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 #![cfg_attr(bootstrap, feature(let_else))]
19 #![feature(if_let_guard)]
20 #![feature(negative_impls)]
21 #![feature(min_specialization)]
22 #![feature(rustc_attrs)]
23 #![deny(rustc::untranslatable_diagnostic)]
24 #![deny(rustc::diagnostic_outside_of_impl)]
27 extern crate rustc_macros;
32 use rustc_data_structures::AtomicRef;
33 use rustc_macros::HashStable_Generic;
34 use rustc_serialize::{Decodable, Decoder, Encodable, Encoder};
36 mod caching_source_map_view;
38 pub use self::caching_source_map_view::CachingSourceMapView;
39 use source_map::SourceMap;
44 use hygiene::Transparency;
45 pub use hygiene::{DesugaringKind, ExpnKind, MacroKind};
46 pub use hygiene::{ExpnData, ExpnHash, ExpnId, LocalExpnId, SyntaxContext};
47 use rustc_data_structures::stable_hasher::HashingControls;
49 use def_id::{CrateNum, DefId, DefPathHash, LocalDefId, LOCAL_CRATE};
52 pub use span_encoding::{Span, DUMMY_SP};
55 pub use symbol::{sym, Symbol};
57 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;
82 // Per-session global variables: this struct is stored in thread-local storage
83 // in such a way that it is accessible without any kind of handle to all
84 // threads within the compilation session, but is not accessible outside the
86 pub struct SessionGlobals {
87 symbol_interner: symbol::Interner,
88 span_interner: Lock<span_encoding::SpanInterner>,
89 hygiene_data: Lock<hygiene::HygieneData>,
90 source_map: Lock<Option<Lrc<SourceMap>>>,
94 pub fn new(edition: Edition) -> SessionGlobals {
96 symbol_interner: symbol::Interner::fresh(),
97 span_interner: Lock::new(span_encoding::SpanInterner::default()),
98 hygiene_data: Lock::new(hygiene::HygieneData::new(edition)),
99 source_map: Lock::new(None),
105 pub fn create_session_globals_then<R>(edition: Edition, f: impl FnOnce() -> R) -> R {
107 !SESSION_GLOBALS.is_set(),
108 "SESSION_GLOBALS should never be overwritten! \
109 Use another thread if you need another SessionGlobals"
111 let session_globals = SessionGlobals::new(edition);
112 SESSION_GLOBALS.set(&session_globals, f)
116 pub fn set_session_globals_then<R>(session_globals: &SessionGlobals, f: impl FnOnce() -> R) -> R {
118 !SESSION_GLOBALS.is_set(),
119 "SESSION_GLOBALS should never be overwritten! \
120 Use another thread if you need another SessionGlobals"
122 SESSION_GLOBALS.set(session_globals, f)
126 pub fn create_default_session_if_not_set_then<R, F>(f: F) -> R
128 F: FnOnce(&SessionGlobals) -> R,
130 create_session_if_not_set_then(edition::DEFAULT_EDITION, f)
134 pub fn create_session_if_not_set_then<R, F>(edition: Edition, f: F) -> R
136 F: FnOnce(&SessionGlobals) -> R,
138 if !SESSION_GLOBALS.is_set() {
139 let session_globals = SessionGlobals::new(edition);
140 SESSION_GLOBALS.set(&session_globals, || SESSION_GLOBALS.with(f))
142 SESSION_GLOBALS.with(f)
147 pub fn with_session_globals<R, F>(f: F) -> R
149 F: FnOnce(&SessionGlobals) -> R,
151 SESSION_GLOBALS.with(f)
155 pub fn create_default_session_globals_then<R>(f: impl FnOnce() -> R) -> R {
156 create_session_globals_then(edition::DEFAULT_EDITION, f)
159 // If this ever becomes non thread-local, `decode_syntax_context`
160 // and `decode_expn_id` will need to be updated to handle concurrent
162 scoped_tls::scoped_thread_local!(static SESSION_GLOBALS: SessionGlobals);
164 // FIXME: We should use this enum or something like it to get rid of the
165 // use of magic `/rust/1.x/...` paths across the board.
166 #[derive(Debug, Eq, PartialEq, Clone, Ord, PartialOrd)]
168 pub enum RealFileName {
170 /// For remapped paths (namely paths into libstd that have been mapped
171 /// to the appropriate spot on the local host's file system, and local file
172 /// system paths that have been remapped with `FilePathMapping`),
174 /// `local_path` is the (host-dependent) local path to the file. This is
175 /// None if the file was imported from another crate
176 local_path: Option<PathBuf>,
177 /// `virtual_name` is the stable path rustc will store internally within
179 virtual_name: PathBuf,
183 impl Hash for RealFileName {
184 fn hash<H: std::hash::Hasher>(&self, state: &mut H) {
185 // To prevent #70924 from happening again we should only hash the
186 // remapped (virtualized) path if that exists. This is because
187 // virtualized paths to sysroot crates (/rust/$hash or /rust/$version)
188 // remain stable even if the corresponding local_path changes
189 self.remapped_path_if_available().hash(state)
193 // This is functionally identical to #[derive(Encodable)], with the exception of
194 // an added assert statement
195 impl<S: Encoder> Encodable<S> for RealFileName {
196 fn encode(&self, encoder: &mut S) {
198 RealFileName::LocalPath(ref local_path) => encoder.emit_enum_variant(0, |encoder| {
199 local_path.encode(encoder);
202 RealFileName::Remapped { ref local_path, ref virtual_name } => encoder
203 .emit_enum_variant(1, |encoder| {
204 // For privacy and build reproducibility, we must not embed host-dependant path in artifacts
205 // if they have been remapped by --remap-path-prefix
206 assert!(local_path.is_none());
207 local_path.encode(encoder);
208 virtual_name.encode(encoder);
215 /// Returns the path suitable for reading from the file system on the local host,
216 /// if this information exists.
217 /// Avoid embedding this in build artifacts; see `remapped_path_if_available()` for that.
218 pub fn local_path(&self) -> Option<&Path> {
220 RealFileName::LocalPath(p) => Some(p),
221 RealFileName::Remapped { local_path: p, virtual_name: _ } => {
222 p.as_ref().map(PathBuf::as_path)
227 /// Returns the path suitable for reading from the file system on the local host,
228 /// if this information exists.
229 /// Avoid embedding this in build artifacts; see `remapped_path_if_available()` for that.
230 pub fn into_local_path(self) -> Option<PathBuf> {
232 RealFileName::LocalPath(p) => Some(p),
233 RealFileName::Remapped { local_path: p, virtual_name: _ } => p,
237 /// Returns the path suitable for embedding into build artifacts. This would still
238 /// be a local path if it has not been remapped. A remapped path will not correspond
239 /// to a valid file system path: see `local_path_if_available()` for something that
240 /// is more likely to return paths into the local host file system.
241 pub fn remapped_path_if_available(&self) -> &Path {
243 RealFileName::LocalPath(p)
244 | RealFileName::Remapped { local_path: _, virtual_name: p } => &p,
248 /// Returns the path suitable for reading from the file system on the local host,
249 /// if this information exists. Otherwise returns the remapped name.
250 /// Avoid embedding this in build artifacts; see `remapped_path_if_available()` for that.
251 pub fn local_path_if_available(&self) -> &Path {
253 RealFileName::LocalPath(path)
254 | RealFileName::Remapped { local_path: None, virtual_name: path }
255 | RealFileName::Remapped { local_path: Some(path), virtual_name: _ } => path,
259 pub fn to_string_lossy(&self, display_pref: FileNameDisplayPreference) -> Cow<'_, str> {
261 FileNameDisplayPreference::Local => self.local_path_if_available().to_string_lossy(),
262 FileNameDisplayPreference::Remapped => {
263 self.remapped_path_if_available().to_string_lossy()
269 /// Differentiates between real files and common virtual files.
270 #[derive(Debug, Eq, PartialEq, Clone, Ord, PartialOrd, Hash)]
271 #[derive(Decodable, Encodable)]
274 /// Call to `quote!`.
278 /// Hack in `src/librustc_ast/parse.rs`.
281 ProcMacroSourceCode(u64),
282 /// Strings provided as `--cfg [cfgspec]` stored in a `crate_cfg`.
284 /// Strings provided as crate attributes in the CLI.
286 /// Custom sources for explicit parser calls from plugins and drivers.
288 DocTest(PathBuf, isize),
289 /// Post-substitution inline assembly from LLVM.
293 impl From<PathBuf> for FileName {
294 fn from(p: PathBuf) -> Self {
295 assert!(!p.to_string_lossy().ends_with('>'));
296 FileName::Real(RealFileName::LocalPath(p))
300 #[derive(Clone, Copy, Eq, PartialEq, Hash, Debug)]
301 pub enum FileNameDisplayPreference {
306 pub struct FileNameDisplay<'a> {
308 display_pref: FileNameDisplayPreference,
311 impl fmt::Display for FileNameDisplay<'_> {
312 fn fmt(&self, fmt: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
316 write!(fmt, "{}", name.to_string_lossy(self.display_pref))
318 QuoteExpansion(_) => write!(fmt, "<quote expansion>"),
319 MacroExpansion(_) => write!(fmt, "<macro expansion>"),
320 Anon(_) => write!(fmt, "<anon>"),
321 ProcMacroSourceCode(_) => write!(fmt, "<proc-macro source code>"),
322 CfgSpec(_) => write!(fmt, "<cfgspec>"),
323 CliCrateAttr(_) => write!(fmt, "<crate attribute>"),
324 Custom(ref s) => write!(fmt, "<{}>", s),
325 DocTest(ref path, _) => write!(fmt, "{}", path.display()),
326 InlineAsm(_) => write!(fmt, "<inline asm>"),
331 impl<'a> FileNameDisplay<'a> {
332 pub fn to_string_lossy(&self) -> Cow<'a, str> {
334 FileName::Real(ref inner) => inner.to_string_lossy(self.display_pref),
335 _ => Cow::from(self.to_string()),
341 pub fn is_real(&self) -> bool {
347 | ProcMacroSourceCode(_)
353 | InlineAsm(_) => false,
357 pub fn prefer_remapped(&self) -> FileNameDisplay<'_> {
358 FileNameDisplay { inner: self, display_pref: FileNameDisplayPreference::Remapped }
361 // This may include transient local filesystem information.
362 // Must not be embedded in build outputs.
363 pub fn prefer_local(&self) -> FileNameDisplay<'_> {
364 FileNameDisplay { inner: self, display_pref: FileNameDisplayPreference::Local }
367 pub fn display(&self, display_pref: FileNameDisplayPreference) -> FileNameDisplay<'_> {
368 FileNameDisplay { inner: self, display_pref }
371 pub fn macro_expansion_source_code(src: &str) -> FileName {
372 let mut hasher = StableHasher::new();
373 src.hash(&mut hasher);
374 FileName::MacroExpansion(hasher.finish())
377 pub fn anon_source_code(src: &str) -> FileName {
378 let mut hasher = StableHasher::new();
379 src.hash(&mut hasher);
380 FileName::Anon(hasher.finish())
383 pub fn proc_macro_source_code(src: &str) -> FileName {
384 let mut hasher = StableHasher::new();
385 src.hash(&mut hasher);
386 FileName::ProcMacroSourceCode(hasher.finish())
389 pub fn cfg_spec_source_code(src: &str) -> FileName {
390 let mut hasher = StableHasher::new();
391 src.hash(&mut hasher);
392 FileName::QuoteExpansion(hasher.finish())
395 pub fn cli_crate_attr_source_code(src: &str) -> FileName {
396 let mut hasher = StableHasher::new();
397 src.hash(&mut hasher);
398 FileName::CliCrateAttr(hasher.finish())
401 pub fn doc_test_source_code(path: PathBuf, line: isize) -> FileName {
402 FileName::DocTest(path, line)
405 pub fn inline_asm_source_code(src: &str) -> FileName {
406 let mut hasher = StableHasher::new();
407 src.hash(&mut hasher);
408 FileName::InlineAsm(hasher.finish())
412 /// Represents a span.
414 /// Spans represent a region of code, used for error reporting. Positions in spans
415 /// are *absolute* positions from the beginning of the [`SourceMap`], not positions
416 /// relative to [`SourceFile`]s. Methods on the `SourceMap` can be used to relate spans back
417 /// to the original source.
419 /// You must be careful if the span crosses more than one file, since you will not be
420 /// able to use many of the functions on spans in source_map and you cannot assume
421 /// that the length of the span is equal to `span.hi - span.lo`; there may be space in the
422 /// [`BytePos`] range between files.
424 /// `SpanData` is public because `Span` uses a thread-local interner and can't be
425 /// sent to other threads, but some pieces of performance infra run in a separate thread.
426 /// Using `Span` is generally preferred.
427 #[derive(Clone, Copy, Hash, PartialEq, Eq)]
428 pub struct SpanData {
431 /// Information about where the macro came from, if this piece of
432 /// code was created by a macro expansion.
433 pub ctxt: SyntaxContext,
434 pub parent: Option<LocalDefId>,
437 // Order spans by position in the file.
438 impl Ord for SpanData {
439 fn cmp(&self, other: &Self) -> Ordering {
444 // `LocalDefId` does not implement `Ord`.
445 // The other fields are enough to determine in-file order.
452 // `LocalDefId` does not implement `Ord`.
453 // The other fields are enough to determine in-file order.
457 (s_lo, s_hi, s_ctxt).cmp(&(o_lo, o_hi, o_ctxt))
461 impl PartialOrd for SpanData {
462 fn partial_cmp(&self, other: &Self) -> Option<Ordering> {
463 Some(self.cmp(other))
469 pub fn span(&self) -> Span {
470 Span::new(self.lo, self.hi, self.ctxt, self.parent)
473 pub fn with_lo(&self, lo: BytePos) -> Span {
474 Span::new(lo, self.hi, self.ctxt, self.parent)
477 pub fn with_hi(&self, hi: BytePos) -> Span {
478 Span::new(self.lo, hi, self.ctxt, self.parent)
481 pub fn with_ctxt(&self, ctxt: SyntaxContext) -> Span {
482 Span::new(self.lo, self.hi, ctxt, self.parent)
485 pub fn with_parent(&self, parent: Option<LocalDefId>) -> Span {
486 Span::new(self.lo, self.hi, self.ctxt, parent)
488 /// Returns `true` if this is a dummy span with any hygienic context.
490 pub fn is_dummy(self) -> bool {
491 self.lo.0 == 0 && self.hi.0 == 0
493 /// Returns `true` if `self` fully encloses `other`.
494 pub fn contains(self, other: Self) -> bool {
495 self.lo <= other.lo && other.hi <= self.hi
499 // The interner is pointed to by a thread local value which is only set on the main thread
500 // with parallelization is disabled. So we don't allow `Span` to transfer between threads
501 // to avoid panics and other errors, even though it would be memory safe to do so.
502 #[cfg(not(parallel_compiler))]
503 impl !Send for Span {}
504 #[cfg(not(parallel_compiler))]
505 impl !Sync for Span {}
507 impl PartialOrd for Span {
508 fn partial_cmp(&self, rhs: &Self) -> Option<Ordering> {
509 PartialOrd::partial_cmp(&self.data(), &rhs.data())
513 fn cmp(&self, rhs: &Self) -> Ordering {
514 Ord::cmp(&self.data(), &rhs.data())
520 pub fn lo(self) -> BytePos {
524 pub fn with_lo(self, lo: BytePos) -> Span {
525 self.data().with_lo(lo)
528 pub fn hi(self) -> BytePos {
532 pub fn with_hi(self, hi: BytePos) -> Span {
533 self.data().with_hi(hi)
536 pub fn ctxt(self) -> SyntaxContext {
537 self.data_untracked().ctxt
539 pub fn eq_ctxt(self, other: Span) -> bool {
540 self.data_untracked().ctxt == other.data_untracked().ctxt
543 pub fn with_ctxt(self, ctxt: SyntaxContext) -> Span {
544 self.data_untracked().with_ctxt(ctxt)
547 pub fn parent(self) -> Option<LocalDefId> {
551 pub fn with_parent(self, ctxt: Option<LocalDefId>) -> Span {
552 self.data().with_parent(ctxt)
555 /// Returns `true` if this is a dummy span with any hygienic context.
557 pub fn is_dummy(self) -> bool {
558 self.data_untracked().is_dummy()
561 /// Returns `true` if this span comes from a macro or desugaring.
563 pub fn from_expansion(self) -> bool {
564 self.ctxt() != SyntaxContext::root()
567 /// Returns `true` if `span` originates in a macro's expansion where debuginfo should be
569 pub fn in_macro_expansion_with_collapse_debuginfo(self) -> bool {
570 let outer_expn = self.ctxt().outer_expn_data();
571 matches!(outer_expn.kind, ExpnKind::Macro(..)) && outer_expn.collapse_debuginfo
574 /// Returns `true` if `span` originates in a derive-macro's expansion.
575 pub fn in_derive_expansion(self) -> bool {
576 matches!(self.ctxt().outer_expn_data().kind, ExpnKind::Macro(MacroKind::Derive, _))
579 /// Gate suggestions that would not be appropriate in a context the user didn't write.
580 pub fn can_be_used_for_suggestions(self) -> bool {
581 !self.from_expansion()
582 // FIXME: If this span comes from a `derive` macro but it points at code the user wrote,
583 // the callsite span and the span will be pointing at different places. It also means that
584 // we can safely provide suggestions on this span.
585 || (matches!(self.ctxt().outer_expn_data().kind, ExpnKind::Macro(MacroKind::Derive, _))
586 && self.parent_callsite().map(|p| (p.lo(), p.hi())) != Some((self.lo(), self.hi())))
590 pub fn with_root_ctxt(lo: BytePos, hi: BytePos) -> Span {
591 Span::new(lo, hi, SyntaxContext::root(), None)
594 /// Returns a new span representing an empty span at the beginning of this span.
596 pub fn shrink_to_lo(self) -> Span {
597 let span = self.data_untracked();
598 span.with_hi(span.lo)
600 /// Returns a new span representing an empty span at the end of this span.
602 pub fn shrink_to_hi(self) -> Span {
603 let span = self.data_untracked();
604 span.with_lo(span.hi)
608 /// Returns `true` if `hi == lo`.
609 pub fn is_empty(self) -> bool {
610 let span = self.data_untracked();
614 /// Returns `self` if `self` is not the dummy span, and `other` otherwise.
615 pub fn substitute_dummy(self, other: Span) -> Span {
616 if self.is_dummy() { other } else { self }
619 /// Returns `true` if `self` fully encloses `other`.
620 pub fn contains(self, other: Span) -> bool {
621 let span = self.data();
622 let other = other.data();
626 /// Returns `true` if `self` touches `other`.
627 pub fn overlaps(self, other: Span) -> bool {
628 let span = self.data();
629 let other = other.data();
630 span.lo < other.hi && other.lo < span.hi
633 /// Returns `true` if the spans are equal with regards to the source text.
635 /// Use this instead of `==` when either span could be generated code,
636 /// and you only care that they point to the same bytes of source text.
637 pub fn source_equal(self, other: Span) -> bool {
638 let span = self.data();
639 let other = other.data();
640 span.lo == other.lo && span.hi == other.hi
643 /// Returns `Some(span)`, where the start is trimmed by the end of `other`.
644 pub fn trim_start(self, other: Span) -> Option<Span> {
645 let span = self.data();
646 let other = other.data();
647 if span.hi > other.hi { Some(span.with_lo(cmp::max(span.lo, other.hi))) } else { None }
650 /// Returns the source span -- this is either the supplied span, or the span for
651 /// the macro callsite that expanded to it.
652 pub fn source_callsite(self) -> Span {
653 let expn_data = self.ctxt().outer_expn_data();
654 if !expn_data.is_root() { expn_data.call_site.source_callsite() } else { self }
657 /// The `Span` for the tokens in the previous macro expansion from which `self` was generated,
659 pub fn parent_callsite(self) -> Option<Span> {
660 let expn_data = self.ctxt().outer_expn_data();
661 if !expn_data.is_root() { Some(expn_data.call_site) } else { None }
664 /// Walk down the expansion ancestors to find a span that's contained within `outer`.
665 pub fn find_ancestor_inside(mut self, outer: Span) -> Option<Span> {
666 while !outer.contains(self) {
667 self = self.parent_callsite()?;
672 /// Like `find_ancestor_inside`, but specifically for when spans might not
673 /// overlaps. Take care when using this, and prefer `find_ancestor_inside`
674 /// when you know that the spans are nested (modulo macro expansion).
675 pub fn find_ancestor_in_same_ctxt(mut self, other: Span) -> Option<Span> {
676 while !Span::eq_ctxt(self, other) {
677 self = self.parent_callsite()?;
682 /// Edition of the crate from which this span came.
683 pub fn edition(self) -> edition::Edition {
684 self.ctxt().edition()
688 pub fn rust_2015(self) -> bool {
689 self.edition() == edition::Edition::Edition2015
693 pub fn rust_2018(self) -> bool {
694 self.edition() >= edition::Edition::Edition2018
698 pub fn rust_2021(self) -> bool {
699 self.edition() >= edition::Edition::Edition2021
703 pub fn rust_2024(self) -> bool {
704 self.edition() >= edition::Edition::Edition2024
707 /// Returns the source callee.
709 /// Returns `None` if the supplied span has no expansion trace,
710 /// else returns the `ExpnData` for the macro definition
711 /// corresponding to the source callsite.
712 pub fn source_callee(self) -> Option<ExpnData> {
713 fn source_callee(expn_data: ExpnData) -> ExpnData {
714 let next_expn_data = expn_data.call_site.ctxt().outer_expn_data();
715 if !next_expn_data.is_root() { source_callee(next_expn_data) } else { expn_data }
717 let expn_data = self.ctxt().outer_expn_data();
718 if !expn_data.is_root() { Some(source_callee(expn_data)) } else { None }
721 /// Checks if a span is "internal" to a macro in which `#[unstable]`
722 /// items can be used (that is, a macro marked with
723 /// `#[allow_internal_unstable]`).
724 pub fn allows_unstable(self, feature: Symbol) -> bool {
727 .allow_internal_unstable
728 .map_or(false, |features| features.iter().any(|&f| f == feature))
731 /// Checks if this span arises from a compiler desugaring of kind `kind`.
732 pub fn is_desugaring(self, kind: DesugaringKind) -> bool {
733 match self.ctxt().outer_expn_data().kind {
734 ExpnKind::Desugaring(k) => k == kind,
739 /// Returns the compiler desugaring that created this span, or `None`
740 /// if this span is not from a desugaring.
741 pub fn desugaring_kind(self) -> Option<DesugaringKind> {
742 match self.ctxt().outer_expn_data().kind {
743 ExpnKind::Desugaring(k) => Some(k),
748 /// Checks if a span is "internal" to a macro in which `unsafe`
749 /// can be used without triggering the `unsafe_code` lint.
750 // (that is, a macro marked with `#[allow_internal_unsafe]`).
751 pub fn allows_unsafe(self) -> bool {
752 self.ctxt().outer_expn_data().allow_internal_unsafe
755 pub fn macro_backtrace(mut self) -> impl Iterator<Item = ExpnData> {
756 let mut prev_span = DUMMY_SP;
757 std::iter::from_fn(move || {
759 let expn_data = self.ctxt().outer_expn_data();
760 if expn_data.is_root() {
764 let is_recursive = expn_data.call_site.source_equal(prev_span);
767 self = expn_data.call_site;
769 // Don't print recursive invocations.
771 return Some(expn_data);
777 /// Returns a `Span` that would enclose both `self` and `end`.
781 /// self lorem ipsum end
782 /// ^^^^^^^^^^^^^^^^^^^^
784 pub fn to(self, end: Span) -> Span {
785 let span_data = self.data();
786 let end_data = end.data();
787 // FIXME(jseyfried): `self.ctxt` should always equal `end.ctxt` here (cf. issue #23480).
788 // Return the macro span on its own to avoid weird diagnostic output. It is preferable to
789 // have an incomplete span than a completely nonsensical one.
790 if span_data.ctxt != end_data.ctxt {
791 if span_data.ctxt == SyntaxContext::root() {
793 } else if end_data.ctxt == SyntaxContext::root() {
796 // Both spans fall within a macro.
797 // FIXME(estebank): check if it is the *same* macro.
800 cmp::min(span_data.lo, end_data.lo),
801 cmp::max(span_data.hi, end_data.hi),
802 if span_data.ctxt == SyntaxContext::root() { end_data.ctxt } else { span_data.ctxt },
803 if span_data.parent == end_data.parent { span_data.parent } else { None },
807 /// Returns a `Span` between the end of `self` to the beginning of `end`.
811 /// self lorem ipsum end
814 pub fn between(self, end: Span) -> Span {
815 let span = self.data();
816 let end = end.data();
820 if end.ctxt == SyntaxContext::root() { end.ctxt } else { span.ctxt },
821 if span.parent == end.parent { span.parent } else { None },
825 /// Returns a `Span` from the beginning of `self` until the beginning of `end`.
829 /// self lorem ipsum end
830 /// ^^^^^^^^^^^^^^^^^
832 pub fn until(self, end: Span) -> Span {
833 // Most of this function's body is copied from `to`.
834 // We can't just do `self.to(end.shrink_to_lo())`,
835 // because to also does some magic where it uses min/max so
836 // it can handle overlapping spans. Some advanced mis-use of
837 // `until` with different ctxts makes this visible.
838 let span_data = self.data();
839 let end_data = end.data();
840 // FIXME(jseyfried): `self.ctxt` should always equal `end.ctxt` here (cf. issue #23480).
841 // Return the macro span on its own to avoid weird diagnostic output. It is preferable to
842 // have an incomplete span than a completely nonsensical one.
843 if span_data.ctxt != end_data.ctxt {
844 if span_data.ctxt == SyntaxContext::root() {
846 } else if end_data.ctxt == SyntaxContext::root() {
849 // Both spans fall within a macro.
850 // FIXME(estebank): check if it is the *same* macro.
855 if end_data.ctxt == SyntaxContext::root() { end_data.ctxt } else { span_data.ctxt },
856 if span_data.parent == end_data.parent { span_data.parent } else { None },
860 pub fn from_inner(self, inner: InnerSpan) -> Span {
861 let span = self.data();
863 span.lo + BytePos::from_usize(inner.start),
864 span.lo + BytePos::from_usize(inner.end),
870 /// Equivalent of `Span::def_site` from the proc macro API,
871 /// except that the location is taken from the `self` span.
872 pub fn with_def_site_ctxt(self, expn_id: ExpnId) -> Span {
873 self.with_ctxt_from_mark(expn_id, Transparency::Opaque)
876 /// Equivalent of `Span::call_site` from the proc macro API,
877 /// except that the location is taken from the `self` span.
878 pub fn with_call_site_ctxt(self, expn_id: ExpnId) -> Span {
879 self.with_ctxt_from_mark(expn_id, Transparency::Transparent)
882 /// Equivalent of `Span::mixed_site` from the proc macro API,
883 /// except that the location is taken from the `self` span.
884 pub fn with_mixed_site_ctxt(self, expn_id: ExpnId) -> Span {
885 self.with_ctxt_from_mark(expn_id, Transparency::SemiTransparent)
888 /// Produces a span with the same location as `self` and context produced by a macro with the
889 /// given ID and transparency, assuming that macro was defined directly and not produced by
890 /// some other macro (which is the case for built-in and procedural macros).
891 pub fn with_ctxt_from_mark(self, expn_id: ExpnId, transparency: Transparency) -> Span {
892 self.with_ctxt(SyntaxContext::root().apply_mark(expn_id, transparency))
896 pub fn apply_mark(self, expn_id: ExpnId, transparency: Transparency) -> Span {
897 let span = self.data();
898 span.with_ctxt(span.ctxt.apply_mark(expn_id, transparency))
902 pub fn remove_mark(&mut self) -> ExpnId {
903 let mut span = self.data();
904 let mark = span.ctxt.remove_mark();
905 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
910 pub fn adjust(&mut self, expn_id: ExpnId) -> Option<ExpnId> {
911 let mut span = self.data();
912 let mark = span.ctxt.adjust(expn_id);
913 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
918 pub fn normalize_to_macros_2_0_and_adjust(&mut self, expn_id: ExpnId) -> Option<ExpnId> {
919 let mut span = self.data();
920 let mark = span.ctxt.normalize_to_macros_2_0_and_adjust(expn_id);
921 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
926 pub fn glob_adjust(&mut self, expn_id: ExpnId, glob_span: Span) -> Option<Option<ExpnId>> {
927 let mut span = self.data();
928 let mark = span.ctxt.glob_adjust(expn_id, glob_span);
929 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
934 pub fn reverse_glob_adjust(
938 ) -> Option<Option<ExpnId>> {
939 let mut span = self.data();
940 let mark = span.ctxt.reverse_glob_adjust(expn_id, glob_span);
941 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
946 pub fn normalize_to_macros_2_0(self) -> Span {
947 let span = self.data();
948 span.with_ctxt(span.ctxt.normalize_to_macros_2_0())
952 pub fn normalize_to_macro_rules(self) -> Span {
953 let span = self.data();
954 span.with_ctxt(span.ctxt.normalize_to_macro_rules())
958 impl Default for Span {
959 fn default() -> Self {
964 impl<E: Encoder> Encodable<E> for Span {
965 default fn encode(&self, s: &mut E) {
966 let span = self.data();
971 impl<D: Decoder> Decodable<D> for Span {
972 default fn decode(s: &mut D) -> Span {
973 let lo = Decodable::decode(s);
974 let hi = Decodable::decode(s);
976 Span::new(lo, hi, SyntaxContext::root(), None)
980 /// Calls the provided closure, using the provided `SourceMap` to format
981 /// any spans that are debug-printed during the closure's execution.
983 /// Normally, the global `TyCtxt` is used to retrieve the `SourceMap`
984 /// (see `rustc_interface::callbacks::span_debug1`). However, some parts
985 /// of the compiler (e.g. `rustc_parse`) may debug-print `Span`s before
986 /// a `TyCtxt` is available. In this case, we fall back to
987 /// the `SourceMap` provided to this function. If that is not available,
988 /// we fall back to printing the raw `Span` field values.
989 pub fn with_source_map<T, F: FnOnce() -> T>(source_map: Lrc<SourceMap>, f: F) -> T {
990 with_session_globals(|session_globals| {
991 *session_globals.source_map.borrow_mut() = Some(source_map);
993 struct ClearSourceMap;
994 impl Drop for ClearSourceMap {
996 with_session_globals(|session_globals| {
997 session_globals.source_map.borrow_mut().take();
1002 let _guard = ClearSourceMap;
1006 impl fmt::Debug for Span {
1007 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
1008 with_session_globals(|session_globals| {
1009 if let Some(source_map) = &*session_globals.source_map.borrow() {
1010 write!(f, "{} ({:?})", source_map.span_to_diagnostic_string(*self), self.ctxt())
1012 f.debug_struct("Span")
1013 .field("lo", &self.lo())
1014 .field("hi", &self.hi())
1015 .field("ctxt", &self.ctxt())
1022 impl fmt::Debug for SpanData {
1023 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
1024 fmt::Debug::fmt(&Span::new(self.lo, self.hi, self.ctxt, self.parent), f)
1028 /// Identifies an offset of a multi-byte character in a `SourceFile`.
1029 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1030 pub struct MultiByteChar {
1031 /// The absolute offset of the character in the `SourceMap`.
1033 /// The number of bytes, `>= 2`.
1037 /// Identifies an offset of a non-narrow character in a `SourceFile`.
1038 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1039 pub enum NonNarrowChar {
1040 /// Represents a zero-width character.
1042 /// Represents a wide (full-width) character.
1044 /// Represents a tab character, represented visually with a width of 4 characters.
1048 impl NonNarrowChar {
1049 fn new(pos: BytePos, width: usize) -> Self {
1051 0 => NonNarrowChar::ZeroWidth(pos),
1052 2 => NonNarrowChar::Wide(pos),
1053 4 => NonNarrowChar::Tab(pos),
1054 _ => panic!("width {} given for non-narrow character", width),
1058 /// Returns the absolute offset of the character in the `SourceMap`.
1059 pub fn pos(&self) -> BytePos {
1061 NonNarrowChar::ZeroWidth(p) | NonNarrowChar::Wide(p) | NonNarrowChar::Tab(p) => p,
1065 /// Returns the width of the character, 0 (zero-width) or 2 (wide).
1066 pub fn width(&self) -> usize {
1068 NonNarrowChar::ZeroWidth(_) => 0,
1069 NonNarrowChar::Wide(_) => 2,
1070 NonNarrowChar::Tab(_) => 4,
1075 impl Add<BytePos> for NonNarrowChar {
1078 fn add(self, rhs: BytePos) -> Self {
1080 NonNarrowChar::ZeroWidth(pos) => NonNarrowChar::ZeroWidth(pos + rhs),
1081 NonNarrowChar::Wide(pos) => NonNarrowChar::Wide(pos + rhs),
1082 NonNarrowChar::Tab(pos) => NonNarrowChar::Tab(pos + rhs),
1087 impl Sub<BytePos> for NonNarrowChar {
1090 fn sub(self, rhs: BytePos) -> Self {
1092 NonNarrowChar::ZeroWidth(pos) => NonNarrowChar::ZeroWidth(pos - rhs),
1093 NonNarrowChar::Wide(pos) => NonNarrowChar::Wide(pos - rhs),
1094 NonNarrowChar::Tab(pos) => NonNarrowChar::Tab(pos - rhs),
1099 /// Identifies an offset of a character that was normalized away from `SourceFile`.
1100 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1101 pub struct NormalizedPos {
1102 /// The absolute offset of the character in the `SourceMap`.
1104 /// The difference between original and normalized string at position.
1108 #[derive(PartialEq, Eq, Clone, Debug)]
1109 pub enum ExternalSource {
1110 /// No external source has to be loaded, since the `SourceFile` represents a local crate.
1113 kind: ExternalSourceKind,
1114 /// Index of the file inside metadata.
1115 metadata_index: u32,
1119 /// The state of the lazy external source loading mechanism of a `SourceFile`.
1120 #[derive(PartialEq, Eq, Clone, Debug)]
1121 pub enum ExternalSourceKind {
1122 /// The external source has been loaded already.
1123 Present(Lrc<String>),
1124 /// No attempt has been made to load the external source.
1126 /// A failed attempt has been made to load the external source.
1131 impl ExternalSource {
1132 pub fn get_source(&self) -> Option<&Lrc<String>> {
1134 ExternalSource::Foreign { kind: ExternalSourceKind::Present(ref src), .. } => Some(src),
1141 pub struct OffsetOverflowError;
1143 #[derive(Copy, Clone, Debug, PartialEq, Eq, PartialOrd, Ord, Hash, Encodable, Decodable)]
1144 #[derive(HashStable_Generic)]
1145 pub enum SourceFileHashAlgorithm {
1151 impl FromStr for SourceFileHashAlgorithm {
1154 fn from_str(s: &str) -> Result<SourceFileHashAlgorithm, ()> {
1156 "md5" => Ok(SourceFileHashAlgorithm::Md5),
1157 "sha1" => Ok(SourceFileHashAlgorithm::Sha1),
1158 "sha256" => Ok(SourceFileHashAlgorithm::Sha256),
1164 /// The hash of the on-disk source file used for debug info.
1165 #[derive(Copy, Clone, PartialEq, Eq, Debug, Hash)]
1166 #[derive(HashStable_Generic, Encodable, Decodable)]
1167 pub struct SourceFileHash {
1168 pub kind: SourceFileHashAlgorithm,
1172 impl SourceFileHash {
1173 pub fn new(kind: SourceFileHashAlgorithm, src: &str) -> SourceFileHash {
1174 let mut hash = SourceFileHash { kind, value: Default::default() };
1175 let len = hash.hash_len();
1176 let value = &mut hash.value[..len];
1177 let data = src.as_bytes();
1179 SourceFileHashAlgorithm::Md5 => {
1180 value.copy_from_slice(&Md5::digest(data));
1182 SourceFileHashAlgorithm::Sha1 => {
1183 value.copy_from_slice(&Sha1::digest(data));
1185 SourceFileHashAlgorithm::Sha256 => {
1186 value.copy_from_slice(&Sha256::digest(data));
1192 /// Check if the stored hash matches the hash of the string.
1193 pub fn matches(&self, src: &str) -> bool {
1194 Self::new(self.kind, src) == *self
1197 /// The bytes of the hash.
1198 pub fn hash_bytes(&self) -> &[u8] {
1199 let len = self.hash_len();
1203 fn hash_len(&self) -> usize {
1205 SourceFileHashAlgorithm::Md5 => 16,
1206 SourceFileHashAlgorithm::Sha1 => 20,
1207 SourceFileHashAlgorithm::Sha256 => 32,
1212 #[derive(HashStable_Generic)]
1213 #[derive(Copy, PartialEq, PartialOrd, Clone, Ord, Eq, Hash, Debug, Encodable, Decodable)]
1214 pub enum DebuggerVisualizerType {
1219 /// A single debugger visualizer file.
1220 #[derive(HashStable_Generic)]
1221 #[derive(Clone, Debug, Hash, PartialEq, Eq, PartialOrd, Ord, Encodable, Decodable)]
1222 pub struct DebuggerVisualizerFile {
1223 /// The complete debugger visualizer source.
1225 /// Indicates which visualizer type this targets.
1226 pub visualizer_type: DebuggerVisualizerType,
1229 impl DebuggerVisualizerFile {
1230 pub fn new(src: Arc<[u8]>, visualizer_type: DebuggerVisualizerType) -> Self {
1231 DebuggerVisualizerFile { src, visualizer_type }
1236 pub enum SourceFileLines {
1237 /// The source file lines, in decoded (random-access) form.
1238 Lines(Vec<BytePos>),
1240 /// The source file lines, in undecoded difference list form.
1241 Diffs(SourceFileDiffs),
1244 impl SourceFileLines {
1245 pub fn is_lines(&self) -> bool {
1246 matches!(self, SourceFileLines::Lines(_))
1250 /// The source file lines in difference list form. This matches the form
1251 /// used within metadata, which saves space by exploiting the fact that the
1252 /// lines list is sorted and individual lines are usually not that long.
1254 /// We read it directly from metadata and only decode it into `Lines` form
1255 /// when necessary. This is a significant performance win, especially for
1256 /// small crates where very little of `std`'s metadata is used.
1258 pub struct SourceFileDiffs {
1259 /// Position of the first line. Note that this is always encoded as a
1260 /// `BytePos` because it is often much larger than any of the
1262 line_start: BytePos,
1264 /// Always 1, 2, or 4. Always as small as possible, while being big
1265 /// enough to hold the length of the longest line in the source file.
1266 /// The 1 case is by far the most common.
1267 bytes_per_diff: usize,
1269 /// The number of diffs encoded in `raw_diffs`. Always one less than
1270 /// the number of lines in the source file.
1273 /// The diffs in "raw" form. Each segment of `bytes_per_diff` length
1274 /// encodes one little-endian diff. Note that they aren't LEB128
1275 /// encoded. This makes for much faster decoding. Besides, the
1276 /// bytes_per_diff==1 case is by far the most common, and LEB128
1277 /// encoding has no effect on that case.
1281 /// A single source in the [`SourceMap`].
1283 pub struct SourceFile {
1284 /// The name of the file that the source came from. Source that doesn't
1285 /// originate from files has names between angle brackets by convention
1286 /// (e.g., `<anon>`).
1288 /// The complete source code.
1289 pub src: Option<Lrc<String>>,
1290 /// The source code's hash.
1291 pub src_hash: SourceFileHash,
1292 /// The external source code (used for external crates, which will have a `None`
1293 /// value as `self.src`.
1294 pub external_src: Lock<ExternalSource>,
1295 /// The start position of this source in the `SourceMap`.
1296 pub start_pos: BytePos,
1297 /// The end position of this source in the `SourceMap`.
1298 pub end_pos: BytePos,
1299 /// Locations of lines beginnings in the source code.
1300 pub lines: Lock<SourceFileLines>,
1301 /// Locations of multi-byte characters in the source code.
1302 pub multibyte_chars: Vec<MultiByteChar>,
1303 /// Width of characters that are not narrow in the source code.
1304 pub non_narrow_chars: Vec<NonNarrowChar>,
1305 /// Locations of characters removed during normalization.
1306 pub normalized_pos: Vec<NormalizedPos>,
1307 /// A hash of the filename, used for speeding up hashing in incremental compilation.
1308 pub name_hash: u128,
1309 /// Indicates which crate this `SourceFile` was imported from.
1313 impl<S: Encoder> Encodable<S> for SourceFile {
1314 fn encode(&self, s: &mut S) {
1315 self.name.encode(s);
1316 self.src_hash.encode(s);
1317 self.start_pos.encode(s);
1318 self.end_pos.encode(s);
1320 // We are always in `Lines` form by the time we reach here.
1321 assert!(self.lines.borrow().is_lines());
1322 self.lines(|lines| {
1323 // Store the length.
1324 s.emit_u32(lines.len() as u32);
1326 // Compute and store the difference list.
1327 if lines.len() != 0 {
1328 let max_line_length = if lines.len() == 1 {
1333 .map(|&[fst, snd]| snd - fst)
1334 .map(|bp| bp.to_usize())
1339 let bytes_per_diff: usize = match max_line_length {
1341 0x100..=0xFFFF => 2,
1345 // Encode the number of bytes used per diff.
1346 s.emit_u8(bytes_per_diff as u8);
1348 // Encode the first element.
1351 // Encode the difference list.
1352 let diff_iter = lines.array_windows().map(|&[fst, snd]| snd - fst);
1353 let num_diffs = lines.len() - 1;
1355 match bytes_per_diff {
1357 raw_diffs = Vec::with_capacity(num_diffs);
1358 for diff in diff_iter {
1359 raw_diffs.push(diff.0 as u8);
1363 raw_diffs = Vec::with_capacity(bytes_per_diff * num_diffs);
1364 for diff in diff_iter {
1365 raw_diffs.extend_from_slice(&(diff.0 as u16).to_le_bytes());
1369 raw_diffs = Vec::with_capacity(bytes_per_diff * num_diffs);
1370 for diff in diff_iter {
1371 raw_diffs.extend_from_slice(&(diff.0 as u32).to_le_bytes());
1374 _ => unreachable!(),
1376 s.emit_raw_bytes(&raw_diffs);
1380 self.multibyte_chars.encode(s);
1381 self.non_narrow_chars.encode(s);
1382 self.name_hash.encode(s);
1383 self.normalized_pos.encode(s);
1384 self.cnum.encode(s);
1388 impl<D: Decoder> Decodable<D> for SourceFile {
1389 fn decode(d: &mut D) -> SourceFile {
1390 let name: FileName = Decodable::decode(d);
1391 let src_hash: SourceFileHash = Decodable::decode(d);
1392 let start_pos: BytePos = Decodable::decode(d);
1393 let end_pos: BytePos = Decodable::decode(d);
1395 let num_lines: u32 = Decodable::decode(d);
1397 // Read the number of bytes used per diff.
1398 let bytes_per_diff = d.read_u8() as usize;
1400 // Read the first element.
1401 let line_start: BytePos = Decodable::decode(d);
1403 // Read the difference list.
1404 let num_diffs = num_lines as usize - 1;
1405 let raw_diffs = d.read_raw_bytes(bytes_per_diff * num_diffs).to_vec();
1406 SourceFileLines::Diffs(SourceFileDiffs {
1413 SourceFileLines::Lines(vec![])
1416 let multibyte_chars: Vec<MultiByteChar> = Decodable::decode(d);
1417 let non_narrow_chars: Vec<NonNarrowChar> = Decodable::decode(d);
1418 let name_hash: u128 = Decodable::decode(d);
1419 let normalized_pos: Vec<NormalizedPos> = Decodable::decode(d);
1420 let cnum: CrateNum = Decodable::decode(d);
1427 // Unused - the metadata decoder will construct
1428 // a new SourceFile, filling in `external_src` properly
1429 external_src: Lock::new(ExternalSource::Unneeded),
1430 lines: Lock::new(lines),
1440 impl fmt::Debug for SourceFile {
1441 fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result {
1442 write!(fmt, "SourceFile({:?})", self.name)
1451 hash_kind: SourceFileHashAlgorithm,
1453 // Compute the file hash before any normalization.
1454 let src_hash = SourceFileHash::new(hash_kind, &src);
1455 let normalized_pos = normalize_src(&mut src, start_pos);
1458 let mut hasher: StableHasher = StableHasher::new();
1459 name.hash(&mut hasher);
1460 hasher.finish::<u128>()
1462 let end_pos = start_pos.to_usize() + src.len();
1463 assert!(end_pos <= u32::MAX as usize);
1465 let (lines, multibyte_chars, non_narrow_chars) =
1466 analyze_source_file::analyze_source_file(&src, start_pos);
1470 src: Some(Lrc::new(src)),
1472 external_src: Lock::new(ExternalSource::Unneeded),
1474 end_pos: Pos::from_usize(end_pos),
1475 lines: Lock::new(SourceFileLines::Lines(lines)),
1484 pub fn lines<F, R>(&self, f: F) -> R
1486 F: FnOnce(&[BytePos]) -> R,
1488 let mut guard = self.lines.borrow_mut();
1490 SourceFileLines::Lines(lines) => f(lines),
1491 SourceFileLines::Diffs(SourceFileDiffs {
1497 // Convert from "diffs" form to "lines" form.
1498 let num_lines = num_diffs + 1;
1499 let mut lines = Vec::with_capacity(num_lines);
1500 lines.push(line_start);
1502 assert_eq!(*num_diffs, raw_diffs.len() / bytes_per_diff);
1503 match bytes_per_diff {
1505 lines.extend(raw_diffs.into_iter().map(|&diff| {
1506 line_start = line_start + BytePos(diff as u32);
1511 lines.extend((0..*num_diffs).map(|i| {
1512 let pos = bytes_per_diff * i;
1513 let bytes = [raw_diffs[pos], raw_diffs[pos + 1]];
1514 let diff = u16::from_le_bytes(bytes);
1515 line_start = line_start + BytePos(diff as u32);
1520 lines.extend((0..*num_diffs).map(|i| {
1521 let pos = bytes_per_diff * i;
1528 let diff = u32::from_le_bytes(bytes);
1529 line_start = line_start + BytePos(diff);
1533 _ => unreachable!(),
1535 let res = f(&lines);
1536 *guard = SourceFileLines::Lines(lines);
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(|lines| 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(|lines| lines.get(line_number).copied())?;
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 {
1621 pub fn is_imported(&self) -> bool {
1625 pub fn count_lines(&self) -> usize {
1626 self.lines(|lines| lines.len())
1629 /// Finds the line containing the given position. The return value is the
1630 /// index into the `lines` array of this `SourceFile`, not the 1-based line
1631 /// number. If the source_file is empty or the position is located before the
1632 /// first line, `None` is returned.
1633 pub fn lookup_line(&self, pos: BytePos) -> Option<usize> {
1634 self.lines(|lines| match lines.partition_point(|x| x <= &pos) {
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 self.lines(|lines| {
1646 assert!(line_index < lines.len());
1647 if line_index == (lines.len() - 1) {
1648 lines[line_index]..self.end_pos
1650 lines[line_index]..lines[line_index + 1]
1655 /// Returns whether or not the file contains the given `SourceMap` byte
1656 /// position. The position one past the end of the file is considered to be
1657 /// contained by the file. This implies that files for which `is_empty`
1658 /// returns true still contain one byte position according to this function.
1660 pub fn contains(&self, byte_pos: BytePos) -> bool {
1661 byte_pos >= self.start_pos && byte_pos <= self.end_pos
1665 pub fn is_empty(&self) -> bool {
1666 self.start_pos == self.end_pos
1669 /// Calculates the original byte position relative to the start of the file
1670 /// based on the given byte position.
1671 pub fn original_relative_byte_pos(&self, pos: BytePos) -> BytePos {
1672 // Diff before any records is 0. Otherwise use the previously recorded
1673 // diff as that applies to the following characters until a new diff
1675 let diff = match self.normalized_pos.binary_search_by(|np| np.pos.cmp(&pos)) {
1676 Ok(i) => self.normalized_pos[i].diff,
1677 Err(i) if i == 0 => 0,
1678 Err(i) => self.normalized_pos[i - 1].diff,
1681 BytePos::from_u32(pos.0 - self.start_pos.0 + diff)
1684 /// Converts an absolute `BytePos` to a `CharPos` relative to the `SourceFile`.
1685 pub fn bytepos_to_file_charpos(&self, bpos: BytePos) -> CharPos {
1686 // The number of extra bytes due to multibyte chars in the `SourceFile`.
1687 let mut total_extra_bytes = 0;
1689 for mbc in self.multibyte_chars.iter() {
1690 debug!("{}-byte char at {:?}", mbc.bytes, mbc.pos);
1692 // Every character is at least one byte, so we only
1693 // count the actual extra bytes.
1694 total_extra_bytes += mbc.bytes as u32 - 1;
1695 // We should never see a byte position in the middle of a
1697 assert!(bpos.to_u32() >= mbc.pos.to_u32() + mbc.bytes as u32);
1703 assert!(self.start_pos.to_u32() + total_extra_bytes <= bpos.to_u32());
1704 CharPos(bpos.to_usize() - self.start_pos.to_usize() - total_extra_bytes as usize)
1707 /// Looks up the file's (1-based) line number and (0-based `CharPos`) column offset, for a
1708 /// given `BytePos`.
1709 pub fn lookup_file_pos(&self, pos: BytePos) -> (usize, CharPos) {
1710 let chpos = self.bytepos_to_file_charpos(pos);
1711 match self.lookup_line(pos) {
1713 let line = a + 1; // Line numbers start at 1
1714 let linebpos = self.lines(|lines| lines[a]);
1715 let linechpos = self.bytepos_to_file_charpos(linebpos);
1716 let col = chpos - linechpos;
1717 debug!("byte pos {:?} is on the line at byte pos {:?}", pos, linebpos);
1718 debug!("char pos {:?} is on the line at char pos {:?}", chpos, linechpos);
1719 debug!("byte is on line: {}", line);
1720 assert!(chpos >= linechpos);
1727 /// Looks up the file's (1-based) line number, (0-based `CharPos`) column offset, and (0-based)
1728 /// column offset when displayed, for a given `BytePos`.
1729 pub fn lookup_file_pos_with_col_display(&self, pos: BytePos) -> (usize, CharPos, usize) {
1730 let (line, col_or_chpos) = self.lookup_file_pos(pos);
1732 let col = col_or_chpos;
1733 let linebpos = self.lines(|lines| lines[line - 1]);
1735 let start_width_idx = self
1737 .binary_search_by_key(&linebpos, |x| x.pos())
1738 .unwrap_or_else(|x| x);
1739 let end_width_idx = self
1741 .binary_search_by_key(&pos, |x| x.pos())
1742 .unwrap_or_else(|x| x);
1743 let special_chars = end_width_idx - start_width_idx;
1744 let non_narrow: usize = self.non_narrow_chars[start_width_idx..end_width_idx]
1748 col.0 - special_chars + non_narrow
1750 (line, col, col_display)
1752 let chpos = col_or_chpos;
1754 let end_width_idx = self
1756 .binary_search_by_key(&pos, |x| x.pos())
1757 .unwrap_or_else(|x| x);
1758 let non_narrow: usize =
1759 self.non_narrow_chars[0..end_width_idx].iter().map(|x| x.width()).sum();
1760 chpos.0 - end_width_idx + non_narrow
1762 (0, chpos, col_display)
1767 /// Normalizes the source code and records the normalizations.
1768 fn normalize_src(src: &mut String, start_pos: BytePos) -> Vec<NormalizedPos> {
1769 let mut normalized_pos = vec![];
1770 remove_bom(src, &mut normalized_pos);
1771 normalize_newlines(src, &mut normalized_pos);
1773 // Offset all the positions by start_pos to match the final file positions.
1774 for np in &mut normalized_pos {
1775 np.pos.0 += start_pos.0;
1781 /// Removes UTF-8 BOM, if any.
1782 fn remove_bom(src: &mut String, normalized_pos: &mut Vec<NormalizedPos>) {
1783 if src.starts_with('\u{feff}') {
1785 normalized_pos.push(NormalizedPos { pos: BytePos(0), diff: 3 });
1789 /// Replaces `\r\n` with `\n` in-place in `src`.
1791 /// Returns error if there's a lone `\r` in the string.
1792 fn normalize_newlines(src: &mut String, normalized_pos: &mut Vec<NormalizedPos>) {
1793 if !src.as_bytes().contains(&b'\r') {
1797 // We replace `\r\n` with `\n` in-place, which doesn't break utf-8 encoding.
1798 // While we *can* call `as_mut_vec` and do surgery on the live string
1799 // directly, let's rather steal the contents of `src`. This makes the code
1800 // safe even if a panic occurs.
1802 let mut buf = std::mem::replace(src, String::new()).into_bytes();
1803 let mut gap_len = 0;
1804 let mut tail = buf.as_mut_slice();
1806 let original_gap = normalized_pos.last().map_or(0, |l| l.diff);
1808 let idx = match find_crlf(&tail[gap_len..]) {
1810 Some(idx) => idx + gap_len,
1812 tail.copy_within(gap_len..idx, 0);
1813 tail = &mut tail[idx - gap_len..];
1814 if tail.len() == gap_len {
1817 cursor += idx - gap_len;
1819 normalized_pos.push(NormalizedPos {
1820 pos: BytePos::from_usize(cursor + 1),
1821 diff: original_gap + gap_len as u32,
1825 // Account for removed `\r`.
1826 // After `set_len`, `buf` is guaranteed to contain utf-8 again.
1827 let new_len = buf.len() - gap_len;
1829 buf.set_len(new_len);
1830 *src = String::from_utf8_unchecked(buf);
1833 fn find_crlf(src: &[u8]) -> Option<usize> {
1834 let mut search_idx = 0;
1835 while let Some(idx) = find_cr(&src[search_idx..]) {
1836 if src[search_idx..].get(idx + 1) != Some(&b'\n') {
1837 search_idx += idx + 1;
1840 return Some(search_idx + idx);
1845 fn find_cr(src: &[u8]) -> Option<usize> {
1846 src.iter().position(|&b| b == b'\r')
1850 // _____________________________________________________________________________
1851 // Pos, BytePos, CharPos
1855 fn from_usize(n: usize) -> Self;
1856 fn to_usize(&self) -> usize;
1857 fn from_u32(n: u32) -> Self;
1858 fn to_u32(&self) -> u32;
1861 macro_rules! impl_pos {
1865 $vis:vis struct $ident:ident($inner_vis:vis $inner_ty:ty);
1870 $vis struct $ident($inner_vis $inner_ty);
1872 impl Pos for $ident {
1874 fn from_usize(n: usize) -> $ident {
1875 $ident(n as $inner_ty)
1879 fn to_usize(&self) -> usize {
1884 fn from_u32(n: u32) -> $ident {
1885 $ident(n as $inner_ty)
1889 fn to_u32(&self) -> u32 {
1894 impl Add for $ident {
1895 type Output = $ident;
1898 fn add(self, rhs: $ident) -> $ident {
1899 $ident(self.0 + rhs.0)
1903 impl Sub for $ident {
1904 type Output = $ident;
1907 fn sub(self, rhs: $ident) -> $ident {
1908 $ident(self.0 - rhs.0)
1918 /// Keep this small (currently 32-bits), as AST contains a lot of them.
1919 #[derive(Clone, Copy, PartialEq, Eq, Hash, PartialOrd, Ord, Debug)]
1920 pub struct BytePos(pub u32);
1922 /// A character offset.
1924 /// Because of multibyte UTF-8 characters, a byte offset
1925 /// is not equivalent to a character offset. The [`SourceMap`] will convert [`BytePos`]
1926 /// values to `CharPos` values as necessary.
1927 #[derive(Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Debug)]
1928 pub struct CharPos(pub usize);
1931 impl<S: Encoder> Encodable<S> for BytePos {
1932 fn encode(&self, s: &mut S) {
1937 impl<D: Decoder> Decodable<D> for BytePos {
1938 fn decode(d: &mut D) -> BytePos {
1939 BytePos(d.read_u32())
1943 // _____________________________________________________________________________
1944 // Loc, SourceFileAndLine, SourceFileAndBytePos
1947 /// A source code location used for error reporting.
1948 #[derive(Debug, Clone)]
1950 /// Information about the original source.
1951 pub file: Lrc<SourceFile>,
1952 /// The (1-based) line number.
1954 /// The (0-based) column offset.
1956 /// The (0-based) column offset when displayed.
1957 pub col_display: usize,
1960 // Used to be structural records.
1962 pub struct SourceFileAndLine {
1963 pub sf: Lrc<SourceFile>,
1964 /// Index of line, starting from 0.
1968 pub struct SourceFileAndBytePos {
1969 pub sf: Lrc<SourceFile>,
1973 #[derive(Copy, Clone, Debug, PartialEq, Eq)]
1974 pub struct LineInfo {
1975 /// Index of line, starting from 0.
1976 pub line_index: usize,
1978 /// Column in line where span begins, starting from 0.
1979 pub start_col: CharPos,
1981 /// Column in line where span ends, starting from 0, exclusive.
1982 pub end_col: CharPos,
1985 pub struct FileLines {
1986 pub file: Lrc<SourceFile>,
1987 pub lines: Vec<LineInfo>,
1990 pub static SPAN_TRACK: AtomicRef<fn(LocalDefId)> = AtomicRef::new(&((|_| {}) as fn(_)));
1992 // _____________________________________________________________________________
1993 // SpanLinesError, SpanSnippetError, DistinctSources, MalformedSourceMapPositions
1996 pub type FileLinesResult = Result<FileLines, SpanLinesError>;
1998 #[derive(Clone, PartialEq, Eq, Debug)]
1999 pub enum SpanLinesError {
2000 DistinctSources(DistinctSources),
2003 #[derive(Clone, PartialEq, Eq, Debug)]
2004 pub enum SpanSnippetError {
2005 IllFormedSpan(Span),
2006 DistinctSources(DistinctSources),
2007 MalformedForSourcemap(MalformedSourceMapPositions),
2008 SourceNotAvailable { filename: FileName },
2011 #[derive(Clone, PartialEq, Eq, Debug)]
2012 pub struct DistinctSources {
2013 pub begin: (FileName, BytePos),
2014 pub end: (FileName, BytePos),
2017 #[derive(Clone, PartialEq, Eq, Debug)]
2018 pub struct MalformedSourceMapPositions {
2020 pub source_len: usize,
2021 pub begin_pos: BytePos,
2022 pub end_pos: BytePos,
2025 /// Range inside of a `Span` used for diagnostics when we only have access to relative positions.
2026 #[derive(Copy, Clone, PartialEq, Eq, Debug)]
2027 pub struct InnerSpan {
2033 pub fn new(start: usize, end: usize) -> InnerSpan {
2034 InnerSpan { start, end }
2038 /// Requirements for a `StableHashingContext` to be used in this crate.
2040 /// This is a hack to allow using the [`HashStable_Generic`] derive macro
2041 /// instead of implementing everything in rustc_middle.
2042 pub trait HashStableContext {
2043 fn def_path_hash(&self, def_id: DefId) -> DefPathHash;
2044 fn hash_spans(&self) -> bool;
2045 /// Accesses `sess.opts.unstable_opts.incremental_ignore_spans` since
2046 /// we don't have easy access to a `Session`
2047 fn unstable_opts_incremental_ignore_spans(&self) -> bool;
2048 fn def_span(&self, def_id: LocalDefId) -> Span;
2049 fn span_data_to_lines_and_cols(
2052 ) -> Option<(Lrc<SourceFile>, usize, BytePos, usize, BytePos)>;
2053 fn hashing_controls(&self) -> HashingControls;
2056 impl<CTX> HashStable<CTX> for Span
2058 CTX: HashStableContext,
2060 /// Hashes a span in a stable way. We can't directly hash the span's `BytePos`
2061 /// fields (that would be similar to hashing pointers, since those are just
2062 /// offsets into the `SourceMap`). Instead, we hash the (file name, line, column)
2063 /// triple, which stays the same even if the containing `SourceFile` has moved
2064 /// within the `SourceMap`.
2066 /// Also note that we are hashing byte offsets for the column, not unicode
2067 /// codepoint offsets. For the purpose of the hash that's sufficient.
2068 /// Also, hashing filenames is expensive so we avoid doing it twice when the
2069 /// span starts and ends in the same file, which is almost always the case.
2070 fn hash_stable(&self, ctx: &mut CTX, hasher: &mut StableHasher) {
2071 const TAG_VALID_SPAN: u8 = 0;
2072 const TAG_INVALID_SPAN: u8 = 1;
2073 const TAG_RELATIVE_SPAN: u8 = 2;
2075 if !ctx.hash_spans() {
2079 let span = self.data_untracked();
2080 span.ctxt.hash_stable(ctx, hasher);
2081 span.parent.hash_stable(ctx, hasher);
2083 if span.is_dummy() {
2084 Hash::hash(&TAG_INVALID_SPAN, hasher);
2088 if let Some(parent) = span.parent {
2089 let def_span = ctx.def_span(parent).data_untracked();
2090 if def_span.contains(span) {
2091 // This span is enclosed in a definition: only hash the relative position.
2092 Hash::hash(&TAG_RELATIVE_SPAN, hasher);
2093 (span.lo - def_span.lo).to_u32().hash_stable(ctx, hasher);
2094 (span.hi - def_span.lo).to_u32().hash_stable(ctx, hasher);
2099 // If this is not an empty or invalid span, we want to hash the last
2100 // position that belongs to it, as opposed to hashing the first
2101 // position past it.
2102 let Some((file, line_lo, col_lo, line_hi, col_hi)) = ctx.span_data_to_lines_and_cols(&span) else {
2103 Hash::hash(&TAG_INVALID_SPAN, hasher);
2107 Hash::hash(&TAG_VALID_SPAN, hasher);
2108 // We truncate the stable ID hash and line and column numbers. The chances
2109 // of causing a collision this way should be minimal.
2110 Hash::hash(&(file.name_hash as u64), hasher);
2112 // Hash both the length and the end location (line/column) of a span. If we
2113 // hash only the length, for example, then two otherwise equal spans with
2114 // different end locations will have the same hash. This can cause a problem
2115 // during incremental compilation wherein a previous result for a query that
2116 // depends on the end location of a span will be incorrectly reused when the
2117 // end location of the span it depends on has changed (see issue #74890). A
2118 // similar analysis applies if some query depends specifically on the length
2119 // of the span, but we only hash the end location. So hash both.
2121 let col_lo_trunc = (col_lo.0 as u64) & 0xFF;
2122 let line_lo_trunc = ((line_lo as u64) & 0xFF_FF_FF) << 8;
2123 let col_hi_trunc = (col_hi.0 as u64) & 0xFF << 32;
2124 let line_hi_trunc = ((line_hi as u64) & 0xFF_FF_FF) << 40;
2125 let col_line = col_lo_trunc | line_lo_trunc | col_hi_trunc | line_hi_trunc;
2126 let len = (span.hi - span.lo).0;
2127 Hash::hash(&col_line, hasher);
2128 Hash::hash(&len, hasher);