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 eq_ctxt(self, other: Span) -> bool {
537 self.data_untracked().ctxt == other.data_untracked().ctxt
540 pub fn with_ctxt(self, ctxt: SyntaxContext) -> Span {
541 self.data_untracked().with_ctxt(ctxt)
544 pub fn parent(self) -> Option<LocalDefId> {
548 pub fn with_parent(self, ctxt: Option<LocalDefId>) -> Span {
549 self.data().with_parent(ctxt)
552 /// Returns `true` if this is a dummy span with any hygienic context.
554 pub fn is_dummy(self) -> bool {
555 self.data_untracked().is_dummy()
558 /// Returns `true` if this span comes from a macro or desugaring.
560 pub fn from_expansion(self) -> bool {
561 self.ctxt() != SyntaxContext::root()
564 /// Returns `true` if `span` originates in a macro's expansion where debuginfo should be
566 pub fn in_macro_expansion_with_collapse_debuginfo(self) -> bool {
567 let outer_expn = self.ctxt().outer_expn_data();
568 matches!(outer_expn.kind, ExpnKind::Macro(..)) && outer_expn.collapse_debuginfo
571 /// Returns `true` if `span` originates in a derive-macro's expansion.
572 pub fn in_derive_expansion(self) -> bool {
573 matches!(self.ctxt().outer_expn_data().kind, ExpnKind::Macro(MacroKind::Derive, _))
576 /// Gate suggestions that would not be appropriate in a context the user didn't write.
577 pub fn can_be_used_for_suggestions(self) -> bool {
578 !self.from_expansion()
579 // FIXME: If this span comes from a `derive` macro but it points at code the user wrote,
580 // the callsite span and the span will be pointing at different places. It also means that
581 // we can safely provide suggestions on this span.
582 || (matches!(self.ctxt().outer_expn_data().kind, ExpnKind::Macro(MacroKind::Derive, _))
583 && self.parent_callsite().map(|p| (p.lo(), p.hi())) != Some((self.lo(), self.hi())))
587 pub fn with_root_ctxt(lo: BytePos, hi: BytePos) -> Span {
588 Span::new(lo, hi, SyntaxContext::root(), None)
591 /// Returns a new span representing an empty span at the beginning of this span.
593 pub fn shrink_to_lo(self) -> Span {
594 let span = self.data_untracked();
595 span.with_hi(span.lo)
597 /// Returns a new span representing an empty span at the end of this span.
599 pub fn shrink_to_hi(self) -> Span {
600 let span = self.data_untracked();
601 span.with_lo(span.hi)
605 /// Returns `true` if `hi == lo`.
606 pub fn is_empty(self) -> bool {
607 let span = self.data_untracked();
611 /// Returns `self` if `self` is not the dummy span, and `other` otherwise.
612 pub fn substitute_dummy(self, other: Span) -> Span {
613 if self.is_dummy() { other } else { self }
616 /// Returns `true` if `self` fully encloses `other`.
617 pub fn contains(self, other: Span) -> bool {
618 let span = self.data();
619 let other = other.data();
623 /// Returns `true` if `self` touches `other`.
624 pub fn overlaps(self, other: Span) -> bool {
625 let span = self.data();
626 let other = other.data();
627 span.lo < other.hi && other.lo < span.hi
630 /// Returns `true` if the spans are equal with regards to the source text.
632 /// Use this instead of `==` when either span could be generated code,
633 /// and you only care that they point to the same bytes of source text.
634 pub fn source_equal(self, other: Span) -> bool {
635 let span = self.data();
636 let other = other.data();
637 span.lo == other.lo && span.hi == other.hi
640 /// Returns `Some(span)`, where the start is trimmed by the end of `other`.
641 pub fn trim_start(self, other: Span) -> Option<Span> {
642 let span = self.data();
643 let other = other.data();
644 if span.hi > other.hi { Some(span.with_lo(cmp::max(span.lo, other.hi))) } else { None }
647 /// Returns the source span -- this is either the supplied span, or the span for
648 /// the macro callsite that expanded to it.
649 pub fn source_callsite(self) -> Span {
650 let expn_data = self.ctxt().outer_expn_data();
651 if !expn_data.is_root() { expn_data.call_site.source_callsite() } else { self }
654 /// The `Span` for the tokens in the previous macro expansion from which `self` was generated,
656 pub fn parent_callsite(self) -> Option<Span> {
657 let expn_data = self.ctxt().outer_expn_data();
658 if !expn_data.is_root() { Some(expn_data.call_site) } else { None }
661 /// Walk down the expansion ancestors to find a span that's contained within `outer`.
662 pub fn find_ancestor_inside(mut self, outer: Span) -> Option<Span> {
663 while !outer.contains(self) {
664 self = self.parent_callsite()?;
669 /// Like `find_ancestor_inside`, but specifically for when spans might not
670 /// overlaps. Take care when using this, and prefer `find_ancestor_inside`
671 /// when you know that the spans are nested (modulo macro expansion).
672 pub fn find_ancestor_in_same_ctxt(mut self, other: Span) -> Option<Span> {
673 while !Span::eq_ctxt(self, other) {
674 self = self.parent_callsite()?;
679 /// Edition of the crate from which this span came.
680 pub fn edition(self) -> edition::Edition {
681 self.ctxt().edition()
685 pub fn rust_2015(self) -> bool {
686 self.edition() == edition::Edition::Edition2015
690 pub fn rust_2018(self) -> bool {
691 self.edition() >= edition::Edition::Edition2018
695 pub fn rust_2021(self) -> bool {
696 self.edition() >= edition::Edition::Edition2021
700 pub fn rust_2024(self) -> bool {
701 self.edition() >= edition::Edition::Edition2024
704 /// Returns the source callee.
706 /// Returns `None` if the supplied span has no expansion trace,
707 /// else returns the `ExpnData` for the macro definition
708 /// corresponding to the source callsite.
709 pub fn source_callee(self) -> Option<ExpnData> {
710 fn source_callee(expn_data: ExpnData) -> ExpnData {
711 let next_expn_data = expn_data.call_site.ctxt().outer_expn_data();
712 if !next_expn_data.is_root() { source_callee(next_expn_data) } else { expn_data }
714 let expn_data = self.ctxt().outer_expn_data();
715 if !expn_data.is_root() { Some(source_callee(expn_data)) } else { None }
718 /// Checks if a span is "internal" to a macro in which `#[unstable]`
719 /// items can be used (that is, a macro marked with
720 /// `#[allow_internal_unstable]`).
721 pub fn allows_unstable(self, feature: Symbol) -> bool {
724 .allow_internal_unstable
725 .map_or(false, |features| features.iter().any(|&f| f == feature))
728 /// Checks if this span arises from a compiler desugaring of kind `kind`.
729 pub fn is_desugaring(self, kind: DesugaringKind) -> bool {
730 match self.ctxt().outer_expn_data().kind {
731 ExpnKind::Desugaring(k) => k == kind,
736 /// Returns the compiler desugaring that created this span, or `None`
737 /// if this span is not from a desugaring.
738 pub fn desugaring_kind(self) -> Option<DesugaringKind> {
739 match self.ctxt().outer_expn_data().kind {
740 ExpnKind::Desugaring(k) => Some(k),
745 /// Checks if a span is "internal" to a macro in which `unsafe`
746 /// can be used without triggering the `unsafe_code` lint.
747 // (that is, a macro marked with `#[allow_internal_unsafe]`).
748 pub fn allows_unsafe(self) -> bool {
749 self.ctxt().outer_expn_data().allow_internal_unsafe
752 pub fn macro_backtrace(mut self) -> impl Iterator<Item = ExpnData> {
753 let mut prev_span = DUMMY_SP;
754 std::iter::from_fn(move || {
756 let expn_data = self.ctxt().outer_expn_data();
757 if expn_data.is_root() {
761 let is_recursive = expn_data.call_site.source_equal(prev_span);
764 self = expn_data.call_site;
766 // Don't print recursive invocations.
768 return Some(expn_data);
774 /// Returns a `Span` that would enclose both `self` and `end`.
778 /// self lorem ipsum end
779 /// ^^^^^^^^^^^^^^^^^^^^
781 pub fn to(self, end: Span) -> Span {
782 let span_data = self.data();
783 let end_data = end.data();
784 // FIXME(jseyfried): `self.ctxt` should always equal `end.ctxt` here (cf. issue #23480).
785 // Return the macro span on its own to avoid weird diagnostic output. It is preferable to
786 // have an incomplete span than a completely nonsensical one.
787 if span_data.ctxt != end_data.ctxt {
788 if span_data.ctxt == SyntaxContext::root() {
790 } else if end_data.ctxt == SyntaxContext::root() {
793 // Both spans fall within a macro.
794 // FIXME(estebank): check if it is the *same* macro.
797 cmp::min(span_data.lo, end_data.lo),
798 cmp::max(span_data.hi, end_data.hi),
799 if span_data.ctxt == SyntaxContext::root() { end_data.ctxt } else { span_data.ctxt },
800 if span_data.parent == end_data.parent { span_data.parent } else { None },
804 /// Returns a `Span` between the end of `self` to the beginning of `end`.
808 /// self lorem ipsum end
811 pub fn between(self, end: Span) -> Span {
812 let span = self.data();
813 let end = end.data();
817 if end.ctxt == SyntaxContext::root() { end.ctxt } else { span.ctxt },
818 if span.parent == end.parent { span.parent } else { None },
822 /// Returns a `Span` from the beginning of `self` until the beginning of `end`.
826 /// self lorem ipsum end
827 /// ^^^^^^^^^^^^^^^^^
829 pub fn until(self, end: Span) -> Span {
830 // Most of this function's body is copied from `to`.
831 // We can't just do `self.to(end.shrink_to_lo())`,
832 // because to also does some magic where it uses min/max so
833 // it can handle overlapping spans. Some advanced mis-use of
834 // `until` with different ctxts makes this visible.
835 let span_data = self.data();
836 let end_data = end.data();
837 // FIXME(jseyfried): `self.ctxt` should always equal `end.ctxt` here (cf. issue #23480).
838 // Return the macro span on its own to avoid weird diagnostic output. It is preferable to
839 // have an incomplete span than a completely nonsensical one.
840 if span_data.ctxt != end_data.ctxt {
841 if span_data.ctxt == SyntaxContext::root() {
843 } else if end_data.ctxt == SyntaxContext::root() {
846 // Both spans fall within a macro.
847 // FIXME(estebank): check if it is the *same* macro.
852 if end_data.ctxt == SyntaxContext::root() { end_data.ctxt } else { span_data.ctxt },
853 if span_data.parent == end_data.parent { span_data.parent } else { None },
857 pub fn from_inner(self, inner: InnerSpan) -> Span {
858 let span = self.data();
860 span.lo + BytePos::from_usize(inner.start),
861 span.lo + BytePos::from_usize(inner.end),
867 /// Equivalent of `Span::def_site` from the proc macro API,
868 /// except that the location is taken from the `self` span.
869 pub fn with_def_site_ctxt(self, expn_id: ExpnId) -> Span {
870 self.with_ctxt_from_mark(expn_id, Transparency::Opaque)
873 /// Equivalent of `Span::call_site` from the proc macro API,
874 /// except that the location is taken from the `self` span.
875 pub fn with_call_site_ctxt(self, expn_id: ExpnId) -> Span {
876 self.with_ctxt_from_mark(expn_id, Transparency::Transparent)
879 /// Equivalent of `Span::mixed_site` from the proc macro API,
880 /// except that the location is taken from the `self` span.
881 pub fn with_mixed_site_ctxt(self, expn_id: ExpnId) -> Span {
882 self.with_ctxt_from_mark(expn_id, Transparency::SemiTransparent)
885 /// Produces a span with the same location as `self` and context produced by a macro with the
886 /// given ID and transparency, assuming that macro was defined directly and not produced by
887 /// some other macro (which is the case for built-in and procedural macros).
888 pub fn with_ctxt_from_mark(self, expn_id: ExpnId, transparency: Transparency) -> Span {
889 self.with_ctxt(SyntaxContext::root().apply_mark(expn_id, transparency))
893 pub fn apply_mark(self, expn_id: ExpnId, transparency: Transparency) -> Span {
894 let span = self.data();
895 span.with_ctxt(span.ctxt.apply_mark(expn_id, transparency))
899 pub fn remove_mark(&mut self) -> ExpnId {
900 let mut span = self.data();
901 let mark = span.ctxt.remove_mark();
902 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
907 pub fn adjust(&mut self, expn_id: ExpnId) -> Option<ExpnId> {
908 let mut span = self.data();
909 let mark = span.ctxt.adjust(expn_id);
910 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
915 pub fn normalize_to_macros_2_0_and_adjust(&mut self, expn_id: ExpnId) -> Option<ExpnId> {
916 let mut span = self.data();
917 let mark = span.ctxt.normalize_to_macros_2_0_and_adjust(expn_id);
918 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
923 pub fn glob_adjust(&mut self, expn_id: ExpnId, glob_span: Span) -> Option<Option<ExpnId>> {
924 let mut span = self.data();
925 let mark = span.ctxt.glob_adjust(expn_id, glob_span);
926 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
931 pub fn reverse_glob_adjust(
935 ) -> Option<Option<ExpnId>> {
936 let mut span = self.data();
937 let mark = span.ctxt.reverse_glob_adjust(expn_id, glob_span);
938 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
943 pub fn normalize_to_macros_2_0(self) -> Span {
944 let span = self.data();
945 span.with_ctxt(span.ctxt.normalize_to_macros_2_0())
949 pub fn normalize_to_macro_rules(self) -> Span {
950 let span = self.data();
951 span.with_ctxt(span.ctxt.normalize_to_macro_rules())
955 impl Default for Span {
956 fn default() -> Self {
961 impl<E: Encoder> Encodable<E> for Span {
962 default fn encode(&self, s: &mut E) {
963 let span = self.data();
968 impl<D: Decoder> Decodable<D> for Span {
969 default fn decode(s: &mut D) -> Span {
970 let lo = Decodable::decode(s);
971 let hi = Decodable::decode(s);
973 Span::new(lo, hi, SyntaxContext::root(), None)
977 /// Calls the provided closure, using the provided `SourceMap` to format
978 /// any spans that are debug-printed during the closure's execution.
980 /// Normally, the global `TyCtxt` is used to retrieve the `SourceMap`
981 /// (see `rustc_interface::callbacks::span_debug1`). However, some parts
982 /// of the compiler (e.g. `rustc_parse`) may debug-print `Span`s before
983 /// a `TyCtxt` is available. In this case, we fall back to
984 /// the `SourceMap` provided to this function. If that is not available,
985 /// we fall back to printing the raw `Span` field values.
986 pub fn with_source_map<T, F: FnOnce() -> T>(source_map: Lrc<SourceMap>, f: F) -> T {
987 with_session_globals(|session_globals| {
988 *session_globals.source_map.borrow_mut() = Some(source_map);
990 struct ClearSourceMap;
991 impl Drop for ClearSourceMap {
993 with_session_globals(|session_globals| {
994 session_globals.source_map.borrow_mut().take();
999 let _guard = ClearSourceMap;
1003 impl fmt::Debug for Span {
1004 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
1005 with_session_globals(|session_globals| {
1006 if let Some(source_map) = &*session_globals.source_map.borrow() {
1007 write!(f, "{} ({:?})", source_map.span_to_diagnostic_string(*self), self.ctxt())
1009 f.debug_struct("Span")
1010 .field("lo", &self.lo())
1011 .field("hi", &self.hi())
1012 .field("ctxt", &self.ctxt())
1019 impl fmt::Debug for SpanData {
1020 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
1021 fmt::Debug::fmt(&Span::new(self.lo, self.hi, self.ctxt, self.parent), f)
1025 /// Identifies an offset of a multi-byte character in a `SourceFile`.
1026 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1027 pub struct MultiByteChar {
1028 /// The absolute offset of the character in the `SourceMap`.
1030 /// The number of bytes, `>= 2`.
1034 /// Identifies an offset of a non-narrow character in a `SourceFile`.
1035 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1036 pub enum NonNarrowChar {
1037 /// Represents a zero-width character.
1039 /// Represents a wide (full-width) character.
1041 /// Represents a tab character, represented visually with a width of 4 characters.
1045 impl NonNarrowChar {
1046 fn new(pos: BytePos, width: usize) -> Self {
1048 0 => NonNarrowChar::ZeroWidth(pos),
1049 2 => NonNarrowChar::Wide(pos),
1050 4 => NonNarrowChar::Tab(pos),
1051 _ => panic!("width {} given for non-narrow character", width),
1055 /// Returns the absolute offset of the character in the `SourceMap`.
1056 pub fn pos(&self) -> BytePos {
1058 NonNarrowChar::ZeroWidth(p) | NonNarrowChar::Wide(p) | NonNarrowChar::Tab(p) => p,
1062 /// Returns the width of the character, 0 (zero-width) or 2 (wide).
1063 pub fn width(&self) -> usize {
1065 NonNarrowChar::ZeroWidth(_) => 0,
1066 NonNarrowChar::Wide(_) => 2,
1067 NonNarrowChar::Tab(_) => 4,
1072 impl Add<BytePos> for NonNarrowChar {
1075 fn add(self, rhs: BytePos) -> Self {
1077 NonNarrowChar::ZeroWidth(pos) => NonNarrowChar::ZeroWidth(pos + rhs),
1078 NonNarrowChar::Wide(pos) => NonNarrowChar::Wide(pos + rhs),
1079 NonNarrowChar::Tab(pos) => NonNarrowChar::Tab(pos + rhs),
1084 impl Sub<BytePos> for NonNarrowChar {
1087 fn sub(self, rhs: BytePos) -> Self {
1089 NonNarrowChar::ZeroWidth(pos) => NonNarrowChar::ZeroWidth(pos - rhs),
1090 NonNarrowChar::Wide(pos) => NonNarrowChar::Wide(pos - rhs),
1091 NonNarrowChar::Tab(pos) => NonNarrowChar::Tab(pos - rhs),
1096 /// Identifies an offset of a character that was normalized away from `SourceFile`.
1097 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1098 pub struct NormalizedPos {
1099 /// The absolute offset of the character in the `SourceMap`.
1101 /// The difference between original and normalized string at position.
1105 #[derive(PartialEq, Eq, Clone, Debug)]
1106 pub enum ExternalSource {
1107 /// No external source has to be loaded, since the `SourceFile` represents a local crate.
1110 kind: ExternalSourceKind,
1111 /// Index of the file inside metadata.
1112 metadata_index: u32,
1116 /// The state of the lazy external source loading mechanism of a `SourceFile`.
1117 #[derive(PartialEq, Eq, Clone, Debug)]
1118 pub enum ExternalSourceKind {
1119 /// The external source has been loaded already.
1120 Present(Lrc<String>),
1121 /// No attempt has been made to load the external source.
1123 /// A failed attempt has been made to load the external source.
1128 impl ExternalSource {
1129 pub fn get_source(&self) -> Option<&Lrc<String>> {
1131 ExternalSource::Foreign { kind: ExternalSourceKind::Present(ref src), .. } => Some(src),
1138 pub struct OffsetOverflowError;
1140 #[derive(Copy, Clone, Debug, PartialEq, Eq, PartialOrd, Ord, Hash, Encodable, Decodable)]
1141 #[derive(HashStable_Generic)]
1142 pub enum SourceFileHashAlgorithm {
1148 impl FromStr for SourceFileHashAlgorithm {
1151 fn from_str(s: &str) -> Result<SourceFileHashAlgorithm, ()> {
1153 "md5" => Ok(SourceFileHashAlgorithm::Md5),
1154 "sha1" => Ok(SourceFileHashAlgorithm::Sha1),
1155 "sha256" => Ok(SourceFileHashAlgorithm::Sha256),
1161 /// The hash of the on-disk source file used for debug info.
1162 #[derive(Copy, Clone, PartialEq, Eq, Debug, Hash)]
1163 #[derive(HashStable_Generic, Encodable, Decodable)]
1164 pub struct SourceFileHash {
1165 pub kind: SourceFileHashAlgorithm,
1169 impl SourceFileHash {
1170 pub fn new(kind: SourceFileHashAlgorithm, src: &str) -> SourceFileHash {
1171 let mut hash = SourceFileHash { kind, value: Default::default() };
1172 let len = hash.hash_len();
1173 let value = &mut hash.value[..len];
1174 let data = src.as_bytes();
1176 SourceFileHashAlgorithm::Md5 => {
1177 value.copy_from_slice(&Md5::digest(data));
1179 SourceFileHashAlgorithm::Sha1 => {
1180 value.copy_from_slice(&Sha1::digest(data));
1182 SourceFileHashAlgorithm::Sha256 => {
1183 value.copy_from_slice(&Sha256::digest(data));
1189 /// Check if the stored hash matches the hash of the string.
1190 pub fn matches(&self, src: &str) -> bool {
1191 Self::new(self.kind, src) == *self
1194 /// The bytes of the hash.
1195 pub fn hash_bytes(&self) -> &[u8] {
1196 let len = self.hash_len();
1200 fn hash_len(&self) -> usize {
1202 SourceFileHashAlgorithm::Md5 => 16,
1203 SourceFileHashAlgorithm::Sha1 => 20,
1204 SourceFileHashAlgorithm::Sha256 => 32,
1209 #[derive(HashStable_Generic)]
1210 #[derive(Copy, PartialEq, PartialOrd, Clone, Ord, Eq, Hash, Debug, Encodable, Decodable)]
1211 pub enum DebuggerVisualizerType {
1216 /// A single debugger visualizer file.
1217 #[derive(HashStable_Generic)]
1218 #[derive(Clone, Debug, Hash, PartialEq, Eq, PartialOrd, Ord, Encodable, Decodable)]
1219 pub struct DebuggerVisualizerFile {
1220 /// The complete debugger visualizer source.
1222 /// Indicates which visualizer type this targets.
1223 pub visualizer_type: DebuggerVisualizerType,
1226 impl DebuggerVisualizerFile {
1227 pub fn new(src: Arc<[u8]>, visualizer_type: DebuggerVisualizerType) -> Self {
1228 DebuggerVisualizerFile { src, visualizer_type }
1233 pub enum SourceFileLines {
1234 /// The source file lines, in decoded (random-access) form.
1235 Lines(Vec<BytePos>),
1237 /// The source file lines, in undecoded difference list form.
1238 Diffs(SourceFileDiffs),
1241 impl SourceFileLines {
1242 pub fn is_lines(&self) -> bool {
1243 matches!(self, SourceFileLines::Lines(_))
1247 /// The source file lines in difference list form. This matches the form
1248 /// used within metadata, which saves space by exploiting the fact that the
1249 /// lines list is sorted and individual lines are usually not that long.
1251 /// We read it directly from metadata and only decode it into `Lines` form
1252 /// when necessary. This is a significant performance win, especially for
1253 /// small crates where very little of `std`'s metadata is used.
1255 pub struct SourceFileDiffs {
1256 /// Position of the first line. Note that this is always encoded as a
1257 /// `BytePos` because it is often much larger than any of the
1259 line_start: BytePos,
1261 /// Always 1, 2, or 4. Always as small as possible, while being big
1262 /// enough to hold the length of the longest line in the source file.
1263 /// The 1 case is by far the most common.
1264 bytes_per_diff: usize,
1266 /// The number of diffs encoded in `raw_diffs`. Always one less than
1267 /// the number of lines in the source file.
1270 /// The diffs in "raw" form. Each segment of `bytes_per_diff` length
1271 /// encodes one little-endian diff. Note that they aren't LEB128
1272 /// encoded. This makes for much faster decoding. Besides, the
1273 /// bytes_per_diff==1 case is by far the most common, and LEB128
1274 /// encoding has no effect on that case.
1278 /// A single source in the [`SourceMap`].
1280 pub struct SourceFile {
1281 /// The name of the file that the source came from. Source that doesn't
1282 /// originate from files has names between angle brackets by convention
1283 /// (e.g., `<anon>`).
1285 /// The complete source code.
1286 pub src: Option<Lrc<String>>,
1287 /// The source code's hash.
1288 pub src_hash: SourceFileHash,
1289 /// The external source code (used for external crates, which will have a `None`
1290 /// value as `self.src`.
1291 pub external_src: Lock<ExternalSource>,
1292 /// The start position of this source in the `SourceMap`.
1293 pub start_pos: BytePos,
1294 /// The end position of this source in the `SourceMap`.
1295 pub end_pos: BytePos,
1296 /// Locations of lines beginnings in the source code.
1297 pub lines: Lock<SourceFileLines>,
1298 /// Locations of multi-byte characters in the source code.
1299 pub multibyte_chars: Vec<MultiByteChar>,
1300 /// Width of characters that are not narrow in the source code.
1301 pub non_narrow_chars: Vec<NonNarrowChar>,
1302 /// Locations of characters removed during normalization.
1303 pub normalized_pos: Vec<NormalizedPos>,
1304 /// A hash of the filename, used for speeding up hashing in incremental compilation.
1305 pub name_hash: u128,
1306 /// Indicates which crate this `SourceFile` was imported from.
1310 impl<S: Encoder> Encodable<S> for SourceFile {
1311 fn encode(&self, s: &mut S) {
1312 self.name.encode(s);
1313 self.src_hash.encode(s);
1314 self.start_pos.encode(s);
1315 self.end_pos.encode(s);
1317 // We are always in `Lines` form by the time we reach here.
1318 assert!(self.lines.borrow().is_lines());
1319 self.lines(|lines| {
1320 // Store the length.
1321 s.emit_u32(lines.len() as u32);
1323 // Compute and store the difference list.
1324 if lines.len() != 0 {
1325 let max_line_length = if lines.len() == 1 {
1330 .map(|&[fst, snd]| snd - fst)
1331 .map(|bp| bp.to_usize())
1336 let bytes_per_diff: usize = match max_line_length {
1338 0x100..=0xFFFF => 2,
1342 // Encode the number of bytes used per diff.
1343 s.emit_u8(bytes_per_diff as u8);
1345 // Encode the first element.
1348 // Encode the difference list.
1349 let diff_iter = lines.array_windows().map(|&[fst, snd]| snd - fst);
1350 let num_diffs = lines.len() - 1;
1352 match bytes_per_diff {
1354 raw_diffs = Vec::with_capacity(num_diffs);
1355 for diff in diff_iter {
1356 raw_diffs.push(diff.0 as u8);
1360 raw_diffs = Vec::with_capacity(bytes_per_diff * num_diffs);
1361 for diff in diff_iter {
1362 raw_diffs.extend_from_slice(&(diff.0 as u16).to_le_bytes());
1366 raw_diffs = Vec::with_capacity(bytes_per_diff * num_diffs);
1367 for diff in diff_iter {
1368 raw_diffs.extend_from_slice(&(diff.0 as u32).to_le_bytes());
1371 _ => unreachable!(),
1373 s.emit_raw_bytes(&raw_diffs);
1377 self.multibyte_chars.encode(s);
1378 self.non_narrow_chars.encode(s);
1379 self.name_hash.encode(s);
1380 self.normalized_pos.encode(s);
1381 self.cnum.encode(s);
1385 impl<D: Decoder> Decodable<D> for SourceFile {
1386 fn decode(d: &mut D) -> SourceFile {
1387 let name: FileName = Decodable::decode(d);
1388 let src_hash: SourceFileHash = Decodable::decode(d);
1389 let start_pos: BytePos = Decodable::decode(d);
1390 let end_pos: BytePos = Decodable::decode(d);
1392 let num_lines: u32 = Decodable::decode(d);
1394 // Read the number of bytes used per diff.
1395 let bytes_per_diff = d.read_u8() as usize;
1397 // Read the first element.
1398 let line_start: BytePos = Decodable::decode(d);
1400 // Read the difference list.
1401 let num_diffs = num_lines as usize - 1;
1402 let raw_diffs = d.read_raw_bytes(bytes_per_diff * num_diffs).to_vec();
1403 SourceFileLines::Diffs(SourceFileDiffs {
1410 SourceFileLines::Lines(vec![])
1413 let multibyte_chars: Vec<MultiByteChar> = Decodable::decode(d);
1414 let non_narrow_chars: Vec<NonNarrowChar> = Decodable::decode(d);
1415 let name_hash: u128 = Decodable::decode(d);
1416 let normalized_pos: Vec<NormalizedPos> = Decodable::decode(d);
1417 let cnum: CrateNum = Decodable::decode(d);
1424 // Unused - the metadata decoder will construct
1425 // a new SourceFile, filling in `external_src` properly
1426 external_src: Lock::new(ExternalSource::Unneeded),
1427 lines: Lock::new(lines),
1437 impl fmt::Debug for SourceFile {
1438 fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result {
1439 write!(fmt, "SourceFile({:?})", self.name)
1448 hash_kind: SourceFileHashAlgorithm,
1450 // Compute the file hash before any normalization.
1451 let src_hash = SourceFileHash::new(hash_kind, &src);
1452 let normalized_pos = normalize_src(&mut src, start_pos);
1455 let mut hasher: StableHasher = StableHasher::new();
1456 name.hash(&mut hasher);
1457 hasher.finish::<u128>()
1459 let end_pos = start_pos.to_usize() + src.len();
1460 assert!(end_pos <= u32::MAX as usize);
1462 let (lines, multibyte_chars, non_narrow_chars) =
1463 analyze_source_file::analyze_source_file(&src, start_pos);
1467 src: Some(Lrc::new(src)),
1469 external_src: Lock::new(ExternalSource::Unneeded),
1471 end_pos: Pos::from_usize(end_pos),
1472 lines: Lock::new(SourceFileLines::Lines(lines)),
1481 pub fn lines<F, R>(&self, f: F) -> R
1483 F: FnOnce(&[BytePos]) -> R,
1485 let mut guard = self.lines.borrow_mut();
1487 SourceFileLines::Lines(lines) => f(lines),
1488 SourceFileLines::Diffs(SourceFileDiffs {
1494 // Convert from "diffs" form to "lines" form.
1495 let num_lines = num_diffs + 1;
1496 let mut lines = Vec::with_capacity(num_lines);
1497 lines.push(line_start);
1499 assert_eq!(*num_diffs, raw_diffs.len() / bytes_per_diff);
1500 match bytes_per_diff {
1502 lines.extend(raw_diffs.into_iter().map(|&diff| {
1503 line_start = line_start + BytePos(diff as u32);
1508 lines.extend((0..*num_diffs).map(|i| {
1509 let pos = bytes_per_diff * i;
1510 let bytes = [raw_diffs[pos], raw_diffs[pos + 1]];
1511 let diff = u16::from_le_bytes(bytes);
1512 line_start = line_start + BytePos(diff as u32);
1517 lines.extend((0..*num_diffs).map(|i| {
1518 let pos = bytes_per_diff * i;
1525 let diff = u32::from_le_bytes(bytes);
1526 line_start = line_start + BytePos(diff);
1530 _ => unreachable!(),
1532 let res = f(&lines);
1533 *guard = SourceFileLines::Lines(lines);
1539 /// Returns the `BytePos` of the beginning of the current line.
1540 pub fn line_begin_pos(&self, pos: BytePos) -> BytePos {
1541 let line_index = self.lookup_line(pos).unwrap();
1542 self.lines(|lines| lines[line_index])
1545 /// Add externally loaded source.
1546 /// If the hash of the input doesn't match or no input is supplied via None,
1547 /// it is interpreted as an error and the corresponding enum variant is set.
1548 /// The return value signifies whether some kind of source is present.
1549 pub fn add_external_src<F>(&self, get_src: F) -> bool
1551 F: FnOnce() -> Option<String>,
1554 *self.external_src.borrow(),
1555 ExternalSource::Foreign { kind: ExternalSourceKind::AbsentOk, .. }
1557 let src = get_src();
1558 let mut external_src = self.external_src.borrow_mut();
1559 // Check that no-one else have provided the source while we were getting it
1560 if let ExternalSource::Foreign {
1561 kind: src_kind @ ExternalSourceKind::AbsentOk, ..
1562 } = &mut *external_src
1564 if let Some(mut src) = src {
1565 // The src_hash needs to be computed on the pre-normalized src.
1566 if self.src_hash.matches(&src) {
1567 normalize_src(&mut src, BytePos::from_usize(0));
1568 *src_kind = ExternalSourceKind::Present(Lrc::new(src));
1572 *src_kind = ExternalSourceKind::AbsentErr;
1577 self.src.is_some() || external_src.get_source().is_some()
1580 self.src.is_some() || self.external_src.borrow().get_source().is_some()
1584 /// Gets a line from the list of pre-computed line-beginnings.
1585 /// The line number here is 0-based.
1586 pub fn get_line(&self, line_number: usize) -> Option<Cow<'_, str>> {
1587 fn get_until_newline(src: &str, begin: usize) -> &str {
1588 // We can't use `lines.get(line_number+1)` because we might
1589 // be parsing when we call this function and thus the current
1590 // line is the last one we have line info for.
1591 let slice = &src[begin..];
1592 match slice.find('\n') {
1593 Some(e) => &slice[..e],
1599 let line = self.lines(|lines| lines.get(line_number).copied())?;
1600 let begin: BytePos = line - self.start_pos;
1604 if let Some(ref src) = self.src {
1605 Some(Cow::from(get_until_newline(src, begin)))
1606 } else if let Some(src) = self.external_src.borrow().get_source() {
1607 Some(Cow::Owned(String::from(get_until_newline(src, begin))))
1613 pub fn is_real_file(&self) -> bool {
1618 pub fn is_imported(&self) -> bool {
1622 pub fn count_lines(&self) -> usize {
1623 self.lines(|lines| lines.len())
1626 /// Finds the line containing the given position. The return value is the
1627 /// index into the `lines` array of this `SourceFile`, not the 1-based line
1628 /// number. If the source_file is empty or the position is located before the
1629 /// first line, `None` is returned.
1630 pub fn lookup_line(&self, pos: BytePos) -> Option<usize> {
1631 self.lines(|lines| match lines.partition_point(|x| x <= &pos) {
1637 pub fn line_bounds(&self, line_index: usize) -> Range<BytePos> {
1638 if self.is_empty() {
1639 return self.start_pos..self.end_pos;
1642 self.lines(|lines| {
1643 assert!(line_index < lines.len());
1644 if line_index == (lines.len() - 1) {
1645 lines[line_index]..self.end_pos
1647 lines[line_index]..lines[line_index + 1]
1652 /// Returns whether or not the file contains the given `SourceMap` byte
1653 /// position. The position one past the end of the file is considered to be
1654 /// contained by the file. This implies that files for which `is_empty`
1655 /// returns true still contain one byte position according to this function.
1657 pub fn contains(&self, byte_pos: BytePos) -> bool {
1658 byte_pos >= self.start_pos && byte_pos <= self.end_pos
1662 pub fn is_empty(&self) -> bool {
1663 self.start_pos == self.end_pos
1666 /// Calculates the original byte position relative to the start of the file
1667 /// based on the given byte position.
1668 pub fn original_relative_byte_pos(&self, pos: BytePos) -> BytePos {
1669 // Diff before any records is 0. Otherwise use the previously recorded
1670 // diff as that applies to the following characters until a new diff
1672 let diff = match self.normalized_pos.binary_search_by(|np| np.pos.cmp(&pos)) {
1673 Ok(i) => self.normalized_pos[i].diff,
1674 Err(i) if i == 0 => 0,
1675 Err(i) => self.normalized_pos[i - 1].diff,
1678 BytePos::from_u32(pos.0 - self.start_pos.0 + diff)
1681 /// Converts an absolute `BytePos` to a `CharPos` relative to the `SourceFile`.
1682 pub fn bytepos_to_file_charpos(&self, bpos: BytePos) -> CharPos {
1683 // The number of extra bytes due to multibyte chars in the `SourceFile`.
1684 let mut total_extra_bytes = 0;
1686 for mbc in self.multibyte_chars.iter() {
1687 debug!("{}-byte char at {:?}", mbc.bytes, mbc.pos);
1689 // Every character is at least one byte, so we only
1690 // count the actual extra bytes.
1691 total_extra_bytes += mbc.bytes as u32 - 1;
1692 // We should never see a byte position in the middle of a
1694 assert!(bpos.to_u32() >= mbc.pos.to_u32() + mbc.bytes as u32);
1700 assert!(self.start_pos.to_u32() + total_extra_bytes <= bpos.to_u32());
1701 CharPos(bpos.to_usize() - self.start_pos.to_usize() - total_extra_bytes as usize)
1704 /// Looks up the file's (1-based) line number and (0-based `CharPos`) column offset, for a
1705 /// given `BytePos`.
1706 pub fn lookup_file_pos(&self, pos: BytePos) -> (usize, CharPos) {
1707 let chpos = self.bytepos_to_file_charpos(pos);
1708 match self.lookup_line(pos) {
1710 let line = a + 1; // Line numbers start at 1
1711 let linebpos = self.lines(|lines| lines[a]);
1712 let linechpos = self.bytepos_to_file_charpos(linebpos);
1713 let col = chpos - linechpos;
1714 debug!("byte pos {:?} is on the line at byte pos {:?}", pos, linebpos);
1715 debug!("char pos {:?} is on the line at char pos {:?}", chpos, linechpos);
1716 debug!("byte is on line: {}", line);
1717 assert!(chpos >= linechpos);
1724 /// Looks up the file's (1-based) line number, (0-based `CharPos`) column offset, and (0-based)
1725 /// column offset when displayed, for a given `BytePos`.
1726 pub fn lookup_file_pos_with_col_display(&self, pos: BytePos) -> (usize, CharPos, usize) {
1727 let (line, col_or_chpos) = self.lookup_file_pos(pos);
1729 let col = col_or_chpos;
1730 let linebpos = self.lines(|lines| lines[line - 1]);
1732 let start_width_idx = self
1734 .binary_search_by_key(&linebpos, |x| x.pos())
1735 .unwrap_or_else(|x| x);
1736 let end_width_idx = self
1738 .binary_search_by_key(&pos, |x| x.pos())
1739 .unwrap_or_else(|x| x);
1740 let special_chars = end_width_idx - start_width_idx;
1741 let non_narrow: usize = self.non_narrow_chars[start_width_idx..end_width_idx]
1745 col.0 - special_chars + non_narrow
1747 (line, col, col_display)
1749 let chpos = col_or_chpos;
1751 let end_width_idx = self
1753 .binary_search_by_key(&pos, |x| x.pos())
1754 .unwrap_or_else(|x| x);
1755 let non_narrow: usize =
1756 self.non_narrow_chars[0..end_width_idx].iter().map(|x| x.width()).sum();
1757 chpos.0 - end_width_idx + non_narrow
1759 (0, chpos, col_display)
1764 /// Normalizes the source code and records the normalizations.
1765 fn normalize_src(src: &mut String, start_pos: BytePos) -> Vec<NormalizedPos> {
1766 let mut normalized_pos = vec![];
1767 remove_bom(src, &mut normalized_pos);
1768 normalize_newlines(src, &mut normalized_pos);
1770 // Offset all the positions by start_pos to match the final file positions.
1771 for np in &mut normalized_pos {
1772 np.pos.0 += start_pos.0;
1778 /// Removes UTF-8 BOM, if any.
1779 fn remove_bom(src: &mut String, normalized_pos: &mut Vec<NormalizedPos>) {
1780 if src.starts_with('\u{feff}') {
1782 normalized_pos.push(NormalizedPos { pos: BytePos(0), diff: 3 });
1786 /// Replaces `\r\n` with `\n` in-place in `src`.
1788 /// Returns error if there's a lone `\r` in the string.
1789 fn normalize_newlines(src: &mut String, normalized_pos: &mut Vec<NormalizedPos>) {
1790 if !src.as_bytes().contains(&b'\r') {
1794 // We replace `\r\n` with `\n` in-place, which doesn't break utf-8 encoding.
1795 // While we *can* call `as_mut_vec` and do surgery on the live string
1796 // directly, let's rather steal the contents of `src`. This makes the code
1797 // safe even if a panic occurs.
1799 let mut buf = std::mem::replace(src, String::new()).into_bytes();
1800 let mut gap_len = 0;
1801 let mut tail = buf.as_mut_slice();
1803 let original_gap = normalized_pos.last().map_or(0, |l| l.diff);
1805 let idx = match find_crlf(&tail[gap_len..]) {
1807 Some(idx) => idx + gap_len,
1809 tail.copy_within(gap_len..idx, 0);
1810 tail = &mut tail[idx - gap_len..];
1811 if tail.len() == gap_len {
1814 cursor += idx - gap_len;
1816 normalized_pos.push(NormalizedPos {
1817 pos: BytePos::from_usize(cursor + 1),
1818 diff: original_gap + gap_len as u32,
1822 // Account for removed `\r`.
1823 // After `set_len`, `buf` is guaranteed to contain utf-8 again.
1824 let new_len = buf.len() - gap_len;
1826 buf.set_len(new_len);
1827 *src = String::from_utf8_unchecked(buf);
1830 fn find_crlf(src: &[u8]) -> Option<usize> {
1831 let mut search_idx = 0;
1832 while let Some(idx) = find_cr(&src[search_idx..]) {
1833 if src[search_idx..].get(idx + 1) != Some(&b'\n') {
1834 search_idx += idx + 1;
1837 return Some(search_idx + idx);
1842 fn find_cr(src: &[u8]) -> Option<usize> {
1843 src.iter().position(|&b| b == b'\r')
1847 // _____________________________________________________________________________
1848 // Pos, BytePos, CharPos
1852 fn from_usize(n: usize) -> Self;
1853 fn to_usize(&self) -> usize;
1854 fn from_u32(n: u32) -> Self;
1855 fn to_u32(&self) -> u32;
1858 macro_rules! impl_pos {
1862 $vis:vis struct $ident:ident($inner_vis:vis $inner_ty:ty);
1867 $vis struct $ident($inner_vis $inner_ty);
1869 impl Pos for $ident {
1871 fn from_usize(n: usize) -> $ident {
1872 $ident(n as $inner_ty)
1876 fn to_usize(&self) -> usize {
1881 fn from_u32(n: u32) -> $ident {
1882 $ident(n as $inner_ty)
1886 fn to_u32(&self) -> u32 {
1891 impl Add for $ident {
1892 type Output = $ident;
1895 fn add(self, rhs: $ident) -> $ident {
1896 $ident(self.0 + rhs.0)
1900 impl Sub for $ident {
1901 type Output = $ident;
1904 fn sub(self, rhs: $ident) -> $ident {
1905 $ident(self.0 - rhs.0)
1915 /// Keep this small (currently 32-bits), as AST contains a lot of them.
1916 #[derive(Clone, Copy, PartialEq, Eq, Hash, PartialOrd, Ord, Debug)]
1917 pub struct BytePos(pub u32);
1919 /// A character offset.
1921 /// Because of multibyte UTF-8 characters, a byte offset
1922 /// is not equivalent to a character offset. The [`SourceMap`] will convert [`BytePos`]
1923 /// values to `CharPos` values as necessary.
1924 #[derive(Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Debug)]
1925 pub struct CharPos(pub usize);
1928 impl<S: Encoder> Encodable<S> for BytePos {
1929 fn encode(&self, s: &mut S) {
1934 impl<D: Decoder> Decodable<D> for BytePos {
1935 fn decode(d: &mut D) -> BytePos {
1936 BytePos(d.read_u32())
1940 // _____________________________________________________________________________
1941 // Loc, SourceFileAndLine, SourceFileAndBytePos
1944 /// A source code location used for error reporting.
1945 #[derive(Debug, Clone)]
1947 /// Information about the original source.
1948 pub file: Lrc<SourceFile>,
1949 /// The (1-based) line number.
1951 /// The (0-based) column offset.
1953 /// The (0-based) column offset when displayed.
1954 pub col_display: usize,
1957 // Used to be structural records.
1959 pub struct SourceFileAndLine {
1960 pub sf: Lrc<SourceFile>,
1961 /// Index of line, starting from 0.
1965 pub struct SourceFileAndBytePos {
1966 pub sf: Lrc<SourceFile>,
1970 #[derive(Copy, Clone, Debug, PartialEq, Eq)]
1971 pub struct LineInfo {
1972 /// Index of line, starting from 0.
1973 pub line_index: usize,
1975 /// Column in line where span begins, starting from 0.
1976 pub start_col: CharPos,
1978 /// Column in line where span ends, starting from 0, exclusive.
1979 pub end_col: CharPos,
1982 pub struct FileLines {
1983 pub file: Lrc<SourceFile>,
1984 pub lines: Vec<LineInfo>,
1987 pub static SPAN_TRACK: AtomicRef<fn(LocalDefId)> = AtomicRef::new(&((|_| {}) as fn(_)));
1989 // _____________________________________________________________________________
1990 // SpanLinesError, SpanSnippetError, DistinctSources, MalformedSourceMapPositions
1993 pub type FileLinesResult = Result<FileLines, SpanLinesError>;
1995 #[derive(Clone, PartialEq, Eq, Debug)]
1996 pub enum SpanLinesError {
1997 DistinctSources(DistinctSources),
2000 #[derive(Clone, PartialEq, Eq, Debug)]
2001 pub enum SpanSnippetError {
2002 IllFormedSpan(Span),
2003 DistinctSources(DistinctSources),
2004 MalformedForSourcemap(MalformedSourceMapPositions),
2005 SourceNotAvailable { filename: FileName },
2008 #[derive(Clone, PartialEq, Eq, Debug)]
2009 pub struct DistinctSources {
2010 pub begin: (FileName, BytePos),
2011 pub end: (FileName, BytePos),
2014 #[derive(Clone, PartialEq, Eq, Debug)]
2015 pub struct MalformedSourceMapPositions {
2017 pub source_len: usize,
2018 pub begin_pos: BytePos,
2019 pub end_pos: BytePos,
2022 /// Range inside of a `Span` used for diagnostics when we only have access to relative positions.
2023 #[derive(Copy, Clone, PartialEq, Eq, Debug)]
2024 pub struct InnerSpan {
2030 pub fn new(start: usize, end: usize) -> InnerSpan {
2031 InnerSpan { start, end }
2035 /// Requirements for a `StableHashingContext` to be used in this crate.
2037 /// This is a hack to allow using the [`HashStable_Generic`] derive macro
2038 /// instead of implementing everything in rustc_middle.
2039 pub trait HashStableContext {
2040 fn def_path_hash(&self, def_id: DefId) -> DefPathHash;
2041 fn hash_spans(&self) -> bool;
2042 /// Accesses `sess.opts.unstable_opts.incremental_ignore_spans` since
2043 /// we don't have easy access to a `Session`
2044 fn unstable_opts_incremental_ignore_spans(&self) -> bool;
2045 fn def_span(&self, def_id: LocalDefId) -> Span;
2046 fn span_data_to_lines_and_cols(
2049 ) -> Option<(Lrc<SourceFile>, usize, BytePos, usize, BytePos)>;
2050 fn hashing_controls(&self) -> HashingControls;
2053 impl<CTX> HashStable<CTX> for Span
2055 CTX: HashStableContext,
2057 /// Hashes a span in a stable way. We can't directly hash the span's `BytePos`
2058 /// fields (that would be similar to hashing pointers, since those are just
2059 /// offsets into the `SourceMap`). Instead, we hash the (file name, line, column)
2060 /// triple, which stays the same even if the containing `SourceFile` has moved
2061 /// within the `SourceMap`.
2063 /// Also note that we are hashing byte offsets for the column, not unicode
2064 /// codepoint offsets. For the purpose of the hash that's sufficient.
2065 /// Also, hashing filenames is expensive so we avoid doing it twice when the
2066 /// span starts and ends in the same file, which is almost always the case.
2067 fn hash_stable(&self, ctx: &mut CTX, hasher: &mut StableHasher) {
2068 const TAG_VALID_SPAN: u8 = 0;
2069 const TAG_INVALID_SPAN: u8 = 1;
2070 const TAG_RELATIVE_SPAN: u8 = 2;
2072 if !ctx.hash_spans() {
2076 let span = self.data_untracked();
2077 span.ctxt.hash_stable(ctx, hasher);
2078 span.parent.hash_stable(ctx, hasher);
2080 if span.is_dummy() {
2081 Hash::hash(&TAG_INVALID_SPAN, hasher);
2085 if let Some(parent) = span.parent {
2086 let def_span = ctx.def_span(parent).data_untracked();
2087 if def_span.contains(span) {
2088 // This span is enclosed in a definition: only hash the relative position.
2089 Hash::hash(&TAG_RELATIVE_SPAN, hasher);
2090 (span.lo - def_span.lo).to_u32().hash_stable(ctx, hasher);
2091 (span.hi - def_span.lo).to_u32().hash_stable(ctx, hasher);
2096 // If this is not an empty or invalid span, we want to hash the last
2097 // position that belongs to it, as opposed to hashing the first
2098 // position past it.
2099 let Some((file, line_lo, col_lo, line_hi, col_hi)) = ctx.span_data_to_lines_and_cols(&span) else {
2100 Hash::hash(&TAG_INVALID_SPAN, hasher);
2104 Hash::hash(&TAG_VALID_SPAN, hasher);
2105 // We truncate the stable ID hash and line and column numbers. The chances
2106 // of causing a collision this way should be minimal.
2107 Hash::hash(&(file.name_hash as u64), hasher);
2109 // Hash both the length and the end location (line/column) of a span. If we
2110 // hash only the length, for example, then two otherwise equal spans with
2111 // different end locations will have the same hash. This can cause a problem
2112 // during incremental compilation wherein a previous result for a query that
2113 // depends on the end location of a span will be incorrectly reused when the
2114 // end location of the span it depends on has changed (see issue #74890). A
2115 // similar analysis applies if some query depends specifically on the length
2116 // of the span, but we only hash the end location. So hash both.
2118 let col_lo_trunc = (col_lo.0 as u64) & 0xFF;
2119 let line_lo_trunc = ((line_lo as u64) & 0xFF_FF_FF) << 8;
2120 let col_hi_trunc = (col_hi.0 as u64) & 0xFF << 32;
2121 let line_hi_trunc = ((line_hi as u64) & 0xFF_FF_FF) << 40;
2122 let col_line = col_lo_trunc | line_lo_trunc | col_hi_trunc | line_hi_trunc;
2123 let len = (span.hi - span.lo).0;
2124 Hash::hash(&col_line, hasher);
2125 Hash::hash(&len, hasher);