id = self.shift(punct.id),
tt::Leaf::Literal(lit) => lit.id = self.shift(lit.id),
},
tt::TokenTree::Subtree(tt) => {
2019-12-20 14:14:30 -06:00
if let Some(it) = tt.delimiter.as_mut() {
it.id = self.shift(it.id);
};
self.shift_all(tt)
}
2019-11-04 11:01:05 -06:00
}
}
}
2019-11-04 09:22:18 -06:00
fn shift(self, id: tt::TokenId) -> tt::TokenId {
if id == tt::TokenId::unspecified() {
return id;
2019-11-04 09:22:18 -06:00
}
tt::TokenId(id.0 + self.0)
}
fn unshift(self, id: tt::TokenId) -> Option<tt::TokenId> {
id.0.checked_sub(self.0).map(tt::TokenId)
2019-11-04 09:22:18 -06:00
}
}
#[derive(Debug, Eq, PartialEq)]
pub enum Origin {
Def,
Call,
}
2019-01-31 13:14:28 -06:00
impl MacroRules {
2019-03-03 03:40:03 -06:00
pub fn parse(tt: &tt::Subtree) -> Result<MacroRules, ParseError> {
2019-09-22 15:39:29 -05:00
// Note: this parsing can be implemented using mbe machinery itself, by
// matching against `$($lhs:tt => $rhs:tt);*` pattern, but implementing
// manually seems easier.
let mut src = TtIter::new(tt);
let mut rules = Vec::new();
while src.len() > 0 {
2021-01-25 15:15:47 -06:00
let rule = Rule::parse(&mut src, true)?;
rules.push(rule);
if let Err(()) = src.expect_char(';') {
if src.len() > 0 {
2020-12-29 12:35:21 -06:00
return Err(ParseError::Expected("expected `;`".to_string()));
}
break;
}
}
2019-09-22 15:39:29 -05:00
2021-06-07 06:59:01 -05:00
for rule in &rules {
2019-09-22 15:39:29 -05:00
validate(&rule.lhs)?;
}
Ok(MacroRules { rules, shift: Shift::new(tt) })
2019-01-31 13:14:28 -06:00
}
2019-11-04 09:22:18 -06:00
pub fn expand(&self, tt: &tt::Subtree) -> ExpandResult<tt::Subtree> {
2019-11-04 09:22:18 -06:00
// apply shift
let mut tt = tt.clone();
self.shift.shift_all(&mut tt);
2021-01-29 06:23:38 -06:00
expander::expand_rules(&self.rules, &tt)
2021-01-25 15:15:47 -06:00
}
pub fn map_id_down(&self, id: tt::TokenId) -> tt::TokenId {
self.shift.shift(id)
}
pub fn map_id_up(&self, id: tt::TokenId) -> (tt::TokenId, Origin) {
match self.shift.unshift(id) {
Some(id) => (id, Origin::Call),
None => (id, Origin::Def),
}
}
}
impl MacroDef {
pub fn parse(tt: &tt::Subtree) -> Result<MacroDef, ParseError> {
let mut src = TtIter::new(tt);
let mut rules = Vec::new();
if Some(tt::DelimiterKind::Brace) == tt.delimiter_kind() {
2021-03-08 14:19:44 -06:00
cov_mark::hit!(parse_macro_def_rules);
2021-01-25 15:15:47 -06:00
while src.len() > 0 {
let rule = Rule::parse(&mut src, true)?;
rules.push(rule);
2021-04-02 20:08:31 -05:00
if let Err(()) = src.expect_any_char(&[';', ',']) {
2021-01-25 15:15:47 -06:00
if src.len() > 0 {
2021-04-02 20:08:31 -05:00
return Err(ParseError::Expected(
"expected `;` or `,` to delimit rules".to_string(),
));
2021-01-25 15:15:47 -06:00
}
break;
}
}
} else {
2021-03-08 14:19:44 -06:00
cov_mark::hit!(parse_macro_def_simple);
2021-01-25 15:15:47 -06:00
let rule = Rule::parse(&mut src, false)?;
if src.len() != 0 {
return Err(ParseError::Expected("remain tokens in macro def".to_string()));
}
rules.push(rule);
}
2021-06-07 06:59:01 -05:00
for rule in &rules {
2021-01-25 15:15:47 -06:00
validate(&rule.lhs)?;
}
Ok(MacroDef { rules, shift: Shift::new(tt) })
}
pub fn expand(&self, tt: &tt::Subtree) -> ExpandResult<tt::Subtree> {
// apply shift
let mut tt = tt.clone();
self.shift.shift_all(&mut tt);
2021-01-29 06:23:38 -06:00
expander::expand_rules(&self.rules, &tt)
2019-01-31 13:14:28 -06:00
}
pub fn map_id_down(&self, id: tt::TokenId) -> tt::TokenId {
self.shift.shift(id)
}
pub fn map_id_up(&self, id: tt::TokenId) -> (tt::TokenId, Origin) {
match self.shift.unshift(id) {
Some(id) => (id, Origin::Call),
None => (id, Origin::Def),
}
}
2019-01-31 13:14:28 -06:00
}
impl Rule {
2021-06-07 06:59:01 -05:00
fn parse(src: &mut TtIter, expect_arrow: bool) -> Result<Self, ParseError> {
2020-12-30 12:38:40 -06:00
let lhs = src
.expect_subtree()
2020-12-30 12:38:40 -06:00
.map_err(|()| ParseError::Expected("expected subtree".to_string()))?;
2021-01-25 15:15:47 -06:00
if expect_arrow {
src.expect_char('=').map_err(|()| ParseError::Expected("expected `=`".to_string()))?;
src.expect_char('>').map_err(|()| ParseError::Expected("expected `>`".to_string()))?;
}
2020-12-30 12:38:40 -06:00
let rhs = src
.expect_subtree()
2020-12-30 12:38:40 -06:00
.map_err(|()| ParseError::Expected("expected subtree".to_string()))?;
2020-12-29 12:35:21 -06:00
2021-06-12 22:54:16 -05:00
let lhs = MetaTemplate(parse_pattern(lhs)?);
let rhs = MetaTemplate(parse_template(rhs)?);
2020-12-29 12:35:21 -06:00
Ok(crate::Rule { lhs, rhs })
}
2019-04-24 10:01:32 -05:00
}
2020-12-29 12:35:21 -06:00
fn validate(pattern: &MetaTemplate) -> Result<(), ParseError> {
for op in pattern.iter() {
match op {
2021-06-12 22:54:16 -05:00
Op::Subtree { tokens, .. } => validate(tokens)?,
2021-01-30 02:12:30 -06:00
Op::Repeat { tokens: subtree, separator, .. } => {
// Checks that no repetition which could match an empty token
// https://github.com/rust-lang/rust/blob/a58b1ed44f5e06976de2bdc4d7dc81c36a96934f/src/librustc_expand/mbe/macro_rules.rs#L558
if separator.is_none()
&& subtree.iter().all(|child_op| {
match child_op {
Op::Var { kind, .. } => {
// vis is optional
2020-12-29 12:35:21 -06:00
if kind.as_ref().map_or(false, |it| it == "vis") {
return true;
}
}
Op::Repeat { kind, .. } => {
return matches!(
kind,
parser::RepeatKind::ZeroOrMore | parser::RepeatKind::ZeroOrOne
)
}
Op::Leaf(_) => {}
2021-01-30 02:12:30 -06:00
Op::Subtree { .. } => {}
}
false
})
{
return Err(ParseError::RepetitionEmptyTokenTree);
}
validate(subtree)?
}
_ => (),
}
}
Ok(())
2019-01-30 14:25:02 -06:00
}
2019-01-31 13:14:28 -06:00
2020-11-26 09:04:23 -06:00
#[derive(Debug, Clone, Eq, PartialEq)]
pub struct ExpandResult<T> {
pub value: T,
pub err: Option<ExpandError>,
}
2020-03-16 06:22:10 -05:00
impl<T> ExpandResult<T> {
2020-11-26 09:04:23 -06:00
pub fn ok(value: T) -> Self {
Self { value, err: None }
2020-03-16 06:22:10 -05:00
}
2020-11-26 09:04:23 -06:00
pub fn only_err(err: ExpandError) -> Self
2020-03-16 06:22:10 -05:00
where
T: Default,
{
2020-11-26 09:04:23 -06:00
Self { value: Default::default(), err: Some(err) }
2020-03-16 06:22:10 -05:00
}
pub fn str_err(err: String) -> Self
where
T: Default,
{
Self::only_err(ExpandError::Other(err))
}
2020-03-16 06:22:10 -05:00
pub fn map<U>(self, f: impl FnOnce(T) -> U) -> ExpandResult<U> {
2020-11-26 09:04:23 -06:00
ExpandResult { value: f(self.value), err: self.err }
2020-03-16 06:22:10 -05:00
}
pub fn result(self) -> Result<T, ExpandError> {
2021-06-07 06:59:01 -05:00
self.err.map_or(Ok(self.value), Err)
2020-03-16 06:22:10 -05:00
}
}
impl<T: Default> From<Result<T, ExpandError>> for ExpandResult<T> {
2020-11-26 09:04:23 -06:00
fn from(result: Result<T, ExpandError>) -> Self {
2021-03-21 09:33:18 -05:00
result.map_or_else(Self::only_err, Self::ok)
2020-03-16 06:22:10 -05:00
}
}