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"
39 using namespace lyx::support;
45 std::string fontToDocBookTag(xml::FontTypes type)
48 case xml::FontTypes::FT_EMPH:
49 case xml::FontTypes::FT_BOLD:
51 case xml::FontTypes::FT_NOUN:
53 case xml::FontTypes::FT_UBAR:
54 case xml::FontTypes::FT_WAVE:
55 case xml::FontTypes::FT_DBAR:
56 case xml::FontTypes::FT_SOUT:
57 case xml::FontTypes::FT_XOUT:
58 case xml::FontTypes::FT_ITALIC:
59 case xml::FontTypes::FT_UPRIGHT:
60 case xml::FontTypes::FT_SLANTED:
61 case xml::FontTypes::FT_SMALLCAPS:
62 case xml::FontTypes::FT_ROMAN:
63 case xml::FontTypes::FT_SANS:
65 case xml::FontTypes::FT_TYPE:
67 case xml::FontTypes::FT_SIZE_TINY:
68 case xml::FontTypes::FT_SIZE_SCRIPT:
69 case xml::FontTypes::FT_SIZE_FOOTNOTE:
70 case xml::FontTypes::FT_SIZE_SMALL:
71 case xml::FontTypes::FT_SIZE_NORMAL:
72 case xml::FontTypes::FT_SIZE_LARGE:
73 case xml::FontTypes::FT_SIZE_LARGER:
74 case xml::FontTypes::FT_SIZE_LARGEST:
75 case xml::FontTypes::FT_SIZE_HUGE:
76 case xml::FontTypes::FT_SIZE_HUGER:
77 case xml::FontTypes::FT_SIZE_INCREASE:
78 case xml::FontTypes::FT_SIZE_DECREASE:
86 string fontToRole(xml::FontTypes type)
88 // Specific fonts are achieved with roles. The only common ones are "" for basic emphasis,
89 // and "bold"/"strong" for bold. With some specific options, other roles are copied into
90 // HTML output (via the DocBook XSLT sheets); otherwise, if not recognised, they are just ignored.
91 // Hence, it is not a problem to have many roles by default here.
92 // See https://www.sourceware.org/ml/docbook/2003-05/msg00269.html
94 case xml::FontTypes::FT_ITALIC:
95 case xml::FontTypes::FT_EMPH:
97 case xml::FontTypes::FT_BOLD:
99 case xml::FontTypes::FT_NOUN: // Outputs a <person>
100 case xml::FontTypes::FT_TYPE: // Outputs a <code>
102 case xml::FontTypes::FT_UBAR:
105 // All other roles are non-standard for DocBook.
107 case xml::FontTypes::FT_WAVE:
109 case xml::FontTypes::FT_DBAR:
111 case xml::FontTypes::FT_SOUT:
113 case xml::FontTypes::FT_XOUT:
115 case xml::FontTypes::FT_UPRIGHT:
117 case xml::FontTypes::FT_SLANTED:
119 case xml::FontTypes::FT_SMALLCAPS:
121 case xml::FontTypes::FT_ROMAN:
123 case xml::FontTypes::FT_SANS:
125 case xml::FontTypes::FT_SIZE_TINY:
127 case xml::FontTypes::FT_SIZE_SCRIPT:
128 return "size_script";
129 case xml::FontTypes::FT_SIZE_FOOTNOTE:
130 return "size_footnote";
131 case xml::FontTypes::FT_SIZE_SMALL:
133 case xml::FontTypes::FT_SIZE_NORMAL:
134 return "size_normal";
135 case xml::FontTypes::FT_SIZE_LARGE:
137 case xml::FontTypes::FT_SIZE_LARGER:
138 return "size_larger";
139 case xml::FontTypes::FT_SIZE_LARGEST:
140 return "size_largest";
141 case xml::FontTypes::FT_SIZE_HUGE:
143 case xml::FontTypes::FT_SIZE_HUGER:
145 case xml::FontTypes::FT_SIZE_INCREASE:
146 return "size_increase";
147 case xml::FontTypes::FT_SIZE_DECREASE:
148 return "size_decrease";
155 string fontToAttribute(xml::FontTypes type) {
156 // If there is a role (i.e. nonstandard use of a tag), output the attribute. Otherwise, the sheer tag is sufficient
158 string role = fontToRole(type);
160 return "role='" + role + "'";
167 // Convenience functions to open and close tags. First, very low-level ones to ensure a consistent new-line behaviour.
171 // Contents of the block.
176 // <paratag>Contents of the paragraph.</paratag>
179 // Content before<inlinetag>Contents of the paragraph.</inlinetag>Content after
181 void openInlineTag(XMLStream & xs, const std::string & tag, const std::string & attr)
183 xs << xml::StartTag(tag, attr);
187 void closeInlineTag(XMLStream & xs, const std::string & tag)
189 xs << xml::EndTag(tag);
193 void openParTag(XMLStream & xs, const std::string & tag, const std::string & attr)
195 if (!xs.isLastTagCR())
197 xs << xml::StartTag(tag, attr);
201 void closeParTag(XMLStream & xs, const std::string & tag)
203 xs << xml::EndTag(tag);
208 void openBlockTag(XMLStream & xs, const std::string & tag, const std::string & attr)
210 if (!xs.isLastTagCR())
212 xs << xml::StartTag(tag, attr);
217 void closeBlockTag(XMLStream & xs, const std::string & tag)
219 if (!xs.isLastTagCR())
221 xs << xml::EndTag(tag);
226 void openTag(XMLStream & xs, const std::string & tag, const std::string & attr, const std::string & tagtype)
228 if (tag.empty() || tag == "NONE")
231 if (tag == "para" || tagtype == "paragraph") // Special case for <para>: always considered as a paragraph.
232 openParTag(xs, tag, attr);
233 else if (tagtype == "block")
234 openBlockTag(xs, tag, attr);
235 else if (tagtype == "inline")
236 openInlineTag(xs, tag, attr);
238 xs.writeError("Unrecognised tag type '" + tagtype + "' for '" + tag + " " + attr + "'");
242 void closeTag(XMLStream & xs, const std::string & tag, const std::string & tagtype)
244 if (tag.empty() || tag == "NONE")
247 if (tag == "para" || tagtype == "paragraph") // Special case for <para>: always considered as a paragraph.
248 closeParTag(xs, tag);
249 else if (tagtype == "block")
250 closeBlockTag(xs, tag);
251 else if (tagtype == "inline")
252 closeInlineTag(xs, tag);
254 xs.writeError("Unrecognised tag type '" + tagtype + "' for '" + tag + "'");
258 // Higher-level convenience functions.
260 void openParTag(XMLStream & xs, const Paragraph * par, const Paragraph * prevpar)
262 Layout const & lay = par->layout();
267 // When should the wrapper be opened here? Only if the previous paragraph has the SAME wrapper tag
268 // (usually, they won't have the same layout) and the CURRENT one allows merging.
269 // The main use case is author information in several paragraphs: if the name of the author is the
270 // first paragraph of an author, then merging with the previous tag does not make sense. Say the
271 // next paragraph is the affiliation, then it should be output in the same <author> tag (different
272 // layout, same wrapper tag).
273 bool openWrapper = lay.docbookwrappertag() != "NONE";
274 if (prevpar != nullptr) {
275 Layout const & prevlay = prevpar->layout();
276 if (prevlay.docbookwrappertag() != "NONE") {
277 openWrapper = prevlay.docbookwrappertag() == lay.docbookwrappertag()
278 && !lay.docbookwrappermergewithprevious();
284 openTag(xs, lay.docbookwrappertag(), lay.docbookwrapperattr(), lay.docbookwrappertagtype());
286 const string & tag = lay.docbooktag();
288 auto xmltag = xml::ParTag(tag, lay.docbookattr());
289 if (!xs.isTagOpen(xmltag, 1)) // Don't nest a paragraph directly in a paragraph.
290 // TODO: required or not?
291 // TODO: avoid creating a ParTag object just for this query...
292 openTag(xs, lay.docbooktag(), lay.docbookattr(), lay.docbooktagtype());
295 openTag(xs, lay.docbookitemtag(), lay.docbookitemattr(), lay.docbookitemtagtype());
296 openTag(xs, lay.docbookiteminnertag(), lay.docbookiteminnerattr(), lay.docbookiteminnertagtype());
300 void closeParTag(XMLStream & xs, Paragraph const * par, Paragraph const * nextpar)
305 // See comment in openParTag.
306 Layout const & lay = par->layout();
307 bool closeWrapper = lay.docbookwrappertag() != "NONE";
308 if (nextpar != nullptr) {
309 Layout const & nextlay = nextpar->layout();
310 if (nextlay.docbookwrappertag() != "NONE") {
311 closeWrapper = nextlay.docbookwrappertag() == lay.docbookwrappertag()
312 && !nextlay.docbookwrappermergewithprevious();
317 closeTag(xs, lay.docbookiteminnertag(), lay.docbookiteminnertagtype());
318 closeTag(xs, lay.docbookitemtag(), lay.docbookitemtagtype());
319 closeTag(xs, lay.docbooktag(), lay.docbooktagtype());
321 closeTag(xs, lay.docbookwrappertag(), lay.docbookwrappertagtype());
325 void openLabelTag(XMLStream & xs, Layout const & lay) // Mostly for definition lists.
327 openTag(xs, lay.docbookitemlabeltag(), lay.docbookitemlabelattr(), lay.docbookitemlabeltagtype());
331 void closeLabelTag(XMLStream & xs, Layout const & lay)
333 closeTag(xs, lay.docbookitemlabeltag(), lay.docbookitemlabeltagtype());
337 void openItemTag(XMLStream & xs, Layout const & lay)
339 openTag(xs, lay.docbookitemtag(), lay.docbookitemattr(), lay.docbookitemtagtype());
343 void closeItemTag(XMLStream & xs, Layout const & lay)
345 closeTag(xs, lay.docbookitemtag(), lay.docbookitemtagtype());
353 OutputParams const &,
354 ParagraphList::const_iterator);
357 void makeParagraphBibliography(
360 OutputParams const & runparams,
362 ParagraphList::const_iterator const & par)
364 // If this is the first paragraph in a bibliography, open the bibliography tag.
365 auto pbegin_before = text.paragraphs().getParagraphBefore(par);
366 if (pbegin_before->layout().latextype != LATEX_BIB_ENVIRONMENT) {
367 xs << xml::StartTag("bibliography");
371 // Start the precooked bibliography entry. This is very much like opening a paragraph tag.
372 // Don't forget the citation ID!
374 for (auto i = 0; i < par->size(); ++i) {
375 Inset const *ip = par->getInset(i);
378 if (const auto * bibitem = dynamic_cast<const InsetBibitem*>(ip)) {
379 attr = from_utf8("xml:id='") + bibitem->getParam("key") + from_utf8("'");
383 xs << xml::StartTag(from_utf8("bibliomixed"), attr);
385 // Generate the entry.
386 auto const begin = text.paragraphs().begin();
387 par->simpleDocBookOnePar(buf, xs, runparams, text.outerFont(std::distance(begin, par)), true, true, 0);
389 // End the precooked bibliography entry.
390 xs << xml::EndTag("bibliomixed");
393 // If this is the last paragraph in a bibliography, close the bibliography tag.
394 auto const end = text.paragraphs().end();
395 bool endBibliography = par == end;
396 if (!endBibliography) {
399 endBibliography = par->layout().latextype != LATEX_BIB_ENVIRONMENT;
402 if (endBibliography) {
403 xs << xml::EndTag("bibliography");
412 OutputParams const & runparams,
414 ParagraphList::const_iterator const & par)
416 auto const begin = text.paragraphs().begin();
417 auto const end = text.paragraphs().end();
418 auto prevpar = text.paragraphs().getParagraphBefore(par);
420 // We want to open the paragraph tag if:
421 // (i) the current layout permits multiple paragraphs
422 // (ii) we are either not already inside a paragraph (HTMLIsBlock) OR
423 // we are, but this is not the first paragraph
425 // But there is also a special case, and we first see whether we are in it.
426 // We do not want to open the paragraph tag if this paragraph contains
427 // only one item, and that item is "inline", i.e., not HTMLIsBlock (such
428 // as a branch). On the other hand, if that single item has a font change
429 // applied to it, then we still do need to open the paragraph.
431 // Obviously, this is very fragile. The main reason we need to do this is
432 // because of branches, e.g., a branch that contains an entire new section.
433 // We do not really want to wrap that whole thing in a <div>...</div>.
434 bool special_case = false;
435 Inset const *specinset = par->size() == 1 ? par->getInset(0) : nullptr;
436 if (specinset && !specinset->getLayout().htmlisblock()) { // TODO: Convert htmlisblock to a DocBook parameter?
437 Layout const &style = par->layout();
438 FontInfo const first_font = style.labeltype == LABEL_MANUAL ?
439 style.labelfont : style.font;
440 FontInfo const our_font =
441 par->getFont(buf.masterBuffer()->params(), 0,
442 text.outerFont(std::distance(begin, par))).fontInfo();
444 if (first_font == our_font)
448 // Plain layouts must be ignored.
449 if (!special_case && buf.params().documentClass().isPlainLayout(par->layout()) && !runparams.docbook_force_pars)
451 // TODO: Could get rid of this with a DocBook equivalent to htmlisblock?
452 if (!special_case && par->size() == 1 && par->getInset(0)) {
453 Inset const * firstInset = par->getInset(0);
455 // Floats cannot be in paragraphs.
456 special_case = to_utf8(firstInset->layoutName()).substr(0, 6) == "Float:";
458 // Bibliographies cannot be in paragraphs.
459 if (!special_case && firstInset->asInsetCommand())
460 special_case = firstInset->asInsetCommand()->params().getCmdName() == "bibtex";
462 // Equations do not deserve their own paragraph (DocBook allows them outside paragraphs).
463 if (!special_case && firstInset->asInsetMath())
466 // ERTs are in comments, not paragraphs.
467 if (!special_case && firstInset->lyxCode() == lyx::ERT_CODE)
470 // Listings should not get into their own paragraph.
471 if (!special_case && firstInset->lyxCode() == lyx::LISTINGS_CODE)
475 bool const open_par = runparams.docbook_make_pars
476 && !runparams.docbook_in_par
479 // We want to issue the closing tag if either:
480 // (i) We opened it, and either docbook_in_par is false,
481 // or we're not in the last paragraph, anyway.
482 // (ii) We didn't open it and docbook_in_par is true,
483 // but we are in the first par, and there is a next par.
486 bool const close_par = open_par && (!runparams.docbook_in_par);
488 // Determine if this paragraph has some real content. Things like new pages are not caught
489 // by Paragraph::empty(), even though they do not generate anything useful in DocBook.
490 // Thus, remove all spaces (including new lines: \r, \n) before checking for emptiness.
491 odocstringstream os2;
493 par->simpleDocBookOnePar(buf, xs2, runparams, text.outerFont(distance(begin, par)), open_par, close_par, 0);
495 docstring cleaned = os2.str();
496 cleaned.erase(std::remove_if(cleaned.begin(), cleaned.end(), ::isspace), cleaned.end());
498 if (!cleaned.empty()) {
500 openParTag(xs, &*par, prevpar);
502 xs << XMLStream::ESCAPE_NONE << os2.str();
505 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
510 void makeEnvironment(
513 OutputParams const &runparams,
515 ParagraphList::const_iterator const & par)
517 auto const end = text.paragraphs().end();
519 // Output the opening tag for this environment, but only if it has not been previously opened (condition
520 // implemented in openParTag).
521 auto prevpar = text.paragraphs().getParagraphBefore(par);
522 openParTag(xs, &*par, prevpar); // TODO: switch in layout for par/block?
524 // Generate the contents of this environment. There is a special case if this is like some environment.
525 Layout const & style = par->layout();
526 if (style.latextype == LATEX_COMMAND) {
527 // Nothing to do (otherwise, infinite loops).
528 } else if (style.latextype == LATEX_ENVIRONMENT ||
529 style.latextype == LATEX_LIST_ENVIRONMENT ||
530 style.latextype == LATEX_ITEM_ENVIRONMENT) {
531 // Open a wrapper tag if needed.
532 if (style.docbookitemwrappertag() != "NONE") {
533 xs << xml::StartTag(style.docbookitemwrappertag(), style.docbookitemwrapperattr());
537 // Generate the label, if need be. If it is taken from the text, sep != 0 and corresponds to the first
538 // character after the label.
540 if (style.labeltype != LABEL_NO_LABEL && style.docbookitemlabeltag() != "NONE") {
541 // At least one condition must be met:
542 // - this environment is not a list
543 // - if this is a list, the label must not be manual (i.e. it must be taken from the layout)
544 if (style.latextype != LATEX_LIST_ENVIRONMENT || style.labeltype != LABEL_MANUAL) {
545 // Usual cases: maybe there is something specified at the layout level. Highly unlikely, though.
546 docstring const lbl = par->params().labelString();
551 openLabelTag(xs, style);
553 closeLabelTag(xs, style);
556 // Only variablelist gets here (or similar items defined as an extension in the layout).
557 openLabelTag(xs, style);
558 sep = par->firstWordDocBook(xs, runparams);
559 closeLabelTag(xs, style);
563 // Maybe the item is completely empty, i.e. if the first word ends at the end of the current paragraph
564 // AND if the next paragraph doesn't have the same depth (if there is such a paragraph).
565 // Common case: there is only the first word on the line, but there is a nested list instead
567 bool emptyItem = false;
568 if (sep == par->size()) { // If the separator is already at the end of this paragraph...
571 if (next_par == text.paragraphs().end()) // There is no next paragraph.
573 else // There is a next paragraph: check depth.
574 emptyItem = par->params().depth() >= next_par->params().depth();
578 // Avoid having an empty item, this is not valid DocBook. A single character is enough to force
579 // generation of a full <para>.
580 // TODO: this always worked only by magic...
583 // Generate the rest of the paragraph, if need be.
584 par->simpleDocBookOnePar(buf, xs, runparams, text.outerFont(std::distance(text.paragraphs().begin(), par)),
588 makeAny(text, buf, xs, runparams, par);
591 // Close the environment.
594 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr); // TODO: switch in layout for par/block?
601 OutputParams const & runparams,
603 ParagraphList::const_iterator const & par)
605 // Unlike XHTML, no need for labels, as they are handled by DocBook tags.
606 auto const begin = text.paragraphs().begin();
607 auto const end = text.paragraphs().end();
611 // Generate this command.
612 auto prevpar = text.paragraphs().getParagraphBefore(par);
613 openParTag(xs, &*par, prevpar);
615 par->simpleDocBookOnePar(buf, xs, runparams,
616 text.outerFont(distance(begin, par)));
618 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
626 OutputParams const &ourparams,
627 ParagraphList::const_iterator par)
629 switch (par->layout().latextype) {
631 makeCommand(buf, xs, ourparams, text, par);
633 case LATEX_ENVIRONMENT:
634 case LATEX_LIST_ENVIRONMENT:
635 case LATEX_ITEM_ENVIRONMENT:
636 makeEnvironment(buf, xs, ourparams, text, par);
638 case LATEX_PARAGRAPH:
639 makeParagraph(buf, xs, ourparams, text, par);
641 case LATEX_BIB_ENVIRONMENT:
642 makeParagraphBibliography(buf, xs, ourparams, text, par);
648 using DocBookDocumentSectioning = tuple<bool, pit_type>;
651 struct DocBookInfoTag
653 const set<pit_type> shouldBeInInfo;
654 const set<pit_type> mustBeInInfo;
655 const set<pit_type> abstract;
659 DocBookInfoTag(const set<pit_type> & shouldBeInInfo, const set<pit_type> & mustBeInInfo,
660 const set<pit_type> & abstract, pit_type bpit, pit_type epit) :
661 shouldBeInInfo(shouldBeInInfo), mustBeInInfo(mustBeInInfo), abstract(abstract),
662 bpit(bpit), epit(epit) {}
666 DocBookDocumentSectioning hasDocumentSectioning(ParagraphList const ¶graphs, pit_type bpit, pit_type const epit) {
667 bool documentHasSections = false;
669 while (bpit < epit) {
670 Layout const &style = paragraphs[bpit].layout();
671 documentHasSections |= style.category() == from_utf8("Sectioning");
673 if (documentHasSections)
677 // Paragraphs before the first section: [ runparams.par_begin ; eppit )
679 return make_tuple(documentHasSections, bpit);
683 bool hasOnlyNotes(Paragraph const & par)
685 // Precondition: the paragraph is not empty. Otherwise, the function will always return true...
686 for (int i = 0; i < par.size(); ++i)
687 // If you find something that is not an inset (like actual text) or an inset that is not a note,
689 if (!par.isInset(i) || !dynamic_cast<InsetNote *>(par.insetList().get(i)))
695 DocBookInfoTag getParagraphsWithInfo(ParagraphList const ¶graphs, pit_type bpit, pit_type const epit) {
696 set<pit_type> shouldBeInInfo;
697 set<pit_type> mustBeInInfo;
698 set<pit_type> abstract;
700 // Find the first non empty paragraph by mutating bpit.
701 while (bpit < epit) {
702 Paragraph const &par = paragraphs[bpit];
703 if (par.empty() || hasOnlyNotes(par))
709 // Find the last info-like paragraph.
710 pit_type cpit = bpit;
711 bool hasAbstractLayout = false;
712 while (cpit < epit) {
713 // Skip paragraphs only containing one note.
714 Paragraph const & par = paragraphs[cpit];
715 if (hasOnlyNotes(par)) {
720 if (par.layout().docbookabstract())
721 hasAbstractLayout = true;
723 // Based on layout information, store this paragraph in one set: should be in <info>, must be.
724 Layout const &style = par.layout();
726 if (style.docbookininfo() == "always") {
727 mustBeInInfo.emplace(cpit);
728 } else if (style.docbookininfo() == "maybe") {
729 shouldBeInInfo.emplace(cpit);
731 // Hypothesis: the <info> parts should be grouped together near the beginning bpit.
732 // There may be notes in between, but nothing else.
737 // Now, cpit points to the last paragraph that has things that could go in <info>.
738 // bpit is the beginning of the <info> part.
740 // Go once again through the list of paragraphs to find the abstract. If there is an abstract
741 // layout, only consider it. Otherwise, an abstract is just a sequence of paragraphs with text.
742 if (hasAbstractLayout) {
744 while (pit < cpit) { // Don't overshoot the <info> part.
745 if (paragraphs[pit].layout().docbookabstract())
746 abstract.emplace(pit);
750 pit_type lastAbstract = epit + 1; // A nonsensical value.
751 docstring lastAbstractLayout;
754 while (pit < cpit) { // Don't overshoot the <info> part.
755 const Paragraph & par = paragraphs.at(pit);
756 if (!par.insetList().empty()) {
757 for (const auto &i : par.insetList()) {
758 if (i.inset->getText(0) != nullptr) {
759 if (lastAbstract == epit + 1) {
760 // First paragraph that matches the heuristic definition of abstract.
762 lastAbstractLayout = par.layout().name();
763 } else if (pit > lastAbstract + 1 || par.layout().name() != lastAbstractLayout) {
764 // This is either too far from the last abstract paragraph or doesn't
765 // have the right layout name, BUT there has already been an abstract
766 // in this document: done with detecting the abstract.
767 goto done; // Easier to get out of two nested loops.
770 abstract.emplace(pit);
780 return DocBookInfoTag(shouldBeInInfo, mustBeInInfo, abstract, bpit, cpit);
783 } // end anonymous namespace
786 xml::FontTag docbookStartFontTag(xml::FontTypes type)
788 return xml::FontTag(from_utf8(fontToDocBookTag(type)), from_utf8(fontToAttribute(type)), type);
792 xml::EndFontTag docbookEndFontTag(xml::FontTypes type)
794 return xml::EndFontTag(from_utf8(fontToDocBookTag(type)), type);
798 void outputDocBookInfo(
802 OutputParams const & runparams,
803 ParagraphList const & paragraphs,
804 DocBookInfoTag const & info)
806 // Perform an additional check on the abstract. Sometimes, there are many paragraphs that should go
807 // into the abstract, but none generates actual content. Thus, first generate to a temporary stream,
808 // then only create the <abstract> tag if these paragraphs generate some content.
809 // This check must be performed *before* a decision on whether or not to output <info> is made.
810 bool hasAbstract = !info.abstract.empty();
813 // Generate the abstract XML into a string before further checks.
814 odocstringstream os2;
817 auto bpit = *std::min_element(info.abstract.begin(), info.abstract.end());
818 auto epit = 1 + *std::max_element(info.abstract.begin(), info.abstract.end());
819 // info.abstract is inclusive, epit is exclusive, hence +1 for looping.
821 while (bpit < epit) {
822 makeAny(text, buf, xs2, runparams, paragraphs.iterator_at(bpit));
827 // Actually output the abstract if there is something to do. Don't count line feeds or spaces in this,
828 // even though they must be properly output if there is some abstract.
829 abstract = os2.str();
830 docstring cleaned = abstract;
831 cleaned.erase(std::remove_if(cleaned.begin(), cleaned.end(), ::isspace), cleaned.end());
833 // Nothing? Then there is no abstract!
838 // The abstract must go in <info>. Otherwise, decide whether to open <info> based on the layouts.
839 bool needInfo = !info.mustBeInInfo.empty() || hasAbstract;
841 // Start the <info> tag if required.
843 xs.startDivision(false);
844 xs << xml::StartTag("info");
848 // Output the elements that should go in <info>, before and after the abstract.
849 for (auto pit : info.shouldBeInInfo) { // Typically, the title: these elements are so important and ubiquitous
850 // that mandating a wrapper like <info> would repel users. Thus, generate them first.
851 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
853 for (auto pit : info.mustBeInInfo) {
854 if (info.abstract.find(pit) == info.abstract.end()) // The abstract must be in info, but is dealt with after.
855 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
858 // Always output the abstract as the last item of the <info>, as it requires special treatment (especially if
859 // it contains several paragraphs that are empty).
861 // string tag = paragraphs[*info.abstract.begin()].layout().docbookforceabstracttag();
862 // if (tag == "NONE")
865 // xs << xml::StartTag(tag);
867 xs << XMLStream::ESCAPE_NONE << abstract;
868 // xs << xml::EndTag(tag);
872 // End the <info> tag if it was started.
874 xs << xml::EndTag("info");
881 void docbookFirstParagraphs(
885 OutputParams const &runparams,
888 // Handle the beginning of the document, supposing it has sections.
889 // Major role: output the first <info> tag.
891 ParagraphList const ¶graphs = text.paragraphs();
892 pit_type bpit = runparams.par_begin;
893 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, epit);
894 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
898 void docbookSimpleAllParagraphs(
902 OutputParams const & runparams)
904 // Handle the given text, supposing it has no sections (i.e. a "simple" text). The input may vary in length
905 // between a single paragraph to a whole document.
907 // First, the <info> tag.
908 ParagraphList const ¶graphs = text.paragraphs();
909 pit_type bpit = runparams.par_begin;
910 pit_type const epit = runparams.par_end;
911 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, epit);
912 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
914 // Then, the content. It starts where the <info> ends.
916 while (bpit < epit) {
917 auto par = paragraphs.iterator_at(bpit);
918 if (!hasOnlyNotes(*par))
919 makeAny(text, buf, xs, runparams, par);
925 void docbookParagraphs(Text const &text,
928 OutputParams const &runparams) {
929 ParagraphList const ¶graphs = text.paragraphs();
930 if (runparams.par_begin == runparams.par_end) {
931 runparams.par_begin = 0;
932 runparams.par_end = paragraphs.size();
934 pit_type bpit = runparams.par_begin;
935 pit_type const epit = runparams.par_end;
938 xs << XMLStream::ESCAPE_NONE << "<!-- DocBook output error! -->\n";
942 std::stack<std::pair<int, string>> headerLevels; // Used to determine when to open/close sections: store the depth
943 // of the section and the tag that was used to open it.
945 // Detect whether the document contains sections. If there are no sections, there can be no automatically
946 // discovered abstract.
947 bool documentHasSections;
949 tie(documentHasSections, eppit) = hasDocumentSectioning(paragraphs, bpit, epit);
951 if (documentHasSections) {
952 docbookFirstParagraphs(text, buf, xs, runparams, eppit);
955 docbookSimpleAllParagraphs(text, buf, xs, runparams);
959 bool currentlyInAppendix = false;
961 while (bpit < epit) {
962 OutputParams ourparams = runparams;
964 auto par = paragraphs.iterator_at(bpit);
965 if (par->params().startOfAppendix())
966 currentlyInAppendix = true;
967 Layout const &style = par->layout();
968 ParagraphList::const_iterator const lastStartedPar = par;
969 ParagraphList::const_iterator send;
971 if (hasOnlyNotes(*par)) {
976 // Think about adding <section> and/or </section>s.
977 const bool isLayoutSectioning = style.category() == from_utf8("Sectioning");
978 if (isLayoutSectioning) {
979 int level = style.toclevel;
981 // Need to close a previous section if it has the same level or a higher one (close <section> if opening a <h2>
982 // after a <h2>, <h3>, <h4>, <h5> or <h6>). More examples:
983 // - current: h2; back: h1; do not close any <section>
984 // - current: h1; back: h2; close two <section> (first the <h2>, then the <h1>, so a new <h1> can come)
985 while (!headerLevels.empty() && level <= headerLevels.top().first) {
986 int stackLevel = headerLevels.top().first;
987 docstring stackTag = from_utf8("</" + headerLevels.top().second + ">");
990 // Output the tag only if it corresponds to a legit section.
991 if (stackLevel != Layout::NOT_IN_TOC)
992 xs << XMLStream::ESCAPE_NONE << stackTag << xml::CR();
995 // Open the new section: first push it onto the stack, then output it in DocBook.
996 string sectionTag = (currentlyInAppendix && style.docbooksectiontag() == "chapter") ?
997 "appendix" : style.docbooksectiontag();
998 headerLevels.push(std::make_pair(level, sectionTag));
1000 // Some sectioning-like elements should not be output (such as FrontMatter).
1001 if (level != Layout::NOT_IN_TOC) {
1002 // Look for a label in the title, i.e. a InsetLabel as a child.
1003 docstring id = docstring();
1004 for (pos_type i = 0; i < par->size(); ++i) {
1005 Inset const *inset = par->getInset(i);
1007 if (auto label = dynamic_cast<InsetLabel const *>(inset)) {
1008 // Generate the attributes for the section if need be.
1009 id += "xml:id=\"" + xml::cleanID(label->screenLabel()) + "\"";
1011 // Don't output the ID as a DocBook <anchor>.
1012 ourparams.docbook_anchors_to_ignore.emplace(label->screenLabel());
1014 // Cannot have multiple IDs per tag.
1020 // Write the open tag for this section.
1021 docstring tag = from_utf8("<" + sectionTag);
1023 tag += from_utf8(" ") + id;
1024 tag += from_utf8(">");
1025 xs << XMLStream::ESCAPE_NONE << tag;
1030 // Close all sections before the bibliography.
1031 // 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)?
1032 auto insetsLength = distance(par->insetList().begin(), par->insetList().end());
1033 if (insetsLength > 0) {
1034 Inset const *firstInset = par->getInset(0);
1035 if (firstInset && dynamic_cast<InsetBibtex const *>(firstInset)) {
1036 while (!headerLevels.empty()) {
1037 int level = headerLevels.top().first;
1038 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1041 // Output the tag only if it corresponds to a legit section.
1042 if (level != Layout::NOT_IN_TOC) {
1043 xs << XMLStream::ESCAPE_NONE << tag;
1050 // Generate this paragraph.
1051 makeAny(text, buf, xs, ourparams, par);
1055 // If need be, close <section>s, but only at the end of the document (otherwise, dealt with at the beginning
1057 while (!headerLevels.empty() && headerLevels.top().first > Layout::NOT_IN_TOC) {
1058 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1060 xs << XMLStream::ESCAPE_NONE << tag;