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. Concatenate the different parts of the paragraph if any.
386 auto const begin = text.paragraphs().begin();
387 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(std::distance(begin, par)), 0);
388 for (auto & parXML : pars)
389 xs << XMLStream::ESCAPE_NONE << parXML;
391 // End the precooked bibliography entry.
392 xs << xml::EndTag("bibliomixed");
395 // If this is the last paragraph in a bibliography, close the bibliography tag.
396 auto const end = text.paragraphs().end();
397 bool endBibliography = par == end;
398 if (!endBibliography) {
401 endBibliography = par->layout().latextype != LATEX_BIB_ENVIRONMENT;
404 if (endBibliography) {
405 xs << xml::EndTag("bibliography");
414 OutputParams const & runparams,
416 ParagraphList::const_iterator const & par)
418 auto const begin = text.paragraphs().begin();
419 auto const end = text.paragraphs().end();
420 auto prevpar = text.paragraphs().getParagraphBefore(par);
422 // We want to open the paragraph tag if:
423 // (i) the current layout permits multiple paragraphs
424 // (ii) we are either not already inside a paragraph (HTMLIsBlock) OR
425 // we are, but this is not the first paragraph
427 // But there is also a special case, and we first see whether we are in it.
428 // We do not want to open the paragraph tag if this paragraph contains
429 // only one item, and that item is "inline", i.e., not HTMLIsBlock (such
430 // as a branch). On the other hand, if that single item has a font change
431 // applied to it, then we still do need to open the paragraph.
433 // Obviously, this is very fragile. The main reason we need to do this is
434 // because of branches, e.g., a branch that contains an entire new section.
435 // We do not really want to wrap that whole thing in a <div>...</div>.
436 bool special_case = false;
437 Inset const *specinset = par->size() == 1 ? par->getInset(0) : nullptr;
438 if (specinset && !specinset->getLayout().htmlisblock()) { // TODO: Convert htmlisblock to a DocBook parameter?
439 Layout const &style = par->layout();
440 FontInfo const first_font = style.labeltype == LABEL_MANUAL ?
441 style.labelfont : style.font;
442 FontInfo const our_font =
443 par->getFont(buf.masterBuffer()->params(), 0,
444 text.outerFont(std::distance(begin, par))).fontInfo();
446 if (first_font == our_font)
450 // Plain layouts must be ignored.
451 if (!special_case && buf.params().documentClass().isPlainLayout(par->layout()) && !runparams.docbook_force_pars)
453 // TODO: Could get rid of this with a DocBook equivalent to htmlisblock?
454 if (!special_case && par->size() == 1 && par->getInset(0)) {
455 Inset const * firstInset = par->getInset(0);
457 // Floats cannot be in paragraphs.
458 special_case = to_utf8(firstInset->layoutName()).substr(0, 6) == "Float:";
460 // Bibliographies cannot be in paragraphs.
461 if (!special_case && firstInset->asInsetCommand())
462 special_case = firstInset->asInsetCommand()->params().getCmdName() == "bibtex";
464 // Equations do not deserve their own paragraph (DocBook allows them outside paragraphs).
465 if (!special_case && firstInset->asInsetMath())
468 // ERTs are in comments, not paragraphs.
469 if (!special_case && firstInset->lyxCode() == lyx::ERT_CODE)
472 // Listings should not get into their own paragraph.
473 if (!special_case && firstInset->lyxCode() == lyx::LISTINGS_CODE)
477 bool const open_par = runparams.docbook_make_pars
478 && !runparams.docbook_in_par
481 // We want to issue the closing tag if either:
482 // (i) We opened it, and either docbook_in_par is false,
483 // or we're not in the last paragraph, anyway.
484 // (ii) We didn't open it and docbook_in_par is true,
485 // but we are in the first par, and there is a next par.
488 bool const close_par = open_par && (!runparams.docbook_in_par);
490 // Determine if this paragraph has some real content. Things like new pages are not caught
491 // by Paragraph::empty(), even though they do not generate anything useful in DocBook.
492 // Thus, remove all spaces (including new lines: \r, \n) before checking for emptiness.
493 // std::all_of allows doing this check without having to copy the string.
494 // Open and close tags around each contained paragraph.
495 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(distance(begin, par)), 0);
496 for (auto & parXML : pars) {
497 if (!std::all_of(parXML.begin(), parXML.end(), ::isspace)) {
499 openParTag(xs, &*par, prevpar);
501 xs << XMLStream::ESCAPE_NONE << parXML;
504 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 openTag(xs, style.docbookitemwrappertag(), style.docbookitemwrapperattr(), style.docbookitemwrappertagtype());
535 // Generate the label, if need be. If it is taken from the text, sep != 0 and corresponds to the first
536 // character after the label.
538 if (style.labeltype != LABEL_NO_LABEL && style.docbookitemlabeltag() != "NONE") {
539 // At least one condition must be met:
540 // - this environment is not a list
541 // - if this is a list, the label must not be manual (i.e. it must be taken from the layout)
542 if (style.latextype != LATEX_LIST_ENVIRONMENT || style.labeltype != LABEL_MANUAL) {
543 // Usual cases: maybe there is something specified at the layout level. Highly unlikely, though.
544 docstring const lbl = par->params().labelString();
549 openLabelTag(xs, style);
551 closeLabelTag(xs, style);
554 // Only variablelist gets here (or similar items defined as an extension in the layout).
555 openLabelTag(xs, style);
556 sep = par->firstWordDocBook(xs, runparams);
557 closeLabelTag(xs, style);
561 // Maybe the item is completely empty, i.e. if the first word ends at the end of the current paragraph
562 // AND if the next paragraph doesn't have the same depth (if there is such a paragraph).
563 // Common case: there is only the first word on the line, but there is a nested list instead
565 bool emptyItem = false;
566 if (sep == par->size()) { // If the separator is already at the end of this paragraph...
569 if (next_par == text.paragraphs().end()) // There is no next paragraph.
571 else // There is a next paragraph: check depth.
572 emptyItem = par->params().depth() >= next_par->params().depth();
576 // Avoid having an empty item, this is not valid DocBook. A single character is enough to force
577 // generation of a full <para>.
578 // TODO: this always worked only by magic...
581 // Generate the rest of the paragraph, if need be. Open as many inner tags as necessary.
582 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(std::distance(text.paragraphs().begin(), par)), sep);
583 auto p = pars.begin();
585 xs << XMLStream::ESCAPE_NONE << *p;
587 if (p != pars.end()) {
588 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
589 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(), par->layout().docbookiteminnertagtype());
595 makeAny(text, buf, xs, runparams, par);
598 // Close the environment.
601 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr); // TODO: switch in layout for par/block?
608 OutputParams const & runparams,
610 ParagraphList::const_iterator const & par)
612 // Unlike XHTML, no need for labels, as they are handled by DocBook tags.
613 auto const begin = text.paragraphs().begin();
614 auto const end = text.paragraphs().end();
618 // Generate this command.
619 auto prevpar = text.paragraphs().getParagraphBefore(par);
620 openParTag(xs, &*par, prevpar);
622 auto pars = par->simpleDocBookOnePar(buf, runparams,text.outerFont(distance(begin, par)));
623 for (auto & parXML : pars)
624 // TODO: decide what to do with openParTag/closeParTag in new lines.
625 xs << XMLStream::ESCAPE_NONE << parXML;
627 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
635 OutputParams const &ourparams,
636 ParagraphList::const_iterator par)
638 switch (par->layout().latextype) {
640 makeCommand(buf, xs, ourparams, text, par);
642 case LATEX_ENVIRONMENT:
643 case LATEX_LIST_ENVIRONMENT:
644 case LATEX_ITEM_ENVIRONMENT:
645 makeEnvironment(buf, xs, ourparams, text, par);
647 case LATEX_PARAGRAPH:
648 makeParagraph(buf, xs, ourparams, text, par);
650 case LATEX_BIB_ENVIRONMENT:
651 makeParagraphBibliography(buf, xs, ourparams, text, par);
657 using DocBookDocumentSectioning = tuple<bool, pit_type>;
660 struct DocBookInfoTag
662 const set<pit_type> shouldBeInInfo;
663 const set<pit_type> mustBeInInfo;
664 const set<pit_type> abstract;
668 DocBookInfoTag(const set<pit_type> & shouldBeInInfo, const set<pit_type> & mustBeInInfo,
669 const set<pit_type> & abstract, pit_type bpit, pit_type epit) :
670 shouldBeInInfo(shouldBeInInfo), mustBeInInfo(mustBeInInfo), abstract(abstract),
671 bpit(bpit), epit(epit) {}
675 DocBookDocumentSectioning hasDocumentSectioning(ParagraphList const ¶graphs, pit_type bpit, pit_type const epit) {
676 bool documentHasSections = false;
678 while (bpit < epit) {
679 Layout const &style = paragraphs[bpit].layout();
680 documentHasSections |= style.category() == from_utf8("Sectioning");
682 if (documentHasSections)
686 // Paragraphs before the first section: [ runparams.par_begin ; eppit )
688 return make_tuple(documentHasSections, bpit);
692 bool hasOnlyNotes(Paragraph const & par)
694 // Precondition: the paragraph is not empty. Otherwise, the function will always return true...
695 for (int i = 0; i < par.size(); ++i)
696 // If you find something that is not an inset (like actual text) or an inset that is not a note,
698 if (!par.isInset(i) || !dynamic_cast<InsetNote *>(par.insetList().get(i)))
704 DocBookInfoTag getParagraphsWithInfo(ParagraphList const ¶graphs, pit_type bpit, pit_type const epit) {
705 set<pit_type> shouldBeInInfo;
706 set<pit_type> mustBeInInfo;
707 set<pit_type> abstract;
709 // Find the first non empty paragraph by mutating bpit.
710 while (bpit < epit) {
711 Paragraph const &par = paragraphs[bpit];
712 if (par.empty() || hasOnlyNotes(par))
718 // Find the last info-like paragraph.
719 pit_type cpit = bpit;
720 bool hasAbstractLayout = false;
721 while (cpit < epit) {
722 // Skip paragraphs only containing one note.
723 Paragraph const & par = paragraphs[cpit];
724 if (hasOnlyNotes(par)) {
729 if (par.layout().docbookabstract())
730 hasAbstractLayout = true;
732 // Based on layout information, store this paragraph in one set: should be in <info>, must be.
733 Layout const &style = par.layout();
735 if (style.docbookininfo() == "always") {
736 mustBeInInfo.emplace(cpit);
737 } else if (style.docbookininfo() == "maybe") {
738 shouldBeInInfo.emplace(cpit);
740 // Hypothesis: the <info> parts should be grouped together near the beginning bpit.
741 // There may be notes in between, but nothing else.
746 // Now, cpit points to the last paragraph that has things that could go in <info>.
747 // bpit is the beginning of the <info> part.
749 // Go once again through the list of paragraphs to find the abstract. If there is an abstract
750 // layout, only consider it. Otherwise, an abstract is just a sequence of paragraphs with text.
751 if (hasAbstractLayout) {
753 while (pit < cpit) { // Don't overshoot the <info> part.
754 if (paragraphs[pit].layout().docbookabstract())
755 abstract.emplace(pit);
759 pit_type lastAbstract = epit + 1; // A nonsensical value.
760 docstring lastAbstractLayout;
763 while (pit < cpit) { // Don't overshoot the <info> part.
764 const Paragraph & par = paragraphs.at(pit);
765 if (!par.insetList().empty()) {
766 for (const auto &i : par.insetList()) {
767 if (i.inset->getText(0) != nullptr) {
768 if (lastAbstract == epit + 1) {
769 // First paragraph that matches the heuristic definition of abstract.
771 lastAbstractLayout = par.layout().name();
772 } else if (pit > lastAbstract + 1 || par.layout().name() != lastAbstractLayout) {
773 // This is either too far from the last abstract paragraph or doesn't
774 // have the right layout name, BUT there has already been an abstract
775 // in this document: done with detecting the abstract.
776 goto done; // Easier to get out of two nested loops.
779 abstract.emplace(pit);
789 return DocBookInfoTag(shouldBeInInfo, mustBeInInfo, abstract, bpit, cpit);
792 } // end anonymous namespace
795 xml::FontTag docbookStartFontTag(xml::FontTypes type)
797 return xml::FontTag(from_utf8(fontToDocBookTag(type)), from_utf8(fontToAttribute(type)), type);
801 xml::EndFontTag docbookEndFontTag(xml::FontTypes type)
803 return xml::EndFontTag(from_utf8(fontToDocBookTag(type)), type);
807 void outputDocBookInfo(
811 OutputParams const & runparams,
812 ParagraphList const & paragraphs,
813 DocBookInfoTag const & info)
815 // Perform an additional check on the abstract. Sometimes, there are many paragraphs that should go
816 // into the abstract, but none generates actual content. Thus, first generate to a temporary stream,
817 // then only create the <abstract> tag if these paragraphs generate some content.
818 // This check must be performed *before* a decision on whether or not to output <info> is made.
819 bool hasAbstract = !info.abstract.empty();
822 // Generate the abstract XML into a string before further checks.
823 odocstringstream os2;
826 auto bpit = *std::min_element(info.abstract.begin(), info.abstract.end());
827 auto epit = 1 + *std::max_element(info.abstract.begin(), info.abstract.end());
828 // info.abstract is inclusive, epit is exclusive, hence +1 for looping.
830 while (bpit < epit) {
831 makeAny(text, buf, xs2, runparams, paragraphs.iterator_at(bpit));
836 // Actually output the abstract if there is something to do. Don't count line feeds or spaces in this,
837 // even though they must be properly output if there is some abstract.
838 abstract = os2.str();
839 docstring cleaned = abstract;
840 cleaned.erase(std::remove_if(cleaned.begin(), cleaned.end(), ::isspace), cleaned.end());
842 // Nothing? Then there is no abstract!
847 // The abstract must go in <info>. Otherwise, decide whether to open <info> based on the layouts.
848 bool needInfo = !info.mustBeInInfo.empty() || hasAbstract;
850 // Start the <info> tag if required.
852 xs.startDivision(false);
853 xs << xml::StartTag("info");
857 // Output the elements that should go in <info>, before and after the abstract.
858 for (auto pit : info.shouldBeInInfo) { // Typically, the title: these elements are so important and ubiquitous
859 // that mandating a wrapper like <info> would repel users. Thus, generate them first.
860 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
862 for (auto pit : info.mustBeInInfo) {
863 if (info.abstract.find(pit) == info.abstract.end()) // The abstract must be in info, but is dealt with after.
864 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
867 // Always output the abstract as the last item of the <info>, as it requires special treatment (especially if
868 // it contains several paragraphs that are empty).
870 // string tag = paragraphs[*info.abstract.begin()].layout().docbookforceabstracttag();
871 // if (tag == "NONE")
874 // xs << xml::StartTag(tag);
876 xs << XMLStream::ESCAPE_NONE << abstract;
877 // xs << xml::EndTag(tag);
881 // End the <info> tag if it was started.
883 xs << xml::EndTag("info");
890 void docbookFirstParagraphs(
894 OutputParams const &runparams,
897 // Handle the beginning of the document, supposing it has sections.
898 // Major role: output the first <info> tag.
900 ParagraphList const ¶graphs = text.paragraphs();
901 pit_type bpit = runparams.par_begin;
902 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, epit);
903 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
907 void docbookSimpleAllParagraphs(
911 OutputParams const & runparams)
913 // Handle the given text, supposing it has no sections (i.e. a "simple" text). The input may vary in length
914 // between a single paragraph to a whole document.
916 // First, the <info> tag.
917 ParagraphList const ¶graphs = text.paragraphs();
918 pit_type bpit = runparams.par_begin;
919 pit_type const epit = runparams.par_end;
920 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, epit);
921 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
923 // Then, the content. It starts where the <info> ends.
925 while (bpit < epit) {
926 auto par = paragraphs.iterator_at(bpit);
927 if (!hasOnlyNotes(*par))
928 makeAny(text, buf, xs, runparams, par);
934 void docbookParagraphs(Text const &text,
937 OutputParams const &runparams) {
938 ParagraphList const ¶graphs = text.paragraphs();
939 if (runparams.par_begin == runparams.par_end) {
940 runparams.par_begin = 0;
941 runparams.par_end = paragraphs.size();
943 pit_type bpit = runparams.par_begin;
944 pit_type const epit = runparams.par_end;
947 xs << XMLStream::ESCAPE_NONE << "<!-- DocBook output error! -->\n";
951 std::stack<std::pair<int, string>> headerLevels; // Used to determine when to open/close sections: store the depth
952 // of the section and the tag that was used to open it.
954 // Detect whether the document contains sections. If there are no sections, there can be no automatically
955 // discovered abstract.
956 bool documentHasSections;
958 tie(documentHasSections, eppit) = hasDocumentSectioning(paragraphs, bpit, epit);
960 if (documentHasSections) {
961 docbookFirstParagraphs(text, buf, xs, runparams, eppit);
964 docbookSimpleAllParagraphs(text, buf, xs, runparams);
968 bool currentlyInAppendix = false;
970 while (bpit < epit) {
971 OutputParams ourparams = runparams;
973 auto par = paragraphs.iterator_at(bpit);
974 if (par->params().startOfAppendix())
975 currentlyInAppendix = true;
976 Layout const &style = par->layout();
977 ParagraphList::const_iterator const lastStartedPar = par;
978 ParagraphList::const_iterator send;
980 if (hasOnlyNotes(*par)) {
985 // Think about adding <section> and/or </section>s.
986 const bool isLayoutSectioning = style.category() == from_utf8("Sectioning");
987 if (isLayoutSectioning) {
988 int level = style.toclevel;
990 // Need to close a previous section if it has the same level or a higher one (close <section> if opening a <h2>
991 // after a <h2>, <h3>, <h4>, <h5> or <h6>). More examples:
992 // - current: h2; back: h1; do not close any <section>
993 // - current: h1; back: h2; close two <section> (first the <h2>, then the <h1>, so a new <h1> can come)
994 while (!headerLevels.empty() && level <= headerLevels.top().first) {
995 int stackLevel = headerLevels.top().first;
996 docstring stackTag = from_utf8("</" + headerLevels.top().second + ">");
999 // Output the tag only if it corresponds to a legit section.
1000 if (stackLevel != Layout::NOT_IN_TOC)
1001 xs << XMLStream::ESCAPE_NONE << stackTag << xml::CR();
1004 // Open the new section: first push it onto the stack, then output it in DocBook.
1005 string sectionTag = (currentlyInAppendix && style.docbooksectiontag() == "chapter") ?
1006 "appendix" : style.docbooksectiontag();
1007 headerLevels.push(std::make_pair(level, sectionTag));
1009 // Some sectioning-like elements should not be output (such as FrontMatter).
1010 if (level != Layout::NOT_IN_TOC) {
1011 // Look for a label in the title, i.e. a InsetLabel as a child.
1012 docstring id = docstring();
1013 for (pos_type i = 0; i < par->size(); ++i) {
1014 Inset const *inset = par->getInset(i);
1016 if (auto label = dynamic_cast<InsetLabel const *>(inset)) {
1017 // Generate the attributes for the section if need be.
1018 id += "xml:id=\"" + xml::cleanID(label->screenLabel()) + "\"";
1020 // Don't output the ID as a DocBook <anchor>.
1021 ourparams.docbook_anchors_to_ignore.emplace(label->screenLabel());
1023 // Cannot have multiple IDs per tag.
1029 // Write the open tag for this section.
1030 docstring tag = from_utf8("<" + sectionTag);
1032 tag += from_utf8(" ") + id;
1033 tag += from_utf8(">");
1034 xs << XMLStream::ESCAPE_NONE << tag;
1039 // Close all sections before the bibliography.
1040 // 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)?
1041 auto insetsLength = distance(par->insetList().begin(), par->insetList().end());
1042 if (insetsLength > 0) {
1043 Inset const *firstInset = par->getInset(0);
1044 if (firstInset && dynamic_cast<InsetBibtex const *>(firstInset)) {
1045 while (!headerLevels.empty()) {
1046 int level = headerLevels.top().first;
1047 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1050 // Output the tag only if it corresponds to a legit section.
1051 if (level != Layout::NOT_IN_TOC) {
1052 xs << XMLStream::ESCAPE_NONE << tag;
1059 // Generate this paragraph.
1060 makeAny(text, buf, xs, ourparams, par);
1064 // If need be, close <section>s, but only at the end of the document (otherwise, dealt with at the beginning
1066 while (!headerLevels.empty() && headerLevels.top().first > Layout::NOT_IN_TOC) {
1067 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1069 xs << XMLStream::ESCAPE_NONE << tag;