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: p, virtual_name: _ } => {
221 p.as_ref().map(PathBuf::as_path)
226 /// Returns the path suitable for reading from the file system on the local host,
227 /// if this information exists.
228 /// Avoid embedding this in build artifacts; see `remapped_path_if_available()` for that.
229 pub fn into_local_path(self) -> Option<PathBuf> {
231 RealFileName::LocalPath(p) => Some(p),
232 RealFileName::Remapped { local_path: p, virtual_name: _ } => p,
236 /// Returns the path suitable for embedding into build artifacts. This would still
237 /// be a local path if it has not been remapped. A remapped path will not correspond
238 /// to a valid file system path: see `local_path_if_available()` for something that
239 /// is more likely to return paths into the local host file system.
240 pub fn remapped_path_if_available(&self) -> &Path {
242 RealFileName::LocalPath(p)
243 | RealFileName::Remapped { local_path: _, virtual_name: p } => &p,
247 /// Returns the path suitable for reading from the file system on the local host,
248 /// if this information exists. Otherwise returns the remapped name.
249 /// Avoid embedding this in build artifacts; see `remapped_path_if_available()` for that.
250 pub fn local_path_if_available(&self) -> &Path {
252 RealFileName::LocalPath(path)
253 | RealFileName::Remapped { local_path: None, virtual_name: path }
254 | RealFileName::Remapped { local_path: Some(path), virtual_name: _ } => path,
258 pub fn to_string_lossy(&self, display_pref: FileNameDisplayPreference) -> Cow<'_, str> {
260 FileNameDisplayPreference::Local => self.local_path_if_available().to_string_lossy(),
261 FileNameDisplayPreference::Remapped => {
262 self.remapped_path_if_available().to_string_lossy()
268 /// Differentiates between real files and common virtual files.
269 #[derive(Debug, Eq, PartialEq, Clone, Ord, PartialOrd, Hash)]
270 #[derive(Decodable, Encodable)]
273 /// Call to `quote!`.
277 /// Hack in `src/librustc_ast/parse.rs`.
280 ProcMacroSourceCode(u64),
281 /// Strings provided as `--cfg [cfgspec]` stored in a `crate_cfg`.
283 /// Strings provided as crate attributes in the CLI.
285 /// Custom sources for explicit parser calls from plugins and drivers.
287 DocTest(PathBuf, isize),
288 /// Post-substitution inline assembly from LLVM.
292 impl From<PathBuf> for FileName {
293 fn from(p: PathBuf) -> Self {
294 assert!(!p.to_string_lossy().ends_with('>'));
295 FileName::Real(RealFileName::LocalPath(p))
299 #[derive(Clone, Copy, Eq, PartialEq, Hash, Debug)]
300 pub enum FileNameDisplayPreference {
301 /// Display the path after the application of rewrite rules provided via `--remap-path-prefix`.
302 /// This is appropriate for paths that get embedded into files produced by the compiler.
304 /// Display the path before the application of rewrite rules provided via `--remap-path-prefix`.
305 /// This is appropriate for use in user-facing output (such as diagnostics).
309 pub struct FileNameDisplay<'a> {
311 display_pref: FileNameDisplayPreference,
314 impl fmt::Display for FileNameDisplay<'_> {
315 fn fmt(&self, fmt: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
319 write!(fmt, "{}", name.to_string_lossy(self.display_pref))
321 QuoteExpansion(_) => write!(fmt, "<quote expansion>"),
322 MacroExpansion(_) => write!(fmt, "<macro expansion>"),
323 Anon(_) => write!(fmt, "<anon>"),
324 ProcMacroSourceCode(_) => write!(fmt, "<proc-macro source code>"),
325 CfgSpec(_) => write!(fmt, "<cfgspec>"),
326 CliCrateAttr(_) => write!(fmt, "<crate attribute>"),
327 Custom(ref s) => write!(fmt, "<{}>", s),
328 DocTest(ref path, _) => write!(fmt, "{}", path.display()),
329 InlineAsm(_) => write!(fmt, "<inline asm>"),
334 impl<'a> FileNameDisplay<'a> {
335 pub fn to_string_lossy(&self) -> Cow<'a, str> {
337 FileName::Real(ref inner) => inner.to_string_lossy(self.display_pref),
338 _ => Cow::from(self.to_string()),
344 pub fn is_real(&self) -> bool {
350 | ProcMacroSourceCode(_)
356 | InlineAsm(_) => false,
360 pub fn prefer_remapped(&self) -> FileNameDisplay<'_> {
361 FileNameDisplay { inner: self, display_pref: FileNameDisplayPreference::Remapped }
364 // This may include transient local filesystem information.
365 // Must not be embedded in build outputs.
366 pub fn prefer_local(&self) -> FileNameDisplay<'_> {
367 FileNameDisplay { inner: self, display_pref: FileNameDisplayPreference::Local }
370 pub fn display(&self, display_pref: FileNameDisplayPreference) -> FileNameDisplay<'_> {
371 FileNameDisplay { inner: self, display_pref }
374 pub fn macro_expansion_source_code(src: &str) -> FileName {
375 let mut hasher = StableHasher::new();
376 src.hash(&mut hasher);
377 FileName::MacroExpansion(hasher.finish())
380 pub fn anon_source_code(src: &str) -> FileName {
381 let mut hasher = StableHasher::new();
382 src.hash(&mut hasher);
383 FileName::Anon(hasher.finish())
386 pub fn proc_macro_source_code(src: &str) -> FileName {
387 let mut hasher = StableHasher::new();
388 src.hash(&mut hasher);
389 FileName::ProcMacroSourceCode(hasher.finish())
392 pub fn cfg_spec_source_code(src: &str) -> FileName {
393 let mut hasher = StableHasher::new();
394 src.hash(&mut hasher);
395 FileName::QuoteExpansion(hasher.finish())
398 pub fn cli_crate_attr_source_code(src: &str) -> FileName {
399 let mut hasher = StableHasher::new();
400 src.hash(&mut hasher);
401 FileName::CliCrateAttr(hasher.finish())
404 pub fn doc_test_source_code(path: PathBuf, line: isize) -> FileName {
405 FileName::DocTest(path, line)
408 pub fn inline_asm_source_code(src: &str) -> FileName {
409 let mut hasher = StableHasher::new();
410 src.hash(&mut hasher);
411 FileName::InlineAsm(hasher.finish())
415 /// Represents a span.
417 /// Spans represent a region of code, used for error reporting. Positions in spans
418 /// are *absolute* positions from the beginning of the [`SourceMap`], not positions
419 /// relative to [`SourceFile`]s. Methods on the `SourceMap` can be used to relate spans back
420 /// to the original source.
422 /// You must be careful if the span crosses more than one file, since you will not be
423 /// able to use many of the functions on spans in source_map and you cannot assume
424 /// that the length of the span is equal to `span.hi - span.lo`; there may be space in the
425 /// [`BytePos`] range between files.
427 /// `SpanData` is public because `Span` uses a thread-local interner and can't be
428 /// sent to other threads, but some pieces of performance infra run in a separate thread.
429 /// Using `Span` is generally preferred.
430 #[derive(Clone, Copy, Hash, PartialEq, Eq)]
431 pub struct SpanData {
434 /// Information about where the macro came from, if this piece of
435 /// code was created by a macro expansion.
436 pub ctxt: SyntaxContext,
437 pub parent: Option<LocalDefId>,
440 // Order spans by position in the file.
441 impl Ord for SpanData {
442 fn cmp(&self, other: &Self) -> Ordering {
447 // `LocalDefId` does not implement `Ord`.
448 // The other fields are enough to determine in-file order.
455 // `LocalDefId` does not implement `Ord`.
456 // The other fields are enough to determine in-file order.
460 (s_lo, s_hi, s_ctxt).cmp(&(o_lo, o_hi, o_ctxt))
464 impl PartialOrd for SpanData {
465 fn partial_cmp(&self, other: &Self) -> Option<Ordering> {
466 Some(self.cmp(other))
472 pub fn span(&self) -> Span {
473 Span::new(self.lo, self.hi, self.ctxt, self.parent)
476 pub fn with_lo(&self, lo: BytePos) -> Span {
477 Span::new(lo, self.hi, self.ctxt, self.parent)
480 pub fn with_hi(&self, hi: BytePos) -> Span {
481 Span::new(self.lo, hi, self.ctxt, self.parent)
484 pub fn with_ctxt(&self, ctxt: SyntaxContext) -> Span {
485 Span::new(self.lo, self.hi, ctxt, self.parent)
488 pub fn with_parent(&self, parent: Option<LocalDefId>) -> Span {
489 Span::new(self.lo, self.hi, self.ctxt, parent)
491 /// Returns `true` if this is a dummy span with any hygienic context.
493 pub fn is_dummy(self) -> bool {
494 self.lo.0 == 0 && self.hi.0 == 0
496 /// Returns `true` if `self` fully encloses `other`.
497 pub fn contains(self, other: Self) -> bool {
498 self.lo <= other.lo && other.hi <= self.hi
502 // The interner is pointed to by a thread local value which is only set on the main thread
503 // with parallelization is disabled. So we don't allow `Span` to transfer between threads
504 // to avoid panics and other errors, even though it would be memory safe to do so.
505 #[cfg(not(parallel_compiler))]
506 impl !Send for Span {}
507 #[cfg(not(parallel_compiler))]
508 impl !Sync for Span {}
510 impl PartialOrd for Span {
511 fn partial_cmp(&self, rhs: &Self) -> Option<Ordering> {
512 PartialOrd::partial_cmp(&self.data(), &rhs.data())
516 fn cmp(&self, rhs: &Self) -> Ordering {
517 Ord::cmp(&self.data(), &rhs.data())
523 pub fn lo(self) -> BytePos {
527 pub fn with_lo(self, lo: BytePos) -> Span {
528 self.data().with_lo(lo)
531 pub fn hi(self) -> BytePos {
535 pub fn with_hi(self, hi: BytePos) -> Span {
536 self.data().with_hi(hi)
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 any kind of macro, desugaring or inlining.
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 this span comes from MIR inlining.
575 pub fn is_inlined(self) -> bool {
576 let outer_expn = self.ctxt().outer_expn_data();
577 matches!(outer_expn.kind, ExpnKind::Inlined)
580 /// Returns `true` if `span` originates in a derive-macro's expansion.
581 pub fn in_derive_expansion(self) -> bool {
582 matches!(self.ctxt().outer_expn_data().kind, ExpnKind::Macro(MacroKind::Derive, _))
585 /// Gate suggestions that would not be appropriate in a context the user didn't write.
586 pub fn can_be_used_for_suggestions(self) -> bool {
587 !self.from_expansion()
588 // FIXME: If this span comes from a `derive` macro but it points at code the user wrote,
589 // the callsite span and the span will be pointing at different places. It also means that
590 // we can safely provide suggestions on this span.
591 || (matches!(self.ctxt().outer_expn_data().kind, ExpnKind::Macro(MacroKind::Derive, _))
592 && self.parent_callsite().map(|p| (p.lo(), p.hi())) != Some((self.lo(), self.hi())))
596 pub fn with_root_ctxt(lo: BytePos, hi: BytePos) -> Span {
597 Span::new(lo, hi, SyntaxContext::root(), None)
600 /// Returns a new span representing an empty span at the beginning of this span.
602 pub fn shrink_to_lo(self) -> Span {
603 let span = self.data_untracked();
604 span.with_hi(span.lo)
606 /// Returns a new span representing an empty span at the end of this span.
608 pub fn shrink_to_hi(self) -> Span {
609 let span = self.data_untracked();
610 span.with_lo(span.hi)
614 /// Returns `true` if `hi == lo`.
615 pub fn is_empty(self) -> bool {
616 let span = self.data_untracked();
620 /// Returns `self` if `self` is not the dummy span, and `other` otherwise.
621 pub fn substitute_dummy(self, other: Span) -> Span {
622 if self.is_dummy() { other } else { self }
625 /// Returns `true` if `self` fully encloses `other`.
626 pub fn contains(self, other: Span) -> bool {
627 let span = self.data();
628 let other = other.data();
632 /// Returns `true` if `self` touches `other`.
633 pub fn overlaps(self, other: Span) -> bool {
634 let span = self.data();
635 let other = other.data();
636 span.lo < other.hi && other.lo < span.hi
639 /// Returns `true` if the spans are equal with regards to the source text.
641 /// Use this instead of `==` when either span could be generated code,
642 /// and you only care that they point to the same bytes of source text.
643 pub fn source_equal(self, other: Span) -> bool {
644 let span = self.data();
645 let other = other.data();
646 span.lo == other.lo && span.hi == other.hi
649 /// Returns `Some(span)`, where the start is trimmed by the end of `other`.
650 pub fn trim_start(self, other: Span) -> Option<Span> {
651 let span = self.data();
652 let other = other.data();
653 if span.hi > other.hi { Some(span.with_lo(cmp::max(span.lo, other.hi))) } else { None }
656 /// Returns the source span -- this is either the supplied span, or the span for
657 /// the macro callsite that expanded to it.
658 pub fn source_callsite(self) -> Span {
659 let expn_data = self.ctxt().outer_expn_data();
660 if !expn_data.is_root() { expn_data.call_site.source_callsite() } else { self }
663 /// The `Span` for the tokens in the previous macro expansion from which `self` was generated,
665 pub fn parent_callsite(self) -> Option<Span> {
666 let expn_data = self.ctxt().outer_expn_data();
667 if !expn_data.is_root() { Some(expn_data.call_site) } else { None }
670 /// Walk down the expansion ancestors to find a span that's contained within `outer`.
671 pub fn find_ancestor_inside(mut self, outer: Span) -> Option<Span> {
672 while !outer.contains(self) {
673 self = self.parent_callsite()?;
678 /// Like `find_ancestor_inside`, but specifically for when spans might not
679 /// overlaps. Take care when using this, and prefer `find_ancestor_inside`
680 /// when you know that the spans are nested (modulo macro expansion).
681 pub fn find_ancestor_in_same_ctxt(mut self, other: Span) -> Option<Span> {
682 while !Span::eq_ctxt(self, other) {
683 self = self.parent_callsite()?;
688 /// Edition of the crate from which this span came.
689 pub fn edition(self) -> edition::Edition {
690 self.ctxt().edition()
694 pub fn rust_2015(self) -> bool {
695 self.edition() == edition::Edition::Edition2015
699 pub fn rust_2018(self) -> bool {
700 self.edition() >= edition::Edition::Edition2018
704 pub fn rust_2021(self) -> bool {
705 self.edition() >= edition::Edition::Edition2021
709 pub fn rust_2024(self) -> bool {
710 self.edition() >= edition::Edition::Edition2024
713 /// Returns the source callee.
715 /// Returns `None` if the supplied span has no expansion trace,
716 /// else returns the `ExpnData` for the macro definition
717 /// corresponding to the source callsite.
718 pub fn source_callee(self) -> Option<ExpnData> {
719 fn source_callee(expn_data: ExpnData) -> ExpnData {
720 let next_expn_data = expn_data.call_site.ctxt().outer_expn_data();
721 if !next_expn_data.is_root() { source_callee(next_expn_data) } else { expn_data }
723 let expn_data = self.ctxt().outer_expn_data();
724 if !expn_data.is_root() { Some(source_callee(expn_data)) } else { None }
727 /// Checks if a span is "internal" to a macro in which `#[unstable]`
728 /// items can be used (that is, a macro marked with
729 /// `#[allow_internal_unstable]`).
730 pub fn allows_unstable(self, feature: Symbol) -> bool {
733 .allow_internal_unstable
734 .map_or(false, |features| features.iter().any(|&f| f == feature))
737 /// Checks if this span arises from a compiler desugaring of kind `kind`.
738 pub fn is_desugaring(self, kind: DesugaringKind) -> bool {
739 match self.ctxt().outer_expn_data().kind {
740 ExpnKind::Desugaring(k) => k == kind,
745 /// Returns the compiler desugaring that created this span, or `None`
746 /// if this span is not from a desugaring.
747 pub fn desugaring_kind(self) -> Option<DesugaringKind> {
748 match self.ctxt().outer_expn_data().kind {
749 ExpnKind::Desugaring(k) => Some(k),
754 /// Checks if a span is "internal" to a macro in which `unsafe`
755 /// can be used without triggering the `unsafe_code` lint.
756 // (that is, a macro marked with `#[allow_internal_unsafe]`).
757 pub fn allows_unsafe(self) -> bool {
758 self.ctxt().outer_expn_data().allow_internal_unsafe
761 pub fn macro_backtrace(mut self) -> impl Iterator<Item = ExpnData> {
762 let mut prev_span = DUMMY_SP;
763 std::iter::from_fn(move || {
765 let expn_data = self.ctxt().outer_expn_data();
766 if expn_data.is_root() {
770 let is_recursive = expn_data.call_site.source_equal(prev_span);
773 self = expn_data.call_site;
775 // Don't print recursive invocations.
777 return Some(expn_data);
783 /// Returns a `Span` that would enclose both `self` and `end`.
787 /// self lorem ipsum end
788 /// ^^^^^^^^^^^^^^^^^^^^
790 pub fn to(self, end: Span) -> Span {
791 let span_data = self.data();
792 let end_data = end.data();
793 // FIXME(jseyfried): `self.ctxt` should always equal `end.ctxt` here (cf. issue #23480).
794 // Return the macro span on its own to avoid weird diagnostic output. It is preferable to
795 // have an incomplete span than a completely nonsensical one.
796 if span_data.ctxt != end_data.ctxt {
797 if span_data.ctxt == SyntaxContext::root() {
799 } else if end_data.ctxt == SyntaxContext::root() {
802 // Both spans fall within a macro.
803 // FIXME(estebank): check if it is the *same* macro.
806 cmp::min(span_data.lo, end_data.lo),
807 cmp::max(span_data.hi, end_data.hi),
808 if span_data.ctxt == SyntaxContext::root() { end_data.ctxt } else { span_data.ctxt },
809 if span_data.parent == end_data.parent { span_data.parent } else { None },
813 /// Returns a `Span` between the end of `self` to the beginning of `end`.
817 /// self lorem ipsum end
820 pub fn between(self, end: Span) -> Span {
821 let span = self.data();
822 let end = end.data();
826 if end.ctxt == SyntaxContext::root() { end.ctxt } else { span.ctxt },
827 if span.parent == end.parent { span.parent } else { None },
831 /// Returns a `Span` from the beginning of `self` until the beginning of `end`.
835 /// self lorem ipsum end
836 /// ^^^^^^^^^^^^^^^^^
838 pub fn until(self, end: Span) -> Span {
839 // Most of this function's body is copied from `to`.
840 // We can't just do `self.to(end.shrink_to_lo())`,
841 // because to also does some magic where it uses min/max so
842 // it can handle overlapping spans. Some advanced mis-use of
843 // `until` with different ctxts makes this visible.
844 let span_data = self.data();
845 let end_data = end.data();
846 // FIXME(jseyfried): `self.ctxt` should always equal `end.ctxt` here (cf. issue #23480).
847 // Return the macro span on its own to avoid weird diagnostic output. It is preferable to
848 // have an incomplete span than a completely nonsensical one.
849 if span_data.ctxt != end_data.ctxt {
850 if span_data.ctxt == SyntaxContext::root() {
852 } else if end_data.ctxt == SyntaxContext::root() {
855 // Both spans fall within a macro.
856 // FIXME(estebank): check if it is the *same* macro.
861 if end_data.ctxt == SyntaxContext::root() { end_data.ctxt } else { span_data.ctxt },
862 if span_data.parent == end_data.parent { span_data.parent } else { None },
866 pub fn from_inner(self, inner: InnerSpan) -> Span {
867 let span = self.data();
869 span.lo + BytePos::from_usize(inner.start),
870 span.lo + BytePos::from_usize(inner.end),
876 /// Equivalent of `Span::def_site` from the proc macro API,
877 /// except that the location is taken from the `self` span.
878 pub fn with_def_site_ctxt(self, expn_id: ExpnId) -> Span {
879 self.with_ctxt_from_mark(expn_id, Transparency::Opaque)
882 /// Equivalent of `Span::call_site` from the proc macro API,
883 /// except that the location is taken from the `self` span.
884 pub fn with_call_site_ctxt(self, expn_id: ExpnId) -> Span {
885 self.with_ctxt_from_mark(expn_id, Transparency::Transparent)
888 /// Equivalent of `Span::mixed_site` from the proc macro API,
889 /// except that the location is taken from the `self` span.
890 pub fn with_mixed_site_ctxt(self, expn_id: ExpnId) -> Span {
891 self.with_ctxt_from_mark(expn_id, Transparency::SemiTransparent)
894 /// Produces a span with the same location as `self` and context produced by a macro with the
895 /// given ID and transparency, assuming that macro was defined directly and not produced by
896 /// some other macro (which is the case for built-in and procedural macros).
897 pub fn with_ctxt_from_mark(self, expn_id: ExpnId, transparency: Transparency) -> Span {
898 self.with_ctxt(SyntaxContext::root().apply_mark(expn_id, transparency))
902 pub fn apply_mark(self, expn_id: ExpnId, transparency: Transparency) -> Span {
903 let span = self.data();
904 span.with_ctxt(span.ctxt.apply_mark(expn_id, transparency))
908 pub fn remove_mark(&mut self) -> ExpnId {
909 let mut span = self.data();
910 let mark = span.ctxt.remove_mark();
911 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
916 pub fn adjust(&mut self, expn_id: ExpnId) -> Option<ExpnId> {
917 let mut span = self.data();
918 let mark = span.ctxt.adjust(expn_id);
919 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
924 pub fn normalize_to_macros_2_0_and_adjust(&mut self, expn_id: ExpnId) -> Option<ExpnId> {
925 let mut span = self.data();
926 let mark = span.ctxt.normalize_to_macros_2_0_and_adjust(expn_id);
927 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
932 pub fn glob_adjust(&mut self, expn_id: ExpnId, glob_span: Span) -> Option<Option<ExpnId>> {
933 let mut span = self.data();
934 let mark = span.ctxt.glob_adjust(expn_id, glob_span);
935 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
940 pub fn reverse_glob_adjust(
944 ) -> Option<Option<ExpnId>> {
945 let mut span = self.data();
946 let mark = span.ctxt.reverse_glob_adjust(expn_id, glob_span);
947 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
952 pub fn normalize_to_macros_2_0(self) -> Span {
953 let span = self.data();
954 span.with_ctxt(span.ctxt.normalize_to_macros_2_0())
958 pub fn normalize_to_macro_rules(self) -> Span {
959 let span = self.data();
960 span.with_ctxt(span.ctxt.normalize_to_macro_rules())
964 impl Default for Span {
965 fn default() -> Self {
970 impl<E: Encoder> Encodable<E> for Span {
971 default fn encode(&self, s: &mut E) {
972 let span = self.data();
977 impl<D: Decoder> Decodable<D> for Span {
978 default fn decode(s: &mut D) -> Span {
979 let lo = Decodable::decode(s);
980 let hi = Decodable::decode(s);
982 Span::new(lo, hi, SyntaxContext::root(), None)
986 /// Calls the provided closure, using the provided `SourceMap` to format
987 /// any spans that are debug-printed during the closure's execution.
989 /// Normally, the global `TyCtxt` is used to retrieve the `SourceMap`
990 /// (see `rustc_interface::callbacks::span_debug1`). However, some parts
991 /// of the compiler (e.g. `rustc_parse`) may debug-print `Span`s before
992 /// a `TyCtxt` is available. In this case, we fall back to
993 /// the `SourceMap` provided to this function. If that is not available,
994 /// we fall back to printing the raw `Span` field values.
995 pub fn with_source_map<T, F: FnOnce() -> T>(source_map: Lrc<SourceMap>, f: F) -> T {
996 with_session_globals(|session_globals| {
997 *session_globals.source_map.borrow_mut() = Some(source_map);
999 struct ClearSourceMap;
1000 impl Drop for ClearSourceMap {
1001 fn drop(&mut self) {
1002 with_session_globals(|session_globals| {
1003 session_globals.source_map.borrow_mut().take();
1008 let _guard = ClearSourceMap;
1012 impl fmt::Debug for Span {
1013 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
1014 with_session_globals(|session_globals| {
1015 if let Some(source_map) = &*session_globals.source_map.borrow() {
1016 write!(f, "{} ({:?})", source_map.span_to_diagnostic_string(*self), self.ctxt())
1018 f.debug_struct("Span")
1019 .field("lo", &self.lo())
1020 .field("hi", &self.hi())
1021 .field("ctxt", &self.ctxt())
1028 impl fmt::Debug for SpanData {
1029 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
1030 fmt::Debug::fmt(&Span::new(self.lo, self.hi, self.ctxt, self.parent), f)
1034 /// Identifies an offset of a multi-byte character in a `SourceFile`.
1035 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1036 pub struct MultiByteChar {
1037 /// The absolute offset of the character in the `SourceMap`.
1039 /// The number of bytes, `>= 2`.
1043 /// Identifies an offset of a non-narrow character in a `SourceFile`.
1044 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1045 pub enum NonNarrowChar {
1046 /// Represents a zero-width character.
1048 /// Represents a wide (full-width) character.
1050 /// Represents a tab character, represented visually with a width of 4 characters.
1054 impl NonNarrowChar {
1055 fn new(pos: BytePos, width: usize) -> Self {
1057 0 => NonNarrowChar::ZeroWidth(pos),
1058 2 => NonNarrowChar::Wide(pos),
1059 4 => NonNarrowChar::Tab(pos),
1060 _ => panic!("width {} given for non-narrow character", width),
1064 /// Returns the absolute offset of the character in the `SourceMap`.
1065 pub fn pos(&self) -> BytePos {
1067 NonNarrowChar::ZeroWidth(p) | NonNarrowChar::Wide(p) | NonNarrowChar::Tab(p) => p,
1071 /// Returns the width of the character, 0 (zero-width) or 2 (wide).
1072 pub fn width(&self) -> usize {
1074 NonNarrowChar::ZeroWidth(_) => 0,
1075 NonNarrowChar::Wide(_) => 2,
1076 NonNarrowChar::Tab(_) => 4,
1081 impl Add<BytePos> for NonNarrowChar {
1084 fn add(self, rhs: BytePos) -> Self {
1086 NonNarrowChar::ZeroWidth(pos) => NonNarrowChar::ZeroWidth(pos + rhs),
1087 NonNarrowChar::Wide(pos) => NonNarrowChar::Wide(pos + rhs),
1088 NonNarrowChar::Tab(pos) => NonNarrowChar::Tab(pos + rhs),
1093 impl Sub<BytePos> for NonNarrowChar {
1096 fn sub(self, rhs: BytePos) -> Self {
1098 NonNarrowChar::ZeroWidth(pos) => NonNarrowChar::ZeroWidth(pos - rhs),
1099 NonNarrowChar::Wide(pos) => NonNarrowChar::Wide(pos - rhs),
1100 NonNarrowChar::Tab(pos) => NonNarrowChar::Tab(pos - rhs),
1105 /// Identifies an offset of a character that was normalized away from `SourceFile`.
1106 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1107 pub struct NormalizedPos {
1108 /// The absolute offset of the character in the `SourceMap`.
1110 /// The difference between original and normalized string at position.
1114 #[derive(PartialEq, Eq, Clone, Debug)]
1115 pub enum ExternalSource {
1116 /// No external source has to be loaded, since the `SourceFile` represents a local crate.
1119 kind: ExternalSourceKind,
1120 /// Index of the file inside metadata.
1121 metadata_index: u32,
1125 /// The state of the lazy external source loading mechanism of a `SourceFile`.
1126 #[derive(PartialEq, Eq, Clone, Debug)]
1127 pub enum ExternalSourceKind {
1128 /// The external source has been loaded already.
1129 Present(Lrc<String>),
1130 /// No attempt has been made to load the external source.
1132 /// A failed attempt has been made to load the external source.
1137 impl ExternalSource {
1138 pub fn get_source(&self) -> Option<&Lrc<String>> {
1140 ExternalSource::Foreign { kind: ExternalSourceKind::Present(ref src), .. } => Some(src),
1147 pub struct OffsetOverflowError;
1149 #[derive(Copy, Clone, Debug, PartialEq, Eq, PartialOrd, Ord, Hash, Encodable, Decodable)]
1150 #[derive(HashStable_Generic)]
1151 pub enum SourceFileHashAlgorithm {
1157 impl FromStr for SourceFileHashAlgorithm {
1160 fn from_str(s: &str) -> Result<SourceFileHashAlgorithm, ()> {
1162 "md5" => Ok(SourceFileHashAlgorithm::Md5),
1163 "sha1" => Ok(SourceFileHashAlgorithm::Sha1),
1164 "sha256" => Ok(SourceFileHashAlgorithm::Sha256),
1170 /// The hash of the on-disk source file used for debug info.
1171 #[derive(Copy, Clone, PartialEq, Eq, Debug, Hash)]
1172 #[derive(HashStable_Generic, Encodable, Decodable)]
1173 pub struct SourceFileHash {
1174 pub kind: SourceFileHashAlgorithm,
1178 impl SourceFileHash {
1179 pub fn new(kind: SourceFileHashAlgorithm, src: &str) -> SourceFileHash {
1180 let mut hash = SourceFileHash { kind, value: Default::default() };
1181 let len = hash.hash_len();
1182 let value = &mut hash.value[..len];
1183 let data = src.as_bytes();
1185 SourceFileHashAlgorithm::Md5 => {
1186 value.copy_from_slice(&Md5::digest(data));
1188 SourceFileHashAlgorithm::Sha1 => {
1189 value.copy_from_slice(&Sha1::digest(data));
1191 SourceFileHashAlgorithm::Sha256 => {
1192 value.copy_from_slice(&Sha256::digest(data));
1198 /// Check if the stored hash matches the hash of the string.
1199 pub fn matches(&self, src: &str) -> bool {
1200 Self::new(self.kind, src) == *self
1203 /// The bytes of the hash.
1204 pub fn hash_bytes(&self) -> &[u8] {
1205 let len = self.hash_len();
1209 fn hash_len(&self) -> usize {
1211 SourceFileHashAlgorithm::Md5 => 16,
1212 SourceFileHashAlgorithm::Sha1 => 20,
1213 SourceFileHashAlgorithm::Sha256 => 32,
1218 #[derive(HashStable_Generic)]
1219 #[derive(Copy, PartialEq, PartialOrd, Clone, Ord, Eq, Hash, Debug, Encodable, Decodable)]
1220 pub enum DebuggerVisualizerType {
1225 /// A single debugger visualizer file.
1226 #[derive(HashStable_Generic)]
1227 #[derive(Clone, Debug, Hash, PartialEq, Eq, PartialOrd, Ord, Encodable, Decodable)]
1228 pub struct DebuggerVisualizerFile {
1229 /// The complete debugger visualizer source.
1231 /// Indicates which visualizer type this targets.
1232 pub visualizer_type: DebuggerVisualizerType,
1235 impl DebuggerVisualizerFile {
1236 pub fn new(src: Arc<[u8]>, visualizer_type: DebuggerVisualizerType) -> Self {
1237 DebuggerVisualizerFile { src, visualizer_type }
1242 pub enum SourceFileLines {
1243 /// The source file lines, in decoded (random-access) form.
1244 Lines(Vec<BytePos>),
1246 /// The source file lines, in undecoded difference list form.
1247 Diffs(SourceFileDiffs),
1250 impl SourceFileLines {
1251 pub fn is_lines(&self) -> bool {
1252 matches!(self, SourceFileLines::Lines(_))
1256 /// The source file lines in difference list form. This matches the form
1257 /// used within metadata, which saves space by exploiting the fact that the
1258 /// lines list is sorted and individual lines are usually not that long.
1260 /// We read it directly from metadata and only decode it into `Lines` form
1261 /// when necessary. This is a significant performance win, especially for
1262 /// small crates where very little of `std`'s metadata is used.
1264 pub struct SourceFileDiffs {
1265 /// Position of the first line. Note that this is always encoded as a
1266 /// `BytePos` because it is often much larger than any of the
1268 line_start: BytePos,
1270 /// Always 1, 2, or 4. Always as small as possible, while being big
1271 /// enough to hold the length of the longest line in the source file.
1272 /// The 1 case is by far the most common.
1273 bytes_per_diff: usize,
1275 /// The number of diffs encoded in `raw_diffs`. Always one less than
1276 /// the number of lines in the source file.
1279 /// The diffs in "raw" form. Each segment of `bytes_per_diff` length
1280 /// encodes one little-endian diff. Note that they aren't LEB128
1281 /// encoded. This makes for much faster decoding. Besides, the
1282 /// bytes_per_diff==1 case is by far the most common, and LEB128
1283 /// encoding has no effect on that case.
1287 /// A single source in the [`SourceMap`].
1289 pub struct SourceFile {
1290 /// The name of the file that the source came from. Source that doesn't
1291 /// originate from files has names between angle brackets by convention
1292 /// (e.g., `<anon>`).
1294 /// The complete source code.
1295 pub src: Option<Lrc<String>>,
1296 /// The source code's hash.
1297 pub src_hash: SourceFileHash,
1298 /// The external source code (used for external crates, which will have a `None`
1299 /// value as `self.src`.
1300 pub external_src: Lock<ExternalSource>,
1301 /// The start position of this source in the `SourceMap`.
1302 pub start_pos: BytePos,
1303 /// The end position of this source in the `SourceMap`.
1304 pub end_pos: BytePos,
1305 /// Locations of lines beginnings in the source code.
1306 pub lines: Lock<SourceFileLines>,
1307 /// Locations of multi-byte characters in the source code.
1308 pub multibyte_chars: Vec<MultiByteChar>,
1309 /// Width of characters that are not narrow in the source code.
1310 pub non_narrow_chars: Vec<NonNarrowChar>,
1311 /// Locations of characters removed during normalization.
1312 pub normalized_pos: Vec<NormalizedPos>,
1313 /// A hash of the filename, used for speeding up hashing in incremental compilation.
1314 pub name_hash: u128,
1315 /// Indicates which crate this `SourceFile` was imported from.
1319 impl<S: Encoder> Encodable<S> for SourceFile {
1320 fn encode(&self, s: &mut S) {
1321 self.name.encode(s);
1322 self.src_hash.encode(s);
1323 self.start_pos.encode(s);
1324 self.end_pos.encode(s);
1326 // We are always in `Lines` form by the time we reach here.
1327 assert!(self.lines.borrow().is_lines());
1328 self.lines(|lines| {
1329 // Store the length.
1330 s.emit_u32(lines.len() as u32);
1332 // Compute and store the difference list.
1333 if lines.len() != 0 {
1334 let max_line_length = if lines.len() == 1 {
1339 .map(|&[fst, snd]| snd - fst)
1340 .map(|bp| bp.to_usize())
1345 let bytes_per_diff: usize = match max_line_length {
1347 0x100..=0xFFFF => 2,
1351 // Encode the number of bytes used per diff.
1352 s.emit_u8(bytes_per_diff as u8);
1354 // Encode the first element.
1357 // Encode the difference list.
1358 let diff_iter = lines.array_windows().map(|&[fst, snd]| snd - fst);
1359 let num_diffs = lines.len() - 1;
1361 match bytes_per_diff {
1363 raw_diffs = Vec::with_capacity(num_diffs);
1364 for diff in diff_iter {
1365 raw_diffs.push(diff.0 as u8);
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 u16).to_le_bytes());
1375 raw_diffs = Vec::with_capacity(bytes_per_diff * num_diffs);
1376 for diff in diff_iter {
1377 raw_diffs.extend_from_slice(&(diff.0 as u32).to_le_bytes());
1380 _ => unreachable!(),
1382 s.emit_raw_bytes(&raw_diffs);
1386 self.multibyte_chars.encode(s);
1387 self.non_narrow_chars.encode(s);
1388 self.name_hash.encode(s);
1389 self.normalized_pos.encode(s);
1390 self.cnum.encode(s);
1394 impl<D: Decoder> Decodable<D> for SourceFile {
1395 fn decode(d: &mut D) -> SourceFile {
1396 let name: FileName = Decodable::decode(d);
1397 let src_hash: SourceFileHash = Decodable::decode(d);
1398 let start_pos: BytePos = Decodable::decode(d);
1399 let end_pos: BytePos = Decodable::decode(d);
1401 let num_lines: u32 = Decodable::decode(d);
1403 // Read the number of bytes used per diff.
1404 let bytes_per_diff = d.read_u8() as usize;
1406 // Read the first element.
1407 let line_start: BytePos = Decodable::decode(d);
1409 // Read the difference list.
1410 let num_diffs = num_lines as usize - 1;
1411 let raw_diffs = d.read_raw_bytes(bytes_per_diff * num_diffs).to_vec();
1412 SourceFileLines::Diffs(SourceFileDiffs {
1419 SourceFileLines::Lines(vec![])
1422 let multibyte_chars: Vec<MultiByteChar> = Decodable::decode(d);
1423 let non_narrow_chars: Vec<NonNarrowChar> = Decodable::decode(d);
1424 let name_hash: u128 = Decodable::decode(d);
1425 let normalized_pos: Vec<NormalizedPos> = Decodable::decode(d);
1426 let cnum: CrateNum = Decodable::decode(d);
1433 // Unused - the metadata decoder will construct
1434 // a new SourceFile, filling in `external_src` properly
1435 external_src: Lock::new(ExternalSource::Unneeded),
1436 lines: Lock::new(lines),
1446 impl fmt::Debug for SourceFile {
1447 fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result {
1448 write!(fmt, "SourceFile({:?})", self.name)
1457 hash_kind: SourceFileHashAlgorithm,
1459 // Compute the file hash before any normalization.
1460 let src_hash = SourceFileHash::new(hash_kind, &src);
1461 let normalized_pos = normalize_src(&mut src, start_pos);
1464 let mut hasher: StableHasher = StableHasher::new();
1465 name.hash(&mut hasher);
1466 hasher.finish::<u128>()
1468 let end_pos = start_pos.to_usize() + src.len();
1469 assert!(end_pos <= u32::MAX as usize);
1471 let (lines, multibyte_chars, non_narrow_chars) =
1472 analyze_source_file::analyze_source_file(&src, start_pos);
1476 src: Some(Lrc::new(src)),
1478 external_src: Lock::new(ExternalSource::Unneeded),
1480 end_pos: Pos::from_usize(end_pos),
1481 lines: Lock::new(SourceFileLines::Lines(lines)),
1490 pub fn lines<F, R>(&self, f: F) -> R
1492 F: FnOnce(&[BytePos]) -> R,
1494 let mut guard = self.lines.borrow_mut();
1496 SourceFileLines::Lines(lines) => f(lines),
1497 SourceFileLines::Diffs(SourceFileDiffs {
1503 // Convert from "diffs" form to "lines" form.
1504 let num_lines = num_diffs + 1;
1505 let mut lines = Vec::with_capacity(num_lines);
1506 lines.push(line_start);
1508 assert_eq!(*num_diffs, raw_diffs.len() / bytes_per_diff);
1509 match bytes_per_diff {
1511 lines.extend(raw_diffs.into_iter().map(|&diff| {
1512 line_start = line_start + BytePos(diff as u32);
1517 lines.extend((0..*num_diffs).map(|i| {
1518 let pos = bytes_per_diff * i;
1519 let bytes = [raw_diffs[pos], raw_diffs[pos + 1]];
1520 let diff = u16::from_le_bytes(bytes);
1521 line_start = line_start + BytePos(diff as u32);
1526 lines.extend((0..*num_diffs).map(|i| {
1527 let pos = bytes_per_diff * i;
1534 let diff = u32::from_le_bytes(bytes);
1535 line_start = line_start + BytePos(diff);
1539 _ => unreachable!(),
1541 let res = f(&lines);
1542 *guard = SourceFileLines::Lines(lines);
1548 /// Returns the `BytePos` of the beginning of the current line.
1549 pub fn line_begin_pos(&self, pos: BytePos) -> BytePos {
1550 let line_index = self.lookup_line(pos).unwrap();
1551 self.lines(|lines| lines[line_index])
1554 /// Add externally loaded source.
1555 /// If the hash of the input doesn't match or no input is supplied via None,
1556 /// it is interpreted as an error and the corresponding enum variant is set.
1557 /// The return value signifies whether some kind of source is present.
1558 pub fn add_external_src<F>(&self, get_src: F) -> bool
1560 F: FnOnce() -> Option<String>,
1563 *self.external_src.borrow(),
1564 ExternalSource::Foreign { kind: ExternalSourceKind::AbsentOk, .. }
1566 let src = get_src();
1567 let mut external_src = self.external_src.borrow_mut();
1568 // Check that no-one else have provided the source while we were getting it
1569 if let ExternalSource::Foreign {
1570 kind: src_kind @ ExternalSourceKind::AbsentOk, ..
1571 } = &mut *external_src
1573 if let Some(mut src) = src {
1574 // The src_hash needs to be computed on the pre-normalized src.
1575 if self.src_hash.matches(&src) {
1576 normalize_src(&mut src, BytePos::from_usize(0));
1577 *src_kind = ExternalSourceKind::Present(Lrc::new(src));
1581 *src_kind = ExternalSourceKind::AbsentErr;
1586 self.src.is_some() || external_src.get_source().is_some()
1589 self.src.is_some() || self.external_src.borrow().get_source().is_some()
1593 /// Gets a line from the list of pre-computed line-beginnings.
1594 /// The line number here is 0-based.
1595 pub fn get_line(&self, line_number: usize) -> Option<Cow<'_, str>> {
1596 fn get_until_newline(src: &str, begin: usize) -> &str {
1597 // We can't use `lines.get(line_number+1)` because we might
1598 // be parsing when we call this function and thus the current
1599 // line is the last one we have line info for.
1600 let slice = &src[begin..];
1601 match slice.find('\n') {
1602 Some(e) => &slice[..e],
1608 let line = self.lines(|lines| lines.get(line_number).copied())?;
1609 let begin: BytePos = line - self.start_pos;
1613 if let Some(ref src) = self.src {
1614 Some(Cow::from(get_until_newline(src, begin)))
1615 } else if let Some(src) = self.external_src.borrow().get_source() {
1616 Some(Cow::Owned(String::from(get_until_newline(src, begin))))
1622 pub fn is_real_file(&self) -> bool {
1627 pub fn is_imported(&self) -> bool {
1631 pub fn count_lines(&self) -> usize {
1632 self.lines(|lines| lines.len())
1635 /// Finds the line containing the given position. The return value is the
1636 /// index into the `lines` array of this `SourceFile`, not the 1-based line
1637 /// number. If the source_file is empty or the position is located before the
1638 /// first line, `None` is returned.
1639 pub fn lookup_line(&self, pos: BytePos) -> Option<usize> {
1640 self.lines(|lines| lines.partition_point(|x| x <= &pos).checked_sub(1))
1643 pub fn line_bounds(&self, line_index: usize) -> Range<BytePos> {
1644 if self.is_empty() {
1645 return self.start_pos..self.end_pos;
1648 self.lines(|lines| {
1649 assert!(line_index < lines.len());
1650 if line_index == (lines.len() - 1) {
1651 lines[line_index]..self.end_pos
1653 lines[line_index]..lines[line_index + 1]
1658 /// Returns whether or not the file contains the given `SourceMap` byte
1659 /// position. The position one past the end of the file is considered to be
1660 /// contained by the file. This implies that files for which `is_empty`
1661 /// returns true still contain one byte position according to this function.
1663 pub fn contains(&self, byte_pos: BytePos) -> bool {
1664 byte_pos >= self.start_pos && byte_pos <= self.end_pos
1668 pub fn is_empty(&self) -> bool {
1669 self.start_pos == self.end_pos
1672 /// Calculates the original byte position relative to the start of the file
1673 /// based on the given byte position.
1674 pub fn original_relative_byte_pos(&self, pos: BytePos) -> BytePos {
1675 // Diff before any records is 0. Otherwise use the previously recorded
1676 // diff as that applies to the following characters until a new diff
1678 let diff = match self.normalized_pos.binary_search_by(|np| np.pos.cmp(&pos)) {
1679 Ok(i) => self.normalized_pos[i].diff,
1680 Err(i) if i == 0 => 0,
1681 Err(i) => self.normalized_pos[i - 1].diff,
1684 BytePos::from_u32(pos.0 - self.start_pos.0 + diff)
1687 /// Converts an absolute `BytePos` to a `CharPos` relative to the `SourceFile`.
1688 pub fn bytepos_to_file_charpos(&self, bpos: BytePos) -> CharPos {
1689 // The number of extra bytes due to multibyte chars in the `SourceFile`.
1690 let mut total_extra_bytes = 0;
1692 for mbc in self.multibyte_chars.iter() {
1693 debug!("{}-byte char at {:?}", mbc.bytes, mbc.pos);
1695 // Every character is at least one byte, so we only
1696 // count the actual extra bytes.
1697 total_extra_bytes += mbc.bytes as u32 - 1;
1698 // We should never see a byte position in the middle of a
1700 assert!(bpos.to_u32() >= mbc.pos.to_u32() + mbc.bytes as u32);
1706 assert!(self.start_pos.to_u32() + total_extra_bytes <= bpos.to_u32());
1707 CharPos(bpos.to_usize() - self.start_pos.to_usize() - total_extra_bytes as usize)
1710 /// Looks up the file's (1-based) line number and (0-based `CharPos`) column offset, for a
1711 /// given `BytePos`.
1712 pub fn lookup_file_pos(&self, pos: BytePos) -> (usize, CharPos) {
1713 let chpos = self.bytepos_to_file_charpos(pos);
1714 match self.lookup_line(pos) {
1716 let line = a + 1; // Line numbers start at 1
1717 let linebpos = self.lines(|lines| lines[a]);
1718 let linechpos = self.bytepos_to_file_charpos(linebpos);
1719 let col = chpos - linechpos;
1720 debug!("byte pos {:?} is on the line at byte pos {:?}", pos, linebpos);
1721 debug!("char pos {:?} is on the line at char pos {:?}", chpos, linechpos);
1722 debug!("byte is on line: {}", line);
1723 assert!(chpos >= linechpos);
1730 /// Looks up the file's (1-based) line number, (0-based `CharPos`) column offset, and (0-based)
1731 /// column offset when displayed, for a given `BytePos`.
1732 pub fn lookup_file_pos_with_col_display(&self, pos: BytePos) -> (usize, CharPos, usize) {
1733 let (line, col_or_chpos) = self.lookup_file_pos(pos);
1735 let col = col_or_chpos;
1736 let linebpos = self.lines(|lines| lines[line - 1]);
1738 let start_width_idx = self
1740 .binary_search_by_key(&linebpos, |x| x.pos())
1741 .unwrap_or_else(|x| x);
1742 let end_width_idx = self
1744 .binary_search_by_key(&pos, |x| x.pos())
1745 .unwrap_or_else(|x| x);
1746 let special_chars = end_width_idx - start_width_idx;
1747 let non_narrow: usize = self.non_narrow_chars[start_width_idx..end_width_idx]
1751 col.0 - special_chars + non_narrow
1753 (line, col, col_display)
1755 let chpos = col_or_chpos;
1757 let end_width_idx = self
1759 .binary_search_by_key(&pos, |x| x.pos())
1760 .unwrap_or_else(|x| x);
1761 let non_narrow: usize =
1762 self.non_narrow_chars[0..end_width_idx].iter().map(|x| x.width()).sum();
1763 chpos.0 - end_width_idx + non_narrow
1765 (0, chpos, col_display)
1770 /// Normalizes the source code and records the normalizations.
1771 fn normalize_src(src: &mut String, start_pos: BytePos) -> Vec<NormalizedPos> {
1772 let mut normalized_pos = vec![];
1773 remove_bom(src, &mut normalized_pos);
1774 normalize_newlines(src, &mut normalized_pos);
1776 // Offset all the positions by start_pos to match the final file positions.
1777 for np in &mut normalized_pos {
1778 np.pos.0 += start_pos.0;
1784 /// Removes UTF-8 BOM, if any.
1785 fn remove_bom(src: &mut String, normalized_pos: &mut Vec<NormalizedPos>) {
1786 if src.starts_with('\u{feff}') {
1788 normalized_pos.push(NormalizedPos { pos: BytePos(0), diff: 3 });
1792 /// Replaces `\r\n` with `\n` in-place in `src`.
1794 /// Returns error if there's a lone `\r` in the string.
1795 fn normalize_newlines(src: &mut String, normalized_pos: &mut Vec<NormalizedPos>) {
1796 if !src.as_bytes().contains(&b'\r') {
1800 // We replace `\r\n` with `\n` in-place, which doesn't break utf-8 encoding.
1801 // While we *can* call `as_mut_vec` and do surgery on the live string
1802 // directly, let's rather steal the contents of `src`. This makes the code
1803 // safe even if a panic occurs.
1805 let mut buf = std::mem::replace(src, String::new()).into_bytes();
1806 let mut gap_len = 0;
1807 let mut tail = buf.as_mut_slice();
1809 let original_gap = normalized_pos.last().map_or(0, |l| l.diff);
1811 let idx = match find_crlf(&tail[gap_len..]) {
1813 Some(idx) => idx + gap_len,
1815 tail.copy_within(gap_len..idx, 0);
1816 tail = &mut tail[idx - gap_len..];
1817 if tail.len() == gap_len {
1820 cursor += idx - gap_len;
1822 normalized_pos.push(NormalizedPos {
1823 pos: BytePos::from_usize(cursor + 1),
1824 diff: original_gap + gap_len as u32,
1828 // Account for removed `\r`.
1829 // After `set_len`, `buf` is guaranteed to contain utf-8 again.
1830 let new_len = buf.len() - gap_len;
1832 buf.set_len(new_len);
1833 *src = String::from_utf8_unchecked(buf);
1836 fn find_crlf(src: &[u8]) -> Option<usize> {
1837 let mut search_idx = 0;
1838 while let Some(idx) = find_cr(&src[search_idx..]) {
1839 if src[search_idx..].get(idx + 1) != Some(&b'\n') {
1840 search_idx += idx + 1;
1843 return Some(search_idx + idx);
1848 fn find_cr(src: &[u8]) -> Option<usize> {
1849 src.iter().position(|&b| b == b'\r')
1853 // _____________________________________________________________________________
1854 // Pos, BytePos, CharPos
1858 fn from_usize(n: usize) -> Self;
1859 fn to_usize(&self) -> usize;
1860 fn from_u32(n: u32) -> Self;
1861 fn to_u32(&self) -> u32;
1864 macro_rules! impl_pos {
1868 $vis:vis struct $ident:ident($inner_vis:vis $inner_ty:ty);
1873 $vis struct $ident($inner_vis $inner_ty);
1875 impl Pos for $ident {
1877 fn from_usize(n: usize) -> $ident {
1878 $ident(n as $inner_ty)
1882 fn to_usize(&self) -> usize {
1887 fn from_u32(n: u32) -> $ident {
1888 $ident(n as $inner_ty)
1892 fn to_u32(&self) -> u32 {
1897 impl Add for $ident {
1898 type Output = $ident;
1901 fn add(self, rhs: $ident) -> $ident {
1902 $ident(self.0 + rhs.0)
1906 impl Sub for $ident {
1907 type Output = $ident;
1910 fn sub(self, rhs: $ident) -> $ident {
1911 $ident(self.0 - rhs.0)
1921 /// Keep this small (currently 32-bits), as AST contains a lot of them.
1922 #[derive(Clone, Copy, PartialEq, Eq, Hash, PartialOrd, Ord, Debug)]
1923 pub struct BytePos(pub u32);
1925 /// A character offset.
1927 /// Because of multibyte UTF-8 characters, a byte offset
1928 /// is not equivalent to a character offset. The [`SourceMap`] will convert [`BytePos`]
1929 /// values to `CharPos` values as necessary.
1930 #[derive(Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Debug)]
1931 pub struct CharPos(pub usize);
1934 impl<S: Encoder> Encodable<S> for BytePos {
1935 fn encode(&self, s: &mut S) {
1940 impl<D: Decoder> Decodable<D> for BytePos {
1941 fn decode(d: &mut D) -> BytePos {
1942 BytePos(d.read_u32())
1946 // _____________________________________________________________________________
1947 // Loc, SourceFileAndLine, SourceFileAndBytePos
1950 /// A source code location used for error reporting.
1951 #[derive(Debug, Clone)]
1953 /// Information about the original source.
1954 pub file: Lrc<SourceFile>,
1955 /// The (1-based) line number.
1957 /// The (0-based) column offset.
1959 /// The (0-based) column offset when displayed.
1960 pub col_display: usize,
1963 // Used to be structural records.
1965 pub struct SourceFileAndLine {
1966 pub sf: Lrc<SourceFile>,
1967 /// Index of line, starting from 0.
1971 pub struct SourceFileAndBytePos {
1972 pub sf: Lrc<SourceFile>,
1976 #[derive(Copy, Clone, Debug, PartialEq, Eq)]
1977 pub struct LineInfo {
1978 /// Index of line, starting from 0.
1979 pub line_index: usize,
1981 /// Column in line where span begins, starting from 0.
1982 pub start_col: CharPos,
1984 /// Column in line where span ends, starting from 0, exclusive.
1985 pub end_col: CharPos,
1988 pub struct FileLines {
1989 pub file: Lrc<SourceFile>,
1990 pub lines: Vec<LineInfo>,
1993 pub static SPAN_TRACK: AtomicRef<fn(LocalDefId)> = AtomicRef::new(&((|_| {}) as fn(_)));
1995 // _____________________________________________________________________________
1996 // SpanLinesError, SpanSnippetError, DistinctSources, MalformedSourceMapPositions
1999 pub type FileLinesResult = Result<FileLines, SpanLinesError>;
2001 #[derive(Clone, PartialEq, Eq, Debug)]
2002 pub enum SpanLinesError {
2003 DistinctSources(DistinctSources),
2006 #[derive(Clone, PartialEq, Eq, Debug)]
2007 pub enum SpanSnippetError {
2008 IllFormedSpan(Span),
2009 DistinctSources(DistinctSources),
2010 MalformedForSourcemap(MalformedSourceMapPositions),
2011 SourceNotAvailable { filename: FileName },
2014 #[derive(Clone, PartialEq, Eq, Debug)]
2015 pub struct DistinctSources {
2016 pub begin: (FileName, BytePos),
2017 pub end: (FileName, BytePos),
2020 #[derive(Clone, PartialEq, Eq, Debug)]
2021 pub struct MalformedSourceMapPositions {
2023 pub source_len: usize,
2024 pub begin_pos: BytePos,
2025 pub end_pos: BytePos,
2028 /// Range inside of a `Span` used for diagnostics when we only have access to relative positions.
2029 #[derive(Copy, Clone, PartialEq, Eq, Debug)]
2030 pub struct InnerSpan {
2036 pub fn new(start: usize, end: usize) -> InnerSpan {
2037 InnerSpan { start, end }
2041 /// Requirements for a `StableHashingContext` to be used in this crate.
2043 /// This is a hack to allow using the [`HashStable_Generic`] derive macro
2044 /// instead of implementing everything in rustc_middle.
2045 pub trait HashStableContext {
2046 fn def_path_hash(&self, def_id: DefId) -> DefPathHash;
2047 fn hash_spans(&self) -> bool;
2048 /// Accesses `sess.opts.unstable_opts.incremental_ignore_spans` since
2049 /// we don't have easy access to a `Session`
2050 fn unstable_opts_incremental_ignore_spans(&self) -> bool;
2051 fn def_span(&self, def_id: LocalDefId) -> Span;
2052 fn span_data_to_lines_and_cols(
2055 ) -> Option<(Lrc<SourceFile>, usize, BytePos, usize, BytePos)>;
2056 fn hashing_controls(&self) -> HashingControls;
2059 impl<CTX> HashStable<CTX> for Span
2061 CTX: HashStableContext,
2063 /// Hashes a span in a stable way. We can't directly hash the span's `BytePos`
2064 /// fields (that would be similar to hashing pointers, since those are just
2065 /// offsets into the `SourceMap`). Instead, we hash the (file name, line, column)
2066 /// triple, which stays the same even if the containing `SourceFile` has moved
2067 /// within the `SourceMap`.
2069 /// Also note that we are hashing byte offsets for the column, not unicode
2070 /// codepoint offsets. For the purpose of the hash that's sufficient.
2071 /// Also, hashing filenames is expensive so we avoid doing it twice when the
2072 /// span starts and ends in the same file, which is almost always the case.
2073 fn hash_stable(&self, ctx: &mut CTX, hasher: &mut StableHasher) {
2074 const TAG_VALID_SPAN: u8 = 0;
2075 const TAG_INVALID_SPAN: u8 = 1;
2076 const TAG_RELATIVE_SPAN: u8 = 2;
2078 if !ctx.hash_spans() {
2082 let span = self.data_untracked();
2083 span.ctxt.hash_stable(ctx, hasher);
2084 span.parent.hash_stable(ctx, hasher);
2086 if span.is_dummy() {
2087 Hash::hash(&TAG_INVALID_SPAN, hasher);
2091 if let Some(parent) = span.parent {
2092 let def_span = ctx.def_span(parent).data_untracked();
2093 if def_span.contains(span) {
2094 // This span is enclosed in a definition: only hash the relative position.
2095 Hash::hash(&TAG_RELATIVE_SPAN, hasher);
2096 (span.lo - def_span.lo).to_u32().hash_stable(ctx, hasher);
2097 (span.hi - def_span.lo).to_u32().hash_stable(ctx, hasher);
2102 // If this is not an empty or invalid span, we want to hash the last
2103 // position that belongs to it, as opposed to hashing the first
2104 // position past it.
2105 let Some((file, line_lo, col_lo, line_hi, col_hi)) = ctx.span_data_to_lines_and_cols(&span) else {
2106 Hash::hash(&TAG_INVALID_SPAN, hasher);
2110 Hash::hash(&TAG_VALID_SPAN, hasher);
2111 // We truncate the stable ID hash and line and column numbers. The chances
2112 // of causing a collision this way should be minimal.
2113 Hash::hash(&(file.name_hash as u64), hasher);
2115 // Hash both the length and the end location (line/column) of a span. If we
2116 // hash only the length, for example, then two otherwise equal spans with
2117 // different end locations will have the same hash. This can cause a problem
2118 // during incremental compilation wherein a previous result for a query that
2119 // depends on the end location of a span will be incorrectly reused when the
2120 // end location of the span it depends on has changed (see issue #74890). A
2121 // similar analysis applies if some query depends specifically on the length
2122 // of the span, but we only hash the end location. So hash both.
2124 let col_lo_trunc = (col_lo.0 as u64) & 0xFF;
2125 let line_lo_trunc = ((line_lo as u64) & 0xFF_FF_FF) << 8;
2126 let col_hi_trunc = (col_hi.0 as u64) & 0xFF << 32;
2127 let line_hi_trunc = ((line_hi as u64) & 0xFF_FF_FF) << 40;
2128 let col_line = col_lo_trunc | line_lo_trunc | col_hi_trunc | line_hi_trunc;
2129 let len = (span.hi - span.lo).0;
2130 Hash::hash(&col_line, hasher);
2131 Hash::hash(&len, hasher);