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(if_let_guard)]
20 #![feature(negative_impls)]
22 #![feature(min_specialization)]
23 #![feature(thread_local_const_init)]
26 extern crate rustc_macros;
28 use rustc_data_structures::AtomicRef;
29 use rustc_macros::HashStable_Generic;
30 use rustc_serialize::{Decodable, Decoder, Encodable, Encoder};
32 mod caching_source_map_view;
34 pub use self::caching_source_map_view::CachingSourceMapView;
35 use source_map::SourceMap;
40 use hygiene::Transparency;
41 pub use hygiene::{DesugaringKind, ExpnKind, ForLoopLoc, MacroKind};
42 pub use hygiene::{ExpnData, ExpnHash, ExpnId, LocalExpnId, SyntaxContext};
44 use def_id::{CrateNum, DefId, DefPathHash, LocalDefId, LOCAL_CRATE};
47 pub use span_encoding::{Span, DUMMY_SP};
50 pub use symbol::{sym, Symbol};
52 mod analyze_source_file;
55 use rustc_data_structures::stable_hasher::{HashStable, StableHasher};
56 use rustc_data_structures::sync::{Lock, Lrc};
59 use std::cmp::{self, Ordering};
62 use std::ops::{Add, Range, Sub};
63 use std::path::{Path, PathBuf};
64 use std::str::FromStr;
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: 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: 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)]
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, display_pref: FileNameDisplayPreference) -> Cow<'_, str> {
264 FileNameDisplayPreference::Local => self.local_path_if_available().to_string_lossy(),
265 FileNameDisplayPreference::Remapped => {
266 self.remapped_path_if_available().to_string_lossy()
272 /// Differentiates between real files and common virtual files.
273 #[derive(Debug, Eq, PartialEq, Clone, Ord, PartialOrd, Hash)]
274 #[derive(Decodable, Encodable)]
277 /// Call to `quote!`.
281 /// Hack in `src/librustc_ast/parse.rs`.
284 ProcMacroSourceCode(u64),
285 /// Strings provided as `--cfg [cfgspec]` stored in a `crate_cfg`.
287 /// Strings provided as crate attributes in the CLI.
289 /// Custom sources for explicit parser calls from plugins and drivers.
291 DocTest(PathBuf, isize),
292 /// Post-substitution inline assembly from LLVM.
296 impl From<PathBuf> for FileName {
297 fn from(p: PathBuf) -> Self {
298 assert!(!p.to_string_lossy().ends_with('>'));
299 FileName::Real(RealFileName::LocalPath(p))
303 #[derive(Clone, Copy, Eq, PartialEq, Hash, Debug)]
304 pub enum FileNameDisplayPreference {
309 pub struct FileNameDisplay<'a> {
311 display_pref: FileNameDisplayPreference,
314 impl fmt::Display for FileNameDisplay<'_> {
315 fn fmt(&self, fmt: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
319 write!(fmt, "{}", name.to_string_lossy(self.display_pref))
321 QuoteExpansion(_) => write!(fmt, "<quote expansion>"),
322 MacroExpansion(_) => write!(fmt, "<macro expansion>"),
323 Anon(_) => write!(fmt, "<anon>"),
324 ProcMacroSourceCode(_) => write!(fmt, "<proc-macro source code>"),
325 CfgSpec(_) => write!(fmt, "<cfgspec>"),
326 CliCrateAttr(_) => write!(fmt, "<crate attribute>"),
327 Custom(ref s) => write!(fmt, "<{}>", s),
328 DocTest(ref path, _) => write!(fmt, "{}", path.display()),
329 InlineAsm(_) => write!(fmt, "<inline asm>"),
334 impl FileNameDisplay<'_> {
335 pub fn to_string_lossy(&self) -> Cow<'_, str> {
337 FileName::Real(ref inner) => inner.to_string_lossy(self.display_pref),
338 _ => Cow::from(format!("{}", self)),
344 pub fn is_real(&self) -> bool {
350 | ProcMacroSourceCode(_)
356 | InlineAsm(_) => false,
360 pub fn prefer_remapped(&self) -> FileNameDisplay<'_> {
361 FileNameDisplay { inner: self, display_pref: FileNameDisplayPreference::Remapped }
364 // This may include transient local filesystem information.
365 // Must not be embedded in build outputs.
366 pub fn prefer_local(&self) -> FileNameDisplay<'_> {
367 FileNameDisplay { inner: self, display_pref: FileNameDisplayPreference::Local }
370 pub fn display(&self, display_pref: FileNameDisplayPreference) -> FileNameDisplay<'_> {
371 FileNameDisplay { inner: self, display_pref }
374 pub fn macro_expansion_source_code(src: &str) -> FileName {
375 let mut hasher = StableHasher::new();
376 src.hash(&mut hasher);
377 FileName::MacroExpansion(hasher.finish())
380 pub fn anon_source_code(src: &str) -> FileName {
381 let mut hasher = StableHasher::new();
382 src.hash(&mut hasher);
383 FileName::Anon(hasher.finish())
386 pub fn proc_macro_source_code(src: &str) -> FileName {
387 let mut hasher = StableHasher::new();
388 src.hash(&mut hasher);
389 FileName::ProcMacroSourceCode(hasher.finish())
392 pub fn cfg_spec_source_code(src: &str) -> FileName {
393 let mut hasher = StableHasher::new();
394 src.hash(&mut hasher);
395 FileName::QuoteExpansion(hasher.finish())
398 pub fn cli_crate_attr_source_code(src: &str) -> FileName {
399 let mut hasher = StableHasher::new();
400 src.hash(&mut hasher);
401 FileName::CliCrateAttr(hasher.finish())
404 pub fn doc_test_source_code(path: PathBuf, line: isize) -> FileName {
405 FileName::DocTest(path, line)
408 pub fn inline_asm_source_code(src: &str) -> FileName {
409 let mut hasher = StableHasher::new();
410 src.hash(&mut hasher);
411 FileName::InlineAsm(hasher.finish())
415 /// Represents a span.
417 /// Spans represent a region of code, used for error reporting. Positions in spans
418 /// are *absolute* positions from the beginning of the [`SourceMap`], not positions
419 /// relative to [`SourceFile`]s. Methods on the `SourceMap` can be used to relate spans back
420 /// to the original source.
422 /// You must be careful if the span crosses more than one file, since you will not be
423 /// able to use many of the functions on spans in source_map and you cannot assume
424 /// that the length of the span is equal to `span.hi - span.lo`; there may be space in the
425 /// [`BytePos`] range between files.
427 /// `SpanData` is public because `Span` uses a thread-local interner and can't be
428 /// sent to other threads, but some pieces of performance infra run in a separate thread.
429 /// Using `Span` is generally preferred.
430 #[derive(Clone, Copy, Hash, PartialEq, Eq, Ord, PartialOrd)]
431 pub struct SpanData {
434 /// Information about where the macro came from, if this piece of
435 /// code was created by a macro expansion.
436 pub ctxt: SyntaxContext,
437 pub parent: Option<LocalDefId>,
442 pub fn span(&self) -> Span {
443 Span::new(self.lo, self.hi, self.ctxt, self.parent)
446 pub fn with_lo(&self, lo: BytePos) -> Span {
447 Span::new(lo, self.hi, self.ctxt, self.parent)
450 pub fn with_hi(&self, hi: BytePos) -> Span {
451 Span::new(self.lo, hi, self.ctxt, self.parent)
454 pub fn with_ctxt(&self, ctxt: SyntaxContext) -> Span {
455 Span::new(self.lo, self.hi, ctxt, self.parent)
458 pub fn with_parent(&self, parent: Option<LocalDefId>) -> Span {
459 Span::new(self.lo, self.hi, self.ctxt, parent)
461 /// Returns `true` if this is a dummy span with any hygienic context.
463 pub fn is_dummy(self) -> bool {
464 self.lo.0 == 0 && self.hi.0 == 0
466 /// Returns `true` if `self` fully encloses `other`.
467 pub fn contains(self, other: Self) -> bool {
468 self.lo <= other.lo && other.hi <= self.hi
472 // The interner is pointed to by a thread local value which is only set on the main thread
473 // with parallelization is disabled. So we don't allow `Span` to transfer between threads
474 // to avoid panics and other errors, even though it would be memory safe to do so.
475 #[cfg(not(parallel_compiler))]
476 impl !Send for Span {}
477 #[cfg(not(parallel_compiler))]
478 impl !Sync for Span {}
480 impl PartialOrd for Span {
481 fn partial_cmp(&self, rhs: &Self) -> Option<Ordering> {
482 PartialOrd::partial_cmp(&self.data(), &rhs.data())
486 fn cmp(&self, rhs: &Self) -> Ordering {
487 Ord::cmp(&self.data(), &rhs.data())
491 /// A collection of `Span`s.
493 /// Spans have two orthogonal attributes:
495 /// - They can be *primary spans*. In this case they are the locus of
496 /// the error, and would be rendered with `^^^`.
497 /// - They can have a *label*. In this case, the label is written next
498 /// to the mark in the snippet when we render.
499 #[derive(Clone, Debug, Hash, PartialEq, Eq, Encodable, Decodable)]
500 pub struct MultiSpan {
501 primary_spans: Vec<Span>,
502 span_labels: Vec<(Span, String)>,
507 pub fn lo(self) -> BytePos {
511 pub fn with_lo(self, lo: BytePos) -> Span {
512 self.data().with_lo(lo)
515 pub fn hi(self) -> BytePos {
519 pub fn with_hi(self, hi: BytePos) -> Span {
520 self.data().with_hi(hi)
523 pub fn ctxt(self) -> SyntaxContext {
524 self.data_untracked().ctxt
527 pub fn with_ctxt(self, ctxt: SyntaxContext) -> Span {
528 self.data_untracked().with_ctxt(ctxt)
531 pub fn parent(self) -> Option<LocalDefId> {
535 pub fn with_parent(self, ctxt: Option<LocalDefId>) -> Span {
536 self.data().with_parent(ctxt)
539 /// Returns `true` if this is a dummy span with any hygienic context.
541 pub fn is_dummy(self) -> bool {
542 self.data_untracked().is_dummy()
545 /// Returns `true` if this span comes from a macro or desugaring.
547 pub fn from_expansion(self) -> bool {
548 self.ctxt() != SyntaxContext::root()
551 /// Returns `true` if `span` originates in a derive-macro's expansion.
552 pub fn in_derive_expansion(self) -> bool {
553 matches!(self.ctxt().outer_expn_data().kind, ExpnKind::Macro(MacroKind::Derive, _))
557 pub fn with_root_ctxt(lo: BytePos, hi: BytePos) -> Span {
558 Span::new(lo, hi, SyntaxContext::root(), None)
561 /// Returns a new span representing an empty span at the beginning of this span.
563 pub fn shrink_to_lo(self) -> Span {
564 let span = self.data_untracked();
565 span.with_hi(span.lo)
567 /// Returns a new span representing an empty span at the end of this span.
569 pub fn shrink_to_hi(self) -> Span {
570 let span = self.data_untracked();
571 span.with_lo(span.hi)
575 /// Returns `true` if `hi == lo`.
576 pub fn is_empty(&self) -> bool {
577 let span = self.data_untracked();
581 /// Returns `self` if `self` is not the dummy span, and `other` otherwise.
582 pub fn substitute_dummy(self, other: Span) -> Span {
583 if self.is_dummy() { other } else { self }
586 /// Returns `true` if `self` fully encloses `other`.
587 pub fn contains(self, other: Span) -> bool {
588 let span = self.data();
589 let other = other.data();
593 /// Returns `true` if `self` touches `other`.
594 pub fn overlaps(self, other: Span) -> bool {
595 let span = self.data();
596 let other = other.data();
597 span.lo < other.hi && other.lo < span.hi
600 /// Returns `true` if the spans are equal with regards to the source text.
602 /// Use this instead of `==` when either span could be generated code,
603 /// and you only care that they point to the same bytes of source text.
604 pub fn source_equal(&self, other: &Span) -> bool {
605 let span = self.data();
606 let other = other.data();
607 span.lo == other.lo && span.hi == other.hi
610 /// Returns `Some(span)`, where the start is trimmed by the end of `other`.
611 pub fn trim_start(self, other: Span) -> Option<Span> {
612 let span = self.data();
613 let other = other.data();
614 if span.hi > other.hi { Some(span.with_lo(cmp::max(span.lo, other.hi))) } else { None }
617 /// Returns the source span -- this is either the supplied span, or the span for
618 /// the macro callsite that expanded to it.
619 pub fn source_callsite(self) -> Span {
620 let expn_data = self.ctxt().outer_expn_data();
621 if !expn_data.is_root() { expn_data.call_site.source_callsite() } else { self }
624 /// The `Span` for the tokens in the previous macro expansion from which `self` was generated,
626 pub fn parent_callsite(self) -> Option<Span> {
627 let expn_data = self.ctxt().outer_expn_data();
628 if !expn_data.is_root() { Some(expn_data.call_site) } else { None }
631 /// Walk down the expansion ancestors to find a span that's contained within `outer`.
632 pub fn find_ancestor_inside(mut self, outer: Span) -> Option<Span> {
633 while !outer.contains(self) {
634 self = self.parent_callsite()?;
639 /// Edition of the crate from which this span came.
640 pub fn edition(self) -> edition::Edition {
641 self.ctxt().edition()
645 pub fn rust_2015(&self) -> bool {
646 self.edition() == edition::Edition::Edition2015
650 pub fn rust_2018(&self) -> bool {
651 self.edition() >= edition::Edition::Edition2018
655 pub fn rust_2021(&self) -> bool {
656 self.edition() >= edition::Edition::Edition2021
659 /// Returns the source callee.
661 /// Returns `None` if the supplied span has no expansion trace,
662 /// else returns the `ExpnData` for the macro definition
663 /// corresponding to the source callsite.
664 pub fn source_callee(self) -> Option<ExpnData> {
665 fn source_callee(expn_data: ExpnData) -> ExpnData {
666 let next_expn_data = expn_data.call_site.ctxt().outer_expn_data();
667 if !next_expn_data.is_root() { source_callee(next_expn_data) } else { expn_data }
669 let expn_data = self.ctxt().outer_expn_data();
670 if !expn_data.is_root() { Some(source_callee(expn_data)) } else { None }
673 /// Checks if a span is "internal" to a macro in which `#[unstable]`
674 /// items can be used (that is, a macro marked with
675 /// `#[allow_internal_unstable]`).
676 pub fn allows_unstable(&self, feature: Symbol) -> bool {
679 .allow_internal_unstable
680 .map_or(false, |features| features.iter().any(|&f| f == feature))
683 /// Checks if this span arises from a compiler desugaring of kind `kind`.
684 pub fn is_desugaring(&self, kind: DesugaringKind) -> bool {
685 match self.ctxt().outer_expn_data().kind {
686 ExpnKind::Desugaring(k) => k == kind,
691 /// Returns the compiler desugaring that created this span, or `None`
692 /// if this span is not from a desugaring.
693 pub fn desugaring_kind(&self) -> Option<DesugaringKind> {
694 match self.ctxt().outer_expn_data().kind {
695 ExpnKind::Desugaring(k) => Some(k),
700 /// Checks if a span is "internal" to a macro in which `unsafe`
701 /// can be used without triggering the `unsafe_code` lint.
702 // (that is, a macro marked with `#[allow_internal_unsafe]`).
703 pub fn allows_unsafe(&self) -> bool {
704 self.ctxt().outer_expn_data().allow_internal_unsafe
707 pub fn macro_backtrace(mut self) -> impl Iterator<Item = ExpnData> {
708 let mut prev_span = DUMMY_SP;
709 std::iter::from_fn(move || {
711 let expn_data = self.ctxt().outer_expn_data();
712 if expn_data.is_root() {
716 let is_recursive = expn_data.call_site.source_equal(&prev_span);
719 self = expn_data.call_site;
721 // Don't print recursive invocations.
723 return Some(expn_data);
729 /// Returns a `Span` that would enclose both `self` and `end`.
733 /// self lorem ipsum end
734 /// ^^^^^^^^^^^^^^^^^^^^
736 pub fn to(self, end: Span) -> Span {
737 let span_data = self.data();
738 let end_data = end.data();
739 // FIXME(jseyfried): `self.ctxt` should always equal `end.ctxt` here (cf. issue #23480).
740 // Return the macro span on its own to avoid weird diagnostic output. It is preferable to
741 // have an incomplete span than a completely nonsensical one.
742 if span_data.ctxt != end_data.ctxt {
743 if span_data.ctxt == SyntaxContext::root() {
745 } else if end_data.ctxt == SyntaxContext::root() {
748 // Both spans fall within a macro.
749 // FIXME(estebank): check if it is the *same* macro.
752 cmp::min(span_data.lo, end_data.lo),
753 cmp::max(span_data.hi, end_data.hi),
754 if span_data.ctxt == SyntaxContext::root() { end_data.ctxt } else { span_data.ctxt },
755 if span_data.parent == end_data.parent { span_data.parent } else { None },
759 /// Returns a `Span` between the end of `self` to the beginning of `end`.
763 /// self lorem ipsum end
766 pub fn between(self, end: Span) -> Span {
767 let span = self.data();
768 let end = end.data();
772 if end.ctxt == SyntaxContext::root() { end.ctxt } else { span.ctxt },
773 if span.parent == end.parent { span.parent } else { None },
777 /// Returns a `Span` from the beginning of `self` until the beginning of `end`.
781 /// self lorem ipsum end
782 /// ^^^^^^^^^^^^^^^^^
784 pub fn until(self, end: Span) -> Span {
785 let span = self.data();
786 let end = end.data();
790 if end.ctxt == SyntaxContext::root() { end.ctxt } else { span.ctxt },
791 if span.parent == end.parent { span.parent } else { None },
795 pub fn from_inner(self, inner: InnerSpan) -> Span {
796 let span = self.data();
798 span.lo + BytePos::from_usize(inner.start),
799 span.lo + BytePos::from_usize(inner.end),
805 /// Equivalent of `Span::def_site` from the proc macro API,
806 /// except that the location is taken from the `self` span.
807 pub fn with_def_site_ctxt(self, expn_id: ExpnId) -> Span {
808 self.with_ctxt_from_mark(expn_id, Transparency::Opaque)
811 /// Equivalent of `Span::call_site` from the proc macro API,
812 /// except that the location is taken from the `self` span.
813 pub fn with_call_site_ctxt(&self, expn_id: ExpnId) -> Span {
814 self.with_ctxt_from_mark(expn_id, Transparency::Transparent)
817 /// Equivalent of `Span::mixed_site` from the proc macro API,
818 /// except that the location is taken from the `self` span.
819 pub fn with_mixed_site_ctxt(&self, expn_id: ExpnId) -> Span {
820 self.with_ctxt_from_mark(expn_id, Transparency::SemiTransparent)
823 /// Produces a span with the same location as `self` and context produced by a macro with the
824 /// given ID and transparency, assuming that macro was defined directly and not produced by
825 /// some other macro (which is the case for built-in and procedural macros).
826 pub fn with_ctxt_from_mark(self, expn_id: ExpnId, transparency: Transparency) -> Span {
827 self.with_ctxt(SyntaxContext::root().apply_mark(expn_id, transparency))
831 pub fn apply_mark(self, expn_id: ExpnId, transparency: Transparency) -> Span {
832 let span = self.data();
833 span.with_ctxt(span.ctxt.apply_mark(expn_id, transparency))
837 pub fn remove_mark(&mut self) -> ExpnId {
838 let mut span = self.data();
839 let mark = span.ctxt.remove_mark();
840 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
845 pub fn adjust(&mut self, expn_id: ExpnId) -> Option<ExpnId> {
846 let mut span = self.data();
847 let mark = span.ctxt.adjust(expn_id);
848 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
853 pub fn normalize_to_macros_2_0_and_adjust(&mut self, expn_id: ExpnId) -> Option<ExpnId> {
854 let mut span = self.data();
855 let mark = span.ctxt.normalize_to_macros_2_0_and_adjust(expn_id);
856 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
861 pub fn glob_adjust(&mut self, expn_id: ExpnId, glob_span: Span) -> Option<Option<ExpnId>> {
862 let mut span = self.data();
863 let mark = span.ctxt.glob_adjust(expn_id, glob_span);
864 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
869 pub fn reverse_glob_adjust(
873 ) -> Option<Option<ExpnId>> {
874 let mut span = self.data();
875 let mark = span.ctxt.reverse_glob_adjust(expn_id, glob_span);
876 *self = Span::new(span.lo, span.hi, span.ctxt, span.parent);
881 pub fn normalize_to_macros_2_0(self) -> Span {
882 let span = self.data();
883 span.with_ctxt(span.ctxt.normalize_to_macros_2_0())
887 pub fn normalize_to_macro_rules(self) -> Span {
888 let span = self.data();
889 span.with_ctxt(span.ctxt.normalize_to_macro_rules())
893 /// A span together with some additional data.
894 #[derive(Clone, Debug)]
895 pub struct SpanLabel {
896 /// The span we are going to include in the final snippet.
899 /// Is this a primary span? This is the "locus" of the message,
900 /// and is indicated with a `^^^^` underline, versus `----`.
901 pub is_primary: bool,
903 /// What label should we attach to this span (if any)?
904 pub label: Option<String>,
907 impl Default for Span {
908 fn default() -> Self {
913 impl<E: Encoder> Encodable<E> for Span {
914 default fn encode(&self, s: &mut E) -> Result<(), E::Error> {
915 let span = self.data();
916 s.emit_struct(false, |s| {
917 s.emit_struct_field("lo", true, |s| span.lo.encode(s))?;
918 s.emit_struct_field("hi", false, |s| span.hi.encode(s))
922 impl<D: Decoder> Decodable<D> for Span {
923 default fn decode(s: &mut D) -> Result<Span, D::Error> {
925 let lo = d.read_struct_field("lo", Decodable::decode)?;
926 let hi = d.read_struct_field("hi", Decodable::decode)?;
928 Ok(Span::new(lo, hi, SyntaxContext::root(), None))
933 /// Calls the provided closure, using the provided `SourceMap` to format
934 /// any spans that are debug-printed during the closure's execution.
936 /// Normally, the global `TyCtxt` is used to retrieve the `SourceMap`
937 /// (see `rustc_interface::callbacks::span_debug1`). However, some parts
938 /// of the compiler (e.g. `rustc_parse`) may debug-print `Span`s before
939 /// a `TyCtxt` is available. In this case, we fall back to
940 /// the `SourceMap` provided to this function. If that is not available,
941 /// we fall back to printing the raw `Span` field values.
942 pub fn with_source_map<T, F: FnOnce() -> T>(source_map: Lrc<SourceMap>, f: F) -> T {
943 with_session_globals(|session_globals| {
944 *session_globals.source_map.borrow_mut() = Some(source_map);
946 struct ClearSourceMap;
947 impl Drop for ClearSourceMap {
949 with_session_globals(|session_globals| {
950 session_globals.source_map.borrow_mut().take();
955 let _guard = ClearSourceMap;
959 pub fn debug_with_source_map(
961 f: &mut fmt::Formatter<'_>,
962 source_map: &SourceMap,
964 write!(f, "{} ({:?})", source_map.span_to_diagnostic_string(span), span.ctxt())
967 pub fn default_span_debug(span: Span, f: &mut fmt::Formatter<'_>) -> fmt::Result {
968 with_session_globals(|session_globals| {
969 if let Some(source_map) = &*session_globals.source_map.borrow() {
970 debug_with_source_map(span, f, source_map)
972 f.debug_struct("Span")
973 .field("lo", &span.lo())
974 .field("hi", &span.hi())
975 .field("ctxt", &span.ctxt())
981 impl fmt::Debug for Span {
982 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
983 (*SPAN_DEBUG)(*self, f)
987 impl fmt::Debug for SpanData {
988 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
989 (*SPAN_DEBUG)(Span::new(self.lo, self.hi, self.ctxt, self.parent), f)
995 pub fn new() -> MultiSpan {
996 MultiSpan { primary_spans: vec![], span_labels: vec![] }
999 pub fn from_span(primary_span: Span) -> MultiSpan {
1000 MultiSpan { primary_spans: vec![primary_span], span_labels: vec![] }
1003 pub fn from_spans(mut vec: Vec<Span>) -> MultiSpan {
1005 MultiSpan { primary_spans: vec, span_labels: vec![] }
1008 pub fn push_span_label(&mut self, span: Span, label: String) {
1009 self.span_labels.push((span, label));
1012 /// Selects the first primary span (if any).
1013 pub fn primary_span(&self) -> Option<Span> {
1014 self.primary_spans.first().cloned()
1017 /// Returns all primary spans.
1018 pub fn primary_spans(&self) -> &[Span] {
1022 /// Returns `true` if any of the primary spans are displayable.
1023 pub fn has_primary_spans(&self) -> bool {
1024 self.primary_spans.iter().any(|sp| !sp.is_dummy())
1027 /// Returns `true` if this contains only a dummy primary span with any hygienic context.
1028 pub fn is_dummy(&self) -> bool {
1029 let mut is_dummy = true;
1030 for span in &self.primary_spans {
1031 if !span.is_dummy() {
1038 /// Replaces all occurrences of one Span with another. Used to move `Span`s in areas that don't
1039 /// display well (like std macros). Returns whether replacements occurred.
1040 pub fn replace(&mut self, before: Span, after: Span) -> bool {
1041 let mut replacements_occurred = false;
1042 for primary_span in &mut self.primary_spans {
1043 if *primary_span == before {
1044 *primary_span = after;
1045 replacements_occurred = true;
1048 for span_label in &mut self.span_labels {
1049 if span_label.0 == before {
1050 span_label.0 = after;
1051 replacements_occurred = true;
1054 replacements_occurred
1057 /// Returns the strings to highlight. We always ensure that there
1058 /// is an entry for each of the primary spans -- for each primary
1059 /// span `P`, if there is at least one label with span `P`, we return
1060 /// those labels (marked as primary). But otherwise we return
1061 /// `SpanLabel` instances with empty labels.
1062 pub fn span_labels(&self) -> Vec<SpanLabel> {
1063 let is_primary = |span| self.primary_spans.contains(&span);
1065 let mut span_labels = self
1068 .map(|&(span, ref label)| SpanLabel {
1070 is_primary: is_primary(span),
1071 label: Some(label.clone()),
1073 .collect::<Vec<_>>();
1075 for &span in &self.primary_spans {
1076 if !span_labels.iter().any(|sl| sl.span == span) {
1077 span_labels.push(SpanLabel { span, is_primary: true, label: None });
1084 /// Returns `true` if any of the span labels is displayable.
1085 pub fn has_span_labels(&self) -> bool {
1086 self.span_labels.iter().any(|(sp, _)| !sp.is_dummy())
1090 impl From<Span> for MultiSpan {
1091 fn from(span: Span) -> MultiSpan {
1092 MultiSpan::from_span(span)
1096 impl From<Vec<Span>> for MultiSpan {
1097 fn from(spans: Vec<Span>) -> MultiSpan {
1098 MultiSpan::from_spans(spans)
1102 /// Identifies an offset of a multi-byte character in a `SourceFile`.
1103 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1104 pub struct MultiByteChar {
1105 /// The absolute offset of the character in the `SourceMap`.
1107 /// The number of bytes, `>= 2`.
1111 /// Identifies an offset of a non-narrow character in a `SourceFile`.
1112 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1113 pub enum NonNarrowChar {
1114 /// Represents a zero-width character.
1116 /// Represents a wide (full-width) character.
1118 /// Represents a tab character, represented visually with a width of 4 characters.
1122 impl NonNarrowChar {
1123 fn new(pos: BytePos, width: usize) -> Self {
1125 0 => NonNarrowChar::ZeroWidth(pos),
1126 2 => NonNarrowChar::Wide(pos),
1127 4 => NonNarrowChar::Tab(pos),
1128 _ => panic!("width {} given for non-narrow character", width),
1132 /// Returns the absolute offset of the character in the `SourceMap`.
1133 pub fn pos(&self) -> BytePos {
1135 NonNarrowChar::ZeroWidth(p) | NonNarrowChar::Wide(p) | NonNarrowChar::Tab(p) => p,
1139 /// Returns the width of the character, 0 (zero-width) or 2 (wide).
1140 pub fn width(&self) -> usize {
1142 NonNarrowChar::ZeroWidth(_) => 0,
1143 NonNarrowChar::Wide(_) => 2,
1144 NonNarrowChar::Tab(_) => 4,
1149 impl Add<BytePos> for NonNarrowChar {
1152 fn add(self, rhs: BytePos) -> Self {
1154 NonNarrowChar::ZeroWidth(pos) => NonNarrowChar::ZeroWidth(pos + rhs),
1155 NonNarrowChar::Wide(pos) => NonNarrowChar::Wide(pos + rhs),
1156 NonNarrowChar::Tab(pos) => NonNarrowChar::Tab(pos + rhs),
1161 impl Sub<BytePos> for NonNarrowChar {
1164 fn sub(self, rhs: BytePos) -> Self {
1166 NonNarrowChar::ZeroWidth(pos) => NonNarrowChar::ZeroWidth(pos - rhs),
1167 NonNarrowChar::Wide(pos) => NonNarrowChar::Wide(pos - rhs),
1168 NonNarrowChar::Tab(pos) => NonNarrowChar::Tab(pos - rhs),
1173 /// Identifies an offset of a character that was normalized away from `SourceFile`.
1174 #[derive(Copy, Clone, Encodable, Decodable, Eq, PartialEq, Debug)]
1175 pub struct NormalizedPos {
1176 /// The absolute offset of the character in the `SourceMap`.
1178 /// The difference between original and normalized string at position.
1182 #[derive(PartialEq, Eq, Clone, Debug)]
1183 pub enum ExternalSource {
1184 /// No external source has to be loaded, since the `SourceFile` represents a local crate.
1187 kind: ExternalSourceKind,
1188 /// This SourceFile's byte-offset within the source_map of its original crate.
1189 original_start_pos: BytePos,
1190 /// The end of this SourceFile within the source_map of its original crate.
1191 original_end_pos: BytePos,
1195 /// The state of the lazy external source loading mechanism of a `SourceFile`.
1196 #[derive(PartialEq, Eq, Clone, Debug)]
1197 pub enum ExternalSourceKind {
1198 /// The external source has been loaded already.
1199 Present(Lrc<String>),
1200 /// No attempt has been made to load the external source.
1202 /// A failed attempt has been made to load the external source.
1207 impl ExternalSource {
1208 pub fn get_source(&self) -> Option<&Lrc<String>> {
1210 ExternalSource::Foreign { kind: ExternalSourceKind::Present(ref src), .. } => Some(src),
1217 pub struct OffsetOverflowError;
1219 #[derive(Copy, Clone, Debug, PartialEq, Eq, PartialOrd, Ord, Hash, Encodable, Decodable)]
1220 pub enum SourceFileHashAlgorithm {
1226 impl FromStr for SourceFileHashAlgorithm {
1229 fn from_str(s: &str) -> Result<SourceFileHashAlgorithm, ()> {
1231 "md5" => Ok(SourceFileHashAlgorithm::Md5),
1232 "sha1" => Ok(SourceFileHashAlgorithm::Sha1),
1233 "sha256" => Ok(SourceFileHashAlgorithm::Sha256),
1239 rustc_data_structures::impl_stable_hash_via_hash!(SourceFileHashAlgorithm);
1241 /// The hash of the on-disk source file used for debug info.
1242 #[derive(Copy, Clone, PartialEq, Eq, Debug)]
1243 #[derive(HashStable_Generic, Encodable, Decodable)]
1244 pub struct SourceFileHash {
1245 pub kind: SourceFileHashAlgorithm,
1249 impl SourceFileHash {
1250 pub fn new(kind: SourceFileHashAlgorithm, src: &str) -> SourceFileHash {
1251 let mut hash = SourceFileHash { kind, value: Default::default() };
1252 let len = hash.hash_len();
1253 let value = &mut hash.value[..len];
1254 let data = src.as_bytes();
1256 SourceFileHashAlgorithm::Md5 => {
1257 value.copy_from_slice(&Md5::digest(data));
1259 SourceFileHashAlgorithm::Sha1 => {
1260 value.copy_from_slice(&Sha1::digest(data));
1262 SourceFileHashAlgorithm::Sha256 => {
1263 value.copy_from_slice(&Sha256::digest(data));
1269 /// Check if the stored hash matches the hash of the string.
1270 pub fn matches(&self, src: &str) -> bool {
1271 Self::new(self.kind, src) == *self
1274 /// The bytes of the hash.
1275 pub fn hash_bytes(&self) -> &[u8] {
1276 let len = self.hash_len();
1280 fn hash_len(&self) -> usize {
1282 SourceFileHashAlgorithm::Md5 => 16,
1283 SourceFileHashAlgorithm::Sha1 => 20,
1284 SourceFileHashAlgorithm::Sha256 => 32,
1289 /// A single source in the [`SourceMap`].
1291 pub struct SourceFile {
1292 /// The name of the file that the source came from. Source that doesn't
1293 /// originate from files has names between angle brackets by convention
1294 /// (e.g., `<anon>`).
1296 /// The complete source code.
1297 pub src: Option<Lrc<String>>,
1298 /// The source code's hash.
1299 pub src_hash: SourceFileHash,
1300 /// The external source code (used for external crates, which will have a `None`
1301 /// value as `self.src`.
1302 pub external_src: Lock<ExternalSource>,
1303 /// The start position of this source in the `SourceMap`.
1304 pub start_pos: BytePos,
1305 /// The end position of this source in the `SourceMap`.
1306 pub end_pos: BytePos,
1307 /// Locations of lines beginnings in the source code.
1308 pub lines: Vec<BytePos>,
1309 /// Locations of multi-byte characters in the source code.
1310 pub multibyte_chars: Vec<MultiByteChar>,
1311 /// Width of characters that are not narrow in the source code.
1312 pub non_narrow_chars: Vec<NonNarrowChar>,
1313 /// Locations of characters removed during normalization.
1314 pub normalized_pos: Vec<NormalizedPos>,
1315 /// A hash of the filename, used for speeding up hashing in incremental compilation.
1316 pub name_hash: u128,
1317 /// Indicates which crate this `SourceFile` was imported from.
1321 impl<S: Encoder> Encodable<S> for SourceFile {
1322 fn encode(&self, s: &mut S) -> Result<(), S::Error> {
1323 s.emit_struct(false, |s| {
1324 s.emit_struct_field("name", true, |s| self.name.encode(s))?;
1325 s.emit_struct_field("src_hash", false, |s| self.src_hash.encode(s))?;
1326 s.emit_struct_field("start_pos", false, |s| self.start_pos.encode(s))?;
1327 s.emit_struct_field("end_pos", false, |s| self.end_pos.encode(s))?;
1328 s.emit_struct_field("lines", false, |s| {
1329 let lines = &self.lines[..];
1330 // Store the length.
1331 s.emit_u32(lines.len() as u32)?;
1333 if !lines.is_empty() {
1334 // In order to preserve some space, we exploit the fact that
1335 // the lines list is sorted and individual lines are
1336 // probably not that long. Because of that we can store lines
1337 // as a difference list, using as little space as possible
1338 // for the differences.
1339 let max_line_length = if lines.len() == 1 {
1344 .map(|&[fst, snd]| snd - fst)
1345 .map(|bp| bp.to_usize())
1350 let bytes_per_diff: u8 = match max_line_length {
1352 0x100..=0xFFFF => 2,
1356 // Encode the number of bytes used per diff.
1357 bytes_per_diff.encode(s)?;
1359 // Encode the first element.
1360 lines[0].encode(s)?;
1362 let diff_iter = lines[..].array_windows().map(|&[fst, snd]| snd - fst);
1364 match bytes_per_diff {
1366 for diff in diff_iter {
1367 (diff.0 as u8).encode(s)?
1371 for diff in diff_iter {
1372 (diff.0 as u16).encode(s)?
1376 for diff in diff_iter {
1380 _ => unreachable!(),
1386 s.emit_struct_field("multibyte_chars", false, |s| self.multibyte_chars.encode(s))?;
1387 s.emit_struct_field("non_narrow_chars", false, |s| self.non_narrow_chars.encode(s))?;
1388 s.emit_struct_field("name_hash", false, |s| self.name_hash.encode(s))?;
1389 s.emit_struct_field("normalized_pos", false, |s| self.normalized_pos.encode(s))?;
1390 s.emit_struct_field("cnum", false, |s| self.cnum.encode(s))
1395 impl<D: Decoder> Decodable<D> for SourceFile {
1396 fn decode(d: &mut D) -> Result<SourceFile, D::Error> {
1398 let name: FileName = d.read_struct_field("name", |d| Decodable::decode(d))?;
1399 let src_hash: SourceFileHash =
1400 d.read_struct_field("src_hash", |d| Decodable::decode(d))?;
1401 let start_pos: BytePos = d.read_struct_field("start_pos", |d| Decodable::decode(d))?;
1402 let end_pos: BytePos = d.read_struct_field("end_pos", |d| Decodable::decode(d))?;
1403 let lines: Vec<BytePos> = d.read_struct_field("lines", |d| {
1404 let num_lines: u32 = Decodable::decode(d)?;
1405 let mut lines = Vec::with_capacity(num_lines as usize);
1408 // Read the number of bytes used per diff.
1409 let bytes_per_diff: u8 = Decodable::decode(d)?;
1411 // Read the first element.
1412 let mut line_start: BytePos = Decodable::decode(d)?;
1413 lines.push(line_start);
1415 for _ in 1..num_lines {
1416 let diff = match bytes_per_diff {
1417 1 => d.read_u8()? as u32,
1418 2 => d.read_u16()? as u32,
1420 _ => unreachable!(),
1423 line_start = line_start + BytePos(diff);
1425 lines.push(line_start);
1431 let multibyte_chars: Vec<MultiByteChar> =
1432 d.read_struct_field("multibyte_chars", |d| Decodable::decode(d))?;
1433 let non_narrow_chars: Vec<NonNarrowChar> =
1434 d.read_struct_field("non_narrow_chars", |d| Decodable::decode(d))?;
1435 let name_hash: u128 = d.read_struct_field("name_hash", |d| Decodable::decode(d))?;
1436 let normalized_pos: Vec<NormalizedPos> =
1437 d.read_struct_field("normalized_pos", |d| Decodable::decode(d))?;
1438 let cnum: CrateNum = d.read_struct_field("cnum", |d| Decodable::decode(d))?;
1445 // Unused - the metadata decoder will construct
1446 // a new SourceFile, filling in `external_src` properly
1447 external_src: Lock::new(ExternalSource::Unneeded),
1459 impl fmt::Debug for SourceFile {
1460 fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result {
1461 write!(fmt, "SourceFile({:?})", self.name)
1470 hash_kind: SourceFileHashAlgorithm,
1472 // Compute the file hash before any normalization.
1473 let src_hash = SourceFileHash::new(hash_kind, &src);
1474 let normalized_pos = normalize_src(&mut src, start_pos);
1477 let mut hasher: StableHasher = StableHasher::new();
1478 name.hash(&mut hasher);
1479 hasher.finish::<u128>()
1481 let end_pos = start_pos.to_usize() + src.len();
1482 assert!(end_pos <= u32::MAX as usize);
1484 let (lines, multibyte_chars, non_narrow_chars) =
1485 analyze_source_file::analyze_source_file(&src[..], start_pos);
1489 src: Some(Lrc::new(src)),
1491 external_src: Lock::new(ExternalSource::Unneeded),
1493 end_pos: Pos::from_usize(end_pos),
1503 /// Returns the `BytePos` of the beginning of the current line.
1504 pub fn line_begin_pos(&self, pos: BytePos) -> BytePos {
1505 let line_index = self.lookup_line(pos).unwrap();
1506 self.lines[line_index]
1509 /// Add externally loaded source.
1510 /// If the hash of the input doesn't match or no input is supplied via None,
1511 /// it is interpreted as an error and the corresponding enum variant is set.
1512 /// The return value signifies whether some kind of source is present.
1513 pub fn add_external_src<F>(&self, get_src: F) -> bool
1515 F: FnOnce() -> Option<String>,
1518 *self.external_src.borrow(),
1519 ExternalSource::Foreign { kind: ExternalSourceKind::AbsentOk, .. }
1521 let src = get_src();
1522 let mut external_src = self.external_src.borrow_mut();
1523 // Check that no-one else have provided the source while we were getting it
1524 if let ExternalSource::Foreign {
1525 kind: src_kind @ ExternalSourceKind::AbsentOk, ..
1526 } = &mut *external_src
1528 if let Some(mut src) = src {
1529 // The src_hash needs to be computed on the pre-normalized src.
1530 if self.src_hash.matches(&src) {
1531 normalize_src(&mut src, BytePos::from_usize(0));
1532 *src_kind = ExternalSourceKind::Present(Lrc::new(src));
1536 *src_kind = ExternalSourceKind::AbsentErr;
1541 self.src.is_some() || external_src.get_source().is_some()
1544 self.src.is_some() || self.external_src.borrow().get_source().is_some()
1548 /// Gets a line from the list of pre-computed line-beginnings.
1549 /// The line number here is 0-based.
1550 pub fn get_line(&self, line_number: usize) -> Option<Cow<'_, str>> {
1551 fn get_until_newline(src: &str, begin: usize) -> &str {
1552 // We can't use `lines.get(line_number+1)` because we might
1553 // be parsing when we call this function and thus the current
1554 // line is the last one we have line info for.
1555 let slice = &src[begin..];
1556 match slice.find('\n') {
1557 Some(e) => &slice[..e],
1563 let line = self.lines.get(line_number)?;
1564 let begin: BytePos = *line - self.start_pos;
1568 if let Some(ref src) = self.src {
1569 Some(Cow::from(get_until_newline(src, begin)))
1570 } else if let Some(src) = self.external_src.borrow().get_source() {
1571 Some(Cow::Owned(String::from(get_until_newline(src, begin))))
1577 pub fn is_real_file(&self) -> bool {
1581 pub fn is_imported(&self) -> bool {
1585 pub fn count_lines(&self) -> usize {
1589 /// Finds the line containing the given position. The return value is the
1590 /// index into the `lines` array of this `SourceFile`, not the 1-based line
1591 /// number. If the source_file is empty or the position is located before the
1592 /// first line, `None` is returned.
1593 pub fn lookup_line(&self, pos: BytePos) -> Option<usize> {
1594 match self.lines.binary_search(&pos) {
1595 Ok(idx) => Some(idx),
1597 Err(idx) => Some(idx - 1),
1601 pub fn line_bounds(&self, line_index: usize) -> Range<BytePos> {
1602 if self.is_empty() {
1603 return self.start_pos..self.end_pos;
1606 assert!(line_index < self.lines.len());
1607 if line_index == (self.lines.len() - 1) {
1608 self.lines[line_index]..self.end_pos
1610 self.lines[line_index]..self.lines[line_index + 1]
1614 /// Returns whether or not the file contains the given `SourceMap` byte
1615 /// position. The position one past the end of the file is considered to be
1616 /// contained by the file. This implies that files for which `is_empty`
1617 /// returns true still contain one byte position according to this function.
1619 pub fn contains(&self, byte_pos: BytePos) -> bool {
1620 byte_pos >= self.start_pos && byte_pos <= self.end_pos
1624 pub fn is_empty(&self) -> bool {
1625 self.start_pos == self.end_pos
1628 /// Calculates the original byte position relative to the start of the file
1629 /// based on the given byte position.
1630 pub fn original_relative_byte_pos(&self, pos: BytePos) -> BytePos {
1631 // Diff before any records is 0. Otherwise use the previously recorded
1632 // diff as that applies to the following characters until a new diff
1634 let diff = match self.normalized_pos.binary_search_by(|np| np.pos.cmp(&pos)) {
1635 Ok(i) => self.normalized_pos[i].diff,
1636 Err(i) if i == 0 => 0,
1637 Err(i) => self.normalized_pos[i - 1].diff,
1640 BytePos::from_u32(pos.0 - self.start_pos.0 + diff)
1643 /// Converts an absolute `BytePos` to a `CharPos` relative to the `SourceFile`.
1644 pub fn bytepos_to_file_charpos(&self, bpos: BytePos) -> CharPos {
1645 // The number of extra bytes due to multibyte chars in the `SourceFile`.
1646 let mut total_extra_bytes = 0;
1648 for mbc in self.multibyte_chars.iter() {
1649 debug!("{}-byte char at {:?}", mbc.bytes, mbc.pos);
1651 // Every character is at least one byte, so we only
1652 // count the actual extra bytes.
1653 total_extra_bytes += mbc.bytes as u32 - 1;
1654 // We should never see a byte position in the middle of a
1656 assert!(bpos.to_u32() >= mbc.pos.to_u32() + mbc.bytes as u32);
1662 assert!(self.start_pos.to_u32() + total_extra_bytes <= bpos.to_u32());
1663 CharPos(bpos.to_usize() - self.start_pos.to_usize() - total_extra_bytes as usize)
1666 /// Looks up the file's (1-based) line number and (0-based `CharPos`) column offset, for a
1667 /// given `BytePos`.
1668 pub fn lookup_file_pos(&self, pos: BytePos) -> (usize, CharPos) {
1669 let chpos = self.bytepos_to_file_charpos(pos);
1670 match self.lookup_line(pos) {
1672 let line = a + 1; // Line numbers start at 1
1673 let linebpos = self.lines[a];
1674 let linechpos = self.bytepos_to_file_charpos(linebpos);
1675 let col = chpos - linechpos;
1676 debug!("byte pos {:?} is on the line at byte pos {:?}", pos, linebpos);
1677 debug!("char pos {:?} is on the line at char pos {:?}", chpos, linechpos);
1678 debug!("byte is on line: {}", line);
1679 assert!(chpos >= linechpos);
1686 /// Looks up the file's (1-based) line number, (0-based `CharPos`) column offset, and (0-based)
1687 /// column offset when displayed, for a given `BytePos`.
1688 pub fn lookup_file_pos_with_col_display(&self, pos: BytePos) -> (usize, CharPos, usize) {
1689 let (line, col_or_chpos) = self.lookup_file_pos(pos);
1691 let col = col_or_chpos;
1692 let linebpos = self.lines[line - 1];
1694 let start_width_idx = self
1696 .binary_search_by_key(&linebpos, |x| x.pos())
1697 .unwrap_or_else(|x| x);
1698 let end_width_idx = self
1700 .binary_search_by_key(&pos, |x| x.pos())
1701 .unwrap_or_else(|x| x);
1702 let special_chars = end_width_idx - start_width_idx;
1703 let non_narrow: usize = self.non_narrow_chars[start_width_idx..end_width_idx]
1707 col.0 - special_chars + non_narrow
1709 (line, col, col_display)
1711 let chpos = col_or_chpos;
1713 let end_width_idx = self
1715 .binary_search_by_key(&pos, |x| x.pos())
1716 .unwrap_or_else(|x| x);
1717 let non_narrow: usize =
1718 self.non_narrow_chars[0..end_width_idx].iter().map(|x| x.width()).sum();
1719 chpos.0 - end_width_idx + non_narrow
1721 (0, chpos, col_display)
1726 /// Normalizes the source code and records the normalizations.
1727 fn normalize_src(src: &mut String, start_pos: BytePos) -> Vec<NormalizedPos> {
1728 let mut normalized_pos = vec![];
1729 remove_bom(src, &mut normalized_pos);
1730 normalize_newlines(src, &mut normalized_pos);
1732 // Offset all the positions by start_pos to match the final file positions.
1733 for np in &mut normalized_pos {
1734 np.pos.0 += start_pos.0;
1740 /// Removes UTF-8 BOM, if any.
1741 fn remove_bom(src: &mut String, normalized_pos: &mut Vec<NormalizedPos>) {
1742 if src.starts_with('\u{feff}') {
1744 normalized_pos.push(NormalizedPos { pos: BytePos(0), diff: 3 });
1748 /// Replaces `\r\n` with `\n` in-place in `src`.
1750 /// Returns error if there's a lone `\r` in the string.
1751 fn normalize_newlines(src: &mut String, normalized_pos: &mut Vec<NormalizedPos>) {
1752 if !src.as_bytes().contains(&b'\r') {
1756 // We replace `\r\n` with `\n` in-place, which doesn't break utf-8 encoding.
1757 // While we *can* call `as_mut_vec` and do surgery on the live string
1758 // directly, let's rather steal the contents of `src`. This makes the code
1759 // safe even if a panic occurs.
1761 let mut buf = std::mem::replace(src, String::new()).into_bytes();
1762 let mut gap_len = 0;
1763 let mut tail = buf.as_mut_slice();
1765 let original_gap = normalized_pos.last().map_or(0, |l| l.diff);
1767 let idx = match find_crlf(&tail[gap_len..]) {
1769 Some(idx) => idx + gap_len,
1771 tail.copy_within(gap_len..idx, 0);
1772 tail = &mut tail[idx - gap_len..];
1773 if tail.len() == gap_len {
1776 cursor += idx - gap_len;
1778 normalized_pos.push(NormalizedPos {
1779 pos: BytePos::from_usize(cursor + 1),
1780 diff: original_gap + gap_len as u32,
1784 // Account for removed `\r`.
1785 // After `set_len`, `buf` is guaranteed to contain utf-8 again.
1786 let new_len = buf.len() - gap_len;
1788 buf.set_len(new_len);
1789 *src = String::from_utf8_unchecked(buf);
1792 fn find_crlf(src: &[u8]) -> Option<usize> {
1793 let mut search_idx = 0;
1794 while let Some(idx) = find_cr(&src[search_idx..]) {
1795 if src[search_idx..].get(idx + 1) != Some(&b'\n') {
1796 search_idx += idx + 1;
1799 return Some(search_idx + idx);
1804 fn find_cr(src: &[u8]) -> Option<usize> {
1805 src.iter().position(|&b| b == b'\r')
1809 // _____________________________________________________________________________
1810 // Pos, BytePos, CharPos
1814 fn from_usize(n: usize) -> Self;
1815 fn to_usize(&self) -> usize;
1816 fn from_u32(n: u32) -> Self;
1817 fn to_u32(&self) -> u32;
1820 macro_rules! impl_pos {
1824 $vis:vis struct $ident:ident($inner_vis:vis $inner_ty:ty);
1829 $vis struct $ident($inner_vis $inner_ty);
1831 impl Pos for $ident {
1833 fn from_usize(n: usize) -> $ident {
1834 $ident(n as $inner_ty)
1838 fn to_usize(&self) -> usize {
1843 fn from_u32(n: u32) -> $ident {
1844 $ident(n as $inner_ty)
1848 fn to_u32(&self) -> u32 {
1853 impl Add for $ident {
1854 type Output = $ident;
1857 fn add(self, rhs: $ident) -> $ident {
1858 $ident(self.0 + rhs.0)
1862 impl Sub for $ident {
1863 type Output = $ident;
1866 fn sub(self, rhs: $ident) -> $ident {
1867 $ident(self.0 - rhs.0)
1877 /// Keep this small (currently 32-bits), as AST contains a lot of them.
1878 #[derive(Clone, Copy, PartialEq, Eq, Hash, PartialOrd, Ord, Debug)]
1879 pub struct BytePos(pub u32);
1881 /// A character offset.
1883 /// Because of multibyte UTF-8 characters, a byte offset
1884 /// is not equivalent to a character offset. The [`SourceMap`] will convert [`BytePos`]
1885 /// values to `CharPos` values as necessary.
1886 #[derive(Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Debug)]
1887 pub struct CharPos(pub usize);
1890 impl<S: rustc_serialize::Encoder> Encodable<S> for BytePos {
1891 fn encode(&self, s: &mut S) -> Result<(), S::Error> {
1896 impl<D: rustc_serialize::Decoder> Decodable<D> for BytePos {
1897 fn decode(d: &mut D) -> Result<BytePos, D::Error> {
1898 Ok(BytePos(d.read_u32()?))
1902 // _____________________________________________________________________________
1903 // Loc, SourceFileAndLine, SourceFileAndBytePos
1906 /// A source code location used for error reporting.
1907 #[derive(Debug, Clone)]
1909 /// Information about the original source.
1910 pub file: Lrc<SourceFile>,
1911 /// The (1-based) line number.
1913 /// The (0-based) column offset.
1915 /// The (0-based) column offset when displayed.
1916 pub col_display: usize,
1919 // Used to be structural records.
1921 pub struct SourceFileAndLine {
1922 pub sf: Lrc<SourceFile>,
1926 pub struct SourceFileAndBytePos {
1927 pub sf: Lrc<SourceFile>,
1931 #[derive(Copy, Clone, Debug, PartialEq, Eq)]
1932 pub struct LineInfo {
1933 /// Index of line, starting from 0.
1934 pub line_index: usize,
1936 /// Column in line where span begins, starting from 0.
1937 pub start_col: CharPos,
1939 /// Column in line where span ends, starting from 0, exclusive.
1940 pub end_col: CharPos,
1943 pub struct FileLines {
1944 pub file: Lrc<SourceFile>,
1945 pub lines: Vec<LineInfo>,
1948 pub static SPAN_DEBUG: AtomicRef<fn(Span, &mut fmt::Formatter<'_>) -> fmt::Result> =
1949 AtomicRef::new(&(default_span_debug as fn(_, &mut fmt::Formatter<'_>) -> _));
1950 pub static SPAN_TRACK: AtomicRef<fn(LocalDefId)> = AtomicRef::new(&((|_| {}) as fn(_)));
1952 // _____________________________________________________________________________
1953 // SpanLinesError, SpanSnippetError, DistinctSources, MalformedSourceMapPositions
1956 pub type FileLinesResult = Result<FileLines, SpanLinesError>;
1958 #[derive(Clone, PartialEq, Eq, Debug)]
1959 pub enum SpanLinesError {
1960 DistinctSources(DistinctSources),
1963 #[derive(Clone, PartialEq, Eq, Debug)]
1964 pub enum SpanSnippetError {
1965 IllFormedSpan(Span),
1966 DistinctSources(DistinctSources),
1967 MalformedForSourcemap(MalformedSourceMapPositions),
1968 SourceNotAvailable { filename: FileName },
1971 #[derive(Clone, PartialEq, Eq, Debug)]
1972 pub struct DistinctSources {
1973 pub begin: (FileName, BytePos),
1974 pub end: (FileName, BytePos),
1977 #[derive(Clone, PartialEq, Eq, Debug)]
1978 pub struct MalformedSourceMapPositions {
1980 pub source_len: usize,
1981 pub begin_pos: BytePos,
1982 pub end_pos: BytePos,
1985 /// Range inside of a `Span` used for diagnostics when we only have access to relative positions.
1986 #[derive(Copy, Clone, PartialEq, Eq, Debug)]
1987 pub struct InnerSpan {
1993 pub fn new(start: usize, end: usize) -> InnerSpan {
1994 InnerSpan { start, end }
1998 /// Requirements for a `StableHashingContext` to be used in this crate.
2000 /// This is a hack to allow using the [`HashStable_Generic`] derive macro
2001 /// instead of implementing everything in rustc_middle.
2002 pub trait HashStableContext {
2003 fn def_path_hash(&self, def_id: DefId) -> DefPathHash;
2004 fn hash_spans(&self) -> bool;
2005 fn def_span(&self, def_id: LocalDefId) -> Span;
2006 fn span_data_to_lines_and_cols(
2009 ) -> Option<(Lrc<SourceFile>, usize, BytePos, usize, BytePos)>;
2012 impl<CTX> HashStable<CTX> for Span
2014 CTX: HashStableContext,
2016 /// Hashes a span in a stable way. We can't directly hash the span's `BytePos`
2017 /// fields (that would be similar to hashing pointers, since those are just
2018 /// offsets into the `SourceMap`). Instead, we hash the (file name, line, column)
2019 /// triple, which stays the same even if the containing `SourceFile` has moved
2020 /// within the `SourceMap`.
2022 /// Also note that we are hashing byte offsets for the column, not unicode
2023 /// codepoint offsets. For the purpose of the hash that's sufficient.
2024 /// Also, hashing filenames is expensive so we avoid doing it twice when the
2025 /// span starts and ends in the same file, which is almost always the case.
2026 fn hash_stable(&self, ctx: &mut CTX, hasher: &mut StableHasher) {
2027 const TAG_VALID_SPAN: u8 = 0;
2028 const TAG_INVALID_SPAN: u8 = 1;
2029 const TAG_RELATIVE_SPAN: u8 = 2;
2031 if !ctx.hash_spans() {
2035 let span = self.data_untracked();
2036 span.ctxt.hash_stable(ctx, hasher);
2037 span.parent.hash_stable(ctx, hasher);
2039 if span.is_dummy() {
2040 Hash::hash(&TAG_INVALID_SPAN, hasher);
2044 if let Some(parent) = span.parent {
2045 let def_span = ctx.def_span(parent).data_untracked();
2046 if def_span.contains(span) {
2047 // This span is enclosed in a definition: only hash the relative position.
2048 Hash::hash(&TAG_RELATIVE_SPAN, hasher);
2049 (span.lo - def_span.lo).to_u32().hash_stable(ctx, hasher);
2050 (span.hi - def_span.lo).to_u32().hash_stable(ctx, hasher);
2055 // If this is not an empty or invalid span, we want to hash the last
2056 // position that belongs to it, as opposed to hashing the first
2057 // position past it.
2058 let (file, line_lo, col_lo, line_hi, col_hi) = match ctx.span_data_to_lines_and_cols(&span)
2062 Hash::hash(&TAG_INVALID_SPAN, hasher);
2067 Hash::hash(&TAG_VALID_SPAN, hasher);
2068 // We truncate the stable ID hash and line and column numbers. The chances
2069 // of causing a collision this way should be minimal.
2070 Hash::hash(&(file.name_hash as u64), hasher);
2072 // Hash both the length and the end location (line/column) of a span. If we
2073 // hash only the length, for example, then two otherwise equal spans with
2074 // different end locations will have the same hash. This can cause a problem
2075 // during incremental compilation wherein a previous result for a query that
2076 // depends on the end location of a span will be incorrectly reused when the
2077 // end location of the span it depends on has changed (see issue #74890). A
2078 // similar analysis applies if some query depends specifically on the length
2079 // of the span, but we only hash the end location. So hash both.
2081 let col_lo_trunc = (col_lo.0 as u64) & 0xFF;
2082 let line_lo_trunc = ((line_lo as u64) & 0xFF_FF_FF) << 8;
2083 let col_hi_trunc = (col_hi.0 as u64) & 0xFF << 32;
2084 let line_hi_trunc = ((line_hi as u64) & 0xFF_FF_FF) << 40;
2085 let col_line = col_lo_trunc | line_lo_trunc | col_hi_trunc | line_hi_trunc;
2086 let len = (span.hi - span.lo).0;
2087 Hash::hash(&col_line, hasher);
2088 Hash::hash(&len, hasher);