rust/src/libsyntax/ext/tt/transcribe.rs

244 lines
8.5 KiB
Rust
Raw Normal View History

2012-09-04 13:37:29 -05:00
use diagnostic::span_handler;
use ast::{token_tree, tt_delim, tt_tok, tt_seq, tt_nonterminal,ident};
use macro_parser::{named_match, matched_seq, matched_nonterminal};
use codemap::span;
use parse::token::{EOF, INTERPOLATED, IDENT, Token, nt_ident, ident_interner};
use std::map::HashMap;
export tt_reader, new_tt_reader, dup_tt_reader, tt_next_token;
enum tt_frame_up { /* to break a circularity */
2012-08-20 14:23:37 -05:00
tt_frame_up(Option<tt_frame>)
}
/* FIXME #2811: figure out how to have a uniquely linked stack, and change to
`~` */
///an unzipping of `token_tree`s
type tt_frame = @{
readme: ~[ast::token_tree],
mut idx: uint,
dotdotdoted: bool,
sep: Option<Token>,
up: tt_frame_up,
};
type tt_reader = @{
sp_diag: span_handler,
interner: @ident_interner,
mut cur: tt_frame,
/* for MBE-style macro transcription */
2012-09-10 17:38:28 -05:00
interpolations: std::map::HashMap<ident, @named_match>,
mut repeat_idx: ~[uint],
mut repeat_len: ~[uint],
/* cached: */
mut cur_tok: Token,
mut cur_span: span
};
/** This can do Macro-By-Example transcription. On the other hand, if
* `src` contains no `tt_seq`s and `tt_nonterminal`s, `interp` can (and
* should) be none. */
fn new_tt_reader(sp_diag: span_handler, itr: @ident_interner,
2012-09-10 17:38:28 -05:00
interp: Option<std::map::HashMap<ident,@named_match>>,
src: ~[ast::token_tree])
-> tt_reader {
let r = @{sp_diag: sp_diag, interner: itr,
mut cur: @{readme: src, mut idx: 0u, dotdotdoted: false,
2012-08-20 14:23:37 -05:00
sep: None, up: tt_frame_up(option::None)},
2012-08-06 14:34:08 -05:00
interpolations: match interp { /* just a convienience */
2012-09-19 11:41:06 -05:00
None => std::map::HashMap(),
2012-08-20 14:23:37 -05:00
Some(x) => x
},
mut repeat_idx: ~[],
mut repeat_len: ~[],
/* dummy values, never read: */
mut cur_tok: EOF,
mut cur_span: ast_util::mk_sp(0u,0u)
};
tt_next_token(r); /* get cur_tok and cur_span set up */
2012-08-01 19:30:05 -05:00
return r;
}
pure fn dup_tt_frame(&&f: tt_frame) -> tt_frame {
@{readme: f.readme, mut idx: f.idx, dotdotdoted: f.dotdotdoted,
2012-08-06 14:34:08 -05:00
sep: f.sep, up: match f.up {
2012-08-20 14:23:37 -05:00
tt_frame_up(Some(up_frame)) => {
tt_frame_up(Some(dup_tt_frame(up_frame)))
}
2012-08-03 21:59:04 -05:00
tt_frame_up(none) => tt_frame_up(none)
}
}
}
pure fn dup_tt_reader(&&r: tt_reader) -> tt_reader {
@{sp_diag: r.sp_diag, interner: r.interner,
mut cur: dup_tt_frame(r.cur),
interpolations: r.interpolations,
mut repeat_idx: copy r.repeat_idx, mut repeat_len: copy r.repeat_len,
mut cur_tok: r.cur_tok, mut cur_span: r.cur_span}
}
pure fn lookup_cur_matched_by_matched(r: tt_reader,
start: @named_match) -> @named_match {
2012-09-28 00:20:47 -05:00
pure fn red(+ad: @named_match, idx: &uint) -> @named_match {
2012-08-06 14:34:08 -05:00
match *ad {
2012-08-03 21:59:04 -05:00
matched_nonterminal(_) => {
// end of the line; duplicate henceforth
ad
}
2012-09-28 00:20:47 -05:00
matched_seq(ads, _) => ads[*idx]
}
}
vec::foldl(start, r.repeat_idx, red)
}
fn lookup_cur_matched(r: tt_reader, name: ident) -> @named_match {
lookup_cur_matched_by_matched(r, r.interpolations.get(name))
}
enum lis {
lis_unconstrained, lis_constraint(uint, ident), lis_contradiction(~str)
}
2012-07-18 18:18:02 -05:00
fn lockstep_iter_size(t: token_tree, r: tt_reader) -> lis {
fn lis_merge(lhs: lis, rhs: lis, r: tt_reader) -> lis {
2012-08-06 14:34:08 -05:00
match lhs {
2012-08-03 21:59:04 -05:00
lis_unconstrained => rhs,
lis_contradiction(_) => lhs,
2012-08-06 14:34:08 -05:00
lis_constraint(l_len, l_id) => match rhs {
2012-08-03 21:59:04 -05:00
lis_unconstrained => lhs,
lis_contradiction(_) => rhs,
lis_constraint(r_len, _) if l_len == r_len => lhs,
lis_constraint(r_len, r_id) => {
2012-07-18 18:18:02 -05:00
let l_n = *r.interner.get(l_id);
let r_n = *r.interner.get(r_id);
2012-08-22 19:24:52 -05:00
lis_contradiction(fmt!("Inconsistent lockstep iteration: \
2012-08-03 21:59:04 -05:00
'%s' has %u items, but '%s' has %u",
2012-08-22 19:24:52 -05:00
l_n, l_len, r_n, r_len))
}
}
}
}
2012-08-06 14:34:08 -05:00
match t {
2012-08-03 21:59:04 -05:00
tt_delim(tts) | tt_seq(_, tts, _, _) => {
2012-09-28 00:20:47 -05:00
vec::foldl(lis_unconstrained, tts, |lis, tt|
lis_merge(lis, lockstep_iter_size(*tt, r), r))
}
2012-08-03 21:59:04 -05:00
tt_tok(*) => lis_unconstrained,
2012-08-06 14:34:08 -05:00
tt_nonterminal(_, name) => match *lookup_cur_matched(r, name) {
2012-08-03 21:59:04 -05:00
matched_nonterminal(_) => lis_unconstrained,
matched_seq(ads, _) => lis_constraint(ads.len(), name)
}
}
}
fn tt_next_token(&&r: tt_reader) -> {tok: Token, sp: span} {
let ret_val = { tok: r.cur_tok, sp: r.cur_span };
while r.cur.idx >= r.cur.readme.len() {
/* done with this set; pop or repeat? */
if ! r.cur.dotdotdoted
|| r.repeat_idx.last() == r.repeat_len.last() - 1 {
2012-08-06 14:34:08 -05:00
match r.cur.up {
2012-08-20 14:23:37 -05:00
tt_frame_up(None) => {
r.cur_tok = EOF;
2012-08-01 19:30:05 -05:00
return ret_val;
}
2012-08-20 14:23:37 -05:00
tt_frame_up(Some(tt_f)) => {
if r.cur.dotdotdoted {
2012-09-28 00:20:47 -05:00
r.repeat_idx.pop();
r.repeat_len.pop();
}
r.cur = tt_f;
r.cur.idx += 1u;
}
}
} else { /* repeat */
r.cur.idx = 0u;
r.repeat_idx[r.repeat_idx.len() - 1u] += 1u;
2012-08-06 14:34:08 -05:00
match r.cur.sep {
2012-08-20 14:23:37 -05:00
Some(tk) => {
r.cur_tok = tk; /* repeat same span, I guess */
2012-08-01 19:30:05 -05:00
return ret_val;
}
2012-08-20 14:23:37 -05:00
None => ()
}
}
}
loop { /* because it's easiest, this handles `tt_delim` not starting
with a `tt_tok`, even though it won't happen */
2012-08-06 14:34:08 -05:00
match r.cur.readme[r.cur.idx] {
2012-08-03 21:59:04 -05:00
tt_delim(tts) => {
r.cur = @{readme: tts, mut idx: 0u, dotdotdoted: false,
2012-08-20 14:23:37 -05:00
sep: None, up: tt_frame_up(option::Some(r.cur)) };
// if this could be 0-length, we'd need to potentially recur here
}
2012-08-03 21:59:04 -05:00
tt_tok(sp, tok) => {
r.cur_span = sp; r.cur_tok = tok;
r.cur.idx += 1u;
2012-08-01 19:30:05 -05:00
return ret_val;
}
2012-08-03 21:59:04 -05:00
tt_seq(sp, tts, sep, zerok) => {
2012-08-06 14:34:08 -05:00
match lockstep_iter_size(tt_seq(sp, tts, sep, zerok), r) {
2012-08-03 21:59:04 -05:00
lis_unconstrained => {
r.sp_diag.span_fatal(
sp, /* blame macro writer */
~"attempted to repeat an expression containing no syntax \
variables matched as repeating at this depth");
}
2012-08-03 21:59:04 -05:00
lis_contradiction(msg) => { /* FIXME #2887 blame macro invoker
instead*/
r.sp_diag.span_fatal(sp, msg);
}
2012-08-03 21:59:04 -05:00
lis_constraint(len, _) => {
if len == 0 {
if !zerok {
r.sp_diag.span_fatal(sp, /* FIXME #2887 blame invoker
*/
~"this must repeat at least \
once");
}
r.cur.idx += 1u;
2012-08-01 19:30:05 -05:00
return tt_next_token(r);
} else {
r.repeat_len.push(len);
r.repeat_idx.push(0u);
r.cur = @{readme: tts, mut idx: 0u, dotdotdoted: true,
2012-08-20 14:23:37 -05:00
sep: sep, up: tt_frame_up(option::Some(r.cur))};
}
}
}
}
// FIXME #2887: think about span stuff here
2012-08-03 21:59:04 -05:00
tt_nonterminal(sp, ident) => {
2012-08-06 14:34:08 -05:00
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. */
2012-08-03 21:59:04 -05:00
matched_nonterminal(nt_ident(sn,b)) => {
r.cur_span = sp; r.cur_tok = IDENT(sn,b);
r.cur.idx += 1u;
2012-08-01 19:30:05 -05:00
return ret_val;
}
2012-08-03 21:59:04 -05:00
matched_nonterminal(other_whole_nt) => {
r.cur_span = sp; r.cur_tok = INTERPOLATED(other_whole_nt);
r.cur.idx += 1u;
2012-08-01 19:30:05 -05:00
return ret_val;
}
2012-08-03 21:59:04 -05:00
matched_seq(*) => {
r.sp_diag.span_fatal(
copy r.cur_span, /* blame the macro writer */
2012-08-22 19:24:52 -05:00
fmt!("variable '%s' is still repeating at this depth",
*r.interner.get(ident)));
}
}
}
}
}
}