summaryrefslogtreecommitdiffstats
path: root/src/librustdoc/html/highlight.rs
blob: 8922bf377858e21981838171b29250bf66acfb5e (plain)
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
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
//! Basic syntax highlighting functionality.
//!
//! This module uses librustc_ast's lexer to provide token-based highlighting for
//! the HTML documentation generated by rustdoc.
//!
//! Use the `render_with_highlighting` to highlight some rust code.

use crate::clean::PrimitiveType;
use crate::html::escape::Escape;
use crate::html::render::Context;

use std::collections::VecDeque;
use std::fmt::{Display, Write};

use rustc_data_structures::fx::FxHashMap;
use rustc_lexer::{LiteralKind, TokenKind};
use rustc_span::edition::Edition;
use rustc_span::symbol::Symbol;
use rustc_span::{BytePos, Span, DUMMY_SP};

use super::format::{self, Buffer};
use super::render::LinkFromSrc;

/// This type is needed in case we want to render links on items to allow to go to their definition.
pub(crate) struct HrefContext<'a, 'b, 'c> {
    pub(crate) context: &'a Context<'b>,
    /// This span contains the current file we're going through.
    pub(crate) file_span: Span,
    /// This field is used to know "how far" from the top of the directory we are to link to either
    /// documentation pages or other source pages.
    pub(crate) root_path: &'c str,
    /// This field is used to calculate precise local URLs.
    pub(crate) current_href: &'c str,
}

/// Decorations are represented as a map from CSS class to vector of character ranges.
/// Each range will be wrapped in a span with that class.
#[derive(Default)]
pub(crate) struct DecorationInfo(pub(crate) FxHashMap<&'static str, Vec<(u32, u32)>>);

#[derive(Eq, PartialEq, Clone, Copy)]
pub(crate) enum Tooltip {
    Ignore,
    CompileFail,
    ShouldPanic,
    Edition(Edition),
    None,
}

/// Highlights `src` as an inline example, returning the HTML output.
pub(crate) fn render_example_with_highlighting(
    src: &str,
    out: &mut Buffer,
    tooltip: Tooltip,
    playground_button: Option<&str>,
) {
    write_header(out, "rust-example-rendered", None, tooltip);
    write_code(out, src, None, None);
    write_footer(out, playground_button);
}

/// Highlights `src` as a macro, returning the HTML output.
pub(crate) fn render_macro_with_highlighting(src: &str, out: &mut Buffer) {
    write_header(out, "macro", None, Tooltip::None);
    write_code(out, src, None, None);
    write_footer(out, None);
}

/// Highlights `src` as a source code page, returning the HTML output.
pub(crate) fn render_source_with_highlighting(
    src: &str,
    out: &mut Buffer,
    line_numbers: Buffer,
    href_context: HrefContext<'_, '_, '_>,
    decoration_info: DecorationInfo,
) {
    write_header(out, "", Some(line_numbers), Tooltip::None);
    write_code(out, src, Some(href_context), Some(decoration_info));
    write_footer(out, None);
}

fn write_header(out: &mut Buffer, class: &str, extra_content: Option<Buffer>, tooltip: Tooltip) {
    write!(
        out,
        "<div class=\"example-wrap{}\">",
        match tooltip {
            Tooltip::Ignore => " ignore",
            Tooltip::CompileFail => " compile_fail",
            Tooltip::ShouldPanic => " should_panic",
            Tooltip::Edition(_) => " edition",
            Tooltip::None => "",
        },
    );

    if tooltip != Tooltip::None {
        write!(
            out,
            "<div class='tooltip'{}>ⓘ</div>",
            if let Tooltip::Edition(edition_info) = tooltip {
                format!(" data-edition=\"{}\"", edition_info)
            } else {
                String::new()
            },
        );
    }

    if let Some(extra) = extra_content {
        out.push_buffer(extra);
    }
    if class.is_empty() {
        write!(out, "<pre class=\"rust\">");
    } else {
        write!(out, "<pre class=\"rust {class}\">");
    }
    write!(out, "<code>");
}

/// Check if two `Class` can be merged together. In the following rules, "unclassified" means `None`
/// basically (since it's `Option<Class>`). The following rules apply:
///
/// * If two `Class` have the same variant, then they can be merged.
/// * If the other `Class` is unclassified and only contains white characters (backline,
///   whitespace, etc), it can be merged.
/// * `Class::Ident` is considered the same as unclassified (because it doesn't have an associated
///    CSS class).
fn can_merge(class1: Option<Class>, class2: Option<Class>, text: &str) -> bool {
    match (class1, class2) {
        (Some(c1), Some(c2)) => c1.is_equal_to(c2),
        (Some(Class::Ident(_)), None) | (None, Some(Class::Ident(_))) => true,
        (Some(_), None) | (None, Some(_)) => text.trim().is_empty(),
        (None, None) => true,
    }
}

/// This type is used as a conveniency to prevent having to pass all its fields as arguments into
/// the various functions (which became its methods).
struct TokenHandler<'a, 'b, 'c, 'd, 'e> {
    out: &'a mut Buffer,
    /// It contains the closing tag and the associated `Class`.
    closing_tags: Vec<(&'static str, Class)>,
    /// This is used because we don't automatically generate the closing tag on `ExitSpan` in
    /// case an `EnterSpan` event with the same class follows.
    pending_exit_span: Option<Class>,
    /// `current_class` and `pending_elems` are used to group HTML elements with same `class`
    /// attributes to reduce the DOM size.
    current_class: Option<Class>,
    /// We need to keep the `Class` for each element because it could contain a `Span` which is
    /// used to generate links.
    pending_elems: Vec<(&'b str, Option<Class>)>,
    href_context: Option<HrefContext<'c, 'd, 'e>>,
}

impl<'a, 'b, 'c, 'd, 'e> TokenHandler<'a, 'b, 'c, 'd, 'e> {
    fn handle_exit_span(&mut self) {
        // We can't get the last `closing_tags` element using `pop()` because `closing_tags` is
        // being used in `write_pending_elems`.
        let class = self.closing_tags.last().expect("ExitSpan without EnterSpan").1;
        // We flush everything just in case...
        self.write_pending_elems(Some(class));

        exit_span(self.out, self.closing_tags.pop().expect("ExitSpan without EnterSpan").0);
        self.pending_exit_span = None;
    }

    /// Write all the pending elements sharing a same (or at mergeable) `Class`.
    ///
    /// If there is a "parent" (if a `EnterSpan` event was encountered) and the parent can be merged
    /// with the elements' class, then we simply write the elements since the `ExitSpan` event will
    /// close the tag.
    ///
    /// Otherwise, if there is only one pending element, we let the `string` function handle both
    /// opening and closing the tag, otherwise we do it into this function.
    ///
    /// It returns `true` if `current_class` must be set to `None` afterwards.
    fn write_pending_elems(&mut self, current_class: Option<Class>) -> bool {
        if self.pending_elems.is_empty() {
            return false;
        }
        if let Some((_, parent_class)) = self.closing_tags.last() &&
            can_merge(current_class, Some(*parent_class), "")
        {
            for (text, class) in self.pending_elems.iter() {
                string(self.out, Escape(text), *class, &self.href_context, false);
            }
        } else {
            // We only want to "open" the tag ourselves if we have more than one pending and if the
            // current parent tag is not the same as our pending content.
            let close_tag = if self.pending_elems.len() > 1 && current_class.is_some() {
                Some(enter_span(self.out, current_class.unwrap(), &self.href_context))
            } else {
                None
            };
            for (text, class) in self.pending_elems.iter() {
                string(self.out, Escape(text), *class, &self.href_context, close_tag.is_none());
            }
            if let Some(close_tag) = close_tag {
                exit_span(self.out, close_tag);
            }
        }
        self.pending_elems.clear();
        true
    }
}

impl<'a, 'b, 'c, 'd, 'e> Drop for TokenHandler<'a, 'b, 'c, 'd, 'e> {
    /// When leaving, we need to flush all pending data to not have missing content.
    fn drop(&mut self) {
        if self.pending_exit_span.is_some() {
            self.handle_exit_span();
        } else {
            self.write_pending_elems(self.current_class);
        }
    }
}

/// Convert the given `src` source code into HTML by adding classes for highlighting.
///
/// This code is used to render code blocks (in the documentation) as well as the source code pages.
///
/// Some explanations on the last arguments:
///
/// In case we are rendering a code block and not a source code file, `href_context` will be `None`.
/// To put it more simply: if `href_context` is `None`, the code won't try to generate links to an
/// item definition.
///
/// More explanations about spans and how we use them here are provided in the
fn write_code(
    out: &mut Buffer,
    src: &str,
    href_context: Option<HrefContext<'_, '_, '_>>,
    decoration_info: Option<DecorationInfo>,
) {
    // This replace allows to fix how the code source with DOS backline characters is displayed.
    let src = src.replace("\r\n", "\n");
    let mut token_handler = TokenHandler {
        out,
        closing_tags: Vec::new(),
        pending_exit_span: None,
        current_class: None,
        pending_elems: Vec::new(),
        href_context,
    };

    Classifier::new(
        &src,
        token_handler.href_context.as_ref().map(|c| c.file_span).unwrap_or(DUMMY_SP),
        decoration_info,
    )
    .highlight(&mut |highlight| {
        match highlight {
            Highlight::Token { text, class } => {
                // If we received a `ExitSpan` event and then have a non-compatible `Class`, we
                // need to close the `<span>`.
                let need_current_class_update = if let Some(pending) = token_handler.pending_exit_span &&
                    !can_merge(Some(pending), class, text) {
                        token_handler.handle_exit_span();
                        true
                // If the two `Class` are different, time to flush the current content and start
                // a new one.
                } else if !can_merge(token_handler.current_class, class, text) {
                    token_handler.write_pending_elems(token_handler.current_class);
                    true
                } else {
                    token_handler.current_class.is_none()
                };

                if need_current_class_update {
                    token_handler.current_class = class.map(Class::dummy);
                }
                token_handler.pending_elems.push((text, class));
            }
            Highlight::EnterSpan { class } => {
                let mut should_add = true;
                if let Some(pending_exit_span) = token_handler.pending_exit_span {
                    if class.is_equal_to(pending_exit_span) {
                        should_add = false;
                    } else {
                        token_handler.handle_exit_span();
                    }
                } else {
                    // We flush everything just in case...
                    if token_handler.write_pending_elems(token_handler.current_class) {
                        token_handler.current_class = None;
                    }
                }
                if should_add {
                    let closing_tag = enter_span(token_handler.out, class, &token_handler.href_context);
                    token_handler.closing_tags.push((closing_tag, class));
                }

                token_handler.current_class = None;
                token_handler.pending_exit_span = None;
            }
            Highlight::ExitSpan => {
                token_handler.current_class = None;
                token_handler.pending_exit_span =
                    Some(token_handler.closing_tags.last().as_ref().expect("ExitSpan without EnterSpan").1);
            }
        };
    });
}

fn write_footer(out: &mut Buffer, playground_button: Option<&str>) {
    writeln!(out, "</code></pre>{}</div>", playground_button.unwrap_or_default());
}

/// How a span of text is classified. Mostly corresponds to token kinds.
#[derive(Clone, Copy, Debug, Eq, PartialEq)]
enum Class {
    Comment,
    DocComment,
    Attribute,
    KeyWord,
    /// Keywords that do pointer/reference stuff.
    RefKeyWord,
    Self_(Span),
    Macro(Span),
    MacroNonTerminal,
    String,
    Number,
    Bool,
    /// `Ident` isn't rendered in the HTML but we still need it for the `Span` it contains.
    Ident(Span),
    Lifetime,
    PreludeTy,
    PreludeVal,
    QuestionMark,
    Decoration(&'static str),
}

impl Class {
    /// It is only looking at the variant, not the variant content.
    ///
    /// It is used mostly to group multiple similar HTML elements into one `<span>` instead of
    /// multiple ones.
    fn is_equal_to(self, other: Self) -> bool {
        match (self, other) {
            (Self::Self_(_), Self::Self_(_))
            | (Self::Macro(_), Self::Macro(_))
            | (Self::Ident(_), Self::Ident(_)) => true,
            (Self::Decoration(c1), Self::Decoration(c2)) => c1 == c2,
            (x, y) => x == y,
        }
    }

    /// If `self` contains a `Span`, it'll be replaced with `DUMMY_SP` to prevent creating links
    /// on "empty content" (because of the attributes merge).
    fn dummy(self) -> Self {
        match self {
            Self::Self_(_) => Self::Self_(DUMMY_SP),
            Self::Macro(_) => Self::Macro(DUMMY_SP),
            Self::Ident(_) => Self::Ident(DUMMY_SP),
            s => s,
        }
    }

    /// Returns the css class expected by rustdoc for each `Class`.
    fn as_html(self) -> &'static str {
        match self {
            Class::Comment => "comment",
            Class::DocComment => "doccomment",
            Class::Attribute => "attribute",
            Class::KeyWord => "kw",
            Class::RefKeyWord => "kw-2",
            Class::Self_(_) => "self",
            Class::Macro(_) => "macro",
            Class::MacroNonTerminal => "macro-nonterminal",
            Class::String => "string",
            Class::Number => "number",
            Class::Bool => "bool-val",
            Class::Ident(_) => "",
            Class::Lifetime => "lifetime",
            Class::PreludeTy => "prelude-ty",
            Class::PreludeVal => "prelude-val",
            Class::QuestionMark => "question-mark",
            Class::Decoration(kind) => kind,
        }
    }

    /// In case this is an item which can be converted into a link to a definition, it'll contain
    /// a "span" (a tuple representing `(lo, hi)` equivalent of `Span`).
    fn get_span(self) -> Option<Span> {
        match self {
            Self::Ident(sp) | Self::Self_(sp) | Self::Macro(sp) => Some(sp),
            Self::Comment
            | Self::DocComment
            | Self::Attribute
            | Self::KeyWord
            | Self::RefKeyWord
            | Self::MacroNonTerminal
            | Self::String
            | Self::Number
            | Self::Bool
            | Self::Lifetime
            | Self::PreludeTy
            | Self::PreludeVal
            | Self::QuestionMark
            | Self::Decoration(_) => None,
        }
    }
}

enum Highlight<'a> {
    Token { text: &'a str, class: Option<Class> },
    EnterSpan { class: Class },
    ExitSpan,
}

struct TokenIter<'a> {
    src: &'a str,
}

impl<'a> Iterator for TokenIter<'a> {
    type Item = (TokenKind, &'a str);
    fn next(&mut self) -> Option<(TokenKind, &'a str)> {
        if self.src.is_empty() {
            return None;
        }
        let token = rustc_lexer::first_token(self.src);
        let (text, rest) = self.src.split_at(token.len as usize);
        self.src = rest;
        Some((token.kind, text))
    }
}

/// Classifies into identifier class; returns `None` if this is a non-keyword identifier.
fn get_real_ident_class(text: &str, allow_path_keywords: bool) -> Option<Class> {
    let ignore: &[&str] =
        if allow_path_keywords { &["self", "Self", "super", "crate"] } else { &["self", "Self"] };
    if ignore.iter().any(|k| *k == text) {
        return None;
    }
    Some(match text {
        "ref" | "mut" => Class::RefKeyWord,
        "false" | "true" => Class::Bool,
        _ if Symbol::intern(text).is_reserved(|| Edition::Edition2021) => Class::KeyWord,
        _ => return None,
    })
}

/// This iterator comes from the same idea than "Peekable" except that it allows to "peek" more than
/// just the next item by using `peek_next`. The `peek` method always returns the next item after
/// the current one whereas `peek_next` will return the next item after the last one peeked.
///
/// You can use both `peek` and `peek_next` at the same time without problem.
struct PeekIter<'a> {
    stored: VecDeque<(TokenKind, &'a str)>,
    /// This position is reinitialized when using `next`. It is used in `peek_next`.
    peek_pos: usize,
    iter: TokenIter<'a>,
}

impl<'a> PeekIter<'a> {
    fn new(iter: TokenIter<'a>) -> Self {
        Self { stored: VecDeque::new(), peek_pos: 0, iter }
    }
    /// Returns the next item after the current one. It doesn't interfere with `peek_next` output.
    fn peek(&mut self) -> Option<&(TokenKind, &'a str)> {
        if self.stored.is_empty() {
            if let Some(next) = self.iter.next() {
                self.stored.push_back(next);
            }
        }
        self.stored.front()
    }
    /// Returns the next item after the last one peeked. It doesn't interfere with `peek` output.
    fn peek_next(&mut self) -> Option<&(TokenKind, &'a str)> {
        self.peek_pos += 1;
        if self.peek_pos - 1 < self.stored.len() {
            self.stored.get(self.peek_pos - 1)
        } else if let Some(next) = self.iter.next() {
            self.stored.push_back(next);
            self.stored.back()
        } else {
            None
        }
    }
}

impl<'a> Iterator for PeekIter<'a> {
    type Item = (TokenKind, &'a str);
    fn next(&mut self) -> Option<Self::Item> {
        self.peek_pos = 0;
        if let Some(first) = self.stored.pop_front() { Some(first) } else { self.iter.next() }
    }
}

/// Custom spans inserted into the source. Eg --scrape-examples uses this to highlight function calls
struct Decorations {
    starts: Vec<(u32, &'static str)>,
    ends: Vec<u32>,
}

impl Decorations {
    fn new(info: DecorationInfo) -> Self {
        // Extract tuples (start, end, kind) into separate sequences of (start, kind) and (end).
        let (mut starts, mut ends): (Vec<_>, Vec<_>) = info
            .0
            .into_iter()
            .flat_map(|(kind, ranges)| ranges.into_iter().map(move |(lo, hi)| ((lo, kind), hi)))
            .unzip();

        // Sort the sequences in document order.
        starts.sort_by_key(|(lo, _)| *lo);
        ends.sort();

        Decorations { starts, ends }
    }
}

/// Processes program tokens, classifying strings of text by highlighting
/// category (`Class`).
struct Classifier<'a> {
    tokens: PeekIter<'a>,
    in_attribute: bool,
    in_macro: bool,
    in_macro_nonterminal: bool,
    byte_pos: u32,
    file_span: Span,
    src: &'a str,
    decorations: Option<Decorations>,
}

impl<'a> Classifier<'a> {
    /// Takes as argument the source code to HTML-ify, the rust edition to use and the source code
    /// file span which will be used later on by the `span_correspondance_map`.
    fn new(src: &str, file_span: Span, decoration_info: Option<DecorationInfo>) -> Classifier<'_> {
        let tokens = PeekIter::new(TokenIter { src });
        let decorations = decoration_info.map(Decorations::new);
        Classifier {
            tokens,
            in_attribute: false,
            in_macro: false,
            in_macro_nonterminal: false,
            byte_pos: 0,
            file_span,
            src,
            decorations,
        }
    }

    /// Convenient wrapper to create a [`Span`] from a position in the file.
    fn new_span(&self, lo: u32, text: &str) -> Span {
        let hi = lo + text.len() as u32;
        let file_lo = self.file_span.lo();
        self.file_span.with_lo(file_lo + BytePos(lo)).with_hi(file_lo + BytePos(hi))
    }

    /// Concatenate colons and idents as one when possible.
    fn get_full_ident_path(&mut self) -> Vec<(TokenKind, usize, usize)> {
        let start = self.byte_pos as usize;
        let mut pos = start;
        let mut has_ident = false;

        loop {
            let mut nb = 0;
            while let Some((TokenKind::Colon, _)) = self.tokens.peek() {
                self.tokens.next();
                nb += 1;
            }
            // Ident path can start with "::" but if we already have content in the ident path,
            // the "::" is mandatory.
            if has_ident && nb == 0 {
                return vec![(TokenKind::Ident, start, pos)];
            } else if nb != 0 && nb != 2 {
                if has_ident {
                    return vec![(TokenKind::Ident, start, pos), (TokenKind::Colon, pos, pos + nb)];
                } else {
                    return vec![(TokenKind::Colon, start, pos + nb)];
                }
            }

            if let Some((None, text)) = self.tokens.peek().map(|(token, text)| {
                if *token == TokenKind::Ident {
                    let class = get_real_ident_class(text, true);
                    (class, text)
                } else {
                    // Doesn't matter which Class we put in here...
                    (Some(Class::Comment), text)
                }
            }) {
                // We only "add" the colon if there is an ident behind.
                pos += text.len() + nb;
                has_ident = true;
                self.tokens.next();
            } else if nb > 0 && has_ident {
                return vec![(TokenKind::Ident, start, pos), (TokenKind::Colon, pos, pos + nb)];
            } else if nb > 0 {
                return vec![(TokenKind::Colon, start, start + nb)];
            } else if has_ident {
                return vec![(TokenKind::Ident, start, pos)];
            } else {
                return Vec::new();
            }
        }
    }

    /// Wraps the tokens iteration to ensure that the `byte_pos` is always correct.
    ///
    /// It returns the token's kind, the token as a string and its byte position in the source
    /// string.
    fn next(&mut self) -> Option<(TokenKind, &'a str, u32)> {
        if let Some((kind, text)) = self.tokens.next() {
            let before = self.byte_pos;
            self.byte_pos += text.len() as u32;
            Some((kind, text, before))
        } else {
            None
        }
    }

    /// Exhausts the `Classifier` writing the output into `sink`.
    ///
    /// The general structure for this method is to iterate over each token,
    /// possibly giving it an HTML span with a class specifying what flavor of
    /// token is used.
    fn highlight(mut self, sink: &mut dyn FnMut(Highlight<'a>)) {
        loop {
            if let Some(decs) = self.decorations.as_mut() {
                let byte_pos = self.byte_pos;
                let n_starts = decs.starts.iter().filter(|(i, _)| byte_pos >= *i).count();
                for (_, kind) in decs.starts.drain(0..n_starts) {
                    sink(Highlight::EnterSpan { class: Class::Decoration(kind) });
                }

                let n_ends = decs.ends.iter().filter(|i| byte_pos >= **i).count();
                for _ in decs.ends.drain(0..n_ends) {
                    sink(Highlight::ExitSpan);
                }
            }

            if self
                .tokens
                .peek()
                .map(|t| matches!(t.0, TokenKind::Colon | TokenKind::Ident))
                .unwrap_or(false)
            {
                let tokens = self.get_full_ident_path();
                for (token, start, end) in &tokens {
                    let text = &self.src[*start..*end];
                    self.advance(*token, text, sink, *start as u32);
                    self.byte_pos += text.len() as u32;
                }
                if !tokens.is_empty() {
                    continue;
                }
            }
            if let Some((token, text, before)) = self.next() {
                self.advance(token, text, sink, before);
            } else {
                break;
            }
        }
    }

    /// Single step of highlighting. This will classify `token`, but maybe also a couple of
    /// following ones as well.
    ///
    /// `before` is the position of the given token in the `source` string and is used as "lo" byte
    /// in case we want to try to generate a link for this token using the
    /// `span_correspondance_map`.
    fn advance(
        &mut self,
        token: TokenKind,
        text: &'a str,
        sink: &mut dyn FnMut(Highlight<'a>),
        before: u32,
    ) {
        let lookahead = self.peek();
        let no_highlight = |sink: &mut dyn FnMut(_)| sink(Highlight::Token { text, class: None });
        let class = match token {
            TokenKind::Whitespace => return no_highlight(sink),
            TokenKind::LineComment { doc_style } | TokenKind::BlockComment { doc_style, .. } => {
                if doc_style.is_some() {
                    Class::DocComment
                } else {
                    Class::Comment
                }
            }
            // Consider this as part of a macro invocation if there was a
            // leading identifier.
            TokenKind::Bang if self.in_macro => {
                self.in_macro = false;
                sink(Highlight::Token { text, class: None });
                sink(Highlight::ExitSpan);
                return;
            }

            // Assume that '&' or '*' is the reference or dereference operator
            // or a reference or pointer type. Unless, of course, it looks like
            // a logical and or a multiplication operator: `&&` or `* `.
            TokenKind::Star => match self.tokens.peek() {
                Some((TokenKind::Whitespace, _)) => return no_highlight(sink),
                Some((TokenKind::Ident, "mut")) => {
                    self.next();
                    sink(Highlight::Token { text: "*mut", class: Some(Class::RefKeyWord) });
                    return;
                }
                Some((TokenKind::Ident, "const")) => {
                    self.next();
                    sink(Highlight::Token { text: "*const", class: Some(Class::RefKeyWord) });
                    return;
                }
                _ => Class::RefKeyWord,
            },
            TokenKind::And => match self.tokens.peek() {
                Some((TokenKind::And, _)) => {
                    self.next();
                    sink(Highlight::Token { text: "&&", class: None });
                    return;
                }
                Some((TokenKind::Eq, _)) => {
                    self.next();
                    sink(Highlight::Token { text: "&=", class: None });
                    return;
                }
                Some((TokenKind::Whitespace, _)) => return no_highlight(sink),
                Some((TokenKind::Ident, "mut")) => {
                    self.next();
                    sink(Highlight::Token { text: "&mut", class: Some(Class::RefKeyWord) });
                    return;
                }
                _ => Class::RefKeyWord,
            },

            // These can either be operators, or arrows.
            TokenKind::Eq => match lookahead {
                Some(TokenKind::Eq) => {
                    self.next();
                    sink(Highlight::Token { text: "==", class: None });
                    return;
                }
                Some(TokenKind::Gt) => {
                    self.next();
                    sink(Highlight::Token { text: "=>", class: None });
                    return;
                }
                _ => return no_highlight(sink),
            },
            TokenKind::Minus if lookahead == Some(TokenKind::Gt) => {
                self.next();
                sink(Highlight::Token { text: "->", class: None });
                return;
            }

            // Other operators.
            TokenKind::Minus
            | TokenKind::Plus
            | TokenKind::Or
            | TokenKind::Slash
            | TokenKind::Caret
            | TokenKind::Percent
            | TokenKind::Bang
            | TokenKind::Lt
            | TokenKind::Gt => return no_highlight(sink),

            // Miscellaneous, no highlighting.
            TokenKind::Dot
            | TokenKind::Semi
            | TokenKind::Comma
            | TokenKind::OpenParen
            | TokenKind::CloseParen
            | TokenKind::OpenBrace
            | TokenKind::CloseBrace
            | TokenKind::OpenBracket
            | TokenKind::At
            | TokenKind::Tilde
            | TokenKind::Colon
            | TokenKind::Unknown => return no_highlight(sink),

            TokenKind::Question => Class::QuestionMark,

            TokenKind::Dollar => match lookahead {
                Some(TokenKind::Ident) => {
                    self.in_macro_nonterminal = true;
                    Class::MacroNonTerminal
                }
                _ => return no_highlight(sink),
            },

            // This might be the start of an attribute. We're going to want to
            // continue highlighting it as an attribute until the ending ']' is
            // seen, so skip out early. Down below we terminate the attribute
            // span when we see the ']'.
            TokenKind::Pound => {
                match lookahead {
                    // Case 1: #![inner_attribute]
                    Some(TokenKind::Bang) => {
                        self.next();
                        if let Some(TokenKind::OpenBracket) = self.peek() {
                            self.in_attribute = true;
                            sink(Highlight::EnterSpan { class: Class::Attribute });
                        }
                        sink(Highlight::Token { text: "#", class: None });
                        sink(Highlight::Token { text: "!", class: None });
                        return;
                    }
                    // Case 2: #[outer_attribute]
                    Some(TokenKind::OpenBracket) => {
                        self.in_attribute = true;
                        sink(Highlight::EnterSpan { class: Class::Attribute });
                    }
                    _ => (),
                }
                return no_highlight(sink);
            }
            TokenKind::CloseBracket => {
                if self.in_attribute {
                    self.in_attribute = false;
                    sink(Highlight::Token { text: "]", class: None });
                    sink(Highlight::ExitSpan);
                    return;
                }
                return no_highlight(sink);
            }
            TokenKind::Literal { kind, .. } => match kind {
                // Text literals.
                LiteralKind::Byte { .. }
                | LiteralKind::Char { .. }
                | LiteralKind::Str { .. }
                | LiteralKind::ByteStr { .. }
                | LiteralKind::RawStr { .. }
                | LiteralKind::RawByteStr { .. } => Class::String,
                // Number literals.
                LiteralKind::Float { .. } | LiteralKind::Int { .. } => Class::Number,
            },
            TokenKind::Ident | TokenKind::RawIdent if lookahead == Some(TokenKind::Bang) => {
                self.in_macro = true;
                sink(Highlight::EnterSpan { class: Class::Macro(self.new_span(before, text)) });
                sink(Highlight::Token { text, class: None });
                return;
            }
            TokenKind::Ident => match get_real_ident_class(text, false) {
                None => match text {
                    "Option" | "Result" => Class::PreludeTy,
                    "Some" | "None" | "Ok" | "Err" => Class::PreludeVal,
                    // "union" is a weak keyword and is only considered as a keyword when declaring
                    // a union type.
                    "union" if self.check_if_is_union_keyword() => Class::KeyWord,
                    _ if self.in_macro_nonterminal => {
                        self.in_macro_nonterminal = false;
                        Class::MacroNonTerminal
                    }
                    "self" | "Self" => Class::Self_(self.new_span(before, text)),
                    _ => Class::Ident(self.new_span(before, text)),
                },
                Some(c) => c,
            },
            TokenKind::RawIdent | TokenKind::UnknownPrefix | TokenKind::InvalidIdent => {
                Class::Ident(self.new_span(before, text))
            }
            TokenKind::Lifetime { .. } => Class::Lifetime,
        };
        // Anything that didn't return above is the simple case where we the
        // class just spans a single token, so we can use the `string` method.
        sink(Highlight::Token { text, class: Some(class) });
    }

    fn peek(&mut self) -> Option<TokenKind> {
        self.tokens.peek().map(|(token_kind, _text)| *token_kind)
    }

    fn check_if_is_union_keyword(&mut self) -> bool {
        while let Some(kind) = self.tokens.peek_next().map(|(token_kind, _text)| token_kind) {
            if *kind == TokenKind::Whitespace {
                continue;
            }
            return *kind == TokenKind::Ident;
        }
        false
    }
}

/// Called when we start processing a span of text that should be highlighted.
/// The `Class` argument specifies how it should be highlighted.
fn enter_span(
    out: &mut Buffer,
    klass: Class,
    href_context: &Option<HrefContext<'_, '_, '_>>,
) -> &'static str {
    string_without_closing_tag(out, "", Some(klass), href_context, true).expect(
        "internal error: enter_span was called with Some(klass) but did not return a \
            closing HTML tag",
    )
}

/// Called at the end of a span of highlighted text.
fn exit_span(out: &mut Buffer, closing_tag: &str) {
    out.write_str(closing_tag);
}

/// Called for a span of text. If the text should be highlighted differently
/// from the surrounding text, then the `Class` argument will be a value other
/// than `None`.
///
/// The following sequences of callbacks are equivalent:
/// ```plain
///     enter_span(Foo), string("text", None), exit_span()
///     string("text", Foo)
/// ```
///
/// The latter can be thought of as a shorthand for the former, which is more
/// flexible.
///
/// Note that if `context` is not `None` and that the given `klass` contains a `Span`, the function
/// will then try to find this `span` in the `span_correspondance_map`. If found, it'll then
/// generate a link for this element (which corresponds to where its definition is located).
fn string<T: Display>(
    out: &mut Buffer,
    text: T,
    klass: Option<Class>,
    href_context: &Option<HrefContext<'_, '_, '_>>,
    open_tag: bool,
) {
    if let Some(closing_tag) = string_without_closing_tag(out, text, klass, href_context, open_tag)
    {
        out.write_str(closing_tag);
    }
}

/// This function writes `text` into `out` with some modifications depending on `klass`:
///
/// * If `klass` is `None`, `text` is written into `out` with no modification.
/// * If `klass` is `Some` but `klass.get_span()` is `None`, it writes the text wrapped in a
///   `<span>` with the provided `klass`.
/// * If `klass` is `Some` and has a [`rustc_span::Span`], it then tries to generate a link (`<a>`
///   element) by retrieving the link information from the `span_correspondance_map` that was filled
///   in `span_map.rs::collect_spans_and_sources`. If it cannot retrieve the information, then it's
///   the same as the second point (`klass` is `Some` but doesn't have a [`rustc_span::Span`]).
fn string_without_closing_tag<T: Display>(
    out: &mut Buffer,
    text: T,
    klass: Option<Class>,
    href_context: &Option<HrefContext<'_, '_, '_>>,
    open_tag: bool,
) -> Option<&'static str> {
    let Some(klass) = klass
    else {
        write!(out, "{}", text);
        return None;
    };
    let Some(def_span) = klass.get_span()
    else {
        if !open_tag {
            write!(out, "{}", text);
            return None;
        }
        write!(out, "<span class=\"{}\">{}", klass.as_html(), text);
        return Some("</span>");
    };

    let mut text_s = text.to_string();
    if text_s.contains("::") {
        text_s = text_s.split("::").intersperse("::").fold(String::new(), |mut path, t| {
            match t {
                "self" | "Self" => write!(
                    &mut path,
                    "<span class=\"{}\">{}</span>",
                    Class::Self_(DUMMY_SP).as_html(),
                    t
                ),
                "crate" | "super" => {
                    write!(&mut path, "<span class=\"{}\">{}</span>", Class::KeyWord.as_html(), t)
                }
                t => write!(&mut path, "{}", t),
            }
            .expect("Failed to build source HTML path");
            path
        });
    }

    if let Some(href_context) = href_context {
        if let Some(href) =
            href_context.context.shared.span_correspondance_map.get(&def_span).and_then(|href| {
                let context = href_context.context;
                // FIXME: later on, it'd be nice to provide two links (if possible) for all items:
                // one to the documentation page and one to the source definition.
                // FIXME: currently, external items only generate a link to their documentation,
                // a link to their definition can be generated using this:
                // https://github.com/rust-lang/rust/blob/60f1a2fc4b535ead9c85ce085fdce49b1b097531/src/librustdoc/html/render/context.rs#L315-L338
                match href {
                    LinkFromSrc::Local(span) => {
                        context.href_from_span_relative(*span, href_context.current_href)
                    }
                    LinkFromSrc::External(def_id) => {
                        format::href_with_root_path(*def_id, context, Some(href_context.root_path))
                            .ok()
                            .map(|(url, _, _)| url)
                    }
                    LinkFromSrc::Primitive(prim) => format::href_with_root_path(
                        PrimitiveType::primitive_locations(context.tcx())[prim],
                        context,
                        Some(href_context.root_path),
                    )
                    .ok()
                    .map(|(url, _, _)| url),
                }
            })
        {
            if !open_tag {
                // We're already inside an element which has the same klass, no need to give it
                // again.
                write!(out, "<a href=\"{}\">{}", href, text_s);
            } else {
                let klass_s = klass.as_html();
                if klass_s.is_empty() {
                    write!(out, "<a href=\"{}\">{}", href, text_s);
                } else {
                    write!(out, "<a class=\"{}\" href=\"{}\">{}", klass_s, href, text_s);
                }
            }
            return Some("</a>");
        }
    }
    if !open_tag {
        write!(out, "{}", text_s);
        return None;
    }
    let klass_s = klass.as_html();
    if klass_s.is_empty() {
        write!(out, "{}", text_s);
        Some("")
    } else {
        write!(out, "<span class=\"{}\">{}", klass_s, text_s);
        Some("</span>")
    }
}

#[cfg(test)]
mod tests;