1 //! The source positions and related helper functions.
5 //! This API is completely unstable and subject to change.
7 #![doc(html_root_url = "https://doc.rust-lang.org/nightly/nightly-rustc/")]
8 #![feature(array_windows)]
9 #![feature(crate_visibility_modifier)]
11 #![feature(const_panic)]
12 #![feature(negative_impls)]
14 #![feature(min_specialization)]
15 #![feature(option_expect_none)]
18 extern crate rustc_macros;
20 use rustc_data_structures::AtomicRef;
21 use rustc_macros::HashStable_Generic;
22 use rustc_serialize::{Decodable, Decoder, Encodable, Encoder};
24 mod caching_source_map_view;
26 pub use self::caching_source_map_view::CachingSourceMapView;
27 use source_map::SourceMap;
32 pub use hygiene::SyntaxContext;
33 use hygiene::Transparency;
34 pub use hygiene::{DesugaringKind, ExpnData, ExpnId, ExpnKind, ForLoopLoc, MacroKind};
36 use def_id::{CrateNum, DefId, LOCAL_CRATE};
38 pub use span_encoding::{Span, DUMMY_SP};
41 pub use symbol::{sym, Symbol};
43 mod analyze_source_file;
46 use rustc_data_structures::fingerprint::Fingerprint;
47 use rustc_data_structures::stable_hasher::{HashStable, StableHasher};
48 use rustc_data_structures::sync::{Lock, Lrc};
51 use std::cell::RefCell;
52 use std::cmp::{self, Ordering};
55 use std::ops::{Add, Range, Sub};
56 use std::path::{Path, PathBuf};
57 use std::str::FromStr;
69 // Per-session global variables: this struct is stored in thread-local storage
70 // in such a way that it is accessible without any kind of handle to all
71 // threads within the compilation session, but is not accessible outside the
73 pub struct SessionGlobals {
74 symbol_interner: Lock<symbol::Interner>,
75 span_interner: Lock<span_encoding::SpanInterner>,
76 hygiene_data: Lock<hygiene::HygieneData>,
77 source_map: Lock<Option<Lrc<SourceMap>>>,
81 pub fn new(edition: Edition) -> SessionGlobals {
83 symbol_interner: Lock::new(symbol::Interner::fresh()),
84 span_interner: Lock::new(span_encoding::SpanInterner::default()),
85 hygiene_data: Lock::new(hygiene::HygieneData::new(edition)),
86 source_map: Lock::new(None),
91 pub fn with_session_globals<R>(edition: Edition, f: impl FnOnce() -> R) -> R {
92 let session_globals = SessionGlobals::new(edition);
93 SESSION_GLOBALS.set(&session_globals, f)
96 pub fn with_default_session_globals<R>(f: impl FnOnce() -> R) -> R {
97 with_session_globals(edition::DEFAULT_EDITION, f)
100 // If this ever becomes non thread-local, `decode_syntax_context`
101 // and `decode_expn_id` will need to be updated to handle concurrent
103 scoped_tls::scoped_thread_local!(pub static SESSION_GLOBALS: SessionGlobals);
105 // FIXME: Perhaps this should not implement Rustc{Decodable, Encodable}
107 // FIXME: We should use this enum or something like it to get rid of the
108 // use of magic `/rust/1.x/...` paths across the board.
109 #[derive(Debug, Eq, PartialEq, Clone, Ord, PartialOrd, Hash)]
110 #[derive(HashStable_Generic, Decodable, Encodable)]
111 pub enum RealFileName {
113 /// For de-virtualized paths (namely paths into libstd that have been mapped
114 /// to the appropriate spot on the local host's file system),
116 /// `local_path` is the (host-dependent) local path to the file.
118 /// `virtual_name` is the stable path rustc will store internally within
120 virtual_name: PathBuf,
125 /// Returns the path suitable for reading from the file system on the local host.
126 /// Avoid embedding this in build artifacts; see `stable_name` for that.
127 pub fn local_path(&self) -> &Path {
129 RealFileName::Named(p)
130 | RealFileName::Devirtualized { local_path: p, virtual_name: _ } => &p,
134 /// Returns the path suitable for reading from the file system on the local host.
135 /// Avoid embedding this in build artifacts; see `stable_name` for that.
136 pub fn into_local_path(self) -> PathBuf {
138 RealFileName::Named(p)
139 | RealFileName::Devirtualized { local_path: p, virtual_name: _ } => p,
143 /// Returns the path suitable for embedding into build artifacts. Note that
144 /// a virtualized path will not correspond to a valid file system path; see
145 /// `local_path` for something that is more likely to return paths into the
146 /// local host file system.
147 pub fn stable_name(&self) -> &Path {
149 RealFileName::Named(p)
150 | RealFileName::Devirtualized { local_path: _, virtual_name: p } => &p,
155 /// Differentiates between real files and common virtual files.
156 #[derive(Debug, Eq, PartialEq, Clone, Ord, PartialOrd, Hash)]
157 #[derive(HashStable_Generic, Decodable, Encodable)]
160 /// Call to `quote!`.
164 /// Hack in `src/librustc_ast/parse.rs`.
167 ProcMacroSourceCode(u64),
168 /// Strings provided as `--cfg [cfgspec]` stored in a `crate_cfg`.
170 /// Strings provided as crate attributes in the CLI.
172 /// Custom sources for explicit parser calls from plugins and drivers.
174 DocTest(PathBuf, isize),
175 /// Post-substitution inline assembly from LLVM
179 impl std::fmt::Display for FileName {
180 fn fmt(&self, fmt: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
183 Real(RealFileName::Named(ref path)) => write!(fmt, "{}", path.display()),
184 // FIXME: might be nice to display both compoments of Devirtualized.
185 // But for now (to backport fix for issue #70924), best to not
186 // perturb diagnostics so its obvious test suite still works.
187 Real(RealFileName::Devirtualized { ref local_path, virtual_name: _ }) => {
188 write!(fmt, "{}", local_path.display())
190 QuoteExpansion(_) => write!(fmt, "<quote expansion>"),
191 MacroExpansion(_) => write!(fmt, "<macro expansion>"),
192 Anon(_) => write!(fmt, "<anon>"),
193 ProcMacroSourceCode(_) => write!(fmt, "<proc-macro source code>"),
194 CfgSpec(_) => write!(fmt, "<cfgspec>"),
195 CliCrateAttr(_) => write!(fmt, "<crate attribute>"),
196 Custom(ref s) => write!(fmt, "<{}>", s),
197 DocTest(ref path, _) => write!(fmt, "{}", path.display()),
198 InlineAsm(_) => write!(fmt, "<inline asm>"),
203 impl From<PathBuf> for FileName {
204 fn from(p: PathBuf) -> Self {
205 assert!(!p.to_string_lossy().ends_with('>'));
206 FileName::Real(RealFileName::Named(p))
211 pub fn is_real(&self) -> bool {
217 | ProcMacroSourceCode(_)
223 | InlineAsm(_) => false,
227 pub fn macro_expansion_source_code(src: &str) -> FileName {
228 let mut hasher = StableHasher::new();
229 src.hash(&mut hasher);
230 FileName::MacroExpansion(hasher.finish())
233 pub fn anon_source_code(src: &str) -> FileName {
234 let mut hasher = StableHasher::new();
235 src.hash(&mut hasher);
236 FileName::Anon(hasher.finish())
239 pub fn proc_macro_source_code(src: &str) -> FileName {
240 let mut hasher = StableHasher::new();
241 src.hash(&mut hasher);
242 FileName::ProcMacroSourceCode(hasher.finish())
245 pub fn cfg_spec_source_code(src: &str) -> FileName {
246 let mut hasher = StableHasher::new();
247 src.hash(&mut hasher);
248 FileName::QuoteExpansion(hasher.finish())
251 pub fn cli_crate_attr_source_code(src: &str) -> FileName {
252 let mut hasher = StableHasher::new();
253 src.hash(&mut hasher);
254 FileName::CliCrateAttr(hasher.finish())
257 pub fn doc_test_source_code(path: PathBuf, line: isize) -> FileName {
258 FileName::DocTest(path, line)
261 pub fn inline_asm_source_code(src: &str) -> FileName {
262 let mut hasher = StableHasher::new();
263 src.hash(&mut hasher);
264 FileName::InlineAsm(hasher.finish())
268 /// Spans represent a region of code, used for error reporting. Positions in spans
269 /// are *absolute* positions from the beginning of the source_map, not positions
270 /// relative to `SourceFile`s. Methods on the `SourceMap` can be used to relate spans back
271 /// to the original source.
272 /// You must be careful if the span crosses more than one file - you will not be
273 /// able to use many of the functions on spans in source_map and you cannot assume
274 /// that the length of the `span = hi - lo`; there may be space in the `BytePos`
275 /// range between files.
277 /// `SpanData` is public because `Span` uses a thread-local interner and can't be
278 /// sent to other threads, but some pieces of performance infra run in a separate thread.
279 /// Using `Span` is generally preferred.
280 #[derive(Clone, Copy, Hash, PartialEq, Eq, Ord, PartialOrd)]
281 pub struct SpanData {
284 /// Information about where the macro came from, if this piece of
285 /// code was created by a macro expansion.
286 pub ctxt: SyntaxContext,
291 pub fn with_lo(&self, lo: BytePos) -> Span {
292 Span::new(lo, self.hi, self.ctxt)
295 pub fn with_hi(&self, hi: BytePos) -> Span {
296 Span::new(self.lo, hi, self.ctxt)
299 pub fn with_ctxt(&self, ctxt: SyntaxContext) -> Span {
300 Span::new(self.lo, self.hi, ctxt)
304 // The interner is pointed to by a thread local value which is only set on the main thread
305 // with parallelization is disabled. So we don't allow `Span` to transfer between threads
306 // to avoid panics and other errors, even though it would be memory safe to do so.
307 #[cfg(not(parallel_compiler))]
308 impl !Send for Span {}
309 #[cfg(not(parallel_compiler))]
310 impl !Sync for Span {}
312 impl PartialOrd for Span {
313 fn partial_cmp(&self, rhs: &Self) -> Option<Ordering> {
314 PartialOrd::partial_cmp(&self.data(), &rhs.data())
318 fn cmp(&self, rhs: &Self) -> Ordering {
319 Ord::cmp(&self.data(), &rhs.data())
323 /// A collection of `Span`s.
325 /// Spans have two orthogonal attributes:
327 /// - They can be *primary spans*. In this case they are the locus of
328 /// the error, and would be rendered with `^^^`.
329 /// - They can have a *label*. In this case, the label is written next
330 /// to the mark in the snippet when we render.
331 #[derive(Clone, Debug, Hash, PartialEq, Eq, Encodable, Decodable)]
332 pub struct MultiSpan {
333 primary_spans: Vec<Span>,
334 span_labels: Vec<(Span, String)>,
339 pub fn lo(self) -> BytePos {
343 pub fn with_lo(self, lo: BytePos) -> Span {
344 self.data().with_lo(lo)
347 pub fn hi(self) -> BytePos {
351 pub fn with_hi(self, hi: BytePos) -> Span {
352 self.data().with_hi(hi)
355 pub fn ctxt(self) -> SyntaxContext {
359 pub fn with_ctxt(self, ctxt: SyntaxContext) -> Span {
360 self.data().with_ctxt(ctxt)
363 /// Returns `true` if this is a dummy span with any hygienic context.
365 pub fn is_dummy(self) -> bool {
366 let span = self.data();
367 span.lo.0 == 0 && span.hi.0 == 0
370 /// Returns `true` if this span comes from a macro or desugaring.
372 pub fn from_expansion(self) -> bool {
373 self.ctxt() != SyntaxContext::root()
376 /// Returns `true` if `span` originates in a derive-macro's expansion.
377 pub fn in_derive_expansion(self) -> bool {
378 matches!(self.ctxt().outer_expn_data().kind, ExpnKind::Macro(MacroKind::Derive, _))
382 pub fn with_root_ctxt(lo: BytePos, hi: BytePos) -> Span {
383 Span::new(lo, hi, SyntaxContext::root())
386 /// Returns a new span representing an empty span at the beginning of this span
388 pub fn shrink_to_lo(self) -> Span {
389 let span = self.data();
390 span.with_hi(span.lo)
392 /// Returns a new span representing an empty span at the end of this span.
394 pub fn shrink_to_hi(self) -> Span {
395 let span = self.data();
396 span.with_lo(span.hi)
400 /// Returns true if hi == lo
401 pub fn is_empty(&self) -> bool {
402 let span = self.data();
406 /// Returns `self` if `self` is not the dummy span, and `other` otherwise.
407 pub fn substitute_dummy(self, other: Span) -> Span {
408 if self.is_dummy() { other } else { self }
411 /// Returns `true` if `self` fully encloses `other`.
412 pub fn contains(self, other: Span) -> bool {
413 let span = self.data();
414 let other = other.data();
415 span.lo <= other.lo && other.hi <= span.hi
418 /// Returns `true` if `self` touches `other`.
419 pub fn overlaps(self, other: Span) -> bool {
420 let span = self.data();
421 let other = other.data();
422 span.lo < other.hi && other.lo < span.hi
425 /// Returns `true` if the spans are equal with regards to the source text.
427 /// Use this instead of `==` when either span could be generated code,
428 /// and you only care that they point to the same bytes of source text.
429 pub fn source_equal(&self, other: &Span) -> bool {
430 let span = self.data();
431 let other = other.data();
432 span.lo == other.lo && span.hi == other.hi
435 /// Returns `Some(span)`, where the start is trimmed by the end of `other`.
436 pub fn trim_start(self, other: Span) -> Option<Span> {
437 let span = self.data();
438 let other = other.data();
439 if span.hi > other.hi { Some(span.with_lo(cmp::max(span.lo, other.hi))) } else { None }
442 /// Returns the source span -- this is either the supplied span, or the span for
443 /// the macro callsite that expanded to it.
444 pub fn source_callsite(self) -> Span {
445 let expn_data = self.ctxt().outer_expn_data();
446 if !expn_data.is_root() { expn_data.call_site.source_callsite() } else { self }
449 /// The `Span` for the tokens in the previous macro expansion from which `self` was generated,
451 pub fn parent(self) -> Option<Span> {
452 let expn_data = self.ctxt().outer_expn_data();
453 if !expn_data.is_root() { Some(expn_data.call_site) } else { None }
456 /// Edition of the crate from which this span came.
457 pub fn edition(self) -> edition::Edition {
458 self.ctxt().outer_expn_data().edition
462 pub fn rust_2015(&self) -> bool {
463 self.edition() == edition::Edition::Edition2015
467 pub fn rust_2018(&self) -> bool {
468 self.edition() >= edition::Edition::Edition2018
471 /// Returns the source callee.
473 /// Returns `None` if the supplied span has no expansion trace,
474 /// else returns the `ExpnData` for the macro definition
475 /// corresponding to the source callsite.
476 pub fn source_callee(self) -> Option<ExpnData> {
477 fn source_callee(expn_data: ExpnData) -> ExpnData {
478 let next_expn_data = expn_data.call_site.ctxt().outer_expn_data();
479 if !next_expn_data.is_root() { source_callee(next_expn_data) } else { expn_data }
481 let expn_data = self.ctxt().outer_expn_data();
482 if !expn_data.is_root() { Some(source_callee(expn_data)) } else { None }
485 /// Checks if a span is "internal" to a macro in which `#[unstable]`
486 /// items can be used (that is, a macro marked with
487 /// `#[allow_internal_unstable]`).
488 pub fn allows_unstable(&self, feature: Symbol) -> bool {
489 self.ctxt().outer_expn_data().allow_internal_unstable.map_or(false, |features| {
492 .any(|&f| f == feature || f == sym::allow_internal_unstable_backcompat_hack)
496 /// Checks if this span arises from a compiler desugaring of kind `kind`.
497 pub fn is_desugaring(&self, kind: DesugaringKind) -> bool {
498 match self.ctxt().outer_expn_data().kind {
499 ExpnKind::Desugaring(k) => k == kind,
504 /// Returns the compiler desugaring that created this span, or `None`
505 /// if this span is not from a desugaring.
506 pub fn desugaring_kind(&self) -> Option<DesugaringKind> {
507 match self.ctxt().outer_expn_data().kind {
508 ExpnKind::Desugaring(k) => Some(k),
513 /// Checks if a span is "internal" to a macro in which `unsafe`
514 /// can be used without triggering the `unsafe_code` lint
515 // (that is, a macro marked with `#[allow_internal_unsafe]`).
516 pub fn allows_unsafe(&self) -> bool {
517 self.ctxt().outer_expn_data().allow_internal_unsafe
520 pub fn macro_backtrace(mut self) -> impl Iterator<Item = ExpnData> {
521 let mut prev_span = DUMMY_SP;
522 std::iter::from_fn(move || {
524 let expn_data = self.ctxt().outer_expn_data();
525 if expn_data.is_root() {
529 let is_recursive = expn_data.call_site.source_equal(&prev_span);
532 self = expn_data.call_site;
534 // Don't print recursive invocations.
536 return Some(expn_data);
542 /// Returns a `Span` that would enclose both `self` and `end`.
546 /// self lorem ipsum end
547 /// ^^^^^^^^^^^^^^^^^^^^
549 pub fn to(self, end: Span) -> Span {
550 let span_data = self.data();
551 let end_data = end.data();
552 // FIXME(jseyfried): `self.ctxt` should always equal `end.ctxt` here (cf. issue #23480).
553 // Return the macro span on its own to avoid weird diagnostic output. It is preferable to
554 // have an incomplete span than a completely nonsensical one.
555 if span_data.ctxt != end_data.ctxt {
556 if span_data.ctxt == SyntaxContext::root() {
558 } else if end_data.ctxt == SyntaxContext::root() {
561 // Both spans fall within a macro.
562 // FIXME(estebank): check if it is the *same* macro.
565 cmp::min(span_data.lo, end_data.lo),
566 cmp::max(span_data.hi, end_data.hi),
567 if span_data.ctxt == SyntaxContext::root() { end_data.ctxt } else { span_data.ctxt },
571 /// Returns a `Span` between the end of `self` to the beginning of `end`.
575 /// self lorem ipsum end
578 pub fn between(self, end: Span) -> Span {
579 let span = self.data();
580 let end = end.data();
584 if end.ctxt == SyntaxContext::root() { end.ctxt } else { span.ctxt },
588 /// Returns a `Span` from the beginning of `self` until the beginning of `end`.
592 /// self lorem ipsum end
593 /// ^^^^^^^^^^^^^^^^^
595 pub fn until(self, end: Span) -> Span {
596 let span = self.data();
597 let end = end.data();
601 if end.ctxt == SyntaxContext::root() { end.ctxt } else { span.ctxt },
605 pub fn from_inner(self, inner: InnerSpan) -> Span {
606 let span = self.data();
608 span.lo + BytePos::from_usize(inner.start),
609 span.lo + BytePos::from_usize(inner.end),
614 /// Equivalent of `Span::def_site` from the proc macro API,
615 /// except that the location is taken from the `self` span.
616 pub fn with_def_site_ctxt(self, expn_id: ExpnId) -> Span {
617 self.with_ctxt_from_mark(expn_id, Transparency::Opaque)
620 /// Equivalent of `Span::call_site` from the proc macro API,
621 /// except that the location is taken from the `self` span.
622 pub fn with_call_site_ctxt(&self, expn_id: ExpnId) -> Span {
623 self.with_ctxt_from_mark(expn_id, Transparency::Transparent)
626 /// Equivalent of `Span::mixed_site` from the proc macro API,
627 /// except that the location is taken from the `self` span.
628 pub fn with_mixed_site_ctxt(&self, expn_id: ExpnId) -> Span {
629 self.with_ctxt_from_mark(expn_id, Transparency::SemiTransparent)
632 /// Produces a span with the same location as `self` and context produced by a macro with the
633 /// given ID and transparency, assuming that macro was defined directly and not produced by
634 /// some other macro (which is the case for built-in and procedural macros).
635 pub fn with_ctxt_from_mark(self, expn_id: ExpnId, transparency: Transparency) -> Span {
636 self.with_ctxt(SyntaxContext::root().apply_mark(expn_id, transparency))
640 pub fn apply_mark(self, expn_id: ExpnId, transparency: Transparency) -> Span {
641 let span = self.data();
642 span.with_ctxt(span.ctxt.apply_mark(expn_id, transparency))
646 pub fn remove_mark(&mut self) -> ExpnId {
647 let mut span = self.data();
648 let mark = span.ctxt.remove_mark();
649 *self = Span::new(span.lo, span.hi, span.ctxt);
654 pub fn adjust(&mut self, expn_id: ExpnId) -> Option<ExpnId> {
655 let mut span = self.data();
656 let mark = span.ctxt.adjust(expn_id);
657 *self = Span::new(span.lo, span.hi, span.ctxt);
662 pub fn normalize_to_macros_2_0_and_adjust(&mut self, expn_id: ExpnId) -> Option<ExpnId> {
663 let mut span = self.data();
664 let mark = span.ctxt.normalize_to_macros_2_0_and_adjust(expn_id);
665 *self = Span::new(span.lo, span.hi, span.ctxt);
670 pub fn glob_adjust(&mut self, expn_id: ExpnId, glob_span: Span) -> Option<Option<ExpnId>> {
671 let mut span = self.data();
672 let mark = span.ctxt.glob_adjust(expn_id, glob_span);
673 *self = Span::new(span.lo, span.hi, span.ctxt);
678 pub fn reverse_glob_adjust(
682 ) -> Option<Option<ExpnId>> {
683 let mut span = self.data();
684 let mark = span.ctxt.reverse_glob_adjust(expn_id, glob_span);
685 *self = Span::new(span.lo, span.hi, span.ctxt);
690 pub fn normalize_to_macros_2_0(self) -> Span {
691 let span = self.data();
692 span.with_ctxt(span.ctxt.normalize_to_macros_2_0())
696 pub fn normalize_to_macro_rules(self) -> Span {
697 let span = self.data();
698 span.with_ctxt(span.ctxt.normalize_to_macro_rules())
702 #[derive(Clone, Debug)]
703 pub struct SpanLabel {
704 /// The span we are going to include in the final snippet.
707 /// Is this a primary span? This is the "locus" of the message,
708 /// and is indicated with a `^^^^` underline, versus `----`.
709 pub is_primary: bool,
711 /// What label should we attach to this span (if any)?
712 pub label: Option<String>,
715 impl Default for Span {
716 fn default() -> Self {
721 impl<E: Encoder> Encodable<E> for Span {
722 default fn encode(&self, s: &mut E) -> Result<(), E::Error> {
723 let span = self.data();
724 s.emit_struct("Span", 2, |s| {
725 s.emit_struct_field("lo", 0, |s| span.lo.encode(s))?;
726 s.emit_struct_field("hi", 1, |s| span.hi.encode(s))
730 impl<D: Decoder> Decodable<D> for Span {
731 default fn decode(s: &mut D) -> Result<Span, D::Error> {
732 s.read_struct("Span", 2, |d| {
733 let lo = d.read_struct_field("lo", 0, Decodable::decode)?;
734 let hi = d.read_struct_field("hi", 1, Decodable::decode)?;
736 Ok(Span::new(lo, hi, SyntaxContext::root()))
741 /// Calls the provided closure, using the provided `SourceMap` to format
742 /// any spans that are debug-printed during the closure's execution.
744 /// Normally, the global `TyCtxt` is used to retrieve the `SourceMap`
745 /// (see `rustc_interface::callbacks::span_debug1). However, some parts
746 /// of the compiler (e.g. `rustc_parse`) may debug-print `Span`s before
747 /// a `TyCtxt` is available. In this case, we fall back to
748 /// the `SourceMap` provided to this function. If that is not available,
749 /// we fall back to printing the raw `Span` field values.
750 pub fn with_source_map<T, F: FnOnce() -> T>(source_map: Lrc<SourceMap>, f: F) -> T {
751 SESSION_GLOBALS.with(|session_globals| {
752 *session_globals.source_map.borrow_mut() = Some(source_map);
754 struct ClearSourceMap;
755 impl Drop for ClearSourceMap {
757 SESSION_GLOBALS.with(|session_globals| {
758 session_globals.source_map.borrow_mut().take();
763 let _guard = ClearSourceMap;
767 pub fn debug_with_source_map(
769 f: &mut fmt::Formatter<'_>,
770 source_map: &SourceMap,
772 write!(f, "{} ({:?})", source_map.span_to_string(span), span.ctxt())
775 pub fn default_span_debug(span: Span, f: &mut fmt::Formatter<'_>) -> fmt::Result {
776 SESSION_GLOBALS.with(|session_globals| {
777 if let Some(source_map) = &*session_globals.source_map.borrow() {
778 debug_with_source_map(span, f, source_map)
780 f.debug_struct("Span")
781 .field("lo", &span.lo())
782 .field("hi", &span.hi())
783 .field("ctxt", &span.ctxt())
789 impl fmt::Debug for Span {
790 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
791 (*SPAN_DEBUG)(*self, f)
795 impl fmt::Debug for SpanData {
796 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
797 (*SPAN_DEBUG)(Span::new(self.lo, self.hi, self.ctxt), f)
803 pub fn new() -> MultiSpan {
804 MultiSpan { primary_spans: vec![], span_labels: vec![] }
807 pub fn from_span(primary_span: Span) -> MultiSpan {
808 MultiSpan { primary_spans: vec![primary_span], span_labels: vec![] }
811 pub fn from_spans(mut vec: Vec<Span>) -> MultiSpan {
813 MultiSpan { primary_spans: vec, span_labels: vec![] }
816 pub fn push_span_label(&mut self, span: Span, label: String) {
817 self.span_labels.push((span, label));
820 /// Selects the first primary span (if any).
821 pub fn primary_span(&self) -> Option<Span> {
822 self.primary_spans.first().cloned()
825 /// Returns all primary spans.
826 pub fn primary_spans(&self) -> &[Span] {
830 /// Returns `true` if any of the primary spans are displayable.
831 pub fn has_primary_spans(&self) -> bool {
832 self.primary_spans.iter().any(|sp| !sp.is_dummy())
835 /// Returns `true` if this contains only a dummy primary span with any hygienic context.
836 pub fn is_dummy(&self) -> bool {
837 let mut is_dummy = true;
838 for span in &self.primary_spans {
839 if !span.is_dummy() {
846 /// Replaces all occurrences of one Span with another. Used to move `Span`s in areas that don't
847 /// display well (like std macros). Returns whether replacements occurred.
848 pub fn replace(&mut self, before: Span, after: Span) -> bool {
849 let mut replacements_occurred = false;
850 for primary_span in &mut self.primary_spans {
851 if *primary_span == before {
852 *primary_span = after;
853 replacements_occurred = true;
856 for span_label in &mut self.span_labels {
857 if span_label.0 == before {
858 span_label.0 = after;
859 replacements_occurred = true;
862 replacements_occurred
865 /// Returns the strings to highlight. We always ensure that there
866 /// is an entry for each of the primary spans -- for each primary
867 /// span `P`, if there is at least one label with span `P`, we return
868 /// those labels (marked as primary). But otherwise we return
869 /// `SpanLabel` instances with empty labels.
870 pub fn span_labels(&self) -> Vec<SpanLabel> {
871 let is_primary = |span| self.primary_spans.contains(&span);
873 let mut span_labels = self
876 .map(|&(span, ref label)| SpanLabel {
878 is_primary: is_primary(span),
879 label: Some(label.clone()),
881 .collect::<Vec<_>>();
883 for &span in &self.primary_spans {
884 if !span_labels.iter().any(|sl| sl.span == span) {
885 span_labels.push(SpanLabel { span, is_primary: true, label: None });
892 /// Returns `true` if any of the span labels is displayable.
893 pub fn has_span_labels(&self) -> bool {
894 self.span_labels.iter().any(|(sp, _)| !sp.is_dummy())
898 impl From<Span> for MultiSpan {
899 fn from(span: Span) -> MultiSpan {
900 MultiSpan::from_span(span)
904 impl From<Vec<Span>> for MultiSpan {
905 fn from(spans: Vec<Span>) -> MultiSpan {
906 MultiSpan::from_spans(spans)
910 /// Identifies an offset of a multi-byte character in a `SourceFile`.
911 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
912 pub struct MultiByteChar {
913 /// The absolute offset of the character in the `SourceMap`.
915 /// The number of bytes, `>= 2`.
919 /// Identifies an offset of a non-narrow character in a `SourceFile`.
920 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
921 pub enum NonNarrowChar {
922 /// Represents a zero-width character.
924 /// Represents a wide (full-width) character.
926 /// Represents a tab character, represented visually with a width of 4 characters.
931 fn new(pos: BytePos, width: usize) -> Self {
933 0 => NonNarrowChar::ZeroWidth(pos),
934 2 => NonNarrowChar::Wide(pos),
935 4 => NonNarrowChar::Tab(pos),
936 _ => panic!("width {} given for non-narrow character", width),
940 /// Returns the absolute offset of the character in the `SourceMap`.
941 pub fn pos(&self) -> BytePos {
943 NonNarrowChar::ZeroWidth(p) | NonNarrowChar::Wide(p) | NonNarrowChar::Tab(p) => p,
947 /// Returns the width of the character, 0 (zero-width) or 2 (wide).
948 pub fn width(&self) -> usize {
950 NonNarrowChar::ZeroWidth(_) => 0,
951 NonNarrowChar::Wide(_) => 2,
952 NonNarrowChar::Tab(_) => 4,
957 impl Add<BytePos> for NonNarrowChar {
960 fn add(self, rhs: BytePos) -> Self {
962 NonNarrowChar::ZeroWidth(pos) => NonNarrowChar::ZeroWidth(pos + rhs),
963 NonNarrowChar::Wide(pos) => NonNarrowChar::Wide(pos + rhs),
964 NonNarrowChar::Tab(pos) => NonNarrowChar::Tab(pos + rhs),
969 impl Sub<BytePos> for NonNarrowChar {
972 fn sub(self, rhs: BytePos) -> Self {
974 NonNarrowChar::ZeroWidth(pos) => NonNarrowChar::ZeroWidth(pos - rhs),
975 NonNarrowChar::Wide(pos) => NonNarrowChar::Wide(pos - rhs),
976 NonNarrowChar::Tab(pos) => NonNarrowChar::Tab(pos - rhs),
981 /// Identifies an offset of a character that was normalized away from `SourceFile`.
982 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
983 pub struct NormalizedPos {
984 /// The absolute offset of the character in the `SourceMap`.
986 /// The difference between original and normalized string at position.
990 #[derive(PartialEq, Eq, Clone, Debug)]
991 pub enum ExternalSource {
992 /// No external source has to be loaded, since the `SourceFile` represents a local crate.
995 kind: ExternalSourceKind,
996 /// This SourceFile's byte-offset within the source_map of its original crate
997 original_start_pos: BytePos,
998 /// The end of this SourceFile within the source_map of its original crate
999 original_end_pos: BytePos,
1003 /// The state of the lazy external source loading mechanism of a `SourceFile`.
1004 #[derive(PartialEq, Eq, Clone, Debug)]
1005 pub enum ExternalSourceKind {
1006 /// The external source has been loaded already.
1007 Present(Lrc<String>),
1008 /// No attempt has been made to load the external source.
1010 /// A failed attempt has been made to load the external source.
1015 impl ExternalSource {
1016 pub fn is_absent(&self) -> bool {
1018 ExternalSource::Foreign { kind: ExternalSourceKind::Present(_), .. } => false,
1023 pub fn get_source(&self) -> Option<&Lrc<String>> {
1025 ExternalSource::Foreign { kind: ExternalSourceKind::Present(ref src), .. } => Some(src),
1032 pub struct OffsetOverflowError;
1034 #[derive(Copy, Clone, Debug, PartialEq, Eq, PartialOrd, Ord, Hash, Encodable, Decodable)]
1035 pub enum SourceFileHashAlgorithm {
1041 impl FromStr for SourceFileHashAlgorithm {
1044 fn from_str(s: &str) -> Result<SourceFileHashAlgorithm, ()> {
1046 "md5" => Ok(SourceFileHashAlgorithm::Md5),
1047 "sha1" => Ok(SourceFileHashAlgorithm::Sha1),
1048 "sha256" => Ok(SourceFileHashAlgorithm::Sha256),
1054 rustc_data_structures::impl_stable_hash_via_hash!(SourceFileHashAlgorithm);
1056 /// The hash of the on-disk source file used for debug info.
1057 #[derive(Copy, Clone, PartialEq, Eq, Debug)]
1058 #[derive(HashStable_Generic, Encodable, Decodable)]
1059 pub struct SourceFileHash {
1060 pub kind: SourceFileHashAlgorithm,
1064 impl SourceFileHash {
1065 pub fn new(kind: SourceFileHashAlgorithm, src: &str) -> SourceFileHash {
1066 let mut hash = SourceFileHash { kind, value: Default::default() };
1067 let len = hash.hash_len();
1068 let value = &mut hash.value[..len];
1069 let data = src.as_bytes();
1071 SourceFileHashAlgorithm::Md5 => {
1072 value.copy_from_slice(&Md5::digest(data));
1074 SourceFileHashAlgorithm::Sha1 => {
1075 value.copy_from_slice(&Sha1::digest(data));
1077 SourceFileHashAlgorithm::Sha256 => {
1078 value.copy_from_slice(&Sha256::digest(data));
1084 /// Check if the stored hash matches the hash of the string.
1085 pub fn matches(&self, src: &str) -> bool {
1086 Self::new(self.kind, src) == *self
1089 /// The bytes of the hash.
1090 pub fn hash_bytes(&self) -> &[u8] {
1091 let len = self.hash_len();
1095 fn hash_len(&self) -> usize {
1097 SourceFileHashAlgorithm::Md5 => 16,
1098 SourceFileHashAlgorithm::Sha1 => 20,
1099 SourceFileHashAlgorithm::Sha256 => 32,
1104 /// A single source in the `SourceMap`.
1106 pub struct SourceFile {
1107 /// The name of the file that the source came from. Source that doesn't
1108 /// originate from files has names between angle brackets by convention
1109 /// (e.g., `<anon>`).
1111 /// `true` if the `name` field above has been modified by `--remap-path-prefix`.
1112 pub name_was_remapped: bool,
1113 /// The unmapped path of the file that the source came from.
1114 /// Set to `None` if the `SourceFile` was imported from an external crate.
1115 pub unmapped_path: Option<FileName>,
1116 /// The complete source code.
1117 pub src: Option<Lrc<String>>,
1118 /// The source code's hash.
1119 pub src_hash: SourceFileHash,
1120 /// The external source code (used for external crates, which will have a `None`
1121 /// value as `self.src`.
1122 pub external_src: Lock<ExternalSource>,
1123 /// The start position of this source in the `SourceMap`.
1124 pub start_pos: BytePos,
1125 /// The end position of this source in the `SourceMap`.
1126 pub end_pos: BytePos,
1127 /// Locations of lines beginnings in the source code.
1128 pub lines: Vec<BytePos>,
1129 /// Locations of multi-byte characters in the source code.
1130 pub multibyte_chars: Vec<MultiByteChar>,
1131 /// Width of characters that are not narrow in the source code.
1132 pub non_narrow_chars: Vec<NonNarrowChar>,
1133 /// Locations of characters removed during normalization.
1134 pub normalized_pos: Vec<NormalizedPos>,
1135 /// A hash of the filename, used for speeding up hashing in incremental compilation.
1136 pub name_hash: u128,
1137 /// Indicates which crate this `SourceFile` was imported from.
1141 impl<S: Encoder> Encodable<S> for SourceFile {
1142 fn encode(&self, s: &mut S) -> Result<(), S::Error> {
1143 s.emit_struct("SourceFile", 8, |s| {
1144 s.emit_struct_field("name", 0, |s| self.name.encode(s))?;
1145 s.emit_struct_field("name_was_remapped", 1, |s| self.name_was_remapped.encode(s))?;
1146 s.emit_struct_field("src_hash", 2, |s| self.src_hash.encode(s))?;
1147 s.emit_struct_field("start_pos", 3, |s| self.start_pos.encode(s))?;
1148 s.emit_struct_field("end_pos", 4, |s| self.end_pos.encode(s))?;
1149 s.emit_struct_field("lines", 5, |s| {
1150 let lines = &self.lines[..];
1151 // Store the length.
1152 s.emit_u32(lines.len() as u32)?;
1154 if !lines.is_empty() {
1155 // In order to preserve some space, we exploit the fact that
1156 // the lines list is sorted and individual lines are
1157 // probably not that long. Because of that we can store lines
1158 // as a difference list, using as little space as possible
1159 // for the differences.
1160 let max_line_length = if lines.len() == 1 {
1165 .map(|&[fst, snd]| snd - fst)
1166 .map(|bp| bp.to_usize())
1171 let bytes_per_diff: u8 = match max_line_length {
1173 0x100..=0xFFFF => 2,
1177 // Encode the number of bytes used per diff.
1178 bytes_per_diff.encode(s)?;
1180 // Encode the first element.
1181 lines[0].encode(s)?;
1183 let diff_iter = lines[..].array_windows().map(|&[fst, snd]| snd - fst);
1185 match bytes_per_diff {
1187 for diff in diff_iter {
1188 (diff.0 as u8).encode(s)?
1192 for diff in diff_iter {
1193 (diff.0 as u16).encode(s)?
1197 for diff in diff_iter {
1201 _ => unreachable!(),
1207 s.emit_struct_field("multibyte_chars", 6, |s| self.multibyte_chars.encode(s))?;
1208 s.emit_struct_field("non_narrow_chars", 7, |s| self.non_narrow_chars.encode(s))?;
1209 s.emit_struct_field("name_hash", 8, |s| self.name_hash.encode(s))?;
1210 s.emit_struct_field("normalized_pos", 9, |s| self.normalized_pos.encode(s))?;
1211 s.emit_struct_field("cnum", 10, |s| self.cnum.encode(s))
1216 impl<D: Decoder> Decodable<D> for SourceFile {
1217 fn decode(d: &mut D) -> Result<SourceFile, D::Error> {
1218 d.read_struct("SourceFile", 8, |d| {
1219 let name: FileName = d.read_struct_field("name", 0, |d| Decodable::decode(d))?;
1220 let name_was_remapped: bool =
1221 d.read_struct_field("name_was_remapped", 1, |d| Decodable::decode(d))?;
1222 let src_hash: SourceFileHash =
1223 d.read_struct_field("src_hash", 2, |d| Decodable::decode(d))?;
1224 let start_pos: BytePos =
1225 d.read_struct_field("start_pos", 3, |d| Decodable::decode(d))?;
1226 let end_pos: BytePos = d.read_struct_field("end_pos", 4, |d| Decodable::decode(d))?;
1227 let lines: Vec<BytePos> = d.read_struct_field("lines", 5, |d| {
1228 let num_lines: u32 = Decodable::decode(d)?;
1229 let mut lines = Vec::with_capacity(num_lines as usize);
1232 // Read the number of bytes used per diff.
1233 let bytes_per_diff: u8 = Decodable::decode(d)?;
1235 // Read the first element.
1236 let mut line_start: BytePos = Decodable::decode(d)?;
1237 lines.push(line_start);
1239 for _ in 1..num_lines {
1240 let diff = match bytes_per_diff {
1241 1 => d.read_u8()? as u32,
1242 2 => d.read_u16()? as u32,
1244 _ => unreachable!(),
1247 line_start = line_start + BytePos(diff);
1249 lines.push(line_start);
1255 let multibyte_chars: Vec<MultiByteChar> =
1256 d.read_struct_field("multibyte_chars", 6, |d| Decodable::decode(d))?;
1257 let non_narrow_chars: Vec<NonNarrowChar> =
1258 d.read_struct_field("non_narrow_chars", 7, |d| Decodable::decode(d))?;
1259 let name_hash: u128 = d.read_struct_field("name_hash", 8, |d| Decodable::decode(d))?;
1260 let normalized_pos: Vec<NormalizedPos> =
1261 d.read_struct_field("normalized_pos", 9, |d| Decodable::decode(d))?;
1262 let cnum: CrateNum = d.read_struct_field("cnum", 10, |d| Decodable::decode(d))?;
1266 unmapped_path: None,
1271 // Unused - the metadata decoder will construct
1272 // a new SourceFile, filling in `external_src` properly
1273 external_src: Lock::new(ExternalSource::Unneeded),
1285 impl fmt::Debug for SourceFile {
1286 fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result {
1287 write!(fmt, "SourceFile({})", self.name)
1294 name_was_remapped: bool,
1295 unmapped_path: FileName,
1298 hash_kind: SourceFileHashAlgorithm,
1300 // Compute the file hash before any normalization.
1301 let src_hash = SourceFileHash::new(hash_kind, &src);
1302 let normalized_pos = normalize_src(&mut src, start_pos);
1305 let mut hasher: StableHasher = StableHasher::new();
1306 name.hash(&mut hasher);
1307 hasher.finish::<u128>()
1309 let end_pos = start_pos.to_usize() + src.len();
1310 assert!(end_pos <= u32::MAX as usize);
1312 let (lines, multibyte_chars, non_narrow_chars) =
1313 analyze_source_file::analyze_source_file(&src[..], start_pos);
1318 unmapped_path: Some(unmapped_path),
1319 src: Some(Lrc::new(src)),
1321 external_src: Lock::new(ExternalSource::Unneeded),
1323 end_pos: Pos::from_usize(end_pos),
1333 /// Returns the `BytePos` of the beginning of the current line.
1334 pub fn line_begin_pos(&self, pos: BytePos) -> BytePos {
1335 let line_index = self.lookup_line(pos).unwrap();
1336 self.lines[line_index]
1339 /// Add externally loaded source.
1340 /// If the hash of the input doesn't match or no input is supplied via None,
1341 /// it is interpreted as an error and the corresponding enum variant is set.
1342 /// The return value signifies whether some kind of source is present.
1343 pub fn add_external_src<F>(&self, get_src: F) -> bool
1345 F: FnOnce() -> Option<String>,
1348 *self.external_src.borrow(),
1349 ExternalSource::Foreign { kind: ExternalSourceKind::AbsentOk, .. }
1351 let src = get_src();
1352 let mut external_src = self.external_src.borrow_mut();
1353 // Check that no-one else have provided the source while we were getting it
1354 if let ExternalSource::Foreign {
1355 kind: src_kind @ ExternalSourceKind::AbsentOk, ..
1356 } = &mut *external_src
1358 if let Some(mut src) = src {
1359 // The src_hash needs to be computed on the pre-normalized src.
1360 if self.src_hash.matches(&src) {
1361 normalize_src(&mut src, BytePos::from_usize(0));
1362 *src_kind = ExternalSourceKind::Present(Lrc::new(src));
1366 *src_kind = ExternalSourceKind::AbsentErr;
1371 self.src.is_some() || external_src.get_source().is_some()
1374 self.src.is_some() || self.external_src.borrow().get_source().is_some()
1378 /// Gets a line from the list of pre-computed line-beginnings.
1379 /// The line number here is 0-based.
1380 pub fn get_line(&self, line_number: usize) -> Option<Cow<'_, str>> {
1381 fn get_until_newline(src: &str, begin: usize) -> &str {
1382 // We can't use `lines.get(line_number+1)` because we might
1383 // be parsing when we call this function and thus the current
1384 // line is the last one we have line info for.
1385 let slice = &src[begin..];
1386 match slice.find('\n') {
1387 Some(e) => &slice[..e],
1393 let line = self.lines.get(line_number)?;
1394 let begin: BytePos = *line - self.start_pos;
1398 if let Some(ref src) = self.src {
1399 Some(Cow::from(get_until_newline(src, begin)))
1400 } else if let Some(src) = self.external_src.borrow().get_source() {
1401 Some(Cow::Owned(String::from(get_until_newline(src, begin))))
1407 pub fn is_real_file(&self) -> bool {
1411 pub fn is_imported(&self) -> bool {
1415 pub fn byte_length(&self) -> u32 {
1416 self.end_pos.0 - self.start_pos.0
1418 pub fn count_lines(&self) -> usize {
1422 /// Finds the line containing the given position. The return value is the
1423 /// index into the `lines` array of this `SourceFile`, not the 1-based line
1424 /// number. If the source_file is empty or the position is located before the
1425 /// first line, `None` is returned.
1426 pub fn lookup_line(&self, pos: BytePos) -> Option<usize> {
1427 if self.lines.is_empty() {
1431 let line_index = lookup_line(&self.lines[..], pos);
1432 assert!(line_index < self.lines.len() as isize);
1433 if line_index >= 0 { Some(line_index as usize) } else { None }
1436 pub fn line_bounds(&self, line_index: usize) -> Range<BytePos> {
1437 if self.is_empty() {
1438 return self.start_pos..self.end_pos;
1441 assert!(line_index < self.lines.len());
1442 if line_index == (self.lines.len() - 1) {
1443 self.lines[line_index]..self.end_pos
1445 self.lines[line_index]..self.lines[line_index + 1]
1449 /// Returns whether or not the file contains the given `SourceMap` byte
1450 /// position. The position one past the end of the file is considered to be
1451 /// contained by the file. This implies that files for which `is_empty`
1452 /// returns true still contain one byte position according to this function.
1454 pub fn contains(&self, byte_pos: BytePos) -> bool {
1455 byte_pos >= self.start_pos && byte_pos <= self.end_pos
1459 pub fn is_empty(&self) -> bool {
1460 self.start_pos == self.end_pos
1463 /// Calculates the original byte position relative to the start of the file
1464 /// based on the given byte position.
1465 pub fn original_relative_byte_pos(&self, pos: BytePos) -> BytePos {
1466 // Diff before any records is 0. Otherwise use the previously recorded
1467 // diff as that applies to the following characters until a new diff
1469 let diff = match self.normalized_pos.binary_search_by(|np| np.pos.cmp(&pos)) {
1470 Ok(i) => self.normalized_pos[i].diff,
1471 Err(i) if i == 0 => 0,
1472 Err(i) => self.normalized_pos[i - 1].diff,
1475 BytePos::from_u32(pos.0 - self.start_pos.0 + diff)
1478 /// Converts an absolute `BytePos` to a `CharPos` relative to the `SourceFile`.
1479 pub fn bytepos_to_file_charpos(&self, bpos: BytePos) -> CharPos {
1480 // The number of extra bytes due to multibyte chars in the `SourceFile`.
1481 let mut total_extra_bytes = 0;
1483 for mbc in self.multibyte_chars.iter() {
1484 debug!("{}-byte char at {:?}", mbc.bytes, mbc.pos);
1486 // Every character is at least one byte, so we only
1487 // count the actual extra bytes.
1488 total_extra_bytes += mbc.bytes as u32 - 1;
1489 // We should never see a byte position in the middle of a
1491 assert!(bpos.to_u32() >= mbc.pos.to_u32() + mbc.bytes as u32);
1497 assert!(self.start_pos.to_u32() + total_extra_bytes <= bpos.to_u32());
1498 CharPos(bpos.to_usize() - self.start_pos.to_usize() - total_extra_bytes as usize)
1501 /// Looks up the file's (1-based) line number and (0-based `CharPos`) column offset, for a
1502 /// given `BytePos`.
1503 pub fn lookup_file_pos(&self, pos: BytePos) -> (usize, CharPos) {
1504 let chpos = self.bytepos_to_file_charpos(pos);
1505 match self.lookup_line(pos) {
1507 let line = a + 1; // Line numbers start at 1
1508 let linebpos = self.lines[a];
1509 let linechpos = self.bytepos_to_file_charpos(linebpos);
1510 let col = chpos - linechpos;
1511 debug!("byte pos {:?} is on the line at byte pos {:?}", pos, linebpos);
1512 debug!("char pos {:?} is on the line at char pos {:?}", chpos, linechpos);
1513 debug!("byte is on line: {}", line);
1514 assert!(chpos >= linechpos);
1521 /// Looks up the file's (1-based) line number, (0-based `CharPos`) column offset, and (0-based)
1522 /// column offset when displayed, for a given `BytePos`.
1523 pub fn lookup_file_pos_with_col_display(&self, pos: BytePos) -> (usize, CharPos, usize) {
1524 let (line, col_or_chpos) = self.lookup_file_pos(pos);
1526 let col = col_or_chpos;
1527 let linebpos = self.lines[line - 1];
1529 let start_width_idx = self
1531 .binary_search_by_key(&linebpos, |x| x.pos())
1532 .unwrap_or_else(|x| x);
1533 let end_width_idx = self
1535 .binary_search_by_key(&pos, |x| x.pos())
1536 .unwrap_or_else(|x| x);
1537 let special_chars = end_width_idx - start_width_idx;
1538 let non_narrow: usize = self.non_narrow_chars[start_width_idx..end_width_idx]
1542 col.0 - special_chars + non_narrow
1544 (line, col, col_display)
1546 let chpos = col_or_chpos;
1548 let end_width_idx = self
1550 .binary_search_by_key(&pos, |x| x.pos())
1551 .unwrap_or_else(|x| x);
1552 let non_narrow: usize =
1553 self.non_narrow_chars[0..end_width_idx].iter().map(|x| x.width()).sum();
1554 chpos.0 - end_width_idx + non_narrow
1556 (0, chpos, col_display)
1561 /// Normalizes the source code and records the normalizations.
1562 fn normalize_src(src: &mut String, start_pos: BytePos) -> Vec<NormalizedPos> {
1563 let mut normalized_pos = vec![];
1564 remove_bom(src, &mut normalized_pos);
1565 normalize_newlines(src, &mut normalized_pos);
1567 // Offset all the positions by start_pos to match the final file positions.
1568 for np in &mut normalized_pos {
1569 np.pos.0 += start_pos.0;
1575 /// Removes UTF-8 BOM, if any.
1576 fn remove_bom(src: &mut String, normalized_pos: &mut Vec<NormalizedPos>) {
1577 if src.starts_with('\u{feff}') {
1579 normalized_pos.push(NormalizedPos { pos: BytePos(0), diff: 3 });
1583 /// Replaces `\r\n` with `\n` in-place in `src`.
1585 /// Returns error if there's a lone `\r` in the string
1586 fn normalize_newlines(src: &mut String, normalized_pos: &mut Vec<NormalizedPos>) {
1587 if !src.as_bytes().contains(&b'\r') {
1591 // We replace `\r\n` with `\n` in-place, which doesn't break utf-8 encoding.
1592 // While we *can* call `as_mut_vec` and do surgery on the live string
1593 // directly, let's rather steal the contents of `src`. This makes the code
1594 // safe even if a panic occurs.
1596 let mut buf = std::mem::replace(src, String::new()).into_bytes();
1597 let mut gap_len = 0;
1598 let mut tail = buf.as_mut_slice();
1600 let original_gap = normalized_pos.last().map_or(0, |l| l.diff);
1602 let idx = match find_crlf(&tail[gap_len..]) {
1604 Some(idx) => idx + gap_len,
1606 tail.copy_within(gap_len..idx, 0);
1607 tail = &mut tail[idx - gap_len..];
1608 if tail.len() == gap_len {
1611 cursor += idx - gap_len;
1613 normalized_pos.push(NormalizedPos {
1614 pos: BytePos::from_usize(cursor + 1),
1615 diff: original_gap + gap_len as u32,
1619 // Account for removed `\r`.
1620 // After `set_len`, `buf` is guaranteed to contain utf-8 again.
1621 let new_len = buf.len() - gap_len;
1623 buf.set_len(new_len);
1624 *src = String::from_utf8_unchecked(buf);
1627 fn find_crlf(src: &[u8]) -> Option<usize> {
1628 let mut search_idx = 0;
1629 while let Some(idx) = find_cr(&src[search_idx..]) {
1630 if src[search_idx..].get(idx + 1) != Some(&b'\n') {
1631 search_idx += idx + 1;
1634 return Some(search_idx + idx);
1639 fn find_cr(src: &[u8]) -> Option<usize> {
1640 src.iter().position(|&b| b == b'\r')
1644 // _____________________________________________________________________________
1645 // Pos, BytePos, CharPos
1649 fn from_usize(n: usize) -> Self;
1650 fn to_usize(&self) -> usize;
1651 fn from_u32(n: u32) -> Self;
1652 fn to_u32(&self) -> u32;
1655 macro_rules! impl_pos {
1659 $vis:vis struct $ident:ident($inner_vis:vis $inner_ty:ty);
1664 $vis struct $ident($inner_vis $inner_ty);
1666 impl Pos for $ident {
1668 fn from_usize(n: usize) -> $ident {
1669 $ident(n as $inner_ty)
1673 fn to_usize(&self) -> usize {
1678 fn from_u32(n: u32) -> $ident {
1679 $ident(n as $inner_ty)
1683 fn to_u32(&self) -> u32 {
1688 impl Add for $ident {
1689 type Output = $ident;
1692 fn add(self, rhs: $ident) -> $ident {
1693 $ident(self.0 + rhs.0)
1697 impl Sub for $ident {
1698 type Output = $ident;
1701 fn sub(self, rhs: $ident) -> $ident {
1702 $ident(self.0 - rhs.0)
1710 /// A byte offset. Keep this small (currently 32-bits), as AST contains
1712 #[derive(Clone, Copy, PartialEq, Eq, Hash, PartialOrd, Ord, Debug)]
1713 pub struct BytePos(pub u32);
1715 /// A character offset. Because of multibyte UTF-8 characters, a byte offset
1716 /// is not equivalent to a character offset. The `SourceMap` will convert `BytePos`
1717 /// values to `CharPos` values as necessary.
1718 #[derive(Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Debug)]
1719 pub struct CharPos(pub usize);
1722 impl<S: rustc_serialize::Encoder> Encodable<S> for BytePos {
1723 fn encode(&self, s: &mut S) -> Result<(), S::Error> {
1728 impl<D: rustc_serialize::Decoder> Decodable<D> for BytePos {
1729 fn decode(d: &mut D) -> Result<BytePos, D::Error> {
1730 Ok(BytePos(d.read_u32()?))
1734 // _____________________________________________________________________________
1735 // Loc, SourceFileAndLine, SourceFileAndBytePos
1738 /// A source code location used for error reporting.
1739 #[derive(Debug, Clone)]
1741 /// Information about the original source.
1742 pub file: Lrc<SourceFile>,
1743 /// The (1-based) line number.
1745 /// The (0-based) column offset.
1747 /// The (0-based) column offset when displayed.
1748 pub col_display: usize,
1751 // Used to be structural records.
1753 pub struct SourceFileAndLine {
1754 pub sf: Lrc<SourceFile>,
1758 pub struct SourceFileAndBytePos {
1759 pub sf: Lrc<SourceFile>,
1763 #[derive(Copy, Clone, Debug, PartialEq, Eq)]
1764 pub struct LineInfo {
1765 /// Index of line, starting from 0.
1766 pub line_index: usize,
1768 /// Column in line where span begins, starting from 0.
1769 pub start_col: CharPos,
1771 /// Column in line where span ends, starting from 0, exclusive.
1772 pub end_col: CharPos,
1775 pub struct FileLines {
1776 pub file: Lrc<SourceFile>,
1777 pub lines: Vec<LineInfo>,
1780 pub static SPAN_DEBUG: AtomicRef<fn(Span, &mut fmt::Formatter<'_>) -> fmt::Result> =
1781 AtomicRef::new(&(default_span_debug as fn(_, &mut fmt::Formatter<'_>) -> _));
1783 // _____________________________________________________________________________
1784 // SpanLinesError, SpanSnippetError, DistinctSources, MalformedSourceMapPositions
1787 pub type FileLinesResult = Result<FileLines, SpanLinesError>;
1789 #[derive(Clone, PartialEq, Eq, Debug)]
1790 pub enum SpanLinesError {
1791 DistinctSources(DistinctSources),
1794 #[derive(Clone, PartialEq, Eq, Debug)]
1795 pub enum SpanSnippetError {
1796 IllFormedSpan(Span),
1797 DistinctSources(DistinctSources),
1798 MalformedForSourcemap(MalformedSourceMapPositions),
1799 SourceNotAvailable { filename: FileName },
1802 #[derive(Clone, PartialEq, Eq, Debug)]
1803 pub struct DistinctSources {
1804 pub begin: (FileName, BytePos),
1805 pub end: (FileName, BytePos),
1808 #[derive(Clone, PartialEq, Eq, Debug)]
1809 pub struct MalformedSourceMapPositions {
1811 pub source_len: usize,
1812 pub begin_pos: BytePos,
1813 pub end_pos: BytePos,
1816 /// Range inside of a `Span` used for diagnostics when we only have access to relative positions.
1817 #[derive(Copy, Clone, PartialEq, Eq, Debug)]
1818 pub struct InnerSpan {
1824 pub fn new(start: usize, end: usize) -> InnerSpan {
1825 InnerSpan { start, end }
1829 // Given a slice of line start positions and a position, returns the index of
1830 // the line the position is on. Returns -1 if the position is located before
1832 fn lookup_line(lines: &[BytePos], pos: BytePos) -> isize {
1833 match lines.binary_search(&pos) {
1834 Ok(line) => line as isize,
1835 Err(line) => line as isize - 1,
1839 /// Requirements for a `StableHashingContext` to be used in this crate.
1840 /// This is a hack to allow using the `HashStable_Generic` derive macro
1841 /// instead of implementing everything in librustc_middle.
1842 pub trait HashStableContext {
1843 fn hash_def_id(&mut self, _: DefId, hasher: &mut StableHasher);
1844 fn hash_crate_num(&mut self, _: CrateNum, hasher: &mut StableHasher);
1845 fn hash_spans(&self) -> bool;
1846 fn byte_pos_to_line_and_col(
1849 ) -> Option<(Lrc<SourceFile>, usize, BytePos)>;
1852 impl<CTX> HashStable<CTX> for Span
1854 CTX: HashStableContext,
1856 /// Hashes a span in a stable way. We can't directly hash the span's `BytePos`
1857 /// fields (that would be similar to hashing pointers, since those are just
1858 /// offsets into the `SourceMap`). Instead, we hash the (file name, line, column)
1859 /// triple, which stays the same even if the containing `SourceFile` has moved
1860 /// within the `SourceMap`.
1861 /// Also note that we are hashing byte offsets for the column, not unicode
1862 /// codepoint offsets. For the purpose of the hash that's sufficient.
1863 /// Also, hashing filenames is expensive so we avoid doing it twice when the
1864 /// span starts and ends in the same file, which is almost always the case.
1865 fn hash_stable(&self, ctx: &mut CTX, hasher: &mut StableHasher) {
1866 const TAG_VALID_SPAN: u8 = 0;
1867 const TAG_INVALID_SPAN: u8 = 1;
1869 if !ctx.hash_spans() {
1873 if *self == DUMMY_SP {
1874 Hash::hash(&TAG_INVALID_SPAN, hasher);
1878 // If this is not an empty or invalid span, we want to hash the last
1879 // position that belongs to it, as opposed to hashing the first
1880 // position past it.
1881 let span = self.data();
1882 let (file_lo, line_lo, col_lo) = match ctx.byte_pos_to_line_and_col(span.lo) {
1885 Hash::hash(&TAG_INVALID_SPAN, hasher);
1886 span.ctxt.hash_stable(ctx, hasher);
1891 if !file_lo.contains(span.hi) {
1892 Hash::hash(&TAG_INVALID_SPAN, hasher);
1893 span.ctxt.hash_stable(ctx, hasher);
1897 let (_, line_hi, col_hi) = match ctx.byte_pos_to_line_and_col(span.hi) {
1900 Hash::hash(&TAG_INVALID_SPAN, hasher);
1901 span.ctxt.hash_stable(ctx, hasher);
1906 Hash::hash(&TAG_VALID_SPAN, hasher);
1907 // We truncate the stable ID hash and line and column numbers. The chances
1908 // of causing a collision this way should be minimal.
1909 Hash::hash(&(file_lo.name_hash as u64), hasher);
1911 // Hash both the length and the end location (line/column) of a span. If we
1912 // hash only the length, for example, then two otherwise equal spans with
1913 // different end locations will have the same hash. This can cause a problem
1914 // during incremental compilation wherein a previous result for a query that
1915 // depends on the end location of a span will be incorrectly reused when the
1916 // end location of the span it depends on has changed (see issue #74890). A
1917 // similar analysis applies if some query depends specifically on the length
1918 // of the span, but we only hash the end location. So hash both.
1920 let col_lo_trunc = (col_lo.0 as u64) & 0xFF;
1921 let line_lo_trunc = ((line_lo as u64) & 0xFF_FF_FF) << 8;
1922 let col_hi_trunc = (col_hi.0 as u64) & 0xFF << 32;
1923 let line_hi_trunc = ((line_hi as u64) & 0xFF_FF_FF) << 40;
1924 let col_line = col_lo_trunc | line_lo_trunc | col_hi_trunc | line_hi_trunc;
1925 let len = (span.hi - span.lo).0;
1926 Hash::hash(&col_line, hasher);
1927 Hash::hash(&len, hasher);
1928 span.ctxt.hash_stable(ctx, hasher);
1932 impl<CTX: HashStableContext> HashStable<CTX> for SyntaxContext {
1933 fn hash_stable(&self, ctx: &mut CTX, hasher: &mut StableHasher) {
1934 const TAG_EXPANSION: u8 = 0;
1935 const TAG_NO_EXPANSION: u8 = 1;
1937 if *self == SyntaxContext::root() {
1938 TAG_NO_EXPANSION.hash_stable(ctx, hasher);
1940 TAG_EXPANSION.hash_stable(ctx, hasher);
1941 let (expn_id, transparency) = self.outer_mark();
1942 expn_id.hash_stable(ctx, hasher);
1943 transparency.hash_stable(ctx, hasher);
1948 impl<CTX: HashStableContext> HashStable<CTX> for ExpnId {
1949 fn hash_stable(&self, ctx: &mut CTX, hasher: &mut StableHasher) {
1950 // Since the same expansion context is usually referenced many
1951 // times, we cache a stable hash of it and hash that instead of
1952 // recursing every time.
1954 static CACHE: RefCell<Vec<Option<Fingerprint>>> = Default::default();
1957 const TAG_ROOT: u8 = 0;
1958 const TAG_NOT_ROOT: u8 = 1;
1960 if *self == ExpnId::root() {
1961 TAG_ROOT.hash_stable(ctx, hasher);
1965 let index = self.as_u32() as usize;
1966 let res = CACHE.with(|cache| cache.borrow().get(index).copied().flatten());
1968 if let Some(res) = res {
1969 res.hash_stable(ctx, hasher);
1971 let new_len = index + 1;
1973 let mut sub_hasher = StableHasher::new();
1974 TAG_NOT_ROOT.hash_stable(ctx, &mut sub_hasher);
1975 self.expn_data().hash_stable(ctx, &mut sub_hasher);
1976 let sub_hash: Fingerprint = sub_hasher.finish();
1978 CACHE.with(|cache| {
1979 let mut cache = cache.borrow_mut();
1980 if cache.len() < new_len {
1981 cache.resize(new_len, None);
1983 cache[index].replace(sub_hash).expect_none("Cache slot was filled");
1985 sub_hash.hash_stable(ctx, hasher);