asciidoc_parser/document/header.rs
1use crate::{
2 HasSpan, Parser, Span,
3 attributes::{Attrlist, AttrlistContext},
4 blocks::metadata::block_title_text,
5 content::{Content, SubstitutionGroup, substitute_attributes_in_reftext},
6 document::{
7 Attribute, Author, AuthorLine, InterpretedValue, RefType, RevisionLine,
8 is_attribute_entry_pass_macro, matches_author_pattern, set_author_metadata,
9 },
10 internal::{debug::DebugSliceReference, opaque_iter::opaque_slice_iter},
11 span::MatchedItem,
12 warnings::{MatchAndWarnings, Warning, WarningType},
13};
14
15opaque_slice_iter! {
16 /// An iterator over the document attributes declared in a [`Header`],
17 /// returned by [`Header::attributes`].
18 pub struct HeaderAttributes<'a> yielding Attribute<'a>;
19}
20
21opaque_slice_iter! {
22 /// An iterator over the comment lines in a [`Header`], returned by
23 /// [`Header::comments`].
24 pub struct Comments<'a> yielding Span<'a>;
25}
26
27/// An AsciiDoc document may begin with a document header. The document header
28/// encapsulates the document title, author and revision information,
29/// document-wide attributes, and other document metadata.
30#[derive(Clone, Eq, PartialEq)]
31pub struct Header<'src> {
32 title_source: Option<Span<'src>>,
33 title: Option<String>,
34 doctitle: Option<String>,
35 main_title: Option<String>,
36 subtitle: Option<String>,
37 id: Option<String>,
38 roles: Vec<String>,
39 attributes: Vec<Attribute<'src>>,
40 author_line: Option<AuthorLine<'src>>,
41 authors: Vec<Author>,
42 revision_line: Option<RevisionLine<'src>>,
43 comments: Vec<Span<'src>>,
44 source: Span<'src>,
45}
46
47impl<'src> Header<'src> {
48 pub(crate) fn parse(
49 mut source: Span<'src>,
50 parser: &mut Parser,
51 ) -> MatchAndWarnings<'src, MatchedItem<'src, Self>> {
52 let original_source = source.discard_empty_lines();
53
54 let mut title_source: Option<Span<'src>> = None;
55 let mut title: Option<String> = None;
56
57 // State that mirrors Asciidoctor's `parse_document_header` doctitle
58 // handling: whether an implicit `= Title` line was seen, the eager
59 // (at-title-line) substitution stored in the `doctitle` attribute, and
60 // whether a `:doctitle:` attribute entry appeared below the title (a
61 // candidate to override the section title).
62 let mut saw_implicit_title = false;
63 let mut implicit_overridden_from_above = false;
64 let mut implicit_doctitle_str: Option<String> = None;
65 let mut doctitle_entry_after_title = false;
66
67 let mut id: Option<String> = None;
68 let mut roles: Vec<String> = vec![];
69 let mut pending_block_title_source: Option<Span<'src>> = None;
70 let mut attributes: Vec<Attribute> = vec![];
71 let mut author_line: Option<AuthorLine<'src>> = None;
72 let mut author_attribute: Option<Author> = None;
73 let mut authorinitials_from_entry = false;
74 let mut revision_line: Option<RevisionLine<'src>> = None;
75 let mut comments: Vec<Span<'src>> = vec![];
76 let mut warnings: Vec<Warning<'src>> = vec![];
77
78 // Aside from the title line, items can appear in almost any order.
79 while !source.is_empty() {
80 let line_mi = source.take_normalized_line();
81 let line = line_mi.item;
82
83 // A blank line after the title ends the header.
84 if line.is_empty() {
85 if title.is_some() {
86 break;
87 }
88 source = line_mi.after;
89 } else if line.starts_with("//") && !line.starts_with("///") {
90 comments.push(line);
91 source = line_mi.after;
92 } else if title.is_some()
93 && let Some((after, terminated)) = skip_block_comment(line, line_mi.after)
94 {
95 // Once a title has been seen, a `////` block comment delimiter
96 // opens a comment block within the header. Skip every line
97 // through the matching closing delimiter (or to the end of the
98 // input if the block is never closed), retaining the whole block
99 // as a single comment so it is not mistaken for the author or
100 // revision line. Blank lines inside the block do not terminate
101 // the header.
102 //
103 // Before a title is seen there is no header author/revision
104 // context to protect, so a leading `////` is left for the block
105 // parser, which retains it as a body-level comment block.
106 comments.push(source.trim_remainder(after).trim_trailing_line_end());
107
108 // An unterminated comment block swallows the rest of the header
109 // (any following attribute entries are never applied), so warn
110 // as Asciidoctor does, anchoring the warning at the opening
111 // delimiter. This mirrors the body-level comment block path,
112 // which reports the same `UnterminatedDelimitedBlock` warning.
113 if !terminated {
114 warnings.push(Warning {
115 source: line,
116 warning: WarningType::UnterminatedDelimitedBlock,
117 origin: None,
118 });
119 }
120
121 source = after;
122 } else if line.starts_with(':')
123 && let Some(attr) = Attribute::parse(source, parser)
124 {
125 // Track an explicit `:authorinitials:` entry so a `:author:`
126 // entry (whether it precedes or follows this one) does not
127 // overwrite it with initials re-derived from the author's name.
128 // Asciidoctor preserves an explicit `authorinitials` for a
129 // single `author`; an empty value (`:authorinitials:`) still
130 // counts as explicit, only an unset (`:authorinitials!:`) does
131 // not.
132 if attr
133 .item
134 .name()
135 .data()
136 .eq_ignore_ascii_case("authorinitials")
137 {
138 authorinitials_from_entry =
139 !matches!(attr.item.value(), InterpretedValue::Unset);
140 }
141
142 // Special handling for :author: attribute to populate individual author
143 // attributes.
144 //
145 // When the value is a plain name, the partitioned name replaces
146 // the stored `author` value. This condenses repeated interior
147 // whitespace and joins a name with four or more parts, matching
148 // Asciidoctor.
149 //
150 // Asciidoctor runs an attribute-entry value through the
151 // substitution pipeline *before* partitioning it into name parts
152 // (`process_authors`, `names_only`). When that substitution
153 // produced inline HTML – for example a `pass:[…]` macro whose
154 // content resolves to a link and inline formatting – the rendered
155 // markup is stripped before partitioning so it does not leak into
156 // `firstname`/`middlename`/`lastname`, while the `author` value
157 // keeps the rendered markup. A value that produced no markup
158 // keeps the original raw-value partitioning, so every other form
159 // is unchanged.
160 let mut author_name_override: Option<String> = None;
161 if attr.item.name().data().eq_ignore_ascii_case("author")
162 && let Some(raw_value) = attr.item.raw_value()
163 && let Some(author) = Author::parse_from_entry(
164 raw_value.data(),
165 attr.item.value().as_maybe_str(),
166 parser,
167 )
168 {
169 // Set individual author attributes.
170 parser.set_attribute_by_value_from_header("firstname", author.firstname());
171 if let Some(middlename) = author.middlename() {
172 parser.set_attribute_by_value_from_header("middlename", middlename);
173 }
174 if let Some(lastname) = author.lastname() {
175 parser.set_attribute_by_value_from_header("lastname", lastname);
176 }
177
178 // Do not re-derive `authorinitials` when the document has
179 // supplied its own via an explicit entry (see above).
180 if !authorinitials_from_entry {
181 parser.set_attribute_by_value_from_header(
182 "authorinitials",
183 author.initials(),
184 );
185 }
186
187 if let Some(email) = author.email() {
188 parser.set_attribute_by_value_from_header("email", email);
189 }
190
191 // Override the stored `author` value with the reconstructed
192 // name when the value was partitioned rather than kept
193 // verbatim. A resolved whole-value `pass:[…]` macro is
194 // partitioned from its substituted value, so the
195 // reconstructed name (the rendered markup with name-joiner
196 // underscores turned to spaces) replaces the stored value.
197 // Otherwise the value is overridden only for a plain name
198 // that was partitioned by the fallback whitespace split (four
199 // or more parts, or punctuation such as a comma); a value that
200 // matches the pattern, carries an inline email (`<…>`), or
201 // holds an attribute reference (`{…}`) keeps the substituted
202 // entry value set below.
203 let raw = raw_value.data();
204
205 if is_attribute_entry_pass_macro(raw)
206 || (!raw.contains('<')
207 && !raw.contains('{')
208 && !matches_author_pattern(raw))
209 {
210 author_name_override = Some(author.name().to_string());
211 }
212
213 // Retain the author parsed from the entry value so the
214 // resolved author list does not have to re-parse the stored
215 // `author` attribute. A later `:author:` entry overrides an
216 // earlier one.
217 author_attribute = Some(author);
218 }
219
220 parser.set_attribute_from_header(&attr.item, &mut warnings);
221
222 if let Some(author_name) = author_name_override {
223 parser.set_attribute_by_value_from_header("author", author_name);
224 }
225
226 // A `:doctitle:` entry below the document title is a candidate to
227 // override the implicit section title (resolved after the header
228 // is fully parsed; see below).
229 if title.is_some() && attr.item.name().data().eq_ignore_ascii_case("doctitle") {
230 doctitle_entry_after_title = true;
231 }
232
233 attributes.push(attr.item);
234 source = attr.after;
235 } else if title.is_none()
236 && line.starts_with('[')
237 && line.ends_with(']')
238 && document_title_follows_block_metadata(source, parser.level_offset())
239 && let Some((metadata, metadata_warnings)) = parse_document_metadata(line, parser)
240 {
241 warnings.extend(metadata_warnings);
242
243 // A block attribute line directly above the document title assigns
244 // metadata to the *document* – its `id`, `reftext`, `role`, and
245 // options – mirroring Asciidoctor's `parse_document_header`. Each
246 // recognized value folds into the document's attributes at this
247 // point in the header, so it follows document order alongside any
248 // equivalent header attribute entry (e.g. `:reftext:`).
249 //
250 // A `separator` sets the subtitle separator; it behaves exactly
251 // like assigning the `title-separator` document attribute here, so
252 // both mechanisms share the same partitioning logic.
253 //
254 // The line is only intercepted when a document title eventually
255 // follows – possibly after further stacked block attribute lines,
256 // each folded on its own pass through this loop, mirroring
257 // Asciidoctor's `parse_block_metadata_lines`. Otherwise it is
258 // block metadata for the body (e.g. a table's `separator`) and is
259 // left for the block parser.
260 if let Some(doc_id) = metadata.id {
261 id = Some(doc_id);
262 }
263 if let Some(separator) = metadata.separator {
264 parser.set_attribute_by_value_from_header("title-separator", separator);
265 }
266 if let Some(reftext) = metadata.reftext {
267 parser.set_attribute_by_value_from_header("reftext", reftext);
268 }
269 if !metadata.roles.is_empty() {
270 // Fold the role(s) into the `role` document attribute
271 // (space-joined, as Asciidoctor stores `attributes['role']`)
272 // and also retain them on the header so `Document::roles()`
273 // agrees with the document attribute (see #820). Roles from
274 // separate stacked block attribute lines accumulate, just as
275 // multiple roles within a single line combine (see #821).
276 roles.extend(metadata.roles);
277 parser.set_attribute_by_value_from_header("role", roles.join(" "));
278 }
279 for option in metadata.options {
280 parser.set_attribute_by_value_from_header(format!("{option}-option"), "");
281 }
282 source = line_mi.after;
283 } else if title.is_none()
284 && let Some(block_title) = block_title_text(line)
285 && document_title_follows_block_metadata(source, parser.level_offset())
286 {
287 // A block title (`.Title`) directly above the document title is
288 // not a title *of* the document – a document has no block title.
289 // Asciidoctor demotes the following `= …` heading to a level-0
290 // section and carries the block title over to the first block of
291 // content. This crate does not represent a body-level document
292 // title as a level-0 section, so the `= …` heading remains the
293 // document title; the block title is still recognized here (so
294 // the heading is not mistaken for a paragraph) and carried over
295 // to the first body block, matching the block-title carryover
296 // above an ordinary section heading (see #782).
297 //
298 // The line is only intercepted when a document title eventually
299 // follows – possibly after further stacked block metadata lines,
300 // each folded on its own pass through this loop. A later block
301 // title overrides an earlier one (last-wins), mirroring
302 // Asciidoctor's `parse_block_metadata_lines`. Otherwise it is an
303 // ordinary block title for the body and is left for the block
304 // parser.
305 pending_block_title_source = Some(block_title);
306 source = line_mi.after;
307 } else if title.is_none()
308 && let Some((marker, count)) = document_title_marker(line, parser.level_offset())
309 {
310 // Strip an optional symmetric close (a trailing ` ==` or ` ##`
311 // matching the opening marker run), mirroring section titles.
312 let title_span = crate::blocks::strip_symmetric_title_close(
313 line.discard(count).discard_whitespace(),
314 marker,
315 count,
316 );
317 saw_implicit_title = true;
318
319 title_source = Some(title_span);
320
321 // A `doctitle` attribute already set above the title – via a
322 // `:doctitle:` entry or the API – overrides the implicit title:
323 // the implicit text is discarded, the existing doctitle stands
324 // as the document title, and the `doctitle` attribute is left
325 // untouched. Otherwise the implicit title is
326 // substituted now (so `{doctitle}` references below resolve to
327 // it) and recorded as the baseline for a later override check.
328 if let InterpretedValue::Value(existing) = parser.attribute_value("doctitle")
329 && !existing.is_empty()
330 {
331 implicit_overridden_from_above = true;
332 implicit_doctitle_str = Some(existing.clone());
333 title = Some(existing);
334 } else {
335 let title_str = apply_header_subs(title_span.data(), parser);
336
337 parser.set_attribute_by_value_from_header("doctitle", &title_str);
338
339 implicit_doctitle_str = Some(title_str.clone());
340 title = Some(title_str);
341 }
342
343 source = line_mi.after;
344 } else if title.is_some() && author_line.is_none() {
345 author_line = Some(AuthorLine::parse(line, parser));
346 source = line_mi.after;
347 } else if title.is_some() && author_line.is_some() && revision_line.is_none() {
348 revision_line = Some(RevisionLine::parse(line, parser));
349 source = line_mi.after;
350 } else {
351 if title.is_some() {
352 warnings.push(Warning {
353 source: line,
354 warning: WarningType::DocumentHeaderNotTerminated,
355 origin: None,
356 });
357 }
358 break;
359 }
360 }
361
362 let after = source.discard_empty_lines();
363 let source = original_source.trim_remainder(source);
364
365 // Finalize the document (section) title, mirroring Asciidoctor's
366 // `parse_document_header` doctitle handling. The `doctitle` attribute
367 // retains the eager, at-title-line substitution; the section title below
368 // is (re)derived from the *final* attribute state so that:
369 //
370 // - an implicit `= Title` referencing an attribute defined later in the
371 // header still resolves ("lazy" resolution),
372 // - a `:doctitle:` attribute entry (above or below the title, or in a
373 // document with no title line at all) can supply or override it.
374 let final_doctitle_attr = match parser.attribute_value("doctitle") {
375 InterpretedValue::Value(v) if !v.is_empty() => Some(v),
376 _ => None,
377 };
378
379 title = if saw_implicit_title {
380 // The base section title is normally the eager (at-title-line)
381 // substitution already held in `title`. It is re-resolved against the
382 // final attribute set only when that eager substitution left an
383 // unresolved attribute reference – an attribute defined later in the
384 // header – so that one-shot substitutions such as a `{counter:…}` in
385 // the title are not evaluated a second time. When the implicit title
386 // was overridden by a `doctitle` set above it, `title` already holds
387 // that (resolved) value and is not re-substituted.
388 //
389 // Residual edge: a title that *mixes* a counter with a later-defined
390 // reference (e.g. `= {counter:n} {project-name}`) still contains a
391 // `{` after the eager pass, so the re-resolution runs and advances the
392 // counter a second time. Re-resolving is done from the raw title (not
393 // the eager result) so that escaped `\{…}` and specialchars stay
394 // correct; the counter here is the price of that. This is a rare
395 // combination and no test exercises it.
396 let base = if !implicit_overridden_from_above
397 && let Some(raw) = title_source
398 && implicit_doctitle_str
399 .as_deref()
400 .is_some_and(|s| s.contains('{'))
401 {
402 Some(apply_header_subs(raw.data(), parser))
403 } else {
404 title
405 };
406
407 // A `:doctitle:` entry below the title overrides the section title
408 // when it sets a new, non-empty value (an empty or unchanged value
409 // leaves the implicit title in place).
410 if doctitle_entry_after_title
411 && let Some(ref dt) = final_doctitle_attr
412 && Some(dt) != implicit_doctitle_str.as_ref()
413 {
414 Some(dt.clone())
415 } else {
416 base
417 }
418 } else {
419 // No `= Title` line: a `:doctitle:` attribute entry, if any, supplies
420 // the implicit document title.
421 final_doctitle_attr
422 };
423
424 // Partition the (fully substituted) document title into a main title and
425 // an optional subtitle. This happens after the header has been fully
426 // parsed so that a `title-separator` attribute takes effect even when it
427 // is defined below the document title line.
428 let (main_title, subtitle) = match &title {
429 Some(title) => {
430 let (main_title, subtitle) = partition_title(title, parser);
431 (Some(main_title), subtitle)
432 }
433 None => (None, None),
434 };
435
436 // The value returned by `Document::doctitle()`: a `title` attribute entry
437 // overrides the section title (Asciidoctor's `Document#doctitle`), even
438 // when it is blank; otherwise the section title is the doctitle.
439 let doctitle = match parser.attribute_value("title") {
440 InterpretedValue::Value(v) => Some(v),
441 InterpretedValue::Set => Some(String::new()),
442 InterpretedValue::Unset => title.clone(),
443 };
444
445 // A document title carrying an explicit ID (`[#id]` above `= Title`)
446 // registers that ID in the catalog, mirroring Asciidoctor, which
447 // registers the document itself under its ID. Without this, a
448 // cross-reference to the document (`<<id>>`) finds no catalog entry and
449 // falls back to the bracketed `[id]` form instead of the title's
450 // reference text. The reference text follows the same precedence as a
451 // whole-document self-reference (see
452 // [`this_document_reference`](crate::content::this_document_reference)):
453 // an explicit `reftext` attribute, otherwise the document title. The
454 // header is parsed before the body, so this ID registers ahead of any
455 // body anchor; a later duplicate is ignored here and reported by the
456 // body parse.
457 if let Some(doc_id) = id.as_deref() {
458 let reftext = match parser.attribute_value("reftext") {
459 InterpretedValue::Value(reftext) if !reftext.is_empty() => Some(reftext),
460 _ => doctitle.clone().filter(|title| !title.is_empty()),
461 };
462
463 let _ = parser.register_ref(doc_id, reftext.as_deref(), RefType::Section);
464 }
465
466 // Resolve the document's author list. The author line, when present, is
467 // the source of truth; otherwise the list is derived from the `author`,
468 // `authors`, and indexed `author_N` document attributes (see
469 // [`resolve_authors`]). Those attributes can only be set by header
470 // attribute entries, so a header with none needs no reconciliation.
471 let authors = resolve_authors(
472 author_line.as_ref(),
473 author_attribute,
474 !attributes.is_empty(),
475 parser,
476 );
477
478 // Asciidoctor exposes the number of resolved authors via the
479 // `authorcount` document attribute. It defaults to `0` (a built-in
480 // default), so only a non-zero count is materialized here – this keeps an
481 // author-less parse from touching the attribute map at all.
482 if !authors.is_empty() {
483 parser.set_attribute_by_value_from_header("authorcount", authors.len().to_string());
484 }
485
486 // A block title recognized directly above the document title is carried
487 // over to the first block of the body. Stash it on the parser as a
488 // pending block title (the same mechanism a section heading uses); the
489 // first block parsed after the header claims it. The title travels as an
490 // owned snapshot with normal (title) substitutions already applied,
491 // matching a `.Title` line on an ordinary block. It is only ever set
492 // when a document title was seen, since the recognizing branch requires
493 // one to follow.
494 if let Some(block_title) = pending_block_title_source {
495 let mut content = Content::from(block_title);
496 SubstitutionGroup::Normal.apply(&mut content, parser, None);
497 parser.pending_block_title = Some(content.to_owned_title());
498 }
499
500 MatchAndWarnings {
501 item: MatchedItem {
502 item: Self {
503 title_source,
504 title,
505 doctitle,
506 main_title,
507 subtitle,
508 id,
509 roles,
510 attributes,
511 author_line,
512 authors,
513 revision_line,
514 comments,
515 source: source.trim_trailing_whitespace(),
516 },
517 after,
518 },
519 warnings,
520 }
521 }
522
523 /// Return a [`Span`] describing the raw document title, if there was one.
524 pub fn title_source(&'src self) -> Option<Span<'src>> {
525 self.title_source
526 }
527
528 /// Return the document's title, if there was one, having applied header
529 /// substitutions.
530 ///
531 /// If the title contains a subtitle (see [`subtitle`]), this returns the
532 /// full, combined title. Use [`main_title`] to obtain only the portion
533 /// preceding the subtitle.
534 ///
535 /// [`subtitle`]: Self::subtitle
536 /// [`main_title`]: Self::main_title
537 pub fn title(&self) -> Option<&str> {
538 self.title.as_deref()
539 }
540
541 /// Return the effective document title, applying the override precedence of
542 /// Asciidoctor's `Document#doctitle`: a `title` attribute entry (even a
543 /// blank one) takes priority over the section [`title`], which in turn may
544 /// have been supplied or overridden by a `:doctitle:` attribute entry.
545 ///
546 /// This backs [`Document::doctitle`] and can differ from [`title`]: for
547 /// `= Document Title` followed by `:title: Override`, [`title`] is
548 /// `Document Title` while this is `Override`.
549 ///
550 /// [`title`]: Self::title
551 /// [`Document::doctitle`]: crate::Document::doctitle
552 pub(crate) fn doctitle(&self) -> Option<&str> {
553 self.doctitle.as_deref()
554 }
555
556 /// Return the main portion of the document title, if there was a title.
557 ///
558 /// When the document title contains a subtitle separator (a colon followed
559 /// by a space, by default), the title is partitioned into a main title and
560 /// a [`subtitle`]. This returns the portion preceding the final separator.
561 /// When there is no subtitle, this is identical to [`title`].
562 ///
563 /// [`subtitle`]: Self::subtitle
564 /// [`title`]: Self::title
565 pub fn main_title(&self) -> Option<&str> {
566 self.main_title.as_deref()
567 }
568
569 /// Return the document's subtitle, if the title contained one.
570 ///
571 /// A subtitle is the text following the final subtitle separator in the
572 /// document title. The separator defaults to a colon followed by a space
573 /// (`:{sp}`) and can be overridden with the `title-separator` document
574 /// attribute. Returns `None` when the title has no subtitle.
575 pub fn subtitle(&self) -> Option<&str> {
576 self.subtitle.as_deref()
577 }
578
579 /// Return the document's ID, if one was assigned.
580 ///
581 /// A document ID is set with a block attribute line directly above the
582 /// document title, using either the shorthand (`[#id]`) or longhand
583 /// (`[id=id]`) syntax. Returns `None` when no such ID was given.
584 pub fn id(&self) -> Option<&str> {
585 self.id.as_deref()
586 }
587
588 /// Return the document's role(s), if any were assigned.
589 ///
590 /// Roles are set with a block attribute line directly above the document
591 /// title, using either the shorthand (`[.role]`) or longhand (`[role=…]`)
592 /// syntax; multiple roles combine. The same role(s) are also folded into
593 /// the `role` document attribute (space-joined), so the block accessor and
594 /// the document attribute agree. Returns an empty vector when no role was
595 /// given.
596 pub fn roles(&self) -> Vec<&str> {
597 self.roles.iter().map(String::as_str).collect()
598 }
599
600 /// Return an iterator over the attributes in this header.
601 pub fn attributes(&'src self) -> HeaderAttributes<'src> {
602 HeaderAttributes::new(&self.attributes)
603 }
604
605 /// Returns the author line, if found.
606 pub fn author_line(&self) -> Option<&AuthorLine<'src>> {
607 self.author_line.as_ref()
608 }
609
610 /// Returns the document's authors.
611 ///
612 /// Authors may be declared on the [author line] or via the `author` /
613 /// `author_N` (and companion `email_N`, …) document attributes; this
614 /// returns the resolved list regardless of which mechanism was used. When
615 /// the document has no author information, the slice is empty.
616 ///
617 /// [author line]: https://docs.asciidoctor.org/asciidoc/latest/document/author-line/
618 pub fn authors(&self) -> &[Author] {
619 &self.authors
620 }
621
622 /// Returns the revision line, if found.
623 pub fn revision_line(&self) -> Option<&RevisionLine<'src>> {
624 self.revision_line.as_ref()
625 }
626
627 /// Return an iterator over the comments in this header.
628 pub fn comments(&'src self) -> Comments<'src> {
629 Comments::new(&self.comments)
630 }
631}
632
633impl<'src> HasSpan<'src> for Header<'src> {
634 fn span(&self) -> Span<'src> {
635 self.source
636 }
637}
638
639/// If `line` opens a `////` block comment, consume the comment block and
640/// return the source position immediately after its closing delimiter together
641/// with a flag reporting whether a closing delimiter was found; otherwise
642/// return `None`.
643///
644/// A block comment delimiter is a line of four or more forward slashes and
645/// nothing else, matching Asciidoctor's comment-block delimiter (a line of
646/// exactly three slashes instead terminates the header and is handled by the
647/// caller). The closing delimiter must repeat the opening line exactly; when
648/// it is absent the block runs to the end of the input, mirroring
649/// Asciidoctor's `read_lines_until`, and the returned flag is `false` so the
650/// caller can warn that the comment block was never terminated.
651///
652/// `after` is the source immediately following `line` (the opening delimiter).
653fn skip_block_comment<'src>(line: Span<'src>, after: Span<'src>) -> Option<(Span<'src>, bool)> {
654 let delimiter = line.data();
655 if delimiter.len() < 4 || !delimiter.bytes().all(|b| b == b'/') {
656 return None;
657 }
658
659 let mut next = after;
660 let mut terminated = false;
661 while !next.is_empty() {
662 let line_mi = next.take_normalized_line();
663 next = line_mi.after;
664 if line_mi.item.data() == delimiter {
665 terminated = true;
666 break;
667 }
668 }
669
670 Some((next, terminated))
671}
672
673/// Returns the ATX marker character and its run length for a `line` that is a
674/// document title under the running `level_offset`, or `None` if the line is
675/// not a document title.
676///
677/// Both the AsciiDoc marker (`=`) and the Markdown-style marker (`#`) are
678/// accepted, mirroring the alternation at the head of Asciidoctor's
679/// section-title regex. A line is the document title when its *effective* level
680/// – the syntactic level (marker run length minus one) shifted by
681/// `level_offset` – is 0, mirroring Asciidoctor's `is_next_line_doctitle?`.
682/// With no offset in effect this is exactly a single `=`/`#` marker; a negative
683/// `:leveloffset:` lets a deeper heading (`==` under `-1`) coerce to the
684/// document title, and a positive offset stops a bare `=` from being one.
685///
686/// The marker character and run length are returned so the caller can strip the
687/// markers and require a symmetric close of the same character and width.
688fn document_title_marker(line: Span<'_>, level_offset: i32) -> Option<(char, usize)> {
689 let data = line.data();
690
691 let marker = if data.starts_with('=') {
692 '='
693 } else if data.starts_with('#') {
694 '#'
695 } else {
696 return None;
697 };
698
699 // Count the leading marker run; a run longer than six is not a heading at
700 // all (`======` is the deepest section title).
701 let count = data.chars().take_while(|&c| c == marker).count();
702 if count > 6 {
703 return None;
704 }
705
706 // The marker run must be followed by a blank – a space or a tab – matching
707 // the section parser's `take_required_whitespace` and Asciidoctor's
708 // `[ \t]+` in the section-title regex. Accepting a tab (not just a space)
709 // keeps a tab-delimited heading shifted to effective level 0 on the same
710 // doctitle path as its space-delimited form.
711 if !data[count..].starts_with([' ', '\t']) {
712 return None;
713 }
714
715 // The effective level – the syntactic level (run length minus one) shifted
716 // by the running `leveloffset` – must be 0 for the line to be the document
717 // title.
718 let syntactic_level = (count as i32) - 1;
719 if syntactic_level.saturating_add(level_offset) != 0 {
720 return None;
721 }
722
723 Some((marker, count))
724}
725
726/// Reports whether a *promotable* document title follows the block metadata run
727/// beginning at `source` (the current bracket-delimited line, included in the
728/// scan).
729///
730/// Consecutive document-metadata block attribute lines (see
731/// [`is_document_metadata_line`]) and block title lines (`.Title`, see
732/// [`block_title_text`]) are consumed, and the first line that is neither is
733/// tested for a document title marker. This generalizes the original
734/// single-line lookahead so that stacked metadata lines above the title are all
735/// folded, mirroring Asciidoctor's `parse_block_metadata_lines`.
736///
737/// A line that starts with `[` and ends with `]` but is *not* a valid
738/// document-metadata line (e.g. a `[[anchor]]` block anchor or a leading-space
739/// form) stops the scan without matching, so the run of foldable lines is only
740/// ever a contiguous prefix of well-formed metadata lines terminated by the
741/// title.
742///
743/// The run's *effective* block style also gates promotion: if it resolves to
744/// `discrete`/`float`, the following level-0 (`=`) heading is a discrete
745/// floating title rather than the document title, so this returns `false` and
746/// neither the metadata nor the heading is folded here – both are left for the
747/// block parser, which produces a `SectionType::Discrete` heading (see #1014).
748/// The effective style is tracked with last-wins semantics that mirror
749/// `Attrlist::merge_block_attribute_line` / `merge_block_style_shorthand` – a
750/// line that specifies a block style overrides the running one, and a line with
751/// none leaves it unchanged – so the header's decision always agrees with the
752/// `BlockMetadata::is_discrete` decision the block parser would make on the
753/// same run.
754fn document_title_follows_block_metadata(source: Span<'_>, level_offset: i32) -> bool {
755 let mut next = source;
756 let mut effective_style_is_discrete = false;
757
758 while !next.is_empty() {
759 let line_mi = next.take_normalized_line();
760 let line = line_mi.item;
761
762 if document_title_marker(line, level_offset).is_some() {
763 return !effective_style_is_discrete;
764 }
765
766 // A block title (`.Title`) may appear anywhere in the metadata run above
767 // the document title; it carries no block style, so it leaves the
768 // running effective style unchanged and the scan continues.
769 if block_title_text(line).is_some() {
770 next = line_mi.after;
771 continue;
772 }
773
774 if !is_document_metadata_line(line) {
775 return false;
776 }
777
778 // Fold this line's block style into the running effective style
779 // (last-wins), leaving it unchanged when the line specifies none.
780 if let Some(is_discrete) = metadata_line_block_style_is_discrete(line) {
781 effective_style_is_discrete = is_discrete;
782 }
783
784 next = line_mi.after;
785 }
786
787 false
788}
789
790/// Classifies the block style of a single bracket-delimited document-metadata
791/// `line`, read structurally (the attribute list is *not* parsed):
792///
793/// * `None` – the line specifies no block style (a `[[id]]` anchor, a shorthand
794/// line whose first positional leads with `.`/`#`/`%`, or a named-first
795/// attribute list such as `[reftext=…]`), so it leaves a running style
796/// unchanged.
797/// * `Some(true)` – the block style is `discrete` or `float`.
798/// * `Some(false)` – the block style is some other value (e.g. `[appendix]`).
799///
800/// Reading the style off the raw text – rather than parsing the attribute list
801/// – keeps an embedded `{counter:…}` in a *value* from being evaluated at
802/// header time (the invariant the `rejected_metadata_run_does_not_fire_counter`
803/// test guards). This is safe because a block style is always the first
804/// positional attribute's leading shorthand token – a bare name that can never
805/// itself be a counter.
806fn metadata_line_block_style_is_discrete(line: Span<'_>) -> Option<bool> {
807 // Drop the enclosing square brackets (the caller has confirmed they are
808 // present via [`is_document_metadata_line`]).
809 let inner = line.slice(1..line.len() - 1).data();
810
811 // A `[[id]]` / `[[id,reftext]]` block anchor still has its inner brackets
812 // and carries no block style.
813 if inner.starts_with('[') {
814 return None;
815 }
816
817 // The block style is the leading shorthand token of the first positional
818 // attribute: a run of name characters terminated by a shorthand delimiter
819 // (`.`, `#`, `%`), a comma, whitespace, or the end of the attribute.
820 let token_len = inner
821 .find(|c: char| !(c.is_ascii_alphanumeric() || c == '-' || c == '_'))
822 .unwrap_or(inner.len());
823
824 // A first positional that leads with a shorthand delimiter (`[#id]`,
825 // `[.role]`, `[%option]`), or a `=` that makes the token a named attribute
826 // name (`[reftext=…]`), carries no block style.
827 if token_len == 0 || inner[token_len..].starts_with('=') {
828 return None;
829 }
830
831 let token = &inner[..token_len];
832 Some(token == "discrete" || token == "float")
833}
834
835/// Reports whether `line` is a block attribute line that this crate folds into
836/// document metadata when it appears above the document title.
837///
838/// This captures the purely syntactic acceptance rules shared by the lookahead
839/// ([`document_title_follows_block_metadata`]) and the folding step
840/// ([`parse_document_metadata`]): the line must be bracket-delimited and its
841/// contents must not be empty and must not begin with whitespace. Both a block
842/// attribute list (`[#id]`, `[reftext=…]`, …) and a `[[id]]` / `[[id,reftext]]`
843/// block anchor are accepted; an empty `[[]]` anchor is not.
844fn is_document_metadata_line(line: Span<'_>) -> bool {
845 if !(line.starts_with('[') && line.ends_with(']')) {
846 return false;
847 }
848
849 let inner = line.slice(1..line.len() - 1);
850
851 if inner.is_empty() || inner.starts_with(' ') || inner.starts_with('\t') {
852 return false;
853 }
854
855 // A `[[anchor]]` block anchor is document metadata when it names a non-empty
856 // anchor; the empty `[[]]` form is not (`inner` would be the two-character
857 // `[]`).
858 if inner.starts_with('[') && inner.ends_with(']') {
859 return inner.len() > 2;
860 }
861
862 true
863}
864
865/// Document metadata folded from a block attribute line appearing directly
866/// above the document title.
867///
868/// Each field holds an already-owned copy of a recognized value, so the caller
869/// can apply them without borrowing the (dropped) attribute list.
870struct DocumentMetadata {
871 id: Option<String>,
872 separator: Option<String>,
873 reftext: Option<String>,
874 roles: Vec<String>,
875 options: Vec<String>,
876}
877
878/// Parse a metadata line appearing directly above the document title into
879/// [`DocumentMetadata`], dispatching on its form: a `[[id]]` / `[[id,reftext]]`
880/// block anchor, or a block attribute list (e.g. `[reftext="…"]`, `[#id]`,
881/// `[role=…]`, `[separator=::]`).
882///
883/// The `line` is expected to begin with `[` and end with `]`. Returns the
884/// folded metadata together with any warnings raised while parsing when the
885/// line is a well-formed metadata line, and `None` otherwise (so the caller can
886/// fall through to its normal handling of the line, which then terminates the
887/// header). The warnings are only surfaced when the line is actually consumed
888/// as document metadata; otherwise the line is left for the block parser, which
889/// reports them on its own path.
890fn parse_document_metadata<'src>(
891 line: Span<'src>,
892 parser: &Parser,
893) -> Option<(DocumentMetadata, Vec<Warning<'src>>)> {
894 // Reject forms that are not document metadata (a leading space or tab, an
895 // empty list, or an empty `[[]]` anchor); see [`is_document_metadata_line`]
896 // for the shared acceptance rules. The caller has already confirmed the
897 // enclosing square brackets are present.
898 if !is_document_metadata_line(line) {
899 return None;
900 }
901
902 // Drop the enclosing square brackets.
903 let inner = line.slice(1..line.len() - 1);
904
905 // A `[[id]]` / `[[id,reftext]]` block anchor still has its inner brackets.
906 if inner.starts_with('[') && inner.ends_with(']') {
907 return parse_document_metadata_anchor(inner.slice(1..inner.len() - 1), parser);
908 }
909
910 let MatchAndWarnings {
911 item: MatchedItem {
912 item: attrlist,
913 after: _,
914 },
915 warnings,
916 } = Attrlist::parse(inner, parser, AttrlistContext::Block);
917
918 let metadata = DocumentMetadata {
919 id: attrlist.id().map(str::to_string),
920 separator: attrlist
921 .named_attribute("separator")
922 .map(|attr| attr.value().to_string()),
923 reftext: attrlist
924 .named_attribute("reftext")
925 .map(|attr| attr.value().to_string()),
926 roles: attrlist.roles().iter().map(|r| r.to_string()).collect(),
927 options: attrlist.options().iter().map(|o| o.to_string()).collect(),
928 };
929
930 Some((metadata, warnings))
931}
932
933/// Fold a `[[id]]` / `[[id,reftext]]` block anchor above the document title
934/// into [`DocumentMetadata`]. `anchor` is the text *between* the inner brackets
935/// (`id` or `id,reftext`).
936///
937/// The anchor ID must be a valid XML name (as the block parser requires of any
938/// block anchor); otherwise `None` is returned so the line falls through and
939/// ends the header. Attribute references in the reftext are resolved against
940/// the attributes in effect at the anchor, mirroring the section/block anchor
941/// path (see [`substitute_attributes_in_reftext`]) rather than the doctitle's
942/// `SpecialCharacters`-plus-references header substitution.
943fn parse_document_metadata_anchor<'src>(
944 anchor: Span<'src>,
945 parser: &Parser,
946) -> Option<(DocumentMetadata, Vec<Warning<'src>>)> {
947 // Split an optional reftext off at the first comma (`id,reftext`). A comma
948 // in the final position leaves the whole span – trailing comma included – as
949 // the ID, which then fails XML-name validation, matching the block parser.
950 let (id, reftext) = match anchor.position(|c| c == ',') {
951 Some(comma) if comma < anchor.len() - 1 => (
952 anchor.slice(0..comma),
953 Some(substitute_attributes_in_reftext(
954 anchor.slice(comma + 1..anchor.len()),
955 parser,
956 )),
957 ),
958 _ => (anchor, None),
959 };
960
961 if !id.is_xml_name() {
962 return None;
963 }
964
965 let metadata = DocumentMetadata {
966 id: Some(id.data().to_string()),
967 separator: None,
968 reftext: reftext.map(|r| r.to_string()),
969 roles: vec![],
970 options: vec![],
971 };
972
973 Some((metadata, vec![]))
974}
975
976/// Partition a document title into its main title and optional subtitle.
977///
978/// The separator is the value of the `title-separator` document attribute
979/// (defaulting to `:`) with a single space appended. The separator is searched
980/// for from the end of the title, so only the last occurrence partitions the
981/// title. When the separator is not present, the entire title is the main
982/// title and there is no subtitle.
983fn partition_title(title: &str, parser: &Parser) -> (String, Option<String>) {
984 // Read the configured `title-separator` document attribute directly. Unlike
985 // `Parser::attribute_value`, this bypasses the counter overlay: the title
986 // separator is a configuration attribute, never a counter, and Asciidoctor
987 // likewise resolves it with a plain attribute lookup.
988 let separator = match parser.effective_attribute("title-separator") {
989 Some(av) => match &av.value {
990 InterpretedValue::Value(value) if !value.is_empty() => value.clone(),
991 _ => ":".to_string(),
992 },
993 None => ":".to_string(),
994 };
995
996 let separator = format!("{separator} ");
997
998 match title.rfind(&separator) {
999 Some(index) => {
1000 let main_title = title[..index].to_string();
1001 let subtitle = title[index + separator.len()..].to_string();
1002 (main_title, Some(subtitle))
1003 }
1004 None => (title.to_string(), None),
1005 }
1006}
1007
1008/// Resolves the document's author list.
1009///
1010/// When an [`AuthorLine`] is present it is normally authoritative (and has
1011/// already populated the `author_N` attributes), except that an explicit
1012/// `:authors:` entry whose value differs from the computed implicit value
1013/// replaces the implicit list. Otherwise the list is
1014/// reconstructed from document attributes, mirroring Asciidoctor's
1015/// `parse_header_metadata` reconciliation in precedence order: a
1016/// directly-assigned `author` attribute
1017/// stands in for a single author; failing that a semicolon-separated `authors`
1018/// attribute is split into individual authors; and failing that a contiguous
1019/// run of indexed `author_N` attributes (`author_1`, `author_2`, …) each
1020/// contributes one author. In each case the email is taken from the companion
1021/// `email`/`email_N` attribute, reflecting its final value.
1022///
1023/// For the `authors` and `author_N` forms this also populates the derived
1024/// author attributes (`author`, `firstname`, `authorinitials`, the `authors`
1025/// list, the per-author `author_N` companions, …) so that references such as
1026/// `{author}` resolve, matching Asciidoctor's `process_authors` (see issue
1027/// #718).
1028///
1029/// `author_attribute` is the author already parsed from the raw `author`
1030/// attribute value (see the header parse loop); it is reused rather than
1031/// re-parsing the HTML-encoded stored value.
1032///
1033/// `header_has_attributes` reports whether the header carried any attribute
1034/// entries. When it did not, none of the `author` / `authors` / `author_N`
1035/// attributes can be set, so the attribute lookups are skipped – the common
1036/// case for a document whose header is just a title (or absent).
1037fn resolve_authors(
1038 author_line: Option<&AuthorLine>,
1039 author_attribute: Option<Author>,
1040 header_has_attributes: bool,
1041 parser: &mut Parser,
1042) -> Vec<Author> {
1043 if let Some(author_line) = author_line {
1044 let implicit_authors: Vec<Author> = author_line.authors().cloned().collect();
1045
1046 // Reconcile an explicit `:authors:` entry against the implicit author
1047 // line, mirroring Asciidoctor's `parse_header_metadata`. When the
1048 // entry's value differs from the computed (comma-joined) value of the
1049 // implicit list, the entry *replaces* that list – re-splitting on `;`,
1050 // updating `authorcount`, and repopulating the derived `author_N` (and
1051 // per-author name-part) attributes. A value that matches the computed
1052 // value leaves the implicit list untouched.
1053 //
1054 // The `:authors:` path is reconciled first (matching Asciidoctor's
1055 // precedence: an explicit `:authors:` entry outranks the indexed
1056 // `:author_N:` entries reconciled just below); the single `:author:`
1057 // entry path keeps its existing inline handling.
1058 if let Some(authors_value) = attribute_string(parser, "authors") {
1059 let computed = implicit_authors
1060 .iter()
1061 .map(Author::name)
1062 .collect::<Vec<_>>()
1063 .join(", ");
1064
1065 if authors_value != computed
1066 && let Some(authors) = authors_from_authors_attribute(&authors_value, parser)
1067 {
1068 // `set_author_metadata` overwrites the derived attributes for
1069 // the replacement authors but does not clear ones the implicit
1070 // list set that the replacement does not (a shorter list leaves
1071 // a stale trailing `author_N`; a replacement author lacking a
1072 // middle name / email leaves the implicit `middlename` /
1073 // `email`). This mirrors Asciidoctor's `doc_attrs.update
1074 // author_metadata` – a merge that overwrites present keys and
1075 // never deletes absent ones – so `{author_3}` and friends can
1076 // outlive an `authorcount` that reflects the shorter list. The
1077 // divergence is Asciidoctor's; see the shrinking-replacement
1078 // regression test in this module.
1079 set_author_metadata(parser, &authors);
1080 return authors;
1081 }
1082 }
1083
1084 // Reconcile explicit indexed `:author_N:` entries against the implicit
1085 // author line, mirroring the indexed branch of Asciidoctor's
1086 // `parse_header_metadata`. Each `author_N` attribute whose value still
1087 // equals the implicit author's name leaves that position untouched; any
1088 // position whose value differs is overridden.
1089 // When at least one position was overridden, the reconciled list
1090 // repopulates the derived attributes – including the combined `authors`
1091 // string – so `{authors}` and `Document::authors()` reflect the
1092 // override, not just the individual `author_N` attribute.
1093 //
1094 // The indexed `author_N` attributes are only assigned once the implicit
1095 // line carries two or more authors, so a single-author line has no
1096 // `author_1` and needs no reconciliation here.
1097 if attribute_string(parser, "author_1").is_some() {
1098 let mut reconciled: Vec<Author> = Vec::new();
1099 let mut any_override = false;
1100 let mut index = 1;
1101
1102 while let Some(current) = attribute_string(parser, &format!("author_{index}")) {
1103 match implicit_authors.get(index - 1) {
1104 // The position is unchanged from the implicit line: reuse the
1105 // already-parsed author so its name partition is preserved. A
1106 // multi-word `lastname` such as `het Draeke` would otherwise
1107 // be re-split into a middle and last name by the names-only
1108 // partitioning below.
1109 Some(implicit) if current == implicit.name() => {
1110 reconciled.push(implicit.clone());
1111 }
1112
1113 // The position was overridden (or added beyond the implicit
1114 // list): partition the entry value with the names-only rules,
1115 // exactly as Asciidoctor does for an attribute-supplied
1116 // author.
1117 _ => {
1118 any_override = true;
1119
1120 if let Some(author) = Author::parse(¤t, parser, true) {
1121 reconciled.push(author);
1122 }
1123 }
1124 }
1125
1126 index += 1;
1127 }
1128
1129 if any_override {
1130 let reconciled = collect_indexed_authors(reconciled.into_iter(), parser);
1131
1132 set_author_metadata(parser, &reconciled);
1133
1134 return reconciled;
1135 }
1136 }
1137
1138 return implicit_authors;
1139 }
1140
1141 if !header_has_attributes {
1142 return vec![];
1143 }
1144
1145 // A directly-assigned `author` attribute describes a single author – but
1146 // only while it remains set. A later `:author!:` unsets the attribute
1147 // without carrying a raw value to refresh `author_attribute`, so consult
1148 // the attribute's final state rather than trusting the cached parse. The
1149 // per-author attributes for this form were already populated inline as the
1150 // `:author:` entry was parsed.
1151 if attribute_string(parser, "author").is_some()
1152 && let Some(author) = author_attribute
1153 {
1154 let author = author.with_email(attribute_string(parser, "email"));
1155
1156 // Mirror Asciidoctor's `process_authors`, which sets the combined
1157 // `authors` attribute to the single author's name. The remaining derived
1158 // keys (`firstname`, `authorinitials`, …) were populated inline as the
1159 // `:author:` entry was parsed – deliberately, so an explicit
1160 // `:authorinitials:` override survives – but `authors` was still left
1161 // unset there (see issue #1027).
1162 parser.set_attribute_by_value_from_header("authors", author.name());
1163
1164 return vec![author];
1165 }
1166
1167 // A semicolon-separated `authors` attribute entry contributes one author
1168 // per entry (Asciidoctor's `process_authors` with `multiple` set).
1169 if let Some(authors_value) = attribute_string(parser, "authors")
1170 && let Some(authors) = authors_from_authors_attribute(&authors_value, parser)
1171 {
1172 set_author_metadata(parser, &authors);
1173 return authors;
1174 }
1175
1176 // Otherwise, walk the indexed `author_N` attributes until one is missing.
1177 let mut raw_names = vec![];
1178 let mut index = 1;
1179
1180 while let Some(name) = attribute_string(parser, &format!("author_{index}")) {
1181 raw_names.push(name);
1182 index += 1;
1183 }
1184
1185 let authors = collect_indexed_authors(
1186 raw_names
1187 .iter()
1188 .filter_map(|name| Author::parse(name, parser, true)),
1189 parser,
1190 );
1191
1192 if !authors.is_empty() {
1193 set_author_metadata(parser, &authors);
1194 }
1195
1196 authors
1197}
1198
1199/// Builds the author list described by an `authors` attribute value, splitting
1200/// it into individual authors (Asciidoctor's `process_authors` with `multiple`
1201/// set) and attaching each author's companion `email_N`. Returns `None` when
1202/// the value yields no authors (so the caller can fall through to its next
1203/// resolution step).
1204fn authors_from_authors_attribute(value: &str, parser: &Parser) -> Option<Vec<Author>> {
1205 let authors = collect_indexed_authors(
1206 split_author_entries(value)
1207 .into_iter()
1208 .filter_map(|entry| Author::parse(entry, parser, true)),
1209 parser,
1210 );
1211
1212 if authors.is_empty() {
1213 None
1214 } else {
1215 Some(authors)
1216 }
1217}
1218
1219/// Reads the string value of a document attribute, or `None` when it is unset
1220/// or set without a value.
1221fn attribute_string(parser: &Parser, name: &str) -> Option<String> {
1222 match parser.attribute_value(name) {
1223 InterpretedValue::Value(value) => Some(value),
1224 _ => None,
1225 }
1226}
1227
1228/// Attaches each parsed author's companion `email_N` attribute (`email_1` for
1229/// the first author, `email_2` for the second, …) so the resolved list carries
1230/// the emails supplied through separate attribute entries.
1231fn collect_indexed_authors(authors: impl Iterator<Item = Author>, parser: &Parser) -> Vec<Author> {
1232 authors
1233 .enumerate()
1234 .map(|(idx, author)| {
1235 author.with_email(attribute_string(parser, &format!("email_{}", idx + 1)))
1236 })
1237 .collect()
1238}
1239
1240/// Splits an `authors` attribute value into raw author entries.
1241///
1242/// A semicolon separates authors only when it is immediately followed by a
1243/// space or the end of the value, matching Asciidoctor's `AuthorDelimiterRx`
1244/// (`/;(?: |$)/`). Blank entries are left in place; [`Author::parse`] trims
1245/// each entry and discards the empty ones.
1246fn split_author_entries(value: &str) -> Vec<&str> {
1247 let bytes = value.as_bytes();
1248 let mut entries: Vec<&str> = Vec::new();
1249 let mut start = 0;
1250
1251 for (index, c) in value.char_indices() {
1252 if c != ';' {
1253 continue;
1254 }
1255
1256 let is_separator = match bytes.get(index + 1) {
1257 Some(next) => *next == b' ',
1258 None => true,
1259 };
1260
1261 if is_separator {
1262 entries.push(&value[start..index]);
1263 start = index + 1;
1264 }
1265 }
1266
1267 entries.push(&value[start..]);
1268 entries
1269}
1270
1271fn apply_header_subs(source: &str, parser: &Parser) -> String {
1272 let span = Span::new(source);
1273
1274 let mut content = Content::from(span);
1275 SubstitutionGroup::Header.apply(&mut content, parser, None);
1276
1277 content.rendered().to_string()
1278}
1279
1280impl std::fmt::Debug for Header<'_> {
1281 fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
1282 f.debug_struct("Header")
1283 .field("title_source", &self.title_source)
1284 .field("title", &self.title)
1285 .field("doctitle", &self.doctitle)
1286 .field("main_title", &self.main_title)
1287 .field("subtitle", &self.subtitle)
1288 .field("id", &self.id)
1289 .field("roles", &self.roles)
1290 .field("attributes", &DebugSliceReference(&self.attributes))
1291 .field("author_line", &self.author_line)
1292 .field("authors", &self.authors)
1293 .field("revision_line", &self.revision_line)
1294 .field("comments", &DebugSliceReference(&self.comments))
1295 .field("source", &self.source)
1296 .finish()
1297 }
1298}
1299
1300#[cfg(test)]
1301mod tests {
1302 #![allow(clippy::unwrap_used)]
1303
1304 use crate::tests::prelude::*;
1305
1306 #[test]
1307 fn attributes_iterator_supports_exact_size_double_ended_and_nth() {
1308 // Exercises the opaque `HeaderAttributes` iterator's full surface:
1309 // `ExactSizeIterator` (and, through its default `len`, `size_hint`),
1310 // `DoubleEndedIterator`, and the `nth` override.
1311 let doc = Parser::default().parse(":alpha: 1\n:bravo: 2\n:charlie: 3\n\nbody\n");
1312 let header = doc.header();
1313
1314 // Collect once to learn the order and length without hard-coding a count.
1315 let names: Vec<_> = header
1316 .attributes()
1317 .map(|a| a.name().data().to_string())
1318 .collect();
1319
1320 assert!(names.len() >= 3);
1321 assert_eq!(names.first().map(String::as_str), Some("alpha"));
1322
1323 assert_eq!(header.attributes().len(), names.len());
1324
1325 assert_eq!(
1326 header.attributes().next_back().map(|a| a.name().data()),
1327 names.last().map(String::as_str),
1328 );
1329
1330 assert_eq!(
1331 header.attributes().nth(1).map(|a| a.name().data()),
1332 Some("bravo"),
1333 );
1334 }
1335
1336 #[test]
1337 fn leveloffset_does_not_coerce_an_over_deep_heading_to_the_doctitle() {
1338 // A marker run deeper than `======` is not a valid heading at all, so
1339 // even a negative `:leveloffset:` whose shift would drive its effective
1340 // level to 0 must not coerce it to the document title. It stays body
1341 // content (which the block parser then reports as exceeding the maximum
1342 // heading level), leaving the document with no doctitle.
1343 let doc = Parser::default().parse(":leveloffset: -6\n======= Not A Title");
1344
1345 assert_eq!(doc.doctitle(), None);
1346 }
1347
1348 #[test]
1349 fn impl_clone() {
1350 // Silly test to mark the #[derive(...)] line as covered.
1351 let mut parser = Parser::default();
1352
1353 let h1 = crate::document::Header::parse(crate::Span::new("= Title"), &mut parser)
1354 .unwrap_if_no_warnings();
1355 let h2 = h1.clone();
1356
1357 assert_eq!(h1, h2);
1358 }
1359
1360 #[test]
1361 fn only_title() {
1362 let mut parser = Parser::default();
1363 let mi = crate::document::Header::parse(crate::Span::new("= Just the Title"), &mut parser)
1364 .unwrap_if_no_warnings();
1365
1366 assert_eq!(
1367 mi.item,
1368 Header {
1369 title_source: Some(Span {
1370 data: "Just the Title",
1371 line: 1,
1372 col: 3,
1373 offset: 2,
1374 }),
1375 title: Some("Just the Title"),
1376 attributes: &[],
1377 author_line: None,
1378 revision_line: None,
1379 comments: &[],
1380 source: Span {
1381 data: "= Just the Title",
1382 line: 1,
1383 col: 1,
1384 offset: 0,
1385 }
1386 }
1387 );
1388
1389 assert_eq!(
1390 mi.after,
1391 Span {
1392 data: "",
1393 line: 1,
1394 col: 17,
1395 offset: 16
1396 }
1397 );
1398 }
1399
1400 #[test]
1401 fn trims_leading_spaces_in_title() {
1402 // This is totally a judgement call on my part. As far as I can tell,
1403 // the language doesn't describe behavior here.
1404 let mut parser = Parser::default();
1405 let mi =
1406 crate::document::Header::parse(crate::Span::new("= Just the Title"), &mut parser)
1407 .unwrap_if_no_warnings();
1408
1409 assert_eq!(
1410 mi.item,
1411 Header {
1412 title_source: Some(Span {
1413 data: "Just the Title",
1414 line: 1,
1415 col: 6,
1416 offset: 5,
1417 }),
1418 title: Some("Just the Title"),
1419 attributes: &[],
1420 author_line: None,
1421 revision_line: None,
1422 comments: &[],
1423 source: Span {
1424 data: "= Just the Title",
1425 line: 1,
1426 col: 1,
1427 offset: 0,
1428 }
1429 }
1430 );
1431
1432 assert_eq!(
1433 mi.after,
1434 Span {
1435 data: "",
1436 line: 1,
1437 col: 20,
1438 offset: 19
1439 }
1440 );
1441 }
1442
1443 #[test]
1444 fn trims_trailing_spaces_in_title() {
1445 let mut parser = Parser::default();
1446 let mi =
1447 crate::document::Header::parse(crate::Span::new("= Just the Title "), &mut parser)
1448 .unwrap_if_no_warnings();
1449
1450 assert_eq!(
1451 mi.item,
1452 Header {
1453 title_source: Some(Span {
1454 data: "Just the Title",
1455 line: 1,
1456 col: 3,
1457 offset: 2,
1458 }),
1459 title: Some("Just the Title"),
1460 attributes: &[],
1461 author_line: None,
1462 revision_line: None,
1463 comments: &[],
1464 source: Span {
1465 data: "= Just the Title",
1466 line: 1,
1467 col: 1,
1468 offset: 0,
1469 }
1470 }
1471 );
1472
1473 assert_eq!(
1474 mi.after,
1475 Span {
1476 data: "",
1477 line: 1,
1478 col: 20,
1479 offset: 19
1480 }
1481 );
1482 }
1483
1484 #[test]
1485 fn title_and_attribute() {
1486 let mut parser = Parser::default();
1487
1488 let mi = crate::document::Header::parse(
1489 crate::Span::new("= Just the Title\n:foo: bar\n\nblah"),
1490 &mut parser,
1491 )
1492 .unwrap_if_no_warnings();
1493
1494 assert_eq!(
1495 mi.item,
1496 Header {
1497 title_source: Some(Span {
1498 data: "Just the Title",
1499 line: 1,
1500 col: 3,
1501 offset: 2,
1502 }),
1503 title: Some("Just the Title"),
1504 attributes: &[Attribute {
1505 name: Span {
1506 data: "foo",
1507 line: 2,
1508 col: 2,
1509 offset: 18,
1510 },
1511 value_source: Some(Span {
1512 data: "bar",
1513 line: 2,
1514 col: 7,
1515 offset: 23,
1516 }),
1517 value: InterpretedValue::Value("bar"),
1518 source: Span {
1519 data: ":foo: bar",
1520 line: 2,
1521 col: 1,
1522 offset: 17,
1523 }
1524 }],
1525 author_line: None,
1526 revision_line: None,
1527 comments: &[],
1528 source: Span {
1529 data: "= Just the Title\n:foo: bar",
1530 line: 1,
1531 col: 1,
1532 offset: 0,
1533 }
1534 }
1535 );
1536
1537 assert_eq!(
1538 mi.after,
1539 Span {
1540 data: "blah",
1541 line: 4,
1542 col: 1,
1543 offset: 28
1544 }
1545 );
1546 }
1547
1548 #[test]
1549 fn title_applies_header_substitutions() {
1550 let mut parser = Parser::default();
1551
1552 let mi = crate::document::Header::parse(
1553 crate::Span::new("= The Title & Some{sp}Nonsense\n:foo: bar\n\nblah"),
1554 &mut parser,
1555 )
1556 .unwrap_if_no_warnings();
1557
1558 assert_eq!(
1559 mi.item,
1560 Header {
1561 title_source: Some(Span {
1562 data: "The Title & Some{sp}Nonsense",
1563 line: 1,
1564 col: 3,
1565 offset: 2,
1566 }),
1567 title: Some("The Title & Some Nonsense"),
1568 attributes: &[Attribute {
1569 name: Span {
1570 data: "foo",
1571 line: 2,
1572 col: 2,
1573 offset: 32,
1574 },
1575 value_source: Some(Span {
1576 data: "bar",
1577 line: 2,
1578 col: 7,
1579 offset: 37,
1580 }),
1581 value: InterpretedValue::Value("bar"),
1582 source: Span {
1583 data: ":foo: bar",
1584 line: 2,
1585 col: 1,
1586 offset: 31,
1587 }
1588 }],
1589 author_line: None,
1590 revision_line: None,
1591 comments: &[],
1592 source: Span {
1593 data: "= The Title & Some{sp}Nonsense\n:foo: bar",
1594 line: 1,
1595 col: 1,
1596 offset: 0,
1597 }
1598 }
1599 );
1600
1601 assert_eq!(
1602 mi.after,
1603 Span {
1604 data: "blah",
1605 line: 4,
1606 col: 1,
1607 offset: 42
1608 }
1609 );
1610 }
1611
1612 #[test]
1613 fn attribute_without_title() {
1614 let mut parser = Parser::default();
1615 let mi = crate::document::Header::parse(crate::Span::new(":foo: bar\n\nblah"), &mut parser)
1616 .unwrap_if_no_warnings();
1617
1618 assert_eq!(
1619 mi.item,
1620 Header {
1621 title_source: None,
1622 title: None,
1623 attributes: &[Attribute {
1624 name: Span {
1625 data: "foo",
1626 line: 1,
1627 col: 2,
1628 offset: 1,
1629 },
1630 value_source: Some(Span {
1631 data: "bar",
1632 line: 1,
1633 col: 7,
1634 offset: 6,
1635 }),
1636 value: InterpretedValue::Value("bar"),
1637 source: Span {
1638 data: ":foo: bar",
1639 line: 1,
1640 col: 1,
1641 offset: 0,
1642 }
1643 }],
1644 author_line: None,
1645 revision_line: None,
1646 comments: &[],
1647 source: Span {
1648 data: ":foo: bar",
1649 line: 1,
1650 col: 1,
1651 offset: 0,
1652 }
1653 }
1654 );
1655
1656 assert_eq!(
1657 mi.after,
1658 Span {
1659 data: "blah",
1660 line: 3,
1661 col: 1,
1662 offset: 11
1663 }
1664 );
1665 }
1666
1667 #[test]
1668 fn sets_doctitle_attribute() {
1669 let mut parser = Parser::default();
1670 let _doc = parser.parse("= Document Title Goes Here");
1671
1672 assert_eq!(
1673 parser.attribute_value("doctitle"),
1674 InterpretedValue::Value("Document Title Goes Here")
1675 );
1676 }
1677
1678 #[test]
1679 fn sets_author_attributes_from_author_attribute() {
1680 let mut parser = Parser::default();
1681 let _doc = parser.parse(":author: John Q. Smith <john@example.com>");
1682
1683 // Verify that individual author attributes are set.
1684 assert_eq!(
1685 parser.attribute_value("firstname"),
1686 InterpretedValue::Value("John")
1687 );
1688 assert_eq!(
1689 parser.attribute_value("middlename"),
1690 InterpretedValue::Value("Q.")
1691 );
1692 assert_eq!(
1693 parser.attribute_value("lastname"),
1694 InterpretedValue::Value("Smith")
1695 );
1696 assert_eq!(
1697 parser.attribute_value("authorinitials"),
1698 InterpretedValue::Value("JQS")
1699 );
1700 assert_eq!(
1701 parser.attribute_value("email"),
1702 InterpretedValue::Value("john@example.com")
1703 );
1704
1705 // Also verify the original author attribute is still set (with HTML encoding).
1706 assert_eq!(
1707 parser.attribute_value("author"),
1708 InterpretedValue::Value("John Q. Smith <john@example.com>")
1709 );
1710 }
1711
1712 #[test]
1713 fn author_attribute_with_four_or_more_parts_is_partitioned() {
1714 // A value with more than three parts does not match the author pattern,
1715 // so it is partitioned by splitting on whitespace into at most three
1716 // parts. The trailing parts are assigned to `lastname` and repeated
1717 // interior whitespace is condensed.
1718 let mut parser = Parser::default();
1719 let _doc = parser.parse(":author: Leroy Harold Scherer, Jr.");
1720
1721 assert_eq!(
1722 parser.attribute_value("author"),
1723 InterpretedValue::Value("Leroy Harold Scherer, Jr.")
1724 );
1725 assert_eq!(
1726 parser.attribute_value("firstname"),
1727 InterpretedValue::Value("Leroy")
1728 );
1729 assert_eq!(
1730 parser.attribute_value("middlename"),
1731 InterpretedValue::Value("Harold")
1732 );
1733 assert_eq!(
1734 parser.attribute_value("lastname"),
1735 InterpretedValue::Value("Scherer, Jr.")
1736 );
1737 assert_eq!(
1738 parser.attribute_value("authorinitials"),
1739 InterpretedValue::Value("LHS")
1740 );
1741 }
1742
1743 #[test]
1744 fn author_attribute_two_part_fallback_partitions_lastname() {
1745 // A two-part value that does not match the author pattern (here because
1746 // of the comma attached to the first part) still partitions into a first
1747 // and last name via the whitespace split.
1748 let mut parser = Parser::default();
1749 let _doc = parser.parse(":author: Jane, Doe");
1750
1751 assert_eq!(
1752 parser.attribute_value("author"),
1753 InterpretedValue::Value("Jane, Doe")
1754 );
1755 assert_eq!(
1756 parser.attribute_value("firstname"),
1757 InterpretedValue::Value("Jane,")
1758 );
1759 assert_eq!(
1760 parser.attribute_value("middlename"),
1761 InterpretedValue::Unset
1762 );
1763 assert_eq!(
1764 parser.attribute_value("lastname"),
1765 InterpretedValue::Value("Doe")
1766 );
1767 }
1768
1769 #[test]
1770 fn author_attribute_single_part_fallback_is_firstname_only() {
1771 // A single-token value that does not match the author pattern partitions
1772 // to `firstname` alone, with no middle or last name.
1773 let mut parser = Parser::default();
1774 let _doc = parser.parse(":author: Jane,");
1775
1776 assert_eq!(
1777 parser.attribute_value("author"),
1778 InterpretedValue::Value("Jane,")
1779 );
1780 assert_eq!(
1781 parser.attribute_value("firstname"),
1782 InterpretedValue::Value("Jane,")
1783 );
1784 assert_eq!(
1785 parser.attribute_value("middlename"),
1786 InterpretedValue::Unset
1787 );
1788 assert_eq!(parser.attribute_value("lastname"), InterpretedValue::Unset);
1789 }
1790
1791 #[test]
1792 fn author_attribute_four_or_more_parts_with_inline_email() {
1793 // A four-plus-part fallback value that carries a trailing `<email>` must
1794 // split the email off before partitioning, so it lands in `email` rather
1795 // than being absorbed into `lastname`.
1796 let mut parser = Parser::default();
1797 let _doc = parser.parse(":author: Leroy Harold Scherer, Jr. <leroy@example.com>");
1798
1799 assert_eq!(
1800 parser.attribute_value("firstname"),
1801 InterpretedValue::Value("Leroy")
1802 );
1803 assert_eq!(
1804 parser.attribute_value("middlename"),
1805 InterpretedValue::Value("Harold")
1806 );
1807 assert_eq!(
1808 parser.attribute_value("lastname"),
1809 InterpretedValue::Value("Scherer, Jr.")
1810 );
1811 assert_eq!(
1812 parser.attribute_value("email"),
1813 InterpretedValue::Value("leroy@example.com")
1814 );
1815 assert_eq!(
1816 parser.attribute_value("authorinitials"),
1817 InterpretedValue::Value("LHS")
1818 );
1819 }
1820
1821 #[test]
1822 fn author_attribute_reference_expands_and_partitions() {
1823 // A `:author:` value given entirely as an attribute reference is expanded
1824 // and then partitioned by the names-only rules, so it yields the same
1825 // metadata as the equivalent literal four-plus-part name.
1826 let mut parser = Parser::default();
1827 let _doc = parser.parse(":full-name: Leroy Harold Scherer, Jr.\n:author: {full-name}");
1828
1829 assert_eq!(
1830 parser.attribute_value("firstname"),
1831 InterpretedValue::Value("Leroy")
1832 );
1833 assert_eq!(
1834 parser.attribute_value("middlename"),
1835 InterpretedValue::Value("Harold")
1836 );
1837 assert_eq!(
1838 parser.attribute_value("lastname"),
1839 InterpretedValue::Value("Scherer, Jr.")
1840 );
1841 assert_eq!(
1842 parser.attribute_value("authorinitials"),
1843 InterpretedValue::Value("LHS")
1844 );
1845 }
1846
1847 #[test]
1848 fn author_attribute_reference_within_larger_value_expands_and_partitions() {
1849 // The same partitioning applies when the reference is only part of the
1850 // value (so the single-attribute fast path is not taken) and the expanded
1851 // result still fails the author pattern.
1852 let mut parser = Parser::default();
1853 let _doc = parser.parse(":rest: Harold Scherer, Jr.\n:author: Leroy {rest}");
1854
1855 assert_eq!(
1856 parser.attribute_value("firstname"),
1857 InterpretedValue::Value("Leroy")
1858 );
1859 assert_eq!(
1860 parser.attribute_value("middlename"),
1861 InterpretedValue::Value("Harold")
1862 );
1863 assert_eq!(
1864 parser.attribute_value("lastname"),
1865 InterpretedValue::Value("Scherer, Jr.")
1866 );
1867 }
1868
1869 #[test]
1870 fn author_attribute_non_breaking_space_is_not_a_name_separator() {
1871 // Only ASCII whitespace separates name parts. A non-breaking space
1872 // (U+00A0) joining two words keeps them as a single first name, matching
1873 // Ruby's whitespace split.
1874 let mut parser = Parser::default();
1875 let _doc = parser.parse(":author: John\u{a0}Doe Scherer, Jr.");
1876
1877 assert_eq!(
1878 parser.attribute_value("firstname"),
1879 InterpretedValue::Value("John\u{a0}Doe")
1880 );
1881 assert_eq!(
1882 parser.attribute_value("middlename"),
1883 InterpretedValue::Value("Scherer,")
1884 );
1885 assert_eq!(
1886 parser.attribute_value("lastname"),
1887 InterpretedValue::Value("Jr.")
1888 );
1889 }
1890
1891 #[test]
1892 fn sets_author_attributes_from_author_attribute_two_names() {
1893 let mut parser = Parser::default();
1894 let _doc = parser.parse(":author: Jane Doe");
1895
1896 // Verify that individual author attributes are set.
1897 assert_eq!(
1898 parser.attribute_value("firstname"),
1899 InterpretedValue::Value("Jane")
1900 );
1901 assert_eq!(
1902 parser.attribute_value("middlename"),
1903 InterpretedValue::Unset
1904 );
1905 assert_eq!(
1906 parser.attribute_value("lastname"),
1907 InterpretedValue::Value("Doe")
1908 );
1909 assert_eq!(
1910 parser.attribute_value("authorinitials"),
1911 InterpretedValue::Value("JD")
1912 );
1913 assert_eq!(parser.attribute_value("email"), InterpretedValue::Unset);
1914 }
1915
1916 #[test]
1917 fn sets_author_attributes_from_author_attribute_single_name() {
1918 let mut parser = Parser::default();
1919 let _doc = parser.parse(":author: Cher");
1920
1921 // Verify that individual author attributes are set.
1922 assert_eq!(
1923 parser.attribute_value("firstname"),
1924 InterpretedValue::Value("Cher")
1925 );
1926 assert_eq!(
1927 parser.attribute_value("middlename"),
1928 InterpretedValue::Unset
1929 );
1930 assert_eq!(parser.attribute_value("lastname"), InterpretedValue::Unset);
1931 assert_eq!(
1932 parser.attribute_value("authorinitials"),
1933 InterpretedValue::Value("C")
1934 );
1935 assert_eq!(parser.attribute_value("email"), InterpretedValue::Unset);
1936 }
1937
1938 #[test]
1939 fn sets_author_attributes_from_empty_string() {
1940 let mut parser = Parser::default();
1941 let _doc = parser.parse(":author:");
1942
1943 // Verify that individual author attributes are set.
1944 assert_eq!(parser.attribute_value("firstname"), InterpretedValue::Unset);
1945 assert_eq!(
1946 parser.attribute_value("middlename"),
1947 InterpretedValue::Unset
1948 );
1949 assert_eq!(parser.attribute_value("lastname"), InterpretedValue::Unset);
1950 assert_eq!(
1951 parser.attribute_value("authorinitials"),
1952 InterpretedValue::Unset
1953 );
1954 assert_eq!(parser.attribute_value("email"), InterpretedValue::Unset);
1955
1956 assert_eq!(parser.attribute_value("author"), InterpretedValue::Set);
1957 }
1958
1959 #[test]
1960 fn authors_from_author_line() {
1961 let doc = Parser::default().parse("= Title\nKismet R. Lee <kismet@asciidoctor.org>");
1962
1963 assert_eq!(doc.authors().len(), 1);
1964
1965 let author = doc.authors().first().unwrap();
1966 assert_eq!(author.name(), "Kismet R. Lee");
1967 assert_eq!(author.email(), Some("kismet@asciidoctor.org"));
1968 assert_eq!(author.initials(), "KRL");
1969 }
1970
1971 #[test]
1972 fn authors_from_author_attribute() {
1973 // With no author line, a directly-assigned `author` attribute stands in
1974 // for a single author, taking its email from the `email` attribute.
1975 let doc =
1976 Parser::default().parse("= Title\n:author: Jane Q. Public\n:email: jane@example.com");
1977
1978 assert_eq!(doc.authors().len(), 1);
1979
1980 let author = doc.authors().first().unwrap();
1981 assert_eq!(author.name(), "Jane Q. Public");
1982 assert_eq!(author.firstname(), "Jane");
1983 assert_eq!(author.middlename(), Some("Q."));
1984 assert_eq!(author.lastname(), Some("Public"));
1985 assert_eq!(author.email(), Some("jane@example.com"));
1986 assert_eq!(author.initials(), "JQP");
1987 }
1988
1989 #[test]
1990 fn authors_from_author_attribute_with_inline_email() {
1991 // The email may be given inline in the `author` attribute value; the
1992 // resolved author reflects the parsed name and that email.
1993 let doc = Parser::default().parse("= Title\n:author: John Q. Smith <john@example.com>");
1994
1995 assert_eq!(doc.authors().len(), 1);
1996
1997 let author = doc.authors().first().unwrap();
1998 assert_eq!(author.name(), "John Q. Smith");
1999 assert_eq!(author.firstname(), "John");
2000 assert_eq!(author.middlename(), Some("Q."));
2001 assert_eq!(author.lastname(), Some("Smith"));
2002 assert_eq!(author.email(), Some("john@example.com"));
2003 assert_eq!(author.initials(), "JQS");
2004 }
2005
2006 #[test]
2007 fn authors_is_empty_without_author_info() {
2008 let doc = Parser::default().parse("= Title\n\nBody.");
2009
2010 assert!(doc.authors().is_empty());
2011 }
2012
2013 #[test]
2014 fn authorcount_reflects_author_line() {
2015 // The `authorcount` attribute counts the resolved authors, whether they
2016 // come from the author line …
2017 let doc = Parser::default().parse("= Title\nJane Doe; John Smith\n\nBody.");
2018
2019 assert_eq!(doc.authors().len(), 2);
2020 assert_eq!(
2021 doc.attribute_value("authorcount"),
2022 InterpretedValue::Value("2")
2023 );
2024
2025 // … or from a single `:author:` attribute entry.
2026 let doc = Parser::default().parse(":author: Jane Doe\n\nBody.");
2027
2028 assert_eq!(
2029 doc.attribute_value("authorcount"),
2030 InterpretedValue::Value("1")
2031 );
2032
2033 // A document with no author information reports a count of zero.
2034 let doc = Parser::default().parse("= Title\n\nBody.");
2035
2036 assert_eq!(
2037 doc.attribute_value("authorcount"),
2038 InterpretedValue::Value("0")
2039 );
2040 }
2041
2042 #[test]
2043 fn explicit_authorinitials_after_author_still_wins() {
2044 // An explicit `:authorinitials:` entry is honored regardless of whether
2045 // it precedes or follows the `:author:` entry.
2046 let doc = Parser::default().parse(":author: Doc Writer\n:authorinitials: DOC\n\nBody.");
2047
2048 assert_eq!(
2049 doc.attribute_value("authorinitials"),
2050 InterpretedValue::Value("DOC")
2051 );
2052
2053 // A second `:author:` entry after the explicit initials does not clobber
2054 // them.
2055 let doc = Parser::default()
2056 .parse(":author: Jane Roe\n:authorinitials: DOC\n:author: Doc Writer\n\nBody.");
2057
2058 assert_eq!(
2059 doc.attribute_value("author"),
2060 InterpretedValue::Value("Doc Writer")
2061 );
2062 assert_eq!(
2063 doc.attribute_value("authorinitials"),
2064 InterpretedValue::Value("DOC")
2065 );
2066 }
2067
2068 #[test]
2069 fn later_author_entry_redrives_initials_without_explicit_override() {
2070 // Without an explicit `:authorinitials:` entry, a later `:author:`
2071 // overwrites the initials derived from the earlier one.
2072 let doc = Parser::default().parse(":author: Jane Roe\n:author: Doc Writer\n\nBody.");
2073
2074 assert_eq!(
2075 doc.attribute_value("authorinitials"),
2076 InterpretedValue::Value("DW")
2077 );
2078 }
2079
2080 #[test]
2081 fn single_author_entry_sets_combined_authors_attribute() {
2082 // A single `:author:` entry sets the combined `authors` attribute to the
2083 // author's name, matching Asciidoctor (`{authors}` equals `{author}`),
2084 // just as the implicit author line already does.
2085 let doc = Parser::default().parse(":author: Doc Writer\n\nBody.");
2086
2087 assert_eq!(
2088 doc.attribute_value("author"),
2089 InterpretedValue::Value("Doc Writer")
2090 );
2091 assert_eq!(
2092 doc.attribute_value("authors"),
2093 InterpretedValue::Value("Doc Writer")
2094 );
2095
2096 // The name is reconstructed the same way in the `authors` string, and an
2097 // explicit `:authorinitials:` override is still preserved (the reason
2098 // this path populates `authors` inline rather than via the shared
2099 // metadata routine).
2100 let doc = Parser::default()
2101 .parse("= T\n:authorinitials: DOC\n:author: Kismet R. Chameleon\n\nBody.");
2102
2103 assert_eq!(
2104 doc.attribute_value("author"),
2105 InterpretedValue::Value("Kismet R. Chameleon")
2106 );
2107 assert_eq!(
2108 doc.attribute_value("authors"),
2109 InterpretedValue::Value("Kismet R. Chameleon")
2110 );
2111 assert_eq!(
2112 doc.attribute_value("authorinitials"),
2113 InterpretedValue::Value("DOC")
2114 );
2115 }
2116
2117 #[test]
2118 fn author_unset_after_entry_leaves_authors_unset() {
2119 // A later `:author!:` removes the single author, so the combined
2120 // `authors` attribute must not be materialized from the earlier entry.
2121 let doc = Parser::default().parse(":author: Jane Doe\n:author!:\n\nBody.");
2122
2123 assert_eq!(doc.attribute_value("author"), InterpretedValue::Unset);
2124 assert_eq!(doc.attribute_value("authors"), InterpretedValue::Unset);
2125 assert!(doc.authors().is_empty());
2126 }
2127
2128 #[test]
2129 fn explicit_authorinitials_not_preserved_for_indexed_or_authors_forms() {
2130 // The explicit-`:authorinitials:` override is honored only for a single
2131 // `:author:` entry. For the indexed `author_N` form (and, as tested
2132 // elsewhere, the `:authors:` form) the derived initials overwrite it,
2133 // matching Asciidoctor (see [`set_author_metadata`]).
2134 let doc = Parser::default().parse(":authorinitials: DOC\n:author_1: Doc Writer\n\nBody.");
2135
2136 assert_eq!(
2137 doc.attribute_value("author"),
2138 InterpretedValue::Value("Doc Writer")
2139 );
2140 assert_eq!(
2141 doc.attribute_value("authorinitials"),
2142 InterpretedValue::Value("DW")
2143 );
2144 }
2145
2146 #[test]
2147 fn authors_attribute_splits_into_indexed_authors() {
2148 // A semicolon-separated `:authors:` entry populates the author list and
2149 // the derived per-author attributes.
2150 let doc = Parser::default().parse(":authors: Jane Doe; John Q. Smith\n\nBody.");
2151
2152 assert_eq!(doc.authors().len(), 2);
2153 assert_eq!(
2154 doc.attribute_value("authors"),
2155 InterpretedValue::Value("Jane Doe, John Q. Smith")
2156 );
2157 assert_eq!(
2158 doc.attribute_value("author"),
2159 InterpretedValue::Value("Jane Doe")
2160 );
2161 assert_eq!(
2162 doc.attribute_value("author_2"),
2163 InterpretedValue::Value("John Q. Smith")
2164 );
2165 assert_eq!(
2166 doc.attribute_value("middlename_2"),
2167 InterpretedValue::Value("Q.")
2168 );
2169 assert_eq!(
2170 doc.attribute_value("authorinitials_2"),
2171 InterpretedValue::Value("JQS")
2172 );
2173 }
2174
2175 #[test]
2176 fn authors_attribute_attaches_companion_emails_and_base_middlename() {
2177 // Companion `:email_N:` entries attach to each split author (`email_1`
2178 // also fills the base `email`), and the first author's middle name lands
2179 // on the unsuffixed `middlename`.
2180 let doc = Parser::default().parse(
2181 ":authors: Jane Q. Doe; John Smith\n:email_1: jane@example.com\n:email_2: john@example.com\n\nBody.",
2182 );
2183
2184 let authors = doc.authors();
2185 assert_eq!(authors.len(), 2);
2186 assert_eq!(authors.first().unwrap().email(), Some("jane@example.com"));
2187 assert_eq!(authors.get(1).unwrap().email(), Some("john@example.com"));
2188
2189 assert_eq!(
2190 doc.attribute_value("middlename"),
2191 InterpretedValue::Value("Q.")
2192 );
2193 assert_eq!(
2194 doc.attribute_value("email"),
2195 InterpretedValue::Value("jane@example.com")
2196 );
2197 assert_eq!(
2198 doc.attribute_value("email_2"),
2199 InterpretedValue::Value("john@example.com")
2200 );
2201 }
2202
2203 #[test]
2204 fn authors_attribute_semicolon_without_space_is_one_author() {
2205 // A semicolon that is not followed by a space (or the end of the value)
2206 // does not separate authors.
2207 let doc = Parser::default().parse(":authors: Joe Doe;Smith Johnson\n\nBody.");
2208
2209 assert_eq!(doc.authors().len(), 1);
2210 assert_eq!(
2211 doc.attribute_value("authorcount"),
2212 InterpretedValue::Value("1")
2213 );
2214 }
2215
2216 #[test]
2217 fn authors_attribute_single_name_authors_and_trailing_separator() {
2218 // Single-name authors carry no last name, and a trailing `;` (a
2219 // separator at the end of the value) contributes no extra author.
2220 let doc = Parser::default().parse(":authors: Cher; Madonna;\n\nBody.");
2221
2222 assert_eq!(doc.authors().len(), 2);
2223 assert_eq!(
2224 doc.attribute_value("authors"),
2225 InterpretedValue::Value("Cher, Madonna")
2226 );
2227 assert_eq!(
2228 doc.attribute_value("author"),
2229 InterpretedValue::Value("Cher")
2230 );
2231 assert_eq!(doc.attribute_value("lastname"), InterpretedValue::Unset);
2232 assert_eq!(
2233 doc.attribute_value("authorinitials"),
2234 InterpretedValue::Value("C")
2235 );
2236 assert_eq!(
2237 doc.attribute_value("author_2"),
2238 InterpretedValue::Value("Madonna")
2239 );
2240 assert_eq!(doc.attribute_value("lastname_2"), InterpretedValue::Unset);
2241 assert_eq!(
2242 doc.attribute_value("authorcount"),
2243 InterpretedValue::Value("2")
2244 );
2245 }
2246
2247 #[test]
2248 fn authors_attribute_with_only_empty_entries_yields_no_authors() {
2249 // An `:authors:` value that splits into only empty entries resolves to no
2250 // authors, so the derived attributes stay unset and `authorcount` is 0.
2251 let doc = Parser::default().parse(":authors: ;\n\nBody.");
2252
2253 assert!(doc.authors().is_empty());
2254 assert_eq!(doc.attribute_value("author"), InterpretedValue::Unset);
2255 assert_eq!(
2256 doc.attribute_value("authorcount"),
2257 InterpretedValue::Value("0")
2258 );
2259
2260 // With no authors resolved, the raw `authors` value is left as written
2261 // (never rewritten to a comma-joined list) – matching Asciidoctor, whose
2262 // `process_authors` returns only `authorcount` for an all-empty value.
2263 assert_eq!(doc.attribute_value("authors"), InterpretedValue::Value(";"));
2264 }
2265
2266 #[test]
2267 fn author_attribute_takes_precedence_over_authors() {
2268 // A base `:author:` entry wins over a semicolon-separated `:authors:`
2269 // entry (matching Asciidoctor's `if author … elsif authors` order), so
2270 // only the single author is resolved.
2271 let doc = Parser::default()
2272 .parse(":author: Solo Writer\n:authors: Jane Doe; John Smith\n\nBody.");
2273
2274 assert_eq!(doc.authors().len(), 1);
2275 assert_eq!(
2276 doc.attribute_value("author"),
2277 InterpretedValue::Value("Solo Writer")
2278 );
2279 assert_eq!(doc.attribute_value("author_2"), InterpretedValue::Unset);
2280 }
2281
2282 #[test]
2283 fn author_unset_after_being_assigned_yields_no_authors() {
2284 // A later `:author!:` unsets the attribute; the resolved author list
2285 // must reflect the removal, not the earlier assignment.
2286 let doc = Parser::default().parse("= Title\n:author: Jane Doe\n:author!:\n\nBody.");
2287
2288 assert_eq!(doc.attribute_value("author"), InterpretedValue::Unset);
2289 assert!(doc.authors().is_empty());
2290 }
2291
2292 #[test]
2293 fn authors_entry_replacing_a_longer_implicit_list_leaves_stale_attributes() {
2294 // When a differing `:authors:` entry replaces a *longer* implicit author
2295 // line, the reconciliation re-derives only the replacement authors'
2296 // attributes. Attributes the implicit line set that the replacement does
2297 // not are left in place: a trailing `author_N` (and its name parts)
2298 // beyond the shorter list, and the base `email` / `middlename` when the
2299 // replacement's first author omits them.
2300 //
2301 // This is not a gap in the port – it is exactly Asciidoctor's behavior.
2302 // Its `parse_header_metadata` reconciles with `doc_attrs.update
2303 // author_metadata`, a merge that overwrites the keys the replacement
2304 // supplies and never deletes the ones it omits, so `{author_3}` (and the
2305 // stale `email` / `middlename`) survive alongside an `authorcount` that
2306 // reflects the shorter list. Verified byte-for-byte against Asciidoctor
2307 // 2.0.26; there is no upstream test for this shrinking case, so this one
2308 // pins the parity to guard against a future "cleanup" that would clear
2309 // the stale keys and diverge.
2310 let mut parser = Parser::default();
2311 let doc = parser.parse(
2312 "= T\nKismet R. Lee <kismet@example.com>; Junior Writer; Third Author\n:authors: Stuart Rackham; Dan Allen\n",
2313 );
2314
2315 // The typed author list and `authorcount` reflect the two-author
2316 // replacement.
2317 let authors = doc.header().authors();
2318 assert_eq!(authors.len(), 2);
2319 assert_eq!(authors.first().unwrap().name(), "Stuart Rackham");
2320 assert_eq!(authors.get(1).unwrap().name(), "Dan Allen");
2321 assert_eq!(
2322 parser.attribute_value("authorcount"),
2323 InterpretedValue::Value("2")
2324 );
2325
2326 // The replacement authors' attributes are re-derived.
2327 assert_eq!(
2328 parser.attribute_value("authors"),
2329 InterpretedValue::Value("Stuart Rackham, Dan Allen")
2330 );
2331 assert_eq!(
2332 parser.attribute_value("author_1"),
2333 InterpretedValue::Value("Stuart Rackham")
2334 );
2335 assert_eq!(
2336 parser.attribute_value("author_2"),
2337 InterpretedValue::Value("Dan Allen")
2338 );
2339
2340 // Stale attributes from the longer/richer implicit list survive, exactly
2341 // as they do in Asciidoctor: the third author, the first implicit
2342 // author's email, and its middle name.
2343 assert_eq!(
2344 parser.attribute_value("author_3"),
2345 InterpretedValue::Value("Third Author")
2346 );
2347 assert_eq!(
2348 parser.attribute_value("email"),
2349 InterpretedValue::Value("kismet@example.com")
2350 );
2351 assert_eq!(
2352 parser.attribute_value("middlename"),
2353 InterpretedValue::Value("R.")
2354 );
2355 }
2356
2357 #[test]
2358 fn impl_debug() {
2359 let doc = Parser::default().parse("= Example Title\n\nabc\n\ndef");
2360 let header = doc.header();
2361
2362 assert_eq!(
2363 format!("{header:#?}"),
2364 r#"Header {
2365 title_source: Some(
2366 Span {
2367 data: "Example Title",
2368 line: 1,
2369 col: 3,
2370 offset: 2,
2371 },
2372 ),
2373 title: Some(
2374 "Example Title",
2375 ),
2376 doctitle: Some(
2377 "Example Title",
2378 ),
2379 main_title: Some(
2380 "Example Title",
2381 ),
2382 subtitle: None,
2383 id: None,
2384 roles: [],
2385 attributes: &[],
2386 author_line: None,
2387 authors: [],
2388 revision_line: None,
2389 comments: &[],
2390 source: Span {
2391 data: "= Example Title",
2392 line: 1,
2393 col: 1,
2394 offset: 0,
2395 },
2396}"#
2397 );
2398 }
2399
2400 #[test]
2401 fn no_subtitle() {
2402 // A title without a colon-space sequence has no subtitle, and its main
2403 // title equals its full title.
2404 let doc = Parser::default().parse("= Just the Title");
2405 let header = doc.header();
2406
2407 assert_eq!(header.title(), Some("Just the Title"));
2408 assert_eq!(header.main_title(), Some("Just the Title"));
2409 assert_eq!(header.subtitle(), None);
2410 }
2411
2412 #[test]
2413 fn no_title() {
2414 // With no document title at all, every title accessor returns `None`.
2415 let doc = Parser::default().parse(":foo: bar\n\nbody");
2416 let header = doc.header();
2417
2418 assert_eq!(header.title(), None);
2419 assert_eq!(header.main_title(), None);
2420 assert_eq!(header.subtitle(), None);
2421 }
2422
2423 #[test]
2424 fn colon_without_space_is_not_a_separator() {
2425 // The separator is a colon *followed by a space*; a bare colon does not
2426 // partition the title.
2427 let doc = Parser::default().parse("= Ratio 3:1 Explained");
2428 let header = doc.header();
2429
2430 assert_eq!(header.main_title(), Some("Ratio 3:1 Explained"));
2431 assert_eq!(header.subtitle(), None);
2432 }
2433
2434 #[test]
2435 fn subtitle_available_on_document() {
2436 // The subtitle is reachable directly from `Document` as well as from its
2437 // `Header`.
2438 let doc = Parser::default().parse("= Main Title: Subtitle");
2439
2440 assert_eq!(doc.doctitle(), Some("Main Title: Subtitle"));
2441 assert_eq!(doc.subtitle(), Some("Subtitle"));
2442 }
2443
2444 #[test]
2445 fn separator_block_attribute_above_title() {
2446 // A `[separator=::]` block attribute above the title changes the
2447 // subtitle separator for that title.
2448 let doc = Parser::default().parse("[separator=::]\n= Main Title:: Subtitle");
2449 let header = doc.header();
2450
2451 assert_eq!(header.main_title(), Some("Main Title"));
2452 assert_eq!(header.subtitle(), Some("Subtitle"));
2453
2454 // The custom separator replaces the default: a plain colon-space no
2455 // longer partitions the title.
2456 let doc = Parser::default().parse("[separator=::]\n= Main: Title:: Subtitle");
2457 let header = doc.header();
2458
2459 assert_eq!(header.main_title(), Some("Main: Title"));
2460 assert_eq!(header.subtitle(), Some("Subtitle"));
2461 }
2462
2463 #[test]
2464 fn separator_attribute_entry_overrides_block_attribute() {
2465 // When both are present, the later assignment wins in document order.
2466 // Here the `:title-separator:` entry follows the block attribute.
2467 let doc = Parser::default()
2468 .parse("[separator=::]\n= Main Title;; Subtitle\n:title-separator: ;;");
2469 let header = doc.header();
2470
2471 assert_eq!(header.main_title(), Some("Main Title"));
2472 assert_eq!(header.subtitle(), Some("Subtitle"));
2473 }
2474
2475 #[test]
2476 fn unrecognized_block_attribute_above_title_is_consumed() {
2477 // A well-formed block attribute line above the document title is now
2478 // parsed as document metadata, so the title that follows is recognized
2479 // even when the line carries no attribute this crate folds. An
2480 // unrecognized attribute (here `foo`) simply contributes no document
2481 // metadata rather than terminating the header.
2482 let doc = Parser::default().parse("[foo=bar]\n= A Header Title");
2483 let header = doc.header();
2484
2485 assert_eq!(header.title(), Some("A Header Title"));
2486 assert_eq!(header.subtitle(), None);
2487 assert_eq!(doc.attribute_value("foo"), InterpretedValue::Unset);
2488 }
2489
2490 #[test]
2491 fn reftext_block_attribute_above_title() {
2492 // A `[reftext="…"]` block attribute above the title recovers the title
2493 // (previously lost) and folds the value into the document's `reftext`
2494 // attribute, matching the `:reftext:` header attribute.
2495 let doc =
2496 Parser::default().parse("[reftext=\"Links and Stuff\"]\n= Links & Stuff\n\nBody.");
2497 let header = doc.header();
2498
2499 assert_eq!(header.title(), Some("Links & Stuff"));
2500 assert_eq!(
2501 doc.attribute_value("reftext"),
2502 InterpretedValue::Value("Links and Stuff")
2503 );
2504 assert_eq!(rendered_paragraphs(&doc), vec!["Body."]);
2505 }
2506
2507 #[test]
2508 fn id_block_attribute_above_title() {
2509 // The `[#id]` shorthand above the title assigns the document ID and
2510 // still recovers the title.
2511 let doc = Parser::default().parse("[#docid]\n= Document Title\n\nBody.");
2512 let header = doc.header();
2513
2514 assert_eq!(header.title(), Some("Document Title"));
2515 assert_eq!(header.id(), Some("docid"));
2516 assert_eq!(doc.id(), Some("docid"));
2517
2518 // The longhand `[id=…]` form is equivalent.
2519 let doc = Parser::default().parse("[id=docid]\n= Document Title");
2520 assert_eq!(doc.header().id(), Some("docid"));
2521 }
2522
2523 #[test]
2524 fn bracket_anchor_above_title() {
2525 // A `[[id]]` block anchor above the title assigns the document ID and
2526 // recovers the title, exactly as the `[#id]` shorthand does.
2527 let doc = Parser::default().parse("[[idname]]\n= Document Title\n\ncontent");
2528 let header = doc.header();
2529
2530 assert_eq!(header.title(), Some("Document Title"));
2531 assert_eq!(header.id(), Some("idname"));
2532 assert_eq!(doc.id(), Some("idname"));
2533 assert_eq!(rendered_paragraphs(&doc), vec!["content"]);
2534
2535 // The `[[id,reftext]]` form additionally folds its reference text into
2536 // the document's `reftext` attribute, resolving attribute references the
2537 // way a section/block anchor reftext does.
2538 let doc = Parser::default()
2539 .parse(":product: Widgets\n[[guide,{product} Guide]]\n= User Guide\n\ncontent");
2540 let header = doc.header();
2541
2542 assert_eq!(header.title(), Some("User Guide"));
2543 assert_eq!(header.id(), Some("guide"));
2544 assert_eq!(
2545 doc.attribute_value("reftext"),
2546 InterpretedValue::Value("Widgets Guide")
2547 );
2548 }
2549
2550 #[test]
2551 fn bracket_anchor_above_title_requires_a_valid_name() {
2552 // A `[[…]]` line whose anchor name is not a valid XML name is not folded
2553 // as document metadata; the header terminates as it does for any other
2554 // unrecognized line.
2555 let doc = Parser::default().parse("[[bad name]]\n= Document Title\n\ncontent");
2556 let header = doc.header();
2557
2558 assert_eq!(header.title(), None);
2559 assert_eq!(header.id(), None);
2560 }
2561
2562 #[test]
2563 fn stacked_block_attributes_above_title() {
2564 // Multiple block attribute lines may stack above the document title;
2565 // each folds into the document's metadata and the title is still
2566 // recovered (see #821). Here an `[#id]` line and a `[reftext="…"]` line
2567 // both sit above the title.
2568 let doc = Parser::default()
2569 .parse("[#docid]\n[reftext=\"Links and Stuff\"]\n= Links & Stuff\n\nBody.");
2570 let header = doc.header();
2571
2572 assert_eq!(header.title(), Some("Links & Stuff"));
2573 assert_eq!(header.id(), Some("docid"));
2574 assert_eq!(doc.id(), Some("docid"));
2575 assert_eq!(
2576 doc.attribute_value("reftext"),
2577 InterpretedValue::Value("Links and Stuff")
2578 );
2579 assert_eq!(rendered_paragraphs(&doc), vec!["Body."]);
2580 }
2581
2582 #[test]
2583 fn stacked_block_attributes_combine_roles() {
2584 // Roles from stacked block attribute lines all fold into the document's
2585 // `role` attribute (space-joined) and are surfaced through the block
2586 // API, alongside an ID set on a separate line.
2587 let doc = Parser::default().parse("[#docid]\n[.one]\n[.two]\n= Document Title");
2588 let header = doc.header();
2589
2590 assert_eq!(header.title(), Some("Document Title"));
2591 assert_eq!(header.id(), Some("docid"));
2592 assert_eq!(
2593 doc.attribute_value("role"),
2594 InterpretedValue::Value("one two")
2595 );
2596 assert_eq!(header.roles(), vec!["one", "two"]);
2597 }
2598
2599 #[test]
2600 fn stacked_block_attributes_require_a_following_title() {
2601 // Stacked block attribute lines are only folded when a document title
2602 // eventually follows. Without one, the run is left for the block parser
2603 // and no title is recognized.
2604 let doc = Parser::default().parse("[#docid]\n[reftext=\"Stuff\"]\n\nBody.");
2605 let header = doc.header();
2606
2607 assert_eq!(header.title(), None);
2608 assert_eq!(header.id(), None);
2609 assert_eq!(doc.attribute_value("reftext"), InterpretedValue::Unset);
2610 }
2611
2612 #[test]
2613 fn stacked_block_attributes_fold_a_block_anchor() {
2614 // A `[[anchor]]` line is a foldable metadata line like the attribute
2615 // lists around it, so a run mixing the two still folds and the title is
2616 // recognized. The ID follows last-wins semantics across the run (the
2617 // block anchor overrides the earlier `[#docid]`), matching Asciidoctor.
2618 let doc = Parser::default().parse("[#docid]\n[[anchor]]\n= Some Title");
2619 let header = doc.header();
2620
2621 assert_eq!(header.title(), Some("Some Title"));
2622 assert_eq!(header.id(), Some("anchor"));
2623 }
2624
2625 #[test]
2626 fn stacked_block_styles_gate_doctitle_by_effective_style() {
2627 // Whether a `[float]`/`[discrete]` style above a level-0 (`=`) title
2628 // suppresses doctitle promotion is decided by the run's *effective*
2629 // (merged, last-wins) block style – the same value the block parser
2630 // computes via `Attrlist::merge_block_attribute_line` – not by any
2631 // single line in isolation.
2632
2633 // A later non-discrete style overrides an earlier `float`, so the
2634 // effective style is not discrete: the title is still promoted rather
2635 // than disappearing.
2636 let doc = Parser::default().parse("[float]\n[normal]\n= Some Title\n\nbody");
2637 assert_eq!(doc.header().title(), Some("Some Title"));
2638 assert!(all_sections(&doc).is_empty());
2639
2640 // A later `float` overrides an earlier non-discrete style, so the
2641 // effective style is discrete: the heading becomes a level-0 discrete
2642 // floating title and is not promoted to the doctitle.
2643 let doc = Parser::default().parse("[normal]\n[float]\n= Some Title\n\nbody");
2644 assert_eq!(doc.header().title(), None);
2645
2646 let sec = first_section(&doc);
2647 assert_eq!(sec.section_type(), SectionType::Discrete);
2648 assert_eq!(sec.level(), 0);
2649 assert_eq!(sec.section_title(), "Some Title");
2650 }
2651
2652 #[test]
2653 fn rejected_metadata_run_does_not_fire_counter() {
2654 // A block attribute line above the title is only parsed as document
2655 // metadata once a title is confirmed to follow it (via
2656 // `document_title_follows_block_metadata`, which scans structurally and
2657 // never parses an attribute list). Here no title follows, so the
2658 // lookahead fails and the `[reftext=…]` line's attribute list is never
2659 // parsed during header parsing – its embedded `{counter:item}` must not
2660 // fire at header time.
2661 //
2662 // The `reftext` line advances the counter exactly once when the block
2663 // parser reaches it (yielding 1), so the following `{counter:item}`
2664 // reference renders 2 – not 3, which is what a leaked header-time
2665 // evaluation would produce.
2666 let doc =
2667 Parser::default().parse("[reftext=\"See {counter:item}\"]\nBody.\n\n{counter:item}");
2668
2669 assert_eq!(doc.header().title(), None);
2670 assert_eq!(rendered_paragraphs(&doc), vec!["Body.", "2"]);
2671 }
2672
2673 #[test]
2674 fn role_block_attribute_above_title() {
2675 // A `[role=…]` block attribute above the title folds into the document's
2676 // `role` attribute; multiple roles are space-joined. The same role(s)
2677 // are surfaced through the block API, so `Header::roles()` and
2678 // `Document::roles()` agree with the document attribute (see #820).
2679 let doc = Parser::default().parse("[role=special]\n= Document Title\n\nBody.");
2680 let header = doc.header();
2681
2682 assert_eq!(header.title(), Some("Document Title"));
2683 assert_eq!(
2684 doc.attribute_value("role"),
2685 InterpretedValue::Value("special")
2686 );
2687 assert_eq!(header.roles(), vec!["special"]);
2688 assert_eq!(doc.roles(), vec!["special"]);
2689
2690 // The dot shorthand assigns roles too, and they combine.
2691 let doc = Parser::default().parse("[.one.two]\n= Document Title");
2692 assert_eq!(
2693 doc.attribute_value("role"),
2694 InterpretedValue::Value("one two")
2695 );
2696 assert_eq!(doc.header().roles(), vec!["one", "two"]);
2697 assert_eq!(doc.roles(), vec!["one", "two"]);
2698 }
2699
2700 #[test]
2701 fn roles_empty_without_block_attribute() {
2702 // With no role assigned above the title, both the block accessor and the
2703 // header accessor report no roles.
2704 let doc = Parser::default().parse("= Document Title\n\nBody.");
2705
2706 assert!(doc.header().roles().is_empty());
2707 assert!(doc.roles().is_empty());
2708 }
2709
2710 #[test]
2711 fn options_block_attribute_above_title() {
2712 // A `[opts=…]` block attribute above the title sets a `<name>-option`
2713 // document attribute for each option.
2714 let doc = Parser::default().parse("[opts=\"noheader,autowidth\"]\n= Document Title");
2715
2716 assert!(doc.is_attribute_set("noheader-option"));
2717 assert!(doc.is_attribute_set("autowidth-option"));
2718
2719 // The `%` shorthand is equivalent.
2720 let doc = Parser::default().parse("[%hardbreaks]\n= Document Title");
2721 assert!(doc.is_attribute_set("hardbreaks-option"));
2722 }
2723
2724 #[test]
2725 fn bracketed_line_that_is_not_a_separator_attribute_list() {
2726 // A `[...]` line above the title that isn't a well-formed block
2727 // attribute list carrying `separator` is not consumed as a separator. An
2728 // empty block anchor (`[[]]`) and a leading-space form are both rejected,
2729 // so the line terminates the header exactly as any other unrecognized
2730 // line would.
2731 let doc = Parser::default().parse("[[]]\n= Some Title: Subtitle");
2732 let header = doc.header();
2733
2734 assert_eq!(header.title(), None);
2735 assert_eq!(header.subtitle(), None);
2736
2737 let doc = Parser::default().parse("[ separator=::]\n= Main Title:: Subtitle");
2738 let header = doc.header();
2739
2740 assert_eq!(header.title(), None);
2741 assert_eq!(header.subtitle(), None);
2742 }
2743
2744 #[test]
2745 fn empty_title_separator_falls_back_to_default() {
2746 // An explicitly empty `title-separator` falls back to the default
2747 // `:{sp}` separator rather than partitioning on an empty string.
2748 let doc = Parser::default().parse("= Main Title: Subtitle\n:title-separator:");
2749 let header = doc.header();
2750
2751 assert_eq!(header.main_title(), Some("Main Title"));
2752 assert_eq!(header.subtitle(), Some("Subtitle"));
2753 }
2754
2755 #[test]
2756 fn counter_does_not_shadow_title_separator() {
2757 // A counter that happens to be named `title-separator` must not be
2758 // mistaken for the configured separator: partitioning reads the document
2759 // attribute directly, ignoring the counter overlay. Here the title
2760 // creates such a counter, but the default `:{sp}` separator still
2761 // applies.
2762 let doc = Parser::default().parse("= Main Title: Subtitle {counter:title-separator}");
2763 let header = doc.header();
2764
2765 assert_eq!(header.main_title(), Some("Main Title"));
2766 assert_eq!(header.subtitle(), Some("Subtitle 1"));
2767 }
2768
2769 #[test]
2770 fn skips_block_comment_before_author() {
2771 // A `////` block comment ahead of the author line is skipped and
2772 // retained as a single header comment; the author line that follows it
2773 // is parsed normally.
2774 let doc = Parser::default()
2775 .parse("= Title\n////\nAsciidoctor\nrelease artist\n////\nRyan Waldron");
2776 let header = doc.header();
2777
2778 let author = header.authors().first().unwrap();
2779 assert_eq!(author.name(), "Ryan Waldron");
2780
2781 assert_eq!(header.comments().count(), 1);
2782 assert_eq!(
2783 header.comments().next().unwrap().data(),
2784 "////\nAsciidoctor\nrelease artist\n////"
2785 );
2786 }
2787
2788 #[test]
2789 fn skips_block_comment_with_blank_lines() {
2790 // Blank lines inside a header block comment do not terminate the header;
2791 // the whole block is skipped and the author line is still recognized.
2792 let doc = Parser::default().parse("= Title\n////\n\nAsciidoctor\n\n////\nRyan Waldron");
2793 let header = doc.header();
2794
2795 assert_eq!(header.authors().first().unwrap().name(), "Ryan Waldron");
2796 assert_eq!(header.comments().count(), 1);
2797 }
2798
2799 #[test]
2800 fn unterminated_block_comment_consumes_rest_of_header() {
2801 // An unterminated `////` block comment runs to the end of the input,
2802 // mirroring Asciidoctor; nothing after it is parsed as an author line.
2803 let doc = Parser::default().parse("= Title\n////\nAsciidoctor\nRyan Waldron");
2804 let header = doc.header();
2805
2806 assert!(header.authors().is_empty());
2807 assert_eq!(header.comments().count(), 1);
2808 }
2809
2810 #[test]
2811 fn longer_block_comment_delimiter_requires_matching_close() {
2812 // The closing delimiter must repeat the opening line exactly: a `////`
2813 // line does not close a `/////` block, so the block runs on until the
2814 // matching `/////` and the author line after it is recognized.
2815 let doc = Parser::default()
2816 .parse("= Title\n/////\nAsciidoctor\n////\nstill comment\n/////\nRyan Waldron");
2817 let header = doc.header();
2818
2819 assert_eq!(header.authors().first().unwrap().name(), "Ryan Waldron");
2820 assert_eq!(header.comments().count(), 1);
2821 }
2822
2823 #[test]
2824 fn three_slashes_is_not_a_block_comment() {
2825 // A line of exactly three slashes is not a block comment delimiter
2826 // (which requires four or more slashes), so it is not skipped: were it
2827 // mistaken for an unterminated block comment it would swallow the rest
2828 // of the header, but instead the author and revision lines before it
2829 // are captured as before.
2830 let mut parser = Parser::default();
2831 let _ = parser.parse("= Title\nJoe Cool\nv1.0\n///\nstuff");
2832
2833 assert_eq!(
2834 parser.attribute_value("author"),
2835 InterpretedValue::Value("Joe Cool")
2836 );
2837 assert_eq!(
2838 parser.attribute_value("revnumber"),
2839 InterpretedValue::Value("1.0")
2840 );
2841 }
2842
2843 mod markdown_style_document_title {
2844 use crate::tests::prelude::*;
2845
2846 #[test]
2847 fn hash_marker_is_a_document_title() {
2848 let mut parser = Parser::default();
2849 let mi =
2850 crate::document::Header::parse(crate::Span::new("# Just the Title"), &mut parser)
2851 .unwrap_if_no_warnings();
2852
2853 assert_eq!(
2854 mi.item,
2855 Header {
2856 title_source: Some(Span {
2857 data: "Just the Title",
2858 line: 1,
2859 col: 3,
2860 offset: 2,
2861 }),
2862 title: Some("Just the Title"),
2863 attributes: &[],
2864 author_line: None,
2865 revision_line: None,
2866 comments: &[],
2867 source: Span {
2868 data: "# Just the Title",
2869 line: 1,
2870 col: 1,
2871 offset: 0,
2872 }
2873 }
2874 );
2875
2876 assert_eq!(
2877 mi.after,
2878 Span {
2879 data: "",
2880 line: 1,
2881 col: 17,
2882 offset: 16
2883 }
2884 );
2885 }
2886
2887 #[test]
2888 fn sets_doctitle_attribute() {
2889 let doc = Parser::default().parse("# Doc Title\n\n{doctitle}");
2890 assert_eq!(doc.header().title(), Some("Doc Title"));
2891 assert_eq!(rendered_paragraphs(&doc), vec!["Doc Title"]);
2892 }
2893
2894 #[test]
2895 fn strips_symmetric_close() {
2896 let doc = Parser::default().parse("# Doc Title #");
2897 assert_eq!(doc.header().title(), Some("Doc Title"));
2898 }
2899
2900 #[test]
2901 fn does_not_strip_mismatched_close() {
2902 // The close must repeat the opening marker, so a trailing `=` after
2903 // a `#` title is title text.
2904 let doc = Parser::default().parse("# Doc Title =");
2905 assert_eq!(doc.header().title(), Some("Doc Title ="));
2906 }
2907
2908 #[test]
2909 fn requires_whitespace_after_marker() {
2910 let doc = Parser::default().parse("#Doc Title");
2911
2912 assert_eq!(doc.header().title(), None);
2913 assert_eq!(rendered_paragraphs(&doc), vec!["#Doc Title"]);
2914 }
2915
2916 #[test]
2917 fn carries_the_rest_of_the_header() {
2918 // Everything that may follow an `=` title – attribute entries, the
2919 // author line, the revision line – follows a `#` title too.
2920 let doc = Parser::default()
2921 .parse("# Doc Title\n:foo: bar\nKismet R. Lee <kismet@asciidoctor.org>\nv1.0\n");
2922 let header = doc.header();
2923
2924 assert_eq!(header.title(), Some("Doc Title"));
2925 assert_eq!(header.authors().first().unwrap().firstname(), "Kismet");
2926 assert_eq!(header.revision_line().unwrap().revnumber().unwrap(), "1.0");
2927 }
2928
2929 #[test]
2930 fn partitions_subtitle() {
2931 let doc = Parser::default().parse("# Main Title: Subtitle");
2932 let header = doc.header();
2933
2934 assert_eq!(header.main_title(), Some("Main Title"));
2935 assert_eq!(header.subtitle(), Some("Subtitle"));
2936 }
2937
2938 #[test]
2939 fn separator_block_attribute_above_title() {
2940 // The `[separator=…]` line is only intercepted when a document title
2941 // follows it; a `#` title qualifies just as an `=` title does.
2942 let doc = Parser::default().parse("[separator=::]\n# Main Title:: Subtitle");
2943 let header = doc.header();
2944
2945 assert_eq!(header.main_title(), Some("Main Title"));
2946 assert_eq!(header.subtitle(), Some("Subtitle"));
2947 }
2948
2949 #[test]
2950 fn markdown_title_followed_by_markdown_sections() {
2951 let doc = Parser::default().parse("# Doc Title\n\n## Section One\n\nblah blah\n");
2952
2953 assert_eq!(doc.header().title(), Some("Doc Title"));
2954
2955 let section = first_section(&doc);
2956
2957 assert_eq!(section.level(), 1);
2958 assert_eq!(section.section_title(), "Section One");
2959 }
2960 }
2961}