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(bool_to_option)]
19 #![feature(crate_visibility_modifier)]
21 #![feature(if_let_guard)]
22 #![feature(negative_impls)]
24 #![feature(min_specialization)]
25 #![feature(rustc_attrs)]
26 #![allow(rustc::potential_query_instability)]
29 extern crate rustc_macros;
34 use rustc_data_structures::AtomicRef;
35 use rustc_macros::HashStable_Generic;
36 use rustc_serialize::{Decodable, Decoder, Encodable, Encoder};
38 mod caching_source_map_view;
40 pub use self::caching_source_map_view::CachingSourceMapView;
41 use source_map::SourceMap;
46 use hygiene::Transparency;
47 pub use hygiene::{DesugaringKind, ExpnKind, MacroKind};
48 pub use hygiene::{ExpnData, ExpnHash, ExpnId, LocalExpnId, SyntaxContext};
49 use rustc_data_structures::stable_hasher::HashingControls;
51 use def_id::{CrateNum, DefId, DefPathHash, LocalDefId, LOCAL_CRATE};
54 pub use span_encoding::{Span, DUMMY_SP};
57 pub use symbol::{sym, Symbol};
59 mod analyze_source_file;
62 use rustc_data_structures::stable_hasher::{HashStable, StableHasher};
63 use rustc_data_structures::sync::{Lock, Lrc};
66 use std::cmp::{self, Ordering};
69 use std::ops::{Add, Range, Sub};
70 use std::path::{Path, PathBuf};
71 use std::str::FromStr;
83 // Per-session global variables: this struct is stored in thread-local storage
84 // in such a way that it is accessible without any kind of handle to all
85 // threads within the compilation session, but is not accessible outside the
87 pub struct SessionGlobals {
88 symbol_interner: symbol::Interner,
89 span_interner: Lock<span_encoding::SpanInterner>,
90 hygiene_data: Lock<hygiene::HygieneData>,
91 source_map: Lock<Option<Lrc<SourceMap>>>,
95 pub fn new(edition: Edition) -> SessionGlobals {
97 symbol_interner: symbol::Interner::fresh(),
98 span_interner: Lock::new(span_encoding::SpanInterner::default()),
99 hygiene_data: Lock::new(hygiene::HygieneData::new(edition)),
100 source_map: Lock::new(None),
106 pub fn create_session_globals_then<R>(edition: Edition, f: impl FnOnce() -> R) -> R {
108 !SESSION_GLOBALS.is_set(),
109 "SESSION_GLOBALS should never be overwritten! \
110 Use another thread if you need another SessionGlobals"
112 let session_globals = SessionGlobals::new(edition);
113 SESSION_GLOBALS.set(&session_globals, f)
117 pub fn set_session_globals_then<R>(session_globals: &SessionGlobals, f: impl FnOnce() -> R) -> R {
119 !SESSION_GLOBALS.is_set(),
120 "SESSION_GLOBALS should never be overwritten! \
121 Use another thread if you need another SessionGlobals"
123 SESSION_GLOBALS.set(session_globals, f)
127 pub fn create_default_session_if_not_set_then<R, F>(f: F) -> R
129 F: FnOnce(&SessionGlobals) -> R,
131 create_session_if_not_set_then(edition::DEFAULT_EDITION, f)
135 pub fn create_session_if_not_set_then<R, F>(edition: Edition, f: F) -> R
137 F: FnOnce(&SessionGlobals) -> R,
139 if !SESSION_GLOBALS.is_set() {
140 let session_globals = SessionGlobals::new(edition);
141 SESSION_GLOBALS.set(&session_globals, || SESSION_GLOBALS.with(f))
143 SESSION_GLOBALS.with(f)
148 pub fn with_session_globals<R, F>(f: F) -> R
150 F: FnOnce(&SessionGlobals) -> R,
152 SESSION_GLOBALS.with(f)
156 pub fn create_default_session_globals_then<R>(f: impl FnOnce() -> R) -> R {
157 create_session_globals_then(edition::DEFAULT_EDITION, f)
160 // If this ever becomes non thread-local, `decode_syntax_context`
161 // and `decode_expn_id` will need to be updated to handle concurrent
163 scoped_tls::scoped_thread_local!(static SESSION_GLOBALS: SessionGlobals);
165 // FIXME: We should use this enum or something like it to get rid of the
166 // use of magic `/rust/1.x/...` paths across the board.
167 #[derive(Debug, Eq, PartialEq, Clone, Ord, PartialOrd)]
169 pub enum RealFileName {
171 /// For remapped paths (namely paths into libstd that have been mapped
172 /// to the appropriate spot on the local host's file system, and local file
173 /// system paths that have been remapped with `FilePathMapping`),
175 /// `local_path` is the (host-dependent) local path to the file. This is
176 /// None if the file was imported from another crate
177 local_path: Option<PathBuf>,
178 /// `virtual_name` is the stable path rustc will store internally within
180 virtual_name: PathBuf,
184 impl Hash for RealFileName {
185 fn hash<H: std::hash::Hasher>(&self, state: &mut H) {
186 // To prevent #70924 from happening again we should only hash the
187 // remapped (virtualized) path if that exists. This is because
188 // virtualized paths to sysroot crates (/rust/$hash or /rust/$version)
189 // remain stable even if the corresponding local_path changes
190 self.remapped_path_if_available().hash(state)
194 // This is functionally identical to #[derive(Encodable)], with the exception of
195 // an added assert statement
196 impl<S: Encoder> Encodable<S> for RealFileName {
197 fn encode(&self, encoder: &mut S) -> Result<(), S::Error> {
198 encoder.emit_enum(|encoder| match *self {
199 RealFileName::LocalPath(ref local_path) => {
200 encoder.emit_enum_variant("LocalPath", 0, 1, |encoder| {
201 encoder.emit_enum_variant_arg(true, |encoder| local_path.encode(encoder))?;
206 RealFileName::Remapped { ref local_path, ref virtual_name } => encoder
207 .emit_enum_variant("Remapped", 1, 2, |encoder| {
208 // For privacy and build reproducibility, we must not embed host-dependant path in artifacts
209 // if they have been remapped by --remap-path-prefix
210 assert!(local_path.is_none());
211 encoder.emit_enum_variant_arg(true, |encoder| local_path.encode(encoder))?;
212 encoder.emit_enum_variant_arg(false, |encoder| virtual_name.encode(encoder))?;
220 /// Returns the path suitable for reading from the file system on the local host,
221 /// if this information exists.
222 /// Avoid embedding this in build artifacts; see `remapped_path_if_available()` for that.
223 pub fn local_path(&self) -> Option<&Path> {
225 RealFileName::LocalPath(p) => Some(p),
226 RealFileName::Remapped { local_path: p, virtual_name: _ } => {
227 p.as_ref().map(PathBuf::as_path)
232 /// Returns the path suitable for reading from the file system on the local host,
233 /// if this information exists.
234 /// Avoid embedding this in build artifacts; see `remapped_path_if_available()` for that.
235 pub fn into_local_path(self) -> Option<PathBuf> {
237 RealFileName::LocalPath(p) => Some(p),
238 RealFileName::Remapped { local_path: p, virtual_name: _ } => p,
242 /// Returns the path suitable for embedding into build artifacts. This would still
243 /// be a local path if it has not been remapped. A remapped path will not correspond
244 /// to a valid file system path: see `local_path_if_available()` for something that
245 /// is more likely to return paths into the local host file system.
246 pub fn remapped_path_if_available(&self) -> &Path {
248 RealFileName::LocalPath(p)
249 | RealFileName::Remapped { local_path: _, virtual_name: p } => &p,
253 /// Returns the path suitable for reading from the file system on the local host,
254 /// if this information exists. Otherwise returns the remapped name.
255 /// Avoid embedding this in build artifacts; see `remapped_path_if_available()` for that.
256 pub fn local_path_if_available(&self) -> &Path {
258 RealFileName::LocalPath(path)
259 | RealFileName::Remapped { local_path: None, virtual_name: path }
260 | RealFileName::Remapped { local_path: Some(path), virtual_name: _ } => path,
264 pub fn to_string_lossy(&self, display_pref: FileNameDisplayPreference) -> Cow<'_, str> {
266 FileNameDisplayPreference::Local => self.local_path_if_available().to_string_lossy(),
267 FileNameDisplayPreference::Remapped => {
268 self.remapped_path_if_available().to_string_lossy()
274 /// Differentiates between real files and common virtual files.
275 #[derive(Debug, Eq, PartialEq, Clone, Ord, PartialOrd, Hash)]
276 #[derive(Decodable, Encodable)]
279 /// Call to `quote!`.
283 /// Hack in `src/librustc_ast/parse.rs`.
286 ProcMacroSourceCode(u64),
287 /// Strings provided as `--cfg [cfgspec]` stored in a `crate_cfg`.
289 /// Strings provided as crate attributes in the CLI.
291 /// Custom sources for explicit parser calls from plugins and drivers.
293 DocTest(PathBuf, isize),
294 /// Post-substitution inline assembly from LLVM.
298 impl From<PathBuf> for FileName {
299 fn from(p: PathBuf) -> Self {
300 assert!(!p.to_string_lossy().ends_with('>'));
301 FileName::Real(RealFileName::LocalPath(p))
305 #[derive(Clone, Copy, Eq, PartialEq, Hash, Debug)]
306 pub enum FileNameDisplayPreference {
311 pub struct FileNameDisplay<'a> {
313 display_pref: FileNameDisplayPreference,
316 impl fmt::Display for FileNameDisplay<'_> {
317 fn fmt(&self, fmt: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
321 write!(fmt, "{}", name.to_string_lossy(self.display_pref))
323 QuoteExpansion(_) => write!(fmt, "<quote expansion>"),
324 MacroExpansion(_) => write!(fmt, "<macro expansion>"),
325 Anon(_) => write!(fmt, "<anon>"),
326 ProcMacroSourceCode(_) => write!(fmt, "<proc-macro source code>"),
327 CfgSpec(_) => write!(fmt, "<cfgspec>"),
328 CliCrateAttr(_) => write!(fmt, "<crate attribute>"),
329 Custom(ref s) => write!(fmt, "<{}>", s),
330 DocTest(ref path, _) => write!(fmt, "{}", path.display()),
331 InlineAsm(_) => write!(fmt, "<inline asm>"),
336 impl FileNameDisplay<'_> {
337 pub fn to_string_lossy(&self) -> Cow<'_, str> {
339 FileName::Real(ref inner) => inner.to_string_lossy(self.display_pref),
340 _ => Cow::from(format!("{}", self)),
346 pub fn is_real(&self) -> bool {
352 | ProcMacroSourceCode(_)
358 | InlineAsm(_) => false,
362 pub fn prefer_remapped(&self) -> FileNameDisplay<'_> {
363 FileNameDisplay { inner: self, display_pref: FileNameDisplayPreference::Remapped }
366 // This may include transient local filesystem information.
367 // Must not be embedded in build outputs.
368 pub fn prefer_local(&self) -> FileNameDisplay<'_> {
369 FileNameDisplay { inner: self, display_pref: FileNameDisplayPreference::Local }
372 pub fn display(&self, display_pref: FileNameDisplayPreference) -> FileNameDisplay<'_> {
373 FileNameDisplay { inner: self, display_pref }
376 pub fn macro_expansion_source_code(src: &str) -> FileName {
377 let mut hasher = StableHasher::new();
378 src.hash(&mut hasher);
379 FileName::MacroExpansion(hasher.finish())
382 pub fn anon_source_code(src: &str) -> FileName {
383 let mut hasher = StableHasher::new();
384 src.hash(&mut hasher);
385 FileName::Anon(hasher.finish())
388 pub fn proc_macro_source_code(src: &str) -> FileName {
389 let mut hasher = StableHasher::new();
390 src.hash(&mut hasher);
391 FileName::ProcMacroSourceCode(hasher.finish())
394 pub fn cfg_spec_source_code(src: &str) -> FileName {
395 let mut hasher = StableHasher::new();
396 src.hash(&mut hasher);
397 FileName::QuoteExpansion(hasher.finish())
400 pub fn cli_crate_attr_source_code(src: &str) -> FileName {
401 let mut hasher = StableHasher::new();
402 src.hash(&mut hasher);
403 FileName::CliCrateAttr(hasher.finish())
406 pub fn doc_test_source_code(path: PathBuf, line: isize) -> FileName {
407 FileName::DocTest(path, line)
410 pub fn inline_asm_source_code(src: &str) -> FileName {
411 let mut hasher = StableHasher::new();
412 src.hash(&mut hasher);
413 FileName::InlineAsm(hasher.finish())
417 /// Represents a span.
419 /// Spans represent a region of code, used for error reporting. Positions in spans
420 /// are *absolute* positions from the beginning of the [`SourceMap`], not positions
421 /// relative to [`SourceFile`]s. Methods on the `SourceMap` can be used to relate spans back
422 /// to the original source.
424 /// You must be careful if the span crosses more than one file, since you will not be
425 /// able to use many of the functions on spans in source_map and you cannot assume
426 /// that the length of the span is equal to `span.hi - span.lo`; there may be space in the
427 /// [`BytePos`] range between files.
429 /// `SpanData` is public because `Span` uses a thread-local interner and can't be
430 /// sent to other threads, but some pieces of performance infra run in a separate thread.
431 /// Using `Span` is generally preferred.
432 #[derive(Clone, Copy, Hash, PartialEq, Eq)]
433 pub struct SpanData {
436 /// Information about where the macro came from, if this piece of
437 /// code was created by a macro expansion.
438 pub ctxt: SyntaxContext,
439 pub parent: Option<LocalDefId>,
442 // Order spans by position in the file.
443 impl Ord for SpanData {
444 fn cmp(&self, other: &Self) -> Ordering {
449 // `LocalDefId` does not implement `Ord`.
450 // The other fields are enough to determine in-file order.
457 // `LocalDefId` does not implement `Ord`.
458 // The other fields are enough to determine in-file order.
462 (s_lo, s_hi, s_ctxt).cmp(&(o_lo, o_hi, o_ctxt))
466 impl PartialOrd for SpanData {
467 fn partial_cmp(&self, other: &Self) -> Option<Ordering> {
468 Some(self.cmp(other))
474 pub fn span(&self) -> Span {
475 Span::new(self.lo, self.hi, self.ctxt, self.parent)
478 pub fn with_lo(&self, lo: BytePos) -> Span {
479 Span::new(lo, self.hi, self.ctxt, self.parent)
482 pub fn with_hi(&self, hi: BytePos) -> Span {
483 Span::new(self.lo, hi, self.ctxt, self.parent)
486 pub fn with_ctxt(&self, ctxt: SyntaxContext) -> Span {
487 Span::new(self.lo, self.hi, ctxt, self.parent)
490 pub fn with_parent(&self, parent: Option<LocalDefId>) -> Span {
491 Span::new(self.lo, self.hi, self.ctxt, parent)
493 /// Returns `true` if this is a dummy span with any hygienic context.
495 pub fn is_dummy(self) -> bool {
496 self.lo.0 == 0 && self.hi.0 == 0
498 /// Returns `true` if `self` fully encloses `other`.
499 pub fn contains(self, other: Self) -> bool {
500 self.lo <= other.lo && other.hi <= self.hi
504 // The interner is pointed to by a thread local value which is only set on the main thread
505 // with parallelization is disabled. So we don't allow `Span` to transfer between threads
506 // to avoid panics and other errors, even though it would be memory safe to do so.
507 #[cfg(not(parallel_compiler))]
508 impl !Send for Span {}
509 #[cfg(not(parallel_compiler))]
510 impl !Sync for Span {}
512 impl PartialOrd for Span {
513 fn partial_cmp(&self, rhs: &Self) -> Option<Ordering> {
514 PartialOrd::partial_cmp(&self.data(), &rhs.data())
518 fn cmp(&self, rhs: &Self) -> Ordering {
519 Ord::cmp(&self.data(), &rhs.data())
523 /// A collection of `Span`s.
525 /// Spans have two orthogonal attributes:
527 /// - They can be *primary spans*. In this case they are the locus of
528 /// the error, and would be rendered with `^^^`.
529 /// - They can have a *label*. In this case, the label is written next
530 /// to the mark in the snippet when we render.
531 #[derive(Clone, Debug, Hash, PartialEq, Eq, Encodable, Decodable)]
532 pub struct MultiSpan {
533 primary_spans: Vec<Span>,
534 span_labels: Vec<(Span, String)>,
539 pub fn lo(self) -> BytePos {
543 pub fn with_lo(self, lo: BytePos) -> Span {
544 self.data().with_lo(lo)
547 pub fn hi(self) -> BytePos {
551 pub fn with_hi(self, hi: BytePos) -> Span {
552 self.data().with_hi(hi)
555 pub fn ctxt(self) -> SyntaxContext {
556 self.data_untracked().ctxt
559 pub fn with_ctxt(self, ctxt: SyntaxContext) -> Span {
560 self.data_untracked().with_ctxt(ctxt)
563 pub fn parent(self) -> Option<LocalDefId> {
567 pub fn with_parent(self, ctxt: Option<LocalDefId>) -> Span {
568 self.data().with_parent(ctxt)
571 /// Returns `true` if this is a dummy span with any hygienic context.
573 pub fn is_dummy(self) -> bool {
574 self.data_untracked().is_dummy()
577 /// Returns `true` if this span comes from a macro or desugaring.
579 pub fn from_expansion(self) -> bool {
580 self.ctxt() != SyntaxContext::root()
583 /// Returns `true` if `span` originates in a derive-macro's expansion.
584 pub fn in_derive_expansion(self) -> bool {
585 matches!(self.ctxt().outer_expn_data().kind, ExpnKind::Macro(MacroKind::Derive, _))
588 /// Gate suggestions that would not be appropriate in a context the user didn't write.
589 pub fn can_be_used_for_suggestions(self) -> bool {
590 !self.from_expansion()
591 // FIXME: If this span comes from a `derive` macro but it points at code the user wrote,
592 // the callsite span and the span will be pointing at different places. It also means that
593 // we can safely provide suggestions on this span.
594 || (matches!(self.ctxt().outer_expn_data().kind, ExpnKind::Macro(MacroKind::Derive, _))
595 && self.parent_callsite().map(|p| (p.lo(), p.hi())) != Some((self.lo(), self.hi())))
599 pub fn with_root_ctxt(lo: BytePos, hi: BytePos) -> Span {
600 Span::new(lo, hi, SyntaxContext::root(), None)
603 /// Returns a new span representing an empty span at the beginning of this span.
605 pub fn shrink_to_lo(self) -> Span {
606 let span = self.data_untracked();
607 span.with_hi(span.lo)
609 /// Returns a new span representing an empty span at the end of this span.
611 pub fn shrink_to_hi(self) -> Span {
612 let span = self.data_untracked();
613 span.with_lo(span.hi)
617 /// Returns `true` if `hi == lo`.
618 pub fn is_empty(self) -> bool {
619 let span = self.data_untracked();
623 /// Returns `self` if `self` is not the dummy span, and `other` otherwise.
624 pub fn substitute_dummy(self, other: Span) -> Span {
625 if self.is_dummy() { other } else { self }
628 /// Returns `true` if `self` fully encloses `other`.
629 pub fn contains(self, other: Span) -> bool {
630 let span = self.data();
631 let other = other.data();
635 /// Returns `true` if `self` touches `other`.
636 pub fn overlaps(self, other: Span) -> bool {
637 let span = self.data();
638 let other = other.data();
639 span.lo < other.hi && other.lo < span.hi
642 /// Returns `true` if the spans are equal with regards to the source text.
644 /// Use this instead of `==` when either span could be generated code,
645 /// and you only care that they point to the same bytes of source text.
646 pub fn source_equal(self, other: Span) -> bool {
647 let span = self.data();
648 let other = other.data();
649 span.lo == other.lo && span.hi == other.hi
652 /// Returns `Some(span)`, where the start is trimmed by the end of `other`.
653 pub fn trim_start(self, other: Span) -> Option<Span> {
654 let span = self.data();
655 let other = other.data();
656 if span.hi > other.hi { Some(span.with_lo(cmp::max(span.lo, other.hi))) } else { None }
659 /// Returns the source span -- this is either the supplied span, or the span for
660 /// the macro callsite that expanded to it.
661 pub fn source_callsite(self) -> Span {
662 let expn_data = self.ctxt().outer_expn_data();
663 if !expn_data.is_root() { expn_data.call_site.source_callsite() } else { self }
666 /// The `Span` for the tokens in the previous macro expansion from which `self` was generated,
668 pub fn parent_callsite(self) -> Option<Span> {
669 let expn_data = self.ctxt().outer_expn_data();
670 if !expn_data.is_root() { Some(expn_data.call_site) } else { None }
673 /// Walk down the expansion ancestors to find a span that's contained within `outer`.
674 pub fn find_ancestor_inside(mut self, outer: Span) -> Option<Span> {
675 while !outer.contains(self) {
676 self = self.parent_callsite()?;
681 /// Edition of the crate from which this span came.
682 pub fn edition(self) -> edition::Edition {
683 self.ctxt().edition()
687 pub fn rust_2015(self) -> bool {
688 self.edition() == edition::Edition::Edition2015
692 pub fn rust_2018(self) -> bool {
693 self.edition() >= edition::Edition::Edition2018
697 pub fn rust_2021(self) -> bool {
698 self.edition() >= edition::Edition::Edition2021
701 /// Returns the source callee.
703 /// Returns `None` if the supplied span has no expansion trace,
704 /// else returns the `ExpnData` for the macro definition
705 /// corresponding to the source callsite.
706 pub fn source_callee(self) -> Option<ExpnData> {
707 fn source_callee(expn_data: ExpnData) -> ExpnData {
708 let next_expn_data = expn_data.call_site.ctxt().outer_expn_data();
709 if !next_expn_data.is_root() { source_callee(next_expn_data) } else { expn_data }
711 let expn_data = self.ctxt().outer_expn_data();
712 if !expn_data.is_root() { Some(source_callee(expn_data)) } else { None }
715 /// Checks if a span is "internal" to a macro in which `#[unstable]`
716 /// items can be used (that is, a macro marked with
717 /// `#[allow_internal_unstable]`).
718 pub fn allows_unstable(self, feature: Symbol) -> bool {
721 .allow_internal_unstable
722 .map_or(false, |features| features.iter().any(|&f| f == feature))
725 /// Checks if this span arises from a compiler desugaring of kind `kind`.
726 pub fn is_desugaring(self, kind: DesugaringKind) -> bool {
727 match self.ctxt().outer_expn_data().kind {
728 ExpnKind::Desugaring(k) => k == kind,
733 /// Returns the compiler desugaring that created this span, or `None`
734 /// if this span is not from a desugaring.
735 pub fn desugaring_kind(self) -> Option<DesugaringKind> {
736 match self.ctxt().outer_expn_data().kind {
737 ExpnKind::Desugaring(k) => Some(k),
742 /// Checks if a span is "internal" to a macro in which `unsafe`
743 /// can be used without triggering the `unsafe_code` lint.
744 // (that is, a macro marked with `#[allow_internal_unsafe]`).
745 pub fn allows_unsafe(self) -> bool {
746 self.ctxt().outer_expn_data().allow_internal_unsafe
749 pub fn macro_backtrace(mut self) -> impl Iterator<Item = ExpnData> {
750 let mut prev_span = DUMMY_SP;
751 std::iter::from_fn(move || {
753 let expn_data = self.ctxt().outer_expn_data();
754 if expn_data.is_root() {
758 let is_recursive = expn_data.call_site.source_equal(prev_span);
761 self = expn_data.call_site;
763 // Don't print recursive invocations.
765 return Some(expn_data);
771 /// Returns a `Span` that would enclose both `self` and `end`.
775 /// self lorem ipsum end
776 /// ^^^^^^^^^^^^^^^^^^^^
778 pub fn to(self, end: Span) -> Span {
779 let span_data = self.data();
780 let end_data = end.data();
781 // FIXME(jseyfried): `self.ctxt` should always equal `end.ctxt` here (cf. issue #23480).
782 // Return the macro span on its own to avoid weird diagnostic output. It is preferable to
783 // have an incomplete span than a completely nonsensical one.
784 if span_data.ctxt != end_data.ctxt {
785 if span_data.ctxt == SyntaxContext::root() {
787 } else if end_data.ctxt == SyntaxContext::root() {
790 // Both spans fall within a macro.
791 // FIXME(estebank): check if it is the *same* macro.
794 cmp::min(span_data.lo, end_data.lo),
795 cmp::max(span_data.hi, end_data.hi),
796 if span_data.ctxt == SyntaxContext::root() { end_data.ctxt } else { span_data.ctxt },
797 if span_data.parent == end_data.parent { span_data.parent } else { None },
801 /// Returns a `Span` between the end of `self` to the beginning of `end`.
805 /// self lorem ipsum end
808 pub fn between(self, end: Span) -> Span {
809 let span = self.data();
810 let end = end.data();
814 if end.ctxt == SyntaxContext::root() { end.ctxt } else { span.ctxt },
815 if span.parent == end.parent { span.parent } else { None },
819 /// Returns a `Span` from the beginning of `self` until the beginning of `end`.
823 /// self lorem ipsum end
824 /// ^^^^^^^^^^^^^^^^^
826 pub fn until(self, end: Span) -> Span {
827 // Most of this function's body is copied from `to`.
828 // We can't just do `self.to(end.shrink_to_lo())`,
829 // because to also does some magic where it uses min/max so
830 // it can handle overlapping spans. Some advanced mis-use of
831 // `until` with different ctxts makes this visible.
832 let span_data = self.data();
833 let end_data = end.data();
834 // FIXME(jseyfried): `self.ctxt` should always equal `end.ctxt` here (cf. issue #23480).
835 // Return the macro span on its own to avoid weird diagnostic output. It is preferable to
836 // have an incomplete span than a completely nonsensical one.
837 if span_data.ctxt != end_data.ctxt {
838 if span_data.ctxt == SyntaxContext::root() {
840 } else if end_data.ctxt == SyntaxContext::root() {
843 // Both spans fall within a macro.
844 // FIXME(estebank): check if it is the *same* macro.
849 if end_data.ctxt == SyntaxContext::root() { end_data.ctxt } else { span_data.ctxt },
850 if span_data.parent == end_data.parent { span_data.parent } else { None },
854 pub fn from_inner(self, inner: InnerSpan) -> Span {
855 let span = self.data();
857 span.lo + BytePos::from_usize(inner.start),
858 span.lo + BytePos::from_usize(inner.end),
864 /// Equivalent of `Span::def_site` from the proc macro API,
865 /// except that the location is taken from the `self` span.
866 pub fn with_def_site_ctxt(self, expn_id: ExpnId) -> Span {
867 self.with_ctxt_from_mark(expn_id, Transparency::Opaque)
870 /// Equivalent of `Span::call_site` from the proc macro API,
871 /// except that the location is taken from the `self` span.
872 pub fn with_call_site_ctxt(self, expn_id: ExpnId) -> Span {
873 self.with_ctxt_from_mark(expn_id, Transparency::Transparent)
876 /// Equivalent of `Span::mixed_site` from the proc macro API,
877 /// except that the location is taken from the `self` span.
878 pub fn with_mixed_site_ctxt(self, expn_id: ExpnId) -> Span {
879 self.with_ctxt_from_mark(expn_id, Transparency::SemiTransparent)
882 /// Produces a span with the same location as `self` and context produced by a macro with the
883 /// given ID and transparency, assuming that macro was defined directly and not produced by
884 /// some other macro (which is the case for built-in and procedural macros).
885 pub fn with_ctxt_from_mark(self, expn_id: ExpnId, transparency: Transparency) -> Span {
886 self.with_ctxt(SyntaxContext::root().apply_mark(expn_id, transparency))
890 pub fn apply_mark(self, expn_id: ExpnId, transparency: Transparency) -> Span {
891 let span = self.data();
892 span.with_ctxt(span.ctxt.apply_mark(expn_id, transparency))
896 pub fn remove_mark(&mut self) -> ExpnId {
897 let mut span = self.data();
898 let mark = span.ctxt.remove_mark();
899 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
904 pub fn adjust(&mut self, expn_id: ExpnId) -> Option<ExpnId> {
905 let mut span = self.data();
906 let mark = span.ctxt.adjust(expn_id);
907 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
912 pub fn normalize_to_macros_2_0_and_adjust(&mut self, expn_id: ExpnId) -> Option<ExpnId> {
913 let mut span = self.data();
914 let mark = span.ctxt.normalize_to_macros_2_0_and_adjust(expn_id);
915 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
920 pub fn glob_adjust(&mut self, expn_id: ExpnId, glob_span: Span) -> Option<Option<ExpnId>> {
921 let mut span = self.data();
922 let mark = span.ctxt.glob_adjust(expn_id, glob_span);
923 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
928 pub fn reverse_glob_adjust(
932 ) -> Option<Option<ExpnId>> {
933 let mut span = self.data();
934 let mark = span.ctxt.reverse_glob_adjust(expn_id, glob_span);
935 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
940 pub fn normalize_to_macros_2_0(self) -> Span {
941 let span = self.data();
942 span.with_ctxt(span.ctxt.normalize_to_macros_2_0())
946 pub fn normalize_to_macro_rules(self) -> Span {
947 let span = self.data();
948 span.with_ctxt(span.ctxt.normalize_to_macro_rules())
952 /// A span together with some additional data.
953 #[derive(Clone, Debug)]
954 pub struct SpanLabel {
955 /// The span we are going to include in the final snippet.
958 /// Is this a primary span? This is the "locus" of the message,
959 /// and is indicated with a `^^^^` underline, versus `----`.
960 pub is_primary: bool,
962 /// What label should we attach to this span (if any)?
963 pub label: Option<String>,
966 impl Default for Span {
967 fn default() -> Self {
972 impl<E: Encoder> Encodable<E> for Span {
973 default fn encode(&self, s: &mut E) -> Result<(), E::Error> {
974 let span = self.data();
975 s.emit_struct(false, |s| {
976 s.emit_struct_field("lo", true, |s| span.lo.encode(s))?;
977 s.emit_struct_field("hi", false, |s| span.hi.encode(s))
981 impl<D: Decoder> Decodable<D> for Span {
982 default fn decode(s: &mut D) -> Span {
983 let lo = Decodable::decode(s);
984 let hi = Decodable::decode(s);
986 Span::new(lo, hi, SyntaxContext::root(), None)
990 /// Calls the provided closure, using the provided `SourceMap` to format
991 /// any spans that are debug-printed during the closure's execution.
993 /// Normally, the global `TyCtxt` is used to retrieve the `SourceMap`
994 /// (see `rustc_interface::callbacks::span_debug1`). However, some parts
995 /// of the compiler (e.g. `rustc_parse`) may debug-print `Span`s before
996 /// a `TyCtxt` is available. In this case, we fall back to
997 /// the `SourceMap` provided to this function. If that is not available,
998 /// we fall back to printing the raw `Span` field values.
999 pub fn with_source_map<T, F: FnOnce() -> T>(source_map: Lrc<SourceMap>, f: F) -> T {
1000 with_session_globals(|session_globals| {
1001 *session_globals.source_map.borrow_mut() = Some(source_map);
1003 struct ClearSourceMap;
1004 impl Drop for ClearSourceMap {
1005 fn drop(&mut self) {
1006 with_session_globals(|session_globals| {
1007 session_globals.source_map.borrow_mut().take();
1012 let _guard = ClearSourceMap;
1016 impl fmt::Debug for Span {
1017 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
1018 with_session_globals(|session_globals| {
1019 if let Some(source_map) = &*session_globals.source_map.borrow() {
1020 write!(f, "{} ({:?})", source_map.span_to_diagnostic_string(*self), self.ctxt())
1022 f.debug_struct("Span")
1023 .field("lo", &self.lo())
1024 .field("hi", &self.hi())
1025 .field("ctxt", &self.ctxt())
1032 impl fmt::Debug for SpanData {
1033 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
1034 fmt::Debug::fmt(&Span::new(self.lo, self.hi, self.ctxt, self.parent), f)
1040 pub fn new() -> MultiSpan {
1041 MultiSpan { primary_spans: vec![], span_labels: vec![] }
1044 pub fn from_span(primary_span: Span) -> MultiSpan {
1045 MultiSpan { primary_spans: vec![primary_span], span_labels: vec![] }
1048 pub fn from_spans(mut vec: Vec<Span>) -> MultiSpan {
1050 MultiSpan { primary_spans: vec, span_labels: vec![] }
1053 pub fn push_span_label(&mut self, span: Span, label: String) {
1054 self.span_labels.push((span, label));
1057 /// Selects the first primary span (if any).
1058 pub fn primary_span(&self) -> Option<Span> {
1059 self.primary_spans.first().cloned()
1062 /// Returns all primary spans.
1063 pub fn primary_spans(&self) -> &[Span] {
1067 /// Returns `true` if any of the primary spans are displayable.
1068 pub fn has_primary_spans(&self) -> bool {
1069 self.primary_spans.iter().any(|sp| !sp.is_dummy())
1072 /// Returns `true` if this contains only a dummy primary span with any hygienic context.
1073 pub fn is_dummy(&self) -> bool {
1074 let mut is_dummy = true;
1075 for span in &self.primary_spans {
1076 if !span.is_dummy() {
1083 /// Replaces all occurrences of one Span with another. Used to move `Span`s in areas that don't
1084 /// display well (like std macros). Returns whether replacements occurred.
1085 pub fn replace(&mut self, before: Span, after: Span) -> bool {
1086 let mut replacements_occurred = false;
1087 for primary_span in &mut self.primary_spans {
1088 if *primary_span == before {
1089 *primary_span = after;
1090 replacements_occurred = true;
1093 for span_label in &mut self.span_labels {
1094 if span_label.0 == before {
1095 span_label.0 = after;
1096 replacements_occurred = true;
1099 replacements_occurred
1102 /// Returns the strings to highlight. We always ensure that there
1103 /// is an entry for each of the primary spans -- for each primary
1104 /// span `P`, if there is at least one label with span `P`, we return
1105 /// those labels (marked as primary). But otherwise we return
1106 /// `SpanLabel` instances with empty labels.
1107 pub fn span_labels(&self) -> Vec<SpanLabel> {
1108 let is_primary = |span| self.primary_spans.contains(&span);
1110 let mut span_labels = self
1113 .map(|&(span, ref label)| SpanLabel {
1115 is_primary: is_primary(span),
1116 label: Some(label.clone()),
1118 .collect::<Vec<_>>();
1120 for &span in &self.primary_spans {
1121 if !span_labels.iter().any(|sl| sl.span == span) {
1122 span_labels.push(SpanLabel { span, is_primary: true, label: None });
1129 /// Returns `true` if any of the span labels is displayable.
1130 pub fn has_span_labels(&self) -> bool {
1131 self.span_labels.iter().any(|(sp, _)| !sp.is_dummy())
1135 impl From<Span> for MultiSpan {
1136 fn from(span: Span) -> MultiSpan {
1137 MultiSpan::from_span(span)
1141 impl From<Vec<Span>> for MultiSpan {
1142 fn from(spans: Vec<Span>) -> MultiSpan {
1143 MultiSpan::from_spans(spans)
1147 /// Identifies an offset of a multi-byte character in a `SourceFile`.
1148 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1149 pub struct MultiByteChar {
1150 /// The absolute offset of the character in the `SourceMap`.
1152 /// The number of bytes, `>= 2`.
1156 /// Identifies an offset of a non-narrow character in a `SourceFile`.
1157 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1158 pub enum NonNarrowChar {
1159 /// Represents a zero-width character.
1161 /// Represents a wide (full-width) character.
1163 /// Represents a tab character, represented visually with a width of 4 characters.
1167 impl NonNarrowChar {
1168 fn new(pos: BytePos, width: usize) -> Self {
1170 0 => NonNarrowChar::ZeroWidth(pos),
1171 2 => NonNarrowChar::Wide(pos),
1172 4 => NonNarrowChar::Tab(pos),
1173 _ => panic!("width {} given for non-narrow character", width),
1177 /// Returns the absolute offset of the character in the `SourceMap`.
1178 pub fn pos(&self) -> BytePos {
1180 NonNarrowChar::ZeroWidth(p) | NonNarrowChar::Wide(p) | NonNarrowChar::Tab(p) => p,
1184 /// Returns the width of the character, 0 (zero-width) or 2 (wide).
1185 pub fn width(&self) -> usize {
1187 NonNarrowChar::ZeroWidth(_) => 0,
1188 NonNarrowChar::Wide(_) => 2,
1189 NonNarrowChar::Tab(_) => 4,
1194 impl Add<BytePos> for NonNarrowChar {
1197 fn add(self, rhs: BytePos) -> Self {
1199 NonNarrowChar::ZeroWidth(pos) => NonNarrowChar::ZeroWidth(pos + rhs),
1200 NonNarrowChar::Wide(pos) => NonNarrowChar::Wide(pos + rhs),
1201 NonNarrowChar::Tab(pos) => NonNarrowChar::Tab(pos + rhs),
1206 impl Sub<BytePos> for NonNarrowChar {
1209 fn sub(self, rhs: BytePos) -> Self {
1211 NonNarrowChar::ZeroWidth(pos) => NonNarrowChar::ZeroWidth(pos - rhs),
1212 NonNarrowChar::Wide(pos) => NonNarrowChar::Wide(pos - rhs),
1213 NonNarrowChar::Tab(pos) => NonNarrowChar::Tab(pos - rhs),
1218 /// Identifies an offset of a character that was normalized away from `SourceFile`.
1219 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1220 pub struct NormalizedPos {
1221 /// The absolute offset of the character in the `SourceMap`.
1223 /// The difference between original and normalized string at position.
1227 #[derive(PartialEq, Eq, Clone, Debug)]
1228 pub enum ExternalSource {
1229 /// No external source has to be loaded, since the `SourceFile` represents a local crate.
1232 kind: ExternalSourceKind,
1233 /// This SourceFile's byte-offset within the source_map of its original crate.
1234 original_start_pos: BytePos,
1235 /// The end of this SourceFile within the source_map of its original crate.
1236 original_end_pos: BytePos,
1240 /// The state of the lazy external source loading mechanism of a `SourceFile`.
1241 #[derive(PartialEq, Eq, Clone, Debug)]
1242 pub enum ExternalSourceKind {
1243 /// The external source has been loaded already.
1244 Present(Lrc<String>),
1245 /// No attempt has been made to load the external source.
1247 /// A failed attempt has been made to load the external source.
1252 impl ExternalSource {
1253 pub fn get_source(&self) -> Option<&Lrc<String>> {
1255 ExternalSource::Foreign { kind: ExternalSourceKind::Present(ref src), .. } => Some(src),
1262 pub struct OffsetOverflowError;
1264 #[derive(Copy, Clone, Debug, PartialEq, Eq, PartialOrd, Ord, Hash, Encodable, Decodable)]
1265 pub enum SourceFileHashAlgorithm {
1271 impl FromStr for SourceFileHashAlgorithm {
1274 fn from_str(s: &str) -> Result<SourceFileHashAlgorithm, ()> {
1276 "md5" => Ok(SourceFileHashAlgorithm::Md5),
1277 "sha1" => Ok(SourceFileHashAlgorithm::Sha1),
1278 "sha256" => Ok(SourceFileHashAlgorithm::Sha256),
1284 rustc_data_structures::impl_stable_hash_via_hash!(SourceFileHashAlgorithm);
1286 /// The hash of the on-disk source file used for debug info.
1287 #[derive(Copy, Clone, PartialEq, Eq, Debug)]
1288 #[derive(HashStable_Generic, Encodable, Decodable)]
1289 pub struct SourceFileHash {
1290 pub kind: SourceFileHashAlgorithm,
1294 impl SourceFileHash {
1295 pub fn new(kind: SourceFileHashAlgorithm, src: &str) -> SourceFileHash {
1296 let mut hash = SourceFileHash { kind, value: Default::default() };
1297 let len = hash.hash_len();
1298 let value = &mut hash.value[..len];
1299 let data = src.as_bytes();
1301 SourceFileHashAlgorithm::Md5 => {
1302 value.copy_from_slice(&Md5::digest(data));
1304 SourceFileHashAlgorithm::Sha1 => {
1305 value.copy_from_slice(&Sha1::digest(data));
1307 SourceFileHashAlgorithm::Sha256 => {
1308 value.copy_from_slice(&Sha256::digest(data));
1314 /// Check if the stored hash matches the hash of the string.
1315 pub fn matches(&self, src: &str) -> bool {
1316 Self::new(self.kind, src) == *self
1319 /// The bytes of the hash.
1320 pub fn hash_bytes(&self) -> &[u8] {
1321 let len = self.hash_len();
1325 fn hash_len(&self) -> usize {
1327 SourceFileHashAlgorithm::Md5 => 16,
1328 SourceFileHashAlgorithm::Sha1 => 20,
1329 SourceFileHashAlgorithm::Sha256 => 32,
1334 /// A single source in the [`SourceMap`].
1336 pub struct SourceFile {
1337 /// The name of the file that the source came from. Source that doesn't
1338 /// originate from files has names between angle brackets by convention
1339 /// (e.g., `<anon>`).
1341 /// The complete source code.
1342 pub src: Option<Lrc<String>>,
1343 /// The source code's hash.
1344 pub src_hash: SourceFileHash,
1345 /// The external source code (used for external crates, which will have a `None`
1346 /// value as `self.src`.
1347 pub external_src: Lock<ExternalSource>,
1348 /// The start position of this source in the `SourceMap`.
1349 pub start_pos: BytePos,
1350 /// The end position of this source in the `SourceMap`.
1351 pub end_pos: BytePos,
1352 /// Locations of lines beginnings in the source code.
1353 pub lines: Vec<BytePos>,
1354 /// Locations of multi-byte characters in the source code.
1355 pub multibyte_chars: Vec<MultiByteChar>,
1356 /// Width of characters that are not narrow in the source code.
1357 pub non_narrow_chars: Vec<NonNarrowChar>,
1358 /// Locations of characters removed during normalization.
1359 pub normalized_pos: Vec<NormalizedPos>,
1360 /// A hash of the filename, used for speeding up hashing in incremental compilation.
1361 pub name_hash: u128,
1362 /// Indicates which crate this `SourceFile` was imported from.
1366 impl<S: Encoder> Encodable<S> for SourceFile {
1367 fn encode(&self, s: &mut S) -> Result<(), S::Error> {
1368 s.emit_struct(false, |s| {
1369 s.emit_struct_field("name", true, |s| self.name.encode(s))?;
1370 s.emit_struct_field("src_hash", false, |s| self.src_hash.encode(s))?;
1371 s.emit_struct_field("start_pos", false, |s| self.start_pos.encode(s))?;
1372 s.emit_struct_field("end_pos", false, |s| self.end_pos.encode(s))?;
1373 s.emit_struct_field("lines", false, |s| {
1374 let lines = &self.lines[..];
1375 // Store the length.
1376 s.emit_u32(lines.len() as u32)?;
1378 if !lines.is_empty() {
1379 // In order to preserve some space, we exploit the fact that
1380 // the lines list is sorted and individual lines are
1381 // probably not that long. Because of that we can store lines
1382 // as a difference list, using as little space as possible
1383 // for the differences.
1384 let max_line_length = if lines.len() == 1 {
1389 .map(|&[fst, snd]| snd - fst)
1390 .map(|bp| bp.to_usize())
1395 let bytes_per_diff: u8 = match max_line_length {
1397 0x100..=0xFFFF => 2,
1401 // Encode the number of bytes used per diff.
1402 bytes_per_diff.encode(s)?;
1404 // Encode the first element.
1405 lines[0].encode(s)?;
1407 let diff_iter = lines.array_windows().map(|&[fst, snd]| snd - fst);
1409 match bytes_per_diff {
1411 for diff in diff_iter {
1412 (diff.0 as u8).encode(s)?
1416 for diff in diff_iter {
1417 (diff.0 as u16).encode(s)?
1421 for diff in diff_iter {
1425 _ => unreachable!(),
1431 s.emit_struct_field("multibyte_chars", false, |s| self.multibyte_chars.encode(s))?;
1432 s.emit_struct_field("non_narrow_chars", false, |s| self.non_narrow_chars.encode(s))?;
1433 s.emit_struct_field("name_hash", false, |s| self.name_hash.encode(s))?;
1434 s.emit_struct_field("normalized_pos", false, |s| self.normalized_pos.encode(s))?;
1435 s.emit_struct_field("cnum", false, |s| self.cnum.encode(s))
1440 impl<D: Decoder> Decodable<D> for SourceFile {
1441 fn decode(d: &mut D) -> SourceFile {
1442 let name: FileName = Decodable::decode(d);
1443 let src_hash: SourceFileHash = Decodable::decode(d);
1444 let start_pos: BytePos = Decodable::decode(d);
1445 let end_pos: BytePos = Decodable::decode(d);
1446 let lines: Vec<BytePos> = {
1447 let num_lines: u32 = Decodable::decode(d);
1448 let mut lines = Vec::with_capacity(num_lines as usize);
1451 // Read the number of bytes used per diff.
1452 let bytes_per_diff: u8 = Decodable::decode(d);
1454 // Read the first element.
1455 let mut line_start: BytePos = Decodable::decode(d);
1456 lines.push(line_start);
1458 for _ in 1..num_lines {
1459 let diff = match bytes_per_diff {
1460 1 => d.read_u8() as u32,
1461 2 => d.read_u16() as u32,
1463 _ => unreachable!(),
1466 line_start = line_start + BytePos(diff);
1468 lines.push(line_start);
1474 let multibyte_chars: Vec<MultiByteChar> = Decodable::decode(d);
1475 let non_narrow_chars: Vec<NonNarrowChar> = Decodable::decode(d);
1476 let name_hash: u128 = Decodable::decode(d);
1477 let normalized_pos: Vec<NormalizedPos> = Decodable::decode(d);
1478 let cnum: CrateNum = Decodable::decode(d);
1485 // Unused - the metadata decoder will construct
1486 // a new SourceFile, filling in `external_src` properly
1487 external_src: Lock::new(ExternalSource::Unneeded),
1498 impl fmt::Debug for SourceFile {
1499 fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result {
1500 write!(fmt, "SourceFile({:?})", self.name)
1509 hash_kind: SourceFileHashAlgorithm,
1511 // Compute the file hash before any normalization.
1512 let src_hash = SourceFileHash::new(hash_kind, &src);
1513 let normalized_pos = normalize_src(&mut src, start_pos);
1516 let mut hasher: StableHasher = StableHasher::new();
1517 name.hash(&mut hasher);
1518 hasher.finish::<u128>()
1520 let end_pos = start_pos.to_usize() + src.len();
1521 assert!(end_pos <= u32::MAX as usize);
1523 let (lines, multibyte_chars, non_narrow_chars) =
1524 analyze_source_file::analyze_source_file(&src, start_pos);
1528 src: Some(Lrc::new(src)),
1530 external_src: Lock::new(ExternalSource::Unneeded),
1532 end_pos: Pos::from_usize(end_pos),
1542 /// Returns the `BytePos` of the beginning of the current line.
1543 pub fn line_begin_pos(&self, pos: BytePos) -> BytePos {
1544 let line_index = self.lookup_line(pos).unwrap();
1545 self.lines[line_index]
1548 /// Add externally loaded source.
1549 /// If the hash of the input doesn't match or no input is supplied via None,
1550 /// it is interpreted as an error and the corresponding enum variant is set.
1551 /// The return value signifies whether some kind of source is present.
1552 pub fn add_external_src<F>(&self, get_src: F) -> bool
1554 F: FnOnce() -> Option<String>,
1557 *self.external_src.borrow(),
1558 ExternalSource::Foreign { kind: ExternalSourceKind::AbsentOk, .. }
1560 let src = get_src();
1561 let mut external_src = self.external_src.borrow_mut();
1562 // Check that no-one else have provided the source while we were getting it
1563 if let ExternalSource::Foreign {
1564 kind: src_kind @ ExternalSourceKind::AbsentOk, ..
1565 } = &mut *external_src
1567 if let Some(mut src) = src {
1568 // The src_hash needs to be computed on the pre-normalized src.
1569 if self.src_hash.matches(&src) {
1570 normalize_src(&mut src, BytePos::from_usize(0));
1571 *src_kind = ExternalSourceKind::Present(Lrc::new(src));
1575 *src_kind = ExternalSourceKind::AbsentErr;
1580 self.src.is_some() || external_src.get_source().is_some()
1583 self.src.is_some() || self.external_src.borrow().get_source().is_some()
1587 /// Gets a line from the list of pre-computed line-beginnings.
1588 /// The line number here is 0-based.
1589 pub fn get_line(&self, line_number: usize) -> Option<Cow<'_, str>> {
1590 fn get_until_newline(src: &str, begin: usize) -> &str {
1591 // We can't use `lines.get(line_number+1)` because we might
1592 // be parsing when we call this function and thus the current
1593 // line is the last one we have line info for.
1594 let slice = &src[begin..];
1595 match slice.find('\n') {
1596 Some(e) => &slice[..e],
1602 let line = self.lines.get(line_number)?;
1603 let begin: BytePos = *line - self.start_pos;
1607 if let Some(ref src) = self.src {
1608 Some(Cow::from(get_until_newline(src, begin)))
1609 } else if let Some(src) = self.external_src.borrow().get_source() {
1610 Some(Cow::Owned(String::from(get_until_newline(src, begin))))
1616 pub fn is_real_file(&self) -> bool {
1620 pub fn is_imported(&self) -> bool {
1624 pub fn count_lines(&self) -> usize {
1628 /// Finds the line containing the given position. The return value is the
1629 /// index into the `lines` array of this `SourceFile`, not the 1-based line
1630 /// number. If the source_file is empty or the position is located before the
1631 /// first line, `None` is returned.
1632 pub fn lookup_line(&self, pos: BytePos) -> Option<usize> {
1633 match self.lines.binary_search(&pos) {
1634 Ok(idx) => Some(idx),
1636 Err(idx) => Some(idx - 1),
1640 pub fn line_bounds(&self, line_index: usize) -> Range<BytePos> {
1641 if self.is_empty() {
1642 return self.start_pos..self.end_pos;
1645 assert!(line_index < self.lines.len());
1646 if line_index == (self.lines.len() - 1) {
1647 self.lines[line_index]..self.end_pos
1649 self.lines[line_index]..self.lines[line_index + 1]
1653 /// Returns whether or not the file contains the given `SourceMap` byte
1654 /// position. The position one past the end of the file is considered to be
1655 /// contained by the file. This implies that files for which `is_empty`
1656 /// returns true still contain one byte position according to this function.
1658 pub fn contains(&self, byte_pos: BytePos) -> bool {
1659 byte_pos >= self.start_pos && byte_pos <= self.end_pos
1663 pub fn is_empty(&self) -> bool {
1664 self.start_pos == self.end_pos
1667 /// Calculates the original byte position relative to the start of the file
1668 /// based on the given byte position.
1669 pub fn original_relative_byte_pos(&self, pos: BytePos) -> BytePos {
1670 // Diff before any records is 0. Otherwise use the previously recorded
1671 // diff as that applies to the following characters until a new diff
1673 let diff = match self.normalized_pos.binary_search_by(|np| np.pos.cmp(&pos)) {
1674 Ok(i) => self.normalized_pos[i].diff,
1675 Err(i) if i == 0 => 0,
1676 Err(i) => self.normalized_pos[i - 1].diff,
1679 BytePos::from_u32(pos.0 - self.start_pos.0 + diff)
1682 /// Converts an absolute `BytePos` to a `CharPos` relative to the `SourceFile`.
1683 pub fn bytepos_to_file_charpos(&self, bpos: BytePos) -> CharPos {
1684 // The number of extra bytes due to multibyte chars in the `SourceFile`.
1685 let mut total_extra_bytes = 0;
1687 for mbc in self.multibyte_chars.iter() {
1688 debug!("{}-byte char at {:?}", mbc.bytes, mbc.pos);
1690 // Every character is at least one byte, so we only
1691 // count the actual extra bytes.
1692 total_extra_bytes += mbc.bytes as u32 - 1;
1693 // We should never see a byte position in the middle of a
1695 assert!(bpos.to_u32() >= mbc.pos.to_u32() + mbc.bytes as u32);
1701 assert!(self.start_pos.to_u32() + total_extra_bytes <= bpos.to_u32());
1702 CharPos(bpos.to_usize() - self.start_pos.to_usize() - total_extra_bytes as usize)
1705 /// Looks up the file's (1-based) line number and (0-based `CharPos`) column offset, for a
1706 /// given `BytePos`.
1707 pub fn lookup_file_pos(&self, pos: BytePos) -> (usize, CharPos) {
1708 let chpos = self.bytepos_to_file_charpos(pos);
1709 match self.lookup_line(pos) {
1711 let line = a + 1; // Line numbers start at 1
1712 let linebpos = self.lines[a];
1713 let linechpos = self.bytepos_to_file_charpos(linebpos);
1714 let col = chpos - linechpos;
1715 debug!("byte pos {:?} is on the line at byte pos {:?}", pos, linebpos);
1716 debug!("char pos {:?} is on the line at char pos {:?}", chpos, linechpos);
1717 debug!("byte is on line: {}", line);
1718 assert!(chpos >= linechpos);
1725 /// Looks up the file's (1-based) line number, (0-based `CharPos`) column offset, and (0-based)
1726 /// column offset when displayed, for a given `BytePos`.
1727 pub fn lookup_file_pos_with_col_display(&self, pos: BytePos) -> (usize, CharPos, usize) {
1728 let (line, col_or_chpos) = self.lookup_file_pos(pos);
1730 let col = col_or_chpos;
1731 let linebpos = self.lines[line - 1];
1733 let start_width_idx = self
1735 .binary_search_by_key(&linebpos, |x| x.pos())
1736 .unwrap_or_else(|x| x);
1737 let end_width_idx = self
1739 .binary_search_by_key(&pos, |x| x.pos())
1740 .unwrap_or_else(|x| x);
1741 let special_chars = end_width_idx - start_width_idx;
1742 let non_narrow: usize = self.non_narrow_chars[start_width_idx..end_width_idx]
1746 col.0 - special_chars + non_narrow
1748 (line, col, col_display)
1750 let chpos = col_or_chpos;
1752 let end_width_idx = self
1754 .binary_search_by_key(&pos, |x| x.pos())
1755 .unwrap_or_else(|x| x);
1756 let non_narrow: usize =
1757 self.non_narrow_chars[0..end_width_idx].iter().map(|x| x.width()).sum();
1758 chpos.0 - end_width_idx + non_narrow
1760 (0, chpos, col_display)
1765 /// Normalizes the source code and records the normalizations.
1766 fn normalize_src(src: &mut String, start_pos: BytePos) -> Vec<NormalizedPos> {
1767 let mut normalized_pos = vec![];
1768 remove_bom(src, &mut normalized_pos);
1769 normalize_newlines(src, &mut normalized_pos);
1771 // Offset all the positions by start_pos to match the final file positions.
1772 for np in &mut normalized_pos {
1773 np.pos.0 += start_pos.0;
1779 /// Removes UTF-8 BOM, if any.
1780 fn remove_bom(src: &mut String, normalized_pos: &mut Vec<NormalizedPos>) {
1781 if src.starts_with('\u{feff}') {
1783 normalized_pos.push(NormalizedPos { pos: BytePos(0), diff: 3 });
1787 /// Replaces `\r\n` with `\n` in-place in `src`.
1789 /// Returns error if there's a lone `\r` in the string.
1790 fn normalize_newlines(src: &mut String, normalized_pos: &mut Vec<NormalizedPos>) {
1791 if !src.as_bytes().contains(&b'\r') {
1795 // We replace `\r\n` with `\n` in-place, which doesn't break utf-8 encoding.
1796 // While we *can* call `as_mut_vec` and do surgery on the live string
1797 // directly, let's rather steal the contents of `src`. This makes the code
1798 // safe even if a panic occurs.
1800 let mut buf = std::mem::replace(src, String::new()).into_bytes();
1801 let mut gap_len = 0;
1802 let mut tail = buf.as_mut_slice();
1804 let original_gap = normalized_pos.last().map_or(0, |l| l.diff);
1806 let idx = match find_crlf(&tail[gap_len..]) {
1808 Some(idx) => idx + gap_len,
1810 tail.copy_within(gap_len..idx, 0);
1811 tail = &mut tail[idx - gap_len..];
1812 if tail.len() == gap_len {
1815 cursor += idx - gap_len;
1817 normalized_pos.push(NormalizedPos {
1818 pos: BytePos::from_usize(cursor + 1),
1819 diff: original_gap + gap_len as u32,
1823 // Account for removed `\r`.
1824 // After `set_len`, `buf` is guaranteed to contain utf-8 again.
1825 let new_len = buf.len() - gap_len;
1827 buf.set_len(new_len);
1828 *src = String::from_utf8_unchecked(buf);
1831 fn find_crlf(src: &[u8]) -> Option<usize> {
1832 let mut search_idx = 0;
1833 while let Some(idx) = find_cr(&src[search_idx..]) {
1834 if src[search_idx..].get(idx + 1) != Some(&b'\n') {
1835 search_idx += idx + 1;
1838 return Some(search_idx + idx);
1843 fn find_cr(src: &[u8]) -> Option<usize> {
1844 src.iter().position(|&b| b == b'\r')
1848 // _____________________________________________________________________________
1849 // Pos, BytePos, CharPos
1853 fn from_usize(n: usize) -> Self;
1854 fn to_usize(&self) -> usize;
1855 fn from_u32(n: u32) -> Self;
1856 fn to_u32(&self) -> u32;
1859 macro_rules! impl_pos {
1863 $vis:vis struct $ident:ident($inner_vis:vis $inner_ty:ty);
1868 $vis struct $ident($inner_vis $inner_ty);
1870 impl Pos for $ident {
1872 fn from_usize(n: usize) -> $ident {
1873 $ident(n as $inner_ty)
1877 fn to_usize(&self) -> usize {
1882 fn from_u32(n: u32) -> $ident {
1883 $ident(n as $inner_ty)
1887 fn to_u32(&self) -> u32 {
1892 impl Add for $ident {
1893 type Output = $ident;
1896 fn add(self, rhs: $ident) -> $ident {
1897 $ident(self.0 + rhs.0)
1901 impl Sub for $ident {
1902 type Output = $ident;
1905 fn sub(self, rhs: $ident) -> $ident {
1906 $ident(self.0 - rhs.0)
1916 /// Keep this small (currently 32-bits), as AST contains a lot of them.
1917 #[derive(Clone, Copy, PartialEq, Eq, Hash, PartialOrd, Ord, Debug)]
1918 pub struct BytePos(pub u32);
1920 /// A character offset.
1922 /// Because of multibyte UTF-8 characters, a byte offset
1923 /// is not equivalent to a character offset. The [`SourceMap`] will convert [`BytePos`]
1924 /// values to `CharPos` values as necessary.
1925 #[derive(Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Debug)]
1926 pub struct CharPos(pub usize);
1929 impl<S: rustc_serialize::Encoder> Encodable<S> for BytePos {
1930 fn encode(&self, s: &mut S) -> Result<(), S::Error> {
1935 impl<D: rustc_serialize::Decoder> Decodable<D> for BytePos {
1936 fn decode(d: &mut D) -> BytePos {
1937 BytePos(d.read_u32())
1941 // _____________________________________________________________________________
1942 // Loc, SourceFileAndLine, SourceFileAndBytePos
1945 /// A source code location used for error reporting.
1946 #[derive(Debug, Clone)]
1948 /// Information about the original source.
1949 pub file: Lrc<SourceFile>,
1950 /// The (1-based) line number.
1952 /// The (0-based) column offset.
1954 /// The (0-based) column offset when displayed.
1955 pub col_display: usize,
1958 // Used to be structural records.
1960 pub struct SourceFileAndLine {
1961 pub sf: Lrc<SourceFile>,
1962 /// Index of line, starting from 0.
1966 pub struct SourceFileAndBytePos {
1967 pub sf: Lrc<SourceFile>,
1971 #[derive(Copy, Clone, Debug, PartialEq, Eq)]
1972 pub struct LineInfo {
1973 /// Index of line, starting from 0.
1974 pub line_index: usize,
1976 /// Column in line where span begins, starting from 0.
1977 pub start_col: CharPos,
1979 /// Column in line where span ends, starting from 0, exclusive.
1980 pub end_col: CharPos,
1983 pub struct FileLines {
1984 pub file: Lrc<SourceFile>,
1985 pub lines: Vec<LineInfo>,
1988 pub static SPAN_TRACK: AtomicRef<fn(LocalDefId)> = AtomicRef::new(&((|_| {}) as fn(_)));
1990 // _____________________________________________________________________________
1991 // SpanLinesError, SpanSnippetError, DistinctSources, MalformedSourceMapPositions
1994 pub type FileLinesResult = Result<FileLines, SpanLinesError>;
1996 #[derive(Clone, PartialEq, Eq, Debug)]
1997 pub enum SpanLinesError {
1998 DistinctSources(DistinctSources),
2001 #[derive(Clone, PartialEq, Eq, Debug)]
2002 pub enum SpanSnippetError {
2003 IllFormedSpan(Span),
2004 DistinctSources(DistinctSources),
2005 MalformedForSourcemap(MalformedSourceMapPositions),
2006 SourceNotAvailable { filename: FileName },
2009 #[derive(Clone, PartialEq, Eq, Debug)]
2010 pub struct DistinctSources {
2011 pub begin: (FileName, BytePos),
2012 pub end: (FileName, BytePos),
2015 #[derive(Clone, PartialEq, Eq, Debug)]
2016 pub struct MalformedSourceMapPositions {
2018 pub source_len: usize,
2019 pub begin_pos: BytePos,
2020 pub end_pos: BytePos,
2023 /// Range inside of a `Span` used for diagnostics when we only have access to relative positions.
2024 #[derive(Copy, Clone, PartialEq, Eq, Debug)]
2025 pub struct InnerSpan {
2031 pub fn new(start: usize, end: usize) -> InnerSpan {
2032 InnerSpan { start, end }
2036 /// Requirements for a `StableHashingContext` to be used in this crate.
2038 /// This is a hack to allow using the [`HashStable_Generic`] derive macro
2039 /// instead of implementing everything in rustc_middle.
2040 pub trait HashStableContext {
2041 fn def_path_hash(&self, def_id: DefId) -> DefPathHash;
2042 fn hash_spans(&self) -> bool;
2043 /// Accesses `sess.opts.debugging_opts.incremental_ignore_spans` since
2044 /// we don't have easy access to a `Session`
2045 fn debug_opts_incremental_ignore_spans(&self) -> bool;
2046 fn def_span(&self, def_id: LocalDefId) -> Span;
2047 fn span_data_to_lines_and_cols(
2050 ) -> Option<(Lrc<SourceFile>, usize, BytePos, usize, BytePos)>;
2051 fn hashing_controls(&self) -> HashingControls;
2054 impl<CTX> HashStable<CTX> for Span
2056 CTX: HashStableContext,
2058 /// Hashes a span in a stable way. We can't directly hash the span's `BytePos`
2059 /// fields (that would be similar to hashing pointers, since those are just
2060 /// offsets into the `SourceMap`). Instead, we hash the (file name, line, column)
2061 /// triple, which stays the same even if the containing `SourceFile` has moved
2062 /// within the `SourceMap`.
2064 /// Also note that we are hashing byte offsets for the column, not unicode
2065 /// codepoint offsets. For the purpose of the hash that's sufficient.
2066 /// Also, hashing filenames is expensive so we avoid doing it twice when the
2067 /// span starts and ends in the same file, which is almost always the case.
2068 fn hash_stable(&self, ctx: &mut CTX, hasher: &mut StableHasher) {
2069 const TAG_VALID_SPAN: u8 = 0;
2070 const TAG_INVALID_SPAN: u8 = 1;
2071 const TAG_RELATIVE_SPAN: u8 = 2;
2073 if !ctx.hash_spans() {
2077 let span = self.data_untracked();
2078 span.ctxt.hash_stable(ctx, hasher);
2079 span.parent.hash_stable(ctx, hasher);
2081 if span.is_dummy() {
2082 Hash::hash(&TAG_INVALID_SPAN, hasher);
2086 if let Some(parent) = span.parent {
2087 let def_span = ctx.def_span(parent).data_untracked();
2088 if def_span.contains(span) {
2089 // This span is enclosed in a definition: only hash the relative position.
2090 Hash::hash(&TAG_RELATIVE_SPAN, hasher);
2091 (span.lo - def_span.lo).to_u32().hash_stable(ctx, hasher);
2092 (span.hi - def_span.lo).to_u32().hash_stable(ctx, hasher);
2097 // If this is not an empty or invalid span, we want to hash the last
2098 // position that belongs to it, as opposed to hashing the first
2099 // position past it.
2100 let Some((file, line_lo, col_lo, line_hi, col_hi)) = ctx.span_data_to_lines_and_cols(&span) else {
2101 Hash::hash(&TAG_INVALID_SPAN, hasher);
2105 Hash::hash(&TAG_VALID_SPAN, hasher);
2106 // We truncate the stable ID hash and line and column numbers. The chances
2107 // of causing a collision this way should be minimal.
2108 Hash::hash(&(file.name_hash as u64), hasher);
2110 // Hash both the length and the end location (line/column) of a span. If we
2111 // hash only the length, for example, then two otherwise equal spans with
2112 // different end locations will have the same hash. This can cause a problem
2113 // during incremental compilation wherein a previous result for a query that
2114 // depends on the end location of a span will be incorrectly reused when the
2115 // end location of the span it depends on has changed (see issue #74890). A
2116 // similar analysis applies if some query depends specifically on the length
2117 // of the span, but we only hash the end location. So hash both.
2119 let col_lo_trunc = (col_lo.0 as u64) & 0xFF;
2120 let line_lo_trunc = ((line_lo as u64) & 0xFF_FF_FF) << 8;
2121 let col_hi_trunc = (col_hi.0 as u64) & 0xFF << 32;
2122 let line_hi_trunc = ((line_hi as u64) & 0xFF_FF_FF) << 40;
2123 let col_line = col_lo_trunc | line_lo_trunc | col_hi_trunc | line_hi_trunc;
2124 let len = (span.hi - span.lo).0;
2125 Hash::hash(&col_line, hasher);
2126 Hash::hash(&len, hasher);