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(if_let_guard)]
19 #![feature(negative_impls)]
20 #![feature(min_specialization)]
21 #![feature(rustc_attrs)]
22 #![deny(rustc::untranslatable_diagnostic)]
23 #![deny(rustc::diagnostic_outside_of_impl)]
26 extern crate rustc_macros;
31 use rustc_data_structures::AtomicRef;
32 use rustc_macros::HashStable_Generic;
33 use rustc_serialize::{Decodable, Decoder, Encodable, Encoder};
35 mod caching_source_map_view;
37 pub use self::caching_source_map_view::CachingSourceMapView;
38 use source_map::SourceMap;
43 use hygiene::Transparency;
44 pub use hygiene::{DesugaringKind, ExpnKind, MacroKind};
45 pub use hygiene::{ExpnData, ExpnHash, ExpnId, LocalExpnId, SyntaxContext};
46 use rustc_data_structures::stable_hasher::HashingControls;
48 use def_id::{CrateNum, DefId, DefPathHash, LocalDefId, LOCAL_CRATE};
51 pub use span_encoding::{Span, DUMMY_SP};
54 pub use symbol::{sym, Symbol};
56 mod analyze_source_file;
61 use rustc_data_structures::stable_hasher::{HashStable, StableHasher};
62 use rustc_data_structures::sync::{Lock, Lrc};
65 use std::cmp::{self, Ordering};
68 use std::ops::{Add, Range, Sub};
69 use std::path::{Path, PathBuf};
70 use std::str::FromStr;
81 /// Per-session global variables: this struct is stored in thread-local storage
82 /// in such a way that it is accessible without any kind of handle to all
83 /// threads within the compilation session, but is not accessible outside the
85 pub struct SessionGlobals {
86 symbol_interner: symbol::Interner,
87 span_interner: Lock<span_encoding::SpanInterner>,
88 hygiene_data: Lock<hygiene::HygieneData>,
89 source_map: Lock<Option<Lrc<SourceMap>>>,
93 pub fn new(edition: Edition) -> SessionGlobals {
95 symbol_interner: symbol::Interner::fresh(),
96 span_interner: Lock::new(span_encoding::SpanInterner::default()),
97 hygiene_data: Lock::new(hygiene::HygieneData::new(edition)),
98 source_map: Lock::new(None),
104 pub fn create_session_globals_then<R>(edition: Edition, f: impl FnOnce() -> R) -> R {
106 !SESSION_GLOBALS.is_set(),
107 "SESSION_GLOBALS should never be overwritten! \
108 Use another thread if you need another SessionGlobals"
110 let session_globals = SessionGlobals::new(edition);
111 SESSION_GLOBALS.set(&session_globals, f)
115 pub fn set_session_globals_then<R>(session_globals: &SessionGlobals, f: impl FnOnce() -> R) -> R {
117 !SESSION_GLOBALS.is_set(),
118 "SESSION_GLOBALS should never be overwritten! \
119 Use another thread if you need another SessionGlobals"
121 SESSION_GLOBALS.set(session_globals, f)
125 pub fn create_default_session_if_not_set_then<R, F>(f: F) -> R
127 F: FnOnce(&SessionGlobals) -> R,
129 create_session_if_not_set_then(edition::DEFAULT_EDITION, f)
133 pub fn create_session_if_not_set_then<R, F>(edition: Edition, f: F) -> R
135 F: FnOnce(&SessionGlobals) -> R,
137 if !SESSION_GLOBALS.is_set() {
138 let session_globals = SessionGlobals::new(edition);
139 SESSION_GLOBALS.set(&session_globals, || SESSION_GLOBALS.with(f))
141 SESSION_GLOBALS.with(f)
146 pub fn with_session_globals<R, F>(f: F) -> R
148 F: FnOnce(&SessionGlobals) -> R,
150 SESSION_GLOBALS.with(f)
154 pub fn create_default_session_globals_then<R>(f: impl FnOnce() -> R) -> R {
155 create_session_globals_then(edition::DEFAULT_EDITION, f)
158 // If this ever becomes non thread-local, `decode_syntax_context`
159 // and `decode_expn_id` will need to be updated to handle concurrent
161 scoped_tls::scoped_thread_local!(static SESSION_GLOBALS: SessionGlobals);
163 // FIXME: We should use this enum or something like it to get rid of the
164 // use of magic `/rust/1.x/...` paths across the board.
165 #[derive(Debug, Eq, PartialEq, Clone, Ord, PartialOrd)]
167 pub enum RealFileName {
169 /// For remapped paths (namely paths into libstd that have been mapped
170 /// to the appropriate spot on the local host's file system, and local file
171 /// system paths that have been remapped with `FilePathMapping`),
173 /// `local_path` is the (host-dependent) local path to the file. This is
174 /// None if the file was imported from another crate
175 local_path: Option<PathBuf>,
176 /// `virtual_name` is the stable path rustc will store internally within
178 virtual_name: PathBuf,
182 impl Hash for RealFileName {
183 fn hash<H: std::hash::Hasher>(&self, state: &mut H) {
184 // To prevent #70924 from happening again we should only hash the
185 // remapped (virtualized) path if that exists. This is because
186 // virtualized paths to sysroot crates (/rust/$hash or /rust/$version)
187 // remain stable even if the corresponding local_path changes
188 self.remapped_path_if_available().hash(state)
192 // This is functionally identical to #[derive(Encodable)], with the exception of
193 // an added assert statement
194 impl<S: Encoder> Encodable<S> for RealFileName {
195 fn encode(&self, encoder: &mut S) {
197 RealFileName::LocalPath(ref local_path) => encoder.emit_enum_variant(0, |encoder| {
198 local_path.encode(encoder);
201 RealFileName::Remapped { ref local_path, ref virtual_name } => encoder
202 .emit_enum_variant(1, |encoder| {
203 // For privacy and build reproducibility, we must not embed host-dependant path in artifacts
204 // if they have been remapped by --remap-path-prefix
205 assert!(local_path.is_none());
206 local_path.encode(encoder);
207 virtual_name.encode(encoder);
214 /// Returns the path suitable for reading from the file system on the local host,
215 /// if this information exists.
216 /// Avoid embedding this in build artifacts; see `remapped_path_if_available()` for that.
217 pub fn local_path(&self) -> Option<&Path> {
219 RealFileName::LocalPath(p) => Some(p),
220 RealFileName::Remapped { local_path, virtual_name: _ } => local_path.as_deref(),
224 /// Returns the path suitable for reading from the file system on the local host,
225 /// if this information exists.
226 /// Avoid embedding this in build artifacts; see `remapped_path_if_available()` for that.
227 pub fn into_local_path(self) -> Option<PathBuf> {
229 RealFileName::LocalPath(p) => Some(p),
230 RealFileName::Remapped { local_path: p, virtual_name: _ } => p,
234 /// Returns the path suitable for embedding into build artifacts. This would still
235 /// be a local path if it has not been remapped. A remapped path will not correspond
236 /// to a valid file system path: see `local_path_if_available()` for something that
237 /// is more likely to return paths into the local host file system.
238 pub fn remapped_path_if_available(&self) -> &Path {
240 RealFileName::LocalPath(p)
241 | RealFileName::Remapped { local_path: _, virtual_name: p } => p,
245 /// Returns the path suitable for reading from the file system on the local host,
246 /// if this information exists. Otherwise returns the remapped name.
247 /// Avoid embedding this in build artifacts; see `remapped_path_if_available()` for that.
248 pub fn local_path_if_available(&self) -> &Path {
250 RealFileName::LocalPath(path)
251 | RealFileName::Remapped { local_path: None, virtual_name: path }
252 | RealFileName::Remapped { local_path: Some(path), virtual_name: _ } => path,
256 pub fn to_string_lossy(&self, display_pref: FileNameDisplayPreference) -> Cow<'_, str> {
258 FileNameDisplayPreference::Local => self.local_path_if_available().to_string_lossy(),
259 FileNameDisplayPreference::Remapped => {
260 self.remapped_path_if_available().to_string_lossy()
266 /// Differentiates between real files and common virtual files.
267 #[derive(Debug, Eq, PartialEq, Clone, Ord, PartialOrd, Hash)]
268 #[derive(Decodable, Encodable)]
271 /// Call to `quote!`.
275 /// Hack in `src/librustc_ast/parse.rs`.
278 ProcMacroSourceCode(u64),
279 /// Strings provided as `--cfg [cfgspec]` stored in a `crate_cfg`.
281 /// Strings provided as crate attributes in the CLI.
283 /// Custom sources for explicit parser calls from plugins and drivers.
285 DocTest(PathBuf, isize),
286 /// Post-substitution inline assembly from LLVM.
290 impl From<PathBuf> for FileName {
291 fn from(p: PathBuf) -> Self {
292 assert!(!p.to_string_lossy().ends_with('>'));
293 FileName::Real(RealFileName::LocalPath(p))
297 #[derive(Clone, Copy, Eq, PartialEq, Hash, Debug)]
298 pub enum FileNameDisplayPreference {
299 /// Display the path after the application of rewrite rules provided via `--remap-path-prefix`.
300 /// This is appropriate for paths that get embedded into files produced by the compiler.
302 /// Display the path before the application of rewrite rules provided via `--remap-path-prefix`.
303 /// This is appropriate for use in user-facing output (such as diagnostics).
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<'a> FileNameDisplay<'a> {
333 pub fn to_string_lossy(&self) -> Cow<'a, str> {
335 FileName::Real(ref inner) => inner.to_string_lossy(self.display_pref),
336 _ => Cow::from(self.to_string()),
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())
521 pub fn lo(self) -> BytePos {
525 pub fn with_lo(self, lo: BytePos) -> Span {
526 self.data().with_lo(lo)
529 pub fn hi(self) -> BytePos {
533 pub fn with_hi(self, hi: BytePos) -> Span {
534 self.data().with_hi(hi)
537 pub fn eq_ctxt(self, other: Span) -> bool {
538 self.data_untracked().ctxt == other.data_untracked().ctxt
541 pub fn with_ctxt(self, ctxt: SyntaxContext) -> Span {
542 self.data_untracked().with_ctxt(ctxt)
545 pub fn parent(self) -> Option<LocalDefId> {
549 pub fn with_parent(self, ctxt: Option<LocalDefId>) -> Span {
550 self.data().with_parent(ctxt)
553 /// Returns `true` if this is a dummy span with any hygienic context.
555 pub fn is_dummy(self) -> bool {
556 self.data_untracked().is_dummy()
559 /// Returns `true` if this span comes from any kind of macro, desugaring or inlining.
561 pub fn from_expansion(self) -> bool {
562 self.ctxt() != SyntaxContext::root()
565 /// Returns `true` if `span` originates in a macro's expansion where debuginfo should be
567 pub fn in_macro_expansion_with_collapse_debuginfo(self) -> bool {
568 let outer_expn = self.ctxt().outer_expn_data();
569 matches!(outer_expn.kind, ExpnKind::Macro(..)) && outer_expn.collapse_debuginfo
572 /// Returns `true` if this span comes from MIR inlining.
573 pub fn is_inlined(self) -> bool {
574 let outer_expn = self.ctxt().outer_expn_data();
575 matches!(outer_expn.kind, ExpnKind::Inlined)
578 /// Returns `true` if `span` originates in a derive-macro's expansion.
579 pub fn in_derive_expansion(self) -> bool {
580 matches!(self.ctxt().outer_expn_data().kind, ExpnKind::Macro(MacroKind::Derive, _))
583 /// Gate suggestions that would not be appropriate in a context the user didn't write.
584 pub fn can_be_used_for_suggestions(self) -> bool {
585 !self.from_expansion()
586 // FIXME: If this span comes from a `derive` macro but it points at code the user wrote,
587 // the callsite span and the span will be pointing at different places. It also means that
588 // we can safely provide suggestions on this span.
589 || (matches!(self.ctxt().outer_expn_data().kind, ExpnKind::Macro(MacroKind::Derive, _))
590 && self.parent_callsite().map(|p| (p.lo(), p.hi())) != Some((self.lo(), self.hi())))
594 pub fn with_root_ctxt(lo: BytePos, hi: BytePos) -> Span {
595 Span::new(lo, hi, SyntaxContext::root(), None)
598 /// Returns a new span representing an empty span at the beginning of this span.
600 pub fn shrink_to_lo(self) -> Span {
601 let span = self.data_untracked();
602 span.with_hi(span.lo)
604 /// Returns a new span representing an empty span at the end of this span.
606 pub fn shrink_to_hi(self) -> Span {
607 let span = self.data_untracked();
608 span.with_lo(span.hi)
612 /// Returns `true` if `hi == lo`.
613 pub fn is_empty(self) -> bool {
614 let span = self.data_untracked();
618 /// Returns `self` if `self` is not the dummy span, and `other` otherwise.
619 pub fn substitute_dummy(self, other: Span) -> Span {
620 if self.is_dummy() { other } else { self }
623 /// Returns `true` if `self` fully encloses `other`.
624 pub fn contains(self, other: Span) -> bool {
625 let span = self.data();
626 let other = other.data();
630 /// Returns `true` if `self` touches `other`.
631 pub fn overlaps(self, other: Span) -> bool {
632 let span = self.data();
633 let other = other.data();
634 span.lo < other.hi && other.lo < span.hi
637 /// Returns `true` if the spans are equal with regards to the source text.
639 /// Use this instead of `==` when either span could be generated code,
640 /// and you only care that they point to the same bytes of source text.
641 pub fn source_equal(self, other: Span) -> bool {
642 let span = self.data();
643 let other = other.data();
644 span.lo == other.lo && span.hi == other.hi
647 /// Returns `Some(span)`, where the start is trimmed by the end of `other`.
648 pub fn trim_start(self, other: Span) -> Option<Span> {
649 let span = self.data();
650 let other = other.data();
651 if span.hi > other.hi { Some(span.with_lo(cmp::max(span.lo, other.hi))) } else { None }
654 /// Returns the source span -- this is either the supplied span, or the span for
655 /// the macro callsite that expanded to it.
656 pub fn source_callsite(self) -> Span {
657 let expn_data = self.ctxt().outer_expn_data();
658 if !expn_data.is_root() { expn_data.call_site.source_callsite() } else { self }
661 /// The `Span` for the tokens in the previous macro expansion from which `self` was generated,
663 pub fn parent_callsite(self) -> Option<Span> {
664 let expn_data = self.ctxt().outer_expn_data();
665 if !expn_data.is_root() { Some(expn_data.call_site) } else { None }
668 /// Walk down the expansion ancestors to find a span that's contained within `outer`.
669 pub fn find_ancestor_inside(mut self, outer: Span) -> Option<Span> {
670 while !outer.contains(self) {
671 self = self.parent_callsite()?;
676 /// Like `find_ancestor_inside`, but specifically for when spans might not
677 /// overlaps. Take care when using this, and prefer `find_ancestor_inside`
678 /// when you know that the spans are nested (modulo macro expansion).
679 pub fn find_ancestor_in_same_ctxt(mut self, other: Span) -> Option<Span> {
680 while !Span::eq_ctxt(self, other) {
681 self = self.parent_callsite()?;
686 /// Edition of the crate from which this span came.
687 pub fn edition(self) -> edition::Edition {
688 self.ctxt().edition()
692 pub fn rust_2015(self) -> bool {
693 self.edition() == edition::Edition::Edition2015
697 pub fn rust_2018(self) -> bool {
698 self.edition() >= edition::Edition::Edition2018
702 pub fn rust_2021(self) -> bool {
703 self.edition() >= edition::Edition::Edition2021
707 pub fn rust_2024(self) -> bool {
708 self.edition() >= edition::Edition::Edition2024
711 /// Returns the source callee.
713 /// Returns `None` if the supplied span has no expansion trace,
714 /// else returns the `ExpnData` for the macro definition
715 /// corresponding to the source callsite.
716 pub fn source_callee(self) -> Option<ExpnData> {
717 fn source_callee(expn_data: ExpnData) -> ExpnData {
718 let next_expn_data = expn_data.call_site.ctxt().outer_expn_data();
719 if !next_expn_data.is_root() { source_callee(next_expn_data) } else { expn_data }
721 let expn_data = self.ctxt().outer_expn_data();
722 if !expn_data.is_root() { Some(source_callee(expn_data)) } else { None }
725 /// Checks if a span is "internal" to a macro in which `#[unstable]`
726 /// items can be used (that is, a macro marked with
727 /// `#[allow_internal_unstable]`).
728 pub fn allows_unstable(self, feature: Symbol) -> bool {
731 .allow_internal_unstable
732 .map_or(false, |features| features.iter().any(|&f| f == feature))
735 /// Checks if this span arises from a compiler desugaring of kind `kind`.
736 pub fn is_desugaring(self, kind: DesugaringKind) -> bool {
737 match self.ctxt().outer_expn_data().kind {
738 ExpnKind::Desugaring(k) => k == kind,
743 /// Returns the compiler desugaring that created this span, or `None`
744 /// if this span is not from a desugaring.
745 pub fn desugaring_kind(self) -> Option<DesugaringKind> {
746 match self.ctxt().outer_expn_data().kind {
747 ExpnKind::Desugaring(k) => Some(k),
752 /// Checks if a span is "internal" to a macro in which `unsafe`
753 /// can be used without triggering the `unsafe_code` lint.
754 /// (that is, a macro marked with `#[allow_internal_unsafe]`).
755 pub fn allows_unsafe(self) -> bool {
756 self.ctxt().outer_expn_data().allow_internal_unsafe
759 pub fn macro_backtrace(mut self) -> impl Iterator<Item = ExpnData> {
760 let mut prev_span = DUMMY_SP;
761 std::iter::from_fn(move || {
763 let expn_data = self.ctxt().outer_expn_data();
764 if expn_data.is_root() {
768 let is_recursive = expn_data.call_site.source_equal(prev_span);
771 self = expn_data.call_site;
773 // Don't print recursive invocations.
775 return Some(expn_data);
781 /// Returns a `Span` that would enclose both `self` and `end`.
785 /// self lorem ipsum end
786 /// ^^^^^^^^^^^^^^^^^^^^
788 pub fn to(self, end: Span) -> Span {
789 let span_data = self.data();
790 let end_data = end.data();
791 // FIXME(jseyfried): `self.ctxt` should always equal `end.ctxt` here (cf. issue #23480).
792 // Return the macro span on its own to avoid weird diagnostic output. It is preferable to
793 // have an incomplete span than a completely nonsensical one.
794 if span_data.ctxt != end_data.ctxt {
795 if span_data.ctxt == SyntaxContext::root() {
797 } else if end_data.ctxt == SyntaxContext::root() {
800 // Both spans fall within a macro.
801 // FIXME(estebank): check if it is the *same* macro.
804 cmp::min(span_data.lo, end_data.lo),
805 cmp::max(span_data.hi, end_data.hi),
806 if span_data.ctxt == SyntaxContext::root() { end_data.ctxt } else { span_data.ctxt },
807 if span_data.parent == end_data.parent { span_data.parent } else { None },
811 /// Returns a `Span` between the end of `self` to the beginning of `end`.
815 /// self lorem ipsum end
818 pub fn between(self, end: Span) -> Span {
819 let span = self.data();
820 let end = end.data();
824 if end.ctxt == SyntaxContext::root() { end.ctxt } else { span.ctxt },
825 if span.parent == end.parent { span.parent } else { None },
829 /// Returns a `Span` from the beginning of `self` until the beginning of `end`.
833 /// self lorem ipsum end
834 /// ^^^^^^^^^^^^^^^^^
836 pub fn until(self, end: Span) -> Span {
837 // Most of this function's body is copied from `to`.
838 // We can't just do `self.to(end.shrink_to_lo())`,
839 // because to also does some magic where it uses min/max so
840 // it can handle overlapping spans. Some advanced mis-use of
841 // `until` with different ctxts makes this visible.
842 let span_data = self.data();
843 let end_data = end.data();
844 // FIXME(jseyfried): `self.ctxt` should always equal `end.ctxt` here (cf. issue #23480).
845 // Return the macro span on its own to avoid weird diagnostic output. It is preferable to
846 // have an incomplete span than a completely nonsensical one.
847 if span_data.ctxt != end_data.ctxt {
848 if span_data.ctxt == SyntaxContext::root() {
850 } else if end_data.ctxt == SyntaxContext::root() {
853 // Both spans fall within a macro.
854 // FIXME(estebank): check if it is the *same* macro.
859 if end_data.ctxt == SyntaxContext::root() { end_data.ctxt } else { span_data.ctxt },
860 if span_data.parent == end_data.parent { span_data.parent } else { None },
864 pub fn from_inner(self, inner: InnerSpan) -> Span {
865 let span = self.data();
867 span.lo + BytePos::from_usize(inner.start),
868 span.lo + BytePos::from_usize(inner.end),
874 /// Equivalent of `Span::def_site` from the proc macro API,
875 /// except that the location is taken from the `self` span.
876 pub fn with_def_site_ctxt(self, expn_id: ExpnId) -> Span {
877 self.with_ctxt_from_mark(expn_id, Transparency::Opaque)
880 /// Equivalent of `Span::call_site` from the proc macro API,
881 /// except that the location is taken from the `self` span.
882 pub fn with_call_site_ctxt(self, expn_id: ExpnId) -> Span {
883 self.with_ctxt_from_mark(expn_id, Transparency::Transparent)
886 /// Equivalent of `Span::mixed_site` from the proc macro API,
887 /// except that the location is taken from the `self` span.
888 pub fn with_mixed_site_ctxt(self, expn_id: ExpnId) -> Span {
889 self.with_ctxt_from_mark(expn_id, Transparency::SemiTransparent)
892 /// Produces a span with the same location as `self` and context produced by a macro with the
893 /// given ID and transparency, assuming that macro was defined directly and not produced by
894 /// some other macro (which is the case for built-in and procedural macros).
895 pub fn with_ctxt_from_mark(self, expn_id: ExpnId, transparency: Transparency) -> Span {
896 self.with_ctxt(SyntaxContext::root().apply_mark(expn_id, transparency))
900 pub fn apply_mark(self, expn_id: ExpnId, transparency: Transparency) -> Span {
901 let span = self.data();
902 span.with_ctxt(span.ctxt.apply_mark(expn_id, transparency))
906 pub fn remove_mark(&mut self) -> ExpnId {
907 let mut span = self.data();
908 let mark = span.ctxt.remove_mark();
909 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
914 pub fn adjust(&mut self, expn_id: ExpnId) -> Option<ExpnId> {
915 let mut span = self.data();
916 let mark = span.ctxt.adjust(expn_id);
917 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
922 pub fn normalize_to_macros_2_0_and_adjust(&mut self, expn_id: ExpnId) -> Option<ExpnId> {
923 let mut span = self.data();
924 let mark = span.ctxt.normalize_to_macros_2_0_and_adjust(expn_id);
925 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
930 pub fn glob_adjust(&mut self, expn_id: ExpnId, glob_span: Span) -> Option<Option<ExpnId>> {
931 let mut span = self.data();
932 let mark = span.ctxt.glob_adjust(expn_id, glob_span);
933 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
938 pub fn reverse_glob_adjust(
942 ) -> Option<Option<ExpnId>> {
943 let mut span = self.data();
944 let mark = span.ctxt.reverse_glob_adjust(expn_id, glob_span);
945 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
950 pub fn normalize_to_macros_2_0(self) -> Span {
951 let span = self.data();
952 span.with_ctxt(span.ctxt.normalize_to_macros_2_0())
956 pub fn normalize_to_macro_rules(self) -> Span {
957 let span = self.data();
958 span.with_ctxt(span.ctxt.normalize_to_macro_rules())
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) {
970 let span = self.data();
975 impl<D: Decoder> Decodable<D> for Span {
976 default fn decode(s: &mut D) -> Span {
977 let lo = Decodable::decode(s);
978 let hi = Decodable::decode(s);
980 Span::new(lo, hi, SyntaxContext::root(), None)
984 /// Calls the provided closure, using the provided `SourceMap` to format
985 /// any spans that are debug-printed during the closure's execution.
987 /// Normally, the global `TyCtxt` is used to retrieve the `SourceMap`
988 /// (see `rustc_interface::callbacks::span_debug1`). However, some parts
989 /// of the compiler (e.g. `rustc_parse`) may debug-print `Span`s before
990 /// a `TyCtxt` is available. In this case, we fall back to
991 /// the `SourceMap` provided to this function. If that is not available,
992 /// we fall back to printing the raw `Span` field values.
993 pub fn with_source_map<T, F: FnOnce() -> T>(source_map: Lrc<SourceMap>, f: F) -> T {
994 with_session_globals(|session_globals| {
995 *session_globals.source_map.borrow_mut() = Some(source_map);
997 struct ClearSourceMap;
998 impl Drop for ClearSourceMap {
1000 with_session_globals(|session_globals| {
1001 session_globals.source_map.borrow_mut().take();
1006 let _guard = ClearSourceMap;
1010 impl fmt::Debug for Span {
1011 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
1012 with_session_globals(|session_globals| {
1013 if let Some(source_map) = &*session_globals.source_map.borrow() {
1014 write!(f, "{} ({:?})", source_map.span_to_diagnostic_string(*self), self.ctxt())
1016 f.debug_struct("Span")
1017 .field("lo", &self.lo())
1018 .field("hi", &self.hi())
1019 .field("ctxt", &self.ctxt())
1026 impl fmt::Debug for SpanData {
1027 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
1028 fmt::Debug::fmt(&Span::new(self.lo, self.hi, self.ctxt, self.parent), f)
1032 /// Identifies an offset of a multi-byte character in a `SourceFile`.
1033 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1034 pub struct MultiByteChar {
1035 /// The absolute offset of the character in the `SourceMap`.
1037 /// The number of bytes, `>= 2`.
1041 /// Identifies an offset of a non-narrow character in a `SourceFile`.
1042 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1043 pub enum NonNarrowChar {
1044 /// Represents a zero-width character.
1046 /// Represents a wide (full-width) character.
1048 /// Represents a tab character, represented visually with a width of 4 characters.
1052 impl NonNarrowChar {
1053 fn new(pos: BytePos, width: usize) -> Self {
1055 0 => NonNarrowChar::ZeroWidth(pos),
1056 2 => NonNarrowChar::Wide(pos),
1057 4 => NonNarrowChar::Tab(pos),
1058 _ => panic!("width {} given for non-narrow character", width),
1062 /// Returns the absolute offset of the character in the `SourceMap`.
1063 pub fn pos(&self) -> BytePos {
1065 NonNarrowChar::ZeroWidth(p) | NonNarrowChar::Wide(p) | NonNarrowChar::Tab(p) => p,
1069 /// Returns the width of the character, 0 (zero-width) or 2 (wide).
1070 pub fn width(&self) -> usize {
1072 NonNarrowChar::ZeroWidth(_) => 0,
1073 NonNarrowChar::Wide(_) => 2,
1074 NonNarrowChar::Tab(_) => 4,
1079 impl Add<BytePos> for NonNarrowChar {
1082 fn add(self, rhs: BytePos) -> Self {
1084 NonNarrowChar::ZeroWidth(pos) => NonNarrowChar::ZeroWidth(pos + rhs),
1085 NonNarrowChar::Wide(pos) => NonNarrowChar::Wide(pos + rhs),
1086 NonNarrowChar::Tab(pos) => NonNarrowChar::Tab(pos + rhs),
1091 impl Sub<BytePos> for NonNarrowChar {
1094 fn sub(self, rhs: BytePos) -> Self {
1096 NonNarrowChar::ZeroWidth(pos) => NonNarrowChar::ZeroWidth(pos - rhs),
1097 NonNarrowChar::Wide(pos) => NonNarrowChar::Wide(pos - rhs),
1098 NonNarrowChar::Tab(pos) => NonNarrowChar::Tab(pos - rhs),
1103 /// Identifies an offset of a character that was normalized away from `SourceFile`.
1104 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1105 pub struct NormalizedPos {
1106 /// The absolute offset of the character in the `SourceMap`.
1108 /// The difference between original and normalized string at position.
1112 #[derive(PartialEq, Eq, Clone, Debug)]
1113 pub enum ExternalSource {
1114 /// No external source has to be loaded, since the `SourceFile` represents a local crate.
1117 kind: ExternalSourceKind,
1118 /// Index of the file inside metadata.
1119 metadata_index: u32,
1123 /// The state of the lazy external source loading mechanism of a `SourceFile`.
1124 #[derive(PartialEq, Eq, Clone, Debug)]
1125 pub enum ExternalSourceKind {
1126 /// The external source has been loaded already.
1127 Present(Lrc<String>),
1128 /// No attempt has been made to load the external source.
1130 /// A failed attempt has been made to load the external source.
1135 impl ExternalSource {
1136 pub fn get_source(&self) -> Option<&Lrc<String>> {
1138 ExternalSource::Foreign { kind: ExternalSourceKind::Present(ref src), .. } => Some(src),
1145 pub struct OffsetOverflowError;
1147 #[derive(Copy, Clone, Debug, PartialEq, Eq, PartialOrd, Ord, Hash, Encodable, Decodable)]
1148 #[derive(HashStable_Generic)]
1149 pub enum SourceFileHashAlgorithm {
1155 impl FromStr for SourceFileHashAlgorithm {
1158 fn from_str(s: &str) -> Result<SourceFileHashAlgorithm, ()> {
1160 "md5" => Ok(SourceFileHashAlgorithm::Md5),
1161 "sha1" => Ok(SourceFileHashAlgorithm::Sha1),
1162 "sha256" => Ok(SourceFileHashAlgorithm::Sha256),
1168 /// The hash of the on-disk source file used for debug info.
1169 #[derive(Copy, Clone, PartialEq, Eq, Debug, Hash)]
1170 #[derive(HashStable_Generic, Encodable, Decodable)]
1171 pub struct SourceFileHash {
1172 pub kind: SourceFileHashAlgorithm,
1176 impl SourceFileHash {
1177 pub fn new(kind: SourceFileHashAlgorithm, src: &str) -> SourceFileHash {
1178 let mut hash = SourceFileHash { kind, value: Default::default() };
1179 let len = hash.hash_len();
1180 let value = &mut hash.value[..len];
1181 let data = src.as_bytes();
1183 SourceFileHashAlgorithm::Md5 => {
1184 value.copy_from_slice(&Md5::digest(data));
1186 SourceFileHashAlgorithm::Sha1 => {
1187 value.copy_from_slice(&Sha1::digest(data));
1189 SourceFileHashAlgorithm::Sha256 => {
1190 value.copy_from_slice(&Sha256::digest(data));
1196 /// Check if the stored hash matches the hash of the string.
1197 pub fn matches(&self, src: &str) -> bool {
1198 Self::new(self.kind, src) == *self
1201 /// The bytes of the hash.
1202 pub fn hash_bytes(&self) -> &[u8] {
1203 let len = self.hash_len();
1207 fn hash_len(&self) -> usize {
1209 SourceFileHashAlgorithm::Md5 => 16,
1210 SourceFileHashAlgorithm::Sha1 => 20,
1211 SourceFileHashAlgorithm::Sha256 => 32,
1216 #[derive(HashStable_Generic)]
1217 #[derive(Copy, PartialEq, PartialOrd, Clone, Ord, Eq, Hash, Debug, Encodable, Decodable)]
1218 pub enum DebuggerVisualizerType {
1223 /// A single debugger visualizer file.
1224 #[derive(HashStable_Generic)]
1225 #[derive(Clone, Debug, Hash, PartialEq, Eq, PartialOrd, Ord, Encodable, Decodable)]
1226 pub struct DebuggerVisualizerFile {
1227 /// The complete debugger visualizer source.
1229 /// Indicates which visualizer type this targets.
1230 pub visualizer_type: DebuggerVisualizerType,
1233 impl DebuggerVisualizerFile {
1234 pub fn new(src: Arc<[u8]>, visualizer_type: DebuggerVisualizerType) -> Self {
1235 DebuggerVisualizerFile { src, visualizer_type }
1240 pub enum SourceFileLines {
1241 /// The source file lines, in decoded (random-access) form.
1242 Lines(Vec<BytePos>),
1244 /// The source file lines, in undecoded difference list form.
1245 Diffs(SourceFileDiffs),
1248 impl SourceFileLines {
1249 pub fn is_lines(&self) -> bool {
1250 matches!(self, SourceFileLines::Lines(_))
1254 /// The source file lines in difference list form. This matches the form
1255 /// used within metadata, which saves space by exploiting the fact that the
1256 /// lines list is sorted and individual lines are usually not that long.
1258 /// We read it directly from metadata and only decode it into `Lines` form
1259 /// when necessary. This is a significant performance win, especially for
1260 /// small crates where very little of `std`'s metadata is used.
1262 pub struct SourceFileDiffs {
1263 /// Position of the first line. Note that this is always encoded as a
1264 /// `BytePos` because it is often much larger than any of the
1266 line_start: BytePos,
1268 /// Always 1, 2, or 4. Always as small as possible, while being big
1269 /// enough to hold the length of the longest line in the source file.
1270 /// The 1 case is by far the most common.
1271 bytes_per_diff: usize,
1273 /// The number of diffs encoded in `raw_diffs`. Always one less than
1274 /// the number of lines in the source file.
1277 /// The diffs in "raw" form. Each segment of `bytes_per_diff` length
1278 /// encodes one little-endian diff. Note that they aren't LEB128
1279 /// encoded. This makes for much faster decoding. Besides, the
1280 /// bytes_per_diff==1 case is by far the most common, and LEB128
1281 /// encoding has no effect on that case.
1285 /// A single source in the [`SourceMap`].
1287 pub struct SourceFile {
1288 /// The name of the file that the source came from. Source that doesn't
1289 /// originate from files has names between angle brackets by convention
1290 /// (e.g., `<anon>`).
1292 /// The complete source code.
1293 pub src: Option<Lrc<String>>,
1294 /// The source code's hash.
1295 pub src_hash: SourceFileHash,
1296 /// The external source code (used for external crates, which will have a `None`
1297 /// value as `self.src`.
1298 pub external_src: Lock<ExternalSource>,
1299 /// The start position of this source in the `SourceMap`.
1300 pub start_pos: BytePos,
1301 /// The end position of this source in the `SourceMap`.
1302 pub end_pos: BytePos,
1303 /// Locations of lines beginnings in the source code.
1304 pub lines: Lock<SourceFileLines>,
1305 /// Locations of multi-byte characters in the source code.
1306 pub multibyte_chars: Vec<MultiByteChar>,
1307 /// Width of characters that are not narrow in the source code.
1308 pub non_narrow_chars: Vec<NonNarrowChar>,
1309 /// Locations of characters removed during normalization.
1310 pub normalized_pos: Vec<NormalizedPos>,
1311 /// A hash of the filename, used for speeding up hashing in incremental compilation.
1312 pub name_hash: u128,
1313 /// Indicates which crate this `SourceFile` was imported from.
1317 impl<S: Encoder> Encodable<S> for SourceFile {
1318 fn encode(&self, s: &mut S) {
1319 self.name.encode(s);
1320 self.src_hash.encode(s);
1321 self.start_pos.encode(s);
1322 self.end_pos.encode(s);
1324 // We are always in `Lines` form by the time we reach here.
1325 assert!(self.lines.borrow().is_lines());
1326 self.lines(|lines| {
1327 // Store the length.
1328 s.emit_u32(lines.len() as u32);
1330 // Compute and store the difference list.
1331 if lines.len() != 0 {
1332 let max_line_length = if lines.len() == 1 {
1337 .map(|&[fst, snd]| snd - fst)
1338 .map(|bp| bp.to_usize())
1343 let bytes_per_diff: usize = match max_line_length {
1345 0x100..=0xFFFF => 2,
1349 // Encode the number of bytes used per diff.
1350 s.emit_u8(bytes_per_diff as u8);
1352 // Encode the first element.
1355 // Encode the difference list.
1356 let diff_iter = lines.array_windows().map(|&[fst, snd]| snd - fst);
1357 let num_diffs = lines.len() - 1;
1359 match bytes_per_diff {
1361 raw_diffs = Vec::with_capacity(num_diffs);
1362 for diff in diff_iter {
1363 raw_diffs.push(diff.0 as u8);
1367 raw_diffs = Vec::with_capacity(bytes_per_diff * num_diffs);
1368 for diff in diff_iter {
1369 raw_diffs.extend_from_slice(&(diff.0 as u16).to_le_bytes());
1373 raw_diffs = Vec::with_capacity(bytes_per_diff * num_diffs);
1374 for diff in diff_iter {
1375 raw_diffs.extend_from_slice(&(diff.0 as u32).to_le_bytes());
1378 _ => unreachable!(),
1380 s.emit_raw_bytes(&raw_diffs);
1384 self.multibyte_chars.encode(s);
1385 self.non_narrow_chars.encode(s);
1386 self.name_hash.encode(s);
1387 self.normalized_pos.encode(s);
1388 self.cnum.encode(s);
1392 impl<D: Decoder> Decodable<D> for SourceFile {
1393 fn decode(d: &mut D) -> SourceFile {
1394 let name: FileName = Decodable::decode(d);
1395 let src_hash: SourceFileHash = Decodable::decode(d);
1396 let start_pos: BytePos = Decodable::decode(d);
1397 let end_pos: BytePos = Decodable::decode(d);
1399 let num_lines: u32 = Decodable::decode(d);
1401 // Read the number of bytes used per diff.
1402 let bytes_per_diff = d.read_u8() as usize;
1404 // Read the first element.
1405 let line_start: BytePos = Decodable::decode(d);
1407 // Read the difference list.
1408 let num_diffs = num_lines as usize - 1;
1409 let raw_diffs = d.read_raw_bytes(bytes_per_diff * num_diffs).to_vec();
1410 SourceFileLines::Diffs(SourceFileDiffs {
1417 SourceFileLines::Lines(vec![])
1420 let multibyte_chars: Vec<MultiByteChar> = Decodable::decode(d);
1421 let non_narrow_chars: Vec<NonNarrowChar> = Decodable::decode(d);
1422 let name_hash: u128 = Decodable::decode(d);
1423 let normalized_pos: Vec<NormalizedPos> = Decodable::decode(d);
1424 let cnum: CrateNum = Decodable::decode(d);
1431 // Unused - the metadata decoder will construct
1432 // a new SourceFile, filling in `external_src` properly
1433 external_src: Lock::new(ExternalSource::Unneeded),
1434 lines: Lock::new(lines),
1444 impl fmt::Debug for SourceFile {
1445 fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result {
1446 write!(fmt, "SourceFile({:?})", self.name)
1455 hash_kind: SourceFileHashAlgorithm,
1457 // Compute the file hash before any normalization.
1458 let src_hash = SourceFileHash::new(hash_kind, &src);
1459 let normalized_pos = normalize_src(&mut src, start_pos);
1462 let mut hasher: StableHasher = StableHasher::new();
1463 name.hash(&mut hasher);
1464 hasher.finish::<u128>()
1466 let end_pos = start_pos.to_usize() + src.len();
1467 assert!(end_pos <= u32::MAX as usize);
1469 let (lines, multibyte_chars, non_narrow_chars) =
1470 analyze_source_file::analyze_source_file(&src, start_pos);
1474 src: Some(Lrc::new(src)),
1476 external_src: Lock::new(ExternalSource::Unneeded),
1478 end_pos: Pos::from_usize(end_pos),
1479 lines: Lock::new(SourceFileLines::Lines(lines)),
1488 pub fn lines<F, R>(&self, f: F) -> R
1490 F: FnOnce(&[BytePos]) -> R,
1492 let mut guard = self.lines.borrow_mut();
1494 SourceFileLines::Lines(lines) => f(lines),
1495 SourceFileLines::Diffs(SourceFileDiffs {
1501 // Convert from "diffs" form to "lines" form.
1502 let num_lines = num_diffs + 1;
1503 let mut lines = Vec::with_capacity(num_lines);
1504 lines.push(line_start);
1506 assert_eq!(*num_diffs, raw_diffs.len() / bytes_per_diff);
1507 match bytes_per_diff {
1509 lines.extend(raw_diffs.into_iter().map(|&diff| {
1510 line_start = line_start + BytePos(diff as u32);
1515 lines.extend((0..*num_diffs).map(|i| {
1516 let pos = bytes_per_diff * i;
1517 let bytes = [raw_diffs[pos], raw_diffs[pos + 1]];
1518 let diff = u16::from_le_bytes(bytes);
1519 line_start = line_start + BytePos(diff as u32);
1524 lines.extend((0..*num_diffs).map(|i| {
1525 let pos = bytes_per_diff * i;
1532 let diff = u32::from_le_bytes(bytes);
1533 line_start = line_start + BytePos(diff);
1537 _ => unreachable!(),
1539 let res = f(&lines);
1540 *guard = SourceFileLines::Lines(lines);
1546 /// Returns the `BytePos` of the beginning of the current line.
1547 pub fn line_begin_pos(&self, pos: BytePos) -> BytePos {
1548 let line_index = self.lookup_line(pos).unwrap();
1549 self.lines(|lines| lines[line_index])
1552 /// Add externally loaded source.
1553 /// If the hash of the input doesn't match or no input is supplied via None,
1554 /// it is interpreted as an error and the corresponding enum variant is set.
1555 /// The return value signifies whether some kind of source is present.
1556 pub fn add_external_src<F>(&self, get_src: F) -> bool
1558 F: FnOnce() -> Option<String>,
1561 *self.external_src.borrow(),
1562 ExternalSource::Foreign { kind: ExternalSourceKind::AbsentOk, .. }
1564 let src = get_src();
1565 let mut external_src = self.external_src.borrow_mut();
1566 // Check that no-one else have provided the source while we were getting it
1567 if let ExternalSource::Foreign {
1568 kind: src_kind @ ExternalSourceKind::AbsentOk, ..
1569 } = &mut *external_src
1571 if let Some(mut src) = src {
1572 // The src_hash needs to be computed on the pre-normalized src.
1573 if self.src_hash.matches(&src) {
1574 normalize_src(&mut src, BytePos::from_usize(0));
1575 *src_kind = ExternalSourceKind::Present(Lrc::new(src));
1579 *src_kind = ExternalSourceKind::AbsentErr;
1584 self.src.is_some() || external_src.get_source().is_some()
1587 self.src.is_some() || self.external_src.borrow().get_source().is_some()
1591 /// Gets a line from the list of pre-computed line-beginnings.
1592 /// The line number here is 0-based.
1593 pub fn get_line(&self, line_number: usize) -> Option<Cow<'_, str>> {
1594 fn get_until_newline(src: &str, begin: usize) -> &str {
1595 // We can't use `lines.get(line_number+1)` because we might
1596 // be parsing when we call this function and thus the current
1597 // line is the last one we have line info for.
1598 let slice = &src[begin..];
1599 match slice.find('\n') {
1600 Some(e) => &slice[..e],
1606 let line = self.lines(|lines| lines.get(line_number).copied())?;
1607 let begin: BytePos = line - self.start_pos;
1611 if let Some(ref src) = self.src {
1612 Some(Cow::from(get_until_newline(src, begin)))
1613 } else if let Some(src) = self.external_src.borrow().get_source() {
1614 Some(Cow::Owned(String::from(get_until_newline(src, begin))))
1620 pub fn is_real_file(&self) -> bool {
1625 pub fn is_imported(&self) -> bool {
1629 pub fn count_lines(&self) -> usize {
1630 self.lines(|lines| lines.len())
1633 /// Finds the line containing the given position. The return value is the
1634 /// index into the `lines` array of this `SourceFile`, not the 1-based line
1635 /// number. If the source_file is empty or the position is located before the
1636 /// first line, `None` is returned.
1637 pub fn lookup_line(&self, pos: BytePos) -> Option<usize> {
1638 self.lines(|lines| lines.partition_point(|x| x <= &pos).checked_sub(1))
1641 pub fn line_bounds(&self, line_index: usize) -> Range<BytePos> {
1642 if self.is_empty() {
1643 return self.start_pos..self.end_pos;
1646 self.lines(|lines| {
1647 assert!(line_index < lines.len());
1648 if line_index == (lines.len() - 1) {
1649 lines[line_index]..self.end_pos
1651 lines[line_index]..lines[line_index + 1]
1656 /// Returns whether or not the file contains the given `SourceMap` byte
1657 /// position. The position one past the end of the file is considered to be
1658 /// contained by the file. This implies that files for which `is_empty`
1659 /// returns true still contain one byte position according to this function.
1661 pub fn contains(&self, byte_pos: BytePos) -> bool {
1662 byte_pos >= self.start_pos && byte_pos <= self.end_pos
1666 pub fn is_empty(&self) -> bool {
1667 self.start_pos == self.end_pos
1670 /// Calculates the original byte position relative to the start of the file
1671 /// based on the given byte position.
1672 pub fn original_relative_byte_pos(&self, pos: BytePos) -> BytePos {
1673 // Diff before any records is 0. Otherwise use the previously recorded
1674 // diff as that applies to the following characters until a new diff
1676 let diff = match self.normalized_pos.binary_search_by(|np| np.pos.cmp(&pos)) {
1677 Ok(i) => self.normalized_pos[i].diff,
1678 Err(i) if i == 0 => 0,
1679 Err(i) => self.normalized_pos[i - 1].diff,
1682 BytePos::from_u32(pos.0 - self.start_pos.0 + diff)
1685 /// Converts an absolute `BytePos` to a `CharPos` relative to the `SourceFile`.
1686 pub fn bytepos_to_file_charpos(&self, bpos: BytePos) -> CharPos {
1687 // The number of extra bytes due to multibyte chars in the `SourceFile`.
1688 let mut total_extra_bytes = 0;
1690 for mbc in self.multibyte_chars.iter() {
1691 debug!("{}-byte char at {:?}", mbc.bytes, mbc.pos);
1693 // Every character is at least one byte, so we only
1694 // count the actual extra bytes.
1695 total_extra_bytes += mbc.bytes as u32 - 1;
1696 // We should never see a byte position in the middle of a
1698 assert!(bpos.to_u32() >= mbc.pos.to_u32() + mbc.bytes as u32);
1704 assert!(self.start_pos.to_u32() + total_extra_bytes <= bpos.to_u32());
1705 CharPos(bpos.to_usize() - self.start_pos.to_usize() - total_extra_bytes as usize)
1708 /// Looks up the file's (1-based) line number and (0-based `CharPos`) column offset, for a
1709 /// given `BytePos`.
1710 pub fn lookup_file_pos(&self, pos: BytePos) -> (usize, CharPos) {
1711 let chpos = self.bytepos_to_file_charpos(pos);
1712 match self.lookup_line(pos) {
1714 let line = a + 1; // Line numbers start at 1
1715 let linebpos = self.lines(|lines| lines[a]);
1716 let linechpos = self.bytepos_to_file_charpos(linebpos);
1717 let col = chpos - linechpos;
1718 debug!("byte pos {:?} is on the line at byte pos {:?}", pos, linebpos);
1719 debug!("char pos {:?} is on the line at char pos {:?}", chpos, linechpos);
1720 debug!("byte is on line: {}", line);
1721 assert!(chpos >= linechpos);
1728 /// Looks up the file's (1-based) line number, (0-based `CharPos`) column offset, and (0-based)
1729 /// column offset when displayed, for a given `BytePos`.
1730 pub fn lookup_file_pos_with_col_display(&self, pos: BytePos) -> (usize, CharPos, usize) {
1731 let (line, col_or_chpos) = self.lookup_file_pos(pos);
1733 let col = col_or_chpos;
1734 let linebpos = self.lines(|lines| lines[line - 1]);
1736 let start_width_idx = self
1738 .binary_search_by_key(&linebpos, |x| x.pos())
1739 .unwrap_or_else(|x| x);
1740 let end_width_idx = self
1742 .binary_search_by_key(&pos, |x| x.pos())
1743 .unwrap_or_else(|x| x);
1744 let special_chars = end_width_idx - start_width_idx;
1745 let non_narrow: usize = self.non_narrow_chars[start_width_idx..end_width_idx]
1749 col.0 - special_chars + non_narrow
1751 (line, col, col_display)
1753 let chpos = col_or_chpos;
1755 let end_width_idx = self
1757 .binary_search_by_key(&pos, |x| x.pos())
1758 .unwrap_or_else(|x| x);
1759 let non_narrow: usize =
1760 self.non_narrow_chars[0..end_width_idx].iter().map(|x| x.width()).sum();
1761 chpos.0 - end_width_idx + non_narrow
1763 (0, chpos, col_display)
1768 /// Normalizes the source code and records the normalizations.
1769 fn normalize_src(src: &mut String, start_pos: BytePos) -> Vec<NormalizedPos> {
1770 let mut normalized_pos = vec![];
1771 remove_bom(src, &mut normalized_pos);
1772 normalize_newlines(src, &mut normalized_pos);
1774 // Offset all the positions by start_pos to match the final file positions.
1775 for np in &mut normalized_pos {
1776 np.pos.0 += start_pos.0;
1782 /// Removes UTF-8 BOM, if any.
1783 fn remove_bom(src: &mut String, normalized_pos: &mut Vec<NormalizedPos>) {
1784 if src.starts_with('\u{feff}') {
1786 normalized_pos.push(NormalizedPos { pos: BytePos(0), diff: 3 });
1790 /// Replaces `\r\n` with `\n` in-place in `src`.
1792 /// Returns error if there's a lone `\r` in the string.
1793 fn normalize_newlines(src: &mut String, normalized_pos: &mut Vec<NormalizedPos>) {
1794 if !src.as_bytes().contains(&b'\r') {
1798 // We replace `\r\n` with `\n` in-place, which doesn't break utf-8 encoding.
1799 // While we *can* call `as_mut_vec` and do surgery on the live string
1800 // directly, let's rather steal the contents of `src`. This makes the code
1801 // safe even if a panic occurs.
1803 let mut buf = std::mem::replace(src, String::new()).into_bytes();
1804 let mut gap_len = 0;
1805 let mut tail = buf.as_mut_slice();
1807 let original_gap = normalized_pos.last().map_or(0, |l| l.diff);
1809 let idx = match find_crlf(&tail[gap_len..]) {
1811 Some(idx) => idx + gap_len,
1813 tail.copy_within(gap_len..idx, 0);
1814 tail = &mut tail[idx - gap_len..];
1815 if tail.len() == gap_len {
1818 cursor += idx - gap_len;
1820 normalized_pos.push(NormalizedPos {
1821 pos: BytePos::from_usize(cursor + 1),
1822 diff: original_gap + gap_len as u32,
1826 // Account for removed `\r`.
1827 // After `set_len`, `buf` is guaranteed to contain utf-8 again.
1828 let new_len = buf.len() - gap_len;
1830 buf.set_len(new_len);
1831 *src = String::from_utf8_unchecked(buf);
1834 fn find_crlf(src: &[u8]) -> Option<usize> {
1835 let mut search_idx = 0;
1836 while let Some(idx) = find_cr(&src[search_idx..]) {
1837 if src[search_idx..].get(idx + 1) != Some(&b'\n') {
1838 search_idx += idx + 1;
1841 return Some(search_idx + idx);
1846 fn find_cr(src: &[u8]) -> Option<usize> {
1847 src.iter().position(|&b| b == b'\r')
1851 // _____________________________________________________________________________
1852 // Pos, BytePos, CharPos
1856 fn from_usize(n: usize) -> Self;
1857 fn to_usize(&self) -> usize;
1858 fn from_u32(n: u32) -> Self;
1859 fn to_u32(&self) -> u32;
1862 macro_rules! impl_pos {
1866 $vis:vis struct $ident:ident($inner_vis:vis $inner_ty:ty);
1871 $vis struct $ident($inner_vis $inner_ty);
1873 impl Pos for $ident {
1875 fn from_usize(n: usize) -> $ident {
1876 $ident(n as $inner_ty)
1880 fn to_usize(&self) -> usize {
1885 fn from_u32(n: u32) -> $ident {
1886 $ident(n as $inner_ty)
1890 fn to_u32(&self) -> u32 {
1895 impl Add for $ident {
1896 type Output = $ident;
1899 fn add(self, rhs: $ident) -> $ident {
1900 $ident(self.0 + rhs.0)
1904 impl Sub for $ident {
1905 type Output = $ident;
1908 fn sub(self, rhs: $ident) -> $ident {
1909 $ident(self.0 - rhs.0)
1919 /// Keep this small (currently 32-bits), as AST contains a lot of them.
1920 #[derive(Clone, Copy, PartialEq, Eq, Hash, PartialOrd, Ord, Debug)]
1921 pub struct BytePos(pub u32);
1923 /// A character offset.
1925 /// Because of multibyte UTF-8 characters, a byte offset
1926 /// is not equivalent to a character offset. The [`SourceMap`] will convert [`BytePos`]
1927 /// values to `CharPos` values as necessary.
1928 #[derive(Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Debug)]
1929 pub struct CharPos(pub usize);
1932 impl<S: Encoder> Encodable<S> for BytePos {
1933 fn encode(&self, s: &mut S) {
1938 impl<D: Decoder> Decodable<D> for BytePos {
1939 fn decode(d: &mut D) -> BytePos {
1940 BytePos(d.read_u32())
1944 // _____________________________________________________________________________
1945 // Loc, SourceFileAndLine, SourceFileAndBytePos
1948 /// A source code location used for error reporting.
1949 #[derive(Debug, Clone)]
1951 /// Information about the original source.
1952 pub file: Lrc<SourceFile>,
1953 /// The (1-based) line number.
1955 /// The (0-based) column offset.
1957 /// The (0-based) column offset when displayed.
1958 pub col_display: usize,
1961 // Used to be structural records.
1963 pub struct SourceFileAndLine {
1964 pub sf: Lrc<SourceFile>,
1965 /// Index of line, starting from 0.
1969 pub struct SourceFileAndBytePos {
1970 pub sf: Lrc<SourceFile>,
1974 #[derive(Copy, Clone, Debug, PartialEq, Eq)]
1975 pub struct LineInfo {
1976 /// Index of line, starting from 0.
1977 pub line_index: usize,
1979 /// Column in line where span begins, starting from 0.
1980 pub start_col: CharPos,
1982 /// Column in line where span ends, starting from 0, exclusive.
1983 pub end_col: CharPos,
1986 pub struct FileLines {
1987 pub file: Lrc<SourceFile>,
1988 pub lines: Vec<LineInfo>,
1991 pub static SPAN_TRACK: AtomicRef<fn(LocalDefId)> = AtomicRef::new(&((|_| {}) as fn(_)));
1993 // _____________________________________________________________________________
1994 // SpanLinesError, SpanSnippetError, DistinctSources, MalformedSourceMapPositions
1997 pub type FileLinesResult = Result<FileLines, SpanLinesError>;
1999 #[derive(Clone, PartialEq, Eq, Debug)]
2000 pub enum SpanLinesError {
2001 DistinctSources(DistinctSources),
2004 #[derive(Clone, PartialEq, Eq, Debug)]
2005 pub enum SpanSnippetError {
2006 IllFormedSpan(Span),
2007 DistinctSources(DistinctSources),
2008 MalformedForSourcemap(MalformedSourceMapPositions),
2009 SourceNotAvailable { filename: FileName },
2012 #[derive(Clone, PartialEq, Eq, Debug)]
2013 pub struct DistinctSources {
2014 pub begin: (FileName, BytePos),
2015 pub end: (FileName, BytePos),
2018 #[derive(Clone, PartialEq, Eq, Debug)]
2019 pub struct MalformedSourceMapPositions {
2021 pub source_len: usize,
2022 pub begin_pos: BytePos,
2023 pub end_pos: BytePos,
2026 /// Range inside of a `Span` used for diagnostics when we only have access to relative positions.
2027 #[derive(Copy, Clone, PartialEq, Eq, Debug)]
2028 pub struct InnerSpan {
2034 pub fn new(start: usize, end: usize) -> InnerSpan {
2035 InnerSpan { start, end }
2039 /// Requirements for a `StableHashingContext` to be used in this crate.
2041 /// This is a hack to allow using the [`HashStable_Generic`] derive macro
2042 /// instead of implementing everything in rustc_middle.
2043 pub trait HashStableContext {
2044 fn def_path_hash(&self, def_id: DefId) -> DefPathHash;
2045 fn hash_spans(&self) -> bool;
2046 /// Accesses `sess.opts.unstable_opts.incremental_ignore_spans` since
2047 /// we don't have easy access to a `Session`
2048 fn unstable_opts_incremental_ignore_spans(&self) -> bool;
2049 fn def_span(&self, def_id: LocalDefId) -> Span;
2050 fn span_data_to_lines_and_cols(
2053 ) -> Option<(Lrc<SourceFile>, usize, BytePos, usize, BytePos)>;
2054 fn hashing_controls(&self) -> HashingControls;
2057 impl<CTX> HashStable<CTX> for Span
2059 CTX: HashStableContext,
2061 /// Hashes a span in a stable way. We can't directly hash the span's `BytePos`
2062 /// fields (that would be similar to hashing pointers, since those are just
2063 /// offsets into the `SourceMap`). Instead, we hash the (file name, line, column)
2064 /// triple, which stays the same even if the containing `SourceFile` has moved
2065 /// within the `SourceMap`.
2067 /// Also note that we are hashing byte offsets for the column, not unicode
2068 /// codepoint offsets. For the purpose of the hash that's sufficient.
2069 /// Also, hashing filenames is expensive so we avoid doing it twice when the
2070 /// span starts and ends in the same file, which is almost always the case.
2071 fn hash_stable(&self, ctx: &mut CTX, hasher: &mut StableHasher) {
2072 const TAG_VALID_SPAN: u8 = 0;
2073 const TAG_INVALID_SPAN: u8 = 1;
2074 const TAG_RELATIVE_SPAN: u8 = 2;
2076 if !ctx.hash_spans() {
2080 let span = self.data_untracked();
2081 span.ctxt.hash_stable(ctx, hasher);
2082 span.parent.hash_stable(ctx, hasher);
2084 if span.is_dummy() {
2085 Hash::hash(&TAG_INVALID_SPAN, hasher);
2089 if let Some(parent) = span.parent {
2090 let def_span = ctx.def_span(parent).data_untracked();
2091 if def_span.contains(span) {
2092 // This span is enclosed in a definition: only hash the relative position.
2093 Hash::hash(&TAG_RELATIVE_SPAN, hasher);
2094 (span.lo - def_span.lo).to_u32().hash_stable(ctx, hasher);
2095 (span.hi - def_span.lo).to_u32().hash_stable(ctx, hasher);
2100 // If this is not an empty or invalid span, we want to hash the last
2101 // position that belongs to it, as opposed to hashing the first
2102 // position past it.
2103 let Some((file, line_lo, col_lo, line_hi, col_hi)) = ctx.span_data_to_lines_and_cols(&span) else {
2104 Hash::hash(&TAG_INVALID_SPAN, hasher);
2108 Hash::hash(&TAG_VALID_SPAN, hasher);
2109 // We truncate the stable ID hash and line and column numbers. The chances
2110 // of causing a collision this way should be minimal.
2111 Hash::hash(&(file.name_hash as u64), hasher);
2113 // Hash both the length and the end location (line/column) of a span. If we
2114 // hash only the length, for example, then two otherwise equal spans with
2115 // different end locations will have the same hash. This can cause a problem
2116 // during incremental compilation wherein a previous result for a query that
2117 // depends on the end location of a span will be incorrectly reused when the
2118 // end location of the span it depends on has changed (see issue #74890). A
2119 // similar analysis applies if some query depends specifically on the length
2120 // of the span, but we only hash the end location. So hash both.
2122 let col_lo_trunc = (col_lo.0 as u64) & 0xFF;
2123 let line_lo_trunc = ((line_lo as u64) & 0xFF_FF_FF) << 8;
2124 let col_hi_trunc = (col_hi.0 as u64) & 0xFF << 32;
2125 let line_hi_trunc = ((line_hi as u64) & 0xFF_FF_FF) << 40;
2126 let col_line = col_lo_trunc | line_lo_trunc | col_hi_trunc | line_hi_trunc;
2127 let len = (span.hi - span.lo).0;
2128 Hash::hash(&col_line, hasher);
2129 Hash::hash(&len, hasher);