zumen-xml 0.3.0

zumen の XML の L0(決定 0209)。要素・属性・文字と名前空間を、元のバイト範囲で読む
Documentation
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
//! XML 1.0 と XML の名前空間 1.0 の文書を読む(決定 0209)。
//!
//! 入れ子は再帰でなく、開いた要素の並びでたどる。要素を予算のレコード、入れ子を深さで数える(決定 0010)。

use core::ops::Range;
use std::collections::{BTreeMap, BTreeSet};

use zumen_base::{Cx, Diagnostic, Error, Kind, Ref, Stop, codes};

use crate::{Attribute, Document, Element, Encoding, Name, Namespace, Node, name, text, utf16};

/// 読めない理由
enum Fail {
    /// XML の形になっていない。範囲はその場所
    Invalid(Range<usize>),
    /// 予算か中断の合図で止めた。範囲は扱っていた場所
    Stop(Stop, Range<usize>),
}

/// XML の文書を読む。XML の形になっていなければ `xml.document.invalid`、UTF-8・Shift_JIS・UTF-16 のほかの文字コードなら
/// `xml.encoding.unsupported` を報告して読まない。文書型宣言は処理せず `xml.dtd.dropped` を報告する
pub fn read(src: &[u8], cx: &mut Cx) -> Result<Document, Error> {
    cx.check(Kind::InputBytes, src.len() as u64)
        .map_err(|stop| stopped(cx, stop, None))?;
    if utf16::is_utf32(src) {
        return Err(unsupported(cx, "UTF-32"));
    }
    if let Some(detection) = utf16::detect(src) {
        return read_utf16(src, detection, cx);
    }
    let mut p = Parser::new(src);
    let encoding = p.encoding(cx)?;
    match p.document(cx) {
        Ok(()) => Ok(Document {
            src: src.to_vec(),
            encoding,
            elements: p.elements,
            doctype: p.doctype,
            decoded: None,
        }),
        Err(Fail::Invalid(range)) => {
            cx.report(Diagnostic::new(codes::XML_DOCUMENT_INVALID).at(at(range)));
            Err(Error::Invalid)
        }
        Err(Fail::Stop(stop, range)) => Err(stopped(cx, stop, Some(range))),
    }
}

fn read_utf16(src: &[u8], detection: utf16::Detection, cx: &mut Cx) -> Result<Document, Error> {
    let decoded = utf16::Decoded::read(src, detection, cx)?;
    let mut p = Parser::new(&decoded.bytes);
    p.mapping = Some(&decoded);
    let encoding = p.encoding_expected(cx, Some(detection))?;
    match p.document(cx) {
        Ok(()) => {}
        Err(Fail::Invalid(range)) => {
            cx.report(Diagnostic::new(codes::XML_DOCUMENT_INVALID).at(p.place(range)));
            return Err(Error::Invalid);
        }
        Err(Fail::Stop(stop, range)) => {
            return Err(stopped(cx, stop, Some(decoded.original_range(range))));
        }
    }
    let mut elements = p.elements;
    let mut doctype = p.doctype;
    // 全公開範囲を原本へ戻す。要素番号と親子の順は動かさない。
    for element in &mut elements {
        let span = element.span.clone();
        map_charge(cx, &decoded, span.clone())?;
        element.span = decoded.original_range(span.clone());
        map_name(&mut element.name, &decoded);
        for attribute in &mut element.attributes {
            map_charge(cx, &decoded, attribute.name.raw.clone())?;
            map_name(&mut attribute.name, &decoded);
            attribute.value = decoded.original_range(attribute.value.clone());
        }
        for node in &mut element.children {
            let range = match node {
                Node::Text(range) | Node::CData(range) => range.clone(),
                Node::Element(_) => span.clone(),
            };
            map_charge(cx, &decoded, range)?;
            if let Node::Text(range) | Node::CData(range) = node {
                *range = decoded.original_range(range.clone());
            }
        }
    }
    if let Some(range) = &mut doctype {
        map_charge(cx, &decoded, range.clone())?;
        *range = decoded.original_range(range.clone());
    }
    Ok(Document {
        src: src.to_vec(),
        encoding,
        elements,
        doctype,
        decoded: Some(decoded),
    })
}

fn map_charge(cx: &mut Cx, decoded: &utf16::Decoded, range: Range<usize>) -> Result<(), Error> {
    cx.check(Kind::Work, 1)
        .map_err(|stop| stopped(cx, stop, Some(decoded.original_range(range))))
}

fn map_name(name: &mut Name, decoded: &utf16::Decoded) {
    name.raw = decoded.original_range(name.raw.clone());
    name.local = decoded.original_range(name.local..name.local).start;
    name.prefix_end = name
        .prefix_end
        .map(|end| decoded.original_range(end..end).start);
    if let Some(Namespace::Declared(range)) = &mut name.namespace {
        *range = decoded.original_range(range.clone());
    }
}

fn at(range: Range<usize>) -> Ref {
    Ref::at_bytes(range.start as u64, range.end as u64)
}

/// 止めた理由を報告する。扱っていたバイト範囲があれば、場所にする(決定 0119)
fn stopped(cx: &mut Cx, stop: Stop, place: Option<Range<usize>>) -> Error {
    let diagnostic = stop.to_diagnostic();
    cx.report(match place {
        Some(range) => diagnostic.at(at(range)),
        None => diagnostic,
    });
    Error::Stopped(stop)
}

fn unsupported(cx: &mut Cx, encoding: &str) -> Error {
    cx.report(Diagnostic::new(codes::XML_ENCODING_UNSUPPORTED).with("encoding", encoding));
    Error::Unsupported
}

fn is_space(byte: u8) -> bool {
    matches!(byte, b' ' | b'\t' | b'\r' | b'\n')
}

/// 名前に使えないバイト。0x80 以上は、どの文字コードでも名前の文字として通す
fn is_delimiter(byte: u8) -> bool {
    is_space(byte)
        || matches!(
            byte,
            b'<' | b'>' | b'/' | b'=' | b'!' | b'?' | b'"' | b'\'' | b'&' | b';' | b'[' | b']'
        )
}

fn find(src: &[u8], pattern: &[u8], from: usize) -> Option<usize> {
    src.get(from..)?
        .windows(pattern.len())
        .position(|window| window == pattern)
        .map(|at| from + at)
}

/// `src` の `range` の中で、`byte` の最初の位置。見たバイトの数を `seen` に足す
fn find_in(src: &[u8], byte: u8, range: Range<usize>, seen: &mut usize) -> Option<usize> {
    let from = range.start;
    let rest = src.get(range).unwrap_or_default();
    let found = rest.iter().position(|b| *b == byte);
    *seen += found.map_or(rest.len(), |at| at + 1);
    found.map(|at| from + at)
}

/// XML 宣言の擬似属性(`version="1.0" encoding="…"`)の名前と値。形が違えば `None`
fn pseudo_attributes(declaration: &[u8]) -> Option<Vec<(&[u8], &[u8])>> {
    let mut pairs = Vec::new();
    let mut rest = declaration;
    loop {
        let trimmed = rest.trim_ascii_start();
        if trimmed.is_empty() {
            return Some(pairs);
        }
        if trimmed.len() == rest.len() && !pairs.is_empty() {
            return None;
        }
        let name_end = trimmed.iter().position(|b| *b == b'=' || is_space(*b))?;
        let (name, after) = trimmed.split_at(name_end);
        let after = after
            .trim_ascii_start()
            .strip_prefix(b"=")?
            .trim_ascii_start();
        let (&quote, after) = after.split_first()?;
        if !matches!(quote, b'"' | b'\'') {
            return None;
        }
        let end = after.iter().position(|b| *b == quote)?;
        let (value, after) = after.split_at(end);
        pairs.push((name, value));
        rest = after.get(1..).unwrap_or_default();
    }
}

/// 名前空間の宣言の接頭辞。既定の名前空間は `None`
type Prefix<'s> = Option<&'s [u8]>;

/// 名前空間の宣言の値の範囲。`xmlns=""` は `None`
type Declared = Option<Range<usize>>;

struct Parser<'s> {
    src: &'s [u8],
    mapping: Option<&'s utf16::Decoded>,
    shift_jis: bool,
    pos: usize,
    elements: Vec<Element>,
    /// 開いている要素の名前空間の宣言を、接頭辞ごとに積む。内側の宣言が後。名前を引く手間は、宣言の数に比例せず、
    /// 接頭辞の種類の数の対数になる(#721)
    namespaces: BTreeMap<Prefix<'s>, Vec<Declared>>,
    /// 開いている要素の名前空間の宣言の接頭辞。内側の宣言が後
    scopes: Vec<Prefix<'s>>,
    /// 開いている要素ごとの、`scopes` の長さ。閉じるときに戻す
    marks: Vec<usize>,
    doctype: Option<Range<usize>>,
}

impl<'s> Parser<'s> {
    fn new(src: &'s [u8]) -> Self {
        Self {
            src,
            mapping: None,
            shift_jis: false,
            pos: 0,
            elements: Vec::new(),
            namespaces: BTreeMap::new(),
            scopes: Vec::new(),
            marks: Vec::new(),
            doctype: None,
        }
    }

    fn peek(&self) -> Option<u8> {
        self.src.get(self.pos).copied()
    }

    fn place(&self, range: Range<usize>) -> Ref {
        at(match self.mapping {
            Some(decoded) => decoded.original_range(range),
            None => range,
        })
    }

    fn starts(&self, pattern: &[u8]) -> bool {
        self.src
            .get(self.pos..)
            .is_some_and(|rest| rest.starts_with(pattern))
    }

    fn bytes(&self, range: Range<usize>) -> &'s [u8] {
        self.src.get(range).unwrap_or_default()
    }

    /// 名前空間を宣言する
    fn declare(&mut self, prefix: Prefix<'s>, value: Declared) {
        self.namespaces.entry(prefix).or_default().push(value);
        self.scopes.push(prefix);
    }

    /// `scopes` を `mark` の長さに戻し、そのあとの宣言を外す
    fn unwind(&mut self, mark: usize) {
        while self.scopes.len() > mark {
            let Some(prefix) = self.scopes.pop() else {
                break;
            };
            if let Some(stack) = self.namespaces.get_mut(&prefix) {
                stack.pop();
                if stack.is_empty() {
                    self.namespaces.remove(&prefix);
                }
            }
        }
    }

    /// 接頭辞のいちばん内側の宣言。宣言がなければ `None`
    fn declared(&self, prefix: Prefix<'s>) -> Option<&Declared> {
        self.namespaces.get(&prefix).and_then(|stack| stack.last())
    }

    /// 空白を飛ばす。飛ばしたら真
    fn skip_space(&mut self) -> bool {
        let start = self.pos;
        while self.peek().is_some_and(is_space) {
            self.pos += 1;
        }
        self.pos > start
    }

    fn work(&self, cx: &mut Cx, amount: u64, start: usize) -> Result<(), Fail> {
        cx.check(Kind::Work, amount)
            .map_err(|stop| Fail::Stop(stop, start..self.pos))
    }

    /// バイト順マークと XML 宣言から、文字コードを決める
    fn encoding(&mut self, cx: &mut Cx) -> Result<Encoding, Error> {
        let encoding = self.encoding_expected(cx, None)?;
        self.shift_jis = encoding == Encoding::ShiftJis;
        Ok(encoding)
    }

    fn encoding_expected(
        &mut self,
        cx: &mut Cx,
        expected: Option<utf16::Detection>,
    ) -> Result<Encoding, Error> {
        // UTF-16 の原本の BOM は復号時に除去済み。残る U+FEFF は文書の文字なので除かない。
        let bom = expected.is_none() && self.starts(&[0xef, 0xbb, 0xbf]);
        if bom {
            self.pos = 3;
        } else if matches!(
            self.src,
            [0xfe, 0xff, ..] | [0xff, 0xfe, ..] | [0, _, ..] | [_, 0, ..]
        ) {
            return Err(unsupported(cx, "UTF-16"));
        }
        if !(self.starts(b"<?xml") && self.src.get(self.pos + 5).copied().is_some_and(is_space)) {
            if let Some(expected) = expected {
                if !expected.bom {
                    cx.report(
                        Diagnostic::new(codes::XML_DOCUMENT_INVALID)
                            .at(self.place(0..self.src.len().min(1))),
                    );
                    return Err(Error::Invalid);
                }
                return Ok(match expected.order {
                    utf16::Order::Little => Encoding::Utf16Le,
                    utf16::Order::Big => Encoding::Utf16Be,
                });
            }
            return Ok(Encoding::Utf8);
        }
        let start = self.pos;
        let Some(end) = find(self.src, b"?>", start) else {
            cx.report(
                Diagnostic::new(codes::XML_DOCUMENT_INVALID).at(self.place(start..self.src.len())),
            );
            return Err(Error::Invalid);
        };
        let src = self.src;
        let declaration = src.get(start + 5..end).unwrap_or_default();
        let Some(pairs) = pseudo_attributes(declaration) else {
            cx.report(Diagnostic::new(codes::XML_DOCUMENT_INVALID).at(self.place(start..end + 2)));
            return Err(Error::Invalid);
        };
        if pairs.first().map(|(name, _)| *name) != Some(&b"version"[..]) {
            cx.report(Diagnostic::new(codes::XML_DOCUMENT_INVALID).at(self.place(start..end + 2)));
            return Err(Error::Invalid);
        }
        self.pos = end + 2;
        if let Some(expected) = expected {
            let label = pairs
                .iter()
                .find(|(name, _)| *name == b"encoding")
                .map(|(_, label)| *label);
            // XML 1.0 4.3.3 と RFC 2781 3.3。generic UTF-16 は BOM、LE/BE ラベルは BOM なし。
            let valid = if expected.bom {
                label.is_none_or(|label| label.eq_ignore_ascii_case(b"UTF-16"))
            } else {
                label.is_some_and(|label| {
                    label.eq_ignore_ascii_case(match expected.order {
                        utf16::Order::Little => b"UTF-16LE",
                        utf16::Order::Big => b"UTF-16BE",
                    })
                })
            };
            if !valid {
                cx.report(
                    Diagnostic::new(codes::XML_DOCUMENT_INVALID).at(self.place(start..end + 2)),
                );
                return Err(Error::Invalid);
            }
            return Ok(match expected.order {
                utf16::Order::Little => Encoding::Utf16Le,
                utf16::Order::Big => Encoding::Utf16Be,
            });
        }
        let Some((_, label)) = pairs.iter().find(|(name, _)| *name == b"encoding") else {
            return Ok(Encoding::Utf8);
        };
        let found = encoding_rs::Encoding::for_label(label);
        let label = String::from_utf8_lossy(label);
        if found == Some(encoding_rs::UTF_8) {
            Ok(Encoding::Utf8)
        } else if found == Some(encoding_rs::SHIFT_JIS) && !bom {
            Ok(Encoding::ShiftJis)
        } else {
            Err(unsupported(cx, &label))
        }
    }

    /// 文書。ルートの要素の前と後には、空白・注記・処理命令と、前に 1 つの文書型宣言だけを置ける
    fn document(&mut self, cx: &mut Cx) -> Result<(), Fail> {
        let mut root = false;
        loop {
            self.skip_space();
            let start = self.pos;
            if self.pos >= self.src.len() {
                return if root {
                    Ok(())
                } else {
                    Err(Fail::Invalid(start..start))
                };
            }
            self.work(cx, 1, start)?;
            if self.starts(b"<?") {
                self.skip_to(b"?>", start)?;
            } else if self.starts(b"<!--") {
                self.skip_to(b"-->", start)?;
            } else if !root && self.doctype.is_none() && self.starts(b"<!DOCTYPE") {
                self.doctype(cx)?;
            } else if !root && self.peek() == Some(b'<') {
                self.tree(cx)?;
                root = true;
            } else {
                return Err(Fail::Invalid(start..self.src.len()));
            }
        }
    }

    /// `pattern` の後まで飛ばす。なければ、`start` から終わりまでを壊れているとする
    fn skip_to(&mut self, pattern: &[u8], start: usize) -> Result<(), Fail> {
        let end =
            find(self.src, pattern, self.pos + 2).ok_or(Fail::Invalid(start..self.src.len()))?;
        self.pos = end + pattern.len();
        Ok(())
    }

    /// 文書型宣言を、内部の部分集合ごと飛ばす。中は処理しない
    fn doctype(&mut self, cx: &mut Cx) -> Result<(), Fail> {
        let start = self.pos;
        self.pos += b"<!DOCTYPE".len();
        let mut depth = 0usize;
        loop {
            let byte = self.peek().ok_or(Fail::Invalid(start..self.src.len()))?;
            if self.shift_jis && matches!(byte, 0x81..=0x9f | 0xe0..=0xfc) {
                self.pos += name::shift_jis_width(self.src, self.pos)
                    .ok_or(Fail::Invalid(self.pos..(self.pos + 2).min(self.src.len())))?;
                continue;
            }
            match byte {
                b'"' | b'\'' => {
                    let end = self
                        .src
                        .get(self.pos + 1..)
                        .and_then(|rest| rest.iter().position(|b| *b == byte))
                        .ok_or(Fail::Invalid(start..self.src.len()))?;
                    self.pos += end + 2;
                    continue;
                }
                b'<' if self.starts(b"<!--") => {
                    self.skip_to(b"-->", start)?;
                    continue;
                }
                b'[' => depth += 1,
                b']' => depth = depth.checked_sub(1).ok_or(Fail::Invalid(start..self.pos))?,
                b'>' if depth == 0 => break,
                _ => {}
            }
            self.pos += 1;
        }
        self.pos += 1;
        self.doctype = Some(start..self.pos);
        cx.report(Diagnostic::new(codes::XML_DTD_DROPPED).at(self.place(start..self.pos)));
        Ok(())
    }

    /// 名前。`:` で接頭辞と分け、接頭辞を除いた名前の始まりを返す
    fn name(&mut self, cx: &mut Cx) -> Result<(Range<usize>, usize), Fail> {
        let start = self.pos;
        while self.peek().is_some_and(|b| !is_delimiter(b)) {
            self.pos += if self.shift_jis {
                name::shift_jis_width(self.src, self.pos)
                    .ok_or(Fail::Invalid(self.pos..(self.pos + 2).min(self.src.len())))?
            } else {
                1
            };
        }
        let raw = start..self.pos;
        let bytes = self.bytes(raw.clone());
        if self.mapping.is_some() {
            self.work(cx, raw.len() as u64, start)?;
            if !utf16::valid_name(bytes) {
                return Err(Fail::Invalid(raw));
            }
        }
        if self.shift_jis && !name::valid_shift_jis(bytes) {
            return Err(Fail::Invalid(raw));
        }
        let first = bytes.first().copied();
        if first.is_none_or(|b| b.is_ascii_digit() || matches!(b, b'-' | b'.' | b':')) {
            return Err(Fail::Invalid(start..self.pos.max(start + 1)));
        }
        let colons: Vec<usize> = bytes
            .iter()
            .enumerate()
            .filter_map(|(at, b)| (*b == b':').then_some(at))
            .collect();
        match colons.as_slice() {
            [] => Ok((raw, start)),
            [colon] if start + colon + 1 < self.pos => Ok((raw, start + colon + 1)),
            _ => Err(Fail::Invalid(raw)),
        }
    }

    /// 文字か属性の値の参照を確かめる。DTD で宣言する実体の参照は、報告して残す。`&` と `;` は `range` の中だけで
    /// 探すので、手間は範囲の長さに比例する(#721)。見たバイトの数を返す
    fn references(&self, range: Range<usize>, cx: &mut Cx) -> Result<usize, Fail> {
        let mut seen = 0;
        let mut at = range.start;
        while let Some(amp) = find_in(self.src, b'&', at..range.end, &mut seen) {
            let end = find_in(self.src, b';', amp + 1..range.end, &mut seen)
                .ok_or(Fail::Invalid(amp..range.end))?;
            let name = core::str::from_utf8(self.bytes(amp + 1..end))
                .map_err(|_| Fail::Invalid(amp..end + 1))?;
            if text::reference(name).is_none() {
                let valid = name
                    .bytes()
                    .next()
                    .is_some_and(|b| !b.is_ascii_digit() && !matches!(b, b'#' | b'-' | b'.'))
                    && name.bytes().all(|b| !is_delimiter(b));
                if !valid {
                    return Err(Fail::Invalid(amp..end + 1));
                }
                cx.report(
                    Diagnostic::new(codes::XML_ENTITY_UNSUPPORTED)
                        .with("name", name)
                        .at(self.place(amp..end + 1)),
                );
            }
            at = end + 1;
        }
        Ok(seen)
    }

    /// 接頭辞の名前空間。`element` なら、接頭辞のない名前は既定の名前空間に入る
    fn resolve(
        &self,
        raw: &Range<usize>,
        local: usize,
        element: bool,
        cx: &mut Cx,
    ) -> Option<Namespace> {
        if local == raw.start {
            if element {
                return self
                    .declared(None)
                    .and_then(|value| value.clone().map(Namespace::Declared));
            }
            return (self.bytes(raw.clone()) == b"xmlns").then_some(Namespace::Xmlns);
        }
        let prefix = self.bytes(raw.start..local - 1);
        match prefix {
            b"xml" => return Some(Namespace::Xml),
            b"xmlns" => return Some(Namespace::Xmlns),
            _ => {}
        }
        match self.declared(Some(prefix)) {
            Some(value) => value.clone().map(Namespace::Declared),
            None => {
                cx.report(
                    Diagnostic::new(codes::XML_NAMESPACE_MISSING)
                        .with(
                            "prefix",
                            if self.shift_jis {
                                encoding_rs::SHIFT_JIS
                                    .decode_without_bom_handling(prefix)
                                    .0
                                    .into_owned()
                            } else {
                                String::from_utf8_lossy(prefix).into_owned()
                            },
                        )
                        .at(self.place(raw.clone())),
                );
                None
            }
        }
    }

    /// ルートの要素と、その中を読む
    fn tree(&mut self, cx: &mut Cx) -> Result<(), Fail> {
        let mut open = Vec::new();
        if self.open(None, cx)? {
            open.push(0);
        }
        while let Some(&current) = open.last() {
            let start = self.pos;
            self.work(cx, 1, start)?;
            let Some(byte) = self.peek() else {
                let from = self.elements.get(current).map_or(start, |e| e.span.start);
                return Err(Fail::Invalid(from..start));
            };
            if byte != b'<' {
                let end = find(self.src, b"<", start).unwrap_or(self.src.len());
                self.pos = end;
                self.references(start..end, cx)?;
                self.child(current, Node::Text(start..end));
            } else if self.starts(b"</") {
                self.pos += 2;
                let (raw, _) = self.name(cx)?;
                self.skip_space();
                let expected = self
                    .elements
                    .get(current)
                    .map(|e| e.name.raw.clone())
                    .unwrap_or_default();
                if self.peek() != Some(b'>') || self.bytes(raw) != self.bytes(expected) {
                    return Err(Fail::Invalid(start..self.pos + 1));
                }
                self.pos += 1;
                if let Some(element) = self.elements.get_mut(current) {
                    element.span.end = self.pos;
                }
                if let Some(mark) = self.marks.pop() {
                    self.unwind(mark);
                }
                cx.leave();
                open.pop();
            } else if self.starts(b"<!--") {
                self.skip_to(b"-->", start)?;
            } else if self.starts(b"<![CDATA[") {
                let inner = start + b"<![CDATA[".len();
                let end =
                    find(self.src, b"]]>", inner).ok_or(Fail::Invalid(start..self.src.len()))?;
                self.pos = end + 3;
                self.child(current, Node::CData(inner..end));
            } else if self.starts(b"<?") {
                self.skip_to(b"?>", start)?;
            } else if self.starts(b"<!") {
                return Err(Fail::Invalid(start..start + 2));
            } else {
                let index = self.elements.len();
                if self.open(Some(current), cx)? {
                    open.push(index);
                }
            }
        }
        Ok(())
    }

    fn child(&mut self, parent: usize, node: Node) {
        if let Some(parent) = self.elements.get_mut(parent) {
            parent.children.push(node);
        }
    }

    /// 開始タグか空要素のタグを読み、要素を足す。中身を持つ要素なら真
    fn open(&mut self, parent: Option<usize>, cx: &mut Cx) -> Result<bool, Fail> {
        let start = self.pos;
        self.pos += 1;
        let (raw, local) = self.name(cx)?;
        let mut pairs: Vec<(Range<usize>, usize, Range<usize>)> = Vec::new();
        let empty = loop {
            let spaced = self.skip_space();
            match self.peek() {
                Some(b'>') => {
                    self.pos += 1;
                    break false;
                }
                Some(b'/') if self.starts(b"/>") => {
                    self.pos += 2;
                    break true;
                }
                Some(_) if spaced => {
                    let (name, name_local) = self.name(cx)?;
                    self.skip_space();
                    if self.peek() != Some(b'=') {
                        return Err(Fail::Invalid(name.start..self.pos + 1));
                    }
                    self.pos += 1;
                    self.skip_space();
                    let quote = self
                        .peek()
                        .filter(|b| matches!(b, b'"' | b'\''))
                        .ok_or(Fail::Invalid(name.start..self.pos + 1))?;
                    let value_start = self.pos + 1;
                    let value_end = self
                        .src
                        .get(value_start..)
                        .and_then(|rest| rest.iter().position(|b| *b == quote))
                        .map(|end| value_start + end)
                        .ok_or(Fail::Invalid(name.start..self.src.len()))?;
                    let value = value_start..value_end;
                    if self.bytes(value.clone()).contains(&b'<') {
                        return Err(Fail::Invalid(name.start..value_end + 1));
                    }
                    self.pos = value_end + 1;
                    self.references(value.clone(), cx)?;
                    pairs.push((name, name_local, value));
                }
                _ => return Err(Fail::Invalid(start..self.pos.max(start + 1))),
            }
        };
        cx.check(Kind::Records, 1)
            .map_err(|stop| Fail::Stop(stop, start..self.pos))?;
        self.work(cx, 1 + pairs.len() as u64, start)?;
        cx.enter()
            .map_err(|stop| Fail::Stop(stop, start..self.pos))?;
        let mut seen = BTreeSet::new();
        for (name, _, _) in &pairs {
            if !seen.insert(self.bytes(name.clone())) {
                return Err(Fail::Invalid(name.clone()));
            }
        }
        let mark = self.scopes.len();
        for (name, _, value) in &pairs {
            let bytes = self.bytes(name.clone());
            let declared = (!value.is_empty()).then(|| value.clone());
            if bytes == b"xmlns" {
                self.declare(None, declared);
            } else if let Some(prefix) = bytes.strip_prefix(b"xmlns:") {
                if matches!(prefix, b"xml" | b"xmlns") {
                    continue;
                }
                if declared.is_none() {
                    return Err(Fail::Invalid(name.clone()));
                }
                self.declare(Some(prefix), declared);
            }
        }
        let namespace = self.resolve(&raw, local, true, cx);
        let attributes = pairs
            .into_iter()
            .map(|(name, name_local, value)| Attribute {
                name: Name {
                    prefix_end: (name_local > name.start).then_some(name_local.saturating_sub(1)),
                    namespace: self.resolve(&name, name_local, false, cx),
                    raw: name,
                    local: name_local,
                },
                value,
            })
            .collect();
        let index = self.elements.len();
        self.elements.push(Element {
            span: start..self.pos,
            name: Name {
                prefix_end: (local > raw.start).then_some(local.saturating_sub(1)),
                raw,
                local,
                namespace,
            },
            attributes,
            children: Vec::new(),
            parent,
        });
        if let Some(parent) = parent.and_then(|parent| self.elements.get_mut(parent)) {
            parent.children.push(Node::Element(index));
        }
        if empty {
            self.unwind(mark);
            cx.leave();
        } else {
            self.marks.push(mark);
        }
        Ok(!empty)
    }
}

#[cfg(test)]
mod tests {
    use super::*;
    use zumen_base::{Budget, CancelToken, Diagnostics};

    /// 参照は範囲の中だけで探し、文書の残りを見ない。小さな範囲を多く確かめても、見たバイトの数は範囲の長さの
    /// 和を超えない(#721)
    #[test]
    fn references_look_only_inside_the_range() {
        let mut src = b"<a>".to_vec();
        src.extend(std::iter::repeat_n(b'x', 100_000));
        src.extend(b"&amp;</a>");
        let parser = Parser::new(&src);
        let mut budget = Budget::default();
        let mut diag = Diagnostics::new();
        let cancel = CancelToken::new();
        let mut cx = Cx::new(&mut budget, &mut diag, &cancel);
        let mut seen = 0;
        let mut ranges = 0;
        for start in (3..100_003).step_by(10) {
            let Ok(found) = parser.references(start..start + 10, &mut cx) else {
                panic!("{start}");
            };
            seen += found;
            ranges += 10;
        }
        assert_eq!(seen, ranges);
        // 範囲の中の参照は、前と同じく読む
        let Ok(found) = parser.references(100_003..100_008, &mut cx) else {
            panic!("&amp;");
        };
        assert_eq!(found, 5);
        assert!(diag.is_empty());
    }

    /// 名前空間の宣言は接頭辞ごとに積み、閉じた要素の宣言を外す。読み終えたら、積みは空になる(#721)
    #[test]
    fn namespaces_are_stacked_by_prefix_and_unwound() {
        let src = b"<r xmlns:p='urn:r' xmlns='urn:d' xmlns:q='urn:q'>\
            <p:a xmlns:p='urn:a'><b xmlns=''/></p:a><c xmlns:s='urn:c'/></r>";
        let mut budget = Budget::default();
        let mut diag = Diagnostics::new();
        let cancel = CancelToken::new();
        let mut cx = Cx::new(&mut budget, &mut diag, &cancel);
        let mut parser = Parser::new(src);
        assert!(parser.encoding(&mut cx).is_ok());
        // ルートの開始タグのあとは、接頭辞ごとに 1 つずつ
        assert!(matches!(parser.open(None, &mut cx), Ok(true)));
        assert_eq!(parser.namespaces.len(), 3);
        assert!(parser.namespaces.values().all(|stack| stack.len() == 1));
        // `p:a` の中では、`p` の宣言が 2 つ積まれ、内側で引く
        assert!(matches!(parser.open(Some(0), &mut cx), Ok(true)));
        let p = Some(&b"p"[..]);
        assert_eq!(parser.namespaces.get(&p).map(Vec::len), Some(2));
        let inner = parser.declared(p).cloned().flatten();
        assert_eq!(inner.map(|value| parser.bytes(value)), Some(&b"urn:a"[..]));
        assert!(diag.is_empty());

        // 読み終えたら、すべて外れる
        let mut budget = Budget::default();
        let mut diag = Diagnostics::new();
        let mut cx = Cx::new(&mut budget, &mut diag, &cancel);
        let mut parser = Parser::new(src);
        assert!(parser.encoding(&mut cx).is_ok());
        assert!(parser.document(&mut cx).is_ok());
        assert!(parser.namespaces.is_empty());
        assert!(parser.scopes.is_empty());
        assert!(parser.marks.is_empty());
        assert!(diag.is_empty());
    }
}