]> git.proxmox.com Git - rustc.git/blob - compiler/rustc_parse_format/src/lib.rs
New upstream version 1.48.0~beta.8+dfsg1
[rustc.git] / compiler / rustc_parse_format / src / lib.rs
1 //! Macro support for format strings
2 //!
3 //! These structures are used when parsing format strings for the compiler.
4 //! Parsing does not happen at runtime: structures of `std::fmt::rt` are
5 //! generated instead.
6
7 #![doc(
8 html_root_url = "https://doc.rust-lang.org/nightly/nightly-rustc/",
9 html_playground_url = "https://play.rust-lang.org/",
10 test(attr(deny(warnings)))
11 )]
12 #![feature(nll)]
13 #![feature(or_patterns)]
14 #![feature(bool_to_option)]
15
16 pub use Alignment::*;
17 pub use Count::*;
18 pub use Flag::*;
19 pub use Piece::*;
20 pub use Position::*;
21
22 use std::iter;
23 use std::str;
24 use std::string;
25
26 use rustc_span::{InnerSpan, Symbol};
27
28 /// The type of format string that we are parsing.
29 #[derive(Copy, Clone, Debug, Eq, PartialEq)]
30 pub enum ParseMode {
31 /// A normal format string as per `format_args!`.
32 Format,
33 /// An inline assembly template string for `asm!`.
34 InlineAsm,
35 }
36
37 #[derive(Copy, Clone)]
38 struct InnerOffset(usize);
39
40 impl InnerOffset {
41 fn to(self, end: InnerOffset) -> InnerSpan {
42 InnerSpan::new(self.0, end.0)
43 }
44 }
45
46 /// A piece is a portion of the format string which represents the next part
47 /// to emit. These are emitted as a stream by the `Parser` class.
48 #[derive(Copy, Clone, Debug, PartialEq)]
49 pub enum Piece<'a> {
50 /// A literal string which should directly be emitted
51 String(&'a str),
52 /// This describes that formatting should process the next argument (as
53 /// specified inside) for emission.
54 NextArgument(Argument<'a>),
55 }
56
57 /// Representation of an argument specification.
58 #[derive(Copy, Clone, Debug, PartialEq)]
59 pub struct Argument<'a> {
60 /// Where to find this argument
61 pub position: Position,
62 /// How to format the argument
63 pub format: FormatSpec<'a>,
64 }
65
66 /// Specification for the formatting of an argument in the format string.
67 #[derive(Copy, Clone, Debug, PartialEq)]
68 pub struct FormatSpec<'a> {
69 /// Optionally specified character to fill alignment with.
70 pub fill: Option<char>,
71 /// Optionally specified alignment.
72 pub align: Alignment,
73 /// Packed version of various flags provided.
74 pub flags: u32,
75 /// The integer precision to use.
76 pub precision: Count,
77 /// The span of the precision formatting flag (for diagnostics).
78 pub precision_span: Option<InnerSpan>,
79 /// The string width requested for the resulting format.
80 pub width: Count,
81 /// The span of the width formatting flag (for diagnostics).
82 pub width_span: Option<InnerSpan>,
83 /// The descriptor string representing the name of the format desired for
84 /// this argument, this can be empty or any number of characters, although
85 /// it is required to be one word.
86 pub ty: &'a str,
87 /// The span of the descriptor string (for diagnostics).
88 pub ty_span: Option<InnerSpan>,
89 }
90
91 /// Enum describing where an argument for a format can be located.
92 #[derive(Copy, Clone, Debug, PartialEq)]
93 pub enum Position {
94 /// The argument is implied to be located at an index
95 ArgumentImplicitlyIs(usize),
96 /// The argument is located at a specific index given in the format
97 ArgumentIs(usize),
98 /// The argument has a name.
99 ArgumentNamed(Symbol),
100 }
101
102 impl Position {
103 pub fn index(&self) -> Option<usize> {
104 match self {
105 ArgumentIs(i) | ArgumentImplicitlyIs(i) => Some(*i),
106 _ => None,
107 }
108 }
109 }
110
111 /// Enum of alignments which are supported.
112 #[derive(Copy, Clone, Debug, PartialEq)]
113 pub enum Alignment {
114 /// The value will be aligned to the left.
115 AlignLeft,
116 /// The value will be aligned to the right.
117 AlignRight,
118 /// The value will be aligned in the center.
119 AlignCenter,
120 /// The value will take on a default alignment.
121 AlignUnknown,
122 }
123
124 /// Various flags which can be applied to format strings. The meaning of these
125 /// flags is defined by the formatters themselves.
126 #[derive(Copy, Clone, Debug, PartialEq)]
127 pub enum Flag {
128 /// A `+` will be used to denote positive numbers.
129 FlagSignPlus,
130 /// A `-` will be used to denote negative numbers. This is the default.
131 FlagSignMinus,
132 /// An alternate form will be used for the value. In the case of numbers,
133 /// this means that the number will be prefixed with the supplied string.
134 FlagAlternate,
135 /// For numbers, this means that the number will be padded with zeroes,
136 /// and the sign (`+` or `-`) will precede them.
137 FlagSignAwareZeroPad,
138 /// For Debug / `?`, format integers in lower-case hexadecimal.
139 FlagDebugLowerHex,
140 /// For Debug / `?`, format integers in upper-case hexadecimal.
141 FlagDebugUpperHex,
142 }
143
144 /// A count is used for the precision and width parameters of an integer, and
145 /// can reference either an argument or a literal integer.
146 #[derive(Copy, Clone, Debug, PartialEq)]
147 pub enum Count {
148 /// The count is specified explicitly.
149 CountIs(usize),
150 /// The count is specified by the argument with the given name.
151 CountIsName(Symbol),
152 /// The count is specified by the argument at the given index.
153 CountIsParam(usize),
154 /// The count is implied and cannot be explicitly specified.
155 CountImplied,
156 }
157
158 pub struct ParseError {
159 pub description: string::String,
160 pub note: Option<string::String>,
161 pub label: string::String,
162 pub span: InnerSpan,
163 pub secondary_label: Option<(string::String, InnerSpan)>,
164 }
165
166 /// The parser structure for interpreting the input format string. This is
167 /// modeled as an iterator over `Piece` structures to form a stream of tokens
168 /// being output.
169 ///
170 /// This is a recursive-descent parser for the sake of simplicity, and if
171 /// necessary there's probably lots of room for improvement performance-wise.
172 pub struct Parser<'a> {
173 mode: ParseMode,
174 input: &'a str,
175 cur: iter::Peekable<str::CharIndices<'a>>,
176 /// Error messages accumulated during parsing
177 pub errors: Vec<ParseError>,
178 /// Current position of implicit positional argument pointer
179 pub curarg: usize,
180 /// `Some(raw count)` when the string is "raw", used to position spans correctly
181 style: Option<usize>,
182 /// Start and end byte offset of every successfully parsed argument
183 pub arg_places: Vec<InnerSpan>,
184 /// Characters that need to be shifted
185 skips: Vec<usize>,
186 /// Span of the last opening brace seen, used for error reporting
187 last_opening_brace: Option<InnerSpan>,
188 /// Whether the source string is comes from `println!` as opposed to `format!` or `print!`
189 append_newline: bool,
190 /// Whether this formatting string is a literal or it comes from a macro.
191 pub is_literal: bool,
192 /// Start position of the current line.
193 cur_line_start: usize,
194 /// Start and end byte offset of every line of the format string. Excludes
195 /// newline characters and leading whitespace.
196 pub line_spans: Vec<InnerSpan>,
197 }
198
199 impl<'a> Iterator for Parser<'a> {
200 type Item = Piece<'a>;
201
202 fn next(&mut self) -> Option<Piece<'a>> {
203 if let Some(&(pos, c)) = self.cur.peek() {
204 match c {
205 '{' => {
206 let curr_last_brace = self.last_opening_brace;
207 let byte_pos = self.to_span_index(pos);
208 self.last_opening_brace = Some(byte_pos.to(InnerOffset(byte_pos.0 + 1)));
209 self.cur.next();
210 if self.consume('{') {
211 self.last_opening_brace = curr_last_brace;
212
213 Some(String(self.string(pos + 1)))
214 } else {
215 let arg = self.argument();
216 if let Some(end) = self.must_consume('}') {
217 let start = self.to_span_index(pos);
218 let end = self.to_span_index(end + 1);
219 if self.is_literal {
220 self.arg_places.push(start.to(end));
221 }
222 }
223 Some(NextArgument(arg))
224 }
225 }
226 '}' => {
227 self.cur.next();
228 if self.consume('}') {
229 Some(String(self.string(pos + 1)))
230 } else {
231 let err_pos = self.to_span_index(pos);
232 self.err_with_note(
233 "unmatched `}` found",
234 "unmatched `}`",
235 "if you intended to print `}`, you can escape it using `}}`",
236 err_pos.to(err_pos),
237 );
238 None
239 }
240 }
241 _ => Some(String(self.string(pos))),
242 }
243 } else {
244 if self.is_literal {
245 let start = self.to_span_index(self.cur_line_start);
246 let end = self.to_span_index(self.input.len());
247 let span = start.to(end);
248 if self.line_spans.last() != Some(&span) {
249 self.line_spans.push(span);
250 }
251 }
252 None
253 }
254 }
255 }
256
257 impl<'a> Parser<'a> {
258 /// Creates a new parser for the given format string
259 pub fn new(
260 s: &'a str,
261 style: Option<usize>,
262 snippet: Option<string::String>,
263 append_newline: bool,
264 mode: ParseMode,
265 ) -> Parser<'a> {
266 let (skips, is_literal) = find_skips_from_snippet(snippet, style);
267 Parser {
268 mode,
269 input: s,
270 cur: s.char_indices().peekable(),
271 errors: vec![],
272 curarg: 0,
273 style,
274 arg_places: vec![],
275 skips,
276 last_opening_brace: None,
277 append_newline,
278 is_literal,
279 cur_line_start: 0,
280 line_spans: vec![],
281 }
282 }
283
284 /// Notifies of an error. The message doesn't actually need to be of type
285 /// String, but I think it does when this eventually uses conditions so it
286 /// might as well start using it now.
287 fn err<S1: Into<string::String>, S2: Into<string::String>>(
288 &mut self,
289 description: S1,
290 label: S2,
291 span: InnerSpan,
292 ) {
293 self.errors.push(ParseError {
294 description: description.into(),
295 note: None,
296 label: label.into(),
297 span,
298 secondary_label: None,
299 });
300 }
301
302 /// Notifies of an error. The message doesn't actually need to be of type
303 /// String, but I think it does when this eventually uses conditions so it
304 /// might as well start using it now.
305 fn err_with_note<
306 S1: Into<string::String>,
307 S2: Into<string::String>,
308 S3: Into<string::String>,
309 >(
310 &mut self,
311 description: S1,
312 label: S2,
313 note: S3,
314 span: InnerSpan,
315 ) {
316 self.errors.push(ParseError {
317 description: description.into(),
318 note: Some(note.into()),
319 label: label.into(),
320 span,
321 secondary_label: None,
322 });
323 }
324
325 /// Optionally consumes the specified character. If the character is not at
326 /// the current position, then the current iterator isn't moved and `false` is
327 /// returned, otherwise the character is consumed and `true` is returned.
328 fn consume(&mut self, c: char) -> bool {
329 self.consume_pos(c).is_some()
330 }
331
332 /// Optionally consumes the specified character. If the character is not at
333 /// the current position, then the current iterator isn't moved and `None` is
334 /// returned, otherwise the character is consumed and the current position is
335 /// returned.
336 fn consume_pos(&mut self, c: char) -> Option<usize> {
337 if let Some(&(pos, maybe)) = self.cur.peek() {
338 if c == maybe {
339 self.cur.next();
340 return Some(pos);
341 }
342 }
343 None
344 }
345
346 fn to_span_index(&self, pos: usize) -> InnerOffset {
347 let mut pos = pos;
348 // This handles the raw string case, the raw argument is the number of #
349 // in r###"..."### (we need to add one because of the `r`).
350 let raw = self.style.map(|raw| raw + 1).unwrap_or(0);
351 for skip in &self.skips {
352 if pos > *skip {
353 pos += 1;
354 } else if pos == *skip && raw == 0 {
355 pos += 1;
356 } else {
357 break;
358 }
359 }
360 InnerOffset(raw + pos + 1)
361 }
362
363 /// Forces consumption of the specified character. If the character is not
364 /// found, an error is emitted.
365 fn must_consume(&mut self, c: char) -> Option<usize> {
366 self.ws();
367
368 if let Some(&(pos, maybe)) = self.cur.peek() {
369 if c == maybe {
370 self.cur.next();
371 Some(pos)
372 } else {
373 let pos = self.to_span_index(pos);
374 let description = format!("expected `'}}'`, found `{:?}`", maybe);
375 let label = "expected `}`".to_owned();
376 let (note, secondary_label) = if c == '}' {
377 (
378 Some(
379 "if you intended to print `{`, you can escape it using `{{`".to_owned(),
380 ),
381 self.last_opening_brace
382 .map(|sp| ("because of this opening brace".to_owned(), sp)),
383 )
384 } else {
385 (None, None)
386 };
387 self.errors.push(ParseError {
388 description,
389 note,
390 label,
391 span: pos.to(pos),
392 secondary_label,
393 });
394 None
395 }
396 } else {
397 let description = format!("expected `{:?}` but string was terminated", c);
398 // point at closing `"`
399 let pos = self.input.len() - if self.append_newline { 1 } else { 0 };
400 let pos = self.to_span_index(pos);
401 if c == '}' {
402 let label = format!("expected `{:?}`", c);
403 let (note, secondary_label) = if c == '}' {
404 (
405 Some(
406 "if you intended to print `{`, you can escape it using `{{`".to_owned(),
407 ),
408 self.last_opening_brace
409 .map(|sp| ("because of this opening brace".to_owned(), sp)),
410 )
411 } else {
412 (None, None)
413 };
414 self.errors.push(ParseError {
415 description,
416 note,
417 label,
418 span: pos.to(pos),
419 secondary_label,
420 });
421 } else {
422 self.err(description, format!("expected `{:?}`", c), pos.to(pos));
423 }
424 None
425 }
426 }
427
428 /// Consumes all whitespace characters until the first non-whitespace character
429 fn ws(&mut self) {
430 while let Some(&(_, c)) = self.cur.peek() {
431 if c.is_whitespace() {
432 self.cur.next();
433 } else {
434 break;
435 }
436 }
437 }
438
439 /// Parses all of a string which is to be considered a "raw literal" in a
440 /// format string. This is everything outside of the braces.
441 fn string(&mut self, start: usize) -> &'a str {
442 // we may not consume the character, peek the iterator
443 while let Some(&(pos, c)) = self.cur.peek() {
444 match c {
445 '{' | '}' => {
446 return &self.input[start..pos];
447 }
448 '\n' if self.is_literal => {
449 let start = self.to_span_index(self.cur_line_start);
450 let end = self.to_span_index(pos);
451 self.line_spans.push(start.to(end));
452 self.cur_line_start = pos + 1;
453 self.cur.next();
454 }
455 _ => {
456 if self.is_literal && pos == self.cur_line_start && c.is_whitespace() {
457 self.cur_line_start = pos + c.len_utf8();
458 }
459 self.cur.next();
460 }
461 }
462 }
463 &self.input[start..self.input.len()]
464 }
465
466 /// Parses an `Argument` structure, or what's contained within braces inside the format string.
467 fn argument(&mut self) -> Argument<'a> {
468 let pos = self.position();
469 let format = match self.mode {
470 ParseMode::Format => self.format(),
471 ParseMode::InlineAsm => self.inline_asm(),
472 };
473
474 // Resolve position after parsing format spec.
475 let pos = match pos {
476 Some(position) => position,
477 None => {
478 let i = self.curarg;
479 self.curarg += 1;
480 ArgumentImplicitlyIs(i)
481 }
482 };
483
484 Argument { position: pos, format }
485 }
486
487 /// Parses a positional argument for a format. This could either be an
488 /// integer index of an argument, a named argument, or a blank string.
489 /// Returns `Some(parsed_position)` if the position is not implicitly
490 /// consuming a macro argument, `None` if it's the case.
491 fn position(&mut self) -> Option<Position> {
492 if let Some(i) = self.integer() {
493 Some(ArgumentIs(i))
494 } else {
495 match self.cur.peek() {
496 Some(&(_, c)) if rustc_lexer::is_id_start(c) => {
497 Some(ArgumentNamed(Symbol::intern(self.word())))
498 }
499
500 // This is an `ArgumentNext`.
501 // Record the fact and do the resolution after parsing the
502 // format spec, to make things like `{:.*}` work.
503 _ => None,
504 }
505 }
506 }
507
508 /// Parses a format specifier at the current position, returning all of the
509 /// relevant information in the `FormatSpec` struct.
510 fn format(&mut self) -> FormatSpec<'a> {
511 let mut spec = FormatSpec {
512 fill: None,
513 align: AlignUnknown,
514 flags: 0,
515 precision: CountImplied,
516 precision_span: None,
517 width: CountImplied,
518 width_span: None,
519 ty: &self.input[..0],
520 ty_span: None,
521 };
522 if !self.consume(':') {
523 return spec;
524 }
525
526 // fill character
527 if let Some(&(_, c)) = self.cur.peek() {
528 if let Some((_, '>' | '<' | '^')) = self.cur.clone().nth(1) {
529 spec.fill = Some(c);
530 self.cur.next();
531 }
532 }
533 // Alignment
534 if self.consume('<') {
535 spec.align = AlignLeft;
536 } else if self.consume('>') {
537 spec.align = AlignRight;
538 } else if self.consume('^') {
539 spec.align = AlignCenter;
540 }
541 // Sign flags
542 if self.consume('+') {
543 spec.flags |= 1 << (FlagSignPlus as u32);
544 } else if self.consume('-') {
545 spec.flags |= 1 << (FlagSignMinus as u32);
546 }
547 // Alternate marker
548 if self.consume('#') {
549 spec.flags |= 1 << (FlagAlternate as u32);
550 }
551 // Width and precision
552 let mut havewidth = false;
553
554 if self.consume('0') {
555 // small ambiguity with '0$' as a format string. In theory this is a
556 // '0' flag and then an ill-formatted format string with just a '$'
557 // and no count, but this is better if we instead interpret this as
558 // no '0' flag and '0$' as the width instead.
559 if self.consume('$') {
560 spec.width = CountIsParam(0);
561 havewidth = true;
562 } else {
563 spec.flags |= 1 << (FlagSignAwareZeroPad as u32);
564 }
565 }
566 if !havewidth {
567 let width_span_start = if let Some((pos, _)) = self.cur.peek() { *pos } else { 0 };
568 let (w, sp) = self.count(width_span_start);
569 spec.width = w;
570 spec.width_span = sp;
571 }
572 if let Some(start) = self.consume_pos('.') {
573 if let Some(end) = self.consume_pos('*') {
574 // Resolve `CountIsNextParam`.
575 // We can do this immediately as `position` is resolved later.
576 let i = self.curarg;
577 self.curarg += 1;
578 spec.precision = CountIsParam(i);
579 spec.precision_span =
580 Some(self.to_span_index(start).to(self.to_span_index(end + 1)));
581 } else {
582 let (p, sp) = self.count(start);
583 spec.precision = p;
584 spec.precision_span = sp;
585 }
586 }
587 let ty_span_start = self.cur.peek().map(|(pos, _)| *pos);
588 // Optional radix followed by the actual format specifier
589 if self.consume('x') {
590 if self.consume('?') {
591 spec.flags |= 1 << (FlagDebugLowerHex as u32);
592 spec.ty = "?";
593 } else {
594 spec.ty = "x";
595 }
596 } else if self.consume('X') {
597 if self.consume('?') {
598 spec.flags |= 1 << (FlagDebugUpperHex as u32);
599 spec.ty = "?";
600 } else {
601 spec.ty = "X";
602 }
603 } else if self.consume('?') {
604 spec.ty = "?";
605 } else {
606 spec.ty = self.word();
607 let ty_span_end = self.cur.peek().map(|(pos, _)| *pos);
608 if !spec.ty.is_empty() {
609 spec.ty_span = ty_span_start
610 .and_then(|s| ty_span_end.map(|e| (s, e)))
611 .map(|(start, end)| self.to_span_index(start).to(self.to_span_index(end)));
612 }
613 }
614 spec
615 }
616
617 /// Parses an inline assembly template modifier at the current position, returning the modifier
618 /// in the `ty` field of the `FormatSpec` struct.
619 fn inline_asm(&mut self) -> FormatSpec<'a> {
620 let mut spec = FormatSpec {
621 fill: None,
622 align: AlignUnknown,
623 flags: 0,
624 precision: CountImplied,
625 precision_span: None,
626 width: CountImplied,
627 width_span: None,
628 ty: &self.input[..0],
629 ty_span: None,
630 };
631 if !self.consume(':') {
632 return spec;
633 }
634
635 let ty_span_start = self.cur.peek().map(|(pos, _)| *pos);
636 spec.ty = self.word();
637 let ty_span_end = self.cur.peek().map(|(pos, _)| *pos);
638 if !spec.ty.is_empty() {
639 spec.ty_span = ty_span_start
640 .and_then(|s| ty_span_end.map(|e| (s, e)))
641 .map(|(start, end)| self.to_span_index(start).to(self.to_span_index(end)));
642 }
643
644 spec
645 }
646
647 /// Parses a `Count` parameter at the current position. This does not check
648 /// for 'CountIsNextParam' because that is only used in precision, not
649 /// width.
650 fn count(&mut self, start: usize) -> (Count, Option<InnerSpan>) {
651 if let Some(i) = self.integer() {
652 if let Some(end) = self.consume_pos('$') {
653 let span = self.to_span_index(start).to(self.to_span_index(end + 1));
654 (CountIsParam(i), Some(span))
655 } else {
656 (CountIs(i), None)
657 }
658 } else {
659 let tmp = self.cur.clone();
660 let word = self.word();
661 if word.is_empty() {
662 self.cur = tmp;
663 (CountImplied, None)
664 } else if self.consume('$') {
665 (CountIsName(Symbol::intern(word)), None)
666 } else {
667 self.cur = tmp;
668 (CountImplied, None)
669 }
670 }
671 }
672
673 /// Parses a word starting at the current position. A word is the same as
674 /// Rust identifier, except that it can't start with `_` character.
675 fn word(&mut self) -> &'a str {
676 let start = match self.cur.peek() {
677 Some(&(pos, c)) if rustc_lexer::is_id_start(c) => {
678 self.cur.next();
679 pos
680 }
681 _ => {
682 return "";
683 }
684 };
685 let mut end = None;
686 while let Some(&(pos, c)) = self.cur.peek() {
687 if rustc_lexer::is_id_continue(c) {
688 self.cur.next();
689 } else {
690 end = Some(pos);
691 break;
692 }
693 }
694 let end = end.unwrap_or(self.input.len());
695 let word = &self.input[start..end];
696 if word == "_" {
697 self.err_with_note(
698 "invalid argument name `_`",
699 "invalid argument name",
700 "argument name cannot be a single underscore",
701 self.to_span_index(start).to(self.to_span_index(end)),
702 );
703 }
704 word
705 }
706
707 /// Optionally parses an integer at the current position. This doesn't deal
708 /// with overflow at all, it's just accumulating digits.
709 fn integer(&mut self) -> Option<usize> {
710 let mut cur = 0;
711 let mut found = false;
712 while let Some(&(_, c)) = self.cur.peek() {
713 if let Some(i) = c.to_digit(10) {
714 cur = cur * 10 + i as usize;
715 found = true;
716 self.cur.next();
717 } else {
718 break;
719 }
720 }
721 found.then_some(cur)
722 }
723 }
724
725 /// Finds the indices of all characters that have been processed and differ between the actual
726 /// written code (code snippet) and the `InternedString` that gets processed in the `Parser`
727 /// in order to properly synthethise the intra-string `Span`s for error diagnostics.
728 fn find_skips_from_snippet(
729 snippet: Option<string::String>,
730 str_style: Option<usize>,
731 ) -> (Vec<usize>, bool) {
732 let snippet = match snippet {
733 Some(ref s) if s.starts_with('"') || s.starts_with("r#") => s,
734 _ => return (vec![], false),
735 };
736
737 fn find_skips(snippet: &str, is_raw: bool) -> Vec<usize> {
738 let mut eat_ws = false;
739 let mut s = snippet.chars().enumerate().peekable();
740 let mut skips = vec![];
741 while let Some((pos, c)) = s.next() {
742 match (c, s.peek()) {
743 // skip whitespace and empty lines ending in '\\'
744 ('\\', Some((next_pos, '\n'))) if !is_raw => {
745 eat_ws = true;
746 skips.push(pos);
747 skips.push(*next_pos);
748 let _ = s.next();
749 }
750 ('\\', Some((next_pos, '\n' | 'n' | 't'))) if eat_ws => {
751 skips.push(pos);
752 skips.push(*next_pos);
753 let _ = s.next();
754 }
755 (' ' | '\n' | '\t', _) if eat_ws => {
756 skips.push(pos);
757 }
758 ('\\', Some((next_pos, 'n' | 't' | 'r' | '0' | '\\' | '\'' | '\"'))) => {
759 skips.push(*next_pos);
760 let _ = s.next();
761 }
762 ('\\', Some((_, 'x'))) if !is_raw => {
763 for _ in 0..3 {
764 // consume `\xAB` literal
765 if let Some((pos, _)) = s.next() {
766 skips.push(pos);
767 } else {
768 break;
769 }
770 }
771 }
772 ('\\', Some((_, 'u'))) if !is_raw => {
773 if let Some((pos, _)) = s.next() {
774 skips.push(pos);
775 }
776 if let Some((next_pos, next_c)) = s.next() {
777 if next_c == '{' {
778 skips.push(next_pos);
779 let mut i = 0; // consume up to 6 hexanumeric chars + closing `}`
780 while let (Some((next_pos, c)), true) = (s.next(), i < 7) {
781 if c.is_digit(16) {
782 skips.push(next_pos);
783 } else if c == '}' {
784 skips.push(next_pos);
785 break;
786 } else {
787 break;
788 }
789 i += 1;
790 }
791 } else if next_c.is_digit(16) {
792 skips.push(next_pos);
793 // We suggest adding `{` and `}` when appropriate, accept it here as if
794 // it were correct
795 let mut i = 0; // consume up to 6 hexanumeric chars
796 while let (Some((next_pos, c)), _) = (s.next(), i < 6) {
797 if c.is_digit(16) {
798 skips.push(next_pos);
799 } else {
800 break;
801 }
802 i += 1;
803 }
804 }
805 }
806 }
807 _ if eat_ws => {
808 // `take_while(|c| c.is_whitespace())`
809 eat_ws = false;
810 }
811 _ => {}
812 }
813 }
814 skips
815 }
816
817 let r_start = str_style.map(|r| r + 1).unwrap_or(0);
818 let r_end = str_style.unwrap_or(0);
819 let s = &snippet[r_start + 1..snippet.len() - r_end - 1];
820 (find_skips(s, str_style.is_some()), true)
821 }
822
823 #[cfg(test)]
824 mod tests;