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(crate_visibility_modifier)]
19 #![feature(negative_impls)]
21 #![feature(min_specialization)]
22 #![feature(thread_local_const_init)]
25 extern crate rustc_macros;
27 use rustc_data_structures::AtomicRef;
28 use rustc_macros::HashStable_Generic;
29 use rustc_serialize::{Decodable, Decoder, Encodable, Encoder};
31 mod caching_source_map_view;
33 pub use self::caching_source_map_view::CachingSourceMapView;
34 use source_map::SourceMap;
39 pub use hygiene::SyntaxContext;
40 use hygiene::Transparency;
41 pub use hygiene::{DesugaringKind, ExpnData, ExpnId, ExpnKind, ForLoopLoc, MacroKind};
43 use def_id::{CrateNum, DefId, DefPathHash, LOCAL_CRATE};
46 pub use span_encoding::{Span, DUMMY_SP};
49 pub use symbol::{sym, Symbol};
51 mod analyze_source_file;
54 use rustc_data_structures::fingerprint::Fingerprint;
55 use rustc_data_structures::stable_hasher::{HashStable, StableHasher};
56 use rustc_data_structures::sync::{Lock, Lrc};
59 use std::cell::RefCell;
60 use std::cmp::{self, Ordering};
63 use std::ops::{Add, Range, Sub};
64 use std::path::{Path, PathBuf};
65 use std::str::FromStr;
66 use std::thread::LocalKey;
78 // Per-session global variables: this struct is stored in thread-local storage
79 // in such a way that it is accessible without any kind of handle to all
80 // threads within the compilation session, but is not accessible outside the
82 pub struct SessionGlobals {
83 symbol_interner: Lock<symbol::Interner>,
84 span_interner: Lock<span_encoding::SpanInterner>,
85 hygiene_data: Lock<hygiene::HygieneData>,
86 source_map: Lock<Option<Lrc<SourceMap>>>,
90 pub fn new(edition: Edition) -> SessionGlobals {
92 symbol_interner: Lock::new(symbol::Interner::fresh()),
93 span_interner: Lock::new(span_encoding::SpanInterner::default()),
94 hygiene_data: Lock::new(hygiene::HygieneData::new(edition)),
95 source_map: Lock::new(None),
101 pub fn create_session_globals_then<R>(edition: Edition, f: impl FnOnce() -> R) -> R {
103 !SESSION_GLOBALS.is_set(),
104 "SESSION_GLOBALS should never be overwritten! \
105 Use another thread if you need another SessionGlobals"
107 let session_globals = SessionGlobals::new(edition);
108 SESSION_GLOBALS.set(&session_globals, f)
112 pub fn set_session_globals_then<R>(session_globals: &SessionGlobals, f: impl FnOnce() -> R) -> R {
114 !SESSION_GLOBALS.is_set(),
115 "SESSION_GLOBALS should never be overwritten! \
116 Use another thread if you need another SessionGlobals"
118 SESSION_GLOBALS.set(session_globals, f)
122 pub fn create_default_session_if_not_set_then<R, F>(f: F) -> R
124 F: FnOnce(&SessionGlobals) -> R,
126 create_session_if_not_set_then(edition::DEFAULT_EDITION, f)
130 pub fn create_session_if_not_set_then<R, F>(edition: Edition, f: F) -> R
132 F: FnOnce(&SessionGlobals) -> R,
134 if !SESSION_GLOBALS.is_set() {
135 let session_globals = SessionGlobals::new(edition);
136 SESSION_GLOBALS.set(&session_globals, || SESSION_GLOBALS.with(f))
138 SESSION_GLOBALS.with(f)
143 pub fn with_session_globals<R, F>(f: F) -> R
145 F: FnOnce(&SessionGlobals) -> R,
147 SESSION_GLOBALS.with(f)
151 pub fn create_default_session_globals_then<R>(f: impl FnOnce() -> R) -> R {
152 create_session_globals_then(edition::DEFAULT_EDITION, f)
155 // If this ever becomes non thread-local, `decode_syntax_context`
156 // and `decode_expn_id` will need to be updated to handle concurrent
158 scoped_tls::scoped_thread_local!(static SESSION_GLOBALS: SessionGlobals);
160 // FIXME: We should use this enum or something like it to get rid of the
161 // use of magic `/rust/1.x/...` paths across the board.
162 #[derive(Debug, Eq, PartialEq, Clone, Ord, PartialOrd)]
163 #[derive(HashStable_Generic, Decodable)]
164 pub enum RealFileName {
166 /// For remapped paths (namely paths into libstd that have been mapped
167 /// to the appropriate spot on the local host's file system, and local file
168 /// system paths that have been remapped with `FilePathMapping`),
170 /// `local_path` is the (host-dependent) local path to the file. This is
171 /// None if the file was imported from another crate
172 local_path: Option<PathBuf>,
173 /// `virtual_name` is the stable path rustc will store internally within
175 virtual_name: PathBuf,
179 impl Hash for RealFileName {
180 fn hash<H: std::hash::Hasher>(&self, state: &mut H) {
181 // To prevent #70924 from happening again we should only hash the
182 // remapped (virtualized) path if that exists. This is because
183 // virtualized paths to sysroot crates (/rust/$hash or /rust/$version)
184 // remain stable even if the corresponding local_path changes
185 self.remapped_path_if_available().hash(state)
189 // This is functionally identical to #[derive(Encodable)], with the exception of
190 // an added assert statement
191 impl<S: Encoder> Encodable<S> for RealFileName {
192 fn encode(&self, encoder: &mut S) -> Result<(), S::Error> {
193 encoder.emit_enum(|encoder| match *self {
194 RealFileName::LocalPath(ref local_path) => {
195 encoder.emit_enum_variant("LocalPath", 0, 1, |encoder| {
198 .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());
210 .emit_enum_variant_arg(true, |encoder| local_path.encode(encoder))?;
212 .emit_enum_variant_arg(false, |encoder| virtual_name.encode(encoder))?;
220 /// Returns the path suitable for reading from the file system on the local host,
221 /// if this information exists.
222 /// Avoid embedding this in build artifacts; see `remapped_path_if_available()` for that.
223 pub fn local_path(&self) -> Option<&Path> {
225 RealFileName::LocalPath(p) => Some(p),
226 RealFileName::Remapped { local_path: p, virtual_name: _ } => {
227 p.as_ref().map(PathBuf::as_path)
232 /// Returns the path suitable for reading from the file system on the local host,
233 /// if this information exists.
234 /// Avoid embedding this in build artifacts; see `remapped_path_if_available()` for that.
235 pub fn into_local_path(self) -> Option<PathBuf> {
237 RealFileName::LocalPath(p) => Some(p),
238 RealFileName::Remapped { local_path: p, virtual_name: _ } => p,
242 /// Returns the path suitable for embedding into build artifacts. This would still
243 /// be a local path if it has not been remapped. A remapped path will not correspond
244 /// to a valid file system path: see `local_path_if_available()` for something that
245 /// is more likely to return paths into the local host file system.
246 pub fn remapped_path_if_available(&self) -> &Path {
248 RealFileName::LocalPath(p)
249 | RealFileName::Remapped { local_path: _, virtual_name: p } => &p,
253 /// Returns the path suitable for reading from the file system on the local host,
254 /// if this information exists. Otherwise returns the remapped name.
255 /// Avoid embedding this in build artifacts; see `remapped_path_if_available()` for that.
256 pub fn local_path_if_available(&self) -> &Path {
258 RealFileName::LocalPath(path)
259 | RealFileName::Remapped { local_path: None, virtual_name: path }
260 | RealFileName::Remapped { local_path: Some(path), virtual_name: _ } => path,
264 pub fn to_string_lossy(&self, prefer_local: bool) -> Cow<'_, str> {
266 self.local_path_if_available().to_string_lossy()
268 self.remapped_path_if_available().to_string_lossy()
273 /// Differentiates between real files and common virtual files.
274 #[derive(Debug, Eq, PartialEq, Clone, Ord, PartialOrd, Hash)]
275 #[derive(HashStable_Generic, Decodable, Encodable)]
278 /// Call to `quote!`.
282 /// Hack in `src/librustc_ast/parse.rs`.
285 ProcMacroSourceCode(u64),
286 /// Strings provided as `--cfg [cfgspec]` stored in a `crate_cfg`.
288 /// Strings provided as crate attributes in the CLI.
290 /// Custom sources for explicit parser calls from plugins and drivers.
292 DocTest(PathBuf, isize),
293 /// Post-substitution inline assembly from LLVM.
297 impl From<PathBuf> for FileName {
298 fn from(p: PathBuf) -> Self {
299 assert!(!p.to_string_lossy().ends_with('>'));
300 FileName::Real(RealFileName::LocalPath(p))
304 pub struct FileNameDisplay<'a> {
309 impl fmt::Display for FileNameDisplay<'_> {
310 fn fmt(&self, fmt: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
314 write!(fmt, "{}", name.to_string_lossy(self.prefer_local))
316 QuoteExpansion(_) => write!(fmt, "<quote expansion>"),
317 MacroExpansion(_) => write!(fmt, "<macro expansion>"),
318 Anon(_) => write!(fmt, "<anon>"),
319 ProcMacroSourceCode(_) => write!(fmt, "<proc-macro source code>"),
320 CfgSpec(_) => write!(fmt, "<cfgspec>"),
321 CliCrateAttr(_) => write!(fmt, "<crate attribute>"),
322 Custom(ref s) => write!(fmt, "<{}>", s),
323 DocTest(ref path, _) => write!(fmt, "{}", path.display()),
324 InlineAsm(_) => write!(fmt, "<inline asm>"),
329 impl FileNameDisplay<'_> {
330 pub fn to_string_lossy(&self) -> Cow<'_, str> {
332 FileName::Real(ref inner) => inner.to_string_lossy(self.prefer_local),
333 _ => Cow::from(format!("{}", self)),
339 pub fn is_real(&self) -> bool {
345 | ProcMacroSourceCode(_)
351 | InlineAsm(_) => false,
355 pub fn prefer_remapped(&self) -> FileNameDisplay<'_> {
356 FileNameDisplay { inner: self, prefer_local: false }
359 // This may include transient local filesystem information.
360 // Must not be embedded in build outputs.
361 pub fn prefer_local(&self) -> FileNameDisplay<'_> {
362 FileNameDisplay { inner: self, prefer_local: true }
365 pub fn macro_expansion_source_code(src: &str) -> FileName {
366 let mut hasher = StableHasher::new();
367 src.hash(&mut hasher);
368 FileName::MacroExpansion(hasher.finish())
371 pub fn anon_source_code(src: &str) -> FileName {
372 let mut hasher = StableHasher::new();
373 src.hash(&mut hasher);
374 FileName::Anon(hasher.finish())
377 pub fn proc_macro_source_code(src: &str) -> FileName {
378 let mut hasher = StableHasher::new();
379 src.hash(&mut hasher);
380 FileName::ProcMacroSourceCode(hasher.finish())
383 pub fn cfg_spec_source_code(src: &str) -> FileName {
384 let mut hasher = StableHasher::new();
385 src.hash(&mut hasher);
386 FileName::QuoteExpansion(hasher.finish())
389 pub fn cli_crate_attr_source_code(src: &str) -> FileName {
390 let mut hasher = StableHasher::new();
391 src.hash(&mut hasher);
392 FileName::CliCrateAttr(hasher.finish())
395 pub fn doc_test_source_code(path: PathBuf, line: isize) -> FileName {
396 FileName::DocTest(path, line)
399 pub fn inline_asm_source_code(src: &str) -> FileName {
400 let mut hasher = StableHasher::new();
401 src.hash(&mut hasher);
402 FileName::InlineAsm(hasher.finish())
406 /// Represents a span.
408 /// Spans represent a region of code, used for error reporting. Positions in spans
409 /// are *absolute* positions from the beginning of the [`SourceMap`], not positions
410 /// relative to [`SourceFile`]s. Methods on the `SourceMap` can be used to relate spans back
411 /// to the original source.
413 /// You must be careful if the span crosses more than one file, since you will not be
414 /// able to use many of the functions on spans in source_map and you cannot assume
415 /// that the length of the span is equal to `span.hi - span.lo`; there may be space in the
416 /// [`BytePos`] range between files.
418 /// `SpanData` is public because `Span` uses a thread-local interner and can't be
419 /// sent to other threads, but some pieces of performance infra run in a separate thread.
420 /// Using `Span` is generally preferred.
421 #[derive(Clone, Copy, Hash, PartialEq, Eq, Ord, PartialOrd)]
422 pub struct SpanData {
425 /// Information about where the macro came from, if this piece of
426 /// code was created by a macro expansion.
427 pub ctxt: SyntaxContext,
432 pub fn span(&self) -> Span {
433 Span::new(self.lo, self.hi, self.ctxt)
436 pub fn with_lo(&self, lo: BytePos) -> Span {
437 Span::new(lo, self.hi, self.ctxt)
440 pub fn with_hi(&self, hi: BytePos) -> Span {
441 Span::new(self.lo, hi, self.ctxt)
444 pub fn with_ctxt(&self, ctxt: SyntaxContext) -> Span {
445 Span::new(self.lo, self.hi, ctxt)
449 // The interner is pointed to by a thread local value which is only set on the main thread
450 // with parallelization is disabled. So we don't allow `Span` to transfer between threads
451 // to avoid panics and other errors, even though it would be memory safe to do so.
452 #[cfg(not(parallel_compiler))]
453 impl !Send for Span {}
454 #[cfg(not(parallel_compiler))]
455 impl !Sync for Span {}
457 impl PartialOrd for Span {
458 fn partial_cmp(&self, rhs: &Self) -> Option<Ordering> {
459 PartialOrd::partial_cmp(&self.data(), &rhs.data())
463 fn cmp(&self, rhs: &Self) -> Ordering {
464 Ord::cmp(&self.data(), &rhs.data())
468 /// A collection of `Span`s.
470 /// Spans have two orthogonal attributes:
472 /// - They can be *primary spans*. In this case they are the locus of
473 /// the error, and would be rendered with `^^^`.
474 /// - They can have a *label*. In this case, the label is written next
475 /// to the mark in the snippet when we render.
476 #[derive(Clone, Debug, Hash, PartialEq, Eq, Encodable, Decodable)]
477 pub struct MultiSpan {
478 primary_spans: Vec<Span>,
479 span_labels: Vec<(Span, String)>,
484 pub fn lo(self) -> BytePos {
488 pub fn with_lo(self, lo: BytePos) -> Span {
489 self.data().with_lo(lo)
492 pub fn hi(self) -> BytePos {
496 pub fn with_hi(self, hi: BytePos) -> Span {
497 self.data().with_hi(hi)
500 pub fn ctxt(self) -> SyntaxContext {
504 pub fn with_ctxt(self, ctxt: SyntaxContext) -> Span {
505 self.data().with_ctxt(ctxt)
508 /// Returns `true` if this is a dummy span with any hygienic context.
510 pub fn is_dummy(self) -> bool {
511 let span = self.data();
512 span.lo.0 == 0 && span.hi.0 == 0
515 /// Returns `true` if this span comes from a macro or desugaring.
517 pub fn from_expansion(self) -> bool {
518 self.ctxt() != SyntaxContext::root()
521 /// Returns `true` if `span` originates in a derive-macro's expansion.
522 pub fn in_derive_expansion(self) -> bool {
523 matches!(self.ctxt().outer_expn_data().kind, ExpnKind::Macro(MacroKind::Derive, _))
527 pub fn with_root_ctxt(lo: BytePos, hi: BytePos) -> Span {
528 Span::new(lo, hi, SyntaxContext::root())
531 /// Returns a new span representing an empty span at the beginning of this span.
533 pub fn shrink_to_lo(self) -> Span {
534 let span = self.data();
535 span.with_hi(span.lo)
537 /// Returns a new span representing an empty span at the end of this span.
539 pub fn shrink_to_hi(self) -> Span {
540 let span = self.data();
541 span.with_lo(span.hi)
545 /// Returns `true` if `hi == lo`.
546 pub fn is_empty(&self) -> bool {
547 let span = self.data();
551 /// Returns `self` if `self` is not the dummy span, and `other` otherwise.
552 pub fn substitute_dummy(self, other: Span) -> Span {
553 if self.is_dummy() { other } else { self }
556 /// Returns `true` if `self` fully encloses `other`.
557 pub fn contains(self, other: Span) -> bool {
558 let span = self.data();
559 let other = other.data();
560 span.lo <= other.lo && other.hi <= span.hi
563 /// Returns `true` if `self` touches `other`.
564 pub fn overlaps(self, other: Span) -> bool {
565 let span = self.data();
566 let other = other.data();
567 span.lo < other.hi && other.lo < span.hi
570 /// Returns `true` if the spans are equal with regards to the source text.
572 /// Use this instead of `==` when either span could be generated code,
573 /// and you only care that they point to the same bytes of source text.
574 pub fn source_equal(&self, other: &Span) -> bool {
575 let span = self.data();
576 let other = other.data();
577 span.lo == other.lo && span.hi == other.hi
580 /// Returns `Some(span)`, where the start is trimmed by the end of `other`.
581 pub fn trim_start(self, other: Span) -> Option<Span> {
582 let span = self.data();
583 let other = other.data();
584 if span.hi > other.hi { Some(span.with_lo(cmp::max(span.lo, other.hi))) } else { None }
587 /// Returns the source span -- this is either the supplied span, or the span for
588 /// the macro callsite that expanded to it.
589 pub fn source_callsite(self) -> Span {
590 let expn_data = self.ctxt().outer_expn_data();
591 if !expn_data.is_root() { expn_data.call_site.source_callsite() } else { self }
594 /// The `Span` for the tokens in the previous macro expansion from which `self` was generated,
596 pub fn parent(self) -> Option<Span> {
597 let expn_data = self.ctxt().outer_expn_data();
598 if !expn_data.is_root() { Some(expn_data.call_site) } else { None }
601 /// Edition of the crate from which this span came.
602 pub fn edition(self) -> edition::Edition {
603 self.ctxt().edition()
607 pub fn rust_2015(&self) -> bool {
608 self.edition() == edition::Edition::Edition2015
612 pub fn rust_2018(&self) -> bool {
613 self.edition() >= edition::Edition::Edition2018
617 pub fn rust_2021(&self) -> bool {
618 self.edition() >= edition::Edition::Edition2021
621 /// Returns the source callee.
623 /// Returns `None` if the supplied span has no expansion trace,
624 /// else returns the `ExpnData` for the macro definition
625 /// corresponding to the source callsite.
626 pub fn source_callee(self) -> Option<ExpnData> {
627 fn source_callee(expn_data: ExpnData) -> ExpnData {
628 let next_expn_data = expn_data.call_site.ctxt().outer_expn_data();
629 if !next_expn_data.is_root() { source_callee(next_expn_data) } else { expn_data }
631 let expn_data = self.ctxt().outer_expn_data();
632 if !expn_data.is_root() { Some(source_callee(expn_data)) } else { None }
635 /// Checks if a span is "internal" to a macro in which `#[unstable]`
636 /// items can be used (that is, a macro marked with
637 /// `#[allow_internal_unstable]`).
638 pub fn allows_unstable(&self, feature: Symbol) -> bool {
641 .allow_internal_unstable
642 .map_or(false, |features| features.iter().any(|&f| f == feature))
645 /// Checks if this span arises from a compiler desugaring of kind `kind`.
646 pub fn is_desugaring(&self, kind: DesugaringKind) -> bool {
647 match self.ctxt().outer_expn_data().kind {
648 ExpnKind::Desugaring(k) => k == kind,
653 /// Returns the compiler desugaring that created this span, or `None`
654 /// if this span is not from a desugaring.
655 pub fn desugaring_kind(&self) -> Option<DesugaringKind> {
656 match self.ctxt().outer_expn_data().kind {
657 ExpnKind::Desugaring(k) => Some(k),
662 /// Checks if a span is "internal" to a macro in which `unsafe`
663 /// can be used without triggering the `unsafe_code` lint.
664 // (that is, a macro marked with `#[allow_internal_unsafe]`).
665 pub fn allows_unsafe(&self) -> bool {
666 self.ctxt().outer_expn_data().allow_internal_unsafe
669 pub fn macro_backtrace(mut self) -> impl Iterator<Item = ExpnData> {
670 let mut prev_span = DUMMY_SP;
671 std::iter::from_fn(move || {
673 let expn_data = self.ctxt().outer_expn_data();
674 if expn_data.is_root() {
678 let is_recursive = expn_data.call_site.source_equal(&prev_span);
681 self = expn_data.call_site;
683 // Don't print recursive invocations.
685 return Some(expn_data);
691 /// Returns a `Span` that would enclose both `self` and `end`.
695 /// self lorem ipsum end
696 /// ^^^^^^^^^^^^^^^^^^^^
698 pub fn to(self, end: Span) -> Span {
699 let span_data = self.data();
700 let end_data = end.data();
701 // FIXME(jseyfried): `self.ctxt` should always equal `end.ctxt` here (cf. issue #23480).
702 // Return the macro span on its own to avoid weird diagnostic output. It is preferable to
703 // have an incomplete span than a completely nonsensical one.
704 if span_data.ctxt != end_data.ctxt {
705 if span_data.ctxt == SyntaxContext::root() {
707 } else if end_data.ctxt == SyntaxContext::root() {
710 // Both spans fall within a macro.
711 // FIXME(estebank): check if it is the *same* macro.
714 cmp::min(span_data.lo, end_data.lo),
715 cmp::max(span_data.hi, end_data.hi),
716 if span_data.ctxt == SyntaxContext::root() { end_data.ctxt } else { span_data.ctxt },
720 /// Returns a `Span` between the end of `self` to the beginning of `end`.
724 /// self lorem ipsum end
727 pub fn between(self, end: Span) -> Span {
728 let span = self.data();
729 let end = end.data();
733 if end.ctxt == SyntaxContext::root() { end.ctxt } else { span.ctxt },
737 /// Returns a `Span` from the beginning of `self` until the beginning of `end`.
741 /// self lorem ipsum end
742 /// ^^^^^^^^^^^^^^^^^
744 pub fn until(self, end: Span) -> Span {
745 let span = self.data();
746 let end = end.data();
750 if end.ctxt == SyntaxContext::root() { end.ctxt } else { span.ctxt },
754 pub fn from_inner(self, inner: InnerSpan) -> Span {
755 let span = self.data();
757 span.lo + BytePos::from_usize(inner.start),
758 span.lo + BytePos::from_usize(inner.end),
763 /// Equivalent of `Span::def_site` from the proc macro API,
764 /// except that the location is taken from the `self` span.
765 pub fn with_def_site_ctxt(self, expn_id: ExpnId) -> Span {
766 self.with_ctxt_from_mark(expn_id, Transparency::Opaque)
769 /// Equivalent of `Span::call_site` from the proc macro API,
770 /// except that the location is taken from the `self` span.
771 pub fn with_call_site_ctxt(&self, expn_id: ExpnId) -> Span {
772 self.with_ctxt_from_mark(expn_id, Transparency::Transparent)
775 /// Equivalent of `Span::mixed_site` from the proc macro API,
776 /// except that the location is taken from the `self` span.
777 pub fn with_mixed_site_ctxt(&self, expn_id: ExpnId) -> Span {
778 self.with_ctxt_from_mark(expn_id, Transparency::SemiTransparent)
781 /// Produces a span with the same location as `self` and context produced by a macro with the
782 /// given ID and transparency, assuming that macro was defined directly and not produced by
783 /// some other macro (which is the case for built-in and procedural macros).
784 pub fn with_ctxt_from_mark(self, expn_id: ExpnId, transparency: Transparency) -> Span {
785 self.with_ctxt(SyntaxContext::root().apply_mark(expn_id, transparency))
789 pub fn apply_mark(self, expn_id: ExpnId, transparency: Transparency) -> Span {
790 let span = self.data();
791 span.with_ctxt(span.ctxt.apply_mark(expn_id, transparency))
795 pub fn remove_mark(&mut self) -> ExpnId {
796 let mut span = self.data();
797 let mark = span.ctxt.remove_mark();
798 *self = Span::new(span.lo, span.hi, span.ctxt);
803 pub fn adjust(&mut self, expn_id: ExpnId) -> Option<ExpnId> {
804 let mut span = self.data();
805 let mark = span.ctxt.adjust(expn_id);
806 *self = Span::new(span.lo, span.hi, span.ctxt);
811 pub fn normalize_to_macros_2_0_and_adjust(&mut self, expn_id: ExpnId) -> Option<ExpnId> {
812 let mut span = self.data();
813 let mark = span.ctxt.normalize_to_macros_2_0_and_adjust(expn_id);
814 *self = Span::new(span.lo, span.hi, span.ctxt);
819 pub fn glob_adjust(&mut self, expn_id: ExpnId, glob_span: Span) -> Option<Option<ExpnId>> {
820 let mut span = self.data();
821 let mark = span.ctxt.glob_adjust(expn_id, glob_span);
822 *self = Span::new(span.lo, span.hi, span.ctxt);
827 pub fn reverse_glob_adjust(
831 ) -> Option<Option<ExpnId>> {
832 let mut span = self.data();
833 let mark = span.ctxt.reverse_glob_adjust(expn_id, glob_span);
834 *self = Span::new(span.lo, span.hi, span.ctxt);
839 pub fn normalize_to_macros_2_0(self) -> Span {
840 let span = self.data();
841 span.with_ctxt(span.ctxt.normalize_to_macros_2_0())
845 pub fn normalize_to_macro_rules(self) -> Span {
846 let span = self.data();
847 span.with_ctxt(span.ctxt.normalize_to_macro_rules())
851 /// A span together with some additional data.
852 #[derive(Clone, Debug)]
853 pub struct SpanLabel {
854 /// The span we are going to include in the final snippet.
857 /// Is this a primary span? This is the "locus" of the message,
858 /// and is indicated with a `^^^^` underline, versus `----`.
859 pub is_primary: bool,
861 /// What label should we attach to this span (if any)?
862 pub label: Option<String>,
865 impl Default for Span {
866 fn default() -> Self {
871 impl<E: Encoder> Encodable<E> for Span {
872 default fn encode(&self, s: &mut E) -> Result<(), E::Error> {
873 let span = self.data();
874 s.emit_struct(false, |s| {
875 s.emit_struct_field("lo", true, |s| span.lo.encode(s))?;
876 s.emit_struct_field("hi", false, |s| span.hi.encode(s))
880 impl<D: Decoder> Decodable<D> for Span {
881 default fn decode(s: &mut D) -> Result<Span, D::Error> {
883 let lo = d.read_struct_field("lo", Decodable::decode)?;
884 let hi = d.read_struct_field("hi", Decodable::decode)?;
886 Ok(Span::new(lo, hi, SyntaxContext::root()))
891 /// Calls the provided closure, using the provided `SourceMap` to format
892 /// any spans that are debug-printed during the closure's execution.
894 /// Normally, the global `TyCtxt` is used to retrieve the `SourceMap`
895 /// (see `rustc_interface::callbacks::span_debug1`). However, some parts
896 /// of the compiler (e.g. `rustc_parse`) may debug-print `Span`s before
897 /// a `TyCtxt` is available. In this case, we fall back to
898 /// the `SourceMap` provided to this function. If that is not available,
899 /// we fall back to printing the raw `Span` field values.
900 pub fn with_source_map<T, F: FnOnce() -> T>(source_map: Lrc<SourceMap>, f: F) -> T {
901 with_session_globals(|session_globals| {
902 *session_globals.source_map.borrow_mut() = Some(source_map);
904 struct ClearSourceMap;
905 impl Drop for ClearSourceMap {
907 with_session_globals(|session_globals| {
908 session_globals.source_map.borrow_mut().take();
913 let _guard = ClearSourceMap;
917 pub fn debug_with_source_map(
919 f: &mut fmt::Formatter<'_>,
920 source_map: &SourceMap,
922 write!(f, "{} ({:?})", source_map.span_to_diagnostic_string(span), span.ctxt())
925 pub fn default_span_debug(span: Span, f: &mut fmt::Formatter<'_>) -> fmt::Result {
926 with_session_globals(|session_globals| {
927 if let Some(source_map) = &*session_globals.source_map.borrow() {
928 debug_with_source_map(span, f, source_map)
930 f.debug_struct("Span")
931 .field("lo", &span.lo())
932 .field("hi", &span.hi())
933 .field("ctxt", &span.ctxt())
939 impl fmt::Debug for Span {
940 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
941 (*SPAN_DEBUG)(*self, f)
945 impl fmt::Debug for SpanData {
946 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
947 (*SPAN_DEBUG)(Span::new(self.lo, self.hi, self.ctxt), f)
953 pub fn new() -> MultiSpan {
954 MultiSpan { primary_spans: vec![], span_labels: vec![] }
957 pub fn from_span(primary_span: Span) -> MultiSpan {
958 MultiSpan { primary_spans: vec![primary_span], span_labels: vec![] }
961 pub fn from_spans(mut vec: Vec<Span>) -> MultiSpan {
963 MultiSpan { primary_spans: vec, span_labels: vec![] }
966 pub fn push_span_label(&mut self, span: Span, label: String) {
967 self.span_labels.push((span, label));
970 /// Selects the first primary span (if any).
971 pub fn primary_span(&self) -> Option<Span> {
972 self.primary_spans.first().cloned()
975 /// Returns all primary spans.
976 pub fn primary_spans(&self) -> &[Span] {
980 /// Returns `true` if any of the primary spans are displayable.
981 pub fn has_primary_spans(&self) -> bool {
982 self.primary_spans.iter().any(|sp| !sp.is_dummy())
985 /// Returns `true` if this contains only a dummy primary span with any hygienic context.
986 pub fn is_dummy(&self) -> bool {
987 let mut is_dummy = true;
988 for span in &self.primary_spans {
989 if !span.is_dummy() {
996 /// Replaces all occurrences of one Span with another. Used to move `Span`s in areas that don't
997 /// display well (like std macros). Returns whether replacements occurred.
998 pub fn replace(&mut self, before: Span, after: Span) -> bool {
999 let mut replacements_occurred = false;
1000 for primary_span in &mut self.primary_spans {
1001 if *primary_span == before {
1002 *primary_span = after;
1003 replacements_occurred = true;
1006 for span_label in &mut self.span_labels {
1007 if span_label.0 == before {
1008 span_label.0 = after;
1009 replacements_occurred = true;
1012 replacements_occurred
1015 /// Returns the strings to highlight. We always ensure that there
1016 /// is an entry for each of the primary spans -- for each primary
1017 /// span `P`, if there is at least one label with span `P`, we return
1018 /// those labels (marked as primary). But otherwise we return
1019 /// `SpanLabel` instances with empty labels.
1020 pub fn span_labels(&self) -> Vec<SpanLabel> {
1021 let is_primary = |span| self.primary_spans.contains(&span);
1023 let mut span_labels = self
1026 .map(|&(span, ref label)| SpanLabel {
1028 is_primary: is_primary(span),
1029 label: Some(label.clone()),
1031 .collect::<Vec<_>>();
1033 for &span in &self.primary_spans {
1034 if !span_labels.iter().any(|sl| sl.span == span) {
1035 span_labels.push(SpanLabel { span, is_primary: true, label: None });
1042 /// Returns `true` if any of the span labels is displayable.
1043 pub fn has_span_labels(&self) -> bool {
1044 self.span_labels.iter().any(|(sp, _)| !sp.is_dummy())
1048 impl From<Span> for MultiSpan {
1049 fn from(span: Span) -> MultiSpan {
1050 MultiSpan::from_span(span)
1054 impl From<Vec<Span>> for MultiSpan {
1055 fn from(spans: Vec<Span>) -> MultiSpan {
1056 MultiSpan::from_spans(spans)
1060 /// Identifies an offset of a multi-byte character in a `SourceFile`.
1061 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1062 pub struct MultiByteChar {
1063 /// The absolute offset of the character in the `SourceMap`.
1065 /// The number of bytes, `>= 2`.
1069 /// Identifies an offset of a non-narrow character in a `SourceFile`.
1070 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1071 pub enum NonNarrowChar {
1072 /// Represents a zero-width character.
1074 /// Represents a wide (full-width) character.
1076 /// Represents a tab character, represented visually with a width of 4 characters.
1080 impl NonNarrowChar {
1081 fn new(pos: BytePos, width: usize) -> Self {
1083 0 => NonNarrowChar::ZeroWidth(pos),
1084 2 => NonNarrowChar::Wide(pos),
1085 4 => NonNarrowChar::Tab(pos),
1086 _ => panic!("width {} given for non-narrow character", width),
1090 /// Returns the absolute offset of the character in the `SourceMap`.
1091 pub fn pos(&self) -> BytePos {
1093 NonNarrowChar::ZeroWidth(p) | NonNarrowChar::Wide(p) | NonNarrowChar::Tab(p) => p,
1097 /// Returns the width of the character, 0 (zero-width) or 2 (wide).
1098 pub fn width(&self) -> usize {
1100 NonNarrowChar::ZeroWidth(_) => 0,
1101 NonNarrowChar::Wide(_) => 2,
1102 NonNarrowChar::Tab(_) => 4,
1107 impl Add<BytePos> for NonNarrowChar {
1110 fn add(self, rhs: BytePos) -> Self {
1112 NonNarrowChar::ZeroWidth(pos) => NonNarrowChar::ZeroWidth(pos + rhs),
1113 NonNarrowChar::Wide(pos) => NonNarrowChar::Wide(pos + rhs),
1114 NonNarrowChar::Tab(pos) => NonNarrowChar::Tab(pos + rhs),
1119 impl Sub<BytePos> for NonNarrowChar {
1122 fn sub(self, rhs: BytePos) -> Self {
1124 NonNarrowChar::ZeroWidth(pos) => NonNarrowChar::ZeroWidth(pos - rhs),
1125 NonNarrowChar::Wide(pos) => NonNarrowChar::Wide(pos - rhs),
1126 NonNarrowChar::Tab(pos) => NonNarrowChar::Tab(pos - rhs),
1131 /// Identifies an offset of a character that was normalized away from `SourceFile`.
1132 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1133 pub struct NormalizedPos {
1134 /// The absolute offset of the character in the `SourceMap`.
1136 /// The difference between original and normalized string at position.
1140 #[derive(PartialEq, Eq, Clone, Debug)]
1141 pub enum ExternalSource {
1142 /// No external source has to be loaded, since the `SourceFile` represents a local crate.
1145 kind: ExternalSourceKind,
1146 /// This SourceFile's byte-offset within the source_map of its original crate.
1147 original_start_pos: BytePos,
1148 /// The end of this SourceFile within the source_map of its original crate.
1149 original_end_pos: BytePos,
1153 /// The state of the lazy external source loading mechanism of a `SourceFile`.
1154 #[derive(PartialEq, Eq, Clone, Debug)]
1155 pub enum ExternalSourceKind {
1156 /// The external source has been loaded already.
1157 Present(Lrc<String>),
1158 /// No attempt has been made to load the external source.
1160 /// A failed attempt has been made to load the external source.
1165 impl ExternalSource {
1166 pub fn get_source(&self) -> Option<&Lrc<String>> {
1168 ExternalSource::Foreign { kind: ExternalSourceKind::Present(ref src), .. } => Some(src),
1175 pub struct OffsetOverflowError;
1177 #[derive(Copy, Clone, Debug, PartialEq, Eq, PartialOrd, Ord, Hash, Encodable, Decodable)]
1178 pub enum SourceFileHashAlgorithm {
1184 impl FromStr for SourceFileHashAlgorithm {
1187 fn from_str(s: &str) -> Result<SourceFileHashAlgorithm, ()> {
1189 "md5" => Ok(SourceFileHashAlgorithm::Md5),
1190 "sha1" => Ok(SourceFileHashAlgorithm::Sha1),
1191 "sha256" => Ok(SourceFileHashAlgorithm::Sha256),
1197 rustc_data_structures::impl_stable_hash_via_hash!(SourceFileHashAlgorithm);
1199 /// The hash of the on-disk source file used for debug info.
1200 #[derive(Copy, Clone, PartialEq, Eq, Debug)]
1201 #[derive(HashStable_Generic, Encodable, Decodable)]
1202 pub struct SourceFileHash {
1203 pub kind: SourceFileHashAlgorithm,
1207 impl SourceFileHash {
1208 pub fn new(kind: SourceFileHashAlgorithm, src: &str) -> SourceFileHash {
1209 let mut hash = SourceFileHash { kind, value: Default::default() };
1210 let len = hash.hash_len();
1211 let value = &mut hash.value[..len];
1212 let data = src.as_bytes();
1214 SourceFileHashAlgorithm::Md5 => {
1215 value.copy_from_slice(&Md5::digest(data));
1217 SourceFileHashAlgorithm::Sha1 => {
1218 value.copy_from_slice(&Sha1::digest(data));
1220 SourceFileHashAlgorithm::Sha256 => {
1221 value.copy_from_slice(&Sha256::digest(data));
1227 /// Check if the stored hash matches the hash of the string.
1228 pub fn matches(&self, src: &str) -> bool {
1229 Self::new(self.kind, src) == *self
1232 /// The bytes of the hash.
1233 pub fn hash_bytes(&self) -> &[u8] {
1234 let len = self.hash_len();
1238 fn hash_len(&self) -> usize {
1240 SourceFileHashAlgorithm::Md5 => 16,
1241 SourceFileHashAlgorithm::Sha1 => 20,
1242 SourceFileHashAlgorithm::Sha256 => 32,
1247 /// A single source in the [`SourceMap`].
1249 pub struct SourceFile {
1250 /// The name of the file that the source came from. Source that doesn't
1251 /// originate from files has names between angle brackets by convention
1252 /// (e.g., `<anon>`).
1254 /// The complete source code.
1255 pub src: Option<Lrc<String>>,
1256 /// The source code's hash.
1257 pub src_hash: SourceFileHash,
1258 /// The external source code (used for external crates, which will have a `None`
1259 /// value as `self.src`.
1260 pub external_src: Lock<ExternalSource>,
1261 /// The start position of this source in the `SourceMap`.
1262 pub start_pos: BytePos,
1263 /// The end position of this source in the `SourceMap`.
1264 pub end_pos: BytePos,
1265 /// Locations of lines beginnings in the source code.
1266 pub lines: Vec<BytePos>,
1267 /// Locations of multi-byte characters in the source code.
1268 pub multibyte_chars: Vec<MultiByteChar>,
1269 /// Width of characters that are not narrow in the source code.
1270 pub non_narrow_chars: Vec<NonNarrowChar>,
1271 /// Locations of characters removed during normalization.
1272 pub normalized_pos: Vec<NormalizedPos>,
1273 /// A hash of the filename, used for speeding up hashing in incremental compilation.
1274 pub name_hash: u128,
1275 /// Indicates which crate this `SourceFile` was imported from.
1279 impl<S: Encoder> Encodable<S> for SourceFile {
1280 fn encode(&self, s: &mut S) -> Result<(), S::Error> {
1281 s.emit_struct(false, |s| {
1282 s.emit_struct_field("name", true, |s| self.name.encode(s))?;
1283 s.emit_struct_field("src_hash", false, |s| self.src_hash.encode(s))?;
1284 s.emit_struct_field("start_pos", false, |s| self.start_pos.encode(s))?;
1285 s.emit_struct_field("end_pos", false, |s| self.end_pos.encode(s))?;
1286 s.emit_struct_field("lines", false, |s| {
1287 let lines = &self.lines[..];
1288 // Store the length.
1289 s.emit_u32(lines.len() as u32)?;
1291 if !lines.is_empty() {
1292 // In order to preserve some space, we exploit the fact that
1293 // the lines list is sorted and individual lines are
1294 // probably not that long. Because of that we can store lines
1295 // as a difference list, using as little space as possible
1296 // for the differences.
1297 let max_line_length = if lines.len() == 1 {
1302 .map(|&[fst, snd]| snd - fst)
1303 .map(|bp| bp.to_usize())
1308 let bytes_per_diff: u8 = match max_line_length {
1310 0x100..=0xFFFF => 2,
1314 // Encode the number of bytes used per diff.
1315 bytes_per_diff.encode(s)?;
1317 // Encode the first element.
1318 lines[0].encode(s)?;
1320 let diff_iter = lines[..].array_windows().map(|&[fst, snd]| snd - fst);
1322 match bytes_per_diff {
1324 for diff in diff_iter {
1325 (diff.0 as u8).encode(s)?
1329 for diff in diff_iter {
1330 (diff.0 as u16).encode(s)?
1334 for diff in diff_iter {
1338 _ => unreachable!(),
1344 s.emit_struct_field("multibyte_chars", false, |s| self.multibyte_chars.encode(s))?;
1345 s.emit_struct_field("non_narrow_chars", false, |s| self.non_narrow_chars.encode(s))?;
1346 s.emit_struct_field("name_hash", false, |s| self.name_hash.encode(s))?;
1347 s.emit_struct_field("normalized_pos", false, |s| self.normalized_pos.encode(s))?;
1348 s.emit_struct_field("cnum", false, |s| self.cnum.encode(s))
1353 impl<D: Decoder> Decodable<D> for SourceFile {
1354 fn decode(d: &mut D) -> Result<SourceFile, D::Error> {
1356 let name: FileName = d.read_struct_field("name", |d| Decodable::decode(d))?;
1357 let src_hash: SourceFileHash =
1358 d.read_struct_field("src_hash", |d| Decodable::decode(d))?;
1359 let start_pos: BytePos = d.read_struct_field("start_pos", |d| Decodable::decode(d))?;
1360 let end_pos: BytePos = d.read_struct_field("end_pos", |d| Decodable::decode(d))?;
1361 let lines: Vec<BytePos> = d.read_struct_field("lines", |d| {
1362 let num_lines: u32 = Decodable::decode(d)?;
1363 let mut lines = Vec::with_capacity(num_lines as usize);
1366 // Read the number of bytes used per diff.
1367 let bytes_per_diff: u8 = Decodable::decode(d)?;
1369 // Read the first element.
1370 let mut line_start: BytePos = Decodable::decode(d)?;
1371 lines.push(line_start);
1373 for _ in 1..num_lines {
1374 let diff = match bytes_per_diff {
1375 1 => d.read_u8()? as u32,
1376 2 => d.read_u16()? as u32,
1378 _ => unreachable!(),
1381 line_start = line_start + BytePos(diff);
1383 lines.push(line_start);
1389 let multibyte_chars: Vec<MultiByteChar> =
1390 d.read_struct_field("multibyte_chars", |d| Decodable::decode(d))?;
1391 let non_narrow_chars: Vec<NonNarrowChar> =
1392 d.read_struct_field("non_narrow_chars", |d| Decodable::decode(d))?;
1393 let name_hash: u128 = d.read_struct_field("name_hash", |d| Decodable::decode(d))?;
1394 let normalized_pos: Vec<NormalizedPos> =
1395 d.read_struct_field("normalized_pos", |d| Decodable::decode(d))?;
1396 let cnum: CrateNum = d.read_struct_field("cnum", |d| Decodable::decode(d))?;
1403 // Unused - the metadata decoder will construct
1404 // a new SourceFile, filling in `external_src` properly
1405 external_src: Lock::new(ExternalSource::Unneeded),
1417 impl fmt::Debug for SourceFile {
1418 fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result {
1419 write!(fmt, "SourceFile({:?})", self.name)
1428 hash_kind: SourceFileHashAlgorithm,
1430 // Compute the file hash before any normalization.
1431 let src_hash = SourceFileHash::new(hash_kind, &src);
1432 let normalized_pos = normalize_src(&mut src, start_pos);
1435 let mut hasher: StableHasher = StableHasher::new();
1436 name.hash(&mut hasher);
1437 hasher.finish::<u128>()
1439 let end_pos = start_pos.to_usize() + src.len();
1440 assert!(end_pos <= u32::MAX as usize);
1442 let (lines, multibyte_chars, non_narrow_chars) =
1443 analyze_source_file::analyze_source_file(&src[..], start_pos);
1447 src: Some(Lrc::new(src)),
1449 external_src: Lock::new(ExternalSource::Unneeded),
1451 end_pos: Pos::from_usize(end_pos),
1461 /// Returns the `BytePos` of the beginning of the current line.
1462 pub fn line_begin_pos(&self, pos: BytePos) -> BytePos {
1463 let line_index = self.lookup_line(pos).unwrap();
1464 self.lines[line_index]
1467 /// Add externally loaded source.
1468 /// If the hash of the input doesn't match or no input is supplied via None,
1469 /// it is interpreted as an error and the corresponding enum variant is set.
1470 /// The return value signifies whether some kind of source is present.
1471 pub fn add_external_src<F>(&self, get_src: F) -> bool
1473 F: FnOnce() -> Option<String>,
1476 *self.external_src.borrow(),
1477 ExternalSource::Foreign { kind: ExternalSourceKind::AbsentOk, .. }
1479 let src = get_src();
1480 let mut external_src = self.external_src.borrow_mut();
1481 // Check that no-one else have provided the source while we were getting it
1482 if let ExternalSource::Foreign {
1483 kind: src_kind @ ExternalSourceKind::AbsentOk, ..
1484 } = &mut *external_src
1486 if let Some(mut src) = src {
1487 // The src_hash needs to be computed on the pre-normalized src.
1488 if self.src_hash.matches(&src) {
1489 normalize_src(&mut src, BytePos::from_usize(0));
1490 *src_kind = ExternalSourceKind::Present(Lrc::new(src));
1494 *src_kind = ExternalSourceKind::AbsentErr;
1499 self.src.is_some() || external_src.get_source().is_some()
1502 self.src.is_some() || self.external_src.borrow().get_source().is_some()
1506 /// Gets a line from the list of pre-computed line-beginnings.
1507 /// The line number here is 0-based.
1508 pub fn get_line(&self, line_number: usize) -> Option<Cow<'_, str>> {
1509 fn get_until_newline(src: &str, begin: usize) -> &str {
1510 // We can't use `lines.get(line_number+1)` because we might
1511 // be parsing when we call this function and thus the current
1512 // line is the last one we have line info for.
1513 let slice = &src[begin..];
1514 match slice.find('\n') {
1515 Some(e) => &slice[..e],
1521 let line = self.lines.get(line_number)?;
1522 let begin: BytePos = *line - self.start_pos;
1526 if let Some(ref src) = self.src {
1527 Some(Cow::from(get_until_newline(src, begin)))
1528 } else if let Some(src) = self.external_src.borrow().get_source() {
1529 Some(Cow::Owned(String::from(get_until_newline(src, begin))))
1535 pub fn is_real_file(&self) -> bool {
1539 pub fn is_imported(&self) -> bool {
1543 pub fn count_lines(&self) -> usize {
1547 /// Finds the line containing the given position. The return value is the
1548 /// index into the `lines` array of this `SourceFile`, not the 1-based line
1549 /// number. If the source_file is empty or the position is located before the
1550 /// first line, `None` is returned.
1551 pub fn lookup_line(&self, pos: BytePos) -> Option<usize> {
1552 match self.lines.binary_search(&pos) {
1553 Ok(idx) => Some(idx),
1555 Err(idx) => Some(idx - 1),
1559 pub fn line_bounds(&self, line_index: usize) -> Range<BytePos> {
1560 if self.is_empty() {
1561 return self.start_pos..self.end_pos;
1564 assert!(line_index < self.lines.len());
1565 if line_index == (self.lines.len() - 1) {
1566 self.lines[line_index]..self.end_pos
1568 self.lines[line_index]..self.lines[line_index + 1]
1572 /// Returns whether or not the file contains the given `SourceMap` byte
1573 /// position. The position one past the end of the file is considered to be
1574 /// contained by the file. This implies that files for which `is_empty`
1575 /// returns true still contain one byte position according to this function.
1577 pub fn contains(&self, byte_pos: BytePos) -> bool {
1578 byte_pos >= self.start_pos && byte_pos <= self.end_pos
1582 pub fn is_empty(&self) -> bool {
1583 self.start_pos == self.end_pos
1586 /// Calculates the original byte position relative to the start of the file
1587 /// based on the given byte position.
1588 pub fn original_relative_byte_pos(&self, pos: BytePos) -> BytePos {
1589 // Diff before any records is 0. Otherwise use the previously recorded
1590 // diff as that applies to the following characters until a new diff
1592 let diff = match self.normalized_pos.binary_search_by(|np| np.pos.cmp(&pos)) {
1593 Ok(i) => self.normalized_pos[i].diff,
1594 Err(i) if i == 0 => 0,
1595 Err(i) => self.normalized_pos[i - 1].diff,
1598 BytePos::from_u32(pos.0 - self.start_pos.0 + diff)
1601 /// Converts an absolute `BytePos` to a `CharPos` relative to the `SourceFile`.
1602 pub fn bytepos_to_file_charpos(&self, bpos: BytePos) -> CharPos {
1603 // The number of extra bytes due to multibyte chars in the `SourceFile`.
1604 let mut total_extra_bytes = 0;
1606 for mbc in self.multibyte_chars.iter() {
1607 debug!("{}-byte char at {:?}", mbc.bytes, mbc.pos);
1609 // Every character is at least one byte, so we only
1610 // count the actual extra bytes.
1611 total_extra_bytes += mbc.bytes as u32 - 1;
1612 // We should never see a byte position in the middle of a
1614 assert!(bpos.to_u32() >= mbc.pos.to_u32() + mbc.bytes as u32);
1620 assert!(self.start_pos.to_u32() + total_extra_bytes <= bpos.to_u32());
1621 CharPos(bpos.to_usize() - self.start_pos.to_usize() - total_extra_bytes as usize)
1624 /// Looks up the file's (1-based) line number and (0-based `CharPos`) column offset, for a
1625 /// given `BytePos`.
1626 pub fn lookup_file_pos(&self, pos: BytePos) -> (usize, CharPos) {
1627 let chpos = self.bytepos_to_file_charpos(pos);
1628 match self.lookup_line(pos) {
1630 let line = a + 1; // Line numbers start at 1
1631 let linebpos = self.lines[a];
1632 let linechpos = self.bytepos_to_file_charpos(linebpos);
1633 let col = chpos - linechpos;
1634 debug!("byte pos {:?} is on the line at byte pos {:?}", pos, linebpos);
1635 debug!("char pos {:?} is on the line at char pos {:?}", chpos, linechpos);
1636 debug!("byte is on line: {}", line);
1637 assert!(chpos >= linechpos);
1644 /// Looks up the file's (1-based) line number, (0-based `CharPos`) column offset, and (0-based)
1645 /// column offset when displayed, for a given `BytePos`.
1646 pub fn lookup_file_pos_with_col_display(&self, pos: BytePos) -> (usize, CharPos, usize) {
1647 let (line, col_or_chpos) = self.lookup_file_pos(pos);
1649 let col = col_or_chpos;
1650 let linebpos = self.lines[line - 1];
1652 let start_width_idx = self
1654 .binary_search_by_key(&linebpos, |x| x.pos())
1655 .unwrap_or_else(|x| x);
1656 let end_width_idx = self
1658 .binary_search_by_key(&pos, |x| x.pos())
1659 .unwrap_or_else(|x| x);
1660 let special_chars = end_width_idx - start_width_idx;
1661 let non_narrow: usize = self.non_narrow_chars[start_width_idx..end_width_idx]
1665 col.0 - special_chars + non_narrow
1667 (line, col, col_display)
1669 let chpos = col_or_chpos;
1671 let end_width_idx = self
1673 .binary_search_by_key(&pos, |x| x.pos())
1674 .unwrap_or_else(|x| x);
1675 let non_narrow: usize =
1676 self.non_narrow_chars[0..end_width_idx].iter().map(|x| x.width()).sum();
1677 chpos.0 - end_width_idx + non_narrow
1679 (0, chpos, col_display)
1684 /// Normalizes the source code and records the normalizations.
1685 fn normalize_src(src: &mut String, start_pos: BytePos) -> Vec<NormalizedPos> {
1686 let mut normalized_pos = vec![];
1687 remove_bom(src, &mut normalized_pos);
1688 normalize_newlines(src, &mut normalized_pos);
1690 // Offset all the positions by start_pos to match the final file positions.
1691 for np in &mut normalized_pos {
1692 np.pos.0 += start_pos.0;
1698 /// Removes UTF-8 BOM, if any.
1699 fn remove_bom(src: &mut String, normalized_pos: &mut Vec<NormalizedPos>) {
1700 if src.starts_with('\u{feff}') {
1702 normalized_pos.push(NormalizedPos { pos: BytePos(0), diff: 3 });
1706 /// Replaces `\r\n` with `\n` in-place in `src`.
1708 /// Returns error if there's a lone `\r` in the string.
1709 fn normalize_newlines(src: &mut String, normalized_pos: &mut Vec<NormalizedPos>) {
1710 if !src.as_bytes().contains(&b'\r') {
1714 // We replace `\r\n` with `\n` in-place, which doesn't break utf-8 encoding.
1715 // While we *can* call `as_mut_vec` and do surgery on the live string
1716 // directly, let's rather steal the contents of `src`. This makes the code
1717 // safe even if a panic occurs.
1719 let mut buf = std::mem::replace(src, String::new()).into_bytes();
1720 let mut gap_len = 0;
1721 let mut tail = buf.as_mut_slice();
1723 let original_gap = normalized_pos.last().map_or(0, |l| l.diff);
1725 let idx = match find_crlf(&tail[gap_len..]) {
1727 Some(idx) => idx + gap_len,
1729 tail.copy_within(gap_len..idx, 0);
1730 tail = &mut tail[idx - gap_len..];
1731 if tail.len() == gap_len {
1734 cursor += idx - gap_len;
1736 normalized_pos.push(NormalizedPos {
1737 pos: BytePos::from_usize(cursor + 1),
1738 diff: original_gap + gap_len as u32,
1742 // Account for removed `\r`.
1743 // After `set_len`, `buf` is guaranteed to contain utf-8 again.
1744 let new_len = buf.len() - gap_len;
1746 buf.set_len(new_len);
1747 *src = String::from_utf8_unchecked(buf);
1750 fn find_crlf(src: &[u8]) -> Option<usize> {
1751 let mut search_idx = 0;
1752 while let Some(idx) = find_cr(&src[search_idx..]) {
1753 if src[search_idx..].get(idx + 1) != Some(&b'\n') {
1754 search_idx += idx + 1;
1757 return Some(search_idx + idx);
1762 fn find_cr(src: &[u8]) -> Option<usize> {
1763 src.iter().position(|&b| b == b'\r')
1767 // _____________________________________________________________________________
1768 // Pos, BytePos, CharPos
1772 fn from_usize(n: usize) -> Self;
1773 fn to_usize(&self) -> usize;
1774 fn from_u32(n: u32) -> Self;
1775 fn to_u32(&self) -> u32;
1778 macro_rules! impl_pos {
1782 $vis:vis struct $ident:ident($inner_vis:vis $inner_ty:ty);
1787 $vis struct $ident($inner_vis $inner_ty);
1789 impl Pos for $ident {
1791 fn from_usize(n: usize) -> $ident {
1792 $ident(n as $inner_ty)
1796 fn to_usize(&self) -> usize {
1801 fn from_u32(n: u32) -> $ident {
1802 $ident(n as $inner_ty)
1806 fn to_u32(&self) -> u32 {
1811 impl Add for $ident {
1812 type Output = $ident;
1815 fn add(self, rhs: $ident) -> $ident {
1816 $ident(self.0 + rhs.0)
1820 impl Sub for $ident {
1821 type Output = $ident;
1824 fn sub(self, rhs: $ident) -> $ident {
1825 $ident(self.0 - rhs.0)
1835 /// Keep this small (currently 32-bits), as AST contains a lot of them.
1836 #[derive(Clone, Copy, PartialEq, Eq, Hash, PartialOrd, Ord, Debug)]
1837 pub struct BytePos(pub u32);
1839 /// A character offset.
1841 /// Because of multibyte UTF-8 characters, a byte offset
1842 /// is not equivalent to a character offset. The [`SourceMap`] will convert [`BytePos`]
1843 /// values to `CharPos` values as necessary.
1844 #[derive(Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Debug)]
1845 pub struct CharPos(pub usize);
1848 impl<S: rustc_serialize::Encoder> Encodable<S> for BytePos {
1849 fn encode(&self, s: &mut S) -> Result<(), S::Error> {
1854 impl<D: rustc_serialize::Decoder> Decodable<D> for BytePos {
1855 fn decode(d: &mut D) -> Result<BytePos, D::Error> {
1856 Ok(BytePos(d.read_u32()?))
1860 // _____________________________________________________________________________
1861 // Loc, SourceFileAndLine, SourceFileAndBytePos
1864 /// A source code location used for error reporting.
1865 #[derive(Debug, Clone)]
1867 /// Information about the original source.
1868 pub file: Lrc<SourceFile>,
1869 /// The (1-based) line number.
1871 /// The (0-based) column offset.
1873 /// The (0-based) column offset when displayed.
1874 pub col_display: usize,
1877 // Used to be structural records.
1879 pub struct SourceFileAndLine {
1880 pub sf: Lrc<SourceFile>,
1884 pub struct SourceFileAndBytePos {
1885 pub sf: Lrc<SourceFile>,
1889 #[derive(Copy, Clone, Debug, PartialEq, Eq)]
1890 pub struct LineInfo {
1891 /// Index of line, starting from 0.
1892 pub line_index: usize,
1894 /// Column in line where span begins, starting from 0.
1895 pub start_col: CharPos,
1897 /// Column in line where span ends, starting from 0, exclusive.
1898 pub end_col: CharPos,
1901 pub struct FileLines {
1902 pub file: Lrc<SourceFile>,
1903 pub lines: Vec<LineInfo>,
1906 pub static SPAN_DEBUG: AtomicRef<fn(Span, &mut fmt::Formatter<'_>) -> fmt::Result> =
1907 AtomicRef::new(&(default_span_debug as fn(_, &mut fmt::Formatter<'_>) -> _));
1909 // _____________________________________________________________________________
1910 // SpanLinesError, SpanSnippetError, DistinctSources, MalformedSourceMapPositions
1913 pub type FileLinesResult = Result<FileLines, SpanLinesError>;
1915 #[derive(Clone, PartialEq, Eq, Debug)]
1916 pub enum SpanLinesError {
1917 DistinctSources(DistinctSources),
1920 #[derive(Clone, PartialEq, Eq, Debug)]
1921 pub enum SpanSnippetError {
1922 IllFormedSpan(Span),
1923 DistinctSources(DistinctSources),
1924 MalformedForSourcemap(MalformedSourceMapPositions),
1925 SourceNotAvailable { filename: FileName },
1928 #[derive(Clone, PartialEq, Eq, Debug)]
1929 pub struct DistinctSources {
1930 pub begin: (FileName, BytePos),
1931 pub end: (FileName, BytePos),
1934 #[derive(Clone, PartialEq, Eq, Debug)]
1935 pub struct MalformedSourceMapPositions {
1937 pub source_len: usize,
1938 pub begin_pos: BytePos,
1939 pub end_pos: BytePos,
1942 /// Range inside of a `Span` used for diagnostics when we only have access to relative positions.
1943 #[derive(Copy, Clone, PartialEq, Eq, Debug)]
1944 pub struct InnerSpan {
1950 pub fn new(start: usize, end: usize) -> InnerSpan {
1951 InnerSpan { start, end }
1955 /// Requirements for a `StableHashingContext` to be used in this crate.
1957 /// This is a hack to allow using the [`HashStable_Generic`] derive macro
1958 /// instead of implementing everything in rustc_middle.
1959 pub trait HashStableContext {
1960 fn def_path_hash(&self, def_id: DefId) -> DefPathHash;
1961 /// Obtains a cache for storing the `Fingerprint` of an `ExpnId`.
1962 /// This method allows us to have multiple `HashStableContext` implementations
1963 /// that hash things in a different way, without the results of one polluting
1964 /// the cache of the other.
1965 fn expn_id_cache() -> &'static LocalKey<ExpnIdCache>;
1966 fn hash_spans(&self) -> bool;
1967 fn span_data_to_lines_and_cols(
1970 ) -> Option<(Lrc<SourceFile>, usize, BytePos, usize, BytePos)>;
1973 impl<CTX> HashStable<CTX> for Span
1975 CTX: HashStableContext,
1977 /// Hashes a span in a stable way. We can't directly hash the span's `BytePos`
1978 /// fields (that would be similar to hashing pointers, since those are just
1979 /// offsets into the `SourceMap`). Instead, we hash the (file name, line, column)
1980 /// triple, which stays the same even if the containing `SourceFile` has moved
1981 /// within the `SourceMap`.
1983 /// Also note that we are hashing byte offsets for the column, not unicode
1984 /// codepoint offsets. For the purpose of the hash that's sufficient.
1985 /// Also, hashing filenames is expensive so we avoid doing it twice when the
1986 /// span starts and ends in the same file, which is almost always the case.
1987 fn hash_stable(&self, ctx: &mut CTX, hasher: &mut StableHasher) {
1988 const TAG_VALID_SPAN: u8 = 0;
1989 const TAG_INVALID_SPAN: u8 = 1;
1991 if !ctx.hash_spans() {
1995 self.ctxt().hash_stable(ctx, hasher);
1997 if self.is_dummy() {
1998 Hash::hash(&TAG_INVALID_SPAN, hasher);
2002 // If this is not an empty or invalid span, we want to hash the last
2003 // position that belongs to it, as opposed to hashing the first
2004 // position past it.
2005 let span = self.data();
2006 let (file, line_lo, col_lo, line_hi, col_hi) = match ctx.span_data_to_lines_and_cols(&span)
2010 Hash::hash(&TAG_INVALID_SPAN, hasher);
2015 Hash::hash(&TAG_VALID_SPAN, hasher);
2016 // We truncate the stable ID hash and line and column numbers. The chances
2017 // of causing a collision this way should be minimal.
2018 Hash::hash(&(file.name_hash as u64), hasher);
2020 // Hash both the length and the end location (line/column) of a span. If we
2021 // hash only the length, for example, then two otherwise equal spans with
2022 // different end locations will have the same hash. This can cause a problem
2023 // during incremental compilation wherein a previous result for a query that
2024 // depends on the end location of a span will be incorrectly reused when the
2025 // end location of the span it depends on has changed (see issue #74890). A
2026 // similar analysis applies if some query depends specifically on the length
2027 // of the span, but we only hash the end location. So hash both.
2029 let col_lo_trunc = (col_lo.0 as u64) & 0xFF;
2030 let line_lo_trunc = ((line_lo as u64) & 0xFF_FF_FF) << 8;
2031 let col_hi_trunc = (col_hi.0 as u64) & 0xFF << 32;
2032 let line_hi_trunc = ((line_hi as u64) & 0xFF_FF_FF) << 40;
2033 let col_line = col_lo_trunc | line_lo_trunc | col_hi_trunc | line_hi_trunc;
2034 let len = (span.hi - span.lo).0;
2035 Hash::hash(&col_line, hasher);
2036 Hash::hash(&len, hasher);
2040 impl<CTX: HashStableContext> HashStable<CTX> for SyntaxContext {
2041 fn hash_stable(&self, ctx: &mut CTX, hasher: &mut StableHasher) {
2042 const TAG_EXPANSION: u8 = 0;
2043 const TAG_NO_EXPANSION: u8 = 1;
2045 if *self == SyntaxContext::root() {
2046 TAG_NO_EXPANSION.hash_stable(ctx, hasher);
2048 TAG_EXPANSION.hash_stable(ctx, hasher);
2049 let (expn_id, transparency) = self.outer_mark();
2050 expn_id.hash_stable(ctx, hasher);
2051 transparency.hash_stable(ctx, hasher);
2056 pub type ExpnIdCache = RefCell<Vec<Option<Fingerprint>>>;
2058 impl<CTX: HashStableContext> HashStable<CTX> for ExpnId {
2059 fn hash_stable(&self, ctx: &mut CTX, hasher: &mut StableHasher) {
2060 const TAG_ROOT: u8 = 0;
2061 const TAG_NOT_ROOT: u8 = 1;
2063 if *self == ExpnId::root() {
2064 TAG_ROOT.hash_stable(ctx, hasher);
2068 // Since the same expansion context is usually referenced many
2069 // times, we cache a stable hash of it and hash that instead of
2070 // recursing every time.
2071 let index = self.as_u32() as usize;
2072 let res = CTX::expn_id_cache().with(|cache| cache.borrow().get(index).copied().flatten());
2074 if let Some(res) = res {
2075 res.hash_stable(ctx, hasher);
2077 let new_len = index + 1;
2079 let mut sub_hasher = StableHasher::new();
2080 TAG_NOT_ROOT.hash_stable(ctx, &mut sub_hasher);
2081 self.expn_data().hash_stable(ctx, &mut sub_hasher);
2082 let sub_hash: Fingerprint = sub_hasher.finish();
2084 CTX::expn_id_cache().with(|cache| {
2085 let mut cache = cache.borrow_mut();
2086 if cache.len() < new_len {
2087 cache.resize(new_len, None);
2089 let prev = cache[index].replace(sub_hash);
2090 assert_eq!(prev, None, "Cache slot was filled");
2092 sub_hash.hash_stable(ctx, hasher);