1 // Copyright 2012 The Rust Project Developers. See the COPYRIGHT
2 // file at the top-level directory of this distribution and at
3 // http://rust-lang.org/COPYRIGHT.
5 // Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
6 // http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
7 // <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
8 // option. This file may not be copied, modified, or distributed
9 // except according to those terms.
11 // ignore-lexer-test FIXME #15679
13 //! The CodeMap tracks all the source code used within a single crate, mapping
14 //! from integer byte positions to the original source code location. Each bit
15 //! of source parsed during crate parsing (typically files, in-memory strings,
16 //! or various bits of macro expansion) cover a continuous range of bytes in the
17 //! CodeMap and are represented by FileMaps. Byte positions are stored in
18 //! `spans` and used pervasively in the compiler. They are absolute positions
19 //! within the CodeMap, which upon request can be converted to line and column
20 //! information, source code snippets, etc.
22 pub use self::MacroFormat::*;
24 use std::cell::RefCell;
25 use std::num::ToPrimitive;
26 use std::ops::{Add, Sub};
30 use serialize::{Encodable, Decodable, Encoder, Decoder};
33 fn from_usize(n: usize) -> Self;
34 fn to_usize(&self) -> usize;
37 /// A byte offset. Keep this small (currently 32-bits), as AST contains
39 #[derive(Clone, Copy, PartialEq, Eq, Hash, PartialOrd, Debug)]
40 pub struct BytePos(pub u32);
42 /// A character offset. Because of multibyte utf8 characters, a byte offset
43 /// is not equivalent to a character offset. The CodeMap will convert BytePos
44 /// values to CharPos values as necessary.
45 #[derive(Copy, PartialEq, Hash, PartialOrd, Debug)]
46 pub struct CharPos(pub usize);
48 // FIXME: Lots of boilerplate in these impls, but so far my attempts to fix
49 // have been unsuccessful
51 impl Pos for BytePos {
52 fn from_usize(n: usize) -> BytePos { BytePos(n as u32) }
53 fn to_usize(&self) -> usize { let BytePos(n) = *self; n as usize }
56 impl Add for BytePos {
57 type Output = BytePos;
59 fn add(self, rhs: BytePos) -> BytePos {
60 BytePos((self.to_usize() + rhs.to_usize()) as u32)
64 impl Sub for BytePos {
65 type Output = BytePos;
67 fn sub(self, rhs: BytePos) -> BytePos {
68 BytePos((self.to_usize() - rhs.to_usize()) as u32)
72 impl Pos for CharPos {
73 fn from_usize(n: usize) -> CharPos { CharPos(n) }
74 fn to_usize(&self) -> usize { let CharPos(n) = *self; n }
77 impl Add for CharPos {
78 type Output = CharPos;
80 fn add(self, rhs: CharPos) -> CharPos {
81 CharPos(self.to_usize() + rhs.to_usize())
85 impl Sub for CharPos {
86 type Output = CharPos;
88 fn sub(self, rhs: CharPos) -> CharPos {
89 CharPos(self.to_usize() - rhs.to_usize())
93 /// Spans represent a region of code, used for error reporting. Positions in spans
94 /// are *absolute* positions from the beginning of the codemap, not positions
95 /// relative to FileMaps. Methods on the CodeMap can be used to relate spans back
96 /// to the original source.
97 #[derive(Clone, Copy, Debug, Hash)]
101 /// Information about where the macro came from, if this piece of
102 /// code was created by a macro expansion.
106 pub const DUMMY_SP: Span = Span { lo: BytePos(0), hi: BytePos(0), expn_id: NO_EXPANSION };
108 // Generic span to be used for code originating from the command line
109 pub const COMMAND_LINE_SP: Span = Span { lo: BytePos(0),
111 expn_id: COMMAND_LINE_EXPN };
113 #[derive(Clone, PartialEq, Eq, RustcEncodable, RustcDecodable, Hash, Debug, Copy)]
114 pub struct Spanned<T> {
119 impl PartialEq for Span {
120 fn eq(&self, other: &Span) -> bool {
121 return (*self).lo == (*other).lo && (*self).hi == (*other).hi;
123 fn ne(&self, other: &Span) -> bool { !(*self).eq(other) }
128 impl Encodable for Span {
129 /* Note #1972 -- spans are encoded but not decoded */
130 fn encode<S: Encoder>(&self, s: &mut S) -> Result<(), S::Error> {
135 impl Decodable for Span {
136 fn decode<D: Decoder>(_d: &mut D) -> Result<Span, D::Error> {
141 pub fn spanned<T>(lo: BytePos, hi: BytePos, t: T) -> Spanned<T> {
142 respan(mk_sp(lo, hi), t)
145 pub fn respan<T>(sp: Span, t: T) -> Spanned<T> {
146 Spanned {node: t, span: sp}
149 pub fn dummy_spanned<T>(t: T) -> Spanned<T> {
153 /* assuming that we're not in macro expansion */
154 pub fn mk_sp(lo: BytePos, hi: BytePos) -> Span {
155 Span {lo: lo, hi: hi, expn_id: NO_EXPANSION}
158 /// Return the span itself if it doesn't come from a macro expansion,
159 /// otherwise return the call site span up to the `enclosing_sp` by
160 /// following the `expn_info` chain.
161 pub fn original_sp(cm: &CodeMap, sp: Span, enclosing_sp: Span) -> Span {
162 let call_site1 = cm.with_expn_info(sp.expn_id, |ei| ei.map(|ei| ei.call_site));
163 let call_site2 = cm.with_expn_info(enclosing_sp.expn_id, |ei| ei.map(|ei| ei.call_site));
164 match (call_site1, call_site2) {
166 (Some(call_site1), Some(call_site2)) if call_site1 == call_site2 => sp,
167 (Some(call_site1), _) => original_sp(cm, call_site1, enclosing_sp),
171 /// A source code location used for error reporting
173 /// Information about the original source
174 pub file: Rc<FileMap>,
175 /// The (1-based) line number
177 /// The (0-based) column offset
181 /// A source code location used as the result of lookup_char_pos_adj
182 // Actually, *none* of the clients use the filename *or* file field;
183 // perhaps they should just be removed.
184 pub struct LocWithOpt {
185 pub filename: FileName,
188 pub file: Option<Rc<FileMap>>,
191 // used to be structural records. Better names, anyone?
192 pub struct FileMapAndLine { pub fm: Rc<FileMap>, pub line: usize }
193 pub struct FileMapAndBytePos { pub fm: Rc<FileMap>, pub pos: BytePos }
195 /// The syntax with which a macro was invoked.
196 #[derive(Clone, Copy, Hash, Debug)]
197 pub enum MacroFormat {
198 /// e.g. #[derive(...)] <item>
204 #[derive(Clone, Hash, Debug)]
205 pub struct NameAndSpan {
206 /// The name of the macro that was invoked to create the thing
209 /// The format with which the macro was invoked.
210 pub format: MacroFormat,
211 /// The span of the macro definition itself. The macro may not
212 /// have a sensible definition span (e.g. something defined
213 /// completely inside libsyntax) in which case this is None.
214 pub span: Option<Span>
217 /// Extra information for tracking macro expansion of spans
218 #[derive(Hash, Debug)]
219 pub struct ExpnInfo {
220 /// The location of the actual macro invocation, e.g. `let x =
223 /// This may recursively refer to other macro invocations, e.g. if
224 /// `foo!()` invoked `bar!()` internally, and there was an
225 /// expression inside `bar!`; the call_site of the expression in
226 /// the expansion would point to the `bar!` invocation; that
227 /// call_site span would have its own ExpnInfo, with the call_site
228 /// pointing to the `foo!` invocation.
230 /// Information about the macro and its definition.
232 /// The `callee` of the inner expression in the `call_site`
233 /// example would point to the `macro_rules! bar { ... }` and that
234 /// of the `bar!()` invocation would point to the `macro_rules!
236 pub callee: NameAndSpan
239 #[derive(PartialEq, Eq, Clone, Debug, Hash, RustcEncodable, RustcDecodable, Copy)]
240 pub struct ExpnId(u32);
242 pub const NO_EXPANSION: ExpnId = ExpnId(-1);
243 // For code appearing from the command line
244 pub const COMMAND_LINE_EXPN: ExpnId = ExpnId(-2);
247 pub fn from_llvm_cookie(cookie: c_uint) -> ExpnId {
248 ExpnId(cookie as u32)
251 pub fn to_llvm_cookie(self) -> i32 {
252 let ExpnId(cookie) = self;
257 pub type FileName = String;
259 pub struct FileLines {
260 pub file: Rc<FileMap>,
261 pub lines: Vec<usize>
264 /// Identifies an offset of a multi-byte character in a FileMap
266 pub struct MultiByteChar {
267 /// The absolute offset of the character in the CodeMap
269 /// The number of bytes, >=2
273 /// A single source in the CodeMap
275 /// The name of the file that the source came from, source that doesn't
276 /// originate from files has names between angle brackets by convention,
279 /// The complete source code
281 /// The start position of this source in the CodeMap
282 pub start_pos: BytePos,
283 /// Locations of lines beginnings in the source code
284 pub lines: RefCell<Vec<BytePos> >,
285 /// Locations of multi-byte characters in the source code
286 pub multibyte_chars: RefCell<Vec<MultiByteChar> >,
290 /// EFFECT: register a start-of-line offset in the
291 /// table of line-beginnings.
292 /// UNCHECKED INVARIANT: these offsets must be added in the right
293 /// order and must be in the right places; there is shared knowledge
294 /// about what ends a line between this file and parse.rs
295 /// WARNING: pos param here is the offset relative to start of CodeMap,
296 /// and CodeMap will append a newline when adding a filemap without a newline at the end,
297 /// so the safe way to call this is with value calculated as
298 /// filemap.start_pos + newline_offset_relative_to_the_start_of_filemap.
299 pub fn next_line(&self, pos: BytePos) {
300 // the new charpos must be > the last one (or it's the first one).
301 let mut lines = self.lines.borrow_mut();
302 let line_len = lines.len();
303 assert!(line_len == 0 || ((*lines)[line_len - 1] < pos));
307 /// get a line from the list of pre-computed line-beginnings
309 pub fn get_line(&self, line_number: usize) -> Option<String> {
310 let lines = self.lines.borrow();
311 lines.get(line_number).map(|&line| {
312 let begin: BytePos = line - self.start_pos;
313 let begin = begin.to_usize();
314 let slice = &self.src[begin..];
315 match slice.find('\n') {
316 Some(e) => &slice[..e],
322 pub fn record_multibyte_char(&self, pos: BytePos, bytes: usize) {
323 assert!(bytes >=2 && bytes <= 4);
324 let mbc = MultiByteChar {
328 self.multibyte_chars.borrow_mut().push(mbc);
331 pub fn is_real_file(&self) -> bool {
332 !(self.name.starts_with("<") &&
333 self.name.ends_with(">"))
338 pub files: RefCell<Vec<Rc<FileMap>>>,
339 expansions: RefCell<Vec<ExpnInfo>>
343 pub fn new() -> CodeMap {
345 files: RefCell::new(Vec::new()),
346 expansions: RefCell::new(Vec::new()),
350 pub fn new_filemap(&self, filename: FileName, src: String) -> Rc<FileMap> {
351 let mut files = self.files.borrow_mut();
352 let start_pos = match files.last() {
354 Some(last) => last.start_pos.to_usize() + last.src.len(),
357 // Remove utf-8 BOM if any.
358 // FIXME #12884: no efficient/safe way to remove from the start of a string
359 // and reuse the allocation.
360 let mut src = if src.starts_with("\u{feff}") {
361 String::from_str(&src[3..])
363 String::from_str(&src[])
366 // Append '\n' in case it's not already there.
367 // This is a workaround to prevent CodeMap.lookup_filemap_idx from
368 // accidentally overflowing into the next filemap in case the last byte
369 // of span is also the last byte of filemap, which leads to incorrect
370 // results from CodeMap.span_to_*.
371 if src.len() > 0 && !src.ends_with("\n") {
375 let filemap = Rc::new(FileMap {
377 src: src.to_string(),
378 start_pos: Pos::from_usize(start_pos),
379 lines: RefCell::new(Vec::new()),
380 multibyte_chars: RefCell::new(Vec::new()),
383 files.push(filemap.clone());
388 pub fn mk_substr_filename(&self, sp: Span) -> String {
389 let pos = self.lookup_char_pos(sp.lo);
390 (format!("<{}:{}:{}>",
393 pos.col.to_usize() + 1)).to_string()
396 /// Lookup source information about a BytePos
397 pub fn lookup_char_pos(&self, pos: BytePos) -> Loc {
401 pub fn lookup_char_pos_adj(&self, pos: BytePos) -> LocWithOpt {
402 let loc = self.lookup_char_pos(pos);
404 filename: loc.file.name.to_string(),
411 pub fn span_to_string(&self, sp: Span) -> String {
412 if self.files.borrow().len() == 0 && sp == DUMMY_SP {
413 return "no-location".to_string();
416 let lo = self.lookup_char_pos_adj(sp.lo);
417 let hi = self.lookup_char_pos_adj(sp.hi);
418 return (format!("{}:{}:{}: {}:{}",
421 lo.col.to_usize() + 1,
423 hi.col.to_usize() + 1)).to_string()
426 pub fn span_to_filename(&self, sp: Span) -> FileName {
427 self.lookup_char_pos(sp.lo).file.name.to_string()
430 pub fn span_to_lines(&self, sp: Span) -> FileLines {
431 let lo = self.lookup_char_pos(sp.lo);
432 let hi = self.lookup_char_pos(sp.hi);
433 let mut lines = Vec::new();
434 for i in lo.line - 1..hi.line as usize {
437 FileLines {file: lo.file, lines: lines}
440 pub fn span_to_snippet(&self, sp: Span) -> Option<String> {
441 let begin = self.lookup_byte_offset(sp.lo);
442 let end = self.lookup_byte_offset(sp.hi);
444 // FIXME #8256: this used to be an assert but whatever precondition
445 // it's testing isn't true for all spans in the AST, so to allow the
446 // caller to not have to panic (and it can't catch it since the CodeMap
447 // isn't sendable), return None
448 if begin.fm.start_pos != end.fm.start_pos {
451 Some((&begin.fm.src[begin.pos.to_usize()..end.pos.to_usize()]).to_string())
455 pub fn get_filemap(&self, filename: &str) -> Rc<FileMap> {
456 for fm in &*self.files.borrow() {
457 if filename == fm.name {
461 panic!("asking for {} which we don't know about", filename);
464 pub fn lookup_byte_offset(&self, bpos: BytePos) -> FileMapAndBytePos {
465 let idx = self.lookup_filemap_idx(bpos);
466 let fm = (*self.files.borrow())[idx].clone();
467 let offset = bpos - fm.start_pos;
468 FileMapAndBytePos {fm: fm, pos: offset}
471 /// Converts an absolute BytePos to a CharPos relative to the filemap and above.
472 pub fn bytepos_to_file_charpos(&self, bpos: BytePos) -> CharPos {
473 let idx = self.lookup_filemap_idx(bpos);
474 let files = self.files.borrow();
475 let map = &(*files)[idx];
477 // The number of extra bytes due to multibyte chars in the FileMap
478 let mut total_extra_bytes = 0;
480 for mbc in &*map.multibyte_chars.borrow() {
481 debug!("{}-byte char at {:?}", mbc.bytes, mbc.pos);
483 // every character is at least one byte, so we only
484 // count the actual extra bytes.
485 total_extra_bytes += mbc.bytes - 1;
486 // We should never see a byte position in the middle of a
488 assert!(bpos.to_usize() >= mbc.pos.to_usize() + mbc.bytes);
494 assert!(map.start_pos.to_usize() + total_extra_bytes <= bpos.to_usize());
495 CharPos(bpos.to_usize() - map.start_pos.to_usize() - total_extra_bytes)
498 fn lookup_filemap_idx(&self, pos: BytePos) -> usize {
499 let files = self.files.borrow();
501 let len = files.len();
506 if files[m].start_pos > pos {
512 // There can be filemaps with length 0. These have the same start_pos as
513 // the previous filemap, but are not the filemaps we want (because they
514 // are length 0, they cannot contain what we are looking for). So,
515 // rewind until we find a useful filemap.
517 let lines = files[a].lines.borrow();
523 panic!("position {} does not resolve to a source location",
529 panic!("position {} does not resolve to a source location",
536 fn lookup_line(&self, pos: BytePos) -> FileMapAndLine {
537 let idx = self.lookup_filemap_idx(pos);
539 let files = self.files.borrow();
540 let f = (*files)[idx].clone();
543 let lines = f.lines.borrow();
544 let mut b = lines.len();
547 if (*lines)[m] > pos { b = m; } else { a = m; }
550 FileMapAndLine {fm: f, line: a}
553 fn lookup_pos(&self, pos: BytePos) -> Loc {
554 let FileMapAndLine {fm: f, line: a} = self.lookup_line(pos);
555 let line = a + 1; // Line numbers start at 1
556 let chpos = self.bytepos_to_file_charpos(pos);
557 let linebpos = (*f.lines.borrow())[a];
558 let linechpos = self.bytepos_to_file_charpos(linebpos);
559 debug!("byte pos {:?} is on the line at byte pos {:?}",
561 debug!("char pos {:?} is on the line at char pos {:?}",
563 debug!("byte is on line: {}", line);
564 assert!(chpos >= linechpos);
568 col: chpos - linechpos
572 pub fn record_expansion(&self, expn_info: ExpnInfo) -> ExpnId {
573 let mut expansions = self.expansions.borrow_mut();
574 expansions.push(expn_info);
575 ExpnId(expansions.len().to_u32().expect("too many ExpnInfo's!") - 1)
578 pub fn with_expn_info<T, F>(&self, id: ExpnId, f: F) -> T where
579 F: FnOnce(Option<&ExpnInfo>) -> T,
582 NO_EXPANSION => f(None),
583 ExpnId(i) => f(Some(&(*self.expansions.borrow())[i as usize]))
587 /// Check if a span is "internal" to a macro. This means that it is entirely generated by a
588 /// macro expansion and contains no code that was passed in as an argument.
589 pub fn span_is_internal(&self, span: Span) -> bool {
590 // first, check if the given expression was generated by a macro or not
591 // we need to go back the expn_info tree to check only the arguments
592 // of the initial macro call, not the nested ones.
593 let mut is_internal = false;
594 let mut expnid = span.expn_id;
595 while self.with_expn_info(expnid, |expninfo| {
598 // save the parent expn_id for next loop iteration
599 expnid = info.call_site.expn_id;
600 if info.callee.name == "format_args" {
601 // This is a hack because the format_args builtin calls unstable APIs.
602 // I spent like 6 hours trying to solve this more generally but am stupid.
605 } else if info.callee.span.is_none() {
606 // it's a compiler built-in, we *really* don't want to mess with it
607 // so we skip it, unless it was called by a regular macro, in which case
608 // we will handle the caller macro next turn
610 true // continue looping
612 // was this expression from the current macro arguments ?
613 is_internal = !( span.lo > info.call_site.lo &&
614 span.hi < info.call_site.hi );
615 true // continue looping
618 _ => false // stop looping
620 }) { /* empty while loop body */ }
631 let cm = CodeMap::new();
632 let fm = cm.new_filemap("blork.rs".to_string(),
633 "first line.\nsecond line".to_string());
634 fm.next_line(BytePos(0));
635 assert_eq!(fm.get_line(0), Some("first line.".to_string()));
636 // TESTING BROKEN BEHAVIOR:
637 fm.next_line(BytePos(10));
638 assert_eq!(fm.get_line(1), Some(".".to_string()));
644 let cm = CodeMap::new();
645 let fm = cm.new_filemap("blork.rs".to_string(),
646 "first line.\nsecond line".to_string());
647 // TESTING *REALLY* BROKEN BEHAVIOR:
648 fm.next_line(BytePos(0));
649 fm.next_line(BytePos(10));
650 fm.next_line(BytePos(2));
653 fn init_code_map() -> CodeMap {
654 let cm = CodeMap::new();
655 let fm1 = cm.new_filemap("blork.rs".to_string(),
656 "first line.\nsecond line".to_string());
657 let fm2 = cm.new_filemap("empty.rs".to_string(),
659 let fm3 = cm.new_filemap("blork2.rs".to_string(),
660 "first line.\nsecond line".to_string());
662 fm1.next_line(BytePos(0));
663 fm1.next_line(BytePos(12));
664 fm2.next_line(BytePos(24));
665 fm3.next_line(BytePos(24));
666 fm3.next_line(BytePos(34));
673 // Test lookup_byte_offset
674 let cm = init_code_map();
676 let fmabp1 = cm.lookup_byte_offset(BytePos(22));
677 assert_eq!(fmabp1.fm.name, "blork.rs");
678 assert_eq!(fmabp1.pos, BytePos(22));
680 let fmabp2 = cm.lookup_byte_offset(BytePos(24));
681 assert_eq!(fmabp2.fm.name, "blork2.rs");
682 assert_eq!(fmabp2.pos, BytePos(0));
687 // Test bytepos_to_file_charpos
688 let cm = init_code_map();
690 let cp1 = cm.bytepos_to_file_charpos(BytePos(22));
691 assert_eq!(cp1, CharPos(22));
693 let cp2 = cm.bytepos_to_file_charpos(BytePos(24));
694 assert_eq!(cp2, CharPos(0));
699 // Test zero-length filemaps.
700 let cm = init_code_map();
702 let loc1 = cm.lookup_char_pos(BytePos(22));
703 assert_eq!(loc1.file.name, "blork.rs");
704 assert_eq!(loc1.line, 2);
705 assert_eq!(loc1.col, CharPos(10));
707 let loc2 = cm.lookup_char_pos(BytePos(24));
708 assert_eq!(loc2.file.name, "blork2.rs");
709 assert_eq!(loc2.line, 1);
710 assert_eq!(loc2.col, CharPos(0));
713 fn init_code_map_mbc() -> CodeMap {
714 let cm = CodeMap::new();
715 // € is a three byte utf8 char.
717 cm.new_filemap("blork.rs".to_string(),
718 "fir€st €€€€ line.\nsecond line".to_string());
719 let fm2 = cm.new_filemap("blork2.rs".to_string(),
720 "first line€€.\n€ second line".to_string());
722 fm1.next_line(BytePos(0));
723 fm1.next_line(BytePos(22));
724 fm2.next_line(BytePos(40));
725 fm2.next_line(BytePos(58));
727 fm1.record_multibyte_char(BytePos(3), 3);
728 fm1.record_multibyte_char(BytePos(9), 3);
729 fm1.record_multibyte_char(BytePos(12), 3);
730 fm1.record_multibyte_char(BytePos(15), 3);
731 fm1.record_multibyte_char(BytePos(18), 3);
732 fm2.record_multibyte_char(BytePos(50), 3);
733 fm2.record_multibyte_char(BytePos(53), 3);
734 fm2.record_multibyte_char(BytePos(58), 3);
741 // Test bytepos_to_file_charpos in the presence of multi-byte chars
742 let cm = init_code_map_mbc();
744 let cp1 = cm.bytepos_to_file_charpos(BytePos(3));
745 assert_eq!(cp1, CharPos(3));
747 let cp2 = cm.bytepos_to_file_charpos(BytePos(6));
748 assert_eq!(cp2, CharPos(4));
750 let cp3 = cm.bytepos_to_file_charpos(BytePos(56));
751 assert_eq!(cp3, CharPos(12));
753 let cp4 = cm.bytepos_to_file_charpos(BytePos(61));
754 assert_eq!(cp4, CharPos(15));
759 // Test span_to_lines for a span ending at the end of filemap
760 let cm = init_code_map();
761 let span = Span {lo: BytePos(12), hi: BytePos(23), expn_id: NO_EXPANSION};
762 let file_lines = cm.span_to_lines(span);
764 assert_eq!(file_lines.file.name, "blork.rs");
765 assert_eq!(file_lines.lines.len(), 1);
766 assert_eq!(file_lines.lines[0], 1);
771 // Test span_to_snippet for a span ending at the end of filemap
772 let cm = init_code_map();
773 let span = Span {lo: BytePos(12), hi: BytePos(23), expn_id: NO_EXPANSION};
774 let snippet = cm.span_to_snippet(span);
776 assert_eq!(snippet, Some("second line".to_string()));
781 // Test span_to_str for a span ending at the end of filemap
782 let cm = init_code_map();
783 let span = Span {lo: BytePos(12), hi: BytePos(23), expn_id: NO_EXPANSION};
784 let sstr = cm.span_to_string(span);
786 assert_eq!(sstr, "blork.rs:2:1: 2:12");