2 * \file output_docbook.cpp
3 * This file is part of LyX, the document processor.
4 * Licence details can be found in the file COPYING.
6 * \author Lars Gullik Bjønnes
9 * Full author contact details are available in file CREDITS.
15 #include "buffer_funcs.h"
16 #include "BufferParams.h"
18 #include "InsetList.h"
19 #include "Paragraph.h"
20 #include "ParagraphList.h"
21 #include "ParagraphParameters.h"
24 #include "TextClass.h"
26 #include "insets/InsetBibtex.h"
27 #include "insets/InsetBibitem.h"
28 #include "insets/InsetLabel.h"
29 #include "insets/InsetNote.h"
31 #include "support/lassert.h"
33 #include "support/regex.h"
41 using namespace lyx::support;
47 std::string fontToDocBookTag(xml::FontTypes type)
50 case xml::FontTypes::FT_EMPH:
51 case xml::FontTypes::FT_BOLD:
53 case xml::FontTypes::FT_NOUN:
55 case xml::FontTypes::FT_UBAR:
56 case xml::FontTypes::FT_WAVE:
57 case xml::FontTypes::FT_DBAR:
58 case xml::FontTypes::FT_SOUT:
59 case xml::FontTypes::FT_XOUT:
60 case xml::FontTypes::FT_ITALIC:
61 case xml::FontTypes::FT_UPRIGHT:
62 case xml::FontTypes::FT_SLANTED:
63 case xml::FontTypes::FT_SMALLCAPS:
64 case xml::FontTypes::FT_ROMAN:
65 case xml::FontTypes::FT_SANS:
67 case xml::FontTypes::FT_TYPE:
69 case xml::FontTypes::FT_SIZE_TINY:
70 case xml::FontTypes::FT_SIZE_SCRIPT:
71 case xml::FontTypes::FT_SIZE_FOOTNOTE:
72 case xml::FontTypes::FT_SIZE_SMALL:
73 case xml::FontTypes::FT_SIZE_NORMAL:
74 case xml::FontTypes::FT_SIZE_LARGE:
75 case xml::FontTypes::FT_SIZE_LARGER:
76 case xml::FontTypes::FT_SIZE_LARGEST:
77 case xml::FontTypes::FT_SIZE_HUGE:
78 case xml::FontTypes::FT_SIZE_HUGER:
79 case xml::FontTypes::FT_SIZE_INCREASE:
80 case xml::FontTypes::FT_SIZE_DECREASE:
88 string fontToRole(xml::FontTypes type)
90 // Specific fonts are achieved with roles. The only common ones are "" for basic emphasis,
91 // and "bold"/"strong" for bold. With some specific options, other roles are copied into
92 // HTML output (via the DocBook XSLT sheets); otherwise, if not recognised, they are just ignored.
93 // Hence, it is not a problem to have many roles by default here.
94 // See https://www.sourceware.org/ml/docbook/2003-05/msg00269.html
96 case xml::FontTypes::FT_ITALIC:
97 case xml::FontTypes::FT_EMPH:
99 case xml::FontTypes::FT_BOLD:
101 case xml::FontTypes::FT_NOUN: // Outputs a <person>
102 case xml::FontTypes::FT_TYPE: // Outputs a <code>
104 case xml::FontTypes::FT_UBAR:
107 // All other roles are non-standard for DocBook.
109 case xml::FontTypes::FT_WAVE:
111 case xml::FontTypes::FT_DBAR:
113 case xml::FontTypes::FT_SOUT:
115 case xml::FontTypes::FT_XOUT:
117 case xml::FontTypes::FT_UPRIGHT:
119 case xml::FontTypes::FT_SLANTED:
121 case xml::FontTypes::FT_SMALLCAPS:
123 case xml::FontTypes::FT_ROMAN:
125 case xml::FontTypes::FT_SANS:
127 case xml::FontTypes::FT_SIZE_TINY:
129 case xml::FontTypes::FT_SIZE_SCRIPT:
130 return "size_script";
131 case xml::FontTypes::FT_SIZE_FOOTNOTE:
132 return "size_footnote";
133 case xml::FontTypes::FT_SIZE_SMALL:
135 case xml::FontTypes::FT_SIZE_NORMAL:
136 return "size_normal";
137 case xml::FontTypes::FT_SIZE_LARGE:
139 case xml::FontTypes::FT_SIZE_LARGER:
140 return "size_larger";
141 case xml::FontTypes::FT_SIZE_LARGEST:
142 return "size_largest";
143 case xml::FontTypes::FT_SIZE_HUGE:
145 case xml::FontTypes::FT_SIZE_HUGER:
147 case xml::FontTypes::FT_SIZE_INCREASE:
148 return "size_increase";
149 case xml::FontTypes::FT_SIZE_DECREASE:
150 return "size_decrease";
157 string fontToAttribute(xml::FontTypes type) {
158 // If there is a role (i.e. nonstandard use of a tag), output the attribute. Otherwise, the sheer tag is sufficient
160 string role = fontToRole(type);
162 return "role='" + role + "'";
169 xml::FontTag docbookStartFontTag(xml::FontTypes type)
171 return xml::FontTag(from_utf8(fontToDocBookTag(type)), from_utf8(fontToAttribute(type)), type);
175 xml::EndFontTag docbookEndFontTag(xml::FontTypes type)
177 return xml::EndFontTag(from_utf8(fontToDocBookTag(type)), type);
181 // Convenience functions to open and close tags. First, very low-level ones to ensure a consistent new-line behaviour.
185 // Contents of the block.
190 // <paratag>Contents of the paragraph.</paratag>
193 // Content before<inlinetag>Contents of the paragraph.</inlinetag>Content after
195 void openInlineTag(XMLStream & xs, const std::string & tag, const std::string & attr)
197 xs << xml::StartTag(tag, attr);
201 void closeInlineTag(XMLStream & xs, const std::string & tag)
203 xs << xml::EndTag(tag);
207 void openParTag(XMLStream & xs, const std::string & tag, const std::string & attr)
209 if (!xs.isLastTagCR())
211 xs << xml::StartTag(tag, attr);
215 void closeParTag(XMLStream & xs, const std::string & tag)
217 xs << xml::EndTag(tag);
222 void openBlockTag(XMLStream & xs, const std::string & tag, const std::string & attr)
224 if (!xs.isLastTagCR())
226 xs << xml::StartTag(tag, attr);
231 void closeBlockTag(XMLStream & xs, const std::string & tag)
233 if (!xs.isLastTagCR())
235 xs << xml::EndTag(tag);
240 void openTag(XMLStream & xs, const std::string & tag, const std::string & attr, const std::string & tagtype)
242 if (tag.empty() || tag == "NONE")
245 if (tag == "para" || tagtype == "paragraph") // Special case for <para>: always considered as a paragraph.
246 openParTag(xs, tag, attr);
247 else if (tagtype == "block")
248 openBlockTag(xs, tag, attr);
249 else if (tagtype == "inline")
250 openInlineTag(xs, tag, attr);
252 xs.writeError("Unrecognised tag type '" + tagtype + "' for '" + tag + " " + attr + "'");
256 void closeTag(XMLStream & xs, const std::string & tag, const std::string & tagtype)
258 if (tag.empty() || tag == "NONE")
261 if (tag == "para" || tagtype == "paragraph") // Special case for <para>: always considered as a paragraph.
262 closeParTag(xs, tag);
263 else if (tagtype == "block")
264 closeBlockTag(xs, tag);
265 else if (tagtype == "inline")
266 closeInlineTag(xs, tag);
268 xs.writeError("Unrecognised tag type '" + tagtype + "' for '" + tag + "'");
272 // Higher-level convenience functions.
274 void openParTag(XMLStream & xs, const Paragraph * par, const Paragraph * prevpar)
276 Layout const & lay = par->layout();
281 // When should the wrapper be opened here? Only if the previous paragraph has the SAME wrapper tag
282 // (usually, they won't have the same layout) and the CURRENT one allows merging.
283 // The main use case is author information in several paragraphs: if the name of the author is the
284 // first paragraph of an author, then merging with the previous tag does not make sense. Say the
285 // next paragraph is the affiliation, then it should be output in the same <author> tag (different
286 // layout, same wrapper tag).
287 bool openWrapper = lay.docbookwrappertag() != "NONE";
288 if (prevpar != nullptr) {
289 Layout const & prevlay = prevpar->layout();
290 if (prevlay.docbookwrappertag() != "NONE") {
291 openWrapper = prevlay.docbookwrappertag() == lay.docbookwrappertag()
292 && !lay.docbookwrappermergewithprevious();
298 openTag(xs, lay.docbookwrappertag(), lay.docbookwrapperattr(), lay.docbookwrappertagtype());
300 const string & tag = lay.docbooktag();
302 auto xmltag = xml::ParTag(tag, lay.docbookattr());
303 if (!xs.isTagOpen(xmltag, 1)) // Don't nest a paragraph directly in a paragraph.
304 // TODO: required or not?
305 // TODO: avoid creating a ParTag object just for this query...
306 openTag(xs, lay.docbooktag(), lay.docbookattr(), lay.docbooktagtype());
309 openTag(xs, lay.docbookitemtag(), lay.docbookitemattr(), lay.docbookitemtagtype());
310 openTag(xs, lay.docbookiteminnertag(), lay.docbookiteminnerattr(), lay.docbookiteminnertagtype());
314 void closeParTag(XMLStream & xs, Paragraph const * par, Paragraph const * nextpar)
319 // See comment in openParTag.
320 Layout const & lay = par->layout();
321 bool closeWrapper = lay.docbookwrappertag() != "NONE";
322 if (nextpar != nullptr) {
323 Layout const & nextlay = nextpar->layout();
324 if (nextlay.docbookwrappertag() != "NONE") {
325 closeWrapper = nextlay.docbookwrappertag() == lay.docbookwrappertag()
326 && !nextlay.docbookwrappermergewithprevious();
331 closeTag(xs, lay.docbookiteminnertag(), lay.docbookiteminnertagtype());
332 closeTag(xs, lay.docbookitemtag(), lay.docbookitemtagtype());
333 closeTag(xs, lay.docbooktag(), lay.docbooktagtype());
335 closeTag(xs, lay.docbookwrappertag(), lay.docbookwrappertagtype());
339 void openLabelTag(XMLStream & xs, Layout const & lay) // Mostly for definition lists.
341 openTag(xs, lay.docbookitemlabeltag(), lay.docbookitemlabelattr(), lay.docbookitemlabeltagtype());
345 void closeLabelTag(XMLStream & xs, Layout const & lay)
347 closeTag(xs, lay.docbookitemlabeltag(), lay.docbookitemlabeltagtype());
351 void openItemTag(XMLStream & xs, Layout const & lay)
353 openTag(xs, lay.docbookitemtag(), lay.docbookitemattr(), lay.docbookitemtagtype());
357 void closeItemTag(XMLStream & xs, Layout const & lay)
359 closeTag(xs, lay.docbookitemtag(), lay.docbookitemtagtype());
367 OutputParams const &,
368 ParagraphList::const_iterator);
371 void makeParagraphBibliography(
374 OutputParams const & runparams,
376 ParagraphList::const_iterator const & par)
378 // If this is the first paragraph in a bibliography, open the bibliography tag.
379 auto pbegin_before = text.paragraphs().getParagraphBefore(par);
380 if (pbegin_before->layout().latextype != LATEX_BIB_ENVIRONMENT) {
381 xs << xml::StartTag("bibliography");
385 // Start the precooked bibliography entry. This is very much like opening a paragraph tag.
386 // Don't forget the citation ID!
388 for (auto i = 0; i < par->size(); ++i) {
389 Inset const *ip = par->getInset(i);
392 if (const auto * bibitem = dynamic_cast<const InsetBibitem*>(ip)) {
393 attr = from_utf8("xml:id='") + bibitem->getParam("key") + from_utf8("'");
397 xs << xml::StartTag(from_utf8("bibliomixed"), attr);
399 // Generate the entry.
400 auto const begin = text.paragraphs().begin();
401 par->simpleDocBookOnePar(buf, xs, runparams, text.outerFont(std::distance(begin, par)), true, true, 0);
403 // End the precooked bibliography entry.
404 xs << xml::EndTag("bibliomixed");
407 // If this is the last paragraph in a bibliography, close the bibliography tag.
408 auto const end = text.paragraphs().end();
409 bool endBibliography = par == end;
410 if (!endBibliography) {
413 endBibliography = par->layout().latextype != LATEX_BIB_ENVIRONMENT;
416 if (endBibliography) {
417 xs << xml::EndTag("bibliography");
426 OutputParams const & runparams,
428 ParagraphList::const_iterator const & par)
430 auto const begin = text.paragraphs().begin();
431 auto const end = text.paragraphs().end();
432 auto prevpar = text.paragraphs().getParagraphBefore(par);
434 // We want to open the paragraph tag if:
435 // (i) the current layout permits multiple paragraphs
436 // (ii) we are either not already inside a paragraph (HTMLIsBlock) OR
437 // we are, but this is not the first paragraph
439 // But there is also a special case, and we first see whether we are in it.
440 // We do not want to open the paragraph tag if this paragraph contains
441 // only one item, and that item is "inline", i.e., not HTMLIsBlock (such
442 // as a branch). On the other hand, if that single item has a font change
443 // applied to it, then we still do need to open the paragraph.
445 // Obviously, this is very fragile. The main reason we need to do this is
446 // because of branches, e.g., a branch that contains an entire new section.
447 // We do not really want to wrap that whole thing in a <div>...</div>.
448 bool special_case = false;
449 Inset const *specinset = par->size() == 1 ? par->getInset(0) : nullptr;
450 if (specinset && !specinset->getLayout().htmlisblock()) { // TODO: Convert htmlisblock to a DocBook parameter?
451 Layout const &style = par->layout();
452 FontInfo const first_font = style.labeltype == LABEL_MANUAL ?
453 style.labelfont : style.font;
454 FontInfo const our_font =
455 par->getFont(buf.masterBuffer()->params(), 0,
456 text.outerFont(std::distance(begin, par))).fontInfo();
458 if (first_font == our_font)
462 // Plain layouts must be ignored.
463 if (!special_case && buf.params().documentClass().isPlainLayout(par->layout()) && !runparams.docbook_force_pars)
465 // TODO: Could get rid of this with a DocBook equivalent to htmlisblock?
466 if (!special_case && par->size() == 1 && par->getInset(0)) {
467 Inset const * firstInset = par->getInset(0);
469 // Floats cannot be in paragraphs.
470 special_case = to_utf8(firstInset->layoutName()).substr(0, 6) == "Float:";
472 // Bibliographies cannot be in paragraphs.
473 if (!special_case && firstInset->asInsetCommand())
474 special_case = firstInset->asInsetCommand()->params().getCmdName() == "bibtex";
476 // Equations do not deserve their own paragraph (DocBook allows them outside paragraphs).
477 if (!special_case && firstInset->asInsetMath())
480 // ERTs are in comments, not paragraphs.
481 if (!special_case && firstInset->lyxCode() == lyx::ERT_CODE)
484 // Listings should not get into their own paragraph.
485 if (!special_case && firstInset->lyxCode() == lyx::LISTINGS_CODE)
489 bool const open_par = runparams.docbook_make_pars
490 && !runparams.docbook_in_par
493 // We want to issue the closing tag if either:
494 // (i) We opened it, and either docbook_in_par is false,
495 // or we're not in the last paragraph, anyway.
496 // (ii) We didn't open it and docbook_in_par is true,
497 // but we are in the first par, and there is a next par.
500 bool const close_par = open_par && (!runparams.docbook_in_par);
502 // Determine if this paragraph has some real content. Things like new pages are not caught
503 // by Paragraph::empty(), even though they do not generate anything useful in DocBook.
504 odocstringstream os2;
506 par->simpleDocBookOnePar(buf, xs2, runparams, text.outerFont(distance(begin, par)), open_par, close_par, 0);
508 docstring cleaned = os2.str();
509 static const lyx::regex reg("[ \\r\\n]*");
510 cleaned = from_utf8(lyx::regex_replace(to_utf8(cleaned), reg, string("")));
512 if (!cleaned.empty()) {
514 openParTag(xs, &*par, prevpar);
516 xs << XMLStream::ESCAPE_NONE << os2.str();
519 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
524 void makeEnvironment(
527 OutputParams const &runparams,
529 ParagraphList::const_iterator const & par)
531 auto const end = text.paragraphs().end();
533 // Output the opening tag for this environment, but only if it has not been previously opened (condition
534 // implemented in openParTag).
535 auto prevpar = text.paragraphs().getParagraphBefore(par);
536 openParTag(xs, &*par, prevpar); // TODO: switch in layout for par/block?
538 // Generate the contents of this environment. There is a special case if this is like some environment.
539 Layout const & style = par->layout();
540 if (style.latextype == LATEX_COMMAND) {
541 // Nothing to do (otherwise, infinite loops).
542 } else if (style.latextype == LATEX_ENVIRONMENT ||
543 style.latextype == LATEX_LIST_ENVIRONMENT ||
544 style.latextype == LATEX_ITEM_ENVIRONMENT) {
545 // Open a wrapper tag if needed.
546 if (style.docbookitemwrappertag() != "NONE") {
547 xs << xml::StartTag(style.docbookitemwrappertag(), style.docbookitemwrapperattr());
551 // Generate the label, if need be. If it is taken from the text, sep != 0 and corresponds to the first
552 // character after the label.
554 if (style.labeltype != LABEL_NO_LABEL && style.docbookitemlabeltag() != "NONE") {
555 // At least one condition must be met:
556 // - this environment is not a list
557 // - if this is a list, the label must not be manual (i.e. it must be taken from the layout)
558 if (style.latextype != LATEX_LIST_ENVIRONMENT || style.labeltype != LABEL_MANUAL) {
559 // Usual cases: maybe there is something specified at the layout level. Highly unlikely, though.
560 docstring const lbl = par->params().labelString();
565 openLabelTag(xs, style);
567 closeLabelTag(xs, style);
570 // Only variablelist gets here (or similar items defined as an extension in the layout).
571 openLabelTag(xs, style);
572 sep = par->firstWordDocBook(xs, runparams);
573 closeLabelTag(xs, style);
577 // Maybe the item is completely empty, i.e. if the first word ends at the end of the current paragraph
578 // AND if the next paragraph doesn't have the same depth (if there is such a paragraph).
579 // Common case: there is only the first word on the line, but there is a nested list instead
581 bool emptyItem = false;
582 if (sep == par->size()) { // If the separator is already at the end of this paragraph...
585 if (next_par == text.paragraphs().end()) // There is no next paragraph.
587 else // There is a next paragraph: check depth.
588 emptyItem = par->params().depth() >= next_par->params().depth();
592 // Avoid having an empty item, this is not valid DocBook. A single character is enough to force
593 // generation of a full <para>.
594 // TODO: this always worked only by magic...
597 // Generate the rest of the paragraph, if need be.
598 par->simpleDocBookOnePar(buf, xs, runparams, text.outerFont(std::distance(text.paragraphs().begin(), par)),
602 makeAny(text, buf, xs, runparams, par);
605 // Close the environment.
608 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr); // TODO: switch in layout for par/block?
615 OutputParams const & runparams,
617 ParagraphList::const_iterator const & par)
619 // Unlike XHTML, no need for labels, as they are handled by DocBook tags.
620 auto const begin = text.paragraphs().begin();
621 auto const end = text.paragraphs().end();
625 // Generate this command.
626 auto prevpar = text.paragraphs().getParagraphBefore(par);
627 openParTag(xs, &*par, prevpar);
629 par->simpleDocBookOnePar(buf, xs, runparams,
630 text.outerFont(distance(begin, par)));
632 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
640 OutputParams const &ourparams,
641 ParagraphList::const_iterator par)
643 switch (par->layout().latextype) {
645 makeCommand(buf, xs, ourparams, text, par);
647 case LATEX_ENVIRONMENT:
648 case LATEX_LIST_ENVIRONMENT:
649 case LATEX_ITEM_ENVIRONMENT:
650 makeEnvironment(buf, xs, ourparams, text, par);
652 case LATEX_PARAGRAPH:
653 makeParagraph(buf, xs, ourparams, text, par);
655 case LATEX_BIB_ENVIRONMENT:
656 makeParagraphBibliography(buf, xs, ourparams, text, par);
661 } // end anonymous namespace
664 using DocBookDocumentSectioning = tuple<bool, pit_type>;
667 struct DocBookInfoTag
669 const set<pit_type> shouldBeInInfo;
670 const set<pit_type> mustBeInInfo;
671 const set<pit_type> abstract;
675 DocBookInfoTag(const set<pit_type> & shouldBeInInfo, const set<pit_type> & mustBeInInfo,
676 const set<pit_type> & abstract, pit_type bpit, pit_type epit) :
677 shouldBeInInfo(shouldBeInInfo), mustBeInInfo(mustBeInInfo), abstract(abstract),
678 bpit(bpit), epit(epit) {}
682 DocBookDocumentSectioning hasDocumentSectioning(ParagraphList const ¶graphs, pit_type bpit, pit_type const epit) {
683 bool documentHasSections = false;
685 while (bpit < epit) {
686 Layout const &style = paragraphs[bpit].layout();
687 documentHasSections |= style.category() == from_utf8("Sectioning");
689 if (documentHasSections)
693 // Paragraphs before the first section: [ runparams.par_begin ; eppit )
695 return make_tuple(documentHasSections, bpit);
699 bool hasOnlyNotes(Paragraph const & par)
701 // Precondition: the paragraph is not empty. Otherwise, the function will always return true...
702 for (int i = 0; i < par.size(); ++i)
703 // If you find something that is not an inset (like actual text) or an inset that is not a note,
705 if (!par.isInset(i) || !dynamic_cast<InsetNote *>(par.insetList().get(i)))
711 DocBookInfoTag getParagraphsWithInfo(ParagraphList const ¶graphs, pit_type bpit, pit_type const epit) {
712 set<pit_type> shouldBeInInfo;
713 set<pit_type> mustBeInInfo;
714 set<pit_type> abstract;
716 // Find the first non empty paragraph by mutating bpit.
717 while (bpit < epit) {
718 Paragraph const &par = paragraphs[bpit];
719 if (par.empty() || hasOnlyNotes(par))
725 // Find the last info-like paragraph.
726 pit_type cpit = bpit;
727 bool hasAbstractLayout = false;
728 while (cpit < epit) {
729 // Skip paragraphs only containing one note.
730 Paragraph const & par = paragraphs[cpit];
731 if (hasOnlyNotes(par)) {
736 if (par.layout().docbookabstract())
737 hasAbstractLayout = true;
739 // Based on layout information, store this paragraph in one set: should be in <info>, must be.
740 Layout const &style = par.layout();
742 if (style.docbookininfo() == "always") {
743 mustBeInInfo.emplace(cpit);
744 } else if (style.docbookininfo() == "maybe") {
745 shouldBeInInfo.emplace(cpit);
747 // Hypothesis: the <info> parts should be grouped together near the beginning bpit.
748 // There may be notes in between, but nothing else.
753 // Now, cpit points to the last paragraph that has things that could go in <info>.
754 // bpit is the beginning of the <info> part.
756 // Go once again through the list of paragraphs to find the abstract. If there is an abstract
757 // layout, only consider it. Otherwise, an abstract is just a sequence of paragraphs with text.
758 if (hasAbstractLayout) {
760 while (pit < cpit) { // Don't overshoot the <info> part.
761 if (paragraphs[pit].layout().docbookabstract())
762 abstract.emplace(pit);
766 pit_type lastAbstract = epit + 1; // A nonsensical value.
767 docstring lastAbstractLayout;
770 while (pit < cpit) { // Don't overshoot the <info> part.
771 const Paragraph & par = paragraphs.at(pit);
772 if (!par.insetList().empty()) {
773 for (const auto &i : par.insetList()) {
774 if (i.inset->getText(0) != nullptr) {
775 if (lastAbstract == epit + 1) {
776 // First paragraph that matches the heuristic definition of abstract.
778 lastAbstractLayout = par.layout().name();
779 } else if (pit > lastAbstract + 1 || par.layout().name() != lastAbstractLayout) {
780 // This is either too far from the last abstract paragraph or doesn't
781 // have the right layout name, BUT there has already been an abstract
782 // in this document: done with detecting the abstract.
783 goto done; // Easier to get out of two nested loops.
786 abstract.emplace(pit);
796 return DocBookInfoTag(shouldBeInInfo, mustBeInInfo, abstract, bpit, cpit);
800 void outputDocBookInfo(
804 OutputParams const & runparams,
805 ParagraphList const & paragraphs,
806 DocBookInfoTag const & info)
808 // Perform an additional check on the abstract. Sometimes, there are many paragraphs that should go
809 // into the abstract, but none generates actual content. Thus, first generate to a temporary stream,
810 // then only create the <abstract> tag if these paragraphs generate some content.
811 // This check must be performed *before* a decision on whether or not to output <info> is made.
812 bool hasAbstract = !info.abstract.empty();
815 // Generate the abstract XML into a string before further checks.
816 odocstringstream os2;
819 auto bpit = *std::min_element(info.abstract.begin(), info.abstract.end());
820 auto epit = 1 + *std::max_element(info.abstract.begin(), info.abstract.end());
821 // info.abstract is inclusive, epit is exclusive, hence +1 for looping.
823 while (bpit < epit) {
824 makeAny(text, buf, xs2, runparams, paragraphs.iterator_at(bpit));
829 // Actually output the abstract if there is something to do. Don't count line feeds or spaces in this,
830 // even though they must be properly output if there is some abstract.
831 abstract = os2.str();
832 static const lyx::regex reg("[ \\r\\n]*");
833 docstring abstractContent = from_utf8(lyx::regex_replace(to_utf8(abstract), reg, string("")));
835 // Nothing? Then there is no abstract!
836 if (abstractContent.empty())
840 // The abstract must go in <info>. Otherwise, decide whether to open <info> based on the layouts.
841 bool needInfo = !info.mustBeInInfo.empty() || hasAbstract;
843 // Start the <info> tag if required.
845 xs.startDivision(false);
846 xs << xml::StartTag("info");
850 // Output the elements that should go in <info>, before and after the abstract.
851 for (auto pit : info.shouldBeInInfo) { // Typically, the title: these elements are so important and ubiquitous
852 // that mandating a wrapper like <info> would repel users. Thus, generate them first.
853 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
855 for (auto pit : info.mustBeInInfo) {
856 if (info.abstract.find(pit) == info.abstract.end()) // The abstract must be in info, but is dealt with after.
857 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
860 // Always output the abstract as the last item of the <info>, as it requires special treatment (especially if
861 // it contains several paragraphs that are empty).
863 // string tag = paragraphs[*info.abstract.begin()].layout().docbookforceabstracttag();
864 // if (tag == "NONE")
867 // xs << xml::StartTag(tag);
869 xs << XMLStream::ESCAPE_NONE << abstract;
870 // xs << xml::EndTag(tag);
874 // End the <info> tag if it was started.
876 xs << xml::EndTag("info");
883 void docbookFirstParagraphs(
887 OutputParams const &runparams,
890 // Handle the beginning of the document, supposing it has sections.
891 // Major role: output the first <info> tag.
893 ParagraphList const ¶graphs = text.paragraphs();
894 pit_type bpit = runparams.par_begin;
895 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, epit);
896 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
900 void docbookSimpleAllParagraphs(
904 OutputParams const & runparams)
906 // Handle the given text, supposing it has no sections (i.e. a "simple" text). The input may vary in length
907 // between a single paragraph to a whole document.
909 // First, the <info> tag.
910 ParagraphList const ¶graphs = text.paragraphs();
911 pit_type bpit = runparams.par_begin;
912 pit_type const epit = runparams.par_end;
913 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, epit);
914 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
916 // Then, the content. It starts where the <info> ends.
918 while (bpit < epit) {
919 auto par = paragraphs.iterator_at(bpit);
920 if (!hasOnlyNotes(*par))
921 makeAny(text, buf, xs, runparams, par);
927 void docbookParagraphs(Text const &text,
930 OutputParams const &runparams) {
931 ParagraphList const ¶graphs = text.paragraphs();
932 if (runparams.par_begin == runparams.par_end) {
933 runparams.par_begin = 0;
934 runparams.par_end = paragraphs.size();
936 pit_type bpit = runparams.par_begin;
937 pit_type const epit = runparams.par_end;
940 xs << XMLStream::ESCAPE_NONE << "<!-- DocBook output error! -->\n";
944 std::stack<std::pair<int, string>> headerLevels; // Used to determine when to open/close sections: store the depth
945 // of the section and the tag that was used to open it.
947 // Detect whether the document contains sections. If there are no sections, there can be no automatically
948 // discovered abstract.
949 bool documentHasSections;
951 tie(documentHasSections, eppit) = hasDocumentSectioning(paragraphs, bpit, epit);
953 if (documentHasSections) {
954 docbookFirstParagraphs(text, buf, xs, runparams, eppit);
957 docbookSimpleAllParagraphs(text, buf, xs, runparams);
961 bool currentlyInAppendix = false;
963 while (bpit < epit) {
964 OutputParams ourparams = runparams;
966 auto par = paragraphs.iterator_at(bpit);
967 if (par->params().startOfAppendix())
968 currentlyInAppendix = true;
969 Layout const &style = par->layout();
970 ParagraphList::const_iterator const lastStartedPar = par;
971 ParagraphList::const_iterator send;
973 if (hasOnlyNotes(*par)) {
978 // Think about adding <section> and/or </section>s.
979 const bool isLayoutSectioning = style.category() == from_utf8("Sectioning");
980 if (isLayoutSectioning) {
981 int level = style.toclevel;
983 // Need to close a previous section if it has the same level or a higher one (close <section> if opening a <h2>
984 // after a <h2>, <h3>, <h4>, <h5> or <h6>). More examples:
985 // - current: h2; back: h1; do not close any <section>
986 // - current: h1; back: h2; close two <section> (first the <h2>, then the <h1>, so a new <h1> can come)
987 while (!headerLevels.empty() && level <= headerLevels.top().first) {
988 int stackLevel = headerLevels.top().first;
989 docstring stackTag = from_utf8("</" + headerLevels.top().second + ">");
992 // Output the tag only if it corresponds to a legit section.
993 if (stackLevel != Layout::NOT_IN_TOC)
994 xs << XMLStream::ESCAPE_NONE << stackTag << xml::CR();
997 // Open the new section: first push it onto the stack, then output it in DocBook.
998 string sectionTag = (currentlyInAppendix && style.docbooksectiontag() == "chapter") ?
999 "appendix" : style.docbooksectiontag();
1000 headerLevels.push(std::make_pair(level, sectionTag));
1002 // Some sectioning-like elements should not be output (such as FrontMatter).
1003 if (level != Layout::NOT_IN_TOC) {
1004 // Look for a label in the title, i.e. a InsetLabel as a child.
1005 docstring id = docstring();
1006 for (pos_type i = 0; i < par->size(); ++i) {
1007 Inset const *inset = par->getInset(i);
1009 if (auto label = dynamic_cast<InsetLabel const *>(inset)) {
1010 // Generate the attributes for the section if need be.
1011 id += "xml:id=\"" + xml::cleanID(label->screenLabel()) + "\"";
1013 // Don't output the ID as a DocBook <anchor>.
1014 ourparams.docbook_anchors_to_ignore.emplace(label->screenLabel());
1016 // Cannot have multiple IDs per tag.
1022 // Write the open tag for this section.
1023 docstring tag = from_utf8("<" + sectionTag);
1025 tag += from_utf8(" ") + id;
1026 tag += from_utf8(">");
1027 xs << XMLStream::ESCAPE_NONE << tag;
1032 // Close all sections before the bibliography.
1033 // TODO: Only close all when the bibliography is at the end of the document? Or force to output the bibliography at the end of the document? Or don't care (as allowed by DocBook)?
1034 auto insetsLength = distance(par->insetList().begin(), par->insetList().end());
1035 if (insetsLength > 0) {
1036 Inset const *firstInset = par->getInset(0);
1037 if (firstInset && dynamic_cast<InsetBibtex const *>(firstInset)) {
1038 while (!headerLevels.empty()) {
1039 int level = headerLevels.top().first;
1040 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1043 // Output the tag only if it corresponds to a legit section.
1044 if (level != Layout::NOT_IN_TOC) {
1045 xs << XMLStream::ESCAPE_NONE << tag;
1052 // Generate this paragraph.
1053 makeAny(text, buf, xs, ourparams, par);
1057 // If need be, close <section>s, but only at the end of the document (otherwise, dealt with at the beginning
1059 while (!headerLevels.empty() && headerLevels.top().first > Layout::NOT_IN_TOC) {
1060 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1062 xs << XMLStream::ESCAPE_NONE << tag;