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 use hygiene::Transparency;
40 pub use hygiene::{DesugaringKind, ExpnData, ExpnId, ExpnKind, ForLoopLoc, MacroKind};
41 pub use hygiene::{ExpnIdCache, SyntaxContext};
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::stable_hasher::{HashStable, StableHasher};
55 use rustc_data_structures::sync::{Lock, Lrc};
58 use std::cmp::{self, Ordering};
61 use std::ops::{Add, Range, Sub};
62 use std::path::{Path, PathBuf};
63 use std::str::FromStr;
64 use std::thread::LocalKey;
76 // Per-session global variables: this struct is stored in thread-local storage
77 // in such a way that it is accessible without any kind of handle to all
78 // threads within the compilation session, but is not accessible outside the
80 pub struct SessionGlobals {
81 symbol_interner: Lock<symbol::Interner>,
82 span_interner: Lock<span_encoding::SpanInterner>,
83 hygiene_data: Lock<hygiene::HygieneData>,
84 source_map: Lock<Option<Lrc<SourceMap>>>,
88 pub fn new(edition: Edition) -> SessionGlobals {
90 symbol_interner: Lock::new(symbol::Interner::fresh()),
91 span_interner: Lock::new(span_encoding::SpanInterner::default()),
92 hygiene_data: Lock::new(hygiene::HygieneData::new(edition)),
93 source_map: Lock::new(None),
99 pub fn create_session_globals_then<R>(edition: Edition, f: impl FnOnce() -> R) -> R {
101 !SESSION_GLOBALS.is_set(),
102 "SESSION_GLOBALS should never be overwritten! \
103 Use another thread if you need another SessionGlobals"
105 let session_globals = SessionGlobals::new(edition);
106 SESSION_GLOBALS.set(&session_globals, f)
110 pub fn set_session_globals_then<R>(session_globals: &SessionGlobals, f: impl FnOnce() -> R) -> R {
112 !SESSION_GLOBALS.is_set(),
113 "SESSION_GLOBALS should never be overwritten! \
114 Use another thread if you need another SessionGlobals"
116 SESSION_GLOBALS.set(session_globals, f)
120 pub fn create_default_session_if_not_set_then<R, F>(f: F) -> R
122 F: FnOnce(&SessionGlobals) -> R,
124 create_session_if_not_set_then(edition::DEFAULT_EDITION, f)
128 pub fn create_session_if_not_set_then<R, F>(edition: Edition, f: F) -> R
130 F: FnOnce(&SessionGlobals) -> R,
132 if !SESSION_GLOBALS.is_set() {
133 let session_globals = SessionGlobals::new(edition);
134 SESSION_GLOBALS.set(&session_globals, || SESSION_GLOBALS.with(f))
136 SESSION_GLOBALS.with(f)
141 pub fn with_session_globals<R, F>(f: F) -> R
143 F: FnOnce(&SessionGlobals) -> R,
145 SESSION_GLOBALS.with(f)
149 pub fn create_default_session_globals_then<R>(f: impl FnOnce() -> R) -> R {
150 create_session_globals_then(edition::DEFAULT_EDITION, f)
153 // If this ever becomes non thread-local, `decode_syntax_context`
154 // and `decode_expn_id` will need to be updated to handle concurrent
156 scoped_tls::scoped_thread_local!(static SESSION_GLOBALS: SessionGlobals);
158 // FIXME: We should use this enum or something like it to get rid of the
159 // use of magic `/rust/1.x/...` paths across the board.
160 #[derive(Debug, Eq, PartialEq, Clone, Ord, PartialOrd)]
161 #[derive(HashStable_Generic, Decodable)]
162 pub enum RealFileName {
164 /// For remapped paths (namely paths into libstd that have been mapped
165 /// to the appropriate spot on the local host's file system, and local file
166 /// system paths that have been remapped with `FilePathMapping`),
168 /// `local_path` is the (host-dependent) local path to the file. This is
169 /// None if the file was imported from another crate
170 local_path: Option<PathBuf>,
171 /// `virtual_name` is the stable path rustc will store internally within
173 virtual_name: PathBuf,
177 impl Hash for RealFileName {
178 fn hash<H: std::hash::Hasher>(&self, state: &mut H) {
179 // To prevent #70924 from happening again we should only hash the
180 // remapped (virtualized) path if that exists. This is because
181 // virtualized paths to sysroot crates (/rust/$hash or /rust/$version)
182 // remain stable even if the corresponding local_path changes
183 self.remapped_path_if_available().hash(state)
187 // This is functionally identical to #[derive(Encodable)], with the exception of
188 // an added assert statement
189 impl<S: Encoder> Encodable<S> for RealFileName {
190 fn encode(&self, encoder: &mut S) -> Result<(), S::Error> {
191 encoder.emit_enum(|encoder| match *self {
192 RealFileName::LocalPath(ref local_path) => {
193 encoder.emit_enum_variant("LocalPath", 0, 1, |encoder| {
196 .emit_enum_variant_arg(true, |encoder| local_path.encode(encoder))?;
201 RealFileName::Remapped { ref local_path, ref virtual_name } => encoder
202 .emit_enum_variant("Remapped", 1, 2, |encoder| {
203 // For privacy and build reproducibility, we must not embed host-dependant path in artifacts
204 // if they have been remapped by --remap-path-prefix
205 assert!(local_path.is_none());
208 .emit_enum_variant_arg(true, |encoder| local_path.encode(encoder))?;
210 .emit_enum_variant_arg(false, |encoder| virtual_name.encode(encoder))?;
218 /// Returns the path suitable for reading from the file system on the local host,
219 /// if this information exists.
220 /// Avoid embedding this in build artifacts; see `remapped_path_if_available()` for that.
221 pub fn local_path(&self) -> Option<&Path> {
223 RealFileName::LocalPath(p) => Some(p),
224 RealFileName::Remapped { local_path: p, virtual_name: _ } => {
225 p.as_ref().map(PathBuf::as_path)
230 /// Returns the path suitable for reading from the file system on the local host,
231 /// if this information exists.
232 /// Avoid embedding this in build artifacts; see `remapped_path_if_available()` for that.
233 pub fn into_local_path(self) -> Option<PathBuf> {
235 RealFileName::LocalPath(p) => Some(p),
236 RealFileName::Remapped { local_path: p, virtual_name: _ } => p,
240 /// Returns the path suitable for embedding into build artifacts. This would still
241 /// be a local path if it has not been remapped. A remapped path will not correspond
242 /// to a valid file system path: see `local_path_if_available()` for something that
243 /// is more likely to return paths into the local host file system.
244 pub fn remapped_path_if_available(&self) -> &Path {
246 RealFileName::LocalPath(p)
247 | RealFileName::Remapped { local_path: _, virtual_name: p } => &p,
251 /// Returns the path suitable for reading from the file system on the local host,
252 /// if this information exists. Otherwise returns the remapped name.
253 /// Avoid embedding this in build artifacts; see `remapped_path_if_available()` for that.
254 pub fn local_path_if_available(&self) -> &Path {
256 RealFileName::LocalPath(path)
257 | RealFileName::Remapped { local_path: None, virtual_name: path }
258 | RealFileName::Remapped { local_path: Some(path), virtual_name: _ } => path,
262 pub fn to_string_lossy(&self, prefer_local: bool) -> Cow<'_, str> {
264 self.local_path_if_available().to_string_lossy()
266 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(HashStable_Generic, 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 pub struct FileNameDisplay<'a> {
307 impl fmt::Display for FileNameDisplay<'_> {
308 fn fmt(&self, fmt: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
312 write!(fmt, "{}", name.to_string_lossy(self.prefer_local))
314 QuoteExpansion(_) => write!(fmt, "<quote expansion>"),
315 MacroExpansion(_) => write!(fmt, "<macro expansion>"),
316 Anon(_) => write!(fmt, "<anon>"),
317 ProcMacroSourceCode(_) => write!(fmt, "<proc-macro source code>"),
318 CfgSpec(_) => write!(fmt, "<cfgspec>"),
319 CliCrateAttr(_) => write!(fmt, "<crate attribute>"),
320 Custom(ref s) => write!(fmt, "<{}>", s),
321 DocTest(ref path, _) => write!(fmt, "{}", path.display()),
322 InlineAsm(_) => write!(fmt, "<inline asm>"),
327 impl FileNameDisplay<'_> {
328 pub fn to_string_lossy(&self) -> Cow<'_, str> {
330 FileName::Real(ref inner) => inner.to_string_lossy(self.prefer_local),
331 _ => Cow::from(format!("{}", self)),
337 pub fn is_real(&self) -> bool {
343 | ProcMacroSourceCode(_)
349 | InlineAsm(_) => false,
353 pub fn prefer_remapped(&self) -> FileNameDisplay<'_> {
354 FileNameDisplay { inner: self, prefer_local: false }
357 // This may include transient local filesystem information.
358 // Must not be embedded in build outputs.
359 pub fn prefer_local(&self) -> FileNameDisplay<'_> {
360 FileNameDisplay { inner: self, prefer_local: true }
363 pub fn macro_expansion_source_code(src: &str) -> FileName {
364 let mut hasher = StableHasher::new();
365 src.hash(&mut hasher);
366 FileName::MacroExpansion(hasher.finish())
369 pub fn anon_source_code(src: &str) -> FileName {
370 let mut hasher = StableHasher::new();
371 src.hash(&mut hasher);
372 FileName::Anon(hasher.finish())
375 pub fn proc_macro_source_code(src: &str) -> FileName {
376 let mut hasher = StableHasher::new();
377 src.hash(&mut hasher);
378 FileName::ProcMacroSourceCode(hasher.finish())
381 pub fn cfg_spec_source_code(src: &str) -> FileName {
382 let mut hasher = StableHasher::new();
383 src.hash(&mut hasher);
384 FileName::QuoteExpansion(hasher.finish())
387 pub fn cli_crate_attr_source_code(src: &str) -> FileName {
388 let mut hasher = StableHasher::new();
389 src.hash(&mut hasher);
390 FileName::CliCrateAttr(hasher.finish())
393 pub fn doc_test_source_code(path: PathBuf, line: isize) -> FileName {
394 FileName::DocTest(path, line)
397 pub fn inline_asm_source_code(src: &str) -> FileName {
398 let mut hasher = StableHasher::new();
399 src.hash(&mut hasher);
400 FileName::InlineAsm(hasher.finish())
404 /// Represents a span.
406 /// Spans represent a region of code, used for error reporting. Positions in spans
407 /// are *absolute* positions from the beginning of the [`SourceMap`], not positions
408 /// relative to [`SourceFile`]s. Methods on the `SourceMap` can be used to relate spans back
409 /// to the original source.
411 /// You must be careful if the span crosses more than one file, since you will not be
412 /// able to use many of the functions on spans in source_map and you cannot assume
413 /// that the length of the span is equal to `span.hi - span.lo`; there may be space in the
414 /// [`BytePos`] range between files.
416 /// `SpanData` is public because `Span` uses a thread-local interner and can't be
417 /// sent to other threads, but some pieces of performance infra run in a separate thread.
418 /// Using `Span` is generally preferred.
419 #[derive(Clone, Copy, Hash, PartialEq, Eq, Ord, PartialOrd)]
420 pub struct SpanData {
423 /// Information about where the macro came from, if this piece of
424 /// code was created by a macro expansion.
425 pub ctxt: SyntaxContext,
430 pub fn span(&self) -> Span {
431 Span::new(self.lo, self.hi, self.ctxt)
434 pub fn with_lo(&self, lo: BytePos) -> Span {
435 Span::new(lo, self.hi, self.ctxt)
438 pub fn with_hi(&self, hi: BytePos) -> Span {
439 Span::new(self.lo, hi, self.ctxt)
442 pub fn with_ctxt(&self, ctxt: SyntaxContext) -> Span {
443 Span::new(self.lo, self.hi, ctxt)
447 // The interner is pointed to by a thread local value which is only set on the main thread
448 // with parallelization is disabled. So we don't allow `Span` to transfer between threads
449 // to avoid panics and other errors, even though it would be memory safe to do so.
450 #[cfg(not(parallel_compiler))]
451 impl !Send for Span {}
452 #[cfg(not(parallel_compiler))]
453 impl !Sync for Span {}
455 impl PartialOrd for Span {
456 fn partial_cmp(&self, rhs: &Self) -> Option<Ordering> {
457 PartialOrd::partial_cmp(&self.data(), &rhs.data())
461 fn cmp(&self, rhs: &Self) -> Ordering {
462 Ord::cmp(&self.data(), &rhs.data())
466 /// A collection of `Span`s.
468 /// Spans have two orthogonal attributes:
470 /// - They can be *primary spans*. In this case they are the locus of
471 /// the error, and would be rendered with `^^^`.
472 /// - They can have a *label*. In this case, the label is written next
473 /// to the mark in the snippet when we render.
474 #[derive(Clone, Debug, Hash, PartialEq, Eq, Encodable, Decodable)]
475 pub struct MultiSpan {
476 primary_spans: Vec<Span>,
477 span_labels: Vec<(Span, String)>,
482 pub fn lo(self) -> BytePos {
486 pub fn with_lo(self, lo: BytePos) -> Span {
487 self.data().with_lo(lo)
490 pub fn hi(self) -> BytePos {
494 pub fn with_hi(self, hi: BytePos) -> Span {
495 self.data().with_hi(hi)
498 pub fn ctxt(self) -> SyntaxContext {
502 pub fn with_ctxt(self, ctxt: SyntaxContext) -> Span {
503 self.data().with_ctxt(ctxt)
506 /// Returns `true` if this is a dummy span with any hygienic context.
508 pub fn is_dummy(self) -> bool {
509 let span = self.data();
510 span.lo.0 == 0 && span.hi.0 == 0
513 /// Returns `true` if this span comes from a macro or desugaring.
515 pub fn from_expansion(self) -> bool {
516 self.ctxt() != SyntaxContext::root()
519 /// Returns `true` if `span` originates in a derive-macro's expansion.
520 pub fn in_derive_expansion(self) -> bool {
521 matches!(self.ctxt().outer_expn_data().kind, ExpnKind::Macro(MacroKind::Derive, _))
525 pub fn with_root_ctxt(lo: BytePos, hi: BytePos) -> Span {
526 Span::new(lo, hi, SyntaxContext::root())
529 /// Returns a new span representing an empty span at the beginning of this span.
531 pub fn shrink_to_lo(self) -> Span {
532 let span = self.data();
533 span.with_hi(span.lo)
535 /// Returns a new span representing an empty span at the end of this span.
537 pub fn shrink_to_hi(self) -> Span {
538 let span = self.data();
539 span.with_lo(span.hi)
543 /// Returns `true` if `hi == lo`.
544 pub fn is_empty(&self) -> bool {
545 let span = self.data();
549 /// Returns `self` if `self` is not the dummy span, and `other` otherwise.
550 pub fn substitute_dummy(self, other: Span) -> Span {
551 if self.is_dummy() { other } else { self }
554 /// Returns `true` if `self` fully encloses `other`.
555 pub fn contains(self, other: Span) -> bool {
556 let span = self.data();
557 let other = other.data();
558 span.lo <= other.lo && other.hi <= span.hi
561 /// Returns `true` if `self` touches `other`.
562 pub fn overlaps(self, other: Span) -> bool {
563 let span = self.data();
564 let other = other.data();
565 span.lo < other.hi && other.lo < span.hi
568 /// Returns `true` if the spans are equal with regards to the source text.
570 /// Use this instead of `==` when either span could be generated code,
571 /// and you only care that they point to the same bytes of source text.
572 pub fn source_equal(&self, other: &Span) -> bool {
573 let span = self.data();
574 let other = other.data();
575 span.lo == other.lo && span.hi == other.hi
578 /// Returns `Some(span)`, where the start is trimmed by the end of `other`.
579 pub fn trim_start(self, other: Span) -> Option<Span> {
580 let span = self.data();
581 let other = other.data();
582 if span.hi > other.hi { Some(span.with_lo(cmp::max(span.lo, other.hi))) } else { None }
585 /// Returns the source span -- this is either the supplied span, or the span for
586 /// the macro callsite that expanded to it.
587 pub fn source_callsite(self) -> Span {
588 let expn_data = self.ctxt().outer_expn_data();
589 if !expn_data.is_root() { expn_data.call_site.source_callsite() } else { self }
592 /// The `Span` for the tokens in the previous macro expansion from which `self` was generated,
594 pub fn parent(self) -> Option<Span> {
595 let expn_data = self.ctxt().outer_expn_data();
596 if !expn_data.is_root() { Some(expn_data.call_site) } else { None }
599 /// Edition of the crate from which this span came.
600 pub fn edition(self) -> edition::Edition {
601 self.ctxt().edition()
605 pub fn rust_2015(&self) -> bool {
606 self.edition() == edition::Edition::Edition2015
610 pub fn rust_2018(&self) -> bool {
611 self.edition() >= edition::Edition::Edition2018
615 pub fn rust_2021(&self) -> bool {
616 self.edition() >= edition::Edition::Edition2021
619 /// Returns the source callee.
621 /// Returns `None` if the supplied span has no expansion trace,
622 /// else returns the `ExpnData` for the macro definition
623 /// corresponding to the source callsite.
624 pub fn source_callee(self) -> Option<ExpnData> {
625 fn source_callee(expn_data: ExpnData) -> ExpnData {
626 let next_expn_data = expn_data.call_site.ctxt().outer_expn_data();
627 if !next_expn_data.is_root() { source_callee(next_expn_data) } else { expn_data }
629 let expn_data = self.ctxt().outer_expn_data();
630 if !expn_data.is_root() { Some(source_callee(expn_data)) } else { None }
633 /// Checks if a span is "internal" to a macro in which `#[unstable]`
634 /// items can be used (that is, a macro marked with
635 /// `#[allow_internal_unstable]`).
636 pub fn allows_unstable(&self, feature: Symbol) -> bool {
639 .allow_internal_unstable
640 .map_or(false, |features| features.iter().any(|&f| f == feature))
643 /// Checks if this span arises from a compiler desugaring of kind `kind`.
644 pub fn is_desugaring(&self, kind: DesugaringKind) -> bool {
645 match self.ctxt().outer_expn_data().kind {
646 ExpnKind::Desugaring(k) => k == kind,
651 /// Returns the compiler desugaring that created this span, or `None`
652 /// if this span is not from a desugaring.
653 pub fn desugaring_kind(&self) -> Option<DesugaringKind> {
654 match self.ctxt().outer_expn_data().kind {
655 ExpnKind::Desugaring(k) => Some(k),
660 /// Checks if a span is "internal" to a macro in which `unsafe`
661 /// can be used without triggering the `unsafe_code` lint.
662 // (that is, a macro marked with `#[allow_internal_unsafe]`).
663 pub fn allows_unsafe(&self) -> bool {
664 self.ctxt().outer_expn_data().allow_internal_unsafe
667 pub fn macro_backtrace(mut self) -> impl Iterator<Item = ExpnData> {
668 let mut prev_span = DUMMY_SP;
669 std::iter::from_fn(move || {
671 let expn_data = self.ctxt().outer_expn_data();
672 if expn_data.is_root() {
676 let is_recursive = expn_data.call_site.source_equal(&prev_span);
679 self = expn_data.call_site;
681 // Don't print recursive invocations.
683 return Some(expn_data);
689 /// Returns a `Span` that would enclose both `self` and `end`.
693 /// self lorem ipsum end
694 /// ^^^^^^^^^^^^^^^^^^^^
696 pub fn to(self, end: Span) -> Span {
697 let span_data = self.data();
698 let end_data = end.data();
699 // FIXME(jseyfried): `self.ctxt` should always equal `end.ctxt` here (cf. issue #23480).
700 // Return the macro span on its own to avoid weird diagnostic output. It is preferable to
701 // have an incomplete span than a completely nonsensical one.
702 if span_data.ctxt != end_data.ctxt {
703 if span_data.ctxt == SyntaxContext::root() {
705 } else if end_data.ctxt == SyntaxContext::root() {
708 // Both spans fall within a macro.
709 // FIXME(estebank): check if it is the *same* macro.
712 cmp::min(span_data.lo, end_data.lo),
713 cmp::max(span_data.hi, end_data.hi),
714 if span_data.ctxt == SyntaxContext::root() { end_data.ctxt } else { span_data.ctxt },
718 /// Returns a `Span` between the end of `self` to the beginning of `end`.
722 /// self lorem ipsum end
725 pub fn between(self, end: Span) -> Span {
726 let span = self.data();
727 let end = end.data();
731 if end.ctxt == SyntaxContext::root() { end.ctxt } else { span.ctxt },
735 /// Returns a `Span` from the beginning of `self` until the beginning of `end`.
739 /// self lorem ipsum end
740 /// ^^^^^^^^^^^^^^^^^
742 pub fn until(self, end: Span) -> Span {
743 let span = self.data();
744 let end = end.data();
748 if end.ctxt == SyntaxContext::root() { end.ctxt } else { span.ctxt },
752 pub fn from_inner(self, inner: InnerSpan) -> Span {
753 let span = self.data();
755 span.lo + BytePos::from_usize(inner.start),
756 span.lo + BytePos::from_usize(inner.end),
761 /// Equivalent of `Span::def_site` from the proc macro API,
762 /// except that the location is taken from the `self` span.
763 pub fn with_def_site_ctxt(self, expn_id: ExpnId) -> Span {
764 self.with_ctxt_from_mark(expn_id, Transparency::Opaque)
767 /// Equivalent of `Span::call_site` from the proc macro API,
768 /// except that the location is taken from the `self` span.
769 pub fn with_call_site_ctxt(&self, expn_id: ExpnId) -> Span {
770 self.with_ctxt_from_mark(expn_id, Transparency::Transparent)
773 /// Equivalent of `Span::mixed_site` from the proc macro API,
774 /// except that the location is taken from the `self` span.
775 pub fn with_mixed_site_ctxt(&self, expn_id: ExpnId) -> Span {
776 self.with_ctxt_from_mark(expn_id, Transparency::SemiTransparent)
779 /// Produces a span with the same location as `self` and context produced by a macro with the
780 /// given ID and transparency, assuming that macro was defined directly and not produced by
781 /// some other macro (which is the case for built-in and procedural macros).
782 pub fn with_ctxt_from_mark(self, expn_id: ExpnId, transparency: Transparency) -> Span {
783 self.with_ctxt(SyntaxContext::root().apply_mark(expn_id, transparency))
787 pub fn apply_mark(self, expn_id: ExpnId, transparency: Transparency) -> Span {
788 let span = self.data();
789 span.with_ctxt(span.ctxt.apply_mark(expn_id, transparency))
793 pub fn remove_mark(&mut self) -> ExpnId {
794 let mut span = self.data();
795 let mark = span.ctxt.remove_mark();
796 *self = Span::new(span.lo, span.hi, span.ctxt);
801 pub fn adjust(&mut self, expn_id: ExpnId) -> Option<ExpnId> {
802 let mut span = self.data();
803 let mark = span.ctxt.adjust(expn_id);
804 *self = Span::new(span.lo, span.hi, span.ctxt);
809 pub fn normalize_to_macros_2_0_and_adjust(&mut self, expn_id: ExpnId) -> Option<ExpnId> {
810 let mut span = self.data();
811 let mark = span.ctxt.normalize_to_macros_2_0_and_adjust(expn_id);
812 *self = Span::new(span.lo, span.hi, span.ctxt);
817 pub fn glob_adjust(&mut self, expn_id: ExpnId, glob_span: Span) -> Option<Option<ExpnId>> {
818 let mut span = self.data();
819 let mark = span.ctxt.glob_adjust(expn_id, glob_span);
820 *self = Span::new(span.lo, span.hi, span.ctxt);
825 pub fn reverse_glob_adjust(
829 ) -> Option<Option<ExpnId>> {
830 let mut span = self.data();
831 let mark = span.ctxt.reverse_glob_adjust(expn_id, glob_span);
832 *self = Span::new(span.lo, span.hi, span.ctxt);
837 pub fn normalize_to_macros_2_0(self) -> Span {
838 let span = self.data();
839 span.with_ctxt(span.ctxt.normalize_to_macros_2_0())
843 pub fn normalize_to_macro_rules(self) -> Span {
844 let span = self.data();
845 span.with_ctxt(span.ctxt.normalize_to_macro_rules())
849 /// A span together with some additional data.
850 #[derive(Clone, Debug)]
851 pub struct SpanLabel {
852 /// The span we are going to include in the final snippet.
855 /// Is this a primary span? This is the "locus" of the message,
856 /// and is indicated with a `^^^^` underline, versus `----`.
857 pub is_primary: bool,
859 /// What label should we attach to this span (if any)?
860 pub label: Option<String>,
863 impl Default for Span {
864 fn default() -> Self {
869 impl<E: Encoder> Encodable<E> for Span {
870 default fn encode(&self, s: &mut E) -> Result<(), E::Error> {
871 let span = self.data();
872 s.emit_struct(false, |s| {
873 s.emit_struct_field("lo", true, |s| span.lo.encode(s))?;
874 s.emit_struct_field("hi", false, |s| span.hi.encode(s))
878 impl<D: Decoder> Decodable<D> for Span {
879 default fn decode(s: &mut D) -> Result<Span, D::Error> {
881 let lo = d.read_struct_field("lo", Decodable::decode)?;
882 let hi = d.read_struct_field("hi", Decodable::decode)?;
884 Ok(Span::new(lo, hi, SyntaxContext::root()))
889 /// Calls the provided closure, using the provided `SourceMap` to format
890 /// any spans that are debug-printed during the closure's execution.
892 /// Normally, the global `TyCtxt` is used to retrieve the `SourceMap`
893 /// (see `rustc_interface::callbacks::span_debug1`). However, some parts
894 /// of the compiler (e.g. `rustc_parse`) may debug-print `Span`s before
895 /// a `TyCtxt` is available. In this case, we fall back to
896 /// the `SourceMap` provided to this function. If that is not available,
897 /// we fall back to printing the raw `Span` field values.
898 pub fn with_source_map<T, F: FnOnce() -> T>(source_map: Lrc<SourceMap>, f: F) -> T {
899 with_session_globals(|session_globals| {
900 *session_globals.source_map.borrow_mut() = Some(source_map);
902 struct ClearSourceMap;
903 impl Drop for ClearSourceMap {
905 with_session_globals(|session_globals| {
906 session_globals.source_map.borrow_mut().take();
911 let _guard = ClearSourceMap;
915 pub fn debug_with_source_map(
917 f: &mut fmt::Formatter<'_>,
918 source_map: &SourceMap,
920 write!(f, "{} ({:?})", source_map.span_to_diagnostic_string(span), span.ctxt())
923 pub fn default_span_debug(span: Span, f: &mut fmt::Formatter<'_>) -> fmt::Result {
924 with_session_globals(|session_globals| {
925 if let Some(source_map) = &*session_globals.source_map.borrow() {
926 debug_with_source_map(span, f, source_map)
928 f.debug_struct("Span")
929 .field("lo", &span.lo())
930 .field("hi", &span.hi())
931 .field("ctxt", &span.ctxt())
937 impl fmt::Debug for Span {
938 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
939 (*SPAN_DEBUG)(*self, f)
943 impl fmt::Debug for SpanData {
944 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
945 (*SPAN_DEBUG)(Span::new(self.lo, self.hi, self.ctxt), f)
951 pub fn new() -> MultiSpan {
952 MultiSpan { primary_spans: vec![], span_labels: vec![] }
955 pub fn from_span(primary_span: Span) -> MultiSpan {
956 MultiSpan { primary_spans: vec![primary_span], span_labels: vec![] }
959 pub fn from_spans(mut vec: Vec<Span>) -> MultiSpan {
961 MultiSpan { primary_spans: vec, span_labels: vec![] }
964 pub fn push_span_label(&mut self, span: Span, label: String) {
965 self.span_labels.push((span, label));
968 /// Selects the first primary span (if any).
969 pub fn primary_span(&self) -> Option<Span> {
970 self.primary_spans.first().cloned()
973 /// Returns all primary spans.
974 pub fn primary_spans(&self) -> &[Span] {
978 /// Returns `true` if any of the primary spans are displayable.
979 pub fn has_primary_spans(&self) -> bool {
980 self.primary_spans.iter().any(|sp| !sp.is_dummy())
983 /// Returns `true` if this contains only a dummy primary span with any hygienic context.
984 pub fn is_dummy(&self) -> bool {
985 let mut is_dummy = true;
986 for span in &self.primary_spans {
987 if !span.is_dummy() {
994 /// Replaces all occurrences of one Span with another. Used to move `Span`s in areas that don't
995 /// display well (like std macros). Returns whether replacements occurred.
996 pub fn replace(&mut self, before: Span, after: Span) -> bool {
997 let mut replacements_occurred = false;
998 for primary_span in &mut self.primary_spans {
999 if *primary_span == before {
1000 *primary_span = after;
1001 replacements_occurred = true;
1004 for span_label in &mut self.span_labels {
1005 if span_label.0 == before {
1006 span_label.0 = after;
1007 replacements_occurred = true;
1010 replacements_occurred
1013 /// Returns the strings to highlight. We always ensure that there
1014 /// is an entry for each of the primary spans -- for each primary
1015 /// span `P`, if there is at least one label with span `P`, we return
1016 /// those labels (marked as primary). But otherwise we return
1017 /// `SpanLabel` instances with empty labels.
1018 pub fn span_labels(&self) -> Vec<SpanLabel> {
1019 let is_primary = |span| self.primary_spans.contains(&span);
1021 let mut span_labels = self
1024 .map(|&(span, ref label)| SpanLabel {
1026 is_primary: is_primary(span),
1027 label: Some(label.clone()),
1029 .collect::<Vec<_>>();
1031 for &span in &self.primary_spans {
1032 if !span_labels.iter().any(|sl| sl.span == span) {
1033 span_labels.push(SpanLabel { span, is_primary: true, label: None });
1040 /// Returns `true` if any of the span labels is displayable.
1041 pub fn has_span_labels(&self) -> bool {
1042 self.span_labels.iter().any(|(sp, _)| !sp.is_dummy())
1046 impl From<Span> for MultiSpan {
1047 fn from(span: Span) -> MultiSpan {
1048 MultiSpan::from_span(span)
1052 impl From<Vec<Span>> for MultiSpan {
1053 fn from(spans: Vec<Span>) -> MultiSpan {
1054 MultiSpan::from_spans(spans)
1058 /// Identifies an offset of a multi-byte character in a `SourceFile`.
1059 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1060 pub struct MultiByteChar {
1061 /// The absolute offset of the character in the `SourceMap`.
1063 /// The number of bytes, `>= 2`.
1067 /// Identifies an offset of a non-narrow character in a `SourceFile`.
1068 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1069 pub enum NonNarrowChar {
1070 /// Represents a zero-width character.
1072 /// Represents a wide (full-width) character.
1074 /// Represents a tab character, represented visually with a width of 4 characters.
1078 impl NonNarrowChar {
1079 fn new(pos: BytePos, width: usize) -> Self {
1081 0 => NonNarrowChar::ZeroWidth(pos),
1082 2 => NonNarrowChar::Wide(pos),
1083 4 => NonNarrowChar::Tab(pos),
1084 _ => panic!("width {} given for non-narrow character", width),
1088 /// Returns the absolute offset of the character in the `SourceMap`.
1089 pub fn pos(&self) -> BytePos {
1091 NonNarrowChar::ZeroWidth(p) | NonNarrowChar::Wide(p) | NonNarrowChar::Tab(p) => p,
1095 /// Returns the width of the character, 0 (zero-width) or 2 (wide).
1096 pub fn width(&self) -> usize {
1098 NonNarrowChar::ZeroWidth(_) => 0,
1099 NonNarrowChar::Wide(_) => 2,
1100 NonNarrowChar::Tab(_) => 4,
1105 impl Add<BytePos> for NonNarrowChar {
1108 fn add(self, rhs: BytePos) -> Self {
1110 NonNarrowChar::ZeroWidth(pos) => NonNarrowChar::ZeroWidth(pos + rhs),
1111 NonNarrowChar::Wide(pos) => NonNarrowChar::Wide(pos + rhs),
1112 NonNarrowChar::Tab(pos) => NonNarrowChar::Tab(pos + rhs),
1117 impl Sub<BytePos> for NonNarrowChar {
1120 fn sub(self, rhs: BytePos) -> Self {
1122 NonNarrowChar::ZeroWidth(pos) => NonNarrowChar::ZeroWidth(pos - rhs),
1123 NonNarrowChar::Wide(pos) => NonNarrowChar::Wide(pos - rhs),
1124 NonNarrowChar::Tab(pos) => NonNarrowChar::Tab(pos - rhs),
1129 /// Identifies an offset of a character that was normalized away from `SourceFile`.
1130 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1131 pub struct NormalizedPos {
1132 /// The absolute offset of the character in the `SourceMap`.
1134 /// The difference between original and normalized string at position.
1138 #[derive(PartialEq, Eq, Clone, Debug)]
1139 pub enum ExternalSource {
1140 /// No external source has to be loaded, since the `SourceFile` represents a local crate.
1143 kind: ExternalSourceKind,
1144 /// This SourceFile's byte-offset within the source_map of its original crate.
1145 original_start_pos: BytePos,
1146 /// The end of this SourceFile within the source_map of its original crate.
1147 original_end_pos: BytePos,
1151 /// The state of the lazy external source loading mechanism of a `SourceFile`.
1152 #[derive(PartialEq, Eq, Clone, Debug)]
1153 pub enum ExternalSourceKind {
1154 /// The external source has been loaded already.
1155 Present(Lrc<String>),
1156 /// No attempt has been made to load the external source.
1158 /// A failed attempt has been made to load the external source.
1163 impl ExternalSource {
1164 pub fn get_source(&self) -> Option<&Lrc<String>> {
1166 ExternalSource::Foreign { kind: ExternalSourceKind::Present(ref src), .. } => Some(src),
1173 pub struct OffsetOverflowError;
1175 #[derive(Copy, Clone, Debug, PartialEq, Eq, PartialOrd, Ord, Hash, Encodable, Decodable)]
1176 pub enum SourceFileHashAlgorithm {
1182 impl FromStr for SourceFileHashAlgorithm {
1185 fn from_str(s: &str) -> Result<SourceFileHashAlgorithm, ()> {
1187 "md5" => Ok(SourceFileHashAlgorithm::Md5),
1188 "sha1" => Ok(SourceFileHashAlgorithm::Sha1),
1189 "sha256" => Ok(SourceFileHashAlgorithm::Sha256),
1195 rustc_data_structures::impl_stable_hash_via_hash!(SourceFileHashAlgorithm);
1197 /// The hash of the on-disk source file used for debug info.
1198 #[derive(Copy, Clone, PartialEq, Eq, Debug)]
1199 #[derive(HashStable_Generic, Encodable, Decodable)]
1200 pub struct SourceFileHash {
1201 pub kind: SourceFileHashAlgorithm,
1205 impl SourceFileHash {
1206 pub fn new(kind: SourceFileHashAlgorithm, src: &str) -> SourceFileHash {
1207 let mut hash = SourceFileHash { kind, value: Default::default() };
1208 let len = hash.hash_len();
1209 let value = &mut hash.value[..len];
1210 let data = src.as_bytes();
1212 SourceFileHashAlgorithm::Md5 => {
1213 value.copy_from_slice(&Md5::digest(data));
1215 SourceFileHashAlgorithm::Sha1 => {
1216 value.copy_from_slice(&Sha1::digest(data));
1218 SourceFileHashAlgorithm::Sha256 => {
1219 value.copy_from_slice(&Sha256::digest(data));
1225 /// Check if the stored hash matches the hash of the string.
1226 pub fn matches(&self, src: &str) -> bool {
1227 Self::new(self.kind, src) == *self
1230 /// The bytes of the hash.
1231 pub fn hash_bytes(&self) -> &[u8] {
1232 let len = self.hash_len();
1236 fn hash_len(&self) -> usize {
1238 SourceFileHashAlgorithm::Md5 => 16,
1239 SourceFileHashAlgorithm::Sha1 => 20,
1240 SourceFileHashAlgorithm::Sha256 => 32,
1245 /// A single source in the [`SourceMap`].
1247 pub struct SourceFile {
1248 /// The name of the file that the source came from. Source that doesn't
1249 /// originate from files has names between angle brackets by convention
1250 /// (e.g., `<anon>`).
1252 /// The complete source code.
1253 pub src: Option<Lrc<String>>,
1254 /// The source code's hash.
1255 pub src_hash: SourceFileHash,
1256 /// The external source code (used for external crates, which will have a `None`
1257 /// value as `self.src`.
1258 pub external_src: Lock<ExternalSource>,
1259 /// The start position of this source in the `SourceMap`.
1260 pub start_pos: BytePos,
1261 /// The end position of this source in the `SourceMap`.
1262 pub end_pos: BytePos,
1263 /// Locations of lines beginnings in the source code.
1264 pub lines: Vec<BytePos>,
1265 /// Locations of multi-byte characters in the source code.
1266 pub multibyte_chars: Vec<MultiByteChar>,
1267 /// Width of characters that are not narrow in the source code.
1268 pub non_narrow_chars: Vec<NonNarrowChar>,
1269 /// Locations of characters removed during normalization.
1270 pub normalized_pos: Vec<NormalizedPos>,
1271 /// A hash of the filename, used for speeding up hashing in incremental compilation.
1272 pub name_hash: u128,
1273 /// Indicates which crate this `SourceFile` was imported from.
1277 impl<S: Encoder> Encodable<S> for SourceFile {
1278 fn encode(&self, s: &mut S) -> Result<(), S::Error> {
1279 s.emit_struct(false, |s| {
1280 s.emit_struct_field("name", true, |s| self.name.encode(s))?;
1281 s.emit_struct_field("src_hash", false, |s| self.src_hash.encode(s))?;
1282 s.emit_struct_field("start_pos", false, |s| self.start_pos.encode(s))?;
1283 s.emit_struct_field("end_pos", false, |s| self.end_pos.encode(s))?;
1284 s.emit_struct_field("lines", false, |s| {
1285 let lines = &self.lines[..];
1286 // Store the length.
1287 s.emit_u32(lines.len() as u32)?;
1289 if !lines.is_empty() {
1290 // In order to preserve some space, we exploit the fact that
1291 // the lines list is sorted and individual lines are
1292 // probably not that long. Because of that we can store lines
1293 // as a difference list, using as little space as possible
1294 // for the differences.
1295 let max_line_length = if lines.len() == 1 {
1300 .map(|&[fst, snd]| snd - fst)
1301 .map(|bp| bp.to_usize())
1306 let bytes_per_diff: u8 = match max_line_length {
1308 0x100..=0xFFFF => 2,
1312 // Encode the number of bytes used per diff.
1313 bytes_per_diff.encode(s)?;
1315 // Encode the first element.
1316 lines[0].encode(s)?;
1318 let diff_iter = lines[..].array_windows().map(|&[fst, snd]| snd - fst);
1320 match bytes_per_diff {
1322 for diff in diff_iter {
1323 (diff.0 as u8).encode(s)?
1327 for diff in diff_iter {
1328 (diff.0 as u16).encode(s)?
1332 for diff in diff_iter {
1336 _ => unreachable!(),
1342 s.emit_struct_field("multibyte_chars", false, |s| self.multibyte_chars.encode(s))?;
1343 s.emit_struct_field("non_narrow_chars", false, |s| self.non_narrow_chars.encode(s))?;
1344 s.emit_struct_field("name_hash", false, |s| self.name_hash.encode(s))?;
1345 s.emit_struct_field("normalized_pos", false, |s| self.normalized_pos.encode(s))?;
1346 s.emit_struct_field("cnum", false, |s| self.cnum.encode(s))
1351 impl<D: Decoder> Decodable<D> for SourceFile {
1352 fn decode(d: &mut D) -> Result<SourceFile, D::Error> {
1354 let name: FileName = d.read_struct_field("name", |d| Decodable::decode(d))?;
1355 let src_hash: SourceFileHash =
1356 d.read_struct_field("src_hash", |d| Decodable::decode(d))?;
1357 let start_pos: BytePos = d.read_struct_field("start_pos", |d| Decodable::decode(d))?;
1358 let end_pos: BytePos = d.read_struct_field("end_pos", |d| Decodable::decode(d))?;
1359 let lines: Vec<BytePos> = d.read_struct_field("lines", |d| {
1360 let num_lines: u32 = Decodable::decode(d)?;
1361 let mut lines = Vec::with_capacity(num_lines as usize);
1364 // Read the number of bytes used per diff.
1365 let bytes_per_diff: u8 = Decodable::decode(d)?;
1367 // Read the first element.
1368 let mut line_start: BytePos = Decodable::decode(d)?;
1369 lines.push(line_start);
1371 for _ in 1..num_lines {
1372 let diff = match bytes_per_diff {
1373 1 => d.read_u8()? as u32,
1374 2 => d.read_u16()? as u32,
1376 _ => unreachable!(),
1379 line_start = line_start + BytePos(diff);
1381 lines.push(line_start);
1387 let multibyte_chars: Vec<MultiByteChar> =
1388 d.read_struct_field("multibyte_chars", |d| Decodable::decode(d))?;
1389 let non_narrow_chars: Vec<NonNarrowChar> =
1390 d.read_struct_field("non_narrow_chars", |d| Decodable::decode(d))?;
1391 let name_hash: u128 = d.read_struct_field("name_hash", |d| Decodable::decode(d))?;
1392 let normalized_pos: Vec<NormalizedPos> =
1393 d.read_struct_field("normalized_pos", |d| Decodable::decode(d))?;
1394 let cnum: CrateNum = d.read_struct_field("cnum", |d| Decodable::decode(d))?;
1401 // Unused - the metadata decoder will construct
1402 // a new SourceFile, filling in `external_src` properly
1403 external_src: Lock::new(ExternalSource::Unneeded),
1415 impl fmt::Debug for SourceFile {
1416 fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result {
1417 write!(fmt, "SourceFile({:?})", self.name)
1426 hash_kind: SourceFileHashAlgorithm,
1428 // Compute the file hash before any normalization.
1429 let src_hash = SourceFileHash::new(hash_kind, &src);
1430 let normalized_pos = normalize_src(&mut src, start_pos);
1433 let mut hasher: StableHasher = StableHasher::new();
1434 name.hash(&mut hasher);
1435 hasher.finish::<u128>()
1437 let end_pos = start_pos.to_usize() + src.len();
1438 assert!(end_pos <= u32::MAX as usize);
1440 let (lines, multibyte_chars, non_narrow_chars) =
1441 analyze_source_file::analyze_source_file(&src[..], start_pos);
1445 src: Some(Lrc::new(src)),
1447 external_src: Lock::new(ExternalSource::Unneeded),
1449 end_pos: Pos::from_usize(end_pos),
1459 /// Returns the `BytePos` of the beginning of the current line.
1460 pub fn line_begin_pos(&self, pos: BytePos) -> BytePos {
1461 let line_index = self.lookup_line(pos).unwrap();
1462 self.lines[line_index]
1465 /// Add externally loaded source.
1466 /// If the hash of the input doesn't match or no input is supplied via None,
1467 /// it is interpreted as an error and the corresponding enum variant is set.
1468 /// The return value signifies whether some kind of source is present.
1469 pub fn add_external_src<F>(&self, get_src: F) -> bool
1471 F: FnOnce() -> Option<String>,
1474 *self.external_src.borrow(),
1475 ExternalSource::Foreign { kind: ExternalSourceKind::AbsentOk, .. }
1477 let src = get_src();
1478 let mut external_src = self.external_src.borrow_mut();
1479 // Check that no-one else have provided the source while we were getting it
1480 if let ExternalSource::Foreign {
1481 kind: src_kind @ ExternalSourceKind::AbsentOk, ..
1482 } = &mut *external_src
1484 if let Some(mut src) = src {
1485 // The src_hash needs to be computed on the pre-normalized src.
1486 if self.src_hash.matches(&src) {
1487 normalize_src(&mut src, BytePos::from_usize(0));
1488 *src_kind = ExternalSourceKind::Present(Lrc::new(src));
1492 *src_kind = ExternalSourceKind::AbsentErr;
1497 self.src.is_some() || external_src.get_source().is_some()
1500 self.src.is_some() || self.external_src.borrow().get_source().is_some()
1504 /// Gets a line from the list of pre-computed line-beginnings.
1505 /// The line number here is 0-based.
1506 pub fn get_line(&self, line_number: usize) -> Option<Cow<'_, str>> {
1507 fn get_until_newline(src: &str, begin: usize) -> &str {
1508 // We can't use `lines.get(line_number+1)` because we might
1509 // be parsing when we call this function and thus the current
1510 // line is the last one we have line info for.
1511 let slice = &src[begin..];
1512 match slice.find('\n') {
1513 Some(e) => &slice[..e],
1519 let line = self.lines.get(line_number)?;
1520 let begin: BytePos = *line - self.start_pos;
1524 if let Some(ref src) = self.src {
1525 Some(Cow::from(get_until_newline(src, begin)))
1526 } else if let Some(src) = self.external_src.borrow().get_source() {
1527 Some(Cow::Owned(String::from(get_until_newline(src, begin))))
1533 pub fn is_real_file(&self) -> bool {
1537 pub fn is_imported(&self) -> bool {
1541 pub fn count_lines(&self) -> usize {
1545 /// Finds the line containing the given position. The return value is the
1546 /// index into the `lines` array of this `SourceFile`, not the 1-based line
1547 /// number. If the source_file is empty or the position is located before the
1548 /// first line, `None` is returned.
1549 pub fn lookup_line(&self, pos: BytePos) -> Option<usize> {
1550 match self.lines.binary_search(&pos) {
1551 Ok(idx) => Some(idx),
1553 Err(idx) => Some(idx - 1),
1557 pub fn line_bounds(&self, line_index: usize) -> Range<BytePos> {
1558 if self.is_empty() {
1559 return self.start_pos..self.end_pos;
1562 assert!(line_index < self.lines.len());
1563 if line_index == (self.lines.len() - 1) {
1564 self.lines[line_index]..self.end_pos
1566 self.lines[line_index]..self.lines[line_index + 1]
1570 /// Returns whether or not the file contains the given `SourceMap` byte
1571 /// position. The position one past the end of the file is considered to be
1572 /// contained by the file. This implies that files for which `is_empty`
1573 /// returns true still contain one byte position according to this function.
1575 pub fn contains(&self, byte_pos: BytePos) -> bool {
1576 byte_pos >= self.start_pos && byte_pos <= self.end_pos
1580 pub fn is_empty(&self) -> bool {
1581 self.start_pos == self.end_pos
1584 /// Calculates the original byte position relative to the start of the file
1585 /// based on the given byte position.
1586 pub fn original_relative_byte_pos(&self, pos: BytePos) -> BytePos {
1587 // Diff before any records is 0. Otherwise use the previously recorded
1588 // diff as that applies to the following characters until a new diff
1590 let diff = match self.normalized_pos.binary_search_by(|np| np.pos.cmp(&pos)) {
1591 Ok(i) => self.normalized_pos[i].diff,
1592 Err(i) if i == 0 => 0,
1593 Err(i) => self.normalized_pos[i - 1].diff,
1596 BytePos::from_u32(pos.0 - self.start_pos.0 + diff)
1599 /// Converts an absolute `BytePos` to a `CharPos` relative to the `SourceFile`.
1600 pub fn bytepos_to_file_charpos(&self, bpos: BytePos) -> CharPos {
1601 // The number of extra bytes due to multibyte chars in the `SourceFile`.
1602 let mut total_extra_bytes = 0;
1604 for mbc in self.multibyte_chars.iter() {
1605 debug!("{}-byte char at {:?}", mbc.bytes, mbc.pos);
1607 // Every character is at least one byte, so we only
1608 // count the actual extra bytes.
1609 total_extra_bytes += mbc.bytes as u32 - 1;
1610 // We should never see a byte position in the middle of a
1612 assert!(bpos.to_u32() >= mbc.pos.to_u32() + mbc.bytes as u32);
1618 assert!(self.start_pos.to_u32() + total_extra_bytes <= bpos.to_u32());
1619 CharPos(bpos.to_usize() - self.start_pos.to_usize() - total_extra_bytes as usize)
1622 /// Looks up the file's (1-based) line number and (0-based `CharPos`) column offset, for a
1623 /// given `BytePos`.
1624 pub fn lookup_file_pos(&self, pos: BytePos) -> (usize, CharPos) {
1625 let chpos = self.bytepos_to_file_charpos(pos);
1626 match self.lookup_line(pos) {
1628 let line = a + 1; // Line numbers start at 1
1629 let linebpos = self.lines[a];
1630 let linechpos = self.bytepos_to_file_charpos(linebpos);
1631 let col = chpos - linechpos;
1632 debug!("byte pos {:?} is on the line at byte pos {:?}", pos, linebpos);
1633 debug!("char pos {:?} is on the line at char pos {:?}", chpos, linechpos);
1634 debug!("byte is on line: {}", line);
1635 assert!(chpos >= linechpos);
1642 /// Looks up the file's (1-based) line number, (0-based `CharPos`) column offset, and (0-based)
1643 /// column offset when displayed, for a given `BytePos`.
1644 pub fn lookup_file_pos_with_col_display(&self, pos: BytePos) -> (usize, CharPos, usize) {
1645 let (line, col_or_chpos) = self.lookup_file_pos(pos);
1647 let col = col_or_chpos;
1648 let linebpos = self.lines[line - 1];
1650 let start_width_idx = self
1652 .binary_search_by_key(&linebpos, |x| x.pos())
1653 .unwrap_or_else(|x| x);
1654 let end_width_idx = self
1656 .binary_search_by_key(&pos, |x| x.pos())
1657 .unwrap_or_else(|x| x);
1658 let special_chars = end_width_idx - start_width_idx;
1659 let non_narrow: usize = self.non_narrow_chars[start_width_idx..end_width_idx]
1663 col.0 - special_chars + non_narrow
1665 (line, col, col_display)
1667 let chpos = col_or_chpos;
1669 let end_width_idx = self
1671 .binary_search_by_key(&pos, |x| x.pos())
1672 .unwrap_or_else(|x| x);
1673 let non_narrow: usize =
1674 self.non_narrow_chars[0..end_width_idx].iter().map(|x| x.width()).sum();
1675 chpos.0 - end_width_idx + non_narrow
1677 (0, chpos, col_display)
1682 /// Normalizes the source code and records the normalizations.
1683 fn normalize_src(src: &mut String, start_pos: BytePos) -> Vec<NormalizedPos> {
1684 let mut normalized_pos = vec![];
1685 remove_bom(src, &mut normalized_pos);
1686 normalize_newlines(src, &mut normalized_pos);
1688 // Offset all the positions by start_pos to match the final file positions.
1689 for np in &mut normalized_pos {
1690 np.pos.0 += start_pos.0;
1696 /// Removes UTF-8 BOM, if any.
1697 fn remove_bom(src: &mut String, normalized_pos: &mut Vec<NormalizedPos>) {
1698 if src.starts_with('\u{feff}') {
1700 normalized_pos.push(NormalizedPos { pos: BytePos(0), diff: 3 });
1704 /// Replaces `\r\n` with `\n` in-place in `src`.
1706 /// Returns error if there's a lone `\r` in the string.
1707 fn normalize_newlines(src: &mut String, normalized_pos: &mut Vec<NormalizedPos>) {
1708 if !src.as_bytes().contains(&b'\r') {
1712 // We replace `\r\n` with `\n` in-place, which doesn't break utf-8 encoding.
1713 // While we *can* call `as_mut_vec` and do surgery on the live string
1714 // directly, let's rather steal the contents of `src`. This makes the code
1715 // safe even if a panic occurs.
1717 let mut buf = std::mem::replace(src, String::new()).into_bytes();
1718 let mut gap_len = 0;
1719 let mut tail = buf.as_mut_slice();
1721 let original_gap = normalized_pos.last().map_or(0, |l| l.diff);
1723 let idx = match find_crlf(&tail[gap_len..]) {
1725 Some(idx) => idx + gap_len,
1727 tail.copy_within(gap_len..idx, 0);
1728 tail = &mut tail[idx - gap_len..];
1729 if tail.len() == gap_len {
1732 cursor += idx - gap_len;
1734 normalized_pos.push(NormalizedPos {
1735 pos: BytePos::from_usize(cursor + 1),
1736 diff: original_gap + gap_len as u32,
1740 // Account for removed `\r`.
1741 // After `set_len`, `buf` is guaranteed to contain utf-8 again.
1742 let new_len = buf.len() - gap_len;
1744 buf.set_len(new_len);
1745 *src = String::from_utf8_unchecked(buf);
1748 fn find_crlf(src: &[u8]) -> Option<usize> {
1749 let mut search_idx = 0;
1750 while let Some(idx) = find_cr(&src[search_idx..]) {
1751 if src[search_idx..].get(idx + 1) != Some(&b'\n') {
1752 search_idx += idx + 1;
1755 return Some(search_idx + idx);
1760 fn find_cr(src: &[u8]) -> Option<usize> {
1761 src.iter().position(|&b| b == b'\r')
1765 // _____________________________________________________________________________
1766 // Pos, BytePos, CharPos
1770 fn from_usize(n: usize) -> Self;
1771 fn to_usize(&self) -> usize;
1772 fn from_u32(n: u32) -> Self;
1773 fn to_u32(&self) -> u32;
1776 macro_rules! impl_pos {
1780 $vis:vis struct $ident:ident($inner_vis:vis $inner_ty:ty);
1785 $vis struct $ident($inner_vis $inner_ty);
1787 impl Pos for $ident {
1789 fn from_usize(n: usize) -> $ident {
1790 $ident(n as $inner_ty)
1794 fn to_usize(&self) -> usize {
1799 fn from_u32(n: u32) -> $ident {
1800 $ident(n as $inner_ty)
1804 fn to_u32(&self) -> u32 {
1809 impl Add for $ident {
1810 type Output = $ident;
1813 fn add(self, rhs: $ident) -> $ident {
1814 $ident(self.0 + rhs.0)
1818 impl Sub for $ident {
1819 type Output = $ident;
1822 fn sub(self, rhs: $ident) -> $ident {
1823 $ident(self.0 - rhs.0)
1833 /// Keep this small (currently 32-bits), as AST contains a lot of them.
1834 #[derive(Clone, Copy, PartialEq, Eq, Hash, PartialOrd, Ord, Debug)]
1835 pub struct BytePos(pub u32);
1837 /// A character offset.
1839 /// Because of multibyte UTF-8 characters, a byte offset
1840 /// is not equivalent to a character offset. The [`SourceMap`] will convert [`BytePos`]
1841 /// values to `CharPos` values as necessary.
1842 #[derive(Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Debug)]
1843 pub struct CharPos(pub usize);
1846 impl<S: rustc_serialize::Encoder> Encodable<S> for BytePos {
1847 fn encode(&self, s: &mut S) -> Result<(), S::Error> {
1852 impl<D: rustc_serialize::Decoder> Decodable<D> for BytePos {
1853 fn decode(d: &mut D) -> Result<BytePos, D::Error> {
1854 Ok(BytePos(d.read_u32()?))
1858 // _____________________________________________________________________________
1859 // Loc, SourceFileAndLine, SourceFileAndBytePos
1862 /// A source code location used for error reporting.
1863 #[derive(Debug, Clone)]
1865 /// Information about the original source.
1866 pub file: Lrc<SourceFile>,
1867 /// The (1-based) line number.
1869 /// The (0-based) column offset.
1871 /// The (0-based) column offset when displayed.
1872 pub col_display: usize,
1875 // Used to be structural records.
1877 pub struct SourceFileAndLine {
1878 pub sf: Lrc<SourceFile>,
1882 pub struct SourceFileAndBytePos {
1883 pub sf: Lrc<SourceFile>,
1887 #[derive(Copy, Clone, Debug, PartialEq, Eq)]
1888 pub struct LineInfo {
1889 /// Index of line, starting from 0.
1890 pub line_index: usize,
1892 /// Column in line where span begins, starting from 0.
1893 pub start_col: CharPos,
1895 /// Column in line where span ends, starting from 0, exclusive.
1896 pub end_col: CharPos,
1899 pub struct FileLines {
1900 pub file: Lrc<SourceFile>,
1901 pub lines: Vec<LineInfo>,
1904 pub static SPAN_DEBUG: AtomicRef<fn(Span, &mut fmt::Formatter<'_>) -> fmt::Result> =
1905 AtomicRef::new(&(default_span_debug as fn(_, &mut fmt::Formatter<'_>) -> _));
1907 // _____________________________________________________________________________
1908 // SpanLinesError, SpanSnippetError, DistinctSources, MalformedSourceMapPositions
1911 pub type FileLinesResult = Result<FileLines, SpanLinesError>;
1913 #[derive(Clone, PartialEq, Eq, Debug)]
1914 pub enum SpanLinesError {
1915 DistinctSources(DistinctSources),
1918 #[derive(Clone, PartialEq, Eq, Debug)]
1919 pub enum SpanSnippetError {
1920 IllFormedSpan(Span),
1921 DistinctSources(DistinctSources),
1922 MalformedForSourcemap(MalformedSourceMapPositions),
1923 SourceNotAvailable { filename: FileName },
1926 #[derive(Clone, PartialEq, Eq, Debug)]
1927 pub struct DistinctSources {
1928 pub begin: (FileName, BytePos),
1929 pub end: (FileName, BytePos),
1932 #[derive(Clone, PartialEq, Eq, Debug)]
1933 pub struct MalformedSourceMapPositions {
1935 pub source_len: usize,
1936 pub begin_pos: BytePos,
1937 pub end_pos: BytePos,
1940 /// Range inside of a `Span` used for diagnostics when we only have access to relative positions.
1941 #[derive(Copy, Clone, PartialEq, Eq, Debug)]
1942 pub struct InnerSpan {
1948 pub fn new(start: usize, end: usize) -> InnerSpan {
1949 InnerSpan { start, end }
1953 /// Requirements for a `StableHashingContext` to be used in this crate.
1955 /// This is a hack to allow using the [`HashStable_Generic`] derive macro
1956 /// instead of implementing everything in rustc_middle.
1957 pub trait HashStableContext {
1958 fn def_path_hash(&self, def_id: DefId) -> DefPathHash;
1959 /// Obtains a cache for storing the `Fingerprint` of an `ExpnId`.
1960 /// This method allows us to have multiple `HashStableContext` implementations
1961 /// that hash things in a different way, without the results of one polluting
1962 /// the cache of the other.
1963 fn expn_id_cache() -> &'static LocalKey<ExpnIdCache>;
1964 fn hash_spans(&self) -> bool;
1965 fn span_data_to_lines_and_cols(
1968 ) -> Option<(Lrc<SourceFile>, usize, BytePos, usize, BytePos)>;
1971 impl<CTX> HashStable<CTX> for Span
1973 CTX: HashStableContext,
1975 /// Hashes a span in a stable way. We can't directly hash the span's `BytePos`
1976 /// fields (that would be similar to hashing pointers, since those are just
1977 /// offsets into the `SourceMap`). Instead, we hash the (file name, line, column)
1978 /// triple, which stays the same even if the containing `SourceFile` has moved
1979 /// within the `SourceMap`.
1981 /// Also note that we are hashing byte offsets for the column, not unicode
1982 /// codepoint offsets. For the purpose of the hash that's sufficient.
1983 /// Also, hashing filenames is expensive so we avoid doing it twice when the
1984 /// span starts and ends in the same file, which is almost always the case.
1985 fn hash_stable(&self, ctx: &mut CTX, hasher: &mut StableHasher) {
1986 const TAG_VALID_SPAN: u8 = 0;
1987 const TAG_INVALID_SPAN: u8 = 1;
1989 if !ctx.hash_spans() {
1993 self.ctxt().hash_stable(ctx, hasher);
1995 if self.is_dummy() {
1996 Hash::hash(&TAG_INVALID_SPAN, hasher);
2000 // If this is not an empty or invalid span, we want to hash the last
2001 // position that belongs to it, as opposed to hashing the first
2002 // position past it.
2003 let span = self.data();
2004 let (file, line_lo, col_lo, line_hi, col_hi) = match ctx.span_data_to_lines_and_cols(&span)
2008 Hash::hash(&TAG_INVALID_SPAN, hasher);
2013 Hash::hash(&TAG_VALID_SPAN, hasher);
2014 // We truncate the stable ID hash and line and column numbers. The chances
2015 // of causing a collision this way should be minimal.
2016 Hash::hash(&(file.name_hash as u64), hasher);
2018 // Hash both the length and the end location (line/column) of a span. If we
2019 // hash only the length, for example, then two otherwise equal spans with
2020 // different end locations will have the same hash. This can cause a problem
2021 // during incremental compilation wherein a previous result for a query that
2022 // depends on the end location of a span will be incorrectly reused when the
2023 // end location of the span it depends on has changed (see issue #74890). A
2024 // similar analysis applies if some query depends specifically on the length
2025 // of the span, but we only hash the end location. So hash both.
2027 let col_lo_trunc = (col_lo.0 as u64) & 0xFF;
2028 let line_lo_trunc = ((line_lo as u64) & 0xFF_FF_FF) << 8;
2029 let col_hi_trunc = (col_hi.0 as u64) & 0xFF << 32;
2030 let line_hi_trunc = ((line_hi as u64) & 0xFF_FF_FF) << 40;
2031 let col_line = col_lo_trunc | line_lo_trunc | col_hi_trunc | line_hi_trunc;
2032 let len = (span.hi - span.lo).0;
2033 Hash::hash(&col_line, hasher);
2034 Hash::hash(&len, hasher);