bjoernager/rust
bjoernager/rust
1
Fork 0
Code Activity

Use TokenTrees in lhs of macros

Browse source
This commit is contained in:
Piotr Czarnecki 2014-10-06 23:00:56 +01:00
parent 5c1fd5f8b7
commit 38ce6d9eac
10 changed files with 387 additions and 249 deletions
Download patch file Download diff file Expand all files Collapse all files

3
src/librustdoc/html/highlight.rs
View file

@ -163,7 +163,8 @@ fn doit(sess: &parse::ParseSess, mut lexer: lexer::StringReader,
token::Lifetime(..) => "lifetime",
token::DocComment(..) => "doccomment",
token::Underscore | token::Eof | token::Interpolated(..) => "",
token::Underscore | token::Eof | token::Interpolated(..) |
token::MatchNt(..) | token::SubstNt(..) => "",
};
// as mentioned above, use the original source code instead of

50
src/libsyntax/ast.rs
View file

@ -10,7 +10,7 @@
// The Rust abstract syntax tree.
use codemap::{Span, Spanned, DUMMY_SP, ExpnId};
use codemap::{Span, Spanned, DUMMY_SP, ExpnId, respan};
use abi::Abi;
use ast_util;
use owned_slice::OwnedSlice;
@ -657,23 +657,55 @@ pub enum TokenTree {
/// A delimited sequence of token trees
TtDelimited(Span, Rc<Delimited>),
// These only make sense for right-hand-sides of MBE macros:
// This only makes sense for right-hand-sides of MBE macros:
/// A Kleene-style repetition sequence with an optional separator.
// FIXME(eddyb) #6308 Use Rc<[TokenTree]> after DST.
TtSequence(Span, Rc<Vec<TokenTree>>, Option<token::Token>, KleeneOp),
/// A syntactic variable that will be filled in by macro expansion.
TtNonterminal(Span, Ident)
TtSequence(Span, Rc<Vec<TokenTree>>, Option<::parse::token::Token>, KleeneOp, uint),
}
impl TokenTree {
pub fn expand_into_tts(self) -> Rc<Vec<TokenTree>> {
match self {
TtToken(sp, token::DocComment(name)) => {
let doc = MetaNameValue(token::intern_and_get_ident("doc"),
respan(sp, LitStr(token::get_name(name), CookedStr)));
let doc = token::NtMeta(P(respan(sp, doc)));
let delimed = Delimited {
delim: token::Bracket,
open_span: sp,
tts: vec![TtToken(sp, token::Interpolated(doc))],
close_span: sp,
};
Rc::new(vec![TtToken(sp, token::Pound),
TtDelimited(sp, Rc::new(delimed))])
}
TtDelimited(_, ref delimed) => {
let mut tts = Vec::with_capacity(1 + delimed.tts.len() + 1);
tts.push(delimed.open_tt());
tts.extend(delimed.tts.iter().map(|tt| tt.clone()));
tts.push(delimed.close_tt());
Rc::new(tts)
}
TtToken(sp, token::SubstNt(name, namep)) => {
Rc::new(vec![TtToken(sp, token::Dollar),
TtToken(sp, token::Ident(name, namep))])
}
TtToken(sp, token::MatchNt(name, kind, namep, kindp)) => {
Rc::new(vec![TtToken(sp, token::SubstNt(name, namep)),
TtToken(sp, token::Colon),
TtToken(sp, token::Ident(kind, kindp))])
}
_ => panic!("Cannot expand a token")
}
}
/// Returns the `Span` corresponding to this token tree.
pub fn get_span(&self) -> Span {
match *self {
TtToken(span, _) => span,
TtDelimited(span, _) => span,
TtSequence(span, _, _, _) => span,
TtNonterminal(span, _) => span,
TtToken(span, _) => span,
TtDelimited(span, _) => span,
TtSequence(span, _, _, _, _) => span,
}
}
}

52
src/libsyntax/ext/quote.rs
View file

@ -616,6 +616,20 @@ fn mk_token(cx: &ExtCtxt, sp: Span, tok: &token::Token) -> P<ast::Expr> {
vec!(mk_name(cx, sp, ident.ident())));
}
token::MatchNt(name, kind, name_style, kind_style) => {
return cx.expr_call(sp,
mk_token_path(cx, sp, "MatchNt"),
vec![mk_ident(cx, sp, name),
mk_ident(cx, sp, kind),
match name_style {
ModName => mk_token_path(cx, sp, "ModName"),
Plain => mk_token_path(cx, sp, "Plain"),
},
match kind_style {
ModName => mk_token_path(cx, sp, "ModName"),
Plain => mk_token_path(cx, sp, "Plain"),
}]);
}
token::Interpolated(_) => panic!("quote! with interpolated token"),
_ => ()
@ -654,6 +668,25 @@ fn mk_token(cx: &ExtCtxt, sp: Span, tok: &token::Token) -> P<ast::Expr> {
fn mk_tt(cx: &ExtCtxt, _: Span, tt: &ast::TokenTree) -> Vec<P<ast::Stmt>> {
match *tt {
ast::TtToken(sp, SubstNt(ident, _)) => {
// tt.extend($ident.to_tokens(ext_cx).into_iter())
let e_to_toks =
cx.expr_method_call(sp,
cx.expr_ident(sp, ident),
id_ext("to_tokens"),
vec!(cx.expr_ident(sp, id_ext("ext_cx"))));
let e_to_toks =
cx.expr_method_call(sp, e_to_toks, id_ext("into_iter"), vec![]);
let e_push =
cx.expr_method_call(sp,
cx.expr_ident(sp, id_ext("tt")),
id_ext("extend"),
vec!(e_to_toks));
vec!(cx.stmt_expr(e_push))
}
ast::TtToken(sp, ref tok) => {
let e_sp = cx.expr_ident(sp, id_ext("_sp"));
let e_tok = cx.expr_call(sp,
@ -673,25 +706,6 @@ fn mk_tt(cx: &ExtCtxt, _: Span, tt: &ast::TokenTree) -> Vec<P<ast::Stmt>> {
.collect()
},
ast::TtSequence(..) => panic!("TtSequence in quote!"),
ast::TtNonterminal(sp, ident) => {
// tt.extend($ident.to_tokens(ext_cx).into_iter())
let e_to_toks =
cx.expr_method_call(sp,
cx.expr_ident(sp, ident),
id_ext("to_tokens"),
vec!(cx.expr_ident(sp, id_ext("ext_cx"))));
let e_to_toks =
cx.expr_method_call(sp, e_to_toks, id_ext("into_iter"), vec![]);
let e_push =
cx.expr_method_call(sp,
cx.expr_ident(sp, id_ext("tt")),
id_ext("extend"),
vec!(e_to_toks));
vec!(cx.stmt_expr(e_push))
},
}
}

283
src/libsyntax/ext/tt/macro_parser.rs
View file

@ -78,69 +78,80 @@
use ast;
use ast::{Matcher, MatchTok, MatchSeq, MatchNonterminal, Ident};
use ast::{Matcher, TokenTree, Ident};
use ast::{TtDelimited, TtSequence, TtToken};
use codemap::{BytePos, mk_sp};
use codemap;
use parse::lexer::*; //resolve bug?
use parse::ParseSess;
use parse::attr::ParserAttr;
use parse::parser::{LifetimeAndTypesWithoutColons, Parser};
use parse::token::{Eof, DocComment, MatchNt, SubstNt};
use parse::token::{Token, Nonterminal};
use parse::token;
use print::pprust;
use ptr::P;
use std::mem;
use std::rc::Rc;
use std::collections::HashMap;
use std::collections::hash_map::{Vacant, Occupied};
/* to avoid costly uniqueness checks, we require that `MatchSeq` always has a
nonempty body. */
/// an unzipping of `TokenTree`s
#[deriving(Clone)]
struct MatcherTtFrame {
elts: Rc<Vec<ast::TokenTree>>,
idx: uint,
}
#[deriving(Clone)]
pub struct MatcherPos {
elts: Vec<ast::Matcher> , // maybe should be <'>? Need to understand regions.
stack: Vec<MatcherTtFrame>,
elts: Rc<Vec<ast::TokenTree>>,
sep: Option<Token>,
idx: uint,
up: Option<Box<MatcherPos>>,
matches: Vec<Vec<Rc<NamedMatch>>>,
match_lo: uint, match_hi: uint,
match_lo: uint,
match_cur: uint,
match_hi: uint,
sp_lo: BytePos,
}
pub fn count_names(ms: &[Matcher]) -> uint {
ms.iter().fold(0, |ct, m| {
ct + match m.node {
MatchTok(_) => 0u,
MatchSeq(ref more_ms, _, _, _, _) => {
count_names(more_ms.as_slice())
pub fn count_names(ms: &[TokenTree]) -> uint {
ms.iter().fold(0, |count, elt| {
count + match elt {
&TtSequence(_, _, _, _, advance_by) => {
advance_by
}
MatchNonterminal(_, _, _) => 1u
}})
&TtDelimited(_, ref delim) => {
count_names(delim.tts.as_slice())
}
&TtToken(_, MatchNt(..)) => {
1
}
&TtToken(_, _) => 0,
}
})
}
pub fn initial_matcher_pos(ms: Vec<Matcher> , sep: Option<Token>, lo: BytePos)
pub fn initial_matcher_pos(ms: Rc<Vec<TokenTree>>, sep: Option<Token>, lo: BytePos)
-> Box<MatcherPos> {
let mut match_idx_hi = 0u;
for elt in ms.iter() {
match elt.node {
MatchTok(_) => (),
MatchSeq(_,_,_,_,hi) => {
match_idx_hi = hi; // it is monotonic...
}
MatchNonterminal(_,_,pos) => {
match_idx_hi = pos+1u; // ...so latest is highest
}
}
}
let matches = Vec::from_fn(count_names(ms.as_slice()), |_i| Vec::new());
let match_idx_hi = count_names(ms.as_slice());
let matches = Vec::from_fn(match_idx_hi, |_i| Vec::new());
box MatcherPos {
stack: vec![],
elts: ms,
sep: sep,
idx: 0u,
up: None,
matches: matches,
match_lo: 0u,
match_cur: 0u,
match_hi: match_idx_hi,
sp_lo: lo
}
@ -152,11 +163,9 @@ pub fn initial_matcher_pos(ms: Vec<Matcher> , sep: Option<Token>, lo: BytePos)
/// (expr, item, etc). All the leaves in a single NamedMatch correspond to a
/// single matcher_nonterminal in the ast::Matcher that produced it.
///
/// It should probably be renamed, it has more or less exact correspondence to
/// ast::match nodes, and the in-memory structure of a particular NamedMatch
/// represents the match that occurred when a particular subset of an
/// ast::match -- those ast::Matcher nodes leading to a single
/// MatchNonterminal -- was applied to a particular token tree.
/// The in-memory structure of a particular NamedMatch represents the match
/// that occurred when a particular subset of a matcher was applied to a
/// particular token tree.
///
/// The width of each MatchedSeq in the NamedMatch, and the identity of the
/// MatchedNonterminal's, will depend on the token tree it was applied to: each
@ -170,34 +179,43 @@ pub enum NamedMatch {
MatchedNonterminal(Nonterminal)
}
pub fn nameize(p_s: &ParseSess, ms: &[Matcher], res: &[Rc<NamedMatch>])
pub fn nameize(p_s: &ParseSess, ms: &[TokenTree], res: &[Rc<NamedMatch>])
-> HashMap<Ident, Rc<NamedMatch>> {
fn n_rec(p_s: &ParseSess, m: &Matcher, res: &[Rc<NamedMatch>],
ret_val: &mut HashMap<Ident, Rc<NamedMatch>>) {
match *m {
codemap::Spanned {node: MatchTok(_), .. } => (),
codemap::Spanned {node: MatchSeq(ref more_ms, _, _, _, _), .. } => {
for next_m in more_ms.iter() {
n_rec(p_s, next_m, res, ret_val)
};
}
codemap::Spanned {
node: MatchNonterminal(bind_name, _, idx),
span
} => {
if ret_val.contains_key(&bind_name) {
let string = token::get_ident(bind_name);
p_s.span_diagnostic
.span_fatal(span,
format!("duplicated bind name: {}",
string.get()).as_slice())
fn n_rec(p_s: &ParseSess, m: &TokenTree, res: &[Rc<NamedMatch>],
ret_val: &mut HashMap<Ident, Rc<NamedMatch>>, idx: &mut uint) {
match m {
&TtSequence(_, ref more_ms, _, _, _) => {
for next_m in more_ms.iter() {
n_rec(p_s, next_m, res, ret_val, idx)
}
}
ret_val.insert(bind_name, res[idx].clone());
}
&TtDelimited(_, ref delim) => {
for next_m in delim.tts.iter() {
n_rec(p_s, next_m, res, ret_val, idx)
}
}
&TtToken(sp, MatchNt(bind_name, _, _, _)) => {
match ret_val.entry(bind_name) {
Vacant(spot) => {
spot.set(res[*idx].clone());
*idx += 1;
}
Occupied(..) => {
let string = token::get_ident(bind_name);
p_s.span_diagnostic
.span_fatal(sp,
format!("duplicated bind name: {}",
string.get()).as_slice())
}
}
}
&TtToken(_, SubstNt(..)) => panic!("Cannot fill in a NT"),
&TtToken(_, _) => (),
}
}
let mut ret_val = HashMap::new();
for m in ms.iter() { n_rec(p_s, m, res, &mut ret_val) }
let mut idx = 0u;
for m in ms.iter() { n_rec(p_s, m, res, &mut ret_val, &mut idx) }
ret_val
}
@ -210,7 +228,7 @@ pub enum ParseResult {
pub fn parse_or_else(sess: &ParseSess,
cfg: ast::CrateConfig,
rdr: TtReader,
ms: Vec<Matcher> )
ms: Vec<TokenTree> )
-> HashMap<Ident, Rc<NamedMatch>> {
match parse(sess, cfg, rdr, ms.as_slice()) {
Success(m) => m,
@ -237,12 +255,12 @@ pub fn token_name_eq(t1 : &Token, t2 : &Token) -> bool {
pub fn parse(sess: &ParseSess,
cfg: ast::CrateConfig,
mut rdr: TtReader,
ms: &[Matcher])
ms: &[TokenTree])
-> ParseResult {
let mut cur_eis = Vec::new();
cur_eis.push(initial_matcher_pos(ms.iter()
.map(|x| (*x).clone())
.collect(),
cur_eis.push(initial_matcher_pos(Rc::new(ms.iter()
.map(|x| (*x).clone())
.collect()),
None,
rdr.peek().sp.lo));
@ -255,11 +273,22 @@ pub fn parse(sess: &ParseSess,
/* we append new items to this while we go */
loop {
let ei = match cur_eis.pop() {
let mut ei = match cur_eis.pop() {
None => break, /* for each Earley Item */
Some(ei) => ei,
};
// When unzipped trees end, remove them
while ei.idx >= ei.elts.len() {
match ei.stack.pop() {
Some(MatcherTtFrame { elts, idx }) => {
ei.elts = elts;
ei.idx = idx + 1;
}
None => break
}
}
let idx = ei.idx;
let len = ei.elts.len();
@ -293,6 +322,7 @@ pub fn parse(sess: &ParseSess,
sp.hi))));
}
new_pos.match_cur = ei.match_hi;
new_pos.idx += 1;
cur_eis.push(new_pos);
}
@ -301,69 +331,88 @@ pub fn parse(sess: &ParseSess,
// the *_t vars are workarounds for the lack of unary move
match ei.sep {
Some(ref t) if idx == len => { // we need a separator
// i'm conflicted about whether this should be hygienic....
// though in this case, if the separators are never legal
// idents, it shouldn't matter.
if token_name_eq(&tok, t) { //pass the separator
let mut ei_t = ei.clone();
ei_t.idx += 1;
next_eis.push(ei_t);
Some(ref t) if idx == len => { // we need a separator
// i'm conflicted about whether this should be hygienic....
// though in this case, if the separators are never legal
// idents, it shouldn't matter.
if token_name_eq(&tok, t) { //pass the separator
let mut ei_t = ei.clone();
// ei_t.match_cur = ei_t.match_lo;
ei_t.idx += 1;
next_eis.push(ei_t);
}
}
_ => { // we don't need a separator
let mut ei_t = ei;
ei_t.match_cur = ei_t.match_lo;
ei_t.idx = 0;
cur_eis.push(ei_t);
}
}
_ => { // we don't need a separator
let mut ei_t = ei;
ei_t.idx = 0;
cur_eis.push(ei_t);
}
}
} else {
eof_eis.push(ei);
}
} else {
match ei.elts[idx].node.clone() {
/* need to descend into sequence */
MatchSeq(ref matchers, ref sep, kleene_op,
match_idx_lo, match_idx_hi) => {
if kleene_op == ast::ZeroOrMore {
let mut new_ei = ei.clone();
new_ei.idx += 1u;
//we specifically matched zero repeats.
for idx in range(match_idx_lo, match_idx_hi) {
new_ei.matches[idx]
.push(Rc::new(MatchedSeq(Vec::new(), sp)));
match (*ei.elts)[idx].clone() {
/* need to descend into sequence */
TtSequence(_, ref matchers, ref sep, kleene_op, match_num) => {
if kleene_op == ast::ZeroOrMore {
let mut new_ei = ei.clone();
new_ei.match_cur += match_num;
new_ei.idx += 1u;
//we specifically matched zero repeats.
for idx in range(ei.match_cur, ei.match_cur + match_num) {
new_ei.matches[idx]
.push(Rc::new(MatchedSeq(Vec::new(), sp)));
}
cur_eis.push(new_ei);
}
cur_eis.push(new_ei);
let matches = Vec::from_elem(ei.matches.len(), Vec::new());
let ei_t = ei;
cur_eis.push(box MatcherPos {
stack: vec![],
elts: matchers.clone(),
sep: (*sep).clone(),
idx: 0u,
matches: matches,
match_lo: ei_t.match_cur,
match_cur: ei_t.match_cur,
match_hi: ei_t.match_cur + match_num,
up: Some(ei_t),
sp_lo: sp.lo
});
}
let matches = Vec::from_elem(ei.matches.len(), Vec::new());
let ei_t = ei;
cur_eis.push(box MatcherPos {
elts: (*matchers).clone(),
sep: (*sep).clone(),
idx: 0u,
up: Some(ei_t),
matches: matches,
match_lo: match_idx_lo, match_hi: match_idx_hi,
sp_lo: sp.lo
});
}
MatchNonterminal(_,_,_) => {
// Built-in nonterminals never start with these tokens,
// so we can eliminate them from consideration.
match tok {
token::CloseDelim(_) => {},
_ => bb_eis.push(ei),
TtToken(_, MatchNt(..)) => {
// Built-in nonterminals never start with these tokens,
// so we can eliminate them from consideration.
match tok {
token::CloseDelim(_) => {},
_ => bb_eis.push(ei),
}
}
}
MatchTok(ref t) => {
let mut ei_t = ei.clone();
if token_name_eq(t,&tok) {
ei_t.idx += 1;
next_eis.push(ei_t);
TtToken(sp, SubstNt(..)) => {
return Error(sp, "Cannot transcribe in macro LHS".into_string())
}
seq @ TtDelimited(..) | seq @ TtToken(_, DocComment(..)) => {
let tts = seq.expand_into_tts();
let elts = mem::replace(&mut ei.elts, tts);
let idx = ei.idx;
ei.stack.push(MatcherTtFrame {
elts: elts,
idx: idx,
});
ei.idx = 0;
cur_eis.push(ei);
}
TtToken(_, ref t) => {
let mut ei_t = ei.clone();
if token_name_eq(t,&tok) {
ei_t.idx += 1;
next_eis.push(ei_t);
}
}
}
}
}
}
@ -385,8 +434,8 @@ pub fn parse(sess: &ParseSess,
if (bb_eis.len() > 0u && next_eis.len() > 0u)
|| bb_eis.len() > 1u {
let nts = bb_eis.iter().map(|ei| {
match ei.elts[ei.idx].node {
MatchNonterminal(bind, name, _) => {
match (*ei.elts)[ei.idx] {
TtToken(_, MatchNt(bind, name, _, _)) => {
(format!("{} ('{}')",
token::get_ident(name),
token::get_ident(bind))).to_string()
@ -410,12 +459,14 @@ pub fn parse(sess: &ParseSess,
let mut rust_parser = Parser::new(sess, cfg.clone(), box rdr.clone());
let mut ei = bb_eis.pop().unwrap();
match ei.elts[ei.idx].node {
MatchNonterminal(_, name, idx) => {
match (*ei.elts)[ei.idx] {
TtToken(_, MatchNt(_, name, _, _)) => {
let name_string = token::get_ident(name);
ei.matches[idx].push(Rc::new(MatchedNonterminal(
let match_cur = ei.match_cur;
ei.matches[match_cur].push(Rc::new(MatchedNonterminal(
parse_nt(&mut rust_parser, name_string.get()))));
ei.idx += 1u;
ei.match_cur += 1;
}
_ => panic!()
}

45
src/libsyntax/ext/tt/macro_rules.rs
View file

@ -9,8 +9,9 @@
// except according to those terms.
use ast::{Ident, Matcher_, Matcher, MatchTok, MatchNonterminal, MatchSeq, TtDelimited};
use ast::{TtSequence, TtToken};
use ast;
use codemap::{Span, Spanned, DUMMY_SP};
use codemap::{Span, DUMMY_SP};
use ext::base::{ExtCtxt, MacResult, MacroDef};
use ext::base::{NormalTT, TTMacroExpander};
use ext::tt::macro_parser::{Success, Error, Failure};
@ -20,7 +21,7 @@ use parse::lexer::new_tt_reader;
use parse::parser::Parser;
use parse::attr::ParserAttr;
use parse::token::{special_idents, gensym_ident};
use parse::token::{NtMatchers, NtTT};
use parse::token::{MatchNt, NtMatchers, NtTT};
use parse::token;
use print;
use ptr::P;
@ -158,14 +159,18 @@ fn generic_extension<'cx>(cx: &'cx ExtCtxt,
for (i, lhs) in lhses.iter().enumerate() { // try each arm's matchers
match **lhs {
MatchedNonterminal(NtMatchers(ref mtcs)) => {
MatchedNonterminal(NtTT(ref lhs_tt)) => {
let lhs_tt = match **lhs_tt {
TtDelimited(_, ref delim) => delim.tts.as_slice(),
_ => cx.span_fatal(sp, "malformed macro lhs")
};
// `None` is because we're not interpolating
let arg_rdr = new_tt_reader(&cx.parse_sess().span_diagnostic,
None,
arg.iter()
.map(|x| (*x).clone())
.collect());
match parse(cx.parse_sess(), cx.cfg(), arg_rdr, mtcs.as_slice()) {
match parse(cx.parse_sess(), cx.cfg(), arg_rdr, lhs_tt) {
Success(named_matches) => {
let rhs = match *rhses[i] {
// okay, what's your transcriber?
@ -210,31 +215,33 @@ pub fn add_new_extension<'cx>(cx: &'cx mut ExtCtxt,
name: Ident,
arg: Vec<ast::TokenTree> )
-> Box<MacResult+'cx> {
// these spans won't matter, anyways
fn ms(m: Matcher_) -> Matcher {
Spanned {
node: m.clone(),
span: DUMMY_SP
}
}
let lhs_nm = gensym_ident("lhs");
let rhs_nm = gensym_ident("rhs");
// The pattern that macro_rules matches.
// The grammar for macro_rules! is:
// $( $lhs:mtcs => $rhs:tt );+
// $( $lhs:tt => $rhs:tt );+
// ...quasiquoting this would be nice.
// These spans won't matter, anyways
let match_lhs_tok = MatchNt(lhs_nm, special_idents::tt, token::Plain, token::Plain);
let match_rhs_tok = MatchNt(rhs_nm, special_idents::tt, token::Plain, token::Plain);
let argument_gram = vec!(
ms(MatchSeq(vec!(
ms(MatchNonterminal(lhs_nm, special_idents::matchers, 0u)),
ms(MatchTok(token::FatArrow)),
ms(MatchNonterminal(rhs_nm, special_idents::tt, 1u))),
Some(token::Semi), ast::OneOrMore, 0u, 2u)),
TtSequence(DUMMY_SP,
Rc::new(vec![
TtToken(DUMMY_SP, match_lhs),
TtToken(DUMMY_SP, token::FatArrow),
TtToken(DUMMY_SP, match_rhs)]),
Some(token::Semi),
ast::OneOrMore,
2),
//to phase into semicolon-termination instead of
//semicolon-separation
ms(MatchSeq(vec!(ms(MatchTok(token::Semi))), None,
ast::ZeroOrMore, 2u, 2u)));
TtSequence(DUMMY_SP,
Rc::new(vec![TtToken(DUMMY_SP, token::Semi)]),
None,
ast::ZeroOrMore,
0));
// Parse the macro_rules! invocation (`none` is for no interpolations):

137
src/libsyntax/ext/tt/transcribe.rs
View file

@ -9,10 +9,11 @@
// except according to those terms.
use ast;
use ast::{TokenTree, TtDelimited, TtToken, TtSequence, TtNonterminal, Ident};
use ast::{TokenTree, TtDelimited, TtToken, TtSequence, Ident};
use codemap::{Span, DUMMY_SP};
use diagnostic::SpanHandler;
use ext::tt::macro_parser::{NamedMatch, MatchedSeq, MatchedNonterminal};
use parse::token::{Eof, DocComment, Interpolated, MatchNt, SubstNt};
use parse::token::{Token, NtIdent};
use parse::token;
use parse::lexer::TokenAndSpan;
@ -85,17 +86,9 @@ fn lookup_cur_matched_by_matched(r: &TtReader, start: Rc<NamedMatch>) -> Rc<Name
})
}
fn lookup_cur_matched(r: &TtReader, name: Ident) -> Rc<NamedMatch> {
fn lookup_cur_matched(r: &TtReader, name: Ident) -> Option<Rc<NamedMatch>> {
let matched_opt = r.interpolations.find_copy(&name);
match matched_opt {
Some(s) => lookup_cur_matched_by_matched(r, s),
None => {
r.sp_diag
.span_fatal(r.cur_span,
format!("unknown macro variable `{}`",
token::get_ident(name)).as_slice());
}
}
matched_opt.map(|s| lookup_cur_matched_by_matched(r, s))
}
#[deriving(Clone)]
@ -133,16 +126,20 @@ fn lockstep_iter_size(t: &TokenTree, r: &TtReader) -> LockstepIterSize {
size + lockstep_iter_size(tt, r)
})
},
TtSequence(_, ref tts, _, _) => {
TtSequence(_, ref tts, _, _, _) => {
tts.iter().fold(LisUnconstrained, |size, tt| {
size + lockstep_iter_size(tt, r)
})
},
TtToken(_, SubstNt(name, _)) | TtToken(_, MatchNt(name, _, _, _)) =>
match lookup_cur_matched(r, name) {
Some(matched) => match *matched {
MatchedNonterminal(_) => LisUnconstrained,
MatchedSeq(ref ads, _) => LisConstraint(ads.len(), name),
},
_ => LisUnconstrained
},
TtToken(..) => LisUnconstrained,
TtNonterminal(_, name) => match *lookup_cur_matched(r, name) {
MatchedNonterminal(_) => LisUnconstrained,
MatchedSeq(ref ads, _) => LisConstraint(ads.len(), name)
},
}
}
@ -205,40 +202,21 @@ pub fn tt_next_token(r: &mut TtReader) -> TokenAndSpan {
(*frame.forest)[frame.idx].clone()
};
match t {
TtDelimited(_, ref delimed) => {
let mut tts = Vec::with_capacity(1 + delimed.tts.len() + 1);
tts.push(delimed.open_tt());
tts.extend(delimed.tts.iter().map(|tt| tt.clone()));
tts.push(delimed.close_tt());
r.stack.push(TtFrame {
forest: Rc::new(tts),
idx: 0,
dotdotdoted: false,
sep: None
});
// if this could be 0-length, we'd need to potentially recur here
}
TtToken(sp, tok) => {
r.cur_span = sp;
r.cur_tok = tok;
r.stack.last_mut().unwrap().idx += 1;
return ret_val;
}
TtSequence(sp, tts, sep, kleene_op) => {
TtSequence(sp, tts, sep, kleene_op, n) => {
// FIXME(pcwalton): Bad copy.
match lockstep_iter_size(&TtSequence(sp, tts.clone(), sep.clone(), kleene_op), r) {
match lockstep_iter_size(&TtSequence(sp, tts.clone(), sep.clone(), kleene_op, n),
r) {
LisUnconstrained => {
r.sp_diag.span_fatal(
sp.clone(), /* blame macro writer */
"attempted to repeat an expression \
containing no syntax \
variables matched as repeating at this depth");
}
LisContradiction(ref msg) => {
// FIXME #2887 blame macro invoker instead
r.sp_diag.span_fatal(sp.clone(), msg.as_slice());
}
}
LisContradiction(ref msg) => {
// FIXME #2887 blame macro invoker instead
r.sp_diag.span_fatal(sp.clone(), msg.as_slice());
}
LisConstraint(len, _) => {
if len == 0 {
if kleene_op == ast::OneOrMore {
@ -262,31 +240,62 @@ pub fn tt_next_token(r: &mut TtReader) -> TokenAndSpan {
}
}
// FIXME #2887: think about span stuff here
TtNonterminal(sp, ident) => {
r.stack.last_mut().unwrap().idx += 1;
match *lookup_cur_matched(r, ident) {
/* sidestep the interpolation tricks for ident because
(a) idents can be in lots of places, so it'd be a pain
(b) we actually can, since it's a token. */
MatchedNonterminal(NtIdent(box sn, b)) => {
r.cur_span = sp;
r.cur_tok = token::Ident(sn,b);
return ret_val;
TtToken(sp, SubstNt(ident, namep)) => {
match lookup_cur_matched(r, ident) {
None => {
r.stack.push(TtFrame {
forest: TtToken(sp, SubstNt(ident, namep)).expand_into_tts(),
idx: 0,
dotdotdoted: false,
sep: None
});
// this can't be 0 length, just like TtDelimited
}
MatchedNonterminal(ref other_whole_nt) => {
// FIXME(pcwalton): Bad copy.
r.cur_span = sp;
r.cur_tok = token::Interpolated((*other_whole_nt).clone());
return ret_val;
}
MatchedSeq(..) => {
r.sp_diag.span_fatal(
r.cur_span, /* blame the macro writer */
format!("variable '{}' is still repeating at this depth",
token::get_ident(ident)).as_slice());
Some(cur_matched) => {
r.stack.last_mut().unwrap().idx += 1;
match *cur_matched {
// sidestep the interpolation tricks for ident because
// (a) idents can be in lots of places, so it'd be a pain
// (b) we actually can, since it's a token.
MatchedNonterminal(NtIdent(box sn, b)) => {
r.cur_span = sp;
r.cur_tok = token::Ident(sn, b);
return ret_val;
}
MatchedNonterminal(ref other_whole_nt) => {
// FIXME(pcwalton): Bad copy.
r.cur_span = sp;
r.cur_tok = token::Interpolated((*other_whole_nt).clone());
return ret_val;
}
MatchedSeq(..) => {
r.sp_diag.span_fatal(
r.cur_span, /* blame the macro writer */
format!("variable '{}' is still repeating at this depth",
token::get_ident(ident)).as_slice());
}
}
}
}
}
// TtDelimited or any token that can be unzipped
seq @ TtDelimited(..) | seq @ TtToken(_, DocComment(..))
| seq @ TtToken(_, MatchNt(..)) => {
// do not advance the idx yet
r.stack.push(TtFrame {
forest: seq.expand_into_tts(),
idx: 0,
dotdotdoted: false,
sep: None
});
// if this could be 0-length, we'd need to potentially recur here
}
TtToken(sp, tok) => {
r.cur_span = sp;
r.cur_tok = tok;
r.stack.last_mut().unwrap().idx += 1;
return ret_val;
}
}
}
}

13
src/libsyntax/fold.rs
View file

@ -581,13 +581,12 @@ pub fn noop_fold_tt<T: Folder>(tt: &TokenTree, fld: &mut T) -> TokenTree {
}
))
},
TtSequence(span, ref pattern, ref sep, is_optional) =>
TtSequence(span, ref pattern, ref sep, is_optional, advance_by) =>
TtSequence(span,
Rc::new(fld.fold_tts(pattern.as_slice())),
sep.clone().map(|tok| fld.fold_token(tok)),
is_optional),
TtNonterminal(sp,ref ident) =>
TtNonterminal(sp,fld.fold_ident(*ident))
is_optional,
advance_by),
}
}
@ -603,6 +602,12 @@ pub fn noop_fold_token<T: Folder>(t: token::Token, fld: &mut T) -> token::Token
}
token::Lifetime(id) => token::Lifetime(fld.fold_ident(id)),
token::Interpolated(nt) => token::Interpolated(fld.fold_interpolated(nt)),
token::SubstNt(ident, namep) => {
token::SubstNt(fld.fold_ident(ident), namep)
}
token::MatchNt(name, kind, namep, kindp) => {
token::MatchNt(fld.fold_ident(name), fld.fold_ident(kind), namep, kindp)
}
_ => t
}
}

23
src/libsyntax/parse/parser.rs
View file

@ -49,7 +49,7 @@ use ast::{StructVariantKind, BiSub};
use ast::StrStyle;
use ast::{SelfExplicit, SelfRegion, SelfStatic, SelfValue};
use ast::{Delimited, TokenTree, TraitItem, TraitRef, TtDelimited, TtSequence, TtToken};
use ast::{TtNonterminal, TupleVariantKind, Ty, Ty_, TyBot};
use ast::{TupleVariantKind, Ty, Ty_, TyBot};
use ast::{TypeField, TyFixedLengthVec, TyClosure, TyProc, TyBareFn};
use ast::{TyTypeof, TyInfer, TypeMethod};
use ast::{TyNil, TyParam, TyParamBound, TyParen, TyPath, TyPtr, TyQPath};
@ -65,6 +65,7 @@ use ast_util::{as_prec, ident_to_path, operator_prec};
use ast_util;
use codemap::{Span, BytePos, Spanned, spanned, mk_sp};
use codemap;
use ext::tt::macro_parser;
use parse;
use parse::attr::ParserAttr;
use parse::classify;
@ -73,7 +74,7 @@ use parse::common::{seq_sep_trailing_allowed};
use parse::lexer::Reader;
use parse::lexer::TokenAndSpan;
use parse::obsolete::*;
use parse::token::InternedString;
use parse::token::{MatchNt, SubstNt, InternedString};
use parse::token::{keywords, special_idents};
use parse::token;
use parse::{new_sub_parser_from_file, ParseSess};
@ -2508,7 +2509,7 @@ impl<'a> Parser<'a> {
pub fn parse_token_tree(&mut self) -> TokenTree {
// FIXME #6994: currently, this is too eager. It
// parses token trees but also identifies TtSequence's
// and TtNonterminal's; it's too early to know yet
// and token::SubstNt's; it's too early to know yet
// whether something will be a nonterminal or a seq
// yet.
maybe_whole!(deref self, NtTT);
@ -2549,9 +2550,21 @@ impl<'a> Parser<'a> {
let seq = match seq {
Spanned { node, .. } => node,
};
TtSequence(mk_sp(sp.lo, p.span.hi), Rc::new(seq), sep, repeat)
let name_num = macro_parser::count_names(seq.as_slice());
TtSequence(mk_sp(sp.lo, p.span.hi), Rc::new(seq), sep, repeat, name_num)
} else {
TtNonterminal(sp, p.parse_ident())
// A nonterminal that matches or not
let namep = match p.token { token::Ident(_, p) => p, _ => token::Plain };
let name = p.parse_ident();
if p.token == token::Colon && p.look_ahead(1, |t| t.is_ident()) {
p.bump();
let kindp = match p.token { token::Ident(_, p) => p, _ => token::Plain };
let nt_kind = p.parse_ident();
let m = TtToken(sp, MatchNt(name, nt_kind, namep, kindp));
m
} else {
TtToken(sp, SubstNt(name, namep))
}
}
}
_ => {

8
src/libsyntax/parse/token.rs
View file

@ -108,7 +108,15 @@ pub enum Token {
/* For interpolation */
Interpolated(Nonterminal),
// Can be expanded into several tokens.
/// Doc comment
DocComment(ast::Name),
// In left-hand-sides of MBE macros:
/// Parse a nonterminal (name to bind, name of NT, styles of their idents)
MatchNt(ast::Ident, ast::Ident, IdentStyle, IdentStyle),
// In right-hand-sides of MBE macros:
/// A syntactic variable that will be filled in by macro expansion.
SubstNt(ast::Ident, IdentStyle),
// Junk. These carry no data because we don't really care about the data
// they *would* carry, and don't really want to allocate a new ident for

22
src/libsyntax/print/pprust.rs
View file

@ -254,6 +254,8 @@ pub fn token_to_string(tok: &Token) -> String {
/* Other */
token::DocComment(s) => s.as_str().into_string(),
token::SubstNt(s, _) => format!("${}", s),
token::MatchNt(s, t, _, _) => format!("${}:{}", s, t),
token::Eof => "<eof>".into_string(),
token::Whitespace => " ".into_string(),
token::Comment => "/* */".into_string(),
@ -1120,13 +1122,6 @@ impl<'a> State<'a> {
/// expression arguments as expressions). It can be done! I think.
pub fn print_tt(&mut self, tt: &ast::TokenTree) -> IoResult<()> {
match *tt {
ast::TtDelimited(_, ref delimed) => {
try!(word(&mut self.s, token_to_string(&delimed.open_token()).as_slice()));
try!(space(&mut self.s));
try!(self.print_tts(delimed.tts.as_slice()));
try!(space(&mut self.s));
word(&mut self.s, token_to_string(&delimed.close_token()).as_slice())
},
ast::TtToken(_, ref tk) => {
try!(word(&mut self.s, token_to_string(tk).as_slice()));
match *tk {
@ -1136,7 +1131,14 @@ impl<'a> State<'a> {
_ => Ok(())
}
}
ast::TtSequence(_, ref tts, ref separator, kleene_op) => {
ast::TtDelimited(_, ref delimed) => {
try!(word(&mut self.s, token_to_string(&delimed.open_token()).as_slice()));
try!(space(&mut self.s));
try!(self.print_tts(delimed.tts.as_slice()));
try!(space(&mut self.s));
word(&mut self.s, token_to_string(&delimed.close_token()).as_slice())
},
ast::TtSequence(_, ref tts, ref separator, kleene_op, _) => {
try!(word(&mut self.s, "$("));
for tt_elt in (*tts).iter() {
try!(self.print_tt(tt_elt));
@ -1153,10 +1155,6 @@ impl<'a> State<'a> {
ast::OneOrMore => word(&mut self.s, "+"),
}
}
ast::TtNonterminal(_, name) => {
try!(word(&mut self.s, "$"));
self.print_ident(name)
}
}
}