parser: Cleanup LazyTokenStream and avoid some clones

by using a named struct instead of a closure.
This commit is contained in:
Vadim Petrochenkov 2020-10-31 00:40:41 +03:00
parent ffe52882ed
commit d0c63bccc5
4 changed files with 77 additions and 87 deletions

View file

@ -22,7 +22,7 @@ use rustc_serialize::{Decodable, Decoder, Encodable, Encoder};
use rustc_span::{Span, DUMMY_SP}; use rustc_span::{Span, DUMMY_SP};
use smallvec::{smallvec, SmallVec}; use smallvec::{smallvec, SmallVec};
use std::{iter, mem}; use std::{fmt, iter, mem};
/// When the main rust parser encounters a syntax-extension invocation, it /// When the main rust parser encounters a syntax-extension invocation, it
/// parses the arguments to the invocation as a token-tree. This is a very /// parses the arguments to the invocation as a token-tree. This is a very
@ -120,72 +120,51 @@ where
} }
} }
// A cloneable callback which produces a `TokenStream`. Each clone pub trait CreateTokenStream: sync::Send + sync::Sync {
// of this should produce the same `TokenStream` fn create_token_stream(&self) -> TokenStream;
pub trait CreateTokenStream: sync::Send + sync::Sync + FnOnce() -> TokenStream {
// Workaround for the fact that `Clone` is not object-safe
fn clone_it(&self) -> Box<dyn CreateTokenStream>;
} }
impl<F: 'static + Clone + sync::Send + sync::Sync + FnOnce() -> TokenStream> CreateTokenStream impl CreateTokenStream for TokenStream {
for F fn create_token_stream(&self) -> TokenStream {
{ self.clone()
fn clone_it(&self) -> Box<dyn CreateTokenStream> {
Box::new(self.clone())
} }
} }
impl Clone for Box<dyn CreateTokenStream> { /// A lazy version of `TokenStream`, which defers creation
fn clone(&self) -> Self {
let val: &(dyn CreateTokenStream) = &**self;
val.clone_it()
}
}
/// A lazy version of `TokenStream`, which may defer creation
/// of an actual `TokenStream` until it is needed. /// of an actual `TokenStream` until it is needed.
pub type LazyTokenStream = Lrc<LazyTokenStreamInner>; /// `Box` is here only to reduce the structure size.
#[derive(Clone)] #[derive(Clone)]
pub enum LazyTokenStreamInner { pub struct LazyTokenStream(Lrc<Box<dyn CreateTokenStream>>);
Lazy(Box<dyn CreateTokenStream>),
Ready(TokenStream),
}
impl std::fmt::Debug for LazyTokenStreamInner { impl LazyTokenStream {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { pub fn new(inner: impl CreateTokenStream + 'static) -> LazyTokenStream {
match self { LazyTokenStream(Lrc::new(Box::new(inner)))
LazyTokenStreamInner::Lazy(..) => f.debug_struct("LazyTokenStream::Lazy").finish(), }
LazyTokenStreamInner::Ready(..) => f.debug_struct("LazyTokenStream::Ready").finish(),
} pub fn create_token_stream(&self) -> TokenStream {
self.0.create_token_stream()
} }
} }
impl LazyTokenStreamInner { impl fmt::Debug for LazyTokenStream {
pub fn into_token_stream(&self) -> TokenStream { fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
match self { fmt::Debug::fmt("LazyTokenStream", f)
// Note that we do not cache this. If this ever becomes a performance
// problem, we should investigate wrapping `LazyTokenStreamInner`
// in a lock
LazyTokenStreamInner::Lazy(cb) => (cb.clone())(),
LazyTokenStreamInner::Ready(stream) => stream.clone(),
}
} }
} }
impl<S: Encoder> Encodable<S> for LazyTokenStreamInner { impl<S: Encoder> Encodable<S> for LazyTokenStream {
fn encode(&self, _s: &mut S) -> Result<(), S::Error> { fn encode(&self, _s: &mut S) -> Result<(), S::Error> {
panic!("Attempted to encode LazyTokenStream"); panic!("Attempted to encode LazyTokenStream");
} }
} }
impl<D: Decoder> Decodable<D> for LazyTokenStreamInner { impl<D: Decoder> Decodable<D> for LazyTokenStream {
fn decode(_d: &mut D) -> Result<Self, D::Error> { fn decode(_d: &mut D) -> Result<Self, D::Error> {
panic!("Attempted to decode LazyTokenStream"); panic!("Attempted to decode LazyTokenStream");
} }
} }
impl<CTX> HashStable<CTX> for LazyTokenStreamInner { impl<CTX> HashStable<CTX> for LazyTokenStream {
fn hash_stable(&self, _hcx: &mut CTX, _hasher: &mut StableHasher) { fn hash_stable(&self, _hcx: &mut CTX, _hasher: &mut StableHasher) {
panic!("Attempted to compute stable hash for LazyTokenStream"); panic!("Attempted to compute stable hash for LazyTokenStream");
} }

View file

@ -4,12 +4,11 @@ use rustc_ast::attr::HasAttrs;
use rustc_ast::mut_visit::*; use rustc_ast::mut_visit::*;
use rustc_ast::ptr::P; use rustc_ast::ptr::P;
use rustc_ast::token::{DelimToken, Token, TokenKind}; use rustc_ast::token::{DelimToken, Token, TokenKind};
use rustc_ast::tokenstream::{DelimSpan, LazyTokenStreamInner, Spacing, TokenStream, TokenTree}; use rustc_ast::tokenstream::{DelimSpan, LazyTokenStream, Spacing, TokenStream, TokenTree};
use rustc_ast::{self as ast, AttrItem, Attribute, MetaItem}; use rustc_ast::{self as ast, AttrItem, Attribute, MetaItem};
use rustc_attr as attr; use rustc_attr as attr;
use rustc_data_structures::fx::FxHashMap; use rustc_data_structures::fx::FxHashMap;
use rustc_data_structures::map_in_place::MapInPlace; use rustc_data_structures::map_in_place::MapInPlace;
use rustc_data_structures::sync::Lrc;
use rustc_errors::{error_code, struct_span_err, Applicability, Handler}; use rustc_errors::{error_code, struct_span_err, Applicability, Handler};
use rustc_feature::{Feature, Features, State as FeatureState}; use rustc_feature::{Feature, Features, State as FeatureState};
use rustc_feature::{ use rustc_feature::{
@ -303,7 +302,7 @@ impl<'a> StripUnconfigured<'a> {
// Use the `#` in `#[cfg_attr(pred, attr)]` as the `#` token // Use the `#` in `#[cfg_attr(pred, attr)]` as the `#` token
// for `attr` when we expand it to `#[attr]` // for `attr` when we expand it to `#[attr]`
let pound_token = orig_tokens.into_token_stream().trees().next().unwrap(); let pound_token = orig_tokens.create_token_stream().trees().next().unwrap();
if !matches!(pound_token, TokenTree::Token(Token { kind: TokenKind::Pound, .. })) { if !matches!(pound_token, TokenTree::Token(Token { kind: TokenKind::Pound, .. })) {
panic!("Bad tokens for attribute {:?}", attr); panic!("Bad tokens for attribute {:?}", attr);
} }
@ -313,16 +312,16 @@ impl<'a> StripUnconfigured<'a> {
DelimSpan::from_single(pound_token.span()), DelimSpan::from_single(pound_token.span()),
DelimToken::Bracket, DelimToken::Bracket,
item.tokens item.tokens
.clone() .as_ref()
.unwrap_or_else(|| panic!("Missing tokens for {:?}", item)) .unwrap_or_else(|| panic!("Missing tokens for {:?}", item))
.into_token_stream(), .create_token_stream(),
); );
let mut attr = attr::mk_attr_from_item(attr.style, item, span); let mut attr = attr::mk_attr_from_item(attr.style, item, span);
attr.tokens = Some(Lrc::new(LazyTokenStreamInner::Ready(TokenStream::new(vec![ attr.tokens = Some(LazyTokenStream::new(TokenStream::new(vec![
(pound_token, Spacing::Alone), (pound_token, Spacing::Alone),
(bracket_group, Spacing::Alone), (bracket_group, Spacing::Alone),
])))); ])));
self.process_cfg_attr(attr) self.process_cfg_attr(attr)
}) })
.collect() .collect()

View file

@ -249,29 +249,30 @@ pub fn nt_to_tokenstream(nt: &Nonterminal, sess: &ParseSess, span: Span) -> Toke
// came from. Here we attempt to extract these lossless token streams // came from. Here we attempt to extract these lossless token streams
// before we fall back to the stringification. // before we fall back to the stringification.
let convert_tokens = |tokens: Option<LazyTokenStream>| tokens.map(|t| t.into_token_stream()); let convert_tokens =
|tokens: &Option<LazyTokenStream>| tokens.as_ref().map(|t| t.create_token_stream());
let tokens = match *nt { let tokens = match *nt {
Nonterminal::NtItem(ref item) => prepend_attrs(&item.attrs, item.tokens.as_ref()), Nonterminal::NtItem(ref item) => prepend_attrs(&item.attrs, item.tokens.as_ref()),
Nonterminal::NtBlock(ref block) => convert_tokens(block.tokens.clone()), Nonterminal::NtBlock(ref block) => convert_tokens(&block.tokens),
Nonterminal::NtStmt(ref stmt) => { Nonterminal::NtStmt(ref stmt) => {
// FIXME: We currently only collect tokens for `:stmt` // FIXME: We currently only collect tokens for `:stmt`
// matchers in `macro_rules!` macros. When we start collecting // matchers in `macro_rules!` macros. When we start collecting
// tokens for attributes on statements, we will need to prepend // tokens for attributes on statements, we will need to prepend
// attributes here // attributes here
convert_tokens(stmt.tokens.clone()) convert_tokens(&stmt.tokens)
} }
Nonterminal::NtPat(ref pat) => convert_tokens(pat.tokens.clone()), Nonterminal::NtPat(ref pat) => convert_tokens(&pat.tokens),
Nonterminal::NtTy(ref ty) => convert_tokens(ty.tokens.clone()), Nonterminal::NtTy(ref ty) => convert_tokens(&ty.tokens),
Nonterminal::NtIdent(ident, is_raw) => { Nonterminal::NtIdent(ident, is_raw) => {
Some(tokenstream::TokenTree::token(token::Ident(ident.name, is_raw), ident.span).into()) Some(tokenstream::TokenTree::token(token::Ident(ident.name, is_raw), ident.span).into())
} }
Nonterminal::NtLifetime(ident) => { Nonterminal::NtLifetime(ident) => {
Some(tokenstream::TokenTree::token(token::Lifetime(ident.name), ident.span).into()) Some(tokenstream::TokenTree::token(token::Lifetime(ident.name), ident.span).into())
} }
Nonterminal::NtMeta(ref attr) => convert_tokens(attr.tokens.clone()), Nonterminal::NtMeta(ref attr) => convert_tokens(&attr.tokens),
Nonterminal::NtPath(ref path) => convert_tokens(path.tokens.clone()), Nonterminal::NtPath(ref path) => convert_tokens(&path.tokens),
Nonterminal::NtVis(ref vis) => convert_tokens(vis.tokens.clone()), Nonterminal::NtVis(ref vis) => convert_tokens(&vis.tokens),
Nonterminal::NtTT(ref tt) => Some(tt.clone().into()), Nonterminal::NtTT(ref tt) => Some(tt.clone().into()),
Nonterminal::NtExpr(ref expr) | Nonterminal::NtLiteral(ref expr) => { Nonterminal::NtExpr(ref expr) | Nonterminal::NtLiteral(ref expr) => {
if expr.tokens.is_none() { if expr.tokens.is_none() {
@ -604,7 +605,7 @@ fn prepend_attrs(
attrs: &[ast::Attribute], attrs: &[ast::Attribute],
tokens: Option<&tokenstream::LazyTokenStream>, tokens: Option<&tokenstream::LazyTokenStream>,
) -> Option<tokenstream::TokenStream> { ) -> Option<tokenstream::TokenStream> {
let tokens = tokens?.clone().into_token_stream(); let tokens = tokens?.create_token_stream();
if attrs.is_empty() { if attrs.is_empty() {
return Some(tokens); return Some(tokens);
} }
@ -617,9 +618,9 @@ fn prepend_attrs(
); );
builder.push( builder.push(
attr.tokens attr.tokens
.clone() .as_ref()
.unwrap_or_else(|| panic!("Attribute {:?} is missing tokens!", attr)) .unwrap_or_else(|| panic!("Attribute {:?} is missing tokens!", attr))
.into_token_stream(), .create_token_stream(),
); );
} }
builder.push(tokens); builder.push(tokens);

View file

@ -16,8 +16,8 @@ pub use path::PathStyle;
use rustc_ast::ptr::P; use rustc_ast::ptr::P;
use rustc_ast::token::{self, DelimToken, Token, TokenKind}; use rustc_ast::token::{self, DelimToken, Token, TokenKind};
use rustc_ast::tokenstream::{self, DelimSpan, LazyTokenStream, LazyTokenStreamInner, Spacing}; use rustc_ast::tokenstream::{self, DelimSpan, LazyTokenStream, Spacing};
use rustc_ast::tokenstream::{TokenStream, TokenTree}; use rustc_ast::tokenstream::{CreateTokenStream, TokenStream, TokenTree};
use rustc_ast::DUMMY_NODE_ID; use rustc_ast::DUMMY_NODE_ID;
use rustc_ast::{self as ast, AnonConst, AttrStyle, AttrVec, Const, CrateSugar, Extern, Unsafe}; use rustc_ast::{self as ast, AnonConst, AttrStyle, AttrVec, Const, CrateSugar, Extern, Unsafe};
use rustc_ast::{Async, Expr, ExprKind, MacArgs, MacDelimiter, Mutability, StrLit}; use rustc_ast::{Async, Expr, ExprKind, MacArgs, MacDelimiter, Mutability, StrLit};
@ -1199,15 +1199,12 @@ impl<'a> Parser<'a> {
f: impl FnOnce(&mut Self) -> PResult<'a, R>, f: impl FnOnce(&mut Self) -> PResult<'a, R>,
) -> PResult<'a, (R, Option<LazyTokenStream>)> { ) -> PResult<'a, (R, Option<LazyTokenStream>)> {
let start_token = (self.token.clone(), self.token_spacing); let start_token = (self.token.clone(), self.token_spacing);
let mut cursor_snapshot = self.token_cursor.clone(); let cursor_snapshot = self.token_cursor.clone();
let ret = f(self)?; let ret = f(self)?;
let new_calls = self.token_cursor.num_next_calls;
let num_calls = new_calls - cursor_snapshot.num_next_calls;
let desugar_doc_comments = self.desugar_doc_comments;
// We didn't capture any tokens // We didn't capture any tokens
let num_calls = self.token_cursor.num_next_calls - cursor_snapshot.num_next_calls;
if num_calls == 0 { if num_calls == 0 {
return Ok((ret, None)); return Ok((ret, None));
} }
@ -1220,27 +1217,41 @@ impl<'a> Parser<'a> {
// //
// This also makes `Parser` very cheap to clone, since // This also makes `Parser` very cheap to clone, since
// there is no intermediate collection buffer to clone. // there is no intermediate collection buffer to clone.
let lazy_cb = move || { struct LazyTokenStreamImpl {
// The token produced by the final call to `next` or `next_desugared` start_token: (Token, Spacing),
// was not actually consumed by the callback. The combination cursor_snapshot: TokenCursor,
// of chaining the initial token and using `take` produces the desired num_calls: usize,
// result - we produce an empty `TokenStream` if no calls were made, desugar_doc_comments: bool,
// and omit the final token otherwise. }
let tokens = std::iter::once(start_token) impl CreateTokenStream for LazyTokenStreamImpl {
.chain((0..num_calls).map(|_| { fn create_token_stream(&self) -> TokenStream {
if desugar_doc_comments { // The token produced by the final call to `next` or `next_desugared`
cursor_snapshot.next_desugared() // was not actually consumed by the callback. The combination
} else { // of chaining the initial token and using `take` produces the desired
cursor_snapshot.next() // result - we produce an empty `TokenStream` if no calls were made,
} // and omit the final token otherwise.
})) let mut cursor_snapshot = self.cursor_snapshot.clone();
.take(num_calls); let tokens = std::iter::once(self.start_token.clone())
.chain((0..self.num_calls).map(|_| {
if self.desugar_doc_comments {
cursor_snapshot.next_desugared()
} else {
cursor_snapshot.next()
}
}))
.take(self.num_calls);
make_token_stream(tokens) make_token_stream(tokens)
}
}
let lazy_impl = LazyTokenStreamImpl {
start_token,
cursor_snapshot,
num_calls,
desugar_doc_comments: self.desugar_doc_comments,
}; };
let stream = LazyTokenStream::new(LazyTokenStreamInner::Lazy(Box::new(lazy_cb))); Ok((ret, Some(LazyTokenStream::new(lazy_impl))))
Ok((ret, Some(stream)))
} }
/// `::{` or `::*` /// `::{` or `::*`