1 //! Source positions and related helper functions.
3 //! Important concepts in this module include:
5 //! - the *span*, represented by [`SpanData`] and related types;
6 //! - source code as represented by a [`SourceMap`]; and
7 //! - interned strings, represented by [`Symbol`]s, with some common symbols available statically in the [`sym`] module.
9 //! Unlike most compilers, the span contains not only the position in the source code, but also various other metadata,
10 //! such as the edition and macro hygiene. This metadata is stored in [`SyntaxContext`] and [`ExpnData`].
14 //! This API is completely unstable and subject to change.
16 #![doc(html_root_url = "https://doc.rust-lang.org/nightly/nightly-rustc/")]
17 #![feature(array_windows)]
18 #![feature(bool_to_option)]
19 #![feature(crate_visibility_modifier)]
21 #![feature(if_let_guard)]
22 #![feature(negative_impls)]
24 #![feature(min_specialization)]
25 #![feature(rustc_attrs)]
26 #![allow(rustc::potential_query_instability)]
29 extern crate rustc_macros;
34 use rustc_data_structures::AtomicRef;
35 use rustc_macros::HashStable_Generic;
36 use rustc_serialize::{Decodable, Decoder, Encodable, Encoder};
38 mod caching_source_map_view;
40 pub use self::caching_source_map_view::CachingSourceMapView;
41 use source_map::SourceMap;
46 use hygiene::Transparency;
47 pub use hygiene::{DesugaringKind, ExpnKind, MacroKind};
48 pub use hygiene::{ExpnData, ExpnHash, ExpnId, LocalExpnId, SyntaxContext};
49 use rustc_data_structures::stable_hasher::HashingControls;
51 use def_id::{CrateNum, DefId, DefPathHash, LocalDefId, LOCAL_CRATE};
54 pub use span_encoding::{Span, DUMMY_SP};
57 pub use symbol::{sym, Symbol};
59 mod analyze_source_file;
64 use rustc_data_structures::stable_hasher::{HashStable, StableHasher};
65 use rustc_data_structures::sync::{Lock, Lrc};
68 use std::cmp::{self, Ordering};
71 use std::ops::{Add, Range, Sub};
72 use std::path::{Path, PathBuf};
73 use std::str::FromStr;
85 // Per-session global variables: this struct is stored in thread-local storage
86 // in such a way that it is accessible without any kind of handle to all
87 // threads within the compilation session, but is not accessible outside the
89 pub struct SessionGlobals {
90 symbol_interner: symbol::Interner,
91 span_interner: Lock<span_encoding::SpanInterner>,
92 hygiene_data: Lock<hygiene::HygieneData>,
93 source_map: Lock<Option<Lrc<SourceMap>>>,
97 pub fn new(edition: Edition) -> SessionGlobals {
99 symbol_interner: symbol::Interner::fresh(),
100 span_interner: Lock::new(span_encoding::SpanInterner::default()),
101 hygiene_data: Lock::new(hygiene::HygieneData::new(edition)),
102 source_map: Lock::new(None),
108 pub fn create_session_globals_then<R>(edition: Edition, f: impl FnOnce() -> R) -> R {
110 !SESSION_GLOBALS.is_set(),
111 "SESSION_GLOBALS should never be overwritten! \
112 Use another thread if you need another SessionGlobals"
114 let session_globals = SessionGlobals::new(edition);
115 SESSION_GLOBALS.set(&session_globals, f)
119 pub fn set_session_globals_then<R>(session_globals: &SessionGlobals, f: impl FnOnce() -> R) -> R {
121 !SESSION_GLOBALS.is_set(),
122 "SESSION_GLOBALS should never be overwritten! \
123 Use another thread if you need another SessionGlobals"
125 SESSION_GLOBALS.set(session_globals, f)
129 pub fn create_default_session_if_not_set_then<R, F>(f: F) -> R
131 F: FnOnce(&SessionGlobals) -> R,
133 create_session_if_not_set_then(edition::DEFAULT_EDITION, f)
137 pub fn create_session_if_not_set_then<R, F>(edition: Edition, f: F) -> R
139 F: FnOnce(&SessionGlobals) -> R,
141 if !SESSION_GLOBALS.is_set() {
142 let session_globals = SessionGlobals::new(edition);
143 SESSION_GLOBALS.set(&session_globals, || SESSION_GLOBALS.with(f))
145 SESSION_GLOBALS.with(f)
150 pub fn with_session_globals<R, F>(f: F) -> R
152 F: FnOnce(&SessionGlobals) -> R,
154 SESSION_GLOBALS.with(f)
158 pub fn create_default_session_globals_then<R>(f: impl FnOnce() -> R) -> R {
159 create_session_globals_then(edition::DEFAULT_EDITION, f)
162 // If this ever becomes non thread-local, `decode_syntax_context`
163 // and `decode_expn_id` will need to be updated to handle concurrent
165 scoped_tls::scoped_thread_local!(static SESSION_GLOBALS: SessionGlobals);
167 // FIXME: We should use this enum or something like it to get rid of the
168 // use of magic `/rust/1.x/...` paths across the board.
169 #[derive(Debug, Eq, PartialEq, Clone, Ord, PartialOrd)]
171 pub enum RealFileName {
173 /// For remapped paths (namely paths into libstd that have been mapped
174 /// to the appropriate spot on the local host's file system, and local file
175 /// system paths that have been remapped with `FilePathMapping`),
177 /// `local_path` is the (host-dependent) local path to the file. This is
178 /// None if the file was imported from another crate
179 local_path: Option<PathBuf>,
180 /// `virtual_name` is the stable path rustc will store internally within
182 virtual_name: PathBuf,
186 impl Hash for RealFileName {
187 fn hash<H: std::hash::Hasher>(&self, state: &mut H) {
188 // To prevent #70924 from happening again we should only hash the
189 // remapped (virtualized) path if that exists. This is because
190 // virtualized paths to sysroot crates (/rust/$hash or /rust/$version)
191 // remain stable even if the corresponding local_path changes
192 self.remapped_path_if_available().hash(state)
196 // This is functionally identical to #[derive(Encodable)], with the exception of
197 // an added assert statement
198 impl<S: Encoder> Encodable<S> for RealFileName {
199 fn encode(&self, encoder: &mut S) -> Result<(), S::Error> {
200 encoder.emit_enum(|encoder| match *self {
201 RealFileName::LocalPath(ref local_path) => {
202 encoder.emit_enum_variant("LocalPath", 0, 1, |encoder| {
203 encoder.emit_enum_variant_arg(true, |encoder| local_path.encode(encoder))?;
208 RealFileName::Remapped { ref local_path, ref virtual_name } => encoder
209 .emit_enum_variant("Remapped", 1, 2, |encoder| {
210 // For privacy and build reproducibility, we must not embed host-dependant path in artifacts
211 // if they have been remapped by --remap-path-prefix
212 assert!(local_path.is_none());
213 encoder.emit_enum_variant_arg(true, |encoder| local_path.encode(encoder))?;
214 encoder.emit_enum_variant_arg(false, |encoder| virtual_name.encode(encoder))?;
222 /// Returns the path suitable for reading from the file system on the local host,
223 /// if this information exists.
224 /// Avoid embedding this in build artifacts; see `remapped_path_if_available()` for that.
225 pub fn local_path(&self) -> Option<&Path> {
227 RealFileName::LocalPath(p) => Some(p),
228 RealFileName::Remapped { local_path: p, virtual_name: _ } => {
229 p.as_ref().map(PathBuf::as_path)
234 /// Returns the path suitable for reading from the file system on the local host,
235 /// if this information exists.
236 /// Avoid embedding this in build artifacts; see `remapped_path_if_available()` for that.
237 pub fn into_local_path(self) -> Option<PathBuf> {
239 RealFileName::LocalPath(p) => Some(p),
240 RealFileName::Remapped { local_path: p, virtual_name: _ } => p,
244 /// Returns the path suitable for embedding into build artifacts. This would still
245 /// be a local path if it has not been remapped. A remapped path will not correspond
246 /// to a valid file system path: see `local_path_if_available()` for something that
247 /// is more likely to return paths into the local host file system.
248 pub fn remapped_path_if_available(&self) -> &Path {
250 RealFileName::LocalPath(p)
251 | RealFileName::Remapped { local_path: _, virtual_name: p } => &p,
255 /// Returns the path suitable for reading from the file system on the local host,
256 /// if this information exists. Otherwise returns the remapped name.
257 /// Avoid embedding this in build artifacts; see `remapped_path_if_available()` for that.
258 pub fn local_path_if_available(&self) -> &Path {
260 RealFileName::LocalPath(path)
261 | RealFileName::Remapped { local_path: None, virtual_name: path }
262 | RealFileName::Remapped { local_path: Some(path), virtual_name: _ } => path,
266 pub fn to_string_lossy(&self, display_pref: FileNameDisplayPreference) -> Cow<'_, str> {
268 FileNameDisplayPreference::Local => self.local_path_if_available().to_string_lossy(),
269 FileNameDisplayPreference::Remapped => {
270 self.remapped_path_if_available().to_string_lossy()
276 /// Differentiates between real files and common virtual files.
277 #[derive(Debug, Eq, PartialEq, Clone, Ord, PartialOrd, Hash)]
278 #[derive(Decodable, Encodable)]
281 /// Call to `quote!`.
285 /// Hack in `src/librustc_ast/parse.rs`.
288 ProcMacroSourceCode(u64),
289 /// Strings provided as `--cfg [cfgspec]` stored in a `crate_cfg`.
291 /// Strings provided as crate attributes in the CLI.
293 /// Custom sources for explicit parser calls from plugins and drivers.
295 DocTest(PathBuf, isize),
296 /// Post-substitution inline assembly from LLVM.
300 impl From<PathBuf> for FileName {
301 fn from(p: PathBuf) -> Self {
302 assert!(!p.to_string_lossy().ends_with('>'));
303 FileName::Real(RealFileName::LocalPath(p))
307 #[derive(Clone, Copy, Eq, PartialEq, Hash, Debug)]
308 pub enum FileNameDisplayPreference {
313 pub struct FileNameDisplay<'a> {
315 display_pref: FileNameDisplayPreference,
318 impl fmt::Display for FileNameDisplay<'_> {
319 fn fmt(&self, fmt: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
323 write!(fmt, "{}", name.to_string_lossy(self.display_pref))
325 QuoteExpansion(_) => write!(fmt, "<quote expansion>"),
326 MacroExpansion(_) => write!(fmt, "<macro expansion>"),
327 Anon(_) => write!(fmt, "<anon>"),
328 ProcMacroSourceCode(_) => write!(fmt, "<proc-macro source code>"),
329 CfgSpec(_) => write!(fmt, "<cfgspec>"),
330 CliCrateAttr(_) => write!(fmt, "<crate attribute>"),
331 Custom(ref s) => write!(fmt, "<{}>", s),
332 DocTest(ref path, _) => write!(fmt, "{}", path.display()),
333 InlineAsm(_) => write!(fmt, "<inline asm>"),
338 impl FileNameDisplay<'_> {
339 pub fn to_string_lossy(&self) -> Cow<'_, str> {
341 FileName::Real(ref inner) => inner.to_string_lossy(self.display_pref),
342 _ => Cow::from(format!("{}", self)),
348 pub fn is_real(&self) -> bool {
354 | ProcMacroSourceCode(_)
360 | InlineAsm(_) => false,
364 pub fn prefer_remapped(&self) -> FileNameDisplay<'_> {
365 FileNameDisplay { inner: self, display_pref: FileNameDisplayPreference::Remapped }
368 // This may include transient local filesystem information.
369 // Must not be embedded in build outputs.
370 pub fn prefer_local(&self) -> FileNameDisplay<'_> {
371 FileNameDisplay { inner: self, display_pref: FileNameDisplayPreference::Local }
374 pub fn display(&self, display_pref: FileNameDisplayPreference) -> FileNameDisplay<'_> {
375 FileNameDisplay { inner: self, display_pref }
378 pub fn macro_expansion_source_code(src: &str) -> FileName {
379 let mut hasher = StableHasher::new();
380 src.hash(&mut hasher);
381 FileName::MacroExpansion(hasher.finish())
384 pub fn anon_source_code(src: &str) -> FileName {
385 let mut hasher = StableHasher::new();
386 src.hash(&mut hasher);
387 FileName::Anon(hasher.finish())
390 pub fn proc_macro_source_code(src: &str) -> FileName {
391 let mut hasher = StableHasher::new();
392 src.hash(&mut hasher);
393 FileName::ProcMacroSourceCode(hasher.finish())
396 pub fn cfg_spec_source_code(src: &str) -> FileName {
397 let mut hasher = StableHasher::new();
398 src.hash(&mut hasher);
399 FileName::QuoteExpansion(hasher.finish())
402 pub fn cli_crate_attr_source_code(src: &str) -> FileName {
403 let mut hasher = StableHasher::new();
404 src.hash(&mut hasher);
405 FileName::CliCrateAttr(hasher.finish())
408 pub fn doc_test_source_code(path: PathBuf, line: isize) -> FileName {
409 FileName::DocTest(path, line)
412 pub fn inline_asm_source_code(src: &str) -> FileName {
413 let mut hasher = StableHasher::new();
414 src.hash(&mut hasher);
415 FileName::InlineAsm(hasher.finish())
419 /// Represents a span.
421 /// Spans represent a region of code, used for error reporting. Positions in spans
422 /// are *absolute* positions from the beginning of the [`SourceMap`], not positions
423 /// relative to [`SourceFile`]s. Methods on the `SourceMap` can be used to relate spans back
424 /// to the original source.
426 /// You must be careful if the span crosses more than one file, since you will not be
427 /// able to use many of the functions on spans in source_map and you cannot assume
428 /// that the length of the span is equal to `span.hi - span.lo`; there may be space in the
429 /// [`BytePos`] range between files.
431 /// `SpanData` is public because `Span` uses a thread-local interner and can't be
432 /// sent to other threads, but some pieces of performance infra run in a separate thread.
433 /// Using `Span` is generally preferred.
434 #[derive(Clone, Copy, Hash, PartialEq, Eq)]
435 pub struct SpanData {
438 /// Information about where the macro came from, if this piece of
439 /// code was created by a macro expansion.
440 pub ctxt: SyntaxContext,
441 pub parent: Option<LocalDefId>,
444 // Order spans by position in the file.
445 impl Ord for SpanData {
446 fn cmp(&self, other: &Self) -> Ordering {
451 // `LocalDefId` does not implement `Ord`.
452 // The other fields are enough to determine in-file order.
459 // `LocalDefId` does not implement `Ord`.
460 // The other fields are enough to determine in-file order.
464 (s_lo, s_hi, s_ctxt).cmp(&(o_lo, o_hi, o_ctxt))
468 impl PartialOrd for SpanData {
469 fn partial_cmp(&self, other: &Self) -> Option<Ordering> {
470 Some(self.cmp(other))
476 pub fn span(&self) -> Span {
477 Span::new(self.lo, self.hi, self.ctxt, self.parent)
480 pub fn with_lo(&self, lo: BytePos) -> Span {
481 Span::new(lo, self.hi, self.ctxt, self.parent)
484 pub fn with_hi(&self, hi: BytePos) -> Span {
485 Span::new(self.lo, hi, self.ctxt, self.parent)
488 pub fn with_ctxt(&self, ctxt: SyntaxContext) -> Span {
489 Span::new(self.lo, self.hi, ctxt, self.parent)
492 pub fn with_parent(&self, parent: Option<LocalDefId>) -> Span {
493 Span::new(self.lo, self.hi, self.ctxt, parent)
495 /// Returns `true` if this is a dummy span with any hygienic context.
497 pub fn is_dummy(self) -> bool {
498 self.lo.0 == 0 && self.hi.0 == 0
500 /// Returns `true` if `self` fully encloses `other`.
501 pub fn contains(self, other: Self) -> bool {
502 self.lo <= other.lo && other.hi <= self.hi
506 // The interner is pointed to by a thread local value which is only set on the main thread
507 // with parallelization is disabled. So we don't allow `Span` to transfer between threads
508 // to avoid panics and other errors, even though it would be memory safe to do so.
509 #[cfg(not(parallel_compiler))]
510 impl !Send for Span {}
511 #[cfg(not(parallel_compiler))]
512 impl !Sync for Span {}
514 impl PartialOrd for Span {
515 fn partial_cmp(&self, rhs: &Self) -> Option<Ordering> {
516 PartialOrd::partial_cmp(&self.data(), &rhs.data())
520 fn cmp(&self, rhs: &Self) -> Ordering {
521 Ord::cmp(&self.data(), &rhs.data())
527 pub fn lo(self) -> BytePos {
531 pub fn with_lo(self, lo: BytePos) -> Span {
532 self.data().with_lo(lo)
535 pub fn hi(self) -> BytePos {
539 pub fn with_hi(self, hi: BytePos) -> Span {
540 self.data().with_hi(hi)
543 pub fn ctxt(self) -> SyntaxContext {
544 self.data_untracked().ctxt
547 pub fn with_ctxt(self, ctxt: SyntaxContext) -> Span {
548 self.data_untracked().with_ctxt(ctxt)
551 pub fn parent(self) -> Option<LocalDefId> {
555 pub fn with_parent(self, ctxt: Option<LocalDefId>) -> Span {
556 self.data().with_parent(ctxt)
559 /// Returns `true` if this is a dummy span with any hygienic context.
561 pub fn is_dummy(self) -> bool {
562 self.data_untracked().is_dummy()
565 /// Returns `true` if this span comes from a macro or desugaring.
567 pub fn from_expansion(self) -> bool {
568 self.ctxt() != SyntaxContext::root()
571 /// Returns `true` if `span` originates in a derive-macro's expansion.
572 pub fn in_derive_expansion(self) -> bool {
573 matches!(self.ctxt().outer_expn_data().kind, ExpnKind::Macro(MacroKind::Derive, _))
576 /// Gate suggestions that would not be appropriate in a context the user didn't write.
577 pub fn can_be_used_for_suggestions(self) -> bool {
578 !self.from_expansion()
579 // FIXME: If this span comes from a `derive` macro but it points at code the user wrote,
580 // the callsite span and the span will be pointing at different places. It also means that
581 // we can safely provide suggestions on this span.
582 || (matches!(self.ctxt().outer_expn_data().kind, ExpnKind::Macro(MacroKind::Derive, _))
583 && self.parent_callsite().map(|p| (p.lo(), p.hi())) != Some((self.lo(), self.hi())))
587 pub fn with_root_ctxt(lo: BytePos, hi: BytePos) -> Span {
588 Span::new(lo, hi, SyntaxContext::root(), None)
591 /// Returns a new span representing an empty span at the beginning of this span.
593 pub fn shrink_to_lo(self) -> Span {
594 let span = self.data_untracked();
595 span.with_hi(span.lo)
597 /// Returns a new span representing an empty span at the end of this span.
599 pub fn shrink_to_hi(self) -> Span {
600 let span = self.data_untracked();
601 span.with_lo(span.hi)
605 /// Returns `true` if `hi == lo`.
606 pub fn is_empty(self) -> bool {
607 let span = self.data_untracked();
611 /// Returns `self` if `self` is not the dummy span, and `other` otherwise.
612 pub fn substitute_dummy(self, other: Span) -> Span {
613 if self.is_dummy() { other } else { self }
616 /// Returns `true` if `self` fully encloses `other`.
617 pub fn contains(self, other: Span) -> bool {
618 let span = self.data();
619 let other = other.data();
623 /// Returns `true` if `self` touches `other`.
624 pub fn overlaps(self, other: Span) -> bool {
625 let span = self.data();
626 let other = other.data();
627 span.lo < other.hi && other.lo < span.hi
630 /// Returns `true` if the spans are equal with regards to the source text.
632 /// Use this instead of `==` when either span could be generated code,
633 /// and you only care that they point to the same bytes of source text.
634 pub fn source_equal(self, other: Span) -> bool {
635 let span = self.data();
636 let other = other.data();
637 span.lo == other.lo && span.hi == other.hi
640 /// Returns `Some(span)`, where the start is trimmed by the end of `other`.
641 pub fn trim_start(self, other: Span) -> Option<Span> {
642 let span = self.data();
643 let other = other.data();
644 if span.hi > other.hi { Some(span.with_lo(cmp::max(span.lo, other.hi))) } else { None }
647 /// Returns the source span -- this is either the supplied span, or the span for
648 /// the macro callsite that expanded to it.
649 pub fn source_callsite(self) -> Span {
650 let expn_data = self.ctxt().outer_expn_data();
651 if !expn_data.is_root() { expn_data.call_site.source_callsite() } else { self }
654 /// The `Span` for the tokens in the previous macro expansion from which `self` was generated,
656 pub fn parent_callsite(self) -> Option<Span> {
657 let expn_data = self.ctxt().outer_expn_data();
658 if !expn_data.is_root() { Some(expn_data.call_site) } else { None }
661 /// Walk down the expansion ancestors to find a span that's contained within `outer`.
662 pub fn find_ancestor_inside(mut self, outer: Span) -> Option<Span> {
663 while !outer.contains(self) {
664 self = self.parent_callsite()?;
669 /// Edition of the crate from which this span came.
670 pub fn edition(self) -> edition::Edition {
671 self.ctxt().edition()
675 pub fn rust_2015(self) -> bool {
676 self.edition() == edition::Edition::Edition2015
680 pub fn rust_2018(self) -> bool {
681 self.edition() >= edition::Edition::Edition2018
685 pub fn rust_2021(self) -> bool {
686 self.edition() >= edition::Edition::Edition2021
690 pub fn rust_2024(self) -> bool {
691 self.edition() >= edition::Edition::Edition2024
694 /// Returns the source callee.
696 /// Returns `None` if the supplied span has no expansion trace,
697 /// else returns the `ExpnData` for the macro definition
698 /// corresponding to the source callsite.
699 pub fn source_callee(self) -> Option<ExpnData> {
700 fn source_callee(expn_data: ExpnData) -> ExpnData {
701 let next_expn_data = expn_data.call_site.ctxt().outer_expn_data();
702 if !next_expn_data.is_root() { source_callee(next_expn_data) } else { expn_data }
704 let expn_data = self.ctxt().outer_expn_data();
705 if !expn_data.is_root() { Some(source_callee(expn_data)) } else { None }
708 /// Checks if a span is "internal" to a macro in which `#[unstable]`
709 /// items can be used (that is, a macro marked with
710 /// `#[allow_internal_unstable]`).
711 pub fn allows_unstable(self, feature: Symbol) -> bool {
714 .allow_internal_unstable
715 .map_or(false, |features| features.iter().any(|&f| f == feature))
718 /// Checks if this span arises from a compiler desugaring of kind `kind`.
719 pub fn is_desugaring(self, kind: DesugaringKind) -> bool {
720 match self.ctxt().outer_expn_data().kind {
721 ExpnKind::Desugaring(k) => k == kind,
726 /// Returns the compiler desugaring that created this span, or `None`
727 /// if this span is not from a desugaring.
728 pub fn desugaring_kind(self) -> Option<DesugaringKind> {
729 match self.ctxt().outer_expn_data().kind {
730 ExpnKind::Desugaring(k) => Some(k),
735 /// Checks if a span is "internal" to a macro in which `unsafe`
736 /// can be used without triggering the `unsafe_code` lint.
737 // (that is, a macro marked with `#[allow_internal_unsafe]`).
738 pub fn allows_unsafe(self) -> bool {
739 self.ctxt().outer_expn_data().allow_internal_unsafe
742 pub fn macro_backtrace(mut self) -> impl Iterator<Item = ExpnData> {
743 let mut prev_span = DUMMY_SP;
744 std::iter::from_fn(move || {
746 let expn_data = self.ctxt().outer_expn_data();
747 if expn_data.is_root() {
751 let is_recursive = expn_data.call_site.source_equal(prev_span);
754 self = expn_data.call_site;
756 // Don't print recursive invocations.
758 return Some(expn_data);
764 /// Returns a `Span` that would enclose both `self` and `end`.
768 /// self lorem ipsum end
769 /// ^^^^^^^^^^^^^^^^^^^^
771 pub fn to(self, end: Span) -> Span {
772 let span_data = self.data();
773 let end_data = end.data();
774 // FIXME(jseyfried): `self.ctxt` should always equal `end.ctxt` here (cf. issue #23480).
775 // Return the macro span on its own to avoid weird diagnostic output. It is preferable to
776 // have an incomplete span than a completely nonsensical one.
777 if span_data.ctxt != end_data.ctxt {
778 if span_data.ctxt == SyntaxContext::root() {
780 } else if end_data.ctxt == SyntaxContext::root() {
783 // Both spans fall within a macro.
784 // FIXME(estebank): check if it is the *same* macro.
787 cmp::min(span_data.lo, end_data.lo),
788 cmp::max(span_data.hi, end_data.hi),
789 if span_data.ctxt == SyntaxContext::root() { end_data.ctxt } else { span_data.ctxt },
790 if span_data.parent == end_data.parent { span_data.parent } else { None },
794 /// Returns a `Span` between the end of `self` to the beginning of `end`.
798 /// self lorem ipsum end
801 pub fn between(self, end: Span) -> Span {
802 let span = self.data();
803 let end = end.data();
807 if end.ctxt == SyntaxContext::root() { end.ctxt } else { span.ctxt },
808 if span.parent == end.parent { span.parent } else { None },
812 /// Returns a `Span` from the beginning of `self` until the beginning of `end`.
816 /// self lorem ipsum end
817 /// ^^^^^^^^^^^^^^^^^
819 pub fn until(self, end: Span) -> Span {
820 // Most of this function's body is copied from `to`.
821 // We can't just do `self.to(end.shrink_to_lo())`,
822 // because to also does some magic where it uses min/max so
823 // it can handle overlapping spans. Some advanced mis-use of
824 // `until` with different ctxts makes this visible.
825 let span_data = self.data();
826 let end_data = end.data();
827 // FIXME(jseyfried): `self.ctxt` should always equal `end.ctxt` here (cf. issue #23480).
828 // Return the macro span on its own to avoid weird diagnostic output. It is preferable to
829 // have an incomplete span than a completely nonsensical one.
830 if span_data.ctxt != end_data.ctxt {
831 if span_data.ctxt == SyntaxContext::root() {
833 } else if end_data.ctxt == SyntaxContext::root() {
836 // Both spans fall within a macro.
837 // FIXME(estebank): check if it is the *same* macro.
842 if end_data.ctxt == SyntaxContext::root() { end_data.ctxt } else { span_data.ctxt },
843 if span_data.parent == end_data.parent { span_data.parent } else { None },
847 pub fn from_inner(self, inner: InnerSpan) -> Span {
848 let span = self.data();
850 span.lo + BytePos::from_usize(inner.start),
851 span.lo + BytePos::from_usize(inner.end),
857 /// Equivalent of `Span::def_site` from the proc macro API,
858 /// except that the location is taken from the `self` span.
859 pub fn with_def_site_ctxt(self, expn_id: ExpnId) -> Span {
860 self.with_ctxt_from_mark(expn_id, Transparency::Opaque)
863 /// Equivalent of `Span::call_site` from the proc macro API,
864 /// except that the location is taken from the `self` span.
865 pub fn with_call_site_ctxt(self, expn_id: ExpnId) -> Span {
866 self.with_ctxt_from_mark(expn_id, Transparency::Transparent)
869 /// Equivalent of `Span::mixed_site` from the proc macro API,
870 /// except that the location is taken from the `self` span.
871 pub fn with_mixed_site_ctxt(self, expn_id: ExpnId) -> Span {
872 self.with_ctxt_from_mark(expn_id, Transparency::SemiTransparent)
875 /// Produces a span with the same location as `self` and context produced by a macro with the
876 /// given ID and transparency, assuming that macro was defined directly and not produced by
877 /// some other macro (which is the case for built-in and procedural macros).
878 pub fn with_ctxt_from_mark(self, expn_id: ExpnId, transparency: Transparency) -> Span {
879 self.with_ctxt(SyntaxContext::root().apply_mark(expn_id, transparency))
883 pub fn apply_mark(self, expn_id: ExpnId, transparency: Transparency) -> Span {
884 let span = self.data();
885 span.with_ctxt(span.ctxt.apply_mark(expn_id, transparency))
889 pub fn remove_mark(&mut self) -> ExpnId {
890 let mut span = self.data();
891 let mark = span.ctxt.remove_mark();
892 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
897 pub fn adjust(&mut self, expn_id: ExpnId) -> Option<ExpnId> {
898 let mut span = self.data();
899 let mark = span.ctxt.adjust(expn_id);
900 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
905 pub fn normalize_to_macros_2_0_and_adjust(&mut self, expn_id: ExpnId) -> Option<ExpnId> {
906 let mut span = self.data();
907 let mark = span.ctxt.normalize_to_macros_2_0_and_adjust(expn_id);
908 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
913 pub fn glob_adjust(&mut self, expn_id: ExpnId, glob_span: Span) -> Option<Option<ExpnId>> {
914 let mut span = self.data();
915 let mark = span.ctxt.glob_adjust(expn_id, glob_span);
916 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
921 pub fn reverse_glob_adjust(
925 ) -> Option<Option<ExpnId>> {
926 let mut span = self.data();
927 let mark = span.ctxt.reverse_glob_adjust(expn_id, glob_span);
928 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
933 pub fn normalize_to_macros_2_0(self) -> Span {
934 let span = self.data();
935 span.with_ctxt(span.ctxt.normalize_to_macros_2_0())
939 pub fn normalize_to_macro_rules(self) -> Span {
940 let span = self.data();
941 span.with_ctxt(span.ctxt.normalize_to_macro_rules())
945 impl Default for Span {
946 fn default() -> Self {
951 impl<E: Encoder> Encodable<E> for Span {
952 default fn encode(&self, s: &mut E) -> Result<(), E::Error> {
953 let span = self.data();
954 s.emit_struct(false, |s| {
955 s.emit_struct_field("lo", true, |s| span.lo.encode(s))?;
956 s.emit_struct_field("hi", false, |s| span.hi.encode(s))
960 impl<D: Decoder> Decodable<D> for Span {
961 default fn decode(s: &mut D) -> Span {
962 let lo = Decodable::decode(s);
963 let hi = Decodable::decode(s);
965 Span::new(lo, hi, SyntaxContext::root(), None)
969 /// Calls the provided closure, using the provided `SourceMap` to format
970 /// any spans that are debug-printed during the closure's execution.
972 /// Normally, the global `TyCtxt` is used to retrieve the `SourceMap`
973 /// (see `rustc_interface::callbacks::span_debug1`). However, some parts
974 /// of the compiler (e.g. `rustc_parse`) may debug-print `Span`s before
975 /// a `TyCtxt` is available. In this case, we fall back to
976 /// the `SourceMap` provided to this function. If that is not available,
977 /// we fall back to printing the raw `Span` field values.
978 pub fn with_source_map<T, F: FnOnce() -> T>(source_map: Lrc<SourceMap>, f: F) -> T {
979 with_session_globals(|session_globals| {
980 *session_globals.source_map.borrow_mut() = Some(source_map);
982 struct ClearSourceMap;
983 impl Drop for ClearSourceMap {
985 with_session_globals(|session_globals| {
986 session_globals.source_map.borrow_mut().take();
991 let _guard = ClearSourceMap;
995 impl fmt::Debug for Span {
996 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
997 with_session_globals(|session_globals| {
998 if let Some(source_map) = &*session_globals.source_map.borrow() {
999 write!(f, "{} ({:?})", source_map.span_to_diagnostic_string(*self), self.ctxt())
1001 f.debug_struct("Span")
1002 .field("lo", &self.lo())
1003 .field("hi", &self.hi())
1004 .field("ctxt", &self.ctxt())
1011 impl fmt::Debug for SpanData {
1012 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
1013 fmt::Debug::fmt(&Span::new(self.lo, self.hi, self.ctxt, self.parent), f)
1017 /// Identifies an offset of a multi-byte character in a `SourceFile`.
1018 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1019 pub struct MultiByteChar {
1020 /// The absolute offset of the character in the `SourceMap`.
1022 /// The number of bytes, `>= 2`.
1026 /// Identifies an offset of a non-narrow character in a `SourceFile`.
1027 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1028 pub enum NonNarrowChar {
1029 /// Represents a zero-width character.
1031 /// Represents a wide (full-width) character.
1033 /// Represents a tab character, represented visually with a width of 4 characters.
1037 impl NonNarrowChar {
1038 fn new(pos: BytePos, width: usize) -> Self {
1040 0 => NonNarrowChar::ZeroWidth(pos),
1041 2 => NonNarrowChar::Wide(pos),
1042 4 => NonNarrowChar::Tab(pos),
1043 _ => panic!("width {} given for non-narrow character", width),
1047 /// Returns the absolute offset of the character in the `SourceMap`.
1048 pub fn pos(&self) -> BytePos {
1050 NonNarrowChar::ZeroWidth(p) | NonNarrowChar::Wide(p) | NonNarrowChar::Tab(p) => p,
1054 /// Returns the width of the character, 0 (zero-width) or 2 (wide).
1055 pub fn width(&self) -> usize {
1057 NonNarrowChar::ZeroWidth(_) => 0,
1058 NonNarrowChar::Wide(_) => 2,
1059 NonNarrowChar::Tab(_) => 4,
1064 impl Add<BytePos> for NonNarrowChar {
1067 fn add(self, rhs: BytePos) -> Self {
1069 NonNarrowChar::ZeroWidth(pos) => NonNarrowChar::ZeroWidth(pos + rhs),
1070 NonNarrowChar::Wide(pos) => NonNarrowChar::Wide(pos + rhs),
1071 NonNarrowChar::Tab(pos) => NonNarrowChar::Tab(pos + rhs),
1076 impl Sub<BytePos> for NonNarrowChar {
1079 fn sub(self, rhs: BytePos) -> Self {
1081 NonNarrowChar::ZeroWidth(pos) => NonNarrowChar::ZeroWidth(pos - rhs),
1082 NonNarrowChar::Wide(pos) => NonNarrowChar::Wide(pos - rhs),
1083 NonNarrowChar::Tab(pos) => NonNarrowChar::Tab(pos - rhs),
1088 /// Identifies an offset of a character that was normalized away from `SourceFile`.
1089 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1090 pub struct NormalizedPos {
1091 /// The absolute offset of the character in the `SourceMap`.
1093 /// The difference between original and normalized string at position.
1097 #[derive(PartialEq, Eq, Clone, Debug)]
1098 pub enum ExternalSource {
1099 /// No external source has to be loaded, since the `SourceFile` represents a local crate.
1102 kind: ExternalSourceKind,
1103 /// This SourceFile's byte-offset within the source_map of its original crate.
1104 original_start_pos: BytePos,
1105 /// The end of this SourceFile within the source_map of its original crate.
1106 original_end_pos: BytePos,
1110 /// The state of the lazy external source loading mechanism of a `SourceFile`.
1111 #[derive(PartialEq, Eq, Clone, Debug)]
1112 pub enum ExternalSourceKind {
1113 /// The external source has been loaded already.
1114 Present(Lrc<String>),
1115 /// No attempt has been made to load the external source.
1117 /// A failed attempt has been made to load the external source.
1122 impl ExternalSource {
1123 pub fn get_source(&self) -> Option<&Lrc<String>> {
1125 ExternalSource::Foreign { kind: ExternalSourceKind::Present(ref src), .. } => Some(src),
1132 pub struct OffsetOverflowError;
1134 #[derive(Copy, Clone, Debug, PartialEq, Eq, PartialOrd, Ord, Hash, Encodable, Decodable)]
1135 #[derive(HashStable_Generic)]
1136 pub enum SourceFileHashAlgorithm {
1142 impl FromStr for SourceFileHashAlgorithm {
1145 fn from_str(s: &str) -> Result<SourceFileHashAlgorithm, ()> {
1147 "md5" => Ok(SourceFileHashAlgorithm::Md5),
1148 "sha1" => Ok(SourceFileHashAlgorithm::Sha1),
1149 "sha256" => Ok(SourceFileHashAlgorithm::Sha256),
1155 /// The hash of the on-disk source file used for debug info.
1156 #[derive(Copy, Clone, PartialEq, Eq, Debug)]
1157 #[derive(HashStable_Generic, Encodable, Decodable)]
1158 pub struct SourceFileHash {
1159 pub kind: SourceFileHashAlgorithm,
1163 impl SourceFileHash {
1164 pub fn new(kind: SourceFileHashAlgorithm, src: &str) -> SourceFileHash {
1165 let mut hash = SourceFileHash { kind, value: Default::default() };
1166 let len = hash.hash_len();
1167 let value = &mut hash.value[..len];
1168 let data = src.as_bytes();
1170 SourceFileHashAlgorithm::Md5 => {
1171 value.copy_from_slice(&Md5::digest(data));
1173 SourceFileHashAlgorithm::Sha1 => {
1174 value.copy_from_slice(&Sha1::digest(data));
1176 SourceFileHashAlgorithm::Sha256 => {
1177 value.copy_from_slice(&Sha256::digest(data));
1183 /// Check if the stored hash matches the hash of the string.
1184 pub fn matches(&self, src: &str) -> bool {
1185 Self::new(self.kind, src) == *self
1188 /// The bytes of the hash.
1189 pub fn hash_bytes(&self) -> &[u8] {
1190 let len = self.hash_len();
1194 fn hash_len(&self) -> usize {
1196 SourceFileHashAlgorithm::Md5 => 16,
1197 SourceFileHashAlgorithm::Sha1 => 20,
1198 SourceFileHashAlgorithm::Sha256 => 32,
1203 /// A single source in the [`SourceMap`].
1205 pub struct SourceFile {
1206 /// The name of the file that the source came from. Source that doesn't
1207 /// originate from files has names between angle brackets by convention
1208 /// (e.g., `<anon>`).
1210 /// The complete source code.
1211 pub src: Option<Lrc<String>>,
1212 /// The source code's hash.
1213 pub src_hash: SourceFileHash,
1214 /// The external source code (used for external crates, which will have a `None`
1215 /// value as `self.src`.
1216 pub external_src: Lock<ExternalSource>,
1217 /// The start position of this source in the `SourceMap`.
1218 pub start_pos: BytePos,
1219 /// The end position of this source in the `SourceMap`.
1220 pub end_pos: BytePos,
1221 /// Locations of lines beginnings in the source code.
1222 pub lines: Vec<BytePos>,
1223 /// Locations of multi-byte characters in the source code.
1224 pub multibyte_chars: Vec<MultiByteChar>,
1225 /// Width of characters that are not narrow in the source code.
1226 pub non_narrow_chars: Vec<NonNarrowChar>,
1227 /// Locations of characters removed during normalization.
1228 pub normalized_pos: Vec<NormalizedPos>,
1229 /// A hash of the filename, used for speeding up hashing in incremental compilation.
1230 pub name_hash: u128,
1231 /// Indicates which crate this `SourceFile` was imported from.
1235 impl<S: Encoder> Encodable<S> for SourceFile {
1236 fn encode(&self, s: &mut S) -> Result<(), S::Error> {
1237 s.emit_struct(false, |s| {
1238 s.emit_struct_field("name", true, |s| self.name.encode(s))?;
1239 s.emit_struct_field("src_hash", false, |s| self.src_hash.encode(s))?;
1240 s.emit_struct_field("start_pos", false, |s| self.start_pos.encode(s))?;
1241 s.emit_struct_field("end_pos", false, |s| self.end_pos.encode(s))?;
1242 s.emit_struct_field("lines", false, |s| {
1243 let lines = &self.lines[..];
1244 // Store the length.
1245 s.emit_u32(lines.len() as u32)?;
1247 if !lines.is_empty() {
1248 // In order to preserve some space, we exploit the fact that
1249 // the lines list is sorted and individual lines are
1250 // probably not that long. Because of that we can store lines
1251 // as a difference list, using as little space as possible
1252 // for the differences.
1253 let max_line_length = if lines.len() == 1 {
1258 .map(|&[fst, snd]| snd - fst)
1259 .map(|bp| bp.to_usize())
1264 let bytes_per_diff: u8 = match max_line_length {
1266 0x100..=0xFFFF => 2,
1270 // Encode the number of bytes used per diff.
1271 bytes_per_diff.encode(s)?;
1273 // Encode the first element.
1274 lines[0].encode(s)?;
1276 let diff_iter = lines.array_windows().map(|&[fst, snd]| snd - fst);
1278 match bytes_per_diff {
1280 for diff in diff_iter {
1281 (diff.0 as u8).encode(s)?
1285 for diff in diff_iter {
1286 (diff.0 as u16).encode(s)?
1290 for diff in diff_iter {
1294 _ => unreachable!(),
1300 s.emit_struct_field("multibyte_chars", false, |s| self.multibyte_chars.encode(s))?;
1301 s.emit_struct_field("non_narrow_chars", false, |s| self.non_narrow_chars.encode(s))?;
1302 s.emit_struct_field("name_hash", false, |s| self.name_hash.encode(s))?;
1303 s.emit_struct_field("normalized_pos", false, |s| self.normalized_pos.encode(s))?;
1304 s.emit_struct_field("cnum", false, |s| self.cnum.encode(s))
1309 impl<D: Decoder> Decodable<D> for SourceFile {
1310 fn decode(d: &mut D) -> SourceFile {
1311 let name: FileName = Decodable::decode(d);
1312 let src_hash: SourceFileHash = Decodable::decode(d);
1313 let start_pos: BytePos = Decodable::decode(d);
1314 let end_pos: BytePos = Decodable::decode(d);
1315 let lines: Vec<BytePos> = {
1316 let num_lines: u32 = Decodable::decode(d);
1317 let mut lines = Vec::with_capacity(num_lines as usize);
1320 // Read the number of bytes used per diff.
1321 let bytes_per_diff: u8 = Decodable::decode(d);
1323 // Read the first element.
1324 let mut line_start: BytePos = Decodable::decode(d);
1325 lines.push(line_start);
1327 match bytes_per_diff {
1328 1 => lines.extend((1..num_lines).map(|_| {
1329 line_start = line_start + BytePos(d.read_u8() as u32);
1332 2 => lines.extend((1..num_lines).map(|_| {
1333 line_start = line_start + BytePos(d.read_u16() as u32);
1336 4 => lines.extend((1..num_lines).map(|_| {
1337 line_start = line_start + BytePos(d.read_u32());
1340 _ => unreachable!(),
1346 let multibyte_chars: Vec<MultiByteChar> = Decodable::decode(d);
1347 let non_narrow_chars: Vec<NonNarrowChar> = Decodable::decode(d);
1348 let name_hash: u128 = Decodable::decode(d);
1349 let normalized_pos: Vec<NormalizedPos> = Decodable::decode(d);
1350 let cnum: CrateNum = Decodable::decode(d);
1357 // Unused - the metadata decoder will construct
1358 // a new SourceFile, filling in `external_src` properly
1359 external_src: Lock::new(ExternalSource::Unneeded),
1370 impl fmt::Debug for SourceFile {
1371 fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result {
1372 write!(fmt, "SourceFile({:?})", self.name)
1381 hash_kind: SourceFileHashAlgorithm,
1383 // Compute the file hash before any normalization.
1384 let src_hash = SourceFileHash::new(hash_kind, &src);
1385 let normalized_pos = normalize_src(&mut src, start_pos);
1388 let mut hasher: StableHasher = StableHasher::new();
1389 name.hash(&mut hasher);
1390 hasher.finish::<u128>()
1392 let end_pos = start_pos.to_usize() + src.len();
1393 assert!(end_pos <= u32::MAX as usize);
1395 let (lines, multibyte_chars, non_narrow_chars) =
1396 analyze_source_file::analyze_source_file(&src, start_pos);
1400 src: Some(Lrc::new(src)),
1402 external_src: Lock::new(ExternalSource::Unneeded),
1404 end_pos: Pos::from_usize(end_pos),
1414 /// Returns the `BytePos` of the beginning of the current line.
1415 pub fn line_begin_pos(&self, pos: BytePos) -> BytePos {
1416 let line_index = self.lookup_line(pos).unwrap();
1417 self.lines[line_index]
1420 /// Add externally loaded source.
1421 /// If the hash of the input doesn't match or no input is supplied via None,
1422 /// it is interpreted as an error and the corresponding enum variant is set.
1423 /// The return value signifies whether some kind of source is present.
1424 pub fn add_external_src<F>(&self, get_src: F) -> bool
1426 F: FnOnce() -> Option<String>,
1429 *self.external_src.borrow(),
1430 ExternalSource::Foreign { kind: ExternalSourceKind::AbsentOk, .. }
1432 let src = get_src();
1433 let mut external_src = self.external_src.borrow_mut();
1434 // Check that no-one else have provided the source while we were getting it
1435 if let ExternalSource::Foreign {
1436 kind: src_kind @ ExternalSourceKind::AbsentOk, ..
1437 } = &mut *external_src
1439 if let Some(mut src) = src {
1440 // The src_hash needs to be computed on the pre-normalized src.
1441 if self.src_hash.matches(&src) {
1442 normalize_src(&mut src, BytePos::from_usize(0));
1443 *src_kind = ExternalSourceKind::Present(Lrc::new(src));
1447 *src_kind = ExternalSourceKind::AbsentErr;
1452 self.src.is_some() || external_src.get_source().is_some()
1455 self.src.is_some() || self.external_src.borrow().get_source().is_some()
1459 /// Gets a line from the list of pre-computed line-beginnings.
1460 /// The line number here is 0-based.
1461 pub fn get_line(&self, line_number: usize) -> Option<Cow<'_, str>> {
1462 fn get_until_newline(src: &str, begin: usize) -> &str {
1463 // We can't use `lines.get(line_number+1)` because we might
1464 // be parsing when we call this function and thus the current
1465 // line is the last one we have line info for.
1466 let slice = &src[begin..];
1467 match slice.find('\n') {
1468 Some(e) => &slice[..e],
1474 let line = self.lines.get(line_number)?;
1475 let begin: BytePos = *line - self.start_pos;
1479 if let Some(ref src) = self.src {
1480 Some(Cow::from(get_until_newline(src, begin)))
1481 } else if let Some(src) = self.external_src.borrow().get_source() {
1482 Some(Cow::Owned(String::from(get_until_newline(src, begin))))
1488 pub fn is_real_file(&self) -> bool {
1492 pub fn is_imported(&self) -> bool {
1496 pub fn count_lines(&self) -> usize {
1500 /// Finds the line containing the given position. The return value is the
1501 /// index into the `lines` array of this `SourceFile`, not the 1-based line
1502 /// number. If the source_file is empty or the position is located before the
1503 /// first line, `None` is returned.
1504 pub fn lookup_line(&self, pos: BytePos) -> Option<usize> {
1505 match self.lines.binary_search(&pos) {
1506 Ok(idx) => Some(idx),
1508 Err(idx) => Some(idx - 1),
1512 pub fn line_bounds(&self, line_index: usize) -> Range<BytePos> {
1513 if self.is_empty() {
1514 return self.start_pos..self.end_pos;
1517 assert!(line_index < self.lines.len());
1518 if line_index == (self.lines.len() - 1) {
1519 self.lines[line_index]..self.end_pos
1521 self.lines[line_index]..self.lines[line_index + 1]
1525 /// Returns whether or not the file contains the given `SourceMap` byte
1526 /// position. The position one past the end of the file is considered to be
1527 /// contained by the file. This implies that files for which `is_empty`
1528 /// returns true still contain one byte position according to this function.
1530 pub fn contains(&self, byte_pos: BytePos) -> bool {
1531 byte_pos >= self.start_pos && byte_pos <= self.end_pos
1535 pub fn is_empty(&self) -> bool {
1536 self.start_pos == self.end_pos
1539 /// Calculates the original byte position relative to the start of the file
1540 /// based on the given byte position.
1541 pub fn original_relative_byte_pos(&self, pos: BytePos) -> BytePos {
1542 // Diff before any records is 0. Otherwise use the previously recorded
1543 // diff as that applies to the following characters until a new diff
1545 let diff = match self.normalized_pos.binary_search_by(|np| np.pos.cmp(&pos)) {
1546 Ok(i) => self.normalized_pos[i].diff,
1547 Err(i) if i == 0 => 0,
1548 Err(i) => self.normalized_pos[i - 1].diff,
1551 BytePos::from_u32(pos.0 - self.start_pos.0 + diff)
1554 /// Converts an absolute `BytePos` to a `CharPos` relative to the `SourceFile`.
1555 pub fn bytepos_to_file_charpos(&self, bpos: BytePos) -> CharPos {
1556 // The number of extra bytes due to multibyte chars in the `SourceFile`.
1557 let mut total_extra_bytes = 0;
1559 for mbc in self.multibyte_chars.iter() {
1560 debug!("{}-byte char at {:?}", mbc.bytes, mbc.pos);
1562 // Every character is at least one byte, so we only
1563 // count the actual extra bytes.
1564 total_extra_bytes += mbc.bytes as u32 - 1;
1565 // We should never see a byte position in the middle of a
1567 assert!(bpos.to_u32() >= mbc.pos.to_u32() + mbc.bytes as u32);
1573 assert!(self.start_pos.to_u32() + total_extra_bytes <= bpos.to_u32());
1574 CharPos(bpos.to_usize() - self.start_pos.to_usize() - total_extra_bytes as usize)
1577 /// Looks up the file's (1-based) line number and (0-based `CharPos`) column offset, for a
1578 /// given `BytePos`.
1579 pub fn lookup_file_pos(&self, pos: BytePos) -> (usize, CharPos) {
1580 let chpos = self.bytepos_to_file_charpos(pos);
1581 match self.lookup_line(pos) {
1583 let line = a + 1; // Line numbers start at 1
1584 let linebpos = self.lines[a];
1585 let linechpos = self.bytepos_to_file_charpos(linebpos);
1586 let col = chpos - linechpos;
1587 debug!("byte pos {:?} is on the line at byte pos {:?}", pos, linebpos);
1588 debug!("char pos {:?} is on the line at char pos {:?}", chpos, linechpos);
1589 debug!("byte is on line: {}", line);
1590 assert!(chpos >= linechpos);
1597 /// Looks up the file's (1-based) line number, (0-based `CharPos`) column offset, and (0-based)
1598 /// column offset when displayed, for a given `BytePos`.
1599 pub fn lookup_file_pos_with_col_display(&self, pos: BytePos) -> (usize, CharPos, usize) {
1600 let (line, col_or_chpos) = self.lookup_file_pos(pos);
1602 let col = col_or_chpos;
1603 let linebpos = self.lines[line - 1];
1605 let start_width_idx = self
1607 .binary_search_by_key(&linebpos, |x| x.pos())
1608 .unwrap_or_else(|x| x);
1609 let end_width_idx = self
1611 .binary_search_by_key(&pos, |x| x.pos())
1612 .unwrap_or_else(|x| x);
1613 let special_chars = end_width_idx - start_width_idx;
1614 let non_narrow: usize = self.non_narrow_chars[start_width_idx..end_width_idx]
1618 col.0 - special_chars + non_narrow
1620 (line, col, col_display)
1622 let chpos = col_or_chpos;
1624 let end_width_idx = self
1626 .binary_search_by_key(&pos, |x| x.pos())
1627 .unwrap_or_else(|x| x);
1628 let non_narrow: usize =
1629 self.non_narrow_chars[0..end_width_idx].iter().map(|x| x.width()).sum();
1630 chpos.0 - end_width_idx + non_narrow
1632 (0, chpos, col_display)
1637 /// Normalizes the source code and records the normalizations.
1638 fn normalize_src(src: &mut String, start_pos: BytePos) -> Vec<NormalizedPos> {
1639 let mut normalized_pos = vec![];
1640 remove_bom(src, &mut normalized_pos);
1641 normalize_newlines(src, &mut normalized_pos);
1643 // Offset all the positions by start_pos to match the final file positions.
1644 for np in &mut normalized_pos {
1645 np.pos.0 += start_pos.0;
1651 /// Removes UTF-8 BOM, if any.
1652 fn remove_bom(src: &mut String, normalized_pos: &mut Vec<NormalizedPos>) {
1653 if src.starts_with('\u{feff}') {
1655 normalized_pos.push(NormalizedPos { pos: BytePos(0), diff: 3 });
1659 /// Replaces `\r\n` with `\n` in-place in `src`.
1661 /// Returns error if there's a lone `\r` in the string.
1662 fn normalize_newlines(src: &mut String, normalized_pos: &mut Vec<NormalizedPos>) {
1663 if !src.as_bytes().contains(&b'\r') {
1667 // We replace `\r\n` with `\n` in-place, which doesn't break utf-8 encoding.
1668 // While we *can* call `as_mut_vec` and do surgery on the live string
1669 // directly, let's rather steal the contents of `src`. This makes the code
1670 // safe even if a panic occurs.
1672 let mut buf = std::mem::replace(src, String::new()).into_bytes();
1673 let mut gap_len = 0;
1674 let mut tail = buf.as_mut_slice();
1676 let original_gap = normalized_pos.last().map_or(0, |l| l.diff);
1678 let idx = match find_crlf(&tail[gap_len..]) {
1680 Some(idx) => idx + gap_len,
1682 tail.copy_within(gap_len..idx, 0);
1683 tail = &mut tail[idx - gap_len..];
1684 if tail.len() == gap_len {
1687 cursor += idx - gap_len;
1689 normalized_pos.push(NormalizedPos {
1690 pos: BytePos::from_usize(cursor + 1),
1691 diff: original_gap + gap_len as u32,
1695 // Account for removed `\r`.
1696 // After `set_len`, `buf` is guaranteed to contain utf-8 again.
1697 let new_len = buf.len() - gap_len;
1699 buf.set_len(new_len);
1700 *src = String::from_utf8_unchecked(buf);
1703 fn find_crlf(src: &[u8]) -> Option<usize> {
1704 let mut search_idx = 0;
1705 while let Some(idx) = find_cr(&src[search_idx..]) {
1706 if src[search_idx..].get(idx + 1) != Some(&b'\n') {
1707 search_idx += idx + 1;
1710 return Some(search_idx + idx);
1715 fn find_cr(src: &[u8]) -> Option<usize> {
1716 src.iter().position(|&b| b == b'\r')
1720 // _____________________________________________________________________________
1721 // Pos, BytePos, CharPos
1725 fn from_usize(n: usize) -> Self;
1726 fn to_usize(&self) -> usize;
1727 fn from_u32(n: u32) -> Self;
1728 fn to_u32(&self) -> u32;
1731 macro_rules! impl_pos {
1735 $vis:vis struct $ident:ident($inner_vis:vis $inner_ty:ty);
1740 $vis struct $ident($inner_vis $inner_ty);
1742 impl Pos for $ident {
1744 fn from_usize(n: usize) -> $ident {
1745 $ident(n as $inner_ty)
1749 fn to_usize(&self) -> usize {
1754 fn from_u32(n: u32) -> $ident {
1755 $ident(n as $inner_ty)
1759 fn to_u32(&self) -> u32 {
1764 impl Add for $ident {
1765 type Output = $ident;
1768 fn add(self, rhs: $ident) -> $ident {
1769 $ident(self.0 + rhs.0)
1773 impl Sub for $ident {
1774 type Output = $ident;
1777 fn sub(self, rhs: $ident) -> $ident {
1778 $ident(self.0 - rhs.0)
1788 /// Keep this small (currently 32-bits), as AST contains a lot of them.
1789 #[derive(Clone, Copy, PartialEq, Eq, Hash, PartialOrd, Ord, Debug)]
1790 pub struct BytePos(pub u32);
1792 /// A character offset.
1794 /// Because of multibyte UTF-8 characters, a byte offset
1795 /// is not equivalent to a character offset. The [`SourceMap`] will convert [`BytePos`]
1796 /// values to `CharPos` values as necessary.
1797 #[derive(Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Debug)]
1798 pub struct CharPos(pub usize);
1801 impl<S: rustc_serialize::Encoder> Encodable<S> for BytePos {
1802 fn encode(&self, s: &mut S) -> Result<(), S::Error> {
1807 impl<D: rustc_serialize::Decoder> Decodable<D> for BytePos {
1808 fn decode(d: &mut D) -> BytePos {
1809 BytePos(d.read_u32())
1813 // _____________________________________________________________________________
1814 // Loc, SourceFileAndLine, SourceFileAndBytePos
1817 /// A source code location used for error reporting.
1818 #[derive(Debug, Clone)]
1820 /// Information about the original source.
1821 pub file: Lrc<SourceFile>,
1822 /// The (1-based) line number.
1824 /// The (0-based) column offset.
1826 /// The (0-based) column offset when displayed.
1827 pub col_display: usize,
1830 // Used to be structural records.
1832 pub struct SourceFileAndLine {
1833 pub sf: Lrc<SourceFile>,
1834 /// Index of line, starting from 0.
1838 pub struct SourceFileAndBytePos {
1839 pub sf: Lrc<SourceFile>,
1843 #[derive(Copy, Clone, Debug, PartialEq, Eq)]
1844 pub struct LineInfo {
1845 /// Index of line, starting from 0.
1846 pub line_index: usize,
1848 /// Column in line where span begins, starting from 0.
1849 pub start_col: CharPos,
1851 /// Column in line where span ends, starting from 0, exclusive.
1852 pub end_col: CharPos,
1855 pub struct FileLines {
1856 pub file: Lrc<SourceFile>,
1857 pub lines: Vec<LineInfo>,
1860 pub static SPAN_TRACK: AtomicRef<fn(LocalDefId)> = AtomicRef::new(&((|_| {}) as fn(_)));
1862 // _____________________________________________________________________________
1863 // SpanLinesError, SpanSnippetError, DistinctSources, MalformedSourceMapPositions
1866 pub type FileLinesResult = Result<FileLines, SpanLinesError>;
1868 #[derive(Clone, PartialEq, Eq, Debug)]
1869 pub enum SpanLinesError {
1870 DistinctSources(DistinctSources),
1873 #[derive(Clone, PartialEq, Eq, Debug)]
1874 pub enum SpanSnippetError {
1875 IllFormedSpan(Span),
1876 DistinctSources(DistinctSources),
1877 MalformedForSourcemap(MalformedSourceMapPositions),
1878 SourceNotAvailable { filename: FileName },
1881 #[derive(Clone, PartialEq, Eq, Debug)]
1882 pub struct DistinctSources {
1883 pub begin: (FileName, BytePos),
1884 pub end: (FileName, BytePos),
1887 #[derive(Clone, PartialEq, Eq, Debug)]
1888 pub struct MalformedSourceMapPositions {
1890 pub source_len: usize,
1891 pub begin_pos: BytePos,
1892 pub end_pos: BytePos,
1895 /// Range inside of a `Span` used for diagnostics when we only have access to relative positions.
1896 #[derive(Copy, Clone, PartialEq, Eq, Debug)]
1897 pub struct InnerSpan {
1903 pub fn new(start: usize, end: usize) -> InnerSpan {
1904 InnerSpan { start, end }
1908 /// Requirements for a `StableHashingContext` to be used in this crate.
1910 /// This is a hack to allow using the [`HashStable_Generic`] derive macro
1911 /// instead of implementing everything in rustc_middle.
1912 pub trait HashStableContext {
1913 fn def_path_hash(&self, def_id: DefId) -> DefPathHash;
1914 fn hash_spans(&self) -> bool;
1915 /// Accesses `sess.opts.debugging_opts.incremental_ignore_spans` since
1916 /// we don't have easy access to a `Session`
1917 fn debug_opts_incremental_ignore_spans(&self) -> bool;
1918 fn def_span(&self, def_id: LocalDefId) -> Span;
1919 fn span_data_to_lines_and_cols(
1922 ) -> Option<(Lrc<SourceFile>, usize, BytePos, usize, BytePos)>;
1923 fn hashing_controls(&self) -> HashingControls;
1926 impl<CTX> HashStable<CTX> for Span
1928 CTX: HashStableContext,
1930 /// Hashes a span in a stable way. We can't directly hash the span's `BytePos`
1931 /// fields (that would be similar to hashing pointers, since those are just
1932 /// offsets into the `SourceMap`). Instead, we hash the (file name, line, column)
1933 /// triple, which stays the same even if the containing `SourceFile` has moved
1934 /// within the `SourceMap`.
1936 /// Also note that we are hashing byte offsets for the column, not unicode
1937 /// codepoint offsets. For the purpose of the hash that's sufficient.
1938 /// Also, hashing filenames is expensive so we avoid doing it twice when the
1939 /// span starts and ends in the same file, which is almost always the case.
1940 fn hash_stable(&self, ctx: &mut CTX, hasher: &mut StableHasher) {
1941 const TAG_VALID_SPAN: u8 = 0;
1942 const TAG_INVALID_SPAN: u8 = 1;
1943 const TAG_RELATIVE_SPAN: u8 = 2;
1945 if !ctx.hash_spans() {
1949 let span = self.data_untracked();
1950 span.ctxt.hash_stable(ctx, hasher);
1951 span.parent.hash_stable(ctx, hasher);
1953 if span.is_dummy() {
1954 Hash::hash(&TAG_INVALID_SPAN, hasher);
1958 if let Some(parent) = span.parent {
1959 let def_span = ctx.def_span(parent).data_untracked();
1960 if def_span.contains(span) {
1961 // This span is enclosed in a definition: only hash the relative position.
1962 Hash::hash(&TAG_RELATIVE_SPAN, hasher);
1963 (span.lo - def_span.lo).to_u32().hash_stable(ctx, hasher);
1964 (span.hi - def_span.lo).to_u32().hash_stable(ctx, hasher);
1969 // If this is not an empty or invalid span, we want to hash the last
1970 // position that belongs to it, as opposed to hashing the first
1971 // position past it.
1972 let Some((file, line_lo, col_lo, line_hi, col_hi)) = ctx.span_data_to_lines_and_cols(&span) else {
1973 Hash::hash(&TAG_INVALID_SPAN, hasher);
1977 Hash::hash(&TAG_VALID_SPAN, hasher);
1978 // We truncate the stable ID hash and line and column numbers. The chances
1979 // of causing a collision this way should be minimal.
1980 Hash::hash(&(file.name_hash as u64), hasher);
1982 // Hash both the length and the end location (line/column) of a span. If we
1983 // hash only the length, for example, then two otherwise equal spans with
1984 // different end locations will have the same hash. This can cause a problem
1985 // during incremental compilation wherein a previous result for a query that
1986 // depends on the end location of a span will be incorrectly reused when the
1987 // end location of the span it depends on has changed (see issue #74890). A
1988 // similar analysis applies if some query depends specifically on the length
1989 // of the span, but we only hash the end location. So hash both.
1991 let col_lo_trunc = (col_lo.0 as u64) & 0xFF;
1992 let line_lo_trunc = ((line_lo as u64) & 0xFF_FF_FF) << 8;
1993 let col_hi_trunc = (col_hi.0 as u64) & 0xFF << 32;
1994 let line_hi_trunc = ((line_hi as u64) & 0xFF_FF_FF) << 40;
1995 let col_line = col_lo_trunc | line_lo_trunc | col_hi_trunc | line_hi_trunc;
1996 let len = (span.hi - span.lo).0;
1997 Hash::hash(&col_line, hasher);
1998 Hash::hash(&len, hasher);