1 //! Source positions and related helper functions.
3 //! Important concepts in this module include:
5 //! - the *span*, represented by [`SpanData`] and related types;
6 //! - source code as represented by a [`SourceMap`]; and
7 //! - interned strings, represented by [`Symbol`]s, with some common symbols available statically in the [`sym`] module.
9 //! Unlike most compilers, the span contains not only the position in the source code, but also various other metadata,
10 //! such as the edition and macro hygiene. This metadata is stored in [`SyntaxContext`] and [`ExpnData`].
14 //! This API is completely unstable and subject to change.
16 #![doc(html_root_url = "https://doc.rust-lang.org/nightly/nightly-rustc/")]
17 #![feature(array_windows)]
18 #![feature(bool_to_option)]
19 #![feature(crate_visibility_modifier)]
20 #![feature(if_let_guard)]
21 #![feature(negative_impls)]
23 #![feature(min_specialization)]
24 #![cfg_attr(not(bootstrap), allow(rustc::potential_query_instability))]
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;
60 use rustc_data_structures::stable_hasher::{HashStable, StableHasher};
61 use rustc_data_structures::sync::{Lock, Lrc};
64 use std::cmp::{self, Ordering};
67 use std::ops::{Add, Range, Sub};
68 use std::path::{Path, PathBuf};
69 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) -> Result<(), S::Error> {
196 encoder.emit_enum(|encoder| match *self {
197 RealFileName::LocalPath(ref local_path) => {
198 encoder.emit_enum_variant("LocalPath", 0, 1, |encoder| {
199 encoder.emit_enum_variant_arg(true, |encoder| local_path.encode(encoder))?;
204 RealFileName::Remapped { ref local_path, ref virtual_name } => encoder
205 .emit_enum_variant("Remapped", 1, 2, |encoder| {
206 // For privacy and build reproducibility, we must not embed host-dependant path in artifacts
207 // if they have been remapped by --remap-path-prefix
208 assert!(local_path.is_none());
209 encoder.emit_enum_variant_arg(true, |encoder| local_path.encode(encoder))?;
210 encoder.emit_enum_variant_arg(false, |encoder| virtual_name.encode(encoder))?;
218 /// Returns the path suitable for reading from the file system on the local host,
219 /// if this information exists.
220 /// Avoid embedding this in build artifacts; see `remapped_path_if_available()` for that.
221 pub fn local_path(&self) -> Option<&Path> {
223 RealFileName::LocalPath(p) => Some(p),
224 RealFileName::Remapped { local_path: p, virtual_name: _ } => {
225 p.as_ref().map(PathBuf::as_path)
230 /// Returns the path suitable for reading from the file system on the local host,
231 /// if this information exists.
232 /// Avoid embedding this in build artifacts; see `remapped_path_if_available()` for that.
233 pub fn into_local_path(self) -> Option<PathBuf> {
235 RealFileName::LocalPath(p) => Some(p),
236 RealFileName::Remapped { local_path: p, virtual_name: _ } => p,
240 /// Returns the path suitable for embedding into build artifacts. This would still
241 /// be a local path if it has not been remapped. A remapped path will not correspond
242 /// to a valid file system path: see `local_path_if_available()` for something that
243 /// is more likely to return paths into the local host file system.
244 pub fn remapped_path_if_available(&self) -> &Path {
246 RealFileName::LocalPath(p)
247 | RealFileName::Remapped { local_path: _, virtual_name: p } => &p,
251 /// Returns the path suitable for reading from the file system on the local host,
252 /// if this information exists. Otherwise returns the remapped name.
253 /// Avoid embedding this in build artifacts; see `remapped_path_if_available()` for that.
254 pub fn local_path_if_available(&self) -> &Path {
256 RealFileName::LocalPath(path)
257 | RealFileName::Remapped { local_path: None, virtual_name: path }
258 | RealFileName::Remapped { local_path: Some(path), virtual_name: _ } => path,
262 pub fn to_string_lossy(&self, display_pref: FileNameDisplayPreference) -> Cow<'_, str> {
264 FileNameDisplayPreference::Local => self.local_path_if_available().to_string_lossy(),
265 FileNameDisplayPreference::Remapped => {
266 self.remapped_path_if_available().to_string_lossy()
272 /// Differentiates between real files and common virtual files.
273 #[derive(Debug, Eq, PartialEq, Clone, Ord, PartialOrd, Hash)]
274 #[derive(Decodable, Encodable)]
277 /// Call to `quote!`.
281 /// Hack in `src/librustc_ast/parse.rs`.
284 ProcMacroSourceCode(u64),
285 /// Strings provided as `--cfg [cfgspec]` stored in a `crate_cfg`.
287 /// Strings provided as crate attributes in the CLI.
289 /// Custom sources for explicit parser calls from plugins and drivers.
291 DocTest(PathBuf, isize),
292 /// Post-substitution inline assembly from LLVM.
296 impl From<PathBuf> for FileName {
297 fn from(p: PathBuf) -> Self {
298 assert!(!p.to_string_lossy().ends_with('>'));
299 FileName::Real(RealFileName::LocalPath(p))
303 #[derive(Clone, Copy, Eq, PartialEq, Hash, Debug)]
304 pub enum FileNameDisplayPreference {
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 FileNameDisplay<'_> {
335 pub fn to_string_lossy(&self) -> Cow<'_, str> {
337 FileName::Real(ref inner) => inner.to_string_lossy(self.display_pref),
338 _ => Cow::from(format!("{}", self)),
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())
521 /// A collection of `Span`s.
523 /// Spans have two orthogonal attributes:
525 /// - They can be *primary spans*. In this case they are the locus of
526 /// the error, and would be rendered with `^^^`.
527 /// - They can have a *label*. In this case, the label is written next
528 /// to the mark in the snippet when we render.
529 #[derive(Clone, Debug, Hash, PartialEq, Eq, Encodable, Decodable)]
530 pub struct MultiSpan {
531 primary_spans: Vec<Span>,
532 span_labels: Vec<(Span, String)>,
537 pub fn lo(self) -> BytePos {
541 pub fn with_lo(self, lo: BytePos) -> Span {
542 self.data().with_lo(lo)
545 pub fn hi(self) -> BytePos {
549 pub fn with_hi(self, hi: BytePos) -> Span {
550 self.data().with_hi(hi)
553 pub fn ctxt(self) -> SyntaxContext {
554 self.data_untracked().ctxt
557 pub fn with_ctxt(self, ctxt: SyntaxContext) -> Span {
558 self.data_untracked().with_ctxt(ctxt)
561 pub fn parent(self) -> Option<LocalDefId> {
565 pub fn with_parent(self, ctxt: Option<LocalDefId>) -> Span {
566 self.data().with_parent(ctxt)
569 /// Returns `true` if this is a dummy span with any hygienic context.
571 pub fn is_dummy(self) -> bool {
572 self.data_untracked().is_dummy()
575 /// Returns `true` if this span comes from a macro or desugaring.
577 pub fn from_expansion(self) -> bool {
578 self.ctxt() != SyntaxContext::root()
581 /// Returns `true` if `span` originates in a derive-macro's expansion.
582 pub fn in_derive_expansion(self) -> bool {
583 matches!(self.ctxt().outer_expn_data().kind, ExpnKind::Macro(MacroKind::Derive, _))
586 /// Gate suggestions that would not be appropriate in a context the user didn't write.
587 pub fn can_be_used_for_suggestions(self) -> bool {
588 !self.from_expansion()
589 // FIXME: If this span comes from a `derive` macro but it points at code the user wrote,
590 // the callsite span and the span will be pointing at different places. It also means that
591 // we can safely provide suggestions on this span.
592 || (matches!(self.ctxt().outer_expn_data().kind, ExpnKind::Macro(MacroKind::Derive, _))
593 && self.parent_callsite().map(|p| (p.lo(), p.hi())) != Some((self.lo(), self.hi())))
597 pub fn with_root_ctxt(lo: BytePos, hi: BytePos) -> Span {
598 Span::new(lo, hi, SyntaxContext::root(), None)
601 /// Returns a new span representing an empty span at the beginning of this span.
603 pub fn shrink_to_lo(self) -> Span {
604 let span = self.data_untracked();
605 span.with_hi(span.lo)
607 /// Returns a new span representing an empty span at the end of this span.
609 pub fn shrink_to_hi(self) -> Span {
610 let span = self.data_untracked();
611 span.with_lo(span.hi)
615 /// Returns `true` if `hi == lo`.
616 pub fn is_empty(self) -> bool {
617 let span = self.data_untracked();
621 /// Returns `self` if `self` is not the dummy span, and `other` otherwise.
622 pub fn substitute_dummy(self, other: Span) -> Span {
623 if self.is_dummy() { other } else { self }
626 /// Returns `true` if `self` fully encloses `other`.
627 pub fn contains(self, other: Span) -> bool {
628 let span = self.data();
629 let other = other.data();
633 /// Returns `true` if `self` touches `other`.
634 pub fn overlaps(self, other: Span) -> bool {
635 let span = self.data();
636 let other = other.data();
637 span.lo < other.hi && other.lo < span.hi
640 /// Returns `true` if the spans are equal with regards to the source text.
642 /// Use this instead of `==` when either span could be generated code,
643 /// and you only care that they point to the same bytes of source text.
644 pub fn source_equal(self, other: Span) -> bool {
645 let span = self.data();
646 let other = other.data();
647 span.lo == other.lo && span.hi == other.hi
650 /// Returns `Some(span)`, where the start is trimmed by the end of `other`.
651 pub fn trim_start(self, other: Span) -> Option<Span> {
652 let span = self.data();
653 let other = other.data();
654 if span.hi > other.hi { Some(span.with_lo(cmp::max(span.lo, other.hi))) } else { None }
657 /// Returns the source span -- this is either the supplied span, or the span for
658 /// the macro callsite that expanded to it.
659 pub fn source_callsite(self) -> Span {
660 let expn_data = self.ctxt().outer_expn_data();
661 if !expn_data.is_root() { expn_data.call_site.source_callsite() } else { self }
664 /// The `Span` for the tokens in the previous macro expansion from which `self` was generated,
666 pub fn parent_callsite(self) -> Option<Span> {
667 let expn_data = self.ctxt().outer_expn_data();
668 if !expn_data.is_root() { Some(expn_data.call_site) } else { None }
671 /// Walk down the expansion ancestors to find a span that's contained within `outer`.
672 pub fn find_ancestor_inside(mut self, outer: Span) -> Option<Span> {
673 while !outer.contains(self) {
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
699 /// Returns the source callee.
701 /// Returns `None` if the supplied span has no expansion trace,
702 /// else returns the `ExpnData` for the macro definition
703 /// corresponding to the source callsite.
704 pub fn source_callee(self) -> Option<ExpnData> {
705 fn source_callee(expn_data: ExpnData) -> ExpnData {
706 let next_expn_data = expn_data.call_site.ctxt().outer_expn_data();
707 if !next_expn_data.is_root() { source_callee(next_expn_data) } else { expn_data }
709 let expn_data = self.ctxt().outer_expn_data();
710 if !expn_data.is_root() { Some(source_callee(expn_data)) } else { None }
713 /// Checks if a span is "internal" to a macro in which `#[unstable]`
714 /// items can be used (that is, a macro marked with
715 /// `#[allow_internal_unstable]`).
716 pub fn allows_unstable(self, feature: Symbol) -> bool {
719 .allow_internal_unstable
720 .map_or(false, |features| features.iter().any(|&f| f == feature))
723 /// Checks if this span arises from a compiler desugaring of kind `kind`.
724 pub fn is_desugaring(self, kind: DesugaringKind) -> bool {
725 match self.ctxt().outer_expn_data().kind {
726 ExpnKind::Desugaring(k) => k == kind,
731 /// Returns the compiler desugaring that created this span, or `None`
732 /// if this span is not from a desugaring.
733 pub fn desugaring_kind(self) -> Option<DesugaringKind> {
734 match self.ctxt().outer_expn_data().kind {
735 ExpnKind::Desugaring(k) => Some(k),
740 /// Checks if a span is "internal" to a macro in which `unsafe`
741 /// can be used without triggering the `unsafe_code` lint.
742 // (that is, a macro marked with `#[allow_internal_unsafe]`).
743 pub fn allows_unsafe(self) -> bool {
744 self.ctxt().outer_expn_data().allow_internal_unsafe
747 pub fn macro_backtrace(mut self) -> impl Iterator<Item = ExpnData> {
748 let mut prev_span = DUMMY_SP;
749 std::iter::from_fn(move || {
751 let expn_data = self.ctxt().outer_expn_data();
752 if expn_data.is_root() {
756 let is_recursive = expn_data.call_site.source_equal(prev_span);
759 self = expn_data.call_site;
761 // Don't print recursive invocations.
763 return Some(expn_data);
769 /// Returns a `Span` that would enclose both `self` and `end`.
773 /// self lorem ipsum end
774 /// ^^^^^^^^^^^^^^^^^^^^
776 pub fn to(self, end: Span) -> Span {
777 let span_data = self.data();
778 let end_data = end.data();
779 // FIXME(jseyfried): `self.ctxt` should always equal `end.ctxt` here (cf. issue #23480).
780 // Return the macro span on its own to avoid weird diagnostic output. It is preferable to
781 // have an incomplete span than a completely nonsensical one.
782 if span_data.ctxt != end_data.ctxt {
783 if span_data.ctxt == SyntaxContext::root() {
785 } else if end_data.ctxt == SyntaxContext::root() {
788 // Both spans fall within a macro.
789 // FIXME(estebank): check if it is the *same* macro.
792 cmp::min(span_data.lo, end_data.lo),
793 cmp::max(span_data.hi, end_data.hi),
794 if span_data.ctxt == SyntaxContext::root() { end_data.ctxt } else { span_data.ctxt },
795 if span_data.parent == end_data.parent { span_data.parent } else { None },
799 /// Returns a `Span` between the end of `self` to the beginning of `end`.
803 /// self lorem ipsum end
806 pub fn between(self, end: Span) -> Span {
807 let span = self.data();
808 let end = end.data();
812 if end.ctxt == SyntaxContext::root() { end.ctxt } else { span.ctxt },
813 if span.parent == end.parent { span.parent } else { None },
817 /// Returns a `Span` from the beginning of `self` until the beginning of `end`.
821 /// self lorem ipsum end
822 /// ^^^^^^^^^^^^^^^^^
824 pub fn until(self, end: Span) -> Span {
825 // Most of this function's body is copied from `to`.
826 // We can't just do `self.to(end.shrink_to_lo())`,
827 // because to also does some magic where it uses min/max so
828 // it can handle overlapping spans. Some advanced mis-use of
829 // `until` with different ctxts makes this visible.
830 let span_data = self.data();
831 let end_data = end.data();
832 // FIXME(jseyfried): `self.ctxt` should always equal `end.ctxt` here (cf. issue #23480).
833 // Return the macro span on its own to avoid weird diagnostic output. It is preferable to
834 // have an incomplete span than a completely nonsensical one.
835 if span_data.ctxt != end_data.ctxt {
836 if span_data.ctxt == SyntaxContext::root() {
838 } else if end_data.ctxt == SyntaxContext::root() {
841 // Both spans fall within a macro.
842 // FIXME(estebank): check if it is the *same* macro.
847 if end_data.ctxt == SyntaxContext::root() { end_data.ctxt } else { span_data.ctxt },
848 if span_data.parent == end_data.parent { span_data.parent } else { None },
852 pub fn from_inner(self, inner: InnerSpan) -> Span {
853 let span = self.data();
855 span.lo + BytePos::from_usize(inner.start),
856 span.lo + BytePos::from_usize(inner.end),
862 /// Equivalent of `Span::def_site` from the proc macro API,
863 /// except that the location is taken from the `self` span.
864 pub fn with_def_site_ctxt(self, expn_id: ExpnId) -> Span {
865 self.with_ctxt_from_mark(expn_id, Transparency::Opaque)
868 /// Equivalent of `Span::call_site` from the proc macro API,
869 /// except that the location is taken from the `self` span.
870 pub fn with_call_site_ctxt(self, expn_id: ExpnId) -> Span {
871 self.with_ctxt_from_mark(expn_id, Transparency::Transparent)
874 /// Equivalent of `Span::mixed_site` from the proc macro API,
875 /// except that the location is taken from the `self` span.
876 pub fn with_mixed_site_ctxt(self, expn_id: ExpnId) -> Span {
877 self.with_ctxt_from_mark(expn_id, Transparency::SemiTransparent)
880 /// Produces a span with the same location as `self` and context produced by a macro with the
881 /// given ID and transparency, assuming that macro was defined directly and not produced by
882 /// some other macro (which is the case for built-in and procedural macros).
883 pub fn with_ctxt_from_mark(self, expn_id: ExpnId, transparency: Transparency) -> Span {
884 self.with_ctxt(SyntaxContext::root().apply_mark(expn_id, transparency))
888 pub fn apply_mark(self, expn_id: ExpnId, transparency: Transparency) -> Span {
889 let span = self.data();
890 span.with_ctxt(span.ctxt.apply_mark(expn_id, transparency))
894 pub fn remove_mark(&mut self) -> ExpnId {
895 let mut span = self.data();
896 let mark = span.ctxt.remove_mark();
897 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
902 pub fn adjust(&mut self, expn_id: ExpnId) -> Option<ExpnId> {
903 let mut span = self.data();
904 let mark = span.ctxt.adjust(expn_id);
905 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
910 pub fn normalize_to_macros_2_0_and_adjust(&mut self, expn_id: ExpnId) -> Option<ExpnId> {
911 let mut span = self.data();
912 let mark = span.ctxt.normalize_to_macros_2_0_and_adjust(expn_id);
913 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
918 pub fn glob_adjust(&mut self, expn_id: ExpnId, glob_span: Span) -> Option<Option<ExpnId>> {
919 let mut span = self.data();
920 let mark = span.ctxt.glob_adjust(expn_id, glob_span);
921 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
926 pub fn reverse_glob_adjust(
930 ) -> Option<Option<ExpnId>> {
931 let mut span = self.data();
932 let mark = span.ctxt.reverse_glob_adjust(expn_id, glob_span);
933 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
938 pub fn normalize_to_macros_2_0(self) -> Span {
939 let span = self.data();
940 span.with_ctxt(span.ctxt.normalize_to_macros_2_0())
944 pub fn normalize_to_macro_rules(self) -> Span {
945 let span = self.data();
946 span.with_ctxt(span.ctxt.normalize_to_macro_rules())
950 /// A span together with some additional data.
951 #[derive(Clone, Debug)]
952 pub struct SpanLabel {
953 /// The span we are going to include in the final snippet.
956 /// Is this a primary span? This is the "locus" of the message,
957 /// and is indicated with a `^^^^` underline, versus `----`.
958 pub is_primary: bool,
960 /// What label should we attach to this span (if any)?
961 pub label: Option<String>,
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) -> Result<(), E::Error> {
972 let span = self.data();
973 s.emit_struct(false, |s| {
974 s.emit_struct_field("lo", true, |s| span.lo.encode(s))?;
975 s.emit_struct_field("hi", false, |s| span.hi.encode(s))
979 impl<D: Decoder> Decodable<D> for Span {
980 default fn decode(s: &mut D) -> Span {
982 let lo = d.read_struct_field("lo", Decodable::decode);
983 let hi = d.read_struct_field("hi", Decodable::decode);
985 Span::new(lo, hi, SyntaxContext::root(), None)
990 /// Calls the provided closure, using the provided `SourceMap` to format
991 /// any spans that are debug-printed during the closure's execution.
993 /// Normally, the global `TyCtxt` is used to retrieve the `SourceMap`
994 /// (see `rustc_interface::callbacks::span_debug1`). However, some parts
995 /// of the compiler (e.g. `rustc_parse`) may debug-print `Span`s before
996 /// a `TyCtxt` is available. In this case, we fall back to
997 /// the `SourceMap` provided to this function. If that is not available,
998 /// we fall back to printing the raw `Span` field values.
999 pub fn with_source_map<T, F: FnOnce() -> T>(source_map: Lrc<SourceMap>, f: F) -> T {
1000 with_session_globals(|session_globals| {
1001 *session_globals.source_map.borrow_mut() = Some(source_map);
1003 struct ClearSourceMap;
1004 impl Drop for ClearSourceMap {
1005 fn drop(&mut self) {
1006 with_session_globals(|session_globals| {
1007 session_globals.source_map.borrow_mut().take();
1012 let _guard = ClearSourceMap;
1016 pub fn debug_with_source_map(
1018 f: &mut fmt::Formatter<'_>,
1019 source_map: &SourceMap,
1021 write!(f, "{} ({:?})", source_map.span_to_diagnostic_string(span), span.ctxt())
1024 pub fn default_span_debug(span: Span, f: &mut fmt::Formatter<'_>) -> fmt::Result {
1025 with_session_globals(|session_globals| {
1026 if let Some(source_map) = &*session_globals.source_map.borrow() {
1027 debug_with_source_map(span, f, source_map)
1029 f.debug_struct("Span")
1030 .field("lo", &span.lo())
1031 .field("hi", &span.hi())
1032 .field("ctxt", &span.ctxt())
1038 impl fmt::Debug for Span {
1039 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
1040 (*SPAN_DEBUG)(*self, f)
1044 impl fmt::Debug for SpanData {
1045 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
1046 (*SPAN_DEBUG)(Span::new(self.lo, self.hi, self.ctxt, self.parent), f)
1052 pub fn new() -> MultiSpan {
1053 MultiSpan { primary_spans: vec![], span_labels: vec![] }
1056 pub fn from_span(primary_span: Span) -> MultiSpan {
1057 MultiSpan { primary_spans: vec![primary_span], span_labels: vec![] }
1060 pub fn from_spans(mut vec: Vec<Span>) -> MultiSpan {
1062 MultiSpan { primary_spans: vec, span_labels: vec![] }
1065 pub fn push_span_label(&mut self, span: Span, label: String) {
1066 self.span_labels.push((span, label));
1069 /// Selects the first primary span (if any).
1070 pub fn primary_span(&self) -> Option<Span> {
1071 self.primary_spans.first().cloned()
1074 /// Returns all primary spans.
1075 pub fn primary_spans(&self) -> &[Span] {
1079 /// Returns `true` if any of the primary spans are displayable.
1080 pub fn has_primary_spans(&self) -> bool {
1081 self.primary_spans.iter().any(|sp| !sp.is_dummy())
1084 /// Returns `true` if this contains only a dummy primary span with any hygienic context.
1085 pub fn is_dummy(&self) -> bool {
1086 let mut is_dummy = true;
1087 for span in &self.primary_spans {
1088 if !span.is_dummy() {
1095 /// Replaces all occurrences of one Span with another. Used to move `Span`s in areas that don't
1096 /// display well (like std macros). Returns whether replacements occurred.
1097 pub fn replace(&mut self, before: Span, after: Span) -> bool {
1098 let mut replacements_occurred = false;
1099 for primary_span in &mut self.primary_spans {
1100 if *primary_span == before {
1101 *primary_span = after;
1102 replacements_occurred = true;
1105 for span_label in &mut self.span_labels {
1106 if span_label.0 == before {
1107 span_label.0 = after;
1108 replacements_occurred = true;
1111 replacements_occurred
1114 /// Returns the strings to highlight. We always ensure that there
1115 /// is an entry for each of the primary spans -- for each primary
1116 /// span `P`, if there is at least one label with span `P`, we return
1117 /// those labels (marked as primary). But otherwise we return
1118 /// `SpanLabel` instances with empty labels.
1119 pub fn span_labels(&self) -> Vec<SpanLabel> {
1120 let is_primary = |span| self.primary_spans.contains(&span);
1122 let mut span_labels = self
1125 .map(|&(span, ref label)| SpanLabel {
1127 is_primary: is_primary(span),
1128 label: Some(label.clone()),
1130 .collect::<Vec<_>>();
1132 for &span in &self.primary_spans {
1133 if !span_labels.iter().any(|sl| sl.span == span) {
1134 span_labels.push(SpanLabel { span, is_primary: true, label: None });
1141 /// Returns `true` if any of the span labels is displayable.
1142 pub fn has_span_labels(&self) -> bool {
1143 self.span_labels.iter().any(|(sp, _)| !sp.is_dummy())
1147 impl From<Span> for MultiSpan {
1148 fn from(span: Span) -> MultiSpan {
1149 MultiSpan::from_span(span)
1153 impl From<Vec<Span>> for MultiSpan {
1154 fn from(spans: Vec<Span>) -> MultiSpan {
1155 MultiSpan::from_spans(spans)
1159 /// Identifies an offset of a multi-byte character in a `SourceFile`.
1160 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1161 pub struct MultiByteChar {
1162 /// The absolute offset of the character in the `SourceMap`.
1164 /// The number of bytes, `>= 2`.
1168 /// Identifies an offset of a non-narrow character in a `SourceFile`.
1169 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1170 pub enum NonNarrowChar {
1171 /// Represents a zero-width character.
1173 /// Represents a wide (full-width) character.
1175 /// Represents a tab character, represented visually with a width of 4 characters.
1179 impl NonNarrowChar {
1180 fn new(pos: BytePos, width: usize) -> Self {
1182 0 => NonNarrowChar::ZeroWidth(pos),
1183 2 => NonNarrowChar::Wide(pos),
1184 4 => NonNarrowChar::Tab(pos),
1185 _ => panic!("width {} given for non-narrow character", width),
1189 /// Returns the absolute offset of the character in the `SourceMap`.
1190 pub fn pos(&self) -> BytePos {
1192 NonNarrowChar::ZeroWidth(p) | NonNarrowChar::Wide(p) | NonNarrowChar::Tab(p) => p,
1196 /// Returns the width of the character, 0 (zero-width) or 2 (wide).
1197 pub fn width(&self) -> usize {
1199 NonNarrowChar::ZeroWidth(_) => 0,
1200 NonNarrowChar::Wide(_) => 2,
1201 NonNarrowChar::Tab(_) => 4,
1206 impl Add<BytePos> for NonNarrowChar {
1209 fn add(self, rhs: BytePos) -> Self {
1211 NonNarrowChar::ZeroWidth(pos) => NonNarrowChar::ZeroWidth(pos + rhs),
1212 NonNarrowChar::Wide(pos) => NonNarrowChar::Wide(pos + rhs),
1213 NonNarrowChar::Tab(pos) => NonNarrowChar::Tab(pos + rhs),
1218 impl Sub<BytePos> for NonNarrowChar {
1221 fn sub(self, rhs: BytePos) -> Self {
1223 NonNarrowChar::ZeroWidth(pos) => NonNarrowChar::ZeroWidth(pos - rhs),
1224 NonNarrowChar::Wide(pos) => NonNarrowChar::Wide(pos - rhs),
1225 NonNarrowChar::Tab(pos) => NonNarrowChar::Tab(pos - rhs),
1230 /// Identifies an offset of a character that was normalized away from `SourceFile`.
1231 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1232 pub struct NormalizedPos {
1233 /// The absolute offset of the character in the `SourceMap`.
1235 /// The difference between original and normalized string at position.
1239 #[derive(PartialEq, Eq, Clone, Debug)]
1240 pub enum ExternalSource {
1241 /// No external source has to be loaded, since the `SourceFile` represents a local crate.
1244 kind: ExternalSourceKind,
1245 /// This SourceFile's byte-offset within the source_map of its original crate.
1246 original_start_pos: BytePos,
1247 /// The end of this SourceFile within the source_map of its original crate.
1248 original_end_pos: BytePos,
1252 /// The state of the lazy external source loading mechanism of a `SourceFile`.
1253 #[derive(PartialEq, Eq, Clone, Debug)]
1254 pub enum ExternalSourceKind {
1255 /// The external source has been loaded already.
1256 Present(Lrc<String>),
1257 /// No attempt has been made to load the external source.
1259 /// A failed attempt has been made to load the external source.
1264 impl ExternalSource {
1265 pub fn get_source(&self) -> Option<&Lrc<String>> {
1267 ExternalSource::Foreign { kind: ExternalSourceKind::Present(ref src), .. } => Some(src),
1274 pub struct OffsetOverflowError;
1276 #[derive(Copy, Clone, Debug, PartialEq, Eq, PartialOrd, Ord, Hash, Encodable, Decodable)]
1277 pub enum SourceFileHashAlgorithm {
1283 impl FromStr for SourceFileHashAlgorithm {
1286 fn from_str(s: &str) -> Result<SourceFileHashAlgorithm, ()> {
1288 "md5" => Ok(SourceFileHashAlgorithm::Md5),
1289 "sha1" => Ok(SourceFileHashAlgorithm::Sha1),
1290 "sha256" => Ok(SourceFileHashAlgorithm::Sha256),
1296 rustc_data_structures::impl_stable_hash_via_hash!(SourceFileHashAlgorithm);
1298 /// The hash of the on-disk source file used for debug info.
1299 #[derive(Copy, Clone, PartialEq, Eq, Debug)]
1300 #[derive(HashStable_Generic, Encodable, Decodable)]
1301 pub struct SourceFileHash {
1302 pub kind: SourceFileHashAlgorithm,
1306 impl SourceFileHash {
1307 pub fn new(kind: SourceFileHashAlgorithm, src: &str) -> SourceFileHash {
1308 let mut hash = SourceFileHash { kind, value: Default::default() };
1309 let len = hash.hash_len();
1310 let value = &mut hash.value[..len];
1311 let data = src.as_bytes();
1313 SourceFileHashAlgorithm::Md5 => {
1314 value.copy_from_slice(&Md5::digest(data));
1316 SourceFileHashAlgorithm::Sha1 => {
1317 value.copy_from_slice(&Sha1::digest(data));
1319 SourceFileHashAlgorithm::Sha256 => {
1320 value.copy_from_slice(&Sha256::digest(data));
1326 /// Check if the stored hash matches the hash of the string.
1327 pub fn matches(&self, src: &str) -> bool {
1328 Self::new(self.kind, src) == *self
1331 /// The bytes of the hash.
1332 pub fn hash_bytes(&self) -> &[u8] {
1333 let len = self.hash_len();
1337 fn hash_len(&self) -> usize {
1339 SourceFileHashAlgorithm::Md5 => 16,
1340 SourceFileHashAlgorithm::Sha1 => 20,
1341 SourceFileHashAlgorithm::Sha256 => 32,
1346 /// A single source in the [`SourceMap`].
1348 pub struct SourceFile {
1349 /// The name of the file that the source came from. Source that doesn't
1350 /// originate from files has names between angle brackets by convention
1351 /// (e.g., `<anon>`).
1353 /// The complete source code.
1354 pub src: Option<Lrc<String>>,
1355 /// The source code's hash.
1356 pub src_hash: SourceFileHash,
1357 /// The external source code (used for external crates, which will have a `None`
1358 /// value as `self.src`.
1359 pub external_src: Lock<ExternalSource>,
1360 /// The start position of this source in the `SourceMap`.
1361 pub start_pos: BytePos,
1362 /// The end position of this source in the `SourceMap`.
1363 pub end_pos: BytePos,
1364 /// Locations of lines beginnings in the source code.
1365 pub lines: Vec<BytePos>,
1366 /// Locations of multi-byte characters in the source code.
1367 pub multibyte_chars: Vec<MultiByteChar>,
1368 /// Width of characters that are not narrow in the source code.
1369 pub non_narrow_chars: Vec<NonNarrowChar>,
1370 /// Locations of characters removed during normalization.
1371 pub normalized_pos: Vec<NormalizedPos>,
1372 /// A hash of the filename, used for speeding up hashing in incremental compilation.
1373 pub name_hash: u128,
1374 /// Indicates which crate this `SourceFile` was imported from.
1378 impl<S: Encoder> Encodable<S> for SourceFile {
1379 fn encode(&self, s: &mut S) -> Result<(), S::Error> {
1380 s.emit_struct(false, |s| {
1381 s.emit_struct_field("name", true, |s| self.name.encode(s))?;
1382 s.emit_struct_field("src_hash", false, |s| self.src_hash.encode(s))?;
1383 s.emit_struct_field("start_pos", false, |s| self.start_pos.encode(s))?;
1384 s.emit_struct_field("end_pos", false, |s| self.end_pos.encode(s))?;
1385 s.emit_struct_field("lines", false, |s| {
1386 let lines = &self.lines[..];
1387 // Store the length.
1388 s.emit_u32(lines.len() as u32)?;
1390 if !lines.is_empty() {
1391 // In order to preserve some space, we exploit the fact that
1392 // the lines list is sorted and individual lines are
1393 // probably not that long. Because of that we can store lines
1394 // as a difference list, using as little space as possible
1395 // for the differences.
1396 let max_line_length = if lines.len() == 1 {
1401 .map(|&[fst, snd]| snd - fst)
1402 .map(|bp| bp.to_usize())
1407 let bytes_per_diff: u8 = match max_line_length {
1409 0x100..=0xFFFF => 2,
1413 // Encode the number of bytes used per diff.
1414 bytes_per_diff.encode(s)?;
1416 // Encode the first element.
1417 lines[0].encode(s)?;
1419 let diff_iter = lines.array_windows().map(|&[fst, snd]| snd - fst);
1421 match bytes_per_diff {
1423 for diff in diff_iter {
1424 (diff.0 as u8).encode(s)?
1428 for diff in diff_iter {
1429 (diff.0 as u16).encode(s)?
1433 for diff in diff_iter {
1437 _ => unreachable!(),
1443 s.emit_struct_field("multibyte_chars", false, |s| self.multibyte_chars.encode(s))?;
1444 s.emit_struct_field("non_narrow_chars", false, |s| self.non_narrow_chars.encode(s))?;
1445 s.emit_struct_field("name_hash", false, |s| self.name_hash.encode(s))?;
1446 s.emit_struct_field("normalized_pos", false, |s| self.normalized_pos.encode(s))?;
1447 s.emit_struct_field("cnum", false, |s| self.cnum.encode(s))
1452 impl<D: Decoder> Decodable<D> for SourceFile {
1453 fn decode(d: &mut D) -> SourceFile {
1455 let name: FileName = d.read_struct_field("name", |d| Decodable::decode(d));
1456 let src_hash: SourceFileHash =
1457 d.read_struct_field("src_hash", |d| Decodable::decode(d));
1458 let start_pos: BytePos = d.read_struct_field("start_pos", |d| Decodable::decode(d));
1459 let end_pos: BytePos = d.read_struct_field("end_pos", |d| Decodable::decode(d));
1460 let lines: Vec<BytePos> = d.read_struct_field("lines", |d| {
1461 let num_lines: u32 = Decodable::decode(d);
1462 let mut lines = Vec::with_capacity(num_lines as usize);
1465 // Read the number of bytes used per diff.
1466 let bytes_per_diff: u8 = Decodable::decode(d);
1468 // Read the first element.
1469 let mut line_start: BytePos = Decodable::decode(d);
1470 lines.push(line_start);
1472 for _ in 1..num_lines {
1473 let diff = match bytes_per_diff {
1474 1 => d.read_u8() as u32,
1475 2 => d.read_u16() as u32,
1477 _ => unreachable!(),
1480 line_start = line_start + BytePos(diff);
1482 lines.push(line_start);
1488 let multibyte_chars: Vec<MultiByteChar> =
1489 d.read_struct_field("multibyte_chars", |d| Decodable::decode(d));
1490 let non_narrow_chars: Vec<NonNarrowChar> =
1491 d.read_struct_field("non_narrow_chars", |d| Decodable::decode(d));
1492 let name_hash: u128 = d.read_struct_field("name_hash", |d| Decodable::decode(d));
1493 let normalized_pos: Vec<NormalizedPos> =
1494 d.read_struct_field("normalized_pos", |d| Decodable::decode(d));
1495 let cnum: CrateNum = d.read_struct_field("cnum", |d| Decodable::decode(d));
1502 // Unused - the metadata decoder will construct
1503 // a new SourceFile, filling in `external_src` properly
1504 external_src: Lock::new(ExternalSource::Unneeded),
1516 impl fmt::Debug for SourceFile {
1517 fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result {
1518 write!(fmt, "SourceFile({:?})", self.name)
1527 hash_kind: SourceFileHashAlgorithm,
1529 // Compute the file hash before any normalization.
1530 let src_hash = SourceFileHash::new(hash_kind, &src);
1531 let normalized_pos = normalize_src(&mut src, start_pos);
1534 let mut hasher: StableHasher = StableHasher::new();
1535 name.hash(&mut hasher);
1536 hasher.finish::<u128>()
1538 let end_pos = start_pos.to_usize() + src.len();
1539 assert!(end_pos <= u32::MAX as usize);
1541 let (lines, multibyte_chars, non_narrow_chars) =
1542 analyze_source_file::analyze_source_file(&src, start_pos);
1546 src: Some(Lrc::new(src)),
1548 external_src: Lock::new(ExternalSource::Unneeded),
1550 end_pos: Pos::from_usize(end_pos),
1560 /// Returns the `BytePos` of the beginning of the current line.
1561 pub fn line_begin_pos(&self, pos: BytePos) -> BytePos {
1562 let line_index = self.lookup_line(pos).unwrap();
1563 self.lines[line_index]
1566 /// Add externally loaded source.
1567 /// If the hash of the input doesn't match or no input is supplied via None,
1568 /// it is interpreted as an error and the corresponding enum variant is set.
1569 /// The return value signifies whether some kind of source is present.
1570 pub fn add_external_src<F>(&self, get_src: F) -> bool
1572 F: FnOnce() -> Option<String>,
1575 *self.external_src.borrow(),
1576 ExternalSource::Foreign { kind: ExternalSourceKind::AbsentOk, .. }
1578 let src = get_src();
1579 let mut external_src = self.external_src.borrow_mut();
1580 // Check that no-one else have provided the source while we were getting it
1581 if let ExternalSource::Foreign {
1582 kind: src_kind @ ExternalSourceKind::AbsentOk, ..
1583 } = &mut *external_src
1585 if let Some(mut src) = src {
1586 // The src_hash needs to be computed on the pre-normalized src.
1587 if self.src_hash.matches(&src) {
1588 normalize_src(&mut src, BytePos::from_usize(0));
1589 *src_kind = ExternalSourceKind::Present(Lrc::new(src));
1593 *src_kind = ExternalSourceKind::AbsentErr;
1598 self.src.is_some() || external_src.get_source().is_some()
1601 self.src.is_some() || self.external_src.borrow().get_source().is_some()
1605 /// Gets a line from the list of pre-computed line-beginnings.
1606 /// The line number here is 0-based.
1607 pub fn get_line(&self, line_number: usize) -> Option<Cow<'_, str>> {
1608 fn get_until_newline(src: &str, begin: usize) -> &str {
1609 // We can't use `lines.get(line_number+1)` because we might
1610 // be parsing when we call this function and thus the current
1611 // line is the last one we have line info for.
1612 let slice = &src[begin..];
1613 match slice.find('\n') {
1614 Some(e) => &slice[..e],
1620 let line = self.lines.get(line_number)?;
1621 let begin: BytePos = *line - self.start_pos;
1625 if let Some(ref src) = self.src {
1626 Some(Cow::from(get_until_newline(src, begin)))
1627 } else if let Some(src) = self.external_src.borrow().get_source() {
1628 Some(Cow::Owned(String::from(get_until_newline(src, begin))))
1634 pub fn is_real_file(&self) -> bool {
1638 pub fn is_imported(&self) -> bool {
1642 pub fn count_lines(&self) -> usize {
1646 /// Finds the line containing the given position. The return value is the
1647 /// index into the `lines` array of this `SourceFile`, not the 1-based line
1648 /// number. If the source_file is empty or the position is located before the
1649 /// first line, `None` is returned.
1650 pub fn lookup_line(&self, pos: BytePos) -> Option<usize> {
1651 match self.lines.binary_search(&pos) {
1652 Ok(idx) => Some(idx),
1654 Err(idx) => Some(idx - 1),
1658 pub fn line_bounds(&self, line_index: usize) -> Range<BytePos> {
1659 if self.is_empty() {
1660 return self.start_pos..self.end_pos;
1663 assert!(line_index < self.lines.len());
1664 if line_index == (self.lines.len() - 1) {
1665 self.lines[line_index]..self.end_pos
1667 self.lines[line_index]..self.lines[line_index + 1]
1671 /// Returns whether or not the file contains the given `SourceMap` byte
1672 /// position. The position one past the end of the file is considered to be
1673 /// contained by the file. This implies that files for which `is_empty`
1674 /// returns true still contain one byte position according to this function.
1676 pub fn contains(&self, byte_pos: BytePos) -> bool {
1677 byte_pos >= self.start_pos && byte_pos <= self.end_pos
1681 pub fn is_empty(&self) -> bool {
1682 self.start_pos == self.end_pos
1685 /// Calculates the original byte position relative to the start of the file
1686 /// based on the given byte position.
1687 pub fn original_relative_byte_pos(&self, pos: BytePos) -> BytePos {
1688 // Diff before any records is 0. Otherwise use the previously recorded
1689 // diff as that applies to the following characters until a new diff
1691 let diff = match self.normalized_pos.binary_search_by(|np| np.pos.cmp(&pos)) {
1692 Ok(i) => self.normalized_pos[i].diff,
1693 Err(i) if i == 0 => 0,
1694 Err(i) => self.normalized_pos[i - 1].diff,
1697 BytePos::from_u32(pos.0 - self.start_pos.0 + diff)
1700 /// Converts an absolute `BytePos` to a `CharPos` relative to the `SourceFile`.
1701 pub fn bytepos_to_file_charpos(&self, bpos: BytePos) -> CharPos {
1702 // The number of extra bytes due to multibyte chars in the `SourceFile`.
1703 let mut total_extra_bytes = 0;
1705 for mbc in self.multibyte_chars.iter() {
1706 debug!("{}-byte char at {:?}", mbc.bytes, mbc.pos);
1708 // Every character is at least one byte, so we only
1709 // count the actual extra bytes.
1710 total_extra_bytes += mbc.bytes as u32 - 1;
1711 // We should never see a byte position in the middle of a
1713 assert!(bpos.to_u32() >= mbc.pos.to_u32() + mbc.bytes as u32);
1719 assert!(self.start_pos.to_u32() + total_extra_bytes <= bpos.to_u32());
1720 CharPos(bpos.to_usize() - self.start_pos.to_usize() - total_extra_bytes as usize)
1723 /// Looks up the file's (1-based) line number and (0-based `CharPos`) column offset, for a
1724 /// given `BytePos`.
1725 pub fn lookup_file_pos(&self, pos: BytePos) -> (usize, CharPos) {
1726 let chpos = self.bytepos_to_file_charpos(pos);
1727 match self.lookup_line(pos) {
1729 let line = a + 1; // Line numbers start at 1
1730 let linebpos = self.lines[a];
1731 let linechpos = self.bytepos_to_file_charpos(linebpos);
1732 let col = chpos - linechpos;
1733 debug!("byte pos {:?} is on the line at byte pos {:?}", pos, linebpos);
1734 debug!("char pos {:?} is on the line at char pos {:?}", chpos, linechpos);
1735 debug!("byte is on line: {}", line);
1736 assert!(chpos >= linechpos);
1743 /// Looks up the file's (1-based) line number, (0-based `CharPos`) column offset, and (0-based)
1744 /// column offset when displayed, for a given `BytePos`.
1745 pub fn lookup_file_pos_with_col_display(&self, pos: BytePos) -> (usize, CharPos, usize) {
1746 let (line, col_or_chpos) = self.lookup_file_pos(pos);
1748 let col = col_or_chpos;
1749 let linebpos = self.lines[line - 1];
1751 let start_width_idx = self
1753 .binary_search_by_key(&linebpos, |x| x.pos())
1754 .unwrap_or_else(|x| x);
1755 let end_width_idx = self
1757 .binary_search_by_key(&pos, |x| x.pos())
1758 .unwrap_or_else(|x| x);
1759 let special_chars = end_width_idx - start_width_idx;
1760 let non_narrow: usize = self.non_narrow_chars[start_width_idx..end_width_idx]
1764 col.0 - special_chars + non_narrow
1766 (line, col, col_display)
1768 let chpos = col_or_chpos;
1770 let end_width_idx = self
1772 .binary_search_by_key(&pos, |x| x.pos())
1773 .unwrap_or_else(|x| x);
1774 let non_narrow: usize =
1775 self.non_narrow_chars[0..end_width_idx].iter().map(|x| x.width()).sum();
1776 chpos.0 - end_width_idx + non_narrow
1778 (0, chpos, col_display)
1783 /// Normalizes the source code and records the normalizations.
1784 fn normalize_src(src: &mut String, start_pos: BytePos) -> Vec<NormalizedPos> {
1785 let mut normalized_pos = vec![];
1786 remove_bom(src, &mut normalized_pos);
1787 normalize_newlines(src, &mut normalized_pos);
1789 // Offset all the positions by start_pos to match the final file positions.
1790 for np in &mut normalized_pos {
1791 np.pos.0 += start_pos.0;
1797 /// Removes UTF-8 BOM, if any.
1798 fn remove_bom(src: &mut String, normalized_pos: &mut Vec<NormalizedPos>) {
1799 if src.starts_with('\u{feff}') {
1801 normalized_pos.push(NormalizedPos { pos: BytePos(0), diff: 3 });
1805 /// Replaces `\r\n` with `\n` in-place in `src`.
1807 /// Returns error if there's a lone `\r` in the string.
1808 fn normalize_newlines(src: &mut String, normalized_pos: &mut Vec<NormalizedPos>) {
1809 if !src.as_bytes().contains(&b'\r') {
1813 // We replace `\r\n` with `\n` in-place, which doesn't break utf-8 encoding.
1814 // While we *can* call `as_mut_vec` and do surgery on the live string
1815 // directly, let's rather steal the contents of `src`. This makes the code
1816 // safe even if a panic occurs.
1818 let mut buf = std::mem::replace(src, String::new()).into_bytes();
1819 let mut gap_len = 0;
1820 let mut tail = buf.as_mut_slice();
1822 let original_gap = normalized_pos.last().map_or(0, |l| l.diff);
1824 let idx = match find_crlf(&tail[gap_len..]) {
1826 Some(idx) => idx + gap_len,
1828 tail.copy_within(gap_len..idx, 0);
1829 tail = &mut tail[idx - gap_len..];
1830 if tail.len() == gap_len {
1833 cursor += idx - gap_len;
1835 normalized_pos.push(NormalizedPos {
1836 pos: BytePos::from_usize(cursor + 1),
1837 diff: original_gap + gap_len as u32,
1841 // Account for removed `\r`.
1842 // After `set_len`, `buf` is guaranteed to contain utf-8 again.
1843 let new_len = buf.len() - gap_len;
1845 buf.set_len(new_len);
1846 *src = String::from_utf8_unchecked(buf);
1849 fn find_crlf(src: &[u8]) -> Option<usize> {
1850 let mut search_idx = 0;
1851 while let Some(idx) = find_cr(&src[search_idx..]) {
1852 if src[search_idx..].get(idx + 1) != Some(&b'\n') {
1853 search_idx += idx + 1;
1856 return Some(search_idx + idx);
1861 fn find_cr(src: &[u8]) -> Option<usize> {
1862 src.iter().position(|&b| b == b'\r')
1866 // _____________________________________________________________________________
1867 // Pos, BytePos, CharPos
1871 fn from_usize(n: usize) -> Self;
1872 fn to_usize(&self) -> usize;
1873 fn from_u32(n: u32) -> Self;
1874 fn to_u32(&self) -> u32;
1877 macro_rules! impl_pos {
1881 $vis:vis struct $ident:ident($inner_vis:vis $inner_ty:ty);
1886 $vis struct $ident($inner_vis $inner_ty);
1888 impl Pos for $ident {
1890 fn from_usize(n: usize) -> $ident {
1891 $ident(n as $inner_ty)
1895 fn to_usize(&self) -> usize {
1900 fn from_u32(n: u32) -> $ident {
1901 $ident(n as $inner_ty)
1905 fn to_u32(&self) -> u32 {
1910 impl Add for $ident {
1911 type Output = $ident;
1914 fn add(self, rhs: $ident) -> $ident {
1915 $ident(self.0 + rhs.0)
1919 impl Sub for $ident {
1920 type Output = $ident;
1923 fn sub(self, rhs: $ident) -> $ident {
1924 $ident(self.0 - rhs.0)
1934 /// Keep this small (currently 32-bits), as AST contains a lot of them.
1935 #[derive(Clone, Copy, PartialEq, Eq, Hash, PartialOrd, Ord, Debug)]
1936 pub struct BytePos(pub u32);
1938 /// A character offset.
1940 /// Because of multibyte UTF-8 characters, a byte offset
1941 /// is not equivalent to a character offset. The [`SourceMap`] will convert [`BytePos`]
1942 /// values to `CharPos` values as necessary.
1943 #[derive(Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Debug)]
1944 pub struct CharPos(pub usize);
1947 impl<S: rustc_serialize::Encoder> Encodable<S> for BytePos {
1948 fn encode(&self, s: &mut S) -> Result<(), S::Error> {
1953 impl<D: rustc_serialize::Decoder> Decodable<D> for BytePos {
1954 fn decode(d: &mut D) -> BytePos {
1955 BytePos(d.read_u32())
1959 // _____________________________________________________________________________
1960 // Loc, SourceFileAndLine, SourceFileAndBytePos
1963 /// A source code location used for error reporting.
1964 #[derive(Debug, Clone)]
1966 /// Information about the original source.
1967 pub file: Lrc<SourceFile>,
1968 /// The (1-based) line number.
1970 /// The (0-based) column offset.
1972 /// The (0-based) column offset when displayed.
1973 pub col_display: usize,
1976 // Used to be structural records.
1978 pub struct SourceFileAndLine {
1979 pub sf: Lrc<SourceFile>,
1980 /// Index of line, starting from 0.
1984 pub struct SourceFileAndBytePos {
1985 pub sf: Lrc<SourceFile>,
1989 #[derive(Copy, Clone, Debug, PartialEq, Eq)]
1990 pub struct LineInfo {
1991 /// Index of line, starting from 0.
1992 pub line_index: usize,
1994 /// Column in line where span begins, starting from 0.
1995 pub start_col: CharPos,
1997 /// Column in line where span ends, starting from 0, exclusive.
1998 pub end_col: CharPos,
2001 pub struct FileLines {
2002 pub file: Lrc<SourceFile>,
2003 pub lines: Vec<LineInfo>,
2006 pub static SPAN_DEBUG: AtomicRef<fn(Span, &mut fmt::Formatter<'_>) -> fmt::Result> =
2007 AtomicRef::new(&(default_span_debug as fn(_, &mut fmt::Formatter<'_>) -> _));
2008 pub static SPAN_TRACK: AtomicRef<fn(LocalDefId)> = AtomicRef::new(&((|_| {}) as fn(_)));
2010 // _____________________________________________________________________________
2011 // SpanLinesError, SpanSnippetError, DistinctSources, MalformedSourceMapPositions
2014 pub type FileLinesResult = Result<FileLines, SpanLinesError>;
2016 #[derive(Clone, PartialEq, Eq, Debug)]
2017 pub enum SpanLinesError {
2018 DistinctSources(DistinctSources),
2021 #[derive(Clone, PartialEq, Eq, Debug)]
2022 pub enum SpanSnippetError {
2023 IllFormedSpan(Span),
2024 DistinctSources(DistinctSources),
2025 MalformedForSourcemap(MalformedSourceMapPositions),
2026 SourceNotAvailable { filename: FileName },
2029 #[derive(Clone, PartialEq, Eq, Debug)]
2030 pub struct DistinctSources {
2031 pub begin: (FileName, BytePos),
2032 pub end: (FileName, BytePos),
2035 #[derive(Clone, PartialEq, Eq, Debug)]
2036 pub struct MalformedSourceMapPositions {
2038 pub source_len: usize,
2039 pub begin_pos: BytePos,
2040 pub end_pos: BytePos,
2043 /// Range inside of a `Span` used for diagnostics when we only have access to relative positions.
2044 #[derive(Copy, Clone, PartialEq, Eq, Debug)]
2045 pub struct InnerSpan {
2051 pub fn new(start: usize, end: usize) -> InnerSpan {
2052 InnerSpan { start, end }
2056 /// Requirements for a `StableHashingContext` to be used in this crate.
2058 /// This is a hack to allow using the [`HashStable_Generic`] derive macro
2059 /// instead of implementing everything in rustc_middle.
2060 pub trait HashStableContext {
2061 fn def_path_hash(&self, def_id: DefId) -> DefPathHash;
2062 fn hash_spans(&self) -> bool;
2063 /// Accesses `sess.opts.debugging_opts.incremental_ignore_spans` since
2064 /// we don't have easy access to a `Session`
2065 fn debug_opts_incremental_ignore_spans(&self) -> bool;
2066 fn def_span(&self, def_id: LocalDefId) -> Span;
2067 fn span_data_to_lines_and_cols(
2070 ) -> Option<(Lrc<SourceFile>, usize, BytePos, usize, BytePos)>;
2071 fn hashing_controls(&self) -> HashingControls;
2074 impl<CTX> HashStable<CTX> for Span
2076 CTX: HashStableContext,
2078 /// Hashes a span in a stable way. We can't directly hash the span's `BytePos`
2079 /// fields (that would be similar to hashing pointers, since those are just
2080 /// offsets into the `SourceMap`). Instead, we hash the (file name, line, column)
2081 /// triple, which stays the same even if the containing `SourceFile` has moved
2082 /// within the `SourceMap`.
2084 /// Also note that we are hashing byte offsets for the column, not unicode
2085 /// codepoint offsets. For the purpose of the hash that's sufficient.
2086 /// Also, hashing filenames is expensive so we avoid doing it twice when the
2087 /// span starts and ends in the same file, which is almost always the case.
2088 fn hash_stable(&self, ctx: &mut CTX, hasher: &mut StableHasher) {
2089 const TAG_VALID_SPAN: u8 = 0;
2090 const TAG_INVALID_SPAN: u8 = 1;
2091 const TAG_RELATIVE_SPAN: u8 = 2;
2093 if !ctx.hash_spans() {
2097 let span = self.data_untracked();
2098 span.ctxt.hash_stable(ctx, hasher);
2099 span.parent.hash_stable(ctx, hasher);
2101 if span.is_dummy() {
2102 Hash::hash(&TAG_INVALID_SPAN, hasher);
2106 if let Some(parent) = span.parent {
2107 let def_span = ctx.def_span(parent).data_untracked();
2108 if def_span.contains(span) {
2109 // This span is enclosed in a definition: only hash the relative position.
2110 Hash::hash(&TAG_RELATIVE_SPAN, hasher);
2111 (span.lo - def_span.lo).to_u32().hash_stable(ctx, hasher);
2112 (span.hi - def_span.lo).to_u32().hash_stable(ctx, hasher);
2117 // If this is not an empty or invalid span, we want to hash the last
2118 // position that belongs to it, as opposed to hashing the first
2119 // position past it.
2120 let (file, line_lo, col_lo, line_hi, col_hi) = match ctx.span_data_to_lines_and_cols(&span)
2124 Hash::hash(&TAG_INVALID_SPAN, hasher);
2129 Hash::hash(&TAG_VALID_SPAN, hasher);
2130 // We truncate the stable ID hash and line and column numbers. The chances
2131 // of causing a collision this way should be minimal.
2132 Hash::hash(&(file.name_hash as u64), hasher);
2134 // Hash both the length and the end location (line/column) of a span. If we
2135 // hash only the length, for example, then two otherwise equal spans with
2136 // different end locations will have the same hash. This can cause a problem
2137 // during incremental compilation wherein a previous result for a query that
2138 // depends on the end location of a span will be incorrectly reused when the
2139 // end location of the span it depends on has changed (see issue #74890). A
2140 // similar analysis applies if some query depends specifically on the length
2141 // of the span, but we only hash the end location. So hash both.
2143 let col_lo_trunc = (col_lo.0 as u64) & 0xFF;
2144 let line_lo_trunc = ((line_lo as u64) & 0xFF_FF_FF) << 8;
2145 let col_hi_trunc = (col_hi.0 as u64) & 0xFF << 32;
2146 let line_hi_trunc = ((line_hi as u64) & 0xFF_FF_FF) << 40;
2147 let col_line = col_lo_trunc | line_lo_trunc | col_hi_trunc | line_hi_trunc;
2148 let len = (span.hi - span.lo).0;
2149 Hash::hash(&col_line, hasher);
2150 Hash::hash(&len, hasher);