regex_lite/hir/
mod.rs

1use alloc::{boxed::Box, string::String, vec, vec::Vec};
2
3use crate::{error::Error, utf8};
4
5mod parse;
6
7/// Escapes all regular expression meta characters in `pattern`.
8///
9/// The string returned may be safely used as a literal in a regular
10/// expression.
11pub fn escape(pattern: &str) -> String {
12    let mut buf = String::new();
13    buf.reserve(pattern.len());
14    for ch in pattern.chars() {
15        if is_meta_character(ch) {
16            buf.push('\\');
17        }
18        buf.push(ch);
19    }
20    buf
21}
22
23/// Returns true if the given character has significance in a regex.
24///
25/// Generally speaking, these are the only characters which _must_ be escaped
26/// in order to match their literal meaning. For example, to match a literal
27/// `|`, one could write `\|`. Sometimes escaping isn't always necessary. For
28/// example, `-` is treated as a meta character because of its significance
29/// for writing ranges inside of character classes, but the regex `-` will
30/// match a literal `-` because `-` has no special meaning outside of character
31/// classes.
32///
33/// In order to determine whether a character may be escaped at all, the
34/// [`is_escapeable_character`] routine should be used. The difference between
35/// `is_meta_character` and `is_escapeable_character` is that the latter will
36/// return true for some characters that are _not_ meta characters. For
37/// example, `%` and `\%` both match a literal `%` in all contexts. In other
38/// words, `is_escapeable_character` includes "superfluous" escapes.
39///
40/// Note that the set of characters for which this function returns `true` or
41/// `false` is fixed and won't change in a semver compatible release. (In this
42/// case, "semver compatible release" actually refers to the `regex` crate
43/// itself, since reducing or expanding the set of meta characters would be a
44/// breaking change for not just `regex-syntax` but also `regex` itself.)
45fn is_meta_character(c: char) -> bool {
46    match c {
47        '\\' | '.' | '+' | '*' | '?' | '(' | ')' | '|' | '[' | ']' | '{'
48        | '}' | '^' | '$' | '#' | '&' | '-' | '~' => true,
49        _ => false,
50    }
51}
52
53/// Returns true if the given character can be escaped in a regex.
54///
55/// This returns true in all cases that `is_meta_character` returns true, but
56/// also returns true in some cases where `is_meta_character` returns false.
57/// For example, `%` is not a meta character, but it is escapeable. That is,
58/// `%` and `\%` both match a literal `%` in all contexts.
59///
60/// The purpose of this routine is to provide knowledge about what characters
61/// may be escaped. Namely, most regex engines permit "superfluous" escapes
62/// where characters without any special significance may be escaped even
63/// though there is no actual _need_ to do so.
64///
65/// This will return false for some characters. For example, `e` is not
66/// escapeable. Therefore, `\e` will either result in a parse error (which is
67/// true today), or it could backwards compatibly evolve into a new construct
68/// with its own meaning. Indeed, that is the purpose of banning _some_
69/// superfluous escapes: it provides a way to evolve the syntax in a compatible
70/// manner.
71fn is_escapeable_character(c: char) -> bool {
72    // Certainly escapeable if it's a meta character.
73    if is_meta_character(c) {
74        return true;
75    }
76    // Any character that isn't ASCII is definitely not escapeable. There's
77    // no real need to allow things like \☃ right?
78    if !c.is_ascii() {
79        return false;
80    }
81    // Otherwise, we basically say that everything is escapeable unless it's a
82    // letter or digit. Things like \3 are either octal (when enabled) or an
83    // error, and we should keep it that way. Otherwise, letters are reserved
84    // for adding new syntax in a backwards compatible way.
85    match c {
86        '0'..='9' | 'A'..='Z' | 'a'..='z' => false,
87        // While not currently supported, we keep these as not escapeable to
88        // give us some flexibility with respect to supporting the \< and
89        // \> word boundary assertions in the future. By rejecting them as
90        // escapeable, \< and \> will result in a parse error. Thus, we can
91        // turn them into something else in the future without it being a
92        // backwards incompatible change.
93        '<' | '>' => false,
94        _ => true,
95    }
96}
97
98/// The configuration for a regex parser.
99#[derive(Clone, Copy, Debug)]
100pub(crate) struct Config {
101    /// The maximum number of times we're allowed to recurse.
102    ///
103    /// Note that unlike the regex-syntax parser, we actually use recursion in
104    /// this parser for simplicity. My hope is that by setting a conservative
105    /// default call limit and providing a way to configure it, that we can
106    /// keep this simplification. But if we must, we can re-work the parser to
107    /// put the call stack on the heap like regex-syntax does.
108    pub(crate) nest_limit: u32,
109    /// Various flags that control how a pattern is interpreted.
110    pub(crate) flags: Flags,
111}
112
113impl Default for Config {
114    fn default() -> Config {
115        Config { nest_limit: 50, flags: Flags::default() }
116    }
117}
118
119/// Various flags that control the interpretation of the pattern.
120///
121/// These can be set via explicit configuration in code, or change dynamically
122/// during parsing via inline flags. For example, `foo(?i:bar)baz` will match
123/// `foo` and `baz` case sensitiviely and `bar` case insensitively (assuming a
124/// default configuration).
125#[derive(Clone, Copy, Debug, Default)]
126pub(crate) struct Flags {
127    /// Whether to match case insensitively.
128    ///
129    /// This is the `i` flag.
130    pub(crate) case_insensitive: bool,
131    /// Whether `^` and `$` should be treated as line anchors or not.
132    ///
133    /// This is the `m` flag.
134    pub(crate) multi_line: bool,
135    /// Whether `.` should match line terminators or not.
136    ///
137    /// This is the `s` flag.
138    pub(crate) dot_matches_new_line: bool,
139    /// Whether to swap the meaning of greedy and non-greedy operators.
140    ///
141    /// This is the `U` flag.
142    pub(crate) swap_greed: bool,
143    /// Whether to enable CRLF mode.
144    ///
145    /// This is the `R` flag.
146    pub(crate) crlf: bool,
147    /// Whether to ignore whitespace. i.e., verbose mode.
148    ///
149    /// This is the `x` flag.
150    pub(crate) ignore_whitespace: bool,
151}
152
153#[derive(Clone, Debug, Eq, PartialEq)]
154pub(crate) struct Hir {
155    kind: HirKind,
156    is_start_anchored: bool,
157    is_match_empty: bool,
158    static_explicit_captures_len: Option<usize>,
159}
160
161#[derive(Clone, Debug, Eq, PartialEq)]
162pub(crate) enum HirKind {
163    Empty,
164    Char(char),
165    Class(Class),
166    Look(Look),
167    Repetition(Repetition),
168    Capture(Capture),
169    Concat(Vec<Hir>),
170    Alternation(Vec<Hir>),
171}
172
173impl Hir {
174    /// Parses the given pattern string with the given configuration into a
175    /// structured representation. If the pattern is invalid, then an error
176    /// is returned.
177    pub(crate) fn parse(config: Config, pattern: &str) -> Result<Hir, Error> {
178        self::parse::Parser::new(config, pattern).parse()
179    }
180
181    /// Returns the underlying kind of this high-level intermediate
182    /// representation.
183    ///
184    /// Note that there is explicitly no way to build an `Hir` directly from
185    /// an `HirKind`. If you need to do that, then you must do case analysis
186    /// on the `HirKind` and call the appropriate smart constructor on `Hir`.
187    pub(crate) fn kind(&self) -> &HirKind {
188        &self.kind
189    }
190
191    /// Returns true if and only if this Hir expression can only match at the
192    /// beginning of a haystack.
193    pub(crate) fn is_start_anchored(&self) -> bool {
194        self.is_start_anchored
195    }
196
197    /// Returns true if and only if this Hir expression can match the empty
198    /// string.
199    pub(crate) fn is_match_empty(&self) -> bool {
200        self.is_match_empty
201    }
202
203    /// If the pattern always reports the same number of matching capture groups
204    /// for every match, then this returns the number of those groups. This
205    /// doesn't include the implicit group found in every pattern.
206    pub(crate) fn static_explicit_captures_len(&self) -> Option<usize> {
207        self.static_explicit_captures_len
208    }
209
210    fn fail() -> Hir {
211        let kind = HirKind::Class(Class { ranges: vec![] });
212        Hir {
213            kind,
214            is_start_anchored: false,
215            is_match_empty: false,
216            static_explicit_captures_len: Some(0),
217        }
218    }
219
220    fn empty() -> Hir {
221        let kind = HirKind::Empty;
222        Hir {
223            kind,
224            is_start_anchored: false,
225            is_match_empty: true,
226            static_explicit_captures_len: Some(0),
227        }
228    }
229
230    fn char(ch: char) -> Hir {
231        let kind = HirKind::Char(ch);
232        Hir {
233            kind,
234            is_start_anchored: false,
235            is_match_empty: false,
236            static_explicit_captures_len: Some(0),
237        }
238    }
239
240    fn class(class: Class) -> Hir {
241        let kind = HirKind::Class(class);
242        Hir {
243            kind,
244            is_start_anchored: false,
245            is_match_empty: false,
246            static_explicit_captures_len: Some(0),
247        }
248    }
249
250    fn look(look: Look) -> Hir {
251        let kind = HirKind::Look(look);
252        Hir {
253            kind,
254            is_start_anchored: matches!(look, Look::Start),
255            is_match_empty: true,
256            static_explicit_captures_len: Some(0),
257        }
258    }
259
260    fn repetition(rep: Repetition) -> Hir {
261        if rep.min == 0 && rep.max == Some(0) {
262            return Hir::empty();
263        } else if rep.min == 1 && rep.max == Some(1) {
264            return *rep.sub;
265        }
266        let is_start_anchored = rep.min > 0 && rep.sub.is_start_anchored;
267        let is_match_empty = rep.min == 0 || rep.sub.is_match_empty;
268        let mut static_explicit_captures_len =
269            rep.sub.static_explicit_captures_len;
270        // If the static captures len of the sub-expression is not known or
271        // is greater than zero, then it automatically propagates to the
272        // repetition, regardless of the repetition. Otherwise, it might
273        // change, but only when the repetition can match 0 times.
274        if rep.min == 0
275            && static_explicit_captures_len.map_or(false, |len| len > 0)
276        {
277            // If we require a match 0 times, then our captures len is
278            // guaranteed to be zero. Otherwise, if we *can* match the empty
279            // string, then it's impossible to know how many captures will be
280            // in the resulting match.
281            if rep.max == Some(0) {
282                static_explicit_captures_len = Some(0);
283            } else {
284                static_explicit_captures_len = None;
285            }
286        }
287        Hir {
288            kind: HirKind::Repetition(rep),
289            is_start_anchored,
290            is_match_empty,
291            static_explicit_captures_len,
292        }
293    }
294
295    fn capture(cap: Capture) -> Hir {
296        let is_start_anchored = cap.sub.is_start_anchored;
297        let is_match_empty = cap.sub.is_match_empty;
298        let static_explicit_captures_len = cap
299            .sub
300            .static_explicit_captures_len
301            .map(|len| len.saturating_add(1));
302        let kind = HirKind::Capture(cap);
303        Hir {
304            kind,
305            is_start_anchored,
306            is_match_empty,
307            static_explicit_captures_len,
308        }
309    }
310
311    fn concat(mut subs: Vec<Hir>) -> Hir {
312        if subs.is_empty() {
313            Hir::empty()
314        } else if subs.len() == 1 {
315            subs.pop().unwrap()
316        } else {
317            let is_start_anchored = subs[0].is_start_anchored;
318            let mut is_match_empty = true;
319            let mut static_explicit_captures_len = Some(0usize);
320            for sub in subs.iter() {
321                is_match_empty = is_match_empty && sub.is_match_empty;
322                static_explicit_captures_len = static_explicit_captures_len
323                    .and_then(|len1| {
324                        Some((len1, sub.static_explicit_captures_len?))
325                    })
326                    .and_then(|(len1, len2)| Some(len1.saturating_add(len2)));
327            }
328            Hir {
329                kind: HirKind::Concat(subs),
330                is_start_anchored,
331                is_match_empty,
332                static_explicit_captures_len,
333            }
334        }
335    }
336
337    fn alternation(mut subs: Vec<Hir>) -> Hir {
338        if subs.is_empty() {
339            Hir::fail()
340        } else if subs.len() == 1 {
341            subs.pop().unwrap()
342        } else {
343            let mut it = subs.iter().peekable();
344            let mut is_start_anchored =
345                it.peek().map_or(false, |sub| sub.is_start_anchored);
346            let mut is_match_empty =
347                it.peek().map_or(false, |sub| sub.is_match_empty);
348            let mut static_explicit_captures_len =
349                it.peek().and_then(|sub| sub.static_explicit_captures_len);
350            for sub in it {
351                is_start_anchored = is_start_anchored && sub.is_start_anchored;
352                is_match_empty = is_match_empty || sub.is_match_empty;
353                if static_explicit_captures_len
354                    != sub.static_explicit_captures_len
355                {
356                    static_explicit_captures_len = None;
357                }
358            }
359            Hir {
360                kind: HirKind::Alternation(subs),
361                is_start_anchored,
362                is_match_empty,
363                static_explicit_captures_len,
364            }
365        }
366    }
367}
368
369impl HirKind {
370    /// Returns a slice of this kind's sub-expressions, if any.
371    fn subs(&self) -> &[Hir] {
372        use core::slice::from_ref;
373
374        match *self {
375            HirKind::Empty
376            | HirKind::Char(_)
377            | HirKind::Class(_)
378            | HirKind::Look(_) => &[],
379            HirKind::Repetition(Repetition { ref sub, .. }) => from_ref(sub),
380            HirKind::Capture(Capture { ref sub, .. }) => from_ref(sub),
381            HirKind::Concat(ref subs) => subs,
382            HirKind::Alternation(ref subs) => subs,
383        }
384    }
385}
386
387#[derive(Clone, Debug, Eq, PartialEq)]
388pub(crate) struct Class {
389    pub(crate) ranges: Vec<ClassRange>,
390}
391
392impl Class {
393    /// Create a new class from the given ranges. The ranges may be provided
394    /// in any order or may even overlap. They will be automatically
395    /// canonicalized.
396    fn new<I: IntoIterator<Item = ClassRange>>(ranges: I) -> Class {
397        let mut class = Class { ranges: ranges.into_iter().collect() };
398        class.canonicalize();
399        class
400    }
401
402    /// Expand this class such that it matches the ASCII codepoints in this set
403    /// case insensitively.
404    fn ascii_case_fold(&mut self) {
405        let len = self.ranges.len();
406        for i in 0..len {
407            if let Some(folded) = self.ranges[i].ascii_case_fold() {
408                self.ranges.push(folded);
409            }
410        }
411        self.canonicalize();
412    }
413
414    /// Negate this set.
415    ///
416    /// For all `x` where `x` is any element, if `x` was in this set, then it
417    /// will not be in this set after negation.
418    fn negate(&mut self) {
419        const MIN: char = '\x00';
420        const MAX: char = char::MAX;
421
422        if self.ranges.is_empty() {
423            self.ranges.push(ClassRange { start: MIN, end: MAX });
424            return;
425        }
426
427        // There should be a way to do this in-place with constant memory,
428        // but I couldn't figure out a simple way to do it. So just append
429        // the negation to the end of this range, and then drain it before
430        // we're done.
431        let drain_end = self.ranges.len();
432
433        // If our class doesn't start the minimum possible char, then negation
434        // needs to include all codepoints up to the minimum in this set.
435        if self.ranges[0].start > MIN {
436            self.ranges.push(ClassRange {
437                start: MIN,
438                // OK because we know it's bigger than MIN.
439                end: prev_char(self.ranges[0].start).unwrap(),
440            });
441        }
442        for i in 1..drain_end {
443            // let lower = self.ranges[i - 1].upper().increment();
444            // let upper = self.ranges[i].lower().decrement();
445            // self.ranges.push(I::create(lower, upper));
446            self.ranges.push(ClassRange {
447                // OK because we know i-1 is never the last range and therefore
448                // there must be a range greater than it. It therefore follows
449                // that 'end' can never be char::MAX, and thus there must be
450                // a next char.
451                start: next_char(self.ranges[i - 1].end).unwrap(),
452                // Since 'i' is guaranteed to never be the first range, it
453                // follows that there is always a range before this and thus
454                // 'start' can never be '\x00'. Thus, there must be a previous
455                // char.
456                end: prev_char(self.ranges[i].start).unwrap(),
457            });
458        }
459        if self.ranges[drain_end - 1].end < MAX {
460            // let lower = self.ranges[drain_end - 1].upper().increment();
461            // self.ranges.push(I::create(lower, I::Bound::max_value()));
462            self.ranges.push(ClassRange {
463                // OK because we know 'end' is less than char::MAX, and thus
464                // there is a next char.
465                start: next_char(self.ranges[drain_end - 1].end).unwrap(),
466                end: MAX,
467            });
468        }
469        self.ranges.drain(..drain_end);
470        // We don't need to canonicalize because we processed the ranges above
471        // in canonical order and the new ranges we added based on those are
472        // also necessarily in canonical order.
473    }
474
475    /// Converts this set into a canonical ordering.
476    fn canonicalize(&mut self) {
477        if self.is_canonical() {
478            return;
479        }
480        self.ranges.sort();
481        assert!(!self.ranges.is_empty());
482
483        // Is there a way to do this in-place with constant memory? I couldn't
484        // figure out a way to do it. So just append the canonicalization to
485        // the end of this range, and then drain it before we're done.
486        let drain_end = self.ranges.len();
487        for oldi in 0..drain_end {
488            // If we've added at least one new range, then check if we can
489            // merge this range in the previously added range.
490            if self.ranges.len() > drain_end {
491                let (last, rest) = self.ranges.split_last_mut().unwrap();
492                if let Some(union) = last.union(&rest[oldi]) {
493                    *last = union;
494                    continue;
495                }
496            }
497            self.ranges.push(self.ranges[oldi]);
498        }
499        self.ranges.drain(..drain_end);
500    }
501
502    /// Returns true if and only if this class is in a canonical ordering.
503    fn is_canonical(&self) -> bool {
504        for pair in self.ranges.windows(2) {
505            if pair[0] >= pair[1] {
506                return false;
507            }
508            if pair[0].is_contiguous(&pair[1]) {
509                return false;
510            }
511        }
512        true
513    }
514}
515
516#[derive(Clone, Copy, Debug, Eq, PartialEq, PartialOrd, Ord)]
517pub(crate) struct ClassRange {
518    pub(crate) start: char,
519    pub(crate) end: char,
520}
521
522impl ClassRange {
523    /// Apply simple case folding to this byte range. Only ASCII case mappings
524    /// (for A-Za-z) are applied.
525    ///
526    /// Additional ranges are appended to the given vector. Canonical ordering
527    /// is *not* maintained in the given vector.
528    fn ascii_case_fold(&self) -> Option<ClassRange> {
529        if !(ClassRange { start: 'a', end: 'z' }).is_intersection_empty(self) {
530            let start = core::cmp::max(self.start, 'a');
531            let end = core::cmp::min(self.end, 'z');
532            return Some(ClassRange {
533                start: char::try_from(u32::from(start) - 32).unwrap(),
534                end: char::try_from(u32::from(end) - 32).unwrap(),
535            });
536        }
537        if !(ClassRange { start: 'A', end: 'Z' }).is_intersection_empty(self) {
538            let start = core::cmp::max(self.start, 'A');
539            let end = core::cmp::min(self.end, 'Z');
540            return Some(ClassRange {
541                start: char::try_from(u32::from(start) + 32).unwrap(),
542                end: char::try_from(u32::from(end) + 32).unwrap(),
543            });
544        }
545        None
546    }
547
548    /// Union the given overlapping range into this range.
549    ///
550    /// If the two ranges aren't contiguous, then this returns `None`.
551    fn union(&self, other: &ClassRange) -> Option<ClassRange> {
552        if !self.is_contiguous(other) {
553            return None;
554        }
555        let start = core::cmp::min(self.start, other.start);
556        let end = core::cmp::max(self.end, other.end);
557        Some(ClassRange { start, end })
558    }
559
560    /// Returns true if and only if the two ranges are contiguous. Two ranges
561    /// are contiguous if and only if the ranges are either overlapping or
562    /// adjacent.
563    fn is_contiguous(&self, other: &ClassRange) -> bool {
564        let (s1, e1) = (u32::from(self.start), u32::from(self.end));
565        let (s2, e2) = (u32::from(other.start), u32::from(other.end));
566        core::cmp::max(s1, s2) <= core::cmp::min(e1, e2).saturating_add(1)
567    }
568
569    /// Returns true if and only if the intersection of this range and the
570    /// other range is empty.
571    fn is_intersection_empty(&self, other: &ClassRange) -> bool {
572        let (s1, e1) = (self.start, self.end);
573        let (s2, e2) = (other.start, other.end);
574        core::cmp::max(s1, s2) > core::cmp::min(e1, e2)
575    }
576}
577
578/// The high-level intermediate representation for a look-around assertion.
579///
580/// An assertion match is always zero-length. Also called an "empty match."
581#[derive(Clone, Copy, Debug, Eq, PartialEq)]
582pub(crate) enum Look {
583    /// Match the beginning of text. Specifically, this matches at the starting
584    /// position of the input.
585    Start = 1 << 0,
586    /// Match the end of text. Specifically, this matches at the ending
587    /// position of the input.
588    End = 1 << 1,
589    /// Match the beginning of a line or the beginning of text. Specifically,
590    /// this matches at the starting position of the input, or at the position
591    /// immediately following a `\n` character.
592    StartLF = 1 << 2,
593    /// Match the end of a line or the end of text. Specifically, this matches
594    /// at the end position of the input, or at the position immediately
595    /// preceding a `\n` character.
596    EndLF = 1 << 3,
597    /// Match the beginning of a line or the beginning of text. Specifically,
598    /// this matches at the starting position of the input, or at the position
599    /// immediately following either a `\r` or `\n` character, but never after
600    /// a `\r` when a `\n` follows.
601    StartCRLF = 1 << 4,
602    /// Match the end of a line or the end of text. Specifically, this matches
603    /// at the end position of the input, or at the position immediately
604    /// preceding a `\r` or `\n` character, but never before a `\n` when a `\r`
605    /// precedes it.
606    EndCRLF = 1 << 5,
607    /// Match an ASCII-only word boundary. That is, this matches a position
608    /// where the left adjacent character and right adjacent character
609    /// correspond to a word and non-word or a non-word and word character.
610    Word = 1 << 6,
611    /// Match an ASCII-only negation of a word boundary.
612    WordNegate = 1 << 7,
613    /// Match the start of an ASCII-only word boundary. That is, this matches a
614    /// position at either the beginning of the haystack or where the previous
615    /// character is not a word character and the following character is a word
616    /// character.
617    WordStart = 1 << 8,
618    /// Match the end of an ASCII-only word boundary. That is, this matches
619    /// a position at either the end of the haystack or where the previous
620    /// character is a word character and the following character is not a word
621    /// character.
622    WordEnd = 1 << 9,
623    /// Match the start half of an ASCII-only word boundary. That is, this
624    /// matches a position at either the beginning of the haystack or where the
625    /// previous character is not a word character.
626    WordStartHalf = 1 << 10,
627    /// Match the end half of an ASCII-only word boundary. That is, this
628    /// matches a position at either the end of the haystack or where the
629    /// following character is not a word character.
630    WordEndHalf = 1 << 11,
631}
632
633impl Look {
634    /// Returns true if the given position in the given haystack matches this
635    /// look-around assertion.
636    pub(crate) fn is_match(&self, haystack: &[u8], at: usize) -> bool {
637        use self::Look::*;
638
639        match *self {
640            Start => at == 0,
641            End => at == haystack.len(),
642            StartLF => at == 0 || haystack[at - 1] == b'\n',
643            EndLF => at == haystack.len() || haystack[at] == b'\n',
644            StartCRLF => {
645                at == 0
646                    || haystack[at - 1] == b'\n'
647                    || (haystack[at - 1] == b'\r'
648                        && (at >= haystack.len() || haystack[at] != b'\n'))
649            }
650            EndCRLF => {
651                at == haystack.len()
652                    || haystack[at] == b'\r'
653                    || (haystack[at] == b'\n'
654                        && (at == 0 || haystack[at - 1] != b'\r'))
655            }
656            Word => {
657                let word_before =
658                    at > 0 && utf8::is_word_byte(haystack[at - 1]);
659                let word_after =
660                    at < haystack.len() && utf8::is_word_byte(haystack[at]);
661                word_before != word_after
662            }
663            WordNegate => {
664                let word_before =
665                    at > 0 && utf8::is_word_byte(haystack[at - 1]);
666                let word_after =
667                    at < haystack.len() && utf8::is_word_byte(haystack[at]);
668                word_before == word_after
669            }
670            WordStart => {
671                let word_before =
672                    at > 0 && utf8::is_word_byte(haystack[at - 1]);
673                let word_after =
674                    at < haystack.len() && utf8::is_word_byte(haystack[at]);
675                !word_before && word_after
676            }
677            WordEnd => {
678                let word_before =
679                    at > 0 && utf8::is_word_byte(haystack[at - 1]);
680                let word_after =
681                    at < haystack.len() && utf8::is_word_byte(haystack[at]);
682                word_before && !word_after
683            }
684            WordStartHalf => {
685                let word_before =
686                    at > 0 && utf8::is_word_byte(haystack[at - 1]);
687                !word_before
688            }
689            WordEndHalf => {
690                let word_after =
691                    at < haystack.len() && utf8::is_word_byte(haystack[at]);
692                !word_after
693            }
694        }
695    }
696}
697
698/// The high-level intermediate representation of a repetition operator.
699///
700/// A repetition operator permits the repetition of an arbitrary
701/// sub-expression.
702#[derive(Clone, Debug, Eq, PartialEq)]
703pub(crate) struct Repetition {
704    /// The minimum range of the repetition.
705    ///
706    /// Note that special cases like `?`, `+` and `*` all get translated into
707    /// the ranges `{0,1}`, `{1,}` and `{0,}`, respectively.
708    ///
709    /// When `min` is zero, this expression can match the empty string
710    /// regardless of what its sub-expression is.
711    pub(crate) min: u32,
712    /// The maximum range of the repetition.
713    ///
714    /// Note that when `max` is `None`, `min` acts as a lower bound but where
715    /// there is no upper bound. For something like `x{5}` where the min and
716    /// max are equivalent, `min` will be set to `5` and `max` will be set to
717    /// `Some(5)`.
718    pub(crate) max: Option<u32>,
719    /// Whether this repetition operator is greedy or not. A greedy operator
720    /// will match as much as it can. A non-greedy operator will match as
721    /// little as it can.
722    ///
723    /// Typically, operators are greedy by default and are only non-greedy when
724    /// a `?` suffix is used, e.g., `(expr)*` is greedy while `(expr)*?` is
725    /// not. However, this can be inverted via the `U` "ungreedy" flag.
726    pub(crate) greedy: bool,
727    /// The expression being repeated.
728    pub(crate) sub: Box<Hir>,
729}
730
731/// The high-level intermediate representation for a capturing group.
732///
733/// A capturing group always has an index and a child expression. It may
734/// also have a name associated with it (e.g., `(?P<foo>\w)`), but it's not
735/// necessary.
736///
737/// Note that there is no explicit representation of a non-capturing group
738/// in a `Hir`. Instead, non-capturing grouping is handled automatically by
739/// the recursive structure of the `Hir` itself.
740#[derive(Clone, Debug, Eq, PartialEq)]
741pub(crate) struct Capture {
742    /// The capture index of the capture.
743    pub(crate) index: u32,
744    /// The name of the capture, if it exists.
745    pub(crate) name: Option<Box<str>>,
746    /// The expression inside the capturing group, which may be empty.
747    pub(crate) sub: Box<Hir>,
748}
749
750fn next_char(ch: char) -> Option<char> {
751    // Skip over the surrogate range.
752    if ch == '\u{D7FF}' {
753        return Some('\u{E000}');
754    }
755    // OK because char::MAX < u32::MAX and we handle U+D7FF above.
756    char::from_u32(u32::from(ch).checked_add(1).unwrap())
757}
758
759fn prev_char(ch: char) -> Option<char> {
760    // Skip over the surrogate range.
761    if ch == '\u{E000}' {
762        return Some('\u{D7FF}');
763    }
764    // OK because subtracting 1 from any valid scalar value other than 0
765    // and U+E000 yields a valid scalar value.
766    Some(char::from_u32(u32::from(ch).checked_sub(1)?).unwrap())
767}
768
769impl Drop for Hir {
770    fn drop(&mut self) {
771        use core::mem;
772
773        match *self.kind() {
774            HirKind::Empty
775            | HirKind::Char(_)
776            | HirKind::Class(_)
777            | HirKind::Look(_) => return,
778            HirKind::Capture(ref x) if x.sub.kind.subs().is_empty() => return,
779            HirKind::Repetition(ref x) if x.sub.kind.subs().is_empty() => {
780                return
781            }
782            HirKind::Concat(ref x) if x.is_empty() => return,
783            HirKind::Alternation(ref x) if x.is_empty() => return,
784            _ => {}
785        }
786
787        let mut stack = vec![mem::replace(self, Hir::empty())];
788        while let Some(mut expr) = stack.pop() {
789            match expr.kind {
790                HirKind::Empty
791                | HirKind::Char(_)
792                | HirKind::Class(_)
793                | HirKind::Look(_) => {}
794                HirKind::Capture(ref mut x) => {
795                    stack.push(mem::replace(&mut x.sub, Hir::empty()));
796                }
797                HirKind::Repetition(ref mut x) => {
798                    stack.push(mem::replace(&mut x.sub, Hir::empty()));
799                }
800                HirKind::Concat(ref mut x) => {
801                    stack.extend(x.drain(..));
802                }
803                HirKind::Alternation(ref mut x) => {
804                    stack.extend(x.drain(..));
805                }
806            }
807        }
808    }
809}