1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
#![doc = include_str!("../README.md")]
#![deny(missing_docs)]
// TODO: Enable when stable
//#![feature(once_cell)]

/// Combinators that allow combining and extending existing parsers.
pub mod combinator;
/// Error types, traits and utilities.
pub mod error;
/// Traits that allow chaining parser outputs together.
pub mod chain;
/// Parser primitives that accept specific token patterns.
pub mod primitive;
/// Types and traits that facilitate error recovery.
pub mod recovery;
/// Recursive parsers (parser that include themselves within their patterns).
pub mod recursive;
/// Types and traits related to spans.
pub mod span;
/// Token streams and behaviours.
pub mod stream;
/// Text-specific parsers and utilities.
pub mod text;

pub use crate::{
    error::Error,
    span::Span,
};

pub use crate::{
    stream::Stream,
};

use crate::{
    chain::Chain,
    combinator::*,
    primitive::*,
    recovery::*,
};

use std::{
    marker::PhantomData,
    rc::Rc,
    cmp::Ordering,
    ops::Range,
    // TODO: Enable when stable
    //lazy::OnceCell,
    fmt,
};

/// Commonly used functions, traits and types.
pub mod prelude {
    pub use super::{
        error::{Error as _, Simple},
        text::TextParser as _,
        span::Span as _,
        primitive::{any, end, filter, filter_map, just, one_of, none_of, seq},
        recovery::{skip_then_retry_until, nested_delimiters},
        recursive::recursive,
        text,
        Parser,
        BoxedParser,
    };
}

// TODO: Replace with `std::ops::ControlFlow` when stable
enum ControlFlow<C, B> {
    Continue(C),
    Break(B),
}

/// An internal type used to facilitate error prioritisation. You shouldn't need to interact with this type during
/// normal use of the crate.
pub struct Located<I, E> {
    at: usize,
    error: E,
    phantom: PhantomData<I>,
}

impl<I, E: Error<I>> Located<I, E> {
    /// Create a new [`Located`] with the give input position and error.
    pub fn at(at: usize, error: E) -> Self {
        Self { at, error, phantom: PhantomData }
    }

    /// Get the maximum of two located errors. If they hold the same position in the input, merge them.
    pub fn max(self, other: impl Into<Option<Self>>) -> Self {
        let other = match other.into() {
            Some(other) => other,
            None => return self,
        };
        match self.at.cmp(&other.at) {
            Ordering::Greater => self,
            Ordering::Less => other,
            Ordering::Equal => Self {
                error: self.error.merge(other.error),
                ..self
            },
        }
    }

    /// Map the error with the given function.
    pub fn map<U, F: FnOnce(E) -> U>(self, f: F) -> Located<I, U> {
        Located {
            at: self.at,
            error: f(self.error),
            phantom: PhantomData,
        }
    }
}

fn merge_alts<I, E: Error<I>>(a: Option<Located<I, E>>, b: Option<Located<I, E>>) -> Option<Located<I, E>> {
    match (a, b) {
        (Some(a), Some(b)) => Some(a.max(b)),
        (a, b) => a.or(b),
    }
}

// ([], Ok((out, recovered, alt_err))) => parsing successful, recovered = whether the false is a false recovered value,
// alt_err = potential alternative error should a different number of optional patterns be parsed
// ([x, ...], Ok(out)) => parsing failed, but recovery occurred so parsing may continue
// ([...], Err(err)) => parsing failed, recovery failed, and one or more errors were produced
type PResult<I, O, E> = (Vec<Located<I, E>>, Result<(O, Option<Located<I, E>>), Located<I, E>>);

type StreamOf<'a, I, E> = Stream<'a, I, <E as Error<I>>::Span>;

/// A trait implemented by parsers.
///
/// Parsers take a stream of tokens of type `I` and attempt to parse them into a value of type `O`. In doing so, they
/// may encounter errors. These need not be fatal to the parsing process: syntactic errors can be recovered from and a
/// valid output may still be generated alongside any syntax errors that were encountered along the way. Usually, this
/// output comes in the form of an [Abstract Syntax Tree](https://en.wikipedia.org/wiki/Abstract_syntax_tree) (AST).
///
/// Parsers currently only support LL(1) grammars. More concretely, this means that the rules that compose this parser
/// are only permitted to 'look' a single token into the future to determine the path through the grammar rules to be
/// taken by the parser. Unlike other techniques, such as recursive decent, arbitrary backtracking is not permitted.
/// The reasons for this are numerous, but perhaps the most obvious is that it makes error detection and recovery
/// significantly simpler and easier. In the future, this crate may be extended to support more complex grammars.
///
/// LL(1) parsers by themselves are not particularly powerful. Indeed, even very old languages such as C cannot parsed
/// by an LL(1) parser in a single pass. However, this limitation quickly vanishes (and, indeed, makes the design of
/// both the language and the parser easier) when one introduces multiple passes. For example, C compilers generally
/// have a lexical pass prior to the main parser that groups the input characters into tokens.
pub trait Parser<I: Clone, O> {
    /// The type of errors emitted by this parser.
    type Error: Error<I>; // TODO when default associated types are stable: = Cheap<I>;

    /// Parse a stream with all the bells & whistles. You can use this to implement your own parser combinators. Note
    /// that both the signature and semantic requirements of this function are very likely to change in later versions.
    /// Where possible, prefer more ergonomic combinators provided elsewhere in the crate rather than implementing your
    /// own.
    #[deprecated(note = "This method is excluded from the semver guarantees of chumsky. If you decide to use it, broken builds are your fault.")]
    fn parse_inner(&self, stream: &mut StreamOf<I, Self::Error>) -> PResult<I, O, Self::Error>;

    /// Parse an iterator of tokens, yielding an output if possible, and any errors encountered along the way.
    ///
    /// If you don't care about producing an output if errors are encountered, use `Parser::parse` instead.
    fn parse_recovery<
        'a,
        Iter: Iterator<Item = (I, <Self::Error as Error<I>>::Span)> + 'a,
        S: Into<Stream<'a, I, <Self::Error as Error<I>>::Span, Iter>>,
    >(&self, stream: S) -> (Option<O>, Vec<Self::Error>) where Self: Sized {
        #[allow(deprecated)]
        let (mut errors, res) = self.parse_inner(&mut stream.into());
        let out = match res {
            Ok((out, _)) => Some(out),
            Err(err) => {
                errors.push(err);
                None
            },
        };
        (out, errors.into_iter().map(|e| e.error).collect())
    }

    /// Parse an iterator of tokens, yielding an output *or* any errors that were encountered along the way.
    ///
    /// If you wish to attempt to produce an output even if errors are encountered, use `Parser::parse_recovery`.
    fn parse<
        'a,
        Iter: Iterator<Item = (I, <Self::Error as Error<I>>::Span)> + 'a,
        S: Into<Stream<'a, I, <Self::Error as Error<I>>::Span, Iter>>,
    >(&self, stream: S) -> Result<O, Vec<Self::Error>> where Self: Sized {
        let (output, errors) = self.parse_recovery(stream);
        if errors.is_empty() {
            Ok(output.expect("Parsing failed, but no errors were emitted. This is troubling, to say the least."))
        } else {
            Err(errors)
        }
    }

    /// Map the output of this parser to aanother value.
    ///
    /// # Examples
    ///
    /// ```
    /// # use chumsky::{prelude::*, error::Cheap};
    ///
    /// #[derive(Debug, PartialEq)]
    /// enum Token { Word(String), Num(u64) }
    ///
    /// let word = filter::<_, _, Cheap<char>>(|c: &char| c.is_alphabetic())
    ///     .repeated().at_least(1)
    ///     .collect::<String>()
    ///     .map(Token::Word);
    ///
    /// let num = filter::<_, _, Cheap<char>>(|c: &char| c.is_ascii_digit())
    ///     .repeated().at_least(1)
    ///     .collect::<String>()
    ///     .map(|s| Token::Num(s.parse().unwrap()));
    ///
    /// let token = word.or(num);
    ///
    /// assert_eq!(token.parse("test"), Ok(Token::Word("test".to_string())));
    /// assert_eq!(token.parse("42"), Ok(Token::Num(42)));
    /// ```
    fn map<U, F: Fn(O) -> U>(self, f: F) -> Map<Self, F, O> where Self: Sized { Map(self, f, PhantomData) }

    /// Map the output of this parser to another value, making use of the pattern's span.
    fn map_with_span<U, F: Fn(O, <Self::Error as Error<I>>::Span) -> U>(self, f: F) -> MapWithSpan<Self, F, O>
        where Self: Sized
        { MapWithSpan(self, f, PhantomData) }

    /// Map the primary error of this parser to another value.
    ///
    /// This does not map error emitted by sub-patterns within the parser.
    fn map_err<F: Fn(Self::Error) -> Self::Error>(self, f: F) -> MapErr<Self, F>
        where Self: Sized
    { MapErr(self, f) }

    /// Label the pattern parsed by this parser for more useful error messages.
    ///
    /// This is useful when you want to give users a more useful description of an expected pattern than simply a list
    /// of possible initial tokens. For example, it's common to use the term "expression" at a catch-all for a number
    /// of different constructs in many languages.
    ///
    /// This does not label recovered errors generated by sub-patterns within the parser, only error *directly* emitted
    /// by the parser.
    ///
    /// This does not label errors where the labelled pattern consumed input (i.e: in unambiguous cases).
    ///
    /// # Examples
    ///
    /// ```
    /// # use chumsky::{prelude::*, error::Cheap};
    ///
    /// let frac = text::digits(10)
    ///     .chain(just('.'))
    ///     .chain::<char, _, _>(text::digits(10))
    ///     .collect::<String>()
    ///     .then_ignore(end())
    ///     .labelled("number");
    ///
    /// assert_eq!(frac.parse("42.3"), Ok("42.3".to_string()));
    /// assert_eq!(frac.parse("hello"), Err(vec![Cheap::expected_input_found(0..1, None, Some('h')).with_label("number")]));
    /// assert_eq!(frac.parse("42!"), Err(vec![Cheap::expected_input_found(2..3, Some('.'), Some('!')).with_label("number")]));
    /// ```
    fn labelled<L: Into<<Self::Error as Error<I>>::Label> + Clone>(self, label: L) -> Label<Self, L>
        where Self: Sized
    { Label(self, label) }

    /// Transform all outputs of this parser to a pretermined value.
    ///
    /// # Examples
    ///
    /// ```
    /// # use chumsky::{prelude::*, error::Cheap};
    ///
    /// #[derive(Clone, Debug, PartialEq)]
    /// enum Op { Add, Sub, Mul, Div }
    ///
    /// let op = just::<_, Cheap<char>>('+').to(Op::Add)
    ///     .or(just('-').to(Op::Sub))
    ///     .or(just('*').to(Op::Mul))
    ///     .or(just('/').to(Op::Div));
    ///
    /// assert_eq!(op.parse("+"), Ok(Op::Add));
    /// assert_eq!(op.parse("/"), Ok(Op::Div));
    /// ```
    fn to<U: Clone>(self, x: U) -> To<Self, O, U> where Self: Sized { To(self, x, PhantomData) }

    /// Left-fold the output of the parser into a single value, where the output is of type `(_, Vec<_>)`.
    ///
    /// # Examples
    ///
    /// ```
    /// # use chumsky::{prelude::*, error::Cheap};
    ///
    /// let int = text::int::<char, Cheap<char>>(10)
    ///     .collect::<String>()
    ///     .map(|s| s.parse().unwrap());
    ///
    /// let sum = int
    ///     .then(just('+').ignore_then(int).repeated())
    ///     .foldl(|a, b| a + b);
    ///
    /// assert_eq!(sum.parse("1+12+3+9"), Ok(25));
    /// assert_eq!(sum.parse("6"), Ok(6));
    /// ```
    fn foldl<A, B, F: Fn(A, B) -> A>(self, f: F) -> Foldl<Self, F, A, B>
    where
        Self: Parser<I, (A, Vec<B>)> + Sized
    { Foldl(self, f, PhantomData) }

    /// Right-fold the output of the parser into a single value, where the output is of type `(Vec<_>, _)`.
    ///
    /// # Examples
    ///
    /// ```
    /// # use chumsky::{prelude::*, error::Cheap};
    ///
    /// let int = text::int::<char, Cheap<char>>(10)
    ///     .collect::<String>()
    ///     .map(|s| s.parse().unwrap());
    ///
    /// let signed = just('+').to(1)
    ///     .or(just('-').to(-1))
    ///     .repeated()
    ///     .then(int)
    ///     .foldr(|a, b| a * b);
    ///
    /// assert_eq!(signed.parse("3"), Ok(3));
    /// assert_eq!(signed.parse("-17"), Ok(-17));
    /// assert_eq!(signed.parse("--+-+-5"), Ok(5));
    /// ```
    fn foldr<'a, A, B, F: Fn(A, B) -> B + 'a>(self, f: F) -> Foldr<Self, F, A, B>
    where
        Self: Parser<I, (Vec<A>, B)> + Sized
    { Foldr(self, f, PhantomData) }

    /// Ignore the output of this parser, yielding `()` as an output instead.
    ///
    /// This can be used to reduce the cost of parsing by avoiding unnecessary allocations (most collections containing
    /// [ZSTs](https://doc.rust-lang.org/nomicon/exotic-sizes.html#zero-sized-types-zsts) do
    /// [not allocate](https://doc.rust-lang.org/std/vec/struct.Vec.html#guarantees)). For example, it's common to want
    /// to ignore whitespace in many grammars.
    ///
    /// # Examples
    ///
    /// ```
    /// # use chumsky::{prelude::*, error::Cheap};
    ///
    /// // A parser that parses any number of whitespace characters without allocating
    /// let whitespace = filter::<_, _, Cheap<char>>(|c: &char| c.is_whitespace())
    ///     .ignored()
    ///     .repeated();
    ///
    /// assert_eq!(whitespace.parse("    "), Ok(vec![(); 4]));
    /// assert_eq!(whitespace.parse("  hello"), Ok(vec![(); 2]));
    /// ```
    fn ignored(self) -> Ignored<Self, O> where Self: Sized { To(self, (), PhantomData) }

    /// Collect the output of this parser into a collection.
    ///
    /// This is commonly useful for collecting [`Vec<char>`] outputs into [`String`]s.
    ///
    /// # Examples
    ///
    /// ```
    /// # use chumsky::{prelude::*, error::Cheap};
    ///
    /// let word = filter::<_, _, Cheap<char>>(|c: &char| c.is_alphabetic())
    ///     .repeated()
    ///     .collect::<String>();
    ///
    /// assert_eq!(word.parse("hello"), Ok("hello".to_string()));
    /// ```
    fn collect<C: core::iter::FromIterator<O::Item>>(self) -> Map<Self, fn(O) -> C, O>
        where Self: Sized, O: IntoIterator
    { self.map(|items| C::from_iter(items.into_iter())) }

    /// Parse one thing and then another thing, yielding a tuple of the two outputs.
    ///
    /// # Examples
    ///
    /// ```
    /// # use chumsky::{prelude::*, error::Cheap};
    ///
    /// let word = filter::<_, _, Cheap<char>>(|c: &char| c.is_alphabetic())
    ///     .repeated().at_least(1)
    ///     .collect::<String>();
    /// let two_words = word.then_ignore(just(' ')).then(word);
    ///
    /// assert_eq!(two_words.parse("dog cat"), Ok(("dog".to_string(), "cat".to_string())));
    /// assert!(two_words.parse("hedgehog").is_err());
    /// ```
    fn then<U, P: Parser<I, U>>(self, other: P) -> Then<Self, P> where Self: Sized { Then(self, other) }

    /// Parse one thing and then another thing, attempting to chain the two outputs into a [`Vec`].
    ///
    /// # Examples
    ///
    /// ```
    /// # use chumsky::{prelude::*, error::Cheap};
    ///
    /// let int = just('-').or_not()
    ///     .chain(filter::<_, _, Cheap<char>>(|c: &char| c.is_ascii_digit() && *c != '0')
    ///         .chain(filter::<_, _, Cheap<char>>(|c: &char| c.is_ascii_digit()).repeated()))
    ///     .or(just('0').map(|c| vec![c]))
    ///     .then_ignore(end())
    ///     .collect::<String>()
    ///     .map(|s| s.parse().unwrap());
    ///
    /// assert_eq!(int.parse("0"), Ok(0));
    /// assert_eq!(int.parse("415"), Ok(415));
    /// assert_eq!(int.parse("-50"), Ok(-50));
    /// assert!(int.parse("-0").is_err());
    /// assert!(int.parse("05").is_err());
    /// ```
    fn chain<T, U, P: Parser<I, U, Error = Self::Error>>(self, other: P) -> Map<Then<Self, P>, fn((O, U)) -> Vec<T>, (O, U)>
    where
        Self: Sized,
        U: Chain<T>,
        O: Chain<T>,
    {
        self.then(other).map(|(a, b)| {
            let mut v = Vec::with_capacity(a.len() + b.len());
            a.append_to(&mut v);
            b.append_to(&mut v);
            v
        })
    }

    /// Flatten a nested collection.
    fn flatten<T, Inner>(self) -> Map<Self, fn(O) -> Vec<T>, O>
    where
        Self: Sized,
        O: IntoIterator<Item = Inner>,
        Inner: IntoIterator<Item = T>,
    { self.map(|xs| xs.into_iter().map(|xs| xs.into_iter()).flatten().collect()) }

    /// Parse one thing and then another thing, yielding only the output of the latter.
    ///
    /// # Examples
    ///
    /// ```
    /// # use chumsky::{prelude::*, error::Cheap};
    ///
    /// let zeroes = filter::<_, _, Cheap<char>>(|c: &char| *c == '0').ignored().repeated();
    /// let digits = filter(|c: &char| c.is_ascii_digit()).repeated();
    /// let integer = zeroes
    ///     .ignore_then(digits)
    ///     .collect::<String>()
    ///     .map(|s| s.parse().unwrap());
    ///
    /// assert_eq!(integer.parse("00064"), Ok(64));
    /// assert_eq!(integer.parse("32"), Ok(32));
    /// ```
    fn ignore_then<U, P: Parser<I, U>>(self, other: P) -> IgnoreThen<Self, P, O, U>
        where Self: Sized
    { Map(Then(self, other), |(_, u)| u, PhantomData) }

    /// Parse one thing and then another thing, yielding only the output of the former.
    ///
    /// # Examples
    ///
    /// ```
    /// # use chumsky::{prelude::*, error::Cheap};
    ///
    /// let word = filter::<_, _, Cheap<char>>(|c: &char| c.is_alphabetic())
    ///     .repeated().at_least(1)
    ///     .collect::<String>();
    ///
    /// let punctuated = word
    ///     .then_ignore(just('!').or(just('?')).or_not());
    ///
    /// let sentence = punctuated
    ///     .padded() // Allow for whitespace gaps
    ///     .repeated();
    ///
    /// assert_eq!(
    ///     sentence.parse("hello! how are you?"),
    ///     Ok(vec![
    ///         "hello".to_string(),
    ///         "how".to_string(),
    ///         "are".to_string(),
    ///         "you".to_string(),
    ///     ]),
    /// );
    /// ```
    fn then_ignore<U, P: Parser<I, U>>(self, other: P) -> ThenIgnore<Self, P, O, U>
        where Self: Sized
    { Map(Then(self, other), |(o, _)| o, PhantomData) }

    // fn then_catch(self, end: I) -> ThenCatch<Self, I> where Self: Sized { ThenCatch(self, end) }

    /// Parse the pattern surrounded by the given delimiters.
    ///
    /// # Examples
    ///
    /// ```
    /// # use chumsky::{prelude::*, error::Cheap};
    ///
    /// // A LISP-style S-expression
    /// #[derive(Debug, PartialEq)]
    /// enum SExpr {
    ///     Ident(String),
    ///     Num(u64),
    ///     List(Vec<SExpr>),
    /// }
    ///
    /// let ident = filter::<_, _, Cheap<char>>(|c: &char| c.is_alphabetic())
    ///     .repeated().at_least(1)
    ///     .collect::<String>();
    ///
    /// let num = text::int(10)
    ///     .collect::<String>()
    ///     .map(|s| s.parse().unwrap());
    ///
    /// let s_expr = recursive(|s_expr| s_expr
    ///     .padded()
    ///     .repeated()
    ///     .map(SExpr::List)
    ///     .delimited_by('(', ')')
    ///     .or(ident.map(SExpr::Ident))
    ///     .or(num.map(SExpr::Num)));
    ///
    /// // A valid input
    /// assert_eq!(
    ///     s_expr.parse_recovery("(add (mul 42 3) 15)"),
    ///     (
    ///         Some(SExpr::List(vec![
    ///             SExpr::Ident("add".to_string()),
    ///             SExpr::List(vec![
    ///                 SExpr::Ident("mul".to_string()),
    ///                 SExpr::Num(42),
    ///                 SExpr::Num(3),
    ///             ]),
    ///             SExpr::Num(15),
    ///         ])),
    ///         Vec::new(), // No errors!
    ///     ),
    /// );
    /// ```
    fn delimited_by(self, start: I, end: I) -> DelimitedBy<Self, I> where Self: Sized { DelimitedBy(self, start, end) }

    /// Parse one thing or, on failure, another thing.
    ///
    /// # Examples
    ///
    /// ```
    /// # use chumsky::{prelude::*, error::Cheap};
    ///
    /// let op = just::<_, Cheap<char>>('+')
    ///     .or(just('-'))
    ///     .or(just('*'))
    ///     .or(just('/'));
    ///
    /// assert_eq!(op.parse("+"), Ok('+'));
    /// assert_eq!(op.parse("/"), Ok('/'));
    /// assert!(op.parse("!").is_err());
    /// ```
    fn or<P: Parser<I, O>>(self, other: P) -> Or<Self, P> where Self: Sized { Or(self, other) }

    /// Apply a fallback recovery strategy to this parser should it fail.
    ///
    /// There is no silver bullet for error recovery, so this function allows you to specify one of several different
    /// strategies at the location of your choice.
    ///
    /// Note that for implementation reasons, adding an error recovery strategy can cause a parser to 'over-commit',
    /// missing potentially valid alternative parse routes (TODO: document this and explain why and how it happens).
    /// Rest assured that this case is generally quite rare and only happens for very loose, almost-ambiguous syntax.
    /// If you run into cases that you believe should parse but do not, try removing or moving recovery strategies to
    /// fix the problem.
    fn recover_with<S: Strategy<I, O>>(self, strategy: S) -> Recovery<Self, S> where Self: Sized {
        Recovery(self, strategy)
    }

    /// Attempt to parse something, but only if it exists.
    ///
    /// If parsing of the pattern is successful, the output is `Some(_)`. Otherwise, the output is `None`.
    ///
    /// # Examples
    ///
    /// ```
    /// # use chumsky::{prelude::*, error::Cheap};
    ///
    /// let word = filter::<_, _, Cheap<char>>(|c: &char| c.is_alphabetic())
    ///     .repeated().at_least(1)
    ///     .collect::<String>();
    ///
    /// let word_or_question = word
    ///     .then(just('?').or_not());
    ///
    /// assert_eq!(word_or_question.parse("hello?"), Ok(("hello".to_string(), Some('?'))));
    /// assert_eq!(word_or_question.parse("wednesday"), Ok(("wednesday".to_string(), None)));
    /// ```
    fn or_not(self) -> OrNot<Self> where Self: Sized { OrNot(self) }

    /// Parse an expression any number of times (including zero times).
    ///
    /// Input is eagerly parsed. Be aware that the parser will accept no occurences of the pattern too. Consider using
    /// [`Parser::repeated().at_least`] instead if it better suits your use-case.
    ///
    /// # Examples
    ///
    /// ```
    /// # use chumsky::{prelude::*, error::Cheap};
    ///
    /// let num = filter::<_, _, Cheap<char>>(|c: &char| c.is_ascii_digit())
    ///     .repeated().at_least(1)
    ///     .collect::<String>()
    ///     .map(|s| s.parse().unwrap());
    ///
    /// let sum = num.then(just('+').ignore_then(num).repeated())
    ///     .foldl(|a, b| a + b);
    ///
    /// assert_eq!(sum.parse("2+13+4+0+5"), Ok(24));
    /// ```
    fn repeated(self) -> Repeated<Self> where Self: Sized { Repeated(self, 0, None) }

    /// Parse an expression, separated by another, any number of times, optionally with a trailing instance of the
    /// other.
    fn separated_by<U, P: Parser<I, U>>(self, other: P) -> SeparatedBy<Self, P, U> where Self: Sized {
        SeparatedBy {
            a: self,
            b: other,
            at_least: 0,
            allow_leading: false,
            allow_trailing: false,
            phantom: PhantomData,
        }
    }

    /// Box the parser, yielding a parser that performs parsing through dynamic dispatch.
    ///
    /// Boxing a parser might be useful for:
    ///
    /// - Passing a parser over an FFI boundary
    ///
    /// - Getting around compiler implementation problems with long types such as
    ///   [this](https://github.com/rust-lang/rust/issues/54540).
    ///
    /// - Places where you need to name the type of a parser
    ///
    /// Boxing a parser is loosely equivalent to boxing other combinators, such as [`Iterator`].
    fn boxed<'a>(self) -> BoxedParser<'a, I, O, Self::Error> where Self: Sized + 'a {
        BoxedParser(Rc::new(self))
    }
}

impl<'a, I: Clone, O, T: Parser<I, O>> Parser<I, O> for &'a T {
    type Error = T::Error;

    fn parse_inner(&self, stream: &mut StreamOf<I, Self::Error>) -> PResult<I, O, Self::Error> {
        #[allow(deprecated)]
        T::parse_inner(*self, stream)
    }
}

/// See [`Parser::boxed`].
///
/// This type is a [`repr(transparent)`](https://doc.rust-lang.org/nomicon/other-reprs.html#reprtransparent) wrapper
/// around its inner value.
///
/// Due to current implementation details, the inner value is not, in fact, a [`Box`], but is an [`Rc`] to facilitate
/// efficient cloning. This is likely to change in the future. Unlike [`Box`], [`Rc`] has no size guarantees: although
/// it is *currently* the same size as a raw pointer.
// TODO: Don't use an Rc
#[repr(transparent)]
pub struct BoxedParser<'a, I, O, E: Error<I>>(Rc<dyn Parser<I, O, Error = E> + 'a>);

impl<'a, I, O, E: Error<I>> Clone for BoxedParser<'a, I, O, E> {
    fn clone(&self) -> Self { Self(self.0.clone()) }
}

impl<'a, I: Clone, O, E: Error<I>> Parser<I, O> for BoxedParser<'a, I, O, E> {
    type Error = E;

    fn parse_inner(&self, stream: &mut StreamOf<I, Self::Error>) -> PResult<I, O, Self::Error> {
        #[allow(deprecated)]
        self.0.parse_inner(stream)
    }
}