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)]
26 extern crate rustc_macros;
31 use rustc_data_structures::AtomicRef;
32 use rustc_macros::HashStable_Generic;
33 use rustc_serialize::{Decodable, Decoder, Encodable, Encoder};
35 mod caching_source_map_view;
37 pub use self::caching_source_map_view::CachingSourceMapView;
38 use source_map::SourceMap;
43 use hygiene::Transparency;
44 pub use hygiene::{DesugaringKind, ExpnKind, MacroKind};
45 pub use hygiene::{ExpnData, ExpnHash, ExpnId, LocalExpnId, SyntaxContext};
46 use rustc_data_structures::stable_hasher::HashingControls;
48 use def_id::{CrateNum, DefId, DefPathHash, LocalDefId, LOCAL_CRATE};
51 pub use span_encoding::{Span, DUMMY_SP};
54 pub use symbol::{sym, Symbol};
56 mod analyze_source_file;
59 use rustc_data_structures::stable_hasher::{HashStable, StableHasher};
60 use rustc_data_structures::sync::{Lock, Lrc};
63 use std::cmp::{self, Ordering};
66 use std::ops::{Add, Range, Sub};
67 use std::path::{Path, PathBuf};
68 use std::str::FromStr;
80 // Per-session global variables: this struct is stored in thread-local storage
81 // in such a way that it is accessible without any kind of handle to all
82 // threads within the compilation session, but is not accessible outside the
84 pub struct SessionGlobals {
85 symbol_interner: symbol::Interner,
86 span_interner: Lock<span_encoding::SpanInterner>,
87 hygiene_data: Lock<hygiene::HygieneData>,
88 source_map: Lock<Option<Lrc<SourceMap>>>,
92 pub fn new(edition: Edition) -> SessionGlobals {
94 symbol_interner: symbol::Interner::fresh(),
95 span_interner: Lock::new(span_encoding::SpanInterner::default()),
96 hygiene_data: Lock::new(hygiene::HygieneData::new(edition)),
97 source_map: Lock::new(None),
103 pub fn create_session_globals_then<R>(edition: Edition, f: impl FnOnce() -> R) -> R {
105 !SESSION_GLOBALS.is_set(),
106 "SESSION_GLOBALS should never be overwritten! \
107 Use another thread if you need another SessionGlobals"
109 let session_globals = SessionGlobals::new(edition);
110 SESSION_GLOBALS.set(&session_globals, f)
114 pub fn set_session_globals_then<R>(session_globals: &SessionGlobals, f: impl FnOnce() -> R) -> R {
116 !SESSION_GLOBALS.is_set(),
117 "SESSION_GLOBALS should never be overwritten! \
118 Use another thread if you need another SessionGlobals"
120 SESSION_GLOBALS.set(session_globals, f)
124 pub fn create_default_session_if_not_set_then<R, F>(f: F) -> R
126 F: FnOnce(&SessionGlobals) -> R,
128 create_session_if_not_set_then(edition::DEFAULT_EDITION, f)
132 pub fn create_session_if_not_set_then<R, F>(edition: Edition, f: F) -> R
134 F: FnOnce(&SessionGlobals) -> R,
136 if !SESSION_GLOBALS.is_set() {
137 let session_globals = SessionGlobals::new(edition);
138 SESSION_GLOBALS.set(&session_globals, || SESSION_GLOBALS.with(f))
140 SESSION_GLOBALS.with(f)
145 pub fn with_session_globals<R, F>(f: F) -> R
147 F: FnOnce(&SessionGlobals) -> R,
149 SESSION_GLOBALS.with(f)
153 pub fn create_default_session_globals_then<R>(f: impl FnOnce() -> R) -> R {
154 create_session_globals_then(edition::DEFAULT_EDITION, f)
157 // If this ever becomes non thread-local, `decode_syntax_context`
158 // and `decode_expn_id` will need to be updated to handle concurrent
160 scoped_tls::scoped_thread_local!(static SESSION_GLOBALS: SessionGlobals);
162 // FIXME: We should use this enum or something like it to get rid of the
163 // use of magic `/rust/1.x/...` paths across the board.
164 #[derive(Debug, Eq, PartialEq, Clone, Ord, PartialOrd)]
166 pub enum RealFileName {
168 /// For remapped paths (namely paths into libstd that have been mapped
169 /// to the appropriate spot on the local host's file system, and local file
170 /// system paths that have been remapped with `FilePathMapping`),
172 /// `local_path` is the (host-dependent) local path to the file. This is
173 /// None if the file was imported from another crate
174 local_path: Option<PathBuf>,
175 /// `virtual_name` is the stable path rustc will store internally within
177 virtual_name: PathBuf,
181 impl Hash for RealFileName {
182 fn hash<H: std::hash::Hasher>(&self, state: &mut H) {
183 // To prevent #70924 from happening again we should only hash the
184 // remapped (virtualized) path if that exists. This is because
185 // virtualized paths to sysroot crates (/rust/$hash or /rust/$version)
186 // remain stable even if the corresponding local_path changes
187 self.remapped_path_if_available().hash(state)
191 // This is functionally identical to #[derive(Encodable)], with the exception of
192 // an added assert statement
193 impl<S: Encoder> Encodable<S> for RealFileName {
194 fn encode(&self, encoder: &mut S) -> Result<(), S::Error> {
195 encoder.emit_enum(|encoder| match *self {
196 RealFileName::LocalPath(ref local_path) => {
197 encoder.emit_enum_variant("LocalPath", 0, 1, |encoder| {
198 encoder.emit_enum_variant_arg(true, |encoder| local_path.encode(encoder))?;
203 RealFileName::Remapped { ref local_path, ref virtual_name } => encoder
204 .emit_enum_variant("Remapped", 1, 2, |encoder| {
205 // For privacy and build reproducibility, we must not embed host-dependant path in artifacts
206 // if they have been remapped by --remap-path-prefix
207 assert!(local_path.is_none());
208 encoder.emit_enum_variant_arg(true, |encoder| local_path.encode(encoder))?;
209 encoder.emit_enum_variant_arg(false, |encoder| virtual_name.encode(encoder))?;
217 /// Returns the path suitable for reading from the file system on the local host,
218 /// if this information exists.
219 /// Avoid embedding this in build artifacts; see `remapped_path_if_available()` for that.
220 pub fn local_path(&self) -> Option<&Path> {
222 RealFileName::LocalPath(p) => Some(p),
223 RealFileName::Remapped { local_path: p, virtual_name: _ } => {
224 p.as_ref().map(PathBuf::as_path)
229 /// Returns the path suitable for reading from the file system on the local host,
230 /// if this information exists.
231 /// Avoid embedding this in build artifacts; see `remapped_path_if_available()` for that.
232 pub fn into_local_path(self) -> Option<PathBuf> {
234 RealFileName::LocalPath(p) => Some(p),
235 RealFileName::Remapped { local_path: p, virtual_name: _ } => p,
239 /// Returns the path suitable for embedding into build artifacts. This would still
240 /// be a local path if it has not been remapped. A remapped path will not correspond
241 /// to a valid file system path: see `local_path_if_available()` for something that
242 /// is more likely to return paths into the local host file system.
243 pub fn remapped_path_if_available(&self) -> &Path {
245 RealFileName::LocalPath(p)
246 | RealFileName::Remapped { local_path: _, virtual_name: p } => &p,
250 /// Returns the path suitable for reading from the file system on the local host,
251 /// if this information exists. Otherwise returns the remapped name.
252 /// Avoid embedding this in build artifacts; see `remapped_path_if_available()` for that.
253 pub fn local_path_if_available(&self) -> &Path {
255 RealFileName::LocalPath(path)
256 | RealFileName::Remapped { local_path: None, virtual_name: path }
257 | RealFileName::Remapped { local_path: Some(path), virtual_name: _ } => path,
261 pub fn to_string_lossy(&self, display_pref: FileNameDisplayPreference) -> Cow<'_, str> {
263 FileNameDisplayPreference::Local => self.local_path_if_available().to_string_lossy(),
264 FileNameDisplayPreference::Remapped => {
265 self.remapped_path_if_available().to_string_lossy()
271 /// Differentiates between real files and common virtual files.
272 #[derive(Debug, Eq, PartialEq, Clone, Ord, PartialOrd, Hash)]
273 #[derive(Decodable, Encodable)]
276 /// Call to `quote!`.
280 /// Hack in `src/librustc_ast/parse.rs`.
283 ProcMacroSourceCode(u64),
284 /// Strings provided as `--cfg [cfgspec]` stored in a `crate_cfg`.
286 /// Strings provided as crate attributes in the CLI.
288 /// Custom sources for explicit parser calls from plugins and drivers.
290 DocTest(PathBuf, isize),
291 /// Post-substitution inline assembly from LLVM.
295 impl From<PathBuf> for FileName {
296 fn from(p: PathBuf) -> Self {
297 assert!(!p.to_string_lossy().ends_with('>'));
298 FileName::Real(RealFileName::LocalPath(p))
302 #[derive(Clone, Copy, Eq, PartialEq, Hash, Debug)]
303 pub enum FileNameDisplayPreference {
308 pub struct FileNameDisplay<'a> {
310 display_pref: FileNameDisplayPreference,
313 impl fmt::Display for FileNameDisplay<'_> {
314 fn fmt(&self, fmt: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
318 write!(fmt, "{}", name.to_string_lossy(self.display_pref))
320 QuoteExpansion(_) => write!(fmt, "<quote expansion>"),
321 MacroExpansion(_) => write!(fmt, "<macro expansion>"),
322 Anon(_) => write!(fmt, "<anon>"),
323 ProcMacroSourceCode(_) => write!(fmt, "<proc-macro source code>"),
324 CfgSpec(_) => write!(fmt, "<cfgspec>"),
325 CliCrateAttr(_) => write!(fmt, "<crate attribute>"),
326 Custom(ref s) => write!(fmt, "<{}>", s),
327 DocTest(ref path, _) => write!(fmt, "{}", path.display()),
328 InlineAsm(_) => write!(fmt, "<inline asm>"),
333 impl FileNameDisplay<'_> {
334 pub fn to_string_lossy(&self) -> Cow<'_, str> {
336 FileName::Real(ref inner) => inner.to_string_lossy(self.display_pref),
337 _ => Cow::from(format!("{}", self)),
343 pub fn is_real(&self) -> bool {
349 | ProcMacroSourceCode(_)
355 | InlineAsm(_) => false,
359 pub fn prefer_remapped(&self) -> FileNameDisplay<'_> {
360 FileNameDisplay { inner: self, display_pref: FileNameDisplayPreference::Remapped }
363 // This may include transient local filesystem information.
364 // Must not be embedded in build outputs.
365 pub fn prefer_local(&self) -> FileNameDisplay<'_> {
366 FileNameDisplay { inner: self, display_pref: FileNameDisplayPreference::Local }
369 pub fn display(&self, display_pref: FileNameDisplayPreference) -> FileNameDisplay<'_> {
370 FileNameDisplay { inner: self, display_pref }
373 pub fn macro_expansion_source_code(src: &str) -> FileName {
374 let mut hasher = StableHasher::new();
375 src.hash(&mut hasher);
376 FileName::MacroExpansion(hasher.finish())
379 pub fn anon_source_code(src: &str) -> FileName {
380 let mut hasher = StableHasher::new();
381 src.hash(&mut hasher);
382 FileName::Anon(hasher.finish())
385 pub fn proc_macro_source_code(src: &str) -> FileName {
386 let mut hasher = StableHasher::new();
387 src.hash(&mut hasher);
388 FileName::ProcMacroSourceCode(hasher.finish())
391 pub fn cfg_spec_source_code(src: &str) -> FileName {
392 let mut hasher = StableHasher::new();
393 src.hash(&mut hasher);
394 FileName::QuoteExpansion(hasher.finish())
397 pub fn cli_crate_attr_source_code(src: &str) -> FileName {
398 let mut hasher = StableHasher::new();
399 src.hash(&mut hasher);
400 FileName::CliCrateAttr(hasher.finish())
403 pub fn doc_test_source_code(path: PathBuf, line: isize) -> FileName {
404 FileName::DocTest(path, line)
407 pub fn inline_asm_source_code(src: &str) -> FileName {
408 let mut hasher = StableHasher::new();
409 src.hash(&mut hasher);
410 FileName::InlineAsm(hasher.finish())
414 /// Represents a span.
416 /// Spans represent a region of code, used for error reporting. Positions in spans
417 /// are *absolute* positions from the beginning of the [`SourceMap`], not positions
418 /// relative to [`SourceFile`]s. Methods on the `SourceMap` can be used to relate spans back
419 /// to the original source.
421 /// You must be careful if the span crosses more than one file, since you will not be
422 /// able to use many of the functions on spans in source_map and you cannot assume
423 /// that the length of the span is equal to `span.hi - span.lo`; there may be space in the
424 /// [`BytePos`] range between files.
426 /// `SpanData` is public because `Span` uses a thread-local interner and can't be
427 /// sent to other threads, but some pieces of performance infra run in a separate thread.
428 /// Using `Span` is generally preferred.
429 #[derive(Clone, Copy, Hash, PartialEq, Eq)]
430 pub struct SpanData {
433 /// Information about where the macro came from, if this piece of
434 /// code was created by a macro expansion.
435 pub ctxt: SyntaxContext,
436 pub parent: Option<LocalDefId>,
439 // Order spans by position in the file.
440 impl Ord for SpanData {
441 fn cmp(&self, other: &Self) -> Ordering {
446 // `LocalDefId` does not implement `Ord`.
447 // The other fields are enough to determine in-file order.
454 // `LocalDefId` does not implement `Ord`.
455 // The other fields are enough to determine in-file order.
459 (s_lo, s_hi, s_ctxt).cmp(&(o_lo, o_hi, o_ctxt))
463 impl PartialOrd for SpanData {
464 fn partial_cmp(&self, other: &Self) -> Option<Ordering> {
465 Some(self.cmp(other))
471 pub fn span(&self) -> Span {
472 Span::new(self.lo, self.hi, self.ctxt, self.parent)
475 pub fn with_lo(&self, lo: BytePos) -> Span {
476 Span::new(lo, self.hi, self.ctxt, self.parent)
479 pub fn with_hi(&self, hi: BytePos) -> Span {
480 Span::new(self.lo, hi, self.ctxt, self.parent)
483 pub fn with_ctxt(&self, ctxt: SyntaxContext) -> Span {
484 Span::new(self.lo, self.hi, ctxt, self.parent)
487 pub fn with_parent(&self, parent: Option<LocalDefId>) -> Span {
488 Span::new(self.lo, self.hi, self.ctxt, parent)
490 /// Returns `true` if this is a dummy span with any hygienic context.
492 pub fn is_dummy(self) -> bool {
493 self.lo.0 == 0 && self.hi.0 == 0
495 /// Returns `true` if `self` fully encloses `other`.
496 pub fn contains(self, other: Self) -> bool {
497 self.lo <= other.lo && other.hi <= self.hi
501 // The interner is pointed to by a thread local value which is only set on the main thread
502 // with parallelization is disabled. So we don't allow `Span` to transfer between threads
503 // to avoid panics and other errors, even though it would be memory safe to do so.
504 #[cfg(not(parallel_compiler))]
505 impl !Send for Span {}
506 #[cfg(not(parallel_compiler))]
507 impl !Sync for Span {}
509 impl PartialOrd for Span {
510 fn partial_cmp(&self, rhs: &Self) -> Option<Ordering> {
511 PartialOrd::partial_cmp(&self.data(), &rhs.data())
515 fn cmp(&self, rhs: &Self) -> Ordering {
516 Ord::cmp(&self.data(), &rhs.data())
520 /// A collection of `Span`s.
522 /// Spans have two orthogonal attributes:
524 /// - They can be *primary spans*. In this case they are the locus of
525 /// the error, and would be rendered with `^^^`.
526 /// - They can have a *label*. In this case, the label is written next
527 /// to the mark in the snippet when we render.
528 #[derive(Clone, Debug, Hash, PartialEq, Eq, Encodable, Decodable)]
529 pub struct MultiSpan {
530 primary_spans: Vec<Span>,
531 span_labels: Vec<(Span, String)>,
536 pub fn lo(self) -> BytePos {
540 pub fn with_lo(self, lo: BytePos) -> Span {
541 self.data().with_lo(lo)
544 pub fn hi(self) -> BytePos {
548 pub fn with_hi(self, hi: BytePos) -> Span {
549 self.data().with_hi(hi)
552 pub fn ctxt(self) -> SyntaxContext {
553 self.data_untracked().ctxt
556 pub fn with_ctxt(self, ctxt: SyntaxContext) -> Span {
557 self.data_untracked().with_ctxt(ctxt)
560 pub fn parent(self) -> Option<LocalDefId> {
564 pub fn with_parent(self, ctxt: Option<LocalDefId>) -> Span {
565 self.data().with_parent(ctxt)
568 /// Returns `true` if this is a dummy span with any hygienic context.
570 pub fn is_dummy(self) -> bool {
571 self.data_untracked().is_dummy()
574 /// Returns `true` if this span comes from a macro or desugaring.
576 pub fn from_expansion(self) -> bool {
577 self.ctxt() != SyntaxContext::root()
580 /// Returns `true` if `span` originates in a derive-macro's expansion.
581 pub fn in_derive_expansion(self) -> bool {
582 matches!(self.ctxt().outer_expn_data().kind, ExpnKind::Macro(MacroKind::Derive, _))
585 /// Gate suggestions that would not be appropriate in a context the user didn't write.
586 pub fn can_be_used_for_suggestions(self) -> bool {
587 !self.from_expansion()
588 // FIXME: If this span comes from a `derive` macro but it points at code the user wrote,
589 // the callsite span and the span will be pointing at different places. It also means that
590 // we can safely provide suggestions on this span.
591 || (matches!(self.ctxt().outer_expn_data().kind, ExpnKind::Macro(MacroKind::Derive, _))
592 && self.parent_callsite().map(|p| (p.lo(), p.hi())) != Some((self.lo(), self.hi())))
596 pub fn with_root_ctxt(lo: BytePos, hi: BytePos) -> Span {
597 Span::new(lo, hi, SyntaxContext::root(), None)
600 /// Returns a new span representing an empty span at the beginning of this span.
602 pub fn shrink_to_lo(self) -> Span {
603 let span = self.data_untracked();
604 span.with_hi(span.lo)
606 /// Returns a new span representing an empty span at the end of this span.
608 pub fn shrink_to_hi(self) -> Span {
609 let span = self.data_untracked();
610 span.with_lo(span.hi)
614 /// Returns `true` if `hi == lo`.
615 pub fn is_empty(self) -> bool {
616 let span = self.data_untracked();
620 /// Returns `self` if `self` is not the dummy span, and `other` otherwise.
621 pub fn substitute_dummy(self, other: Span) -> Span {
622 if self.is_dummy() { other } else { self }
625 /// Returns `true` if `self` fully encloses `other`.
626 pub fn contains(self, other: Span) -> bool {
627 let span = self.data();
628 let other = other.data();
632 /// Returns `true` if `self` touches `other`.
633 pub fn overlaps(self, other: Span) -> bool {
634 let span = self.data();
635 let other = other.data();
636 span.lo < other.hi && other.lo < span.hi
639 /// Returns `true` if the spans are equal with regards to the source text.
641 /// Use this instead of `==` when either span could be generated code,
642 /// and you only care that they point to the same bytes of source text.
643 pub fn source_equal(self, other: Span) -> bool {
644 let span = self.data();
645 let other = other.data();
646 span.lo == other.lo && span.hi == other.hi
649 /// Returns `Some(span)`, where the start is trimmed by the end of `other`.
650 pub fn trim_start(self, other: Span) -> Option<Span> {
651 let span = self.data();
652 let other = other.data();
653 if span.hi > other.hi { Some(span.with_lo(cmp::max(span.lo, other.hi))) } else { None }
656 /// Returns the source span -- this is either the supplied span, or the span for
657 /// the macro callsite that expanded to it.
658 pub fn source_callsite(self) -> Span {
659 let expn_data = self.ctxt().outer_expn_data();
660 if !expn_data.is_root() { expn_data.call_site.source_callsite() } else { self }
663 /// The `Span` for the tokens in the previous macro expansion from which `self` was generated,
665 pub fn parent_callsite(self) -> Option<Span> {
666 let expn_data = self.ctxt().outer_expn_data();
667 if !expn_data.is_root() { Some(expn_data.call_site) } else { None }
670 /// Walk down the expansion ancestors to find a span that's contained within `outer`.
671 pub fn find_ancestor_inside(mut self, outer: Span) -> Option<Span> {
672 while !outer.contains(self) {
673 self = self.parent_callsite()?;
678 /// Edition of the crate from which this span came.
679 pub fn edition(self) -> edition::Edition {
680 self.ctxt().edition()
684 pub fn rust_2015(self) -> bool {
685 self.edition() == edition::Edition::Edition2015
689 pub fn rust_2018(self) -> bool {
690 self.edition() >= edition::Edition::Edition2018
694 pub fn rust_2021(self) -> bool {
695 self.edition() >= edition::Edition::Edition2021
698 /// Returns the source callee.
700 /// Returns `None` if the supplied span has no expansion trace,
701 /// else returns the `ExpnData` for the macro definition
702 /// corresponding to the source callsite.
703 pub fn source_callee(self) -> Option<ExpnData> {
704 fn source_callee(expn_data: ExpnData) -> ExpnData {
705 let next_expn_data = expn_data.call_site.ctxt().outer_expn_data();
706 if !next_expn_data.is_root() { source_callee(next_expn_data) } else { expn_data }
708 let expn_data = self.ctxt().outer_expn_data();
709 if !expn_data.is_root() { Some(source_callee(expn_data)) } else { None }
712 /// Checks if a span is "internal" to a macro in which `#[unstable]`
713 /// items can be used (that is, a macro marked with
714 /// `#[allow_internal_unstable]`).
715 pub fn allows_unstable(self, feature: Symbol) -> bool {
718 .allow_internal_unstable
719 .map_or(false, |features| features.iter().any(|&f| f == feature))
722 /// Checks if this span arises from a compiler desugaring of kind `kind`.
723 pub fn is_desugaring(self, kind: DesugaringKind) -> bool {
724 match self.ctxt().outer_expn_data().kind {
725 ExpnKind::Desugaring(k) => k == kind,
730 /// Returns the compiler desugaring that created this span, or `None`
731 /// if this span is not from a desugaring.
732 pub fn desugaring_kind(self) -> Option<DesugaringKind> {
733 match self.ctxt().outer_expn_data().kind {
734 ExpnKind::Desugaring(k) => Some(k),
739 /// Checks if a span is "internal" to a macro in which `unsafe`
740 /// can be used without triggering the `unsafe_code` lint.
741 // (that is, a macro marked with `#[allow_internal_unsafe]`).
742 pub fn allows_unsafe(self) -> bool {
743 self.ctxt().outer_expn_data().allow_internal_unsafe
746 pub fn macro_backtrace(mut self) -> impl Iterator<Item = ExpnData> {
747 let mut prev_span = DUMMY_SP;
748 std::iter::from_fn(move || {
750 let expn_data = self.ctxt().outer_expn_data();
751 if expn_data.is_root() {
755 let is_recursive = expn_data.call_site.source_equal(prev_span);
758 self = expn_data.call_site;
760 // Don't print recursive invocations.
762 return Some(expn_data);
768 /// Returns a `Span` that would enclose both `self` and `end`.
772 /// self lorem ipsum end
773 /// ^^^^^^^^^^^^^^^^^^^^
775 pub fn to(self, end: Span) -> Span {
776 let span_data = self.data();
777 let end_data = end.data();
778 // FIXME(jseyfried): `self.ctxt` should always equal `end.ctxt` here (cf. issue #23480).
779 // Return the macro span on its own to avoid weird diagnostic output. It is preferable to
780 // have an incomplete span than a completely nonsensical one.
781 if span_data.ctxt != end_data.ctxt {
782 if span_data.ctxt == SyntaxContext::root() {
784 } else if end_data.ctxt == SyntaxContext::root() {
787 // Both spans fall within a macro.
788 // FIXME(estebank): check if it is the *same* macro.
791 cmp::min(span_data.lo, end_data.lo),
792 cmp::max(span_data.hi, end_data.hi),
793 if span_data.ctxt == SyntaxContext::root() { end_data.ctxt } else { span_data.ctxt },
794 if span_data.parent == end_data.parent { span_data.parent } else { None },
798 /// Returns a `Span` between the end of `self` to the beginning of `end`.
802 /// self lorem ipsum end
805 pub fn between(self, end: Span) -> Span {
806 let span = self.data();
807 let end = end.data();
811 if end.ctxt == SyntaxContext::root() { end.ctxt } else { span.ctxt },
812 if span.parent == end.parent { span.parent } else { None },
816 /// Returns a `Span` from the beginning of `self` until the beginning of `end`.
820 /// self lorem ipsum end
821 /// ^^^^^^^^^^^^^^^^^
823 pub fn until(self, end: Span) -> Span {
824 // Most of this function's body is copied from `to`.
825 // We can't just do `self.to(end.shrink_to_lo())`,
826 // because to also does some magic where it uses min/max so
827 // it can handle overlapping spans. Some advanced mis-use of
828 // `until` with different ctxts makes this visible.
829 let span_data = self.data();
830 let end_data = end.data();
831 // FIXME(jseyfried): `self.ctxt` should always equal `end.ctxt` here (cf. issue #23480).
832 // Return the macro span on its own to avoid weird diagnostic output. It is preferable to
833 // have an incomplete span than a completely nonsensical one.
834 if span_data.ctxt != end_data.ctxt {
835 if span_data.ctxt == SyntaxContext::root() {
837 } else if end_data.ctxt == SyntaxContext::root() {
840 // Both spans fall within a macro.
841 // FIXME(estebank): check if it is the *same* macro.
846 if end_data.ctxt == SyntaxContext::root() { end_data.ctxt } else { span_data.ctxt },
847 if span_data.parent == end_data.parent { span_data.parent } else { None },
851 pub fn from_inner(self, inner: InnerSpan) -> Span {
852 let span = self.data();
854 span.lo + BytePos::from_usize(inner.start),
855 span.lo + BytePos::from_usize(inner.end),
861 /// Equivalent of `Span::def_site` from the proc macro API,
862 /// except that the location is taken from the `self` span.
863 pub fn with_def_site_ctxt(self, expn_id: ExpnId) -> Span {
864 self.with_ctxt_from_mark(expn_id, Transparency::Opaque)
867 /// Equivalent of `Span::call_site` from the proc macro API,
868 /// except that the location is taken from the `self` span.
869 pub fn with_call_site_ctxt(self, expn_id: ExpnId) -> Span {
870 self.with_ctxt_from_mark(expn_id, Transparency::Transparent)
873 /// Equivalent of `Span::mixed_site` from the proc macro API,
874 /// except that the location is taken from the `self` span.
875 pub fn with_mixed_site_ctxt(self, expn_id: ExpnId) -> Span {
876 self.with_ctxt_from_mark(expn_id, Transparency::SemiTransparent)
879 /// Produces a span with the same location as `self` and context produced by a macro with the
880 /// given ID and transparency, assuming that macro was defined directly and not produced by
881 /// some other macro (which is the case for built-in and procedural macros).
882 pub fn with_ctxt_from_mark(self, expn_id: ExpnId, transparency: Transparency) -> Span {
883 self.with_ctxt(SyntaxContext::root().apply_mark(expn_id, transparency))
887 pub fn apply_mark(self, expn_id: ExpnId, transparency: Transparency) -> Span {
888 let span = self.data();
889 span.with_ctxt(span.ctxt.apply_mark(expn_id, transparency))
893 pub fn remove_mark(&mut self) -> ExpnId {
894 let mut span = self.data();
895 let mark = span.ctxt.remove_mark();
896 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
901 pub fn adjust(&mut self, expn_id: ExpnId) -> Option<ExpnId> {
902 let mut span = self.data();
903 let mark = span.ctxt.adjust(expn_id);
904 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
909 pub fn normalize_to_macros_2_0_and_adjust(&mut self, expn_id: ExpnId) -> Option<ExpnId> {
910 let mut span = self.data();
911 let mark = span.ctxt.normalize_to_macros_2_0_and_adjust(expn_id);
912 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
917 pub fn glob_adjust(&mut self, expn_id: ExpnId, glob_span: Span) -> Option<Option<ExpnId>> {
918 let mut span = self.data();
919 let mark = span.ctxt.glob_adjust(expn_id, glob_span);
920 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
925 pub fn reverse_glob_adjust(
929 ) -> Option<Option<ExpnId>> {
930 let mut span = self.data();
931 let mark = span.ctxt.reverse_glob_adjust(expn_id, glob_span);
932 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
937 pub fn normalize_to_macros_2_0(self) -> Span {
938 let span = self.data();
939 span.with_ctxt(span.ctxt.normalize_to_macros_2_0())
943 pub fn normalize_to_macro_rules(self) -> Span {
944 let span = self.data();
945 span.with_ctxt(span.ctxt.normalize_to_macro_rules())
949 /// A span together with some additional data.
950 #[derive(Clone, Debug)]
951 pub struct SpanLabel {
952 /// The span we are going to include in the final snippet.
955 /// Is this a primary span? This is the "locus" of the message,
956 /// and is indicated with a `^^^^` underline, versus `----`.
957 pub is_primary: bool,
959 /// What label should we attach to this span (if any)?
960 pub label: Option<String>,
963 impl Default for Span {
964 fn default() -> Self {
969 impl<E: Encoder> Encodable<E> for Span {
970 default fn encode(&self, s: &mut E) -> Result<(), E::Error> {
971 let span = self.data();
972 s.emit_struct(false, |s| {
973 s.emit_struct_field("lo", true, |s| span.lo.encode(s))?;
974 s.emit_struct_field("hi", false, |s| span.hi.encode(s))
978 impl<D: Decoder> Decodable<D> for Span {
979 default fn decode(s: &mut D) -> Span {
981 let lo = d.read_struct_field("lo", Decodable::decode);
982 let hi = d.read_struct_field("hi", Decodable::decode);
984 Span::new(lo, hi, SyntaxContext::root(), None)
989 /// Calls the provided closure, using the provided `SourceMap` to format
990 /// any spans that are debug-printed during the closure's execution.
992 /// Normally, the global `TyCtxt` is used to retrieve the `SourceMap`
993 /// (see `rustc_interface::callbacks::span_debug1`). However, some parts
994 /// of the compiler (e.g. `rustc_parse`) may debug-print `Span`s before
995 /// a `TyCtxt` is available. In this case, we fall back to
996 /// the `SourceMap` provided to this function. If that is not available,
997 /// we fall back to printing the raw `Span` field values.
998 pub fn with_source_map<T, F: FnOnce() -> T>(source_map: Lrc<SourceMap>, f: F) -> T {
999 with_session_globals(|session_globals| {
1000 *session_globals.source_map.borrow_mut() = Some(source_map);
1002 struct ClearSourceMap;
1003 impl Drop for ClearSourceMap {
1004 fn drop(&mut self) {
1005 with_session_globals(|session_globals| {
1006 session_globals.source_map.borrow_mut().take();
1011 let _guard = ClearSourceMap;
1015 pub fn debug_with_source_map(
1017 f: &mut fmt::Formatter<'_>,
1018 source_map: &SourceMap,
1020 write!(f, "{} ({:?})", source_map.span_to_diagnostic_string(span), span.ctxt())
1023 pub fn default_span_debug(span: Span, f: &mut fmt::Formatter<'_>) -> fmt::Result {
1024 with_session_globals(|session_globals| {
1025 if let Some(source_map) = &*session_globals.source_map.borrow() {
1026 debug_with_source_map(span, f, source_map)
1028 f.debug_struct("Span")
1029 .field("lo", &span.lo())
1030 .field("hi", &span.hi())
1031 .field("ctxt", &span.ctxt())
1037 impl fmt::Debug for Span {
1038 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
1039 (*SPAN_DEBUG)(*self, f)
1043 impl fmt::Debug for SpanData {
1044 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
1045 (*SPAN_DEBUG)(Span::new(self.lo, self.hi, self.ctxt, self.parent), f)
1051 pub fn new() -> MultiSpan {
1052 MultiSpan { primary_spans: vec![], span_labels: vec![] }
1055 pub fn from_span(primary_span: Span) -> MultiSpan {
1056 MultiSpan { primary_spans: vec![primary_span], span_labels: vec![] }
1059 pub fn from_spans(mut vec: Vec<Span>) -> MultiSpan {
1061 MultiSpan { primary_spans: vec, span_labels: vec![] }
1064 pub fn push_span_label(&mut self, span: Span, label: String) {
1065 self.span_labels.push((span, label));
1068 /// Selects the first primary span (if any).
1069 pub fn primary_span(&self) -> Option<Span> {
1070 self.primary_spans.first().cloned()
1073 /// Returns all primary spans.
1074 pub fn primary_spans(&self) -> &[Span] {
1078 /// Returns `true` if any of the primary spans are displayable.
1079 pub fn has_primary_spans(&self) -> bool {
1080 self.primary_spans.iter().any(|sp| !sp.is_dummy())
1083 /// Returns `true` if this contains only a dummy primary span with any hygienic context.
1084 pub fn is_dummy(&self) -> bool {
1085 let mut is_dummy = true;
1086 for span in &self.primary_spans {
1087 if !span.is_dummy() {
1094 /// Replaces all occurrences of one Span with another. Used to move `Span`s in areas that don't
1095 /// display well (like std macros). Returns whether replacements occurred.
1096 pub fn replace(&mut self, before: Span, after: Span) -> bool {
1097 let mut replacements_occurred = false;
1098 for primary_span in &mut self.primary_spans {
1099 if *primary_span == before {
1100 *primary_span = after;
1101 replacements_occurred = true;
1104 for span_label in &mut self.span_labels {
1105 if span_label.0 == before {
1106 span_label.0 = after;
1107 replacements_occurred = true;
1110 replacements_occurred
1113 /// Returns the strings to highlight. We always ensure that there
1114 /// is an entry for each of the primary spans -- for each primary
1115 /// span `P`, if there is at least one label with span `P`, we return
1116 /// those labels (marked as primary). But otherwise we return
1117 /// `SpanLabel` instances with empty labels.
1118 pub fn span_labels(&self) -> Vec<SpanLabel> {
1119 let is_primary = |span| self.primary_spans.contains(&span);
1121 let mut span_labels = self
1124 .map(|&(span, ref label)| SpanLabel {
1126 is_primary: is_primary(span),
1127 label: Some(label.clone()),
1129 .collect::<Vec<_>>();
1131 for &span in &self.primary_spans {
1132 if !span_labels.iter().any(|sl| sl.span == span) {
1133 span_labels.push(SpanLabel { span, is_primary: true, label: None });
1140 /// Returns `true` if any of the span labels is displayable.
1141 pub fn has_span_labels(&self) -> bool {
1142 self.span_labels.iter().any(|(sp, _)| !sp.is_dummy())
1146 impl From<Span> for MultiSpan {
1147 fn from(span: Span) -> MultiSpan {
1148 MultiSpan::from_span(span)
1152 impl From<Vec<Span>> for MultiSpan {
1153 fn from(spans: Vec<Span>) -> MultiSpan {
1154 MultiSpan::from_spans(spans)
1158 /// Identifies an offset of a multi-byte character in a `SourceFile`.
1159 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1160 pub struct MultiByteChar {
1161 /// The absolute offset of the character in the `SourceMap`.
1163 /// The number of bytes, `>= 2`.
1167 /// Identifies an offset of a non-narrow character in a `SourceFile`.
1168 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1169 pub enum NonNarrowChar {
1170 /// Represents a zero-width character.
1172 /// Represents a wide (full-width) character.
1174 /// Represents a tab character, represented visually with a width of 4 characters.
1178 impl NonNarrowChar {
1179 fn new(pos: BytePos, width: usize) -> Self {
1181 0 => NonNarrowChar::ZeroWidth(pos),
1182 2 => NonNarrowChar::Wide(pos),
1183 4 => NonNarrowChar::Tab(pos),
1184 _ => panic!("width {} given for non-narrow character", width),
1188 /// Returns the absolute offset of the character in the `SourceMap`.
1189 pub fn pos(&self) -> BytePos {
1191 NonNarrowChar::ZeroWidth(p) | NonNarrowChar::Wide(p) | NonNarrowChar::Tab(p) => p,
1195 /// Returns the width of the character, 0 (zero-width) or 2 (wide).
1196 pub fn width(&self) -> usize {
1198 NonNarrowChar::ZeroWidth(_) => 0,
1199 NonNarrowChar::Wide(_) => 2,
1200 NonNarrowChar::Tab(_) => 4,
1205 impl Add<BytePos> for NonNarrowChar {
1208 fn add(self, rhs: BytePos) -> Self {
1210 NonNarrowChar::ZeroWidth(pos) => NonNarrowChar::ZeroWidth(pos + rhs),
1211 NonNarrowChar::Wide(pos) => NonNarrowChar::Wide(pos + rhs),
1212 NonNarrowChar::Tab(pos) => NonNarrowChar::Tab(pos + rhs),
1217 impl Sub<BytePos> for NonNarrowChar {
1220 fn sub(self, rhs: BytePos) -> Self {
1222 NonNarrowChar::ZeroWidth(pos) => NonNarrowChar::ZeroWidth(pos - rhs),
1223 NonNarrowChar::Wide(pos) => NonNarrowChar::Wide(pos - rhs),
1224 NonNarrowChar::Tab(pos) => NonNarrowChar::Tab(pos - rhs),
1229 /// Identifies an offset of a character that was normalized away from `SourceFile`.
1230 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1231 pub struct NormalizedPos {
1232 /// The absolute offset of the character in the `SourceMap`.
1234 /// The difference between original and normalized string at position.
1238 #[derive(PartialEq, Eq, Clone, Debug)]
1239 pub enum ExternalSource {
1240 /// No external source has to be loaded, since the `SourceFile` represents a local crate.
1243 kind: ExternalSourceKind,
1244 /// This SourceFile's byte-offset within the source_map of its original crate.
1245 original_start_pos: BytePos,
1246 /// The end of this SourceFile within the source_map of its original crate.
1247 original_end_pos: BytePos,
1251 /// The state of the lazy external source loading mechanism of a `SourceFile`.
1252 #[derive(PartialEq, Eq, Clone, Debug)]
1253 pub enum ExternalSourceKind {
1254 /// The external source has been loaded already.
1255 Present(Lrc<String>),
1256 /// No attempt has been made to load the external source.
1258 /// A failed attempt has been made to load the external source.
1263 impl ExternalSource {
1264 pub fn get_source(&self) -> Option<&Lrc<String>> {
1266 ExternalSource::Foreign { kind: ExternalSourceKind::Present(ref src), .. } => Some(src),
1273 pub struct OffsetOverflowError;
1275 #[derive(Copy, Clone, Debug, PartialEq, Eq, PartialOrd, Ord, Hash, Encodable, Decodable)]
1276 pub enum SourceFileHashAlgorithm {
1282 impl FromStr for SourceFileHashAlgorithm {
1285 fn from_str(s: &str) -> Result<SourceFileHashAlgorithm, ()> {
1287 "md5" => Ok(SourceFileHashAlgorithm::Md5),
1288 "sha1" => Ok(SourceFileHashAlgorithm::Sha1),
1289 "sha256" => Ok(SourceFileHashAlgorithm::Sha256),
1295 rustc_data_structures::impl_stable_hash_via_hash!(SourceFileHashAlgorithm);
1297 /// The hash of the on-disk source file used for debug info.
1298 #[derive(Copy, Clone, PartialEq, Eq, Debug)]
1299 #[derive(HashStable_Generic, Encodable, Decodable)]
1300 pub struct SourceFileHash {
1301 pub kind: SourceFileHashAlgorithm,
1305 impl SourceFileHash {
1306 pub fn new(kind: SourceFileHashAlgorithm, src: &str) -> SourceFileHash {
1307 let mut hash = SourceFileHash { kind, value: Default::default() };
1308 let len = hash.hash_len();
1309 let value = &mut hash.value[..len];
1310 let data = src.as_bytes();
1312 SourceFileHashAlgorithm::Md5 => {
1313 value.copy_from_slice(&Md5::digest(data));
1315 SourceFileHashAlgorithm::Sha1 => {
1316 value.copy_from_slice(&Sha1::digest(data));
1318 SourceFileHashAlgorithm::Sha256 => {
1319 value.copy_from_slice(&Sha256::digest(data));
1325 /// Check if the stored hash matches the hash of the string.
1326 pub fn matches(&self, src: &str) -> bool {
1327 Self::new(self.kind, src) == *self
1330 /// The bytes of the hash.
1331 pub fn hash_bytes(&self) -> &[u8] {
1332 let len = self.hash_len();
1336 fn hash_len(&self) -> usize {
1338 SourceFileHashAlgorithm::Md5 => 16,
1339 SourceFileHashAlgorithm::Sha1 => 20,
1340 SourceFileHashAlgorithm::Sha256 => 32,
1345 /// A single source in the [`SourceMap`].
1347 pub struct SourceFile {
1348 /// The name of the file that the source came from. Source that doesn't
1349 /// originate from files has names between angle brackets by convention
1350 /// (e.g., `<anon>`).
1352 /// The complete source code.
1353 pub src: Option<Lrc<String>>,
1354 /// The source code's hash.
1355 pub src_hash: SourceFileHash,
1356 /// The external source code (used for external crates, which will have a `None`
1357 /// value as `self.src`.
1358 pub external_src: Lock<ExternalSource>,
1359 /// The start position of this source in the `SourceMap`.
1360 pub start_pos: BytePos,
1361 /// The end position of this source in the `SourceMap`.
1362 pub end_pos: BytePos,
1363 /// Locations of lines beginnings in the source code.
1364 pub lines: Vec<BytePos>,
1365 /// Locations of multi-byte characters in the source code.
1366 pub multibyte_chars: Vec<MultiByteChar>,
1367 /// Width of characters that are not narrow in the source code.
1368 pub non_narrow_chars: Vec<NonNarrowChar>,
1369 /// Locations of characters removed during normalization.
1370 pub normalized_pos: Vec<NormalizedPos>,
1371 /// A hash of the filename, used for speeding up hashing in incremental compilation.
1372 pub name_hash: u128,
1373 /// Indicates which crate this `SourceFile` was imported from.
1377 impl<S: Encoder> Encodable<S> for SourceFile {
1378 fn encode(&self, s: &mut S) -> Result<(), S::Error> {
1379 s.emit_struct(false, |s| {
1380 s.emit_struct_field("name", true, |s| self.name.encode(s))?;
1381 s.emit_struct_field("src_hash", false, |s| self.src_hash.encode(s))?;
1382 s.emit_struct_field("start_pos", false, |s| self.start_pos.encode(s))?;
1383 s.emit_struct_field("end_pos", false, |s| self.end_pos.encode(s))?;
1384 s.emit_struct_field("lines", false, |s| {
1385 let lines = &self.lines[..];
1386 // Store the length.
1387 s.emit_u32(lines.len() as u32)?;
1389 if !lines.is_empty() {
1390 // In order to preserve some space, we exploit the fact that
1391 // the lines list is sorted and individual lines are
1392 // probably not that long. Because of that we can store lines
1393 // as a difference list, using as little space as possible
1394 // for the differences.
1395 let max_line_length = if lines.len() == 1 {
1400 .map(|&[fst, snd]| snd - fst)
1401 .map(|bp| bp.to_usize())
1406 let bytes_per_diff: u8 = match max_line_length {
1408 0x100..=0xFFFF => 2,
1412 // Encode the number of bytes used per diff.
1413 bytes_per_diff.encode(s)?;
1415 // Encode the first element.
1416 lines[0].encode(s)?;
1418 let diff_iter = lines.array_windows().map(|&[fst, snd]| snd - fst);
1420 match bytes_per_diff {
1422 for diff in diff_iter {
1423 (diff.0 as u8).encode(s)?
1427 for diff in diff_iter {
1428 (diff.0 as u16).encode(s)?
1432 for diff in diff_iter {
1436 _ => unreachable!(),
1442 s.emit_struct_field("multibyte_chars", false, |s| self.multibyte_chars.encode(s))?;
1443 s.emit_struct_field("non_narrow_chars", false, |s| self.non_narrow_chars.encode(s))?;
1444 s.emit_struct_field("name_hash", false, |s| self.name_hash.encode(s))?;
1445 s.emit_struct_field("normalized_pos", false, |s| self.normalized_pos.encode(s))?;
1446 s.emit_struct_field("cnum", false, |s| self.cnum.encode(s))
1451 impl<D: Decoder> Decodable<D> for SourceFile {
1452 fn decode(d: &mut D) -> SourceFile {
1454 let name: FileName = d.read_struct_field("name", |d| Decodable::decode(d));
1455 let src_hash: SourceFileHash =
1456 d.read_struct_field("src_hash", |d| Decodable::decode(d));
1457 let start_pos: BytePos = d.read_struct_field("start_pos", |d| Decodable::decode(d));
1458 let end_pos: BytePos = d.read_struct_field("end_pos", |d| Decodable::decode(d));
1459 let lines: Vec<BytePos> = d.read_struct_field("lines", |d| {
1460 let num_lines: u32 = Decodable::decode(d);
1461 let mut lines = Vec::with_capacity(num_lines as usize);
1464 // Read the number of bytes used per diff.
1465 let bytes_per_diff: u8 = Decodable::decode(d);
1467 // Read the first element.
1468 let mut line_start: BytePos = Decodable::decode(d);
1469 lines.push(line_start);
1471 for _ in 1..num_lines {
1472 let diff = match bytes_per_diff {
1473 1 => d.read_u8() as u32,
1474 2 => d.read_u16() as u32,
1476 _ => unreachable!(),
1479 line_start = line_start + BytePos(diff);
1481 lines.push(line_start);
1487 let multibyte_chars: Vec<MultiByteChar> =
1488 d.read_struct_field("multibyte_chars", |d| Decodable::decode(d));
1489 let non_narrow_chars: Vec<NonNarrowChar> =
1490 d.read_struct_field("non_narrow_chars", |d| Decodable::decode(d));
1491 let name_hash: u128 = d.read_struct_field("name_hash", |d| Decodable::decode(d));
1492 let normalized_pos: Vec<NormalizedPos> =
1493 d.read_struct_field("normalized_pos", |d| Decodable::decode(d));
1494 let cnum: CrateNum = d.read_struct_field("cnum", |d| Decodable::decode(d));
1501 // Unused - the metadata decoder will construct
1502 // a new SourceFile, filling in `external_src` properly
1503 external_src: Lock::new(ExternalSource::Unneeded),
1515 impl fmt::Debug for SourceFile {
1516 fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result {
1517 write!(fmt, "SourceFile({:?})", self.name)
1526 hash_kind: SourceFileHashAlgorithm,
1528 // Compute the file hash before any normalization.
1529 let src_hash = SourceFileHash::new(hash_kind, &src);
1530 let normalized_pos = normalize_src(&mut src, start_pos);
1533 let mut hasher: StableHasher = StableHasher::new();
1534 name.hash(&mut hasher);
1535 hasher.finish::<u128>()
1537 let end_pos = start_pos.to_usize() + src.len();
1538 assert!(end_pos <= u32::MAX as usize);
1540 let (lines, multibyte_chars, non_narrow_chars) =
1541 analyze_source_file::analyze_source_file(&src, start_pos);
1545 src: Some(Lrc::new(src)),
1547 external_src: Lock::new(ExternalSource::Unneeded),
1549 end_pos: Pos::from_usize(end_pos),
1559 /// Returns the `BytePos` of the beginning of the current line.
1560 pub fn line_begin_pos(&self, pos: BytePos) -> BytePos {
1561 let line_index = self.lookup_line(pos).unwrap();
1562 self.lines[line_index]
1565 /// Add externally loaded source.
1566 /// If the hash of the input doesn't match or no input is supplied via None,
1567 /// it is interpreted as an error and the corresponding enum variant is set.
1568 /// The return value signifies whether some kind of source is present.
1569 pub fn add_external_src<F>(&self, get_src: F) -> bool
1571 F: FnOnce() -> Option<String>,
1574 *self.external_src.borrow(),
1575 ExternalSource::Foreign { kind: ExternalSourceKind::AbsentOk, .. }
1577 let src = get_src();
1578 let mut external_src = self.external_src.borrow_mut();
1579 // Check that no-one else have provided the source while we were getting it
1580 if let ExternalSource::Foreign {
1581 kind: src_kind @ ExternalSourceKind::AbsentOk, ..
1582 } = &mut *external_src
1584 if let Some(mut src) = src {
1585 // The src_hash needs to be computed on the pre-normalized src.
1586 if self.src_hash.matches(&src) {
1587 normalize_src(&mut src, BytePos::from_usize(0));
1588 *src_kind = ExternalSourceKind::Present(Lrc::new(src));
1592 *src_kind = ExternalSourceKind::AbsentErr;
1597 self.src.is_some() || external_src.get_source().is_some()
1600 self.src.is_some() || self.external_src.borrow().get_source().is_some()
1604 /// Gets a line from the list of pre-computed line-beginnings.
1605 /// The line number here is 0-based.
1606 pub fn get_line(&self, line_number: usize) -> Option<Cow<'_, str>> {
1607 fn get_until_newline(src: &str, begin: usize) -> &str {
1608 // We can't use `lines.get(line_number+1)` because we might
1609 // be parsing when we call this function and thus the current
1610 // line is the last one we have line info for.
1611 let slice = &src[begin..];
1612 match slice.find('\n') {
1613 Some(e) => &slice[..e],
1619 let line = self.lines.get(line_number)?;
1620 let begin: BytePos = *line - self.start_pos;
1624 if let Some(ref src) = self.src {
1625 Some(Cow::from(get_until_newline(src, begin)))
1626 } else if let Some(src) = self.external_src.borrow().get_source() {
1627 Some(Cow::Owned(String::from(get_until_newline(src, begin))))
1633 pub fn is_real_file(&self) -> bool {
1637 pub fn is_imported(&self) -> bool {
1641 pub fn count_lines(&self) -> usize {
1645 /// Finds the line containing the given position. The return value is the
1646 /// index into the `lines` array of this `SourceFile`, not the 1-based line
1647 /// number. If the source_file is empty or the position is located before the
1648 /// first line, `None` is returned.
1649 pub fn lookup_line(&self, pos: BytePos) -> Option<usize> {
1650 match self.lines.binary_search(&pos) {
1651 Ok(idx) => Some(idx),
1653 Err(idx) => Some(idx - 1),
1657 pub fn line_bounds(&self, line_index: usize) -> Range<BytePos> {
1658 if self.is_empty() {
1659 return self.start_pos..self.end_pos;
1662 assert!(line_index < self.lines.len());
1663 if line_index == (self.lines.len() - 1) {
1664 self.lines[line_index]..self.end_pos
1666 self.lines[line_index]..self.lines[line_index + 1]
1670 /// Returns whether or not the file contains the given `SourceMap` byte
1671 /// position. The position one past the end of the file is considered to be
1672 /// contained by the file. This implies that files for which `is_empty`
1673 /// returns true still contain one byte position according to this function.
1675 pub fn contains(&self, byte_pos: BytePos) -> bool {
1676 byte_pos >= self.start_pos && byte_pos <= self.end_pos
1680 pub fn is_empty(&self) -> bool {
1681 self.start_pos == self.end_pos
1684 /// Calculates the original byte position relative to the start of the file
1685 /// based on the given byte position.
1686 pub fn original_relative_byte_pos(&self, pos: BytePos) -> BytePos {
1687 // Diff before any records is 0. Otherwise use the previously recorded
1688 // diff as that applies to the following characters until a new diff
1690 let diff = match self.normalized_pos.binary_search_by(|np| np.pos.cmp(&pos)) {
1691 Ok(i) => self.normalized_pos[i].diff,
1692 Err(i) if i == 0 => 0,
1693 Err(i) => self.normalized_pos[i - 1].diff,
1696 BytePos::from_u32(pos.0 - self.start_pos.0 + diff)
1699 /// Converts an absolute `BytePos` to a `CharPos` relative to the `SourceFile`.
1700 pub fn bytepos_to_file_charpos(&self, bpos: BytePos) -> CharPos {
1701 // The number of extra bytes due to multibyte chars in the `SourceFile`.
1702 let mut total_extra_bytes = 0;
1704 for mbc in self.multibyte_chars.iter() {
1705 debug!("{}-byte char at {:?}", mbc.bytes, mbc.pos);
1707 // Every character is at least one byte, so we only
1708 // count the actual extra bytes.
1709 total_extra_bytes += mbc.bytes as u32 - 1;
1710 // We should never see a byte position in the middle of a
1712 assert!(bpos.to_u32() >= mbc.pos.to_u32() + mbc.bytes as u32);
1718 assert!(self.start_pos.to_u32() + total_extra_bytes <= bpos.to_u32());
1719 CharPos(bpos.to_usize() - self.start_pos.to_usize() - total_extra_bytes as usize)
1722 /// Looks up the file's (1-based) line number and (0-based `CharPos`) column offset, for a
1723 /// given `BytePos`.
1724 pub fn lookup_file_pos(&self, pos: BytePos) -> (usize, CharPos) {
1725 let chpos = self.bytepos_to_file_charpos(pos);
1726 match self.lookup_line(pos) {
1728 let line = a + 1; // Line numbers start at 1
1729 let linebpos = self.lines[a];
1730 let linechpos = self.bytepos_to_file_charpos(linebpos);
1731 let col = chpos - linechpos;
1732 debug!("byte pos {:?} is on the line at byte pos {:?}", pos, linebpos);
1733 debug!("char pos {:?} is on the line at char pos {:?}", chpos, linechpos);
1734 debug!("byte is on line: {}", line);
1735 assert!(chpos >= linechpos);
1742 /// Looks up the file's (1-based) line number, (0-based `CharPos`) column offset, and (0-based)
1743 /// column offset when displayed, for a given `BytePos`.
1744 pub fn lookup_file_pos_with_col_display(&self, pos: BytePos) -> (usize, CharPos, usize) {
1745 let (line, col_or_chpos) = self.lookup_file_pos(pos);
1747 let col = col_or_chpos;
1748 let linebpos = self.lines[line - 1];
1750 let start_width_idx = self
1752 .binary_search_by_key(&linebpos, |x| x.pos())
1753 .unwrap_or_else(|x| x);
1754 let end_width_idx = self
1756 .binary_search_by_key(&pos, |x| x.pos())
1757 .unwrap_or_else(|x| x);
1758 let special_chars = end_width_idx - start_width_idx;
1759 let non_narrow: usize = self.non_narrow_chars[start_width_idx..end_width_idx]
1763 col.0 - special_chars + non_narrow
1765 (line, col, col_display)
1767 let chpos = col_or_chpos;
1769 let end_width_idx = self
1771 .binary_search_by_key(&pos, |x| x.pos())
1772 .unwrap_or_else(|x| x);
1773 let non_narrow: usize =
1774 self.non_narrow_chars[0..end_width_idx].iter().map(|x| x.width()).sum();
1775 chpos.0 - end_width_idx + non_narrow
1777 (0, chpos, col_display)
1782 /// Normalizes the source code and records the normalizations.
1783 fn normalize_src(src: &mut String, start_pos: BytePos) -> Vec<NormalizedPos> {
1784 let mut normalized_pos = vec![];
1785 remove_bom(src, &mut normalized_pos);
1786 normalize_newlines(src, &mut normalized_pos);
1788 // Offset all the positions by start_pos to match the final file positions.
1789 for np in &mut normalized_pos {
1790 np.pos.0 += start_pos.0;
1796 /// Removes UTF-8 BOM, if any.
1797 fn remove_bom(src: &mut String, normalized_pos: &mut Vec<NormalizedPos>) {
1798 if src.starts_with('\u{feff}') {
1800 normalized_pos.push(NormalizedPos { pos: BytePos(0), diff: 3 });
1804 /// Replaces `\r\n` with `\n` in-place in `src`.
1806 /// Returns error if there's a lone `\r` in the string.
1807 fn normalize_newlines(src: &mut String, normalized_pos: &mut Vec<NormalizedPos>) {
1808 if !src.as_bytes().contains(&b'\r') {
1812 // We replace `\r\n` with `\n` in-place, which doesn't break utf-8 encoding.
1813 // While we *can* call `as_mut_vec` and do surgery on the live string
1814 // directly, let's rather steal the contents of `src`. This makes the code
1815 // safe even if a panic occurs.
1817 let mut buf = std::mem::replace(src, String::new()).into_bytes();
1818 let mut gap_len = 0;
1819 let mut tail = buf.as_mut_slice();
1821 let original_gap = normalized_pos.last().map_or(0, |l| l.diff);
1823 let idx = match find_crlf(&tail[gap_len..]) {
1825 Some(idx) => idx + gap_len,
1827 tail.copy_within(gap_len..idx, 0);
1828 tail = &mut tail[idx - gap_len..];
1829 if tail.len() == gap_len {
1832 cursor += idx - gap_len;
1834 normalized_pos.push(NormalizedPos {
1835 pos: BytePos::from_usize(cursor + 1),
1836 diff: original_gap + gap_len as u32,
1840 // Account for removed `\r`.
1841 // After `set_len`, `buf` is guaranteed to contain utf-8 again.
1842 let new_len = buf.len() - gap_len;
1844 buf.set_len(new_len);
1845 *src = String::from_utf8_unchecked(buf);
1848 fn find_crlf(src: &[u8]) -> Option<usize> {
1849 let mut search_idx = 0;
1850 while let Some(idx) = find_cr(&src[search_idx..]) {
1851 if src[search_idx..].get(idx + 1) != Some(&b'\n') {
1852 search_idx += idx + 1;
1855 return Some(search_idx + idx);
1860 fn find_cr(src: &[u8]) -> Option<usize> {
1861 src.iter().position(|&b| b == b'\r')
1865 // _____________________________________________________________________________
1866 // Pos, BytePos, CharPos
1870 fn from_usize(n: usize) -> Self;
1871 fn to_usize(&self) -> usize;
1872 fn from_u32(n: u32) -> Self;
1873 fn to_u32(&self) -> u32;
1876 macro_rules! impl_pos {
1880 $vis:vis struct $ident:ident($inner_vis:vis $inner_ty:ty);
1885 $vis struct $ident($inner_vis $inner_ty);
1887 impl Pos for $ident {
1889 fn from_usize(n: usize) -> $ident {
1890 $ident(n as $inner_ty)
1894 fn to_usize(&self) -> usize {
1899 fn from_u32(n: u32) -> $ident {
1900 $ident(n as $inner_ty)
1904 fn to_u32(&self) -> u32 {
1909 impl Add for $ident {
1910 type Output = $ident;
1913 fn add(self, rhs: $ident) -> $ident {
1914 $ident(self.0 + rhs.0)
1918 impl Sub for $ident {
1919 type Output = $ident;
1922 fn sub(self, rhs: $ident) -> $ident {
1923 $ident(self.0 - rhs.0)
1933 /// Keep this small (currently 32-bits), as AST contains a lot of them.
1934 #[derive(Clone, Copy, PartialEq, Eq, Hash, PartialOrd, Ord, Debug)]
1935 pub struct BytePos(pub u32);
1937 /// A character offset.
1939 /// Because of multibyte UTF-8 characters, a byte offset
1940 /// is not equivalent to a character offset. The [`SourceMap`] will convert [`BytePos`]
1941 /// values to `CharPos` values as necessary.
1942 #[derive(Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Debug)]
1943 pub struct CharPos(pub usize);
1946 impl<S: rustc_serialize::Encoder> Encodable<S> for BytePos {
1947 fn encode(&self, s: &mut S) -> Result<(), S::Error> {
1952 impl<D: rustc_serialize::Decoder> Decodable<D> for BytePos {
1953 fn decode(d: &mut D) -> BytePos {
1954 BytePos(d.read_u32())
1958 // _____________________________________________________________________________
1959 // Loc, SourceFileAndLine, SourceFileAndBytePos
1962 /// A source code location used for error reporting.
1963 #[derive(Debug, Clone)]
1965 /// Information about the original source.
1966 pub file: Lrc<SourceFile>,
1967 /// The (1-based) line number.
1969 /// The (0-based) column offset.
1971 /// The (0-based) column offset when displayed.
1972 pub col_display: usize,
1975 // Used to be structural records.
1977 pub struct SourceFileAndLine {
1978 pub sf: Lrc<SourceFile>,
1979 /// Index of line, starting from 0.
1983 pub struct SourceFileAndBytePos {
1984 pub sf: Lrc<SourceFile>,
1988 #[derive(Copy, Clone, Debug, PartialEq, Eq)]
1989 pub struct LineInfo {
1990 /// Index of line, starting from 0.
1991 pub line_index: usize,
1993 /// Column in line where span begins, starting from 0.
1994 pub start_col: CharPos,
1996 /// Column in line where span ends, starting from 0, exclusive.
1997 pub end_col: CharPos,
2000 pub struct FileLines {
2001 pub file: Lrc<SourceFile>,
2002 pub lines: Vec<LineInfo>,
2005 pub static SPAN_DEBUG: AtomicRef<fn(Span, &mut fmt::Formatter<'_>) -> fmt::Result> =
2006 AtomicRef::new(&(default_span_debug as fn(_, &mut fmt::Formatter<'_>) -> _));
2007 pub static SPAN_TRACK: AtomicRef<fn(LocalDefId)> = AtomicRef::new(&((|_| {}) as fn(_)));
2009 // _____________________________________________________________________________
2010 // SpanLinesError, SpanSnippetError, DistinctSources, MalformedSourceMapPositions
2013 pub type FileLinesResult = Result<FileLines, SpanLinesError>;
2015 #[derive(Clone, PartialEq, Eq, Debug)]
2016 pub enum SpanLinesError {
2017 DistinctSources(DistinctSources),
2020 #[derive(Clone, PartialEq, Eq, Debug)]
2021 pub enum SpanSnippetError {
2022 IllFormedSpan(Span),
2023 DistinctSources(DistinctSources),
2024 MalformedForSourcemap(MalformedSourceMapPositions),
2025 SourceNotAvailable { filename: FileName },
2028 #[derive(Clone, PartialEq, Eq, Debug)]
2029 pub struct DistinctSources {
2030 pub begin: (FileName, BytePos),
2031 pub end: (FileName, BytePos),
2034 #[derive(Clone, PartialEq, Eq, Debug)]
2035 pub struct MalformedSourceMapPositions {
2037 pub source_len: usize,
2038 pub begin_pos: BytePos,
2039 pub end_pos: BytePos,
2042 /// Range inside of a `Span` used for diagnostics when we only have access to relative positions.
2043 #[derive(Copy, Clone, PartialEq, Eq, Debug)]
2044 pub struct InnerSpan {
2050 pub fn new(start: usize, end: usize) -> InnerSpan {
2051 InnerSpan { start, end }
2055 /// Requirements for a `StableHashingContext` to be used in this crate.
2057 /// This is a hack to allow using the [`HashStable_Generic`] derive macro
2058 /// instead of implementing everything in rustc_middle.
2059 pub trait HashStableContext {
2060 fn def_path_hash(&self, def_id: DefId) -> DefPathHash;
2061 fn hash_spans(&self) -> bool;
2062 /// Accesses `sess.opts.debugging_opts.incremental_ignore_spans` since
2063 /// we don't have easy access to a `Session`
2064 fn debug_opts_incremental_ignore_spans(&self) -> bool;
2065 fn def_span(&self, def_id: LocalDefId) -> Span;
2066 fn span_data_to_lines_and_cols(
2069 ) -> Option<(Lrc<SourceFile>, usize, BytePos, usize, BytePos)>;
2070 fn hashing_controls(&self) -> HashingControls;
2073 impl<CTX> HashStable<CTX> for Span
2075 CTX: HashStableContext,
2077 /// Hashes a span in a stable way. We can't directly hash the span's `BytePos`
2078 /// fields (that would be similar to hashing pointers, since those are just
2079 /// offsets into the `SourceMap`). Instead, we hash the (file name, line, column)
2080 /// triple, which stays the same even if the containing `SourceFile` has moved
2081 /// within the `SourceMap`.
2083 /// Also note that we are hashing byte offsets for the column, not unicode
2084 /// codepoint offsets. For the purpose of the hash that's sufficient.
2085 /// Also, hashing filenames is expensive so we avoid doing it twice when the
2086 /// span starts and ends in the same file, which is almost always the case.
2087 fn hash_stable(&self, ctx: &mut CTX, hasher: &mut StableHasher) {
2088 const TAG_VALID_SPAN: u8 = 0;
2089 const TAG_INVALID_SPAN: u8 = 1;
2090 const TAG_RELATIVE_SPAN: u8 = 2;
2092 if !ctx.hash_spans() {
2096 let span = self.data_untracked();
2097 span.ctxt.hash_stable(ctx, hasher);
2098 span.parent.hash_stable(ctx, hasher);
2100 if span.is_dummy() {
2101 Hash::hash(&TAG_INVALID_SPAN, hasher);
2105 if let Some(parent) = span.parent {
2106 let def_span = ctx.def_span(parent).data_untracked();
2107 if def_span.contains(span) {
2108 // This span is enclosed in a definition: only hash the relative position.
2109 Hash::hash(&TAG_RELATIVE_SPAN, hasher);
2110 (span.lo - def_span.lo).to_u32().hash_stable(ctx, hasher);
2111 (span.hi - def_span.lo).to_u32().hash_stable(ctx, hasher);
2116 // If this is not an empty or invalid span, we want to hash the last
2117 // position that belongs to it, as opposed to hashing the first
2118 // position past it.
2119 let (file, line_lo, col_lo, line_hi, col_hi) = match ctx.span_data_to_lines_and_cols(&span)
2123 Hash::hash(&TAG_INVALID_SPAN, hasher);
2128 Hash::hash(&TAG_VALID_SPAN, hasher);
2129 // We truncate the stable ID hash and line and column numbers. The chances
2130 // of causing a collision this way should be minimal.
2131 Hash::hash(&(file.name_hash as u64), hasher);
2133 // Hash both the length and the end location (line/column) of a span. If we
2134 // hash only the length, for example, then two otherwise equal spans with
2135 // different end locations will have the same hash. This can cause a problem
2136 // during incremental compilation wherein a previous result for a query that
2137 // depends on the end location of a span will be incorrectly reused when the
2138 // end location of the span it depends on has changed (see issue #74890). A
2139 // similar analysis applies if some query depends specifically on the length
2140 // of the span, but we only hash the end location. So hash both.
2142 let col_lo_trunc = (col_lo.0 as u64) & 0xFF;
2143 let line_lo_trunc = ((line_lo as u64) & 0xFF_FF_FF) << 8;
2144 let col_hi_trunc = (col_hi.0 as u64) & 0xFF << 32;
2145 let line_hi_trunc = ((line_hi as u64) & 0xFF_FF_FF) << 40;
2146 let col_line = col_lo_trunc | line_lo_trunc | col_hi_trunc | line_hi_trunc;
2147 let len = (span.hi - span.lo).0;
2148 Hash::hash(&col_line, hasher);
2149 Hash::hash(&len, hasher);