1 //! Source positions and related helper functions.
3 //! Important concepts in this module include:
5 //! - the *span*, represented by [`SpanData`] and related types;
6 //! - source code as represented by a [`SourceMap`]; and
7 //! - interned strings, represented by [`Symbol`]s, with some common symbols available statically in the [`sym`] module.
9 //! Unlike most compilers, the span contains not only the position in the source code, but also various other metadata,
10 //! such as the edition and macro hygiene. This metadata is stored in [`SyntaxContext`] and [`ExpnData`].
14 //! This API is completely unstable and subject to change.
16 #![doc(html_root_url = "https://doc.rust-lang.org/nightly/nightly-rustc/")]
17 #![feature(array_windows)]
18 #![feature(crate_visibility_modifier)]
19 #![feature(if_let_guard)]
20 #![feature(negative_impls)]
22 #![feature(min_specialization)]
25 extern crate rustc_macros;
30 use rustc_data_structures::AtomicRef;
31 use rustc_macros::HashStable_Generic;
32 use rustc_serialize::{Decodable, Decoder, Encodable, Encoder};
34 mod caching_source_map_view;
36 pub use self::caching_source_map_view::CachingSourceMapView;
37 use source_map::SourceMap;
42 use hygiene::Transparency;
43 pub use hygiene::{DesugaringKind, ExpnKind, MacroKind};
44 pub use hygiene::{ExpnData, ExpnHash, ExpnId, LocalExpnId, SyntaxContext};
45 use rustc_data_structures::stable_hasher::HashingControls;
47 use def_id::{CrateNum, DefId, DefPathHash, LocalDefId, LOCAL_CRATE};
50 pub use span_encoding::{Span, DUMMY_SP};
53 pub use symbol::{sym, Symbol};
55 mod analyze_source_file;
58 use rustc_data_structures::stable_hasher::{HashStable, StableHasher};
59 use rustc_data_structures::sync::{Lock, Lrc};
62 use std::cmp::{self, Ordering};
65 use std::ops::{Add, Range, Sub};
66 use std::path::{Path, PathBuf};
67 use std::str::FromStr;
79 // Per-session global variables: this struct is stored in thread-local storage
80 // in such a way that it is accessible without any kind of handle to all
81 // threads within the compilation session, but is not accessible outside the
83 pub struct SessionGlobals {
84 symbol_interner: symbol::Interner,
85 span_interner: Lock<span_encoding::SpanInterner>,
86 hygiene_data: Lock<hygiene::HygieneData>,
87 source_map: Lock<Option<Lrc<SourceMap>>>,
91 pub fn new(edition: Edition) -> SessionGlobals {
93 symbol_interner: symbol::Interner::fresh(),
94 span_interner: Lock::new(span_encoding::SpanInterner::default()),
95 hygiene_data: Lock::new(hygiene::HygieneData::new(edition)),
96 source_map: Lock::new(None),
102 pub fn create_session_globals_then<R>(edition: Edition, f: impl FnOnce() -> R) -> R {
104 !SESSION_GLOBALS.is_set(),
105 "SESSION_GLOBALS should never be overwritten! \
106 Use another thread if you need another SessionGlobals"
108 let session_globals = SessionGlobals::new(edition);
109 SESSION_GLOBALS.set(&session_globals, f)
113 pub fn set_session_globals_then<R>(session_globals: &SessionGlobals, f: impl FnOnce() -> R) -> R {
115 !SESSION_GLOBALS.is_set(),
116 "SESSION_GLOBALS should never be overwritten! \
117 Use another thread if you need another SessionGlobals"
119 SESSION_GLOBALS.set(session_globals, f)
123 pub fn create_default_session_if_not_set_then<R, F>(f: F) -> R
125 F: FnOnce(&SessionGlobals) -> R,
127 create_session_if_not_set_then(edition::DEFAULT_EDITION, f)
131 pub fn create_session_if_not_set_then<R, F>(edition: Edition, f: F) -> R
133 F: FnOnce(&SessionGlobals) -> R,
135 if !SESSION_GLOBALS.is_set() {
136 let session_globals = SessionGlobals::new(edition);
137 SESSION_GLOBALS.set(&session_globals, || SESSION_GLOBALS.with(f))
139 SESSION_GLOBALS.with(f)
144 pub fn with_session_globals<R, F>(f: F) -> R
146 F: FnOnce(&SessionGlobals) -> R,
148 SESSION_GLOBALS.with(f)
152 pub fn create_default_session_globals_then<R>(f: impl FnOnce() -> R) -> R {
153 create_session_globals_then(edition::DEFAULT_EDITION, f)
156 // If this ever becomes non thread-local, `decode_syntax_context`
157 // and `decode_expn_id` will need to be updated to handle concurrent
159 scoped_tls::scoped_thread_local!(static SESSION_GLOBALS: SessionGlobals);
161 // FIXME: We should use this enum or something like it to get rid of the
162 // use of magic `/rust/1.x/...` paths across the board.
163 #[derive(Debug, Eq, PartialEq, Clone, Ord, PartialOrd)]
165 pub enum RealFileName {
167 /// For remapped paths (namely paths into libstd that have been mapped
168 /// to the appropriate spot on the local host's file system, and local file
169 /// system paths that have been remapped with `FilePathMapping`),
171 /// `local_path` is the (host-dependent) local path to the file. This is
172 /// None if the file was imported from another crate
173 local_path: Option<PathBuf>,
174 /// `virtual_name` is the stable path rustc will store internally within
176 virtual_name: PathBuf,
180 impl Hash for RealFileName {
181 fn hash<H: std::hash::Hasher>(&self, state: &mut H) {
182 // To prevent #70924 from happening again we should only hash the
183 // remapped (virtualized) path if that exists. This is because
184 // virtualized paths to sysroot crates (/rust/$hash or /rust/$version)
185 // remain stable even if the corresponding local_path changes
186 self.remapped_path_if_available().hash(state)
190 // This is functionally identical to #[derive(Encodable)], with the exception of
191 // an added assert statement
192 impl<S: Encoder> Encodable<S> for RealFileName {
193 fn encode(&self, encoder: &mut S) -> Result<(), S::Error> {
194 encoder.emit_enum(|encoder| match *self {
195 RealFileName::LocalPath(ref local_path) => {
196 encoder.emit_enum_variant("LocalPath", 0, 1, |encoder| {
197 encoder.emit_enum_variant_arg(true, |encoder| local_path.encode(encoder))?;
202 RealFileName::Remapped { ref local_path, ref virtual_name } => encoder
203 .emit_enum_variant("Remapped", 1, 2, |encoder| {
204 // For privacy and build reproducibility, we must not embed host-dependant path in artifacts
205 // if they have been remapped by --remap-path-prefix
206 assert!(local_path.is_none());
207 encoder.emit_enum_variant_arg(true, |encoder| local_path.encode(encoder))?;
208 encoder.emit_enum_variant_arg(false, |encoder| virtual_name.encode(encoder))?;
216 /// Returns the path suitable for reading from the file system on the local host,
217 /// if this information exists.
218 /// Avoid embedding this in build artifacts; see `remapped_path_if_available()` for that.
219 pub fn local_path(&self) -> Option<&Path> {
221 RealFileName::LocalPath(p) => Some(p),
222 RealFileName::Remapped { local_path: p, virtual_name: _ } => {
223 p.as_ref().map(PathBuf::as_path)
228 /// Returns the path suitable for reading from the file system on the local host,
229 /// if this information exists.
230 /// Avoid embedding this in build artifacts; see `remapped_path_if_available()` for that.
231 pub fn into_local_path(self) -> Option<PathBuf> {
233 RealFileName::LocalPath(p) => Some(p),
234 RealFileName::Remapped { local_path: p, virtual_name: _ } => p,
238 /// Returns the path suitable for embedding into build artifacts. This would still
239 /// be a local path if it has not been remapped. A remapped path will not correspond
240 /// to a valid file system path: see `local_path_if_available()` for something that
241 /// is more likely to return paths into the local host file system.
242 pub fn remapped_path_if_available(&self) -> &Path {
244 RealFileName::LocalPath(p)
245 | RealFileName::Remapped { local_path: _, virtual_name: p } => &p,
249 /// Returns the path suitable for reading from the file system on the local host,
250 /// if this information exists. Otherwise returns the remapped name.
251 /// Avoid embedding this in build artifacts; see `remapped_path_if_available()` for that.
252 pub fn local_path_if_available(&self) -> &Path {
254 RealFileName::LocalPath(path)
255 | RealFileName::Remapped { local_path: None, virtual_name: path }
256 | RealFileName::Remapped { local_path: Some(path), virtual_name: _ } => path,
260 pub fn to_string_lossy(&self, display_pref: FileNameDisplayPreference) -> Cow<'_, str> {
262 FileNameDisplayPreference::Local => self.local_path_if_available().to_string_lossy(),
263 FileNameDisplayPreference::Remapped => {
264 self.remapped_path_if_available().to_string_lossy()
270 /// Differentiates between real files and common virtual files.
271 #[derive(Debug, Eq, PartialEq, Clone, Ord, PartialOrd, Hash)]
272 #[derive(Decodable, Encodable)]
275 /// Call to `quote!`.
279 /// Hack in `src/librustc_ast/parse.rs`.
282 ProcMacroSourceCode(u64),
283 /// Strings provided as `--cfg [cfgspec]` stored in a `crate_cfg`.
285 /// Strings provided as crate attributes in the CLI.
287 /// Custom sources for explicit parser calls from plugins and drivers.
289 DocTest(PathBuf, isize),
290 /// Post-substitution inline assembly from LLVM.
294 impl From<PathBuf> for FileName {
295 fn from(p: PathBuf) -> Self {
296 assert!(!p.to_string_lossy().ends_with('>'));
297 FileName::Real(RealFileName::LocalPath(p))
301 #[derive(Clone, Copy, Eq, PartialEq, Hash, Debug)]
302 pub enum FileNameDisplayPreference {
307 pub struct FileNameDisplay<'a> {
309 display_pref: FileNameDisplayPreference,
312 impl fmt::Display for FileNameDisplay<'_> {
313 fn fmt(&self, fmt: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
317 write!(fmt, "{}", name.to_string_lossy(self.display_pref))
319 QuoteExpansion(_) => write!(fmt, "<quote expansion>"),
320 MacroExpansion(_) => write!(fmt, "<macro expansion>"),
321 Anon(_) => write!(fmt, "<anon>"),
322 ProcMacroSourceCode(_) => write!(fmt, "<proc-macro source code>"),
323 CfgSpec(_) => write!(fmt, "<cfgspec>"),
324 CliCrateAttr(_) => write!(fmt, "<crate attribute>"),
325 Custom(ref s) => write!(fmt, "<{}>", s),
326 DocTest(ref path, _) => write!(fmt, "{}", path.display()),
327 InlineAsm(_) => write!(fmt, "<inline asm>"),
332 impl FileNameDisplay<'_> {
333 pub fn to_string_lossy(&self) -> Cow<'_, str> {
335 FileName::Real(ref inner) => inner.to_string_lossy(self.display_pref),
336 _ => Cow::from(format!("{}", self)),
342 pub fn is_real(&self) -> bool {
348 | ProcMacroSourceCode(_)
354 | InlineAsm(_) => false,
358 pub fn prefer_remapped(&self) -> FileNameDisplay<'_> {
359 FileNameDisplay { inner: self, display_pref: FileNameDisplayPreference::Remapped }
362 // This may include transient local filesystem information.
363 // Must not be embedded in build outputs.
364 pub fn prefer_local(&self) -> FileNameDisplay<'_> {
365 FileNameDisplay { inner: self, display_pref: FileNameDisplayPreference::Local }
368 pub fn display(&self, display_pref: FileNameDisplayPreference) -> FileNameDisplay<'_> {
369 FileNameDisplay { inner: self, display_pref }
372 pub fn macro_expansion_source_code(src: &str) -> FileName {
373 let mut hasher = StableHasher::new();
374 src.hash(&mut hasher);
375 FileName::MacroExpansion(hasher.finish())
378 pub fn anon_source_code(src: &str) -> FileName {
379 let mut hasher = StableHasher::new();
380 src.hash(&mut hasher);
381 FileName::Anon(hasher.finish())
384 pub fn proc_macro_source_code(src: &str) -> FileName {
385 let mut hasher = StableHasher::new();
386 src.hash(&mut hasher);
387 FileName::ProcMacroSourceCode(hasher.finish())
390 pub fn cfg_spec_source_code(src: &str) -> FileName {
391 let mut hasher = StableHasher::new();
392 src.hash(&mut hasher);
393 FileName::QuoteExpansion(hasher.finish())
396 pub fn cli_crate_attr_source_code(src: &str) -> FileName {
397 let mut hasher = StableHasher::new();
398 src.hash(&mut hasher);
399 FileName::CliCrateAttr(hasher.finish())
402 pub fn doc_test_source_code(path: PathBuf, line: isize) -> FileName {
403 FileName::DocTest(path, line)
406 pub fn inline_asm_source_code(src: &str) -> FileName {
407 let mut hasher = StableHasher::new();
408 src.hash(&mut hasher);
409 FileName::InlineAsm(hasher.finish())
413 /// Represents a span.
415 /// Spans represent a region of code, used for error reporting. Positions in spans
416 /// are *absolute* positions from the beginning of the [`SourceMap`], not positions
417 /// relative to [`SourceFile`]s. Methods on the `SourceMap` can be used to relate spans back
418 /// to the original source.
420 /// You must be careful if the span crosses more than one file, since you will not be
421 /// able to use many of the functions on spans in source_map and you cannot assume
422 /// that the length of the span is equal to `span.hi - span.lo`; there may be space in the
423 /// [`BytePos`] range between files.
425 /// `SpanData` is public because `Span` uses a thread-local interner and can't be
426 /// sent to other threads, but some pieces of performance infra run in a separate thread.
427 /// Using `Span` is generally preferred.
428 #[derive(Clone, Copy, Hash, PartialEq, Eq)]
429 pub struct SpanData {
432 /// Information about where the macro came from, if this piece of
433 /// code was created by a macro expansion.
434 pub ctxt: SyntaxContext,
435 pub parent: Option<LocalDefId>,
438 // Order spans by position in the file.
439 impl Ord for SpanData {
440 fn cmp(&self, other: &Self) -> Ordering {
445 // `LocalDefId` does not implement `Ord`.
446 // The other fields are enough to determine in-file order.
453 // `LocalDefId` does not implement `Ord`.
454 // The other fields are enough to determine in-file order.
458 (s_lo, s_hi, s_ctxt).cmp(&(o_lo, o_hi, o_ctxt))
462 impl PartialOrd for SpanData {
463 fn partial_cmp(&self, other: &Self) -> Option<Ordering> {
464 Some(self.cmp(other))
470 pub fn span(&self) -> Span {
471 Span::new(self.lo, self.hi, self.ctxt, self.parent)
474 pub fn with_lo(&self, lo: BytePos) -> Span {
475 Span::new(lo, self.hi, self.ctxt, self.parent)
478 pub fn with_hi(&self, hi: BytePos) -> Span {
479 Span::new(self.lo, hi, self.ctxt, self.parent)
482 pub fn with_ctxt(&self, ctxt: SyntaxContext) -> Span {
483 Span::new(self.lo, self.hi, ctxt, self.parent)
486 pub fn with_parent(&self, parent: Option<LocalDefId>) -> Span {
487 Span::new(self.lo, self.hi, self.ctxt, parent)
489 /// Returns `true` if this is a dummy span with any hygienic context.
491 pub fn is_dummy(self) -> bool {
492 self.lo.0 == 0 && self.hi.0 == 0
494 /// Returns `true` if `self` fully encloses `other`.
495 pub fn contains(self, other: Self) -> bool {
496 self.lo <= other.lo && other.hi <= self.hi
500 // The interner is pointed to by a thread local value which is only set on the main thread
501 // with parallelization is disabled. So we don't allow `Span` to transfer between threads
502 // to avoid panics and other errors, even though it would be memory safe to do so.
503 #[cfg(not(parallel_compiler))]
504 impl !Send for Span {}
505 #[cfg(not(parallel_compiler))]
506 impl !Sync for Span {}
508 impl PartialOrd for Span {
509 fn partial_cmp(&self, rhs: &Self) -> Option<Ordering> {
510 PartialOrd::partial_cmp(&self.data(), &rhs.data())
514 fn cmp(&self, rhs: &Self) -> Ordering {
515 Ord::cmp(&self.data(), &rhs.data())
519 /// A collection of `Span`s.
521 /// Spans have two orthogonal attributes:
523 /// - They can be *primary spans*. In this case they are the locus of
524 /// the error, and would be rendered with `^^^`.
525 /// - They can have a *label*. In this case, the label is written next
526 /// to the mark in the snippet when we render.
527 #[derive(Clone, Debug, Hash, PartialEq, Eq, Encodable, Decodable)]
528 pub struct MultiSpan {
529 primary_spans: Vec<Span>,
530 span_labels: Vec<(Span, String)>,
535 pub fn lo(self) -> BytePos {
539 pub fn with_lo(self, lo: BytePos) -> Span {
540 self.data().with_lo(lo)
543 pub fn hi(self) -> BytePos {
547 pub fn with_hi(self, hi: BytePos) -> Span {
548 self.data().with_hi(hi)
551 pub fn ctxt(self) -> SyntaxContext {
552 self.data_untracked().ctxt
555 pub fn with_ctxt(self, ctxt: SyntaxContext) -> Span {
556 self.data_untracked().with_ctxt(ctxt)
559 pub fn parent(self) -> Option<LocalDefId> {
563 pub fn with_parent(self, ctxt: Option<LocalDefId>) -> Span {
564 self.data().with_parent(ctxt)
567 /// Returns `true` if this is a dummy span with any hygienic context.
569 pub fn is_dummy(self) -> bool {
570 self.data_untracked().is_dummy()
573 /// Returns `true` if this span comes from a macro or desugaring.
575 pub fn from_expansion(self) -> bool {
576 self.ctxt() != SyntaxContext::root()
579 /// Returns `true` if `span` originates in a derive-macro's expansion.
580 pub fn in_derive_expansion(self) -> bool {
581 matches!(self.ctxt().outer_expn_data().kind, ExpnKind::Macro(MacroKind::Derive, _))
584 /// Gate suggestions that would not be appropriate in a context the user didn't write.
585 pub fn can_be_used_for_suggestions(self) -> bool {
586 !self.from_expansion()
587 // FIXME: If this span comes from a `derive` macro but it points at code the user wrote,
588 // the callsite span and the span will be pointing at different places. It also means that
589 // we can safely provide suggestions on this span.
590 || (matches!(self.ctxt().outer_expn_data().kind, ExpnKind::Macro(MacroKind::Derive, _))
591 && self.parent_callsite().map(|p| (p.lo(), p.hi())) != Some((self.lo(), self.hi())))
595 pub fn with_root_ctxt(lo: BytePos, hi: BytePos) -> Span {
596 Span::new(lo, hi, SyntaxContext::root(), None)
599 /// Returns a new span representing an empty span at the beginning of this span.
601 pub fn shrink_to_lo(self) -> Span {
602 let span = self.data_untracked();
603 span.with_hi(span.lo)
605 /// Returns a new span representing an empty span at the end of this span.
607 pub fn shrink_to_hi(self) -> Span {
608 let span = self.data_untracked();
609 span.with_lo(span.hi)
613 /// Returns `true` if `hi == lo`.
614 pub fn is_empty(&self) -> bool {
615 let span = self.data_untracked();
619 /// Returns `self` if `self` is not the dummy span, and `other` otherwise.
620 pub fn substitute_dummy(self, other: Span) -> Span {
621 if self.is_dummy() { other } else { self }
624 /// Returns `true` if `self` fully encloses `other`.
625 pub fn contains(self, other: Span) -> bool {
626 let span = self.data();
627 let other = other.data();
631 /// Returns `true` if `self` touches `other`.
632 pub fn overlaps(self, other: Span) -> bool {
633 let span = self.data();
634 let other = other.data();
635 span.lo < other.hi && other.lo < span.hi
638 /// Returns `true` if the spans are equal with regards to the source text.
640 /// Use this instead of `==` when either span could be generated code,
641 /// and you only care that they point to the same bytes of source text.
642 pub fn source_equal(&self, other: &Span) -> bool {
643 let span = self.data();
644 let other = other.data();
645 span.lo == other.lo && span.hi == other.hi
648 /// Returns `Some(span)`, where the start is trimmed by the end of `other`.
649 pub fn trim_start(self, other: Span) -> Option<Span> {
650 let span = self.data();
651 let other = other.data();
652 if span.hi > other.hi { Some(span.with_lo(cmp::max(span.lo, other.hi))) } else { None }
655 /// Returns the source span -- this is either the supplied span, or the span for
656 /// the macro callsite that expanded to it.
657 pub fn source_callsite(self) -> Span {
658 let expn_data = self.ctxt().outer_expn_data();
659 if !expn_data.is_root() { expn_data.call_site.source_callsite() } else { self }
662 /// The `Span` for the tokens in the previous macro expansion from which `self` was generated,
664 pub fn parent_callsite(self) -> Option<Span> {
665 let expn_data = self.ctxt().outer_expn_data();
666 if !expn_data.is_root() { Some(expn_data.call_site) } else { None }
669 /// Walk down the expansion ancestors to find a span that's contained within `outer`.
670 pub fn find_ancestor_inside(mut self, outer: Span) -> Option<Span> {
671 while !outer.contains(self) {
672 self = self.parent_callsite()?;
677 /// Edition of the crate from which this span came.
678 pub fn edition(self) -> edition::Edition {
679 self.ctxt().edition()
683 pub fn rust_2015(&self) -> bool {
684 self.edition() == edition::Edition::Edition2015
688 pub fn rust_2018(&self) -> bool {
689 self.edition() >= edition::Edition::Edition2018
693 pub fn rust_2021(&self) -> bool {
694 self.edition() >= edition::Edition::Edition2021
697 /// Returns the source callee.
699 /// Returns `None` if the supplied span has no expansion trace,
700 /// else returns the `ExpnData` for the macro definition
701 /// corresponding to the source callsite.
702 pub fn source_callee(self) -> Option<ExpnData> {
703 fn source_callee(expn_data: ExpnData) -> ExpnData {
704 let next_expn_data = expn_data.call_site.ctxt().outer_expn_data();
705 if !next_expn_data.is_root() { source_callee(next_expn_data) } else { expn_data }
707 let expn_data = self.ctxt().outer_expn_data();
708 if !expn_data.is_root() { Some(source_callee(expn_data)) } else { None }
711 /// Checks if a span is "internal" to a macro in which `#[unstable]`
712 /// items can be used (that is, a macro marked with
713 /// `#[allow_internal_unstable]`).
714 pub fn allows_unstable(&self, feature: Symbol) -> bool {
717 .allow_internal_unstable
718 .map_or(false, |features| features.iter().any(|&f| f == feature))
721 /// Checks if this span arises from a compiler desugaring of kind `kind`.
722 pub fn is_desugaring(&self, kind: DesugaringKind) -> bool {
723 match self.ctxt().outer_expn_data().kind {
724 ExpnKind::Desugaring(k) => k == kind,
729 /// Returns the compiler desugaring that created this span, or `None`
730 /// if this span is not from a desugaring.
731 pub fn desugaring_kind(&self) -> Option<DesugaringKind> {
732 match self.ctxt().outer_expn_data().kind {
733 ExpnKind::Desugaring(k) => Some(k),
738 /// Checks if a span is "internal" to a macro in which `unsafe`
739 /// can be used without triggering the `unsafe_code` lint.
740 // (that is, a macro marked with `#[allow_internal_unsafe]`).
741 pub fn allows_unsafe(&self) -> bool {
742 self.ctxt().outer_expn_data().allow_internal_unsafe
745 pub fn macro_backtrace(mut self) -> impl Iterator<Item = ExpnData> {
746 let mut prev_span = DUMMY_SP;
747 std::iter::from_fn(move || {
749 let expn_data = self.ctxt().outer_expn_data();
750 if expn_data.is_root() {
754 let is_recursive = expn_data.call_site.source_equal(&prev_span);
757 self = expn_data.call_site;
759 // Don't print recursive invocations.
761 return Some(expn_data);
767 /// Returns a `Span` that would enclose both `self` and `end`.
771 /// self lorem ipsum end
772 /// ^^^^^^^^^^^^^^^^^^^^
774 pub fn to(self, end: Span) -> Span {
775 let span_data = self.data();
776 let end_data = end.data();
777 // FIXME(jseyfried): `self.ctxt` should always equal `end.ctxt` here (cf. issue #23480).
778 // Return the macro span on its own to avoid weird diagnostic output. It is preferable to
779 // have an incomplete span than a completely nonsensical one.
780 if span_data.ctxt != end_data.ctxt {
781 if span_data.ctxt == SyntaxContext::root() {
783 } else if end_data.ctxt == SyntaxContext::root() {
786 // Both spans fall within a macro.
787 // FIXME(estebank): check if it is the *same* macro.
790 cmp::min(span_data.lo, end_data.lo),
791 cmp::max(span_data.hi, end_data.hi),
792 if span_data.ctxt == SyntaxContext::root() { end_data.ctxt } else { span_data.ctxt },
793 if span_data.parent == end_data.parent { span_data.parent } else { None },
797 /// Returns a `Span` between the end of `self` to the beginning of `end`.
801 /// self lorem ipsum end
804 pub fn between(self, end: Span) -> Span {
805 let span = self.data();
806 let end = end.data();
810 if end.ctxt == SyntaxContext::root() { end.ctxt } else { span.ctxt },
811 if span.parent == end.parent { span.parent } else { None },
815 /// Returns a `Span` from the beginning of `self` until the beginning of `end`.
819 /// self lorem ipsum end
820 /// ^^^^^^^^^^^^^^^^^
822 pub fn until(self, end: Span) -> Span {
823 // Most of this function's body is copied from `to`.
824 // We can't just do `self.to(end.shrink_to_lo())`,
825 // because to also does some magic where it uses min/max so
826 // it can handle overlapping spans. Some advanced mis-use of
827 // `until` with different ctxts makes this visible.
828 let span_data = self.data();
829 let end_data = end.data();
830 // FIXME(jseyfried): `self.ctxt` should always equal `end.ctxt` here (cf. issue #23480).
831 // Return the macro span on its own to avoid weird diagnostic output. It is preferable to
832 // have an incomplete span than a completely nonsensical one.
833 if span_data.ctxt != end_data.ctxt {
834 if span_data.ctxt == SyntaxContext::root() {
836 } else if end_data.ctxt == SyntaxContext::root() {
839 // Both spans fall within a macro.
840 // FIXME(estebank): check if it is the *same* macro.
845 if end_data.ctxt == SyntaxContext::root() { end_data.ctxt } else { span_data.ctxt },
846 if span_data.parent == end_data.parent { span_data.parent } else { None },
850 pub fn from_inner(self, inner: InnerSpan) -> Span {
851 let span = self.data();
853 span.lo + BytePos::from_usize(inner.start),
854 span.lo + BytePos::from_usize(inner.end),
860 /// Equivalent of `Span::def_site` from the proc macro API,
861 /// except that the location is taken from the `self` span.
862 pub fn with_def_site_ctxt(self, expn_id: ExpnId) -> Span {
863 self.with_ctxt_from_mark(expn_id, Transparency::Opaque)
866 /// Equivalent of `Span::call_site` from the proc macro API,
867 /// except that the location is taken from the `self` span.
868 pub fn with_call_site_ctxt(&self, expn_id: ExpnId) -> Span {
869 self.with_ctxt_from_mark(expn_id, Transparency::Transparent)
872 /// Equivalent of `Span::mixed_site` from the proc macro API,
873 /// except that the location is taken from the `self` span.
874 pub fn with_mixed_site_ctxt(&self, expn_id: ExpnId) -> Span {
875 self.with_ctxt_from_mark(expn_id, Transparency::SemiTransparent)
878 /// Produces a span with the same location as `self` and context produced by a macro with the
879 /// given ID and transparency, assuming that macro was defined directly and not produced by
880 /// some other macro (which is the case for built-in and procedural macros).
881 pub fn with_ctxt_from_mark(self, expn_id: ExpnId, transparency: Transparency) -> Span {
882 self.with_ctxt(SyntaxContext::root().apply_mark(expn_id, transparency))
886 pub fn apply_mark(self, expn_id: ExpnId, transparency: Transparency) -> Span {
887 let span = self.data();
888 span.with_ctxt(span.ctxt.apply_mark(expn_id, transparency))
892 pub fn remove_mark(&mut self) -> ExpnId {
893 let mut span = self.data();
894 let mark = span.ctxt.remove_mark();
895 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
900 pub fn adjust(&mut self, expn_id: ExpnId) -> Option<ExpnId> {
901 let mut span = self.data();
902 let mark = span.ctxt.adjust(expn_id);
903 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
908 pub fn normalize_to_macros_2_0_and_adjust(&mut self, expn_id: ExpnId) -> Option<ExpnId> {
909 let mut span = self.data();
910 let mark = span.ctxt.normalize_to_macros_2_0_and_adjust(expn_id);
911 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
916 pub fn glob_adjust(&mut self, expn_id: ExpnId, glob_span: Span) -> Option<Option<ExpnId>> {
917 let mut span = self.data();
918 let mark = span.ctxt.glob_adjust(expn_id, glob_span);
919 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
924 pub fn reverse_glob_adjust(
928 ) -> Option<Option<ExpnId>> {
929 let mut span = self.data();
930 let mark = span.ctxt.reverse_glob_adjust(expn_id, glob_span);
931 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
936 pub fn normalize_to_macros_2_0(self) -> Span {
937 let span = self.data();
938 span.with_ctxt(span.ctxt.normalize_to_macros_2_0())
942 pub fn normalize_to_macro_rules(self) -> Span {
943 let span = self.data();
944 span.with_ctxt(span.ctxt.normalize_to_macro_rules())
948 /// A span together with some additional data.
949 #[derive(Clone, Debug)]
950 pub struct SpanLabel {
951 /// The span we are going to include in the final snippet.
954 /// Is this a primary span? This is the "locus" of the message,
955 /// and is indicated with a `^^^^` underline, versus `----`.
956 pub is_primary: bool,
958 /// What label should we attach to this span (if any)?
959 pub label: Option<String>,
962 impl Default for Span {
963 fn default() -> Self {
968 impl<E: Encoder> Encodable<E> for Span {
969 default fn encode(&self, s: &mut E) -> Result<(), E::Error> {
970 let span = self.data();
971 s.emit_struct(false, |s| {
972 s.emit_struct_field("lo", true, |s| span.lo.encode(s))?;
973 s.emit_struct_field("hi", false, |s| span.hi.encode(s))
977 impl<D: Decoder> Decodable<D> for Span {
978 default fn decode(s: &mut D) -> Span {
980 let lo = d.read_struct_field("lo", Decodable::decode);
981 let hi = d.read_struct_field("hi", Decodable::decode);
983 Span::new(lo, hi, SyntaxContext::root(), None)
988 /// Calls the provided closure, using the provided `SourceMap` to format
989 /// any spans that are debug-printed during the closure's execution.
991 /// Normally, the global `TyCtxt` is used to retrieve the `SourceMap`
992 /// (see `rustc_interface::callbacks::span_debug1`). However, some parts
993 /// of the compiler (e.g. `rustc_parse`) may debug-print `Span`s before
994 /// a `TyCtxt` is available. In this case, we fall back to
995 /// the `SourceMap` provided to this function. If that is not available,
996 /// we fall back to printing the raw `Span` field values.
997 pub fn with_source_map<T, F: FnOnce() -> T>(source_map: Lrc<SourceMap>, f: F) -> T {
998 with_session_globals(|session_globals| {
999 *session_globals.source_map.borrow_mut() = Some(source_map);
1001 struct ClearSourceMap;
1002 impl Drop for ClearSourceMap {
1003 fn drop(&mut self) {
1004 with_session_globals(|session_globals| {
1005 session_globals.source_map.borrow_mut().take();
1010 let _guard = ClearSourceMap;
1014 pub fn debug_with_source_map(
1016 f: &mut fmt::Formatter<'_>,
1017 source_map: &SourceMap,
1019 write!(f, "{} ({:?})", source_map.span_to_diagnostic_string(span), span.ctxt())
1022 pub fn default_span_debug(span: Span, f: &mut fmt::Formatter<'_>) -> fmt::Result {
1023 with_session_globals(|session_globals| {
1024 if let Some(source_map) = &*session_globals.source_map.borrow() {
1025 debug_with_source_map(span, f, source_map)
1027 f.debug_struct("Span")
1028 .field("lo", &span.lo())
1029 .field("hi", &span.hi())
1030 .field("ctxt", &span.ctxt())
1036 impl fmt::Debug for Span {
1037 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
1038 (*SPAN_DEBUG)(*self, f)
1042 impl fmt::Debug for SpanData {
1043 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
1044 (*SPAN_DEBUG)(Span::new(self.lo, self.hi, self.ctxt, self.parent), f)
1050 pub fn new() -> MultiSpan {
1051 MultiSpan { primary_spans: vec![], span_labels: vec![] }
1054 pub fn from_span(primary_span: Span) -> MultiSpan {
1055 MultiSpan { primary_spans: vec![primary_span], span_labels: vec![] }
1058 pub fn from_spans(mut vec: Vec<Span>) -> MultiSpan {
1060 MultiSpan { primary_spans: vec, span_labels: vec![] }
1063 pub fn push_span_label(&mut self, span: Span, label: String) {
1064 self.span_labels.push((span, label));
1067 /// Selects the first primary span (if any).
1068 pub fn primary_span(&self) -> Option<Span> {
1069 self.primary_spans.first().cloned()
1072 /// Returns all primary spans.
1073 pub fn primary_spans(&self) -> &[Span] {
1077 /// Returns `true` if any of the primary spans are displayable.
1078 pub fn has_primary_spans(&self) -> bool {
1079 self.primary_spans.iter().any(|sp| !sp.is_dummy())
1082 /// Returns `true` if this contains only a dummy primary span with any hygienic context.
1083 pub fn is_dummy(&self) -> bool {
1084 let mut is_dummy = true;
1085 for span in &self.primary_spans {
1086 if !span.is_dummy() {
1093 /// Replaces all occurrences of one Span with another. Used to move `Span`s in areas that don't
1094 /// display well (like std macros). Returns whether replacements occurred.
1095 pub fn replace(&mut self, before: Span, after: Span) -> bool {
1096 let mut replacements_occurred = false;
1097 for primary_span in &mut self.primary_spans {
1098 if *primary_span == before {
1099 *primary_span = after;
1100 replacements_occurred = true;
1103 for span_label in &mut self.span_labels {
1104 if span_label.0 == before {
1105 span_label.0 = after;
1106 replacements_occurred = true;
1109 replacements_occurred
1112 /// Returns the strings to highlight. We always ensure that there
1113 /// is an entry for each of the primary spans -- for each primary
1114 /// span `P`, if there is at least one label with span `P`, we return
1115 /// those labels (marked as primary). But otherwise we return
1116 /// `SpanLabel` instances with empty labels.
1117 pub fn span_labels(&self) -> Vec<SpanLabel> {
1118 let is_primary = |span| self.primary_spans.contains(&span);
1120 let mut span_labels = self
1123 .map(|&(span, ref label)| SpanLabel {
1125 is_primary: is_primary(span),
1126 label: Some(label.clone()),
1128 .collect::<Vec<_>>();
1130 for &span in &self.primary_spans {
1131 if !span_labels.iter().any(|sl| sl.span == span) {
1132 span_labels.push(SpanLabel { span, is_primary: true, label: None });
1139 /// Returns `true` if any of the span labels is displayable.
1140 pub fn has_span_labels(&self) -> bool {
1141 self.span_labels.iter().any(|(sp, _)| !sp.is_dummy())
1145 impl From<Span> for MultiSpan {
1146 fn from(span: Span) -> MultiSpan {
1147 MultiSpan::from_span(span)
1151 impl From<Vec<Span>> for MultiSpan {
1152 fn from(spans: Vec<Span>) -> MultiSpan {
1153 MultiSpan::from_spans(spans)
1157 /// Identifies an offset of a multi-byte character in a `SourceFile`.
1158 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1159 pub struct MultiByteChar {
1160 /// The absolute offset of the character in the `SourceMap`.
1162 /// The number of bytes, `>= 2`.
1166 /// Identifies an offset of a non-narrow character in a `SourceFile`.
1167 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1168 pub enum NonNarrowChar {
1169 /// Represents a zero-width character.
1171 /// Represents a wide (full-width) character.
1173 /// Represents a tab character, represented visually with a width of 4 characters.
1177 impl NonNarrowChar {
1178 fn new(pos: BytePos, width: usize) -> Self {
1180 0 => NonNarrowChar::ZeroWidth(pos),
1181 2 => NonNarrowChar::Wide(pos),
1182 4 => NonNarrowChar::Tab(pos),
1183 _ => panic!("width {} given for non-narrow character", width),
1187 /// Returns the absolute offset of the character in the `SourceMap`.
1188 pub fn pos(&self) -> BytePos {
1190 NonNarrowChar::ZeroWidth(p) | NonNarrowChar::Wide(p) | NonNarrowChar::Tab(p) => p,
1194 /// Returns the width of the character, 0 (zero-width) or 2 (wide).
1195 pub fn width(&self) -> usize {
1197 NonNarrowChar::ZeroWidth(_) => 0,
1198 NonNarrowChar::Wide(_) => 2,
1199 NonNarrowChar::Tab(_) => 4,
1204 impl Add<BytePos> for NonNarrowChar {
1207 fn add(self, rhs: BytePos) -> Self {
1209 NonNarrowChar::ZeroWidth(pos) => NonNarrowChar::ZeroWidth(pos + rhs),
1210 NonNarrowChar::Wide(pos) => NonNarrowChar::Wide(pos + rhs),
1211 NonNarrowChar::Tab(pos) => NonNarrowChar::Tab(pos + rhs),
1216 impl Sub<BytePos> for NonNarrowChar {
1219 fn sub(self, rhs: BytePos) -> Self {
1221 NonNarrowChar::ZeroWidth(pos) => NonNarrowChar::ZeroWidth(pos - rhs),
1222 NonNarrowChar::Wide(pos) => NonNarrowChar::Wide(pos - rhs),
1223 NonNarrowChar::Tab(pos) => NonNarrowChar::Tab(pos - rhs),
1228 /// Identifies an offset of a character that was normalized away from `SourceFile`.
1229 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1230 pub struct NormalizedPos {
1231 /// The absolute offset of the character in the `SourceMap`.
1233 /// The difference between original and normalized string at position.
1237 #[derive(PartialEq, Eq, Clone, Debug)]
1238 pub enum ExternalSource {
1239 /// No external source has to be loaded, since the `SourceFile` represents a local crate.
1242 kind: ExternalSourceKind,
1243 /// This SourceFile's byte-offset within the source_map of its original crate.
1244 original_start_pos: BytePos,
1245 /// The end of this SourceFile within the source_map of its original crate.
1246 original_end_pos: BytePos,
1250 /// The state of the lazy external source loading mechanism of a `SourceFile`.
1251 #[derive(PartialEq, Eq, Clone, Debug)]
1252 pub enum ExternalSourceKind {
1253 /// The external source has been loaded already.
1254 Present(Lrc<String>),
1255 /// No attempt has been made to load the external source.
1257 /// A failed attempt has been made to load the external source.
1262 impl ExternalSource {
1263 pub fn get_source(&self) -> Option<&Lrc<String>> {
1265 ExternalSource::Foreign { kind: ExternalSourceKind::Present(ref src), .. } => Some(src),
1272 pub struct OffsetOverflowError;
1274 #[derive(Copy, Clone, Debug, PartialEq, Eq, PartialOrd, Ord, Hash, Encodable, Decodable)]
1275 pub enum SourceFileHashAlgorithm {
1281 impl FromStr for SourceFileHashAlgorithm {
1284 fn from_str(s: &str) -> Result<SourceFileHashAlgorithm, ()> {
1286 "md5" => Ok(SourceFileHashAlgorithm::Md5),
1287 "sha1" => Ok(SourceFileHashAlgorithm::Sha1),
1288 "sha256" => Ok(SourceFileHashAlgorithm::Sha256),
1294 rustc_data_structures::impl_stable_hash_via_hash!(SourceFileHashAlgorithm);
1296 /// The hash of the on-disk source file used for debug info.
1297 #[derive(Copy, Clone, PartialEq, Eq, Debug)]
1298 #[derive(HashStable_Generic, Encodable, Decodable)]
1299 pub struct SourceFileHash {
1300 pub kind: SourceFileHashAlgorithm,
1304 impl SourceFileHash {
1305 pub fn new(kind: SourceFileHashAlgorithm, src: &str) -> SourceFileHash {
1306 let mut hash = SourceFileHash { kind, value: Default::default() };
1307 let len = hash.hash_len();
1308 let value = &mut hash.value[..len];
1309 let data = src.as_bytes();
1311 SourceFileHashAlgorithm::Md5 => {
1312 value.copy_from_slice(&Md5::digest(data));
1314 SourceFileHashAlgorithm::Sha1 => {
1315 value.copy_from_slice(&Sha1::digest(data));
1317 SourceFileHashAlgorithm::Sha256 => {
1318 value.copy_from_slice(&Sha256::digest(data));
1324 /// Check if the stored hash matches the hash of the string.
1325 pub fn matches(&self, src: &str) -> bool {
1326 Self::new(self.kind, src) == *self
1329 /// The bytes of the hash.
1330 pub fn hash_bytes(&self) -> &[u8] {
1331 let len = self.hash_len();
1335 fn hash_len(&self) -> usize {
1337 SourceFileHashAlgorithm::Md5 => 16,
1338 SourceFileHashAlgorithm::Sha1 => 20,
1339 SourceFileHashAlgorithm::Sha256 => 32,
1344 /// A single source in the [`SourceMap`].
1346 pub struct SourceFile {
1347 /// The name of the file that the source came from. Source that doesn't
1348 /// originate from files has names between angle brackets by convention
1349 /// (e.g., `<anon>`).
1351 /// The complete source code.
1352 pub src: Option<Lrc<String>>,
1353 /// The source code's hash.
1354 pub src_hash: SourceFileHash,
1355 /// The external source code (used for external crates, which will have a `None`
1356 /// value as `self.src`.
1357 pub external_src: Lock<ExternalSource>,
1358 /// The start position of this source in the `SourceMap`.
1359 pub start_pos: BytePos,
1360 /// The end position of this source in the `SourceMap`.
1361 pub end_pos: BytePos,
1362 /// Locations of lines beginnings in the source code.
1363 pub lines: Vec<BytePos>,
1364 /// Locations of multi-byte characters in the source code.
1365 pub multibyte_chars: Vec<MultiByteChar>,
1366 /// Width of characters that are not narrow in the source code.
1367 pub non_narrow_chars: Vec<NonNarrowChar>,
1368 /// Locations of characters removed during normalization.
1369 pub normalized_pos: Vec<NormalizedPos>,
1370 /// A hash of the filename, used for speeding up hashing in incremental compilation.
1371 pub name_hash: u128,
1372 /// Indicates which crate this `SourceFile` was imported from.
1376 impl<S: Encoder> Encodable<S> for SourceFile {
1377 fn encode(&self, s: &mut S) -> Result<(), S::Error> {
1378 s.emit_struct(false, |s| {
1379 s.emit_struct_field("name", true, |s| self.name.encode(s))?;
1380 s.emit_struct_field("src_hash", false, |s| self.src_hash.encode(s))?;
1381 s.emit_struct_field("start_pos", false, |s| self.start_pos.encode(s))?;
1382 s.emit_struct_field("end_pos", false, |s| self.end_pos.encode(s))?;
1383 s.emit_struct_field("lines", false, |s| {
1384 let lines = &self.lines[..];
1385 // Store the length.
1386 s.emit_u32(lines.len() as u32)?;
1388 if !lines.is_empty() {
1389 // In order to preserve some space, we exploit the fact that
1390 // the lines list is sorted and individual lines are
1391 // probably not that long. Because of that we can store lines
1392 // as a difference list, using as little space as possible
1393 // for the differences.
1394 let max_line_length = if lines.len() == 1 {
1399 .map(|&[fst, snd]| snd - fst)
1400 .map(|bp| bp.to_usize())
1405 let bytes_per_diff: u8 = match max_line_length {
1407 0x100..=0xFFFF => 2,
1411 // Encode the number of bytes used per diff.
1412 bytes_per_diff.encode(s)?;
1414 // Encode the first element.
1415 lines[0].encode(s)?;
1417 let diff_iter = lines.array_windows().map(|&[fst, snd]| snd - fst);
1419 match bytes_per_diff {
1421 for diff in diff_iter {
1422 (diff.0 as u8).encode(s)?
1426 for diff in diff_iter {
1427 (diff.0 as u16).encode(s)?
1431 for diff in diff_iter {
1435 _ => unreachable!(),
1441 s.emit_struct_field("multibyte_chars", false, |s| self.multibyte_chars.encode(s))?;
1442 s.emit_struct_field("non_narrow_chars", false, |s| self.non_narrow_chars.encode(s))?;
1443 s.emit_struct_field("name_hash", false, |s| self.name_hash.encode(s))?;
1444 s.emit_struct_field("normalized_pos", false, |s| self.normalized_pos.encode(s))?;
1445 s.emit_struct_field("cnum", false, |s| self.cnum.encode(s))
1450 impl<D: Decoder> Decodable<D> for SourceFile {
1451 fn decode(d: &mut D) -> SourceFile {
1453 let name: FileName = d.read_struct_field("name", |d| Decodable::decode(d));
1454 let src_hash: SourceFileHash =
1455 d.read_struct_field("src_hash", |d| Decodable::decode(d));
1456 let start_pos: BytePos = d.read_struct_field("start_pos", |d| Decodable::decode(d));
1457 let end_pos: BytePos = d.read_struct_field("end_pos", |d| Decodable::decode(d));
1458 let lines: Vec<BytePos> = d.read_struct_field("lines", |d| {
1459 let num_lines: u32 = Decodable::decode(d);
1460 let mut lines = Vec::with_capacity(num_lines as usize);
1463 // Read the number of bytes used per diff.
1464 let bytes_per_diff: u8 = Decodable::decode(d);
1466 // Read the first element.
1467 let mut line_start: BytePos = Decodable::decode(d);
1468 lines.push(line_start);
1470 for _ in 1..num_lines {
1471 let diff = match bytes_per_diff {
1472 1 => d.read_u8() as u32,
1473 2 => d.read_u16() as u32,
1475 _ => unreachable!(),
1478 line_start = line_start + BytePos(diff);
1480 lines.push(line_start);
1486 let multibyte_chars: Vec<MultiByteChar> =
1487 d.read_struct_field("multibyte_chars", |d| Decodable::decode(d));
1488 let non_narrow_chars: Vec<NonNarrowChar> =
1489 d.read_struct_field("non_narrow_chars", |d| Decodable::decode(d));
1490 let name_hash: u128 = d.read_struct_field("name_hash", |d| Decodable::decode(d));
1491 let normalized_pos: Vec<NormalizedPos> =
1492 d.read_struct_field("normalized_pos", |d| Decodable::decode(d));
1493 let cnum: CrateNum = d.read_struct_field("cnum", |d| Decodable::decode(d));
1500 // Unused - the metadata decoder will construct
1501 // a new SourceFile, filling in `external_src` properly
1502 external_src: Lock::new(ExternalSource::Unneeded),
1514 impl fmt::Debug for SourceFile {
1515 fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result {
1516 write!(fmt, "SourceFile({:?})", self.name)
1525 hash_kind: SourceFileHashAlgorithm,
1527 // Compute the file hash before any normalization.
1528 let src_hash = SourceFileHash::new(hash_kind, &src);
1529 let normalized_pos = normalize_src(&mut src, start_pos);
1532 let mut hasher: StableHasher = StableHasher::new();
1533 name.hash(&mut hasher);
1534 hasher.finish::<u128>()
1536 let end_pos = start_pos.to_usize() + src.len();
1537 assert!(end_pos <= u32::MAX as usize);
1539 let (lines, multibyte_chars, non_narrow_chars) =
1540 analyze_source_file::analyze_source_file(&src, start_pos);
1544 src: Some(Lrc::new(src)),
1546 external_src: Lock::new(ExternalSource::Unneeded),
1548 end_pos: Pos::from_usize(end_pos),
1558 /// Returns the `BytePos` of the beginning of the current line.
1559 pub fn line_begin_pos(&self, pos: BytePos) -> BytePos {
1560 let line_index = self.lookup_line(pos).unwrap();
1561 self.lines[line_index]
1564 /// Add externally loaded source.
1565 /// If the hash of the input doesn't match or no input is supplied via None,
1566 /// it is interpreted as an error and the corresponding enum variant is set.
1567 /// The return value signifies whether some kind of source is present.
1568 pub fn add_external_src<F>(&self, get_src: F) -> bool
1570 F: FnOnce() -> Option<String>,
1573 *self.external_src.borrow(),
1574 ExternalSource::Foreign { kind: ExternalSourceKind::AbsentOk, .. }
1576 let src = get_src();
1577 let mut external_src = self.external_src.borrow_mut();
1578 // Check that no-one else have provided the source while we were getting it
1579 if let ExternalSource::Foreign {
1580 kind: src_kind @ ExternalSourceKind::AbsentOk, ..
1581 } = &mut *external_src
1583 if let Some(mut src) = src {
1584 // The src_hash needs to be computed on the pre-normalized src.
1585 if self.src_hash.matches(&src) {
1586 normalize_src(&mut src, BytePos::from_usize(0));
1587 *src_kind = ExternalSourceKind::Present(Lrc::new(src));
1591 *src_kind = ExternalSourceKind::AbsentErr;
1596 self.src.is_some() || external_src.get_source().is_some()
1599 self.src.is_some() || self.external_src.borrow().get_source().is_some()
1603 /// Gets a line from the list of pre-computed line-beginnings.
1604 /// The line number here is 0-based.
1605 pub fn get_line(&self, line_number: usize) -> Option<Cow<'_, str>> {
1606 fn get_until_newline(src: &str, begin: usize) -> &str {
1607 // We can't use `lines.get(line_number+1)` because we might
1608 // be parsing when we call this function and thus the current
1609 // line is the last one we have line info for.
1610 let slice = &src[begin..];
1611 match slice.find('\n') {
1612 Some(e) => &slice[..e],
1618 let line = self.lines.get(line_number)?;
1619 let begin: BytePos = *line - self.start_pos;
1623 if let Some(ref src) = self.src {
1624 Some(Cow::from(get_until_newline(src, begin)))
1625 } else if let Some(src) = self.external_src.borrow().get_source() {
1626 Some(Cow::Owned(String::from(get_until_newline(src, begin))))
1632 pub fn is_real_file(&self) -> bool {
1636 pub fn is_imported(&self) -> bool {
1640 pub fn count_lines(&self) -> usize {
1644 /// Finds the line containing the given position. The return value is the
1645 /// index into the `lines` array of this `SourceFile`, not the 1-based line
1646 /// number. If the source_file is empty or the position is located before the
1647 /// first line, `None` is returned.
1648 pub fn lookup_line(&self, pos: BytePos) -> Option<usize> {
1649 match self.lines.binary_search(&pos) {
1650 Ok(idx) => Some(idx),
1652 Err(idx) => Some(idx - 1),
1656 pub fn line_bounds(&self, line_index: usize) -> Range<BytePos> {
1657 if self.is_empty() {
1658 return self.start_pos..self.end_pos;
1661 assert!(line_index < self.lines.len());
1662 if line_index == (self.lines.len() - 1) {
1663 self.lines[line_index]..self.end_pos
1665 self.lines[line_index]..self.lines[line_index + 1]
1669 /// Returns whether or not the file contains the given `SourceMap` byte
1670 /// position. The position one past the end of the file is considered to be
1671 /// contained by the file. This implies that files for which `is_empty`
1672 /// returns true still contain one byte position according to this function.
1674 pub fn contains(&self, byte_pos: BytePos) -> bool {
1675 byte_pos >= self.start_pos && byte_pos <= self.end_pos
1679 pub fn is_empty(&self) -> bool {
1680 self.start_pos == self.end_pos
1683 /// Calculates the original byte position relative to the start of the file
1684 /// based on the given byte position.
1685 pub fn original_relative_byte_pos(&self, pos: BytePos) -> BytePos {
1686 // Diff before any records is 0. Otherwise use the previously recorded
1687 // diff as that applies to the following characters until a new diff
1689 let diff = match self.normalized_pos.binary_search_by(|np| np.pos.cmp(&pos)) {
1690 Ok(i) => self.normalized_pos[i].diff,
1691 Err(i) if i == 0 => 0,
1692 Err(i) => self.normalized_pos[i - 1].diff,
1695 BytePos::from_u32(pos.0 - self.start_pos.0 + diff)
1698 /// Converts an absolute `BytePos` to a `CharPos` relative to the `SourceFile`.
1699 pub fn bytepos_to_file_charpos(&self, bpos: BytePos) -> CharPos {
1700 // The number of extra bytes due to multibyte chars in the `SourceFile`.
1701 let mut total_extra_bytes = 0;
1703 for mbc in self.multibyte_chars.iter() {
1704 debug!("{}-byte char at {:?}", mbc.bytes, mbc.pos);
1706 // Every character is at least one byte, so we only
1707 // count the actual extra bytes.
1708 total_extra_bytes += mbc.bytes as u32 - 1;
1709 // We should never see a byte position in the middle of a
1711 assert!(bpos.to_u32() >= mbc.pos.to_u32() + mbc.bytes as u32);
1717 assert!(self.start_pos.to_u32() + total_extra_bytes <= bpos.to_u32());
1718 CharPos(bpos.to_usize() - self.start_pos.to_usize() - total_extra_bytes as usize)
1721 /// Looks up the file's (1-based) line number and (0-based `CharPos`) column offset, for a
1722 /// given `BytePos`.
1723 pub fn lookup_file_pos(&self, pos: BytePos) -> (usize, CharPos) {
1724 let chpos = self.bytepos_to_file_charpos(pos);
1725 match self.lookup_line(pos) {
1727 let line = a + 1; // Line numbers start at 1
1728 let linebpos = self.lines[a];
1729 let linechpos = self.bytepos_to_file_charpos(linebpos);
1730 let col = chpos - linechpos;
1731 debug!("byte pos {:?} is on the line at byte pos {:?}", pos, linebpos);
1732 debug!("char pos {:?} is on the line at char pos {:?}", chpos, linechpos);
1733 debug!("byte is on line: {}", line);
1734 assert!(chpos >= linechpos);
1741 /// Looks up the file's (1-based) line number, (0-based `CharPos`) column offset, and (0-based)
1742 /// column offset when displayed, for a given `BytePos`.
1743 pub fn lookup_file_pos_with_col_display(&self, pos: BytePos) -> (usize, CharPos, usize) {
1744 let (line, col_or_chpos) = self.lookup_file_pos(pos);
1746 let col = col_or_chpos;
1747 let linebpos = self.lines[line - 1];
1749 let start_width_idx = self
1751 .binary_search_by_key(&linebpos, |x| x.pos())
1752 .unwrap_or_else(|x| x);
1753 let end_width_idx = self
1755 .binary_search_by_key(&pos, |x| x.pos())
1756 .unwrap_or_else(|x| x);
1757 let special_chars = end_width_idx - start_width_idx;
1758 let non_narrow: usize = self.non_narrow_chars[start_width_idx..end_width_idx]
1762 col.0 - special_chars + non_narrow
1764 (line, col, col_display)
1766 let chpos = col_or_chpos;
1768 let end_width_idx = self
1770 .binary_search_by_key(&pos, |x| x.pos())
1771 .unwrap_or_else(|x| x);
1772 let non_narrow: usize =
1773 self.non_narrow_chars[0..end_width_idx].iter().map(|x| x.width()).sum();
1774 chpos.0 - end_width_idx + non_narrow
1776 (0, chpos, col_display)
1781 /// Normalizes the source code and records the normalizations.
1782 fn normalize_src(src: &mut String, start_pos: BytePos) -> Vec<NormalizedPos> {
1783 let mut normalized_pos = vec![];
1784 remove_bom(src, &mut normalized_pos);
1785 normalize_newlines(src, &mut normalized_pos);
1787 // Offset all the positions by start_pos to match the final file positions.
1788 for np in &mut normalized_pos {
1789 np.pos.0 += start_pos.0;
1795 /// Removes UTF-8 BOM, if any.
1796 fn remove_bom(src: &mut String, normalized_pos: &mut Vec<NormalizedPos>) {
1797 if src.starts_with('\u{feff}') {
1799 normalized_pos.push(NormalizedPos { pos: BytePos(0), diff: 3 });
1803 /// Replaces `\r\n` with `\n` in-place in `src`.
1805 /// Returns error if there's a lone `\r` in the string.
1806 fn normalize_newlines(src: &mut String, normalized_pos: &mut Vec<NormalizedPos>) {
1807 if !src.as_bytes().contains(&b'\r') {
1811 // We replace `\r\n` with `\n` in-place, which doesn't break utf-8 encoding.
1812 // While we *can* call `as_mut_vec` and do surgery on the live string
1813 // directly, let's rather steal the contents of `src`. This makes the code
1814 // safe even if a panic occurs.
1816 let mut buf = std::mem::replace(src, String::new()).into_bytes();
1817 let mut gap_len = 0;
1818 let mut tail = buf.as_mut_slice();
1820 let original_gap = normalized_pos.last().map_or(0, |l| l.diff);
1822 let idx = match find_crlf(&tail[gap_len..]) {
1824 Some(idx) => idx + gap_len,
1826 tail.copy_within(gap_len..idx, 0);
1827 tail = &mut tail[idx - gap_len..];
1828 if tail.len() == gap_len {
1831 cursor += idx - gap_len;
1833 normalized_pos.push(NormalizedPos {
1834 pos: BytePos::from_usize(cursor + 1),
1835 diff: original_gap + gap_len as u32,
1839 // Account for removed `\r`.
1840 // After `set_len`, `buf` is guaranteed to contain utf-8 again.
1841 let new_len = buf.len() - gap_len;
1843 buf.set_len(new_len);
1844 *src = String::from_utf8_unchecked(buf);
1847 fn find_crlf(src: &[u8]) -> Option<usize> {
1848 let mut search_idx = 0;
1849 while let Some(idx) = find_cr(&src[search_idx..]) {
1850 if src[search_idx..].get(idx + 1) != Some(&b'\n') {
1851 search_idx += idx + 1;
1854 return Some(search_idx + idx);
1859 fn find_cr(src: &[u8]) -> Option<usize> {
1860 src.iter().position(|&b| b == b'\r')
1864 // _____________________________________________________________________________
1865 // Pos, BytePos, CharPos
1869 fn from_usize(n: usize) -> Self;
1870 fn to_usize(&self) -> usize;
1871 fn from_u32(n: u32) -> Self;
1872 fn to_u32(&self) -> u32;
1875 macro_rules! impl_pos {
1879 $vis:vis struct $ident:ident($inner_vis:vis $inner_ty:ty);
1884 $vis struct $ident($inner_vis $inner_ty);
1886 impl Pos for $ident {
1888 fn from_usize(n: usize) -> $ident {
1889 $ident(n as $inner_ty)
1893 fn to_usize(&self) -> usize {
1898 fn from_u32(n: u32) -> $ident {
1899 $ident(n as $inner_ty)
1903 fn to_u32(&self) -> u32 {
1908 impl Add for $ident {
1909 type Output = $ident;
1912 fn add(self, rhs: $ident) -> $ident {
1913 $ident(self.0 + rhs.0)
1917 impl Sub for $ident {
1918 type Output = $ident;
1921 fn sub(self, rhs: $ident) -> $ident {
1922 $ident(self.0 - rhs.0)
1932 /// Keep this small (currently 32-bits), as AST contains a lot of them.
1933 #[derive(Clone, Copy, PartialEq, Eq, Hash, PartialOrd, Ord, Debug)]
1934 pub struct BytePos(pub u32);
1936 /// A character offset.
1938 /// Because of multibyte UTF-8 characters, a byte offset
1939 /// is not equivalent to a character offset. The [`SourceMap`] will convert [`BytePos`]
1940 /// values to `CharPos` values as necessary.
1941 #[derive(Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Debug)]
1942 pub struct CharPos(pub usize);
1945 impl<S: rustc_serialize::Encoder> Encodable<S> for BytePos {
1946 fn encode(&self, s: &mut S) -> Result<(), S::Error> {
1951 impl<D: rustc_serialize::Decoder> Decodable<D> for BytePos {
1952 fn decode(d: &mut D) -> BytePos {
1953 BytePos(d.read_u32())
1957 // _____________________________________________________________________________
1958 // Loc, SourceFileAndLine, SourceFileAndBytePos
1961 /// A source code location used for error reporting.
1962 #[derive(Debug, Clone)]
1964 /// Information about the original source.
1965 pub file: Lrc<SourceFile>,
1966 /// The (1-based) line number.
1968 /// The (0-based) column offset.
1970 /// The (0-based) column offset when displayed.
1971 pub col_display: usize,
1974 // Used to be structural records.
1976 pub struct SourceFileAndLine {
1977 pub sf: Lrc<SourceFile>,
1978 /// Index of line, starting from 0.
1982 pub struct SourceFileAndBytePos {
1983 pub sf: Lrc<SourceFile>,
1987 #[derive(Copy, Clone, Debug, PartialEq, Eq)]
1988 pub struct LineInfo {
1989 /// Index of line, starting from 0.
1990 pub line_index: usize,
1992 /// Column in line where span begins, starting from 0.
1993 pub start_col: CharPos,
1995 /// Column in line where span ends, starting from 0, exclusive.
1996 pub end_col: CharPos,
1999 pub struct FileLines {
2000 pub file: Lrc<SourceFile>,
2001 pub lines: Vec<LineInfo>,
2004 pub static SPAN_DEBUG: AtomicRef<fn(Span, &mut fmt::Formatter<'_>) -> fmt::Result> =
2005 AtomicRef::new(&(default_span_debug as fn(_, &mut fmt::Formatter<'_>) -> _));
2006 pub static SPAN_TRACK: AtomicRef<fn(LocalDefId)> = AtomicRef::new(&((|_| {}) as fn(_)));
2008 // _____________________________________________________________________________
2009 // SpanLinesError, SpanSnippetError, DistinctSources, MalformedSourceMapPositions
2012 pub type FileLinesResult = Result<FileLines, SpanLinesError>;
2014 #[derive(Clone, PartialEq, Eq, Debug)]
2015 pub enum SpanLinesError {
2016 DistinctSources(DistinctSources),
2019 #[derive(Clone, PartialEq, Eq, Debug)]
2020 pub enum SpanSnippetError {
2021 IllFormedSpan(Span),
2022 DistinctSources(DistinctSources),
2023 MalformedForSourcemap(MalformedSourceMapPositions),
2024 SourceNotAvailable { filename: FileName },
2027 #[derive(Clone, PartialEq, Eq, Debug)]
2028 pub struct DistinctSources {
2029 pub begin: (FileName, BytePos),
2030 pub end: (FileName, BytePos),
2033 #[derive(Clone, PartialEq, Eq, Debug)]
2034 pub struct MalformedSourceMapPositions {
2036 pub source_len: usize,
2037 pub begin_pos: BytePos,
2038 pub end_pos: BytePos,
2041 /// Range inside of a `Span` used for diagnostics when we only have access to relative positions.
2042 #[derive(Copy, Clone, PartialEq, Eq, Debug)]
2043 pub struct InnerSpan {
2049 pub fn new(start: usize, end: usize) -> InnerSpan {
2050 InnerSpan { start, end }
2054 /// Requirements for a `StableHashingContext` to be used in this crate.
2056 /// This is a hack to allow using the [`HashStable_Generic`] derive macro
2057 /// instead of implementing everything in rustc_middle.
2058 pub trait HashStableContext {
2059 fn def_path_hash(&self, def_id: DefId) -> DefPathHash;
2060 fn hash_spans(&self) -> bool;
2061 /// Accesses `sess.opts.debugging_opts.incremental_ignore_spans` since
2062 /// we don't have easy access to a `Session`
2063 fn debug_opts_incremental_ignore_spans(&self) -> bool;
2064 fn def_span(&self, def_id: LocalDefId) -> Span;
2065 fn span_data_to_lines_and_cols(
2068 ) -> Option<(Lrc<SourceFile>, usize, BytePos, usize, BytePos)>;
2069 fn hashing_controls(&self) -> HashingControls;
2072 impl<CTX> HashStable<CTX> for Span
2074 CTX: HashStableContext,
2076 /// Hashes a span in a stable way. We can't directly hash the span's `BytePos`
2077 /// fields (that would be similar to hashing pointers, since those are just
2078 /// offsets into the `SourceMap`). Instead, we hash the (file name, line, column)
2079 /// triple, which stays the same even if the containing `SourceFile` has moved
2080 /// within the `SourceMap`.
2082 /// Also note that we are hashing byte offsets for the column, not unicode
2083 /// codepoint offsets. For the purpose of the hash that's sufficient.
2084 /// Also, hashing filenames is expensive so we avoid doing it twice when the
2085 /// span starts and ends in the same file, which is almost always the case.
2086 fn hash_stable(&self, ctx: &mut CTX, hasher: &mut StableHasher) {
2087 const TAG_VALID_SPAN: u8 = 0;
2088 const TAG_INVALID_SPAN: u8 = 1;
2089 const TAG_RELATIVE_SPAN: u8 = 2;
2091 if !ctx.hash_spans() {
2095 let span = self.data_untracked();
2096 span.ctxt.hash_stable(ctx, hasher);
2097 span.parent.hash_stable(ctx, hasher);
2099 if span.is_dummy() {
2100 Hash::hash(&TAG_INVALID_SPAN, hasher);
2104 if let Some(parent) = span.parent {
2105 let def_span = ctx.def_span(parent).data_untracked();
2106 if def_span.contains(span) {
2107 // This span is enclosed in a definition: only hash the relative position.
2108 Hash::hash(&TAG_RELATIVE_SPAN, hasher);
2109 (span.lo - def_span.lo).to_u32().hash_stable(ctx, hasher);
2110 (span.hi - def_span.lo).to_u32().hash_stable(ctx, hasher);
2115 // If this is not an empty or invalid span, we want to hash the last
2116 // position that belongs to it, as opposed to hashing the first
2117 // position past it.
2118 let (file, line_lo, col_lo, line_hi, col_hi) = match ctx.span_data_to_lines_and_cols(&span)
2122 Hash::hash(&TAG_INVALID_SPAN, hasher);
2127 Hash::hash(&TAG_VALID_SPAN, hasher);
2128 // We truncate the stable ID hash and line and column numbers. The chances
2129 // of causing a collision this way should be minimal.
2130 Hash::hash(&(file.name_hash as u64), hasher);
2132 // Hash both the length and the end location (line/column) of a span. If we
2133 // hash only the length, for example, then two otherwise equal spans with
2134 // different end locations will have the same hash. This can cause a problem
2135 // during incremental compilation wherein a previous result for a query that
2136 // depends on the end location of a span will be incorrectly reused when the
2137 // end location of the span it depends on has changed (see issue #74890). A
2138 // similar analysis applies if some query depends specifically on the length
2139 // of the span, but we only hash the end location. So hash both.
2141 let col_lo_trunc = (col_lo.0 as u64) & 0xFF;
2142 let line_lo_trunc = ((line_lo as u64) & 0xFF_FF_FF) << 8;
2143 let col_hi_trunc = (col_hi.0 as u64) & 0xFF << 32;
2144 let line_hi_trunc = ((line_hi as u64) & 0xFF_FF_FF) << 40;
2145 let col_line = col_lo_trunc | line_lo_trunc | col_hi_trunc | line_hi_trunc;
2146 let len = (span.hi - span.lo).0;
2147 Hash::hash(&col_line, hasher);
2148 Hash::hash(&len, hasher);