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") // Common check to be performed elsewhere, if it was not here.
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 void compTag(XMLStream & xs, const std::string & tag, const std::string & attr, const std::string & tagtype)
260 if (tag.empty() || tag == "NONE")
263 // Special case for <para>: always considered as a paragraph.
264 if (tag == "para" || tagtype == "paragraph" || tagtype == "block") {
265 if (!xs.isLastTagCR())
267 xs << xml::CompTag(tag, attr);
269 } else if (tagtype == "inline") {
270 xs << xml::CompTag(tag, attr);
272 xs.writeError("Unrecognised tag type '" + tagtype + "' for '" + tag + "'");
277 // Higher-level convenience functions.
279 void openParTag(XMLStream & xs, const Paragraph * par, const Paragraph * prevpar)
281 Layout const & lay = par->layout();
286 // When should the wrapper be opened here? Only if the previous paragraph has the SAME wrapper tag
287 // (usually, they won't have the same layout) and the CURRENT one allows merging.
288 // The main use case is author information in several paragraphs: if the name of the author is the
289 // first paragraph of an author, then merging with the previous tag does not make sense. Say the
290 // next paragraph is the affiliation, then it should be output in the same <author> tag (different
291 // layout, same wrapper tag).
292 bool openWrapper = lay.docbookwrappertag() != "NONE";
293 if (prevpar != nullptr) {
294 Layout const & prevlay = prevpar->layout();
295 if (prevlay.docbookwrappertag() != "NONE") {
296 if (prevlay.docbookwrappertag() == lay.docbookwrappertag() &&
297 prevlay.docbookwrapperattr() == lay.docbookwrapperattr())
298 openWrapper = !lay.docbookwrappermergewithprevious();
306 openTag(xs, lay.docbookwrappertag(), lay.docbookwrapperattr(), lay.docbookwrappertagtype());
308 const string & tag = lay.docbooktag();
310 auto xmltag = xml::ParTag(tag, lay.docbookattr());
311 if (!xs.isTagOpen(xmltag, 1)) // Don't nest a paragraph directly in a paragraph.
312 // TODO: required or not?
313 // TODO: avoid creating a ParTag object just for this query...
314 openTag(xs, lay.docbooktag(), lay.docbookattr(), lay.docbooktagtype());
317 openTag(xs, lay.docbookitemtag(), lay.docbookitemattr(), lay.docbookitemtagtype());
318 openTag(xs, lay.docbookiteminnertag(), lay.docbookiteminnerattr(), lay.docbookiteminnertagtype());
322 void closeParTag(XMLStream & xs, Paragraph const * par, Paragraph const * nextpar)
327 // See comment in openParTag.
328 Layout const & lay = par->layout();
329 bool closeWrapper = lay.docbookwrappertag() != "NONE";
330 if (nextpar != nullptr) {
331 Layout const & nextlay = nextpar->layout();
332 if (nextlay.docbookwrappertag() != "NONE") {
333 if (nextlay.docbookwrappertag() == lay.docbookwrappertag() &&
334 nextlay.docbookwrapperattr() == lay.docbookwrapperattr())
335 closeWrapper = !nextlay.docbookwrappermergewithprevious();
342 closeTag(xs, lay.docbookiteminnertag(), lay.docbookiteminnertagtype());
343 closeTag(xs, lay.docbookitemtag(), lay.docbookitemtagtype());
344 closeTag(xs, lay.docbooktag(), lay.docbooktagtype());
346 closeTag(xs, lay.docbookwrappertag(), lay.docbookwrappertagtype());
350 ParagraphList::const_iterator makeAny(Text const &,
353 OutputParams const &,
354 ParagraphList::const_iterator);
357 void makeBibliography(
361 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 const * pbegin_before = text.paragraphs().getParagraphBefore(par);
366 if (pbegin_before == nullptr || (pbegin_before && 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();
399 bool endBibliography = nextpar == end || nextpar->layout().latextype != LATEX_BIB_ENVIRONMENT;
401 if (endBibliography) {
402 xs << xml::EndTag("bibliography");
412 OutputParams const & runparams,
413 ParagraphList::const_iterator const & par)
415 auto const begin = text.paragraphs().begin();
416 auto const end = text.paragraphs().end();
417 auto prevpar = text.paragraphs().getParagraphBefore(par);
419 // We want to open the paragraph tag if:
420 // (i) the current layout permits multiple paragraphs
421 // (ii) we are either not already inside a paragraph (HTMLIsBlock) OR
422 // we are, but this is not the first paragraph
424 // But there is also a special case, and we first see whether we are in it.
425 // We do not want to open the paragraph tag if this paragraph contains
426 // only one item, and that item is "inline", i.e., not HTMLIsBlock (such
427 // as a branch). On the other hand, if that single item has a font change
428 // applied to it, then we still do need to open the paragraph.
430 // Obviously, this is very fragile. The main reason we need to do this is
431 // because of branches, e.g., a branch that contains an entire new section.
432 // We do not really want to wrap that whole thing in a <div>...</div>.
433 bool special_case = false;
434 Inset const *specinset = par->size() == 1 ? par->getInset(0) : nullptr;
435 if (specinset && !specinset->getLayout().htmlisblock()) { // TODO: Convert htmlisblock to a DocBook parameter?
436 Layout const &style = par->layout();
437 FontInfo const first_font = style.labeltype == LABEL_MANUAL ?
438 style.labelfont : style.font;
439 FontInfo const our_font =
440 par->getFont(buf.masterBuffer()->params(), 0,
441 text.outerFont(std::distance(begin, par))).fontInfo();
443 if (first_font == our_font)
447 size_t nInsets = std::distance(par->insetList().begin(), par->insetList().end());
449 // Plain layouts must be ignored.
450 special_case |= buf.params().documentClass().isPlainLayout(par->layout()) && !runparams.docbook_force_pars;
451 // Equations do not deserve their own paragraph (DocBook allows them outside paragraphs).
452 special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
453 return inset.inset && inset.inset->asInsetMath();
456 // TODO: Could get rid of this with a DocBook equivalent to htmlisblock? Not for all cases, unfortunately... See above for those that have been determined not to be allowable for this potential refactoring.
457 if (!special_case && par->size() == 1 && par->getInset(0)) {
458 Inset const * firstInset = par->getInset(0);
460 // Floats cannot be in paragraphs.
461 special_case = to_utf8(firstInset->layoutName()).substr(0, 6) == "Float:";
463 // Bibliographies cannot be in paragraphs.
464 if (!special_case && firstInset->asInsetCommand())
465 special_case = firstInset->asInsetCommand()->params().getCmdName() == "bibtex";
467 // ERTs are in comments, not paragraphs.
468 if (!special_case && firstInset->lyxCode() == lyx::ERT_CODE)
471 // Listings should not get into their own paragraph.
472 if (!special_case && firstInset->lyxCode() == lyx::LISTINGS_CODE)
475 // Boxes cannot get into their own paragraph.
476 if (!special_case && firstInset->lyxCode() == lyx::BOX_CODE)
480 bool const open_par = runparams.docbook_make_pars
481 && !runparams.docbook_in_par
484 // We want to issue the closing tag if either:
485 // (i) We opened it, and either docbook_in_par is false,
486 // or we're not in the last paragraph, anyway.
487 // (ii) We didn't open it and docbook_in_par is true,
488 // but we are in the first par, and there is a next par.
489 bool const close_par = open_par && (!runparams.docbook_in_par);
491 // Determine if this paragraph has some real content. Things like new pages are not caught
492 // by Paragraph::empty(), even though they do not generate anything useful in DocBook.
493 // Thus, remove all spaces (including new lines: \r, \n) before checking for emptiness.
494 // std::all_of allows doing this check without having to copy the string.
495 // Open and close tags around each contained paragraph.
498 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(distance(begin, par)), 0, nextpar == end, special_case);
499 for (docstring const & parXML : pars) {
500 if (xml::isNotOnlySpace(parXML)) {
502 openParTag(xs, &*par, prevpar);
504 xs << XMLStream::ESCAPE_NONE << parXML;
507 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
513 void makeEnvironment(Text const &text,
516 OutputParams const &runparams,
517 ParagraphList::const_iterator const & par)
519 auto const end = text.paragraphs().end();
523 // Special cases for listing-like environments provided in layouts. This is quite ad-hoc, but provides a useful
524 // default. This should not be used by too many environments (only LyX-Code right now).
525 // This would be much simpler if LyX-Code was implemented as InsetListings...
526 bool mimicListing = false;
527 bool ignoreFonts = false;
528 if (par->layout().docbooktag() == "programlisting") {
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 // Generate the paragraph, if need be.
544 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(std::distance(text.paragraphs().begin(), par)), 0, false, ignoreFonts);
547 auto p = pars.begin();
548 while (p != pars.end()) {
549 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(), par->layout().docbookiteminnertagtype());
550 xs << XMLStream::ESCAPE_NONE << *p;
551 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
558 for (auto const & p : pars) {
559 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(), par->layout().docbookiteminnertagtype());
560 xs << XMLStream::ESCAPE_NONE << p;
561 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
565 makeAny(text, buf, xs, runparams, par);
568 // Close the environment.
569 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr); // TODO: switch in layout for par/block?
573 ParagraphList::const_iterator findEndOfEnvironment(
574 ParagraphList::const_iterator const & pstart,
575 ParagraphList::const_iterator const & pend)
577 // Copy-paste from XHTML. Should be factored out at some point...
578 ParagraphList::const_iterator p = pstart;
579 Layout const & bstyle = p->layout();
580 size_t const depth = p->params().depth();
581 for (++p; p != pend; ++p) {
582 Layout const & style = p->layout();
583 // It shouldn't happen that e.g. a section command occurs inside
584 // a quotation environment, at a higher depth, but as of 6/2009,
585 // it can happen. We pretend that it's just at lowest depth.
586 if (style.latextype == LATEX_COMMAND)
589 // If depth is down, we're done
590 if (p->params().depth() < depth)
593 // If depth is up, we're not done
594 if (p->params().depth() > depth)
597 // FIXME I am not sure about the first check.
598 // Surely we *could* have different layouts that count as
599 // LATEX_PARAGRAPH, right?
600 if (style.latextype == LATEX_PARAGRAPH || style != bstyle)
607 ParagraphList::const_iterator makeListEnvironment(Text const &text,
610 OutputParams const &runparams,
611 ParagraphList::const_iterator const & begin)
614 auto const end = text.paragraphs().end();
615 auto const envend = findEndOfEnvironment(par, end);
617 // Output the opening tag for this environment.
618 Layout const & envstyle = par->layout();
619 openTag(xs, envstyle.docbookwrappertag(), envstyle.docbookwrapperattr(), envstyle.docbookwrappertagtype());
620 openTag(xs, envstyle.docbooktag(), envstyle.docbookattr(), envstyle.docbooktagtype());
622 // Handle the content of the list environment, item by item.
623 while (par != envend) {
624 Layout const & style = par->layout();
626 // Open the item wrapper.
627 openTag(xs, style.docbookitemwrappertag(), style.docbookitemwrapperattr(), style.docbookitemwrappertagtype());
629 // Generate the label, if need be. If it is taken from the text, sep != 0 and corresponds to the first
630 // character after the label.
632 if (style.labeltype != LABEL_NO_LABEL && style.docbookitemlabeltag() != "NONE") {
633 if (style.labeltype == LABEL_MANUAL) {
634 // Only variablelist gets here (or similar items defined as an extension in the layout).
635 openTag(xs, style.docbookitemlabeltag(), style.docbookitemlabelattr(), style.docbookitemlabeltagtype());
636 sep = 1 + par->firstWordDocBook(xs, runparams);
637 closeTag(xs, style.docbookitemlabeltag(), style.docbookitemlabeltagtype());
639 // Usual cases: maybe there is something specified at the layout level. Highly unlikely, though.
640 docstring const lbl = par->params().labelString();
643 openTag(xs, style.docbookitemlabeltag(), style.docbookitemlabelattr(), style.docbookitemlabeltagtype());
645 closeTag(xs, style.docbookitemlabeltag(), style.docbookitemlabeltagtype());
650 // Open the item (after the wrapper and the label).
651 openTag(xs, style.docbookitemtag(), style.docbookitemattr(), style.docbookitemtagtype());
653 // Generate the content of the item.
654 if (sep < par->size()) {
655 auto pars = par->simpleDocBookOnePar(buf, runparams,
656 text.outerFont(std::distance(text.paragraphs().begin(), par)), sep);
657 for (auto &p : pars) {
658 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
659 par->layout().docbookiteminnertagtype());
660 xs << XMLStream::ESCAPE_NONE << p;
661 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
664 // DocBook doesn't like emptiness.
665 compTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
666 par->layout().docbookiteminnertagtype());
669 // If the next item is deeper, it must go entirely within this item (do it recursively).
670 // By construction, with findEndOfEnvironment, depth can only stay constant or increase, never decrease.
671 depth_type currentDepth = par->getDepth();
673 while (par != envend && par->getDepth() != currentDepth)
674 par = makeAny(text, buf, xs, runparams, par);
675 // Usually, this loop only makes one iteration, except in complex scenarios, like an item with a paragraph,
676 // a list, and another paragraph; or an item with two types of list (itemise then enumerate, for instance).
679 closeTag(xs, style.docbookitemtag(), style.docbookitemtagtype());
680 closeTag(xs, style.docbookitemwrappertag(), style.docbookitemwrappertagtype());
683 // Close this environment in exactly the same way as it was opened.
684 closeTag(xs, envstyle.docbooktag(), envstyle.docbooktagtype());
685 closeTag(xs, envstyle.docbookwrappertag(), envstyle.docbookwrappertagtype());
695 OutputParams const & runparams,
696 ParagraphList::const_iterator const & par)
698 // Unlike XHTML, no need for labels, as they are handled by DocBook tags.
699 auto const begin = text.paragraphs().begin();
700 auto const end = text.paragraphs().end();
704 // Generate this command.
705 auto prevpar = text.paragraphs().getParagraphBefore(par);
706 openParTag(xs, &*par, prevpar);
708 auto pars = par->simpleDocBookOnePar(buf, runparams,text.outerFont(distance(begin, par)));
709 for (auto & parXML : pars)
710 // TODO: decide what to do with openParTag/closeParTag in new lines.
711 xs << XMLStream::ESCAPE_NONE << parXML;
713 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
717 ParagraphList::const_iterator makeAny(Text const &text,
720 OutputParams const &runparams,
721 ParagraphList::const_iterator par)
723 switch (par->layout().latextype) {
725 makeCommand(text, buf, xs, runparams, par);
727 case LATEX_ENVIRONMENT:
728 makeEnvironment(text, buf, xs, runparams, par);
730 case LATEX_LIST_ENVIRONMENT:
731 case LATEX_ITEM_ENVIRONMENT:
732 // Only case when makeAny() might consume more than one paragraph.
733 return makeListEnvironment(text, buf, xs, runparams, par);
734 case LATEX_PARAGRAPH:
735 makeParagraph(text, buf, xs, runparams, par);
737 case LATEX_BIB_ENVIRONMENT:
738 makeBibliography(text, buf, xs, runparams, par);
746 bool isLayoutSectioning(Layout const & lay)
748 return lay.category() == from_utf8("Sectioning");
752 using DocBookDocumentSectioning = tuple<bool, pit_type>;
755 struct DocBookInfoTag
757 const set<pit_type> shouldBeInInfo;
758 const set<pit_type> mustBeInInfo; // With the notable exception of the abstract!
759 const set<pit_type> abstract;
760 const bool abstractLayout;
764 DocBookInfoTag(const set<pit_type> & shouldBeInInfo, const set<pit_type> & mustBeInInfo,
765 const set<pit_type> & abstract, bool abstractLayout, pit_type bpit, pit_type epit) :
766 shouldBeInInfo(shouldBeInInfo), mustBeInInfo(mustBeInInfo), abstract(abstract),
767 abstractLayout(abstractLayout), bpit(bpit), epit(epit) {}
771 DocBookDocumentSectioning hasDocumentSectioning(ParagraphList const ¶graphs, pit_type bpit, pit_type const epit) {
772 bool documentHasSections = false;
774 while (bpit < epit) {
775 Layout const &style = paragraphs[bpit].layout();
776 documentHasSections |= isLayoutSectioning(style);
778 if (documentHasSections)
782 // Paragraphs before the first section: [ runparams.par_begin ; eppit )
784 return make_tuple(documentHasSections, bpit);
788 bool hasOnlyNotes(Paragraph const & par)
790 // Precondition: the paragraph is not empty. Otherwise, the function will always return true...
791 for (int i = 0; i < par.size(); ++i)
792 // If you find something that is not an inset (like actual text) or an inset that is not a note,
794 if (!par.isInset(i) || par.getInset(i)->lyxCode() != NOTE_CODE)
800 DocBookInfoTag getParagraphsWithInfo(ParagraphList const ¶graphs,
801 pit_type bpit, pit_type const epit,
802 // Typically, bpit is the beginning of the document and epit the end *or* the first section.
803 bool documentHasSections) {
804 set<pit_type> shouldBeInInfo;
805 set<pit_type> mustBeInInfo;
806 set<pit_type> abstractWithLayout;
807 set<pit_type> abstractNoLayout;
809 // Find the first non empty paragraph by mutating bpit.
810 while (bpit < epit) {
811 Paragraph const &par = paragraphs[bpit];
812 if (par.empty() || hasOnlyNotes(par))
818 // Traverse everything that might belong to <info>.
819 bool hasAbstractLayout = false;
820 pit_type cpit = bpit;
821 for (; cpit < epit; ++cpit) {
822 // Skip paragraphs that don't generate anything in DocBook.
823 Paragraph const & par = paragraphs[cpit];
824 if (hasOnlyNotes(par))
827 // There should never be any section here. (Just a sanity check: if this fails, this function could end up
828 // processing the whole document.)
829 if (isLayoutSectioning(par.layout())) {
830 LYXERR0("Assertion failed: section found in potential <info> paragraphs.");
834 // If this is marked as an abstract by the layout, put it in the right set.
835 if (par.layout().docbookabstract()) {
836 hasAbstractLayout = true;
837 abstractWithLayout.emplace(cpit);
841 // Based on layout information, store this paragraph in one set: should be in <info>, must be,
842 // or abstract (either because of layout or of position).
843 Layout const &style = par.layout();
845 if (style.docbookininfo() == "always")
846 mustBeInInfo.emplace(cpit);
847 else if (style.docbookininfo() == "maybe")
848 shouldBeInInfo.emplace(cpit);
849 else if (documentHasSections && !hasAbstractLayout)
850 abstractNoLayout.emplace(cpit);
851 else // This should definitely not be in <info>.
854 // Now, cpit points to the first paragraph that no more has things that could go in <info>.
855 // bpit is the beginning of the <info> part.
857 return DocBookInfoTag(shouldBeInInfo, mustBeInInfo,
858 hasAbstractLayout ? abstractWithLayout : abstractNoLayout,
859 hasAbstractLayout, bpit, cpit);
862 } // end anonymous namespace
865 xml::FontTag docbookStartFontTag(xml::FontTypes type)
867 return xml::FontTag(from_utf8(fontToDocBookTag(type)), from_utf8(fontToAttribute(type)), type);
871 xml::EndFontTag docbookEndFontTag(xml::FontTypes type)
873 return xml::EndFontTag(from_utf8(fontToDocBookTag(type)), type);
877 void outputDocBookInfo(
881 OutputParams const & runparams,
882 ParagraphList const & paragraphs,
883 DocBookInfoTag const & info)
885 // Perform an additional check on the abstract. Sometimes, there are many paragraphs that should go
886 // into the abstract, but none generates actual content. Thus, first generate to a temporary stream,
887 // then only create the <abstract> tag if these paragraphs generate some content.
888 // This check must be performed *before* a decision on whether or not to output <info> is made.
889 bool hasAbstract = !info.abstract.empty();
892 // Generate the abstract XML into a string before further checks.
893 odocstringstream os2;
895 for (auto const & p : info.abstract)
896 makeAny(text, buf, xs2, runparams, paragraphs.iterator_at(p));
898 // Actually output the abstract if there is something to do. Don't count line feeds or spaces in this,
899 // even though they must be properly output if there is some abstract.
900 abstract = os2.str();
901 docstring cleaned = abstract;
902 cleaned.erase(std::remove_if(cleaned.begin(), cleaned.end(), ::isspace), cleaned.end());
904 // Nothing? Then there is no abstract!
909 // The abstract must go in <info>. Otherwise, decide whether to open <info> based on the layouts.
910 bool needInfo = !info.mustBeInInfo.empty() || hasAbstract;
912 // Start the <info> tag if required.
914 xs.startDivision(false);
915 xs << xml::StartTag("info");
919 // Output the elements that should go in <info>, before and after the abstract.
920 for (auto pit : info.shouldBeInInfo) // Typically, the title: these elements are so important and ubiquitous
921 // that mandating a wrapper like <info> would repel users. Thus, generate them first.
922 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
923 for (auto pit : info.mustBeInInfo)
924 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
926 // Always output the abstract as the last item of the <info>, as it requires special treatment (especially if
927 // it contains several paragraphs that are empty).
929 if (info.abstractLayout) {
930 xs << XMLStream::ESCAPE_NONE << abstract;
933 string tag = paragraphs[*info.abstract.begin()].layout().docbookforceabstracttag();
937 if (!xs.isLastTagCR())
940 xs << xml::StartTag(tag);
942 xs << XMLStream::ESCAPE_NONE << abstract;
943 xs << xml::EndTag(tag);
948 // End the <info> tag if it was started.
950 if (!xs.isLastTagCR())
953 xs << xml::EndTag("info");
960 void docbookSimpleAllParagraphs(
964 OutputParams const & runparams)
966 // Handle the given text, supposing it has no sections (i.e. a "simple" text). The input may vary in length
967 // between a single paragraph to a whole document.
968 pit_type const bpit = runparams.par_begin;
969 pit_type const epit = runparams.par_end;
970 ParagraphList const ¶graphs = text.paragraphs();
972 // First, the <info> tag.
973 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, epit, false);
974 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
976 // Then, the content. It starts where the <info> ends.
977 auto par = paragraphs.iterator_at(info.epit);
978 auto end = paragraphs.iterator_at(epit);
980 if (!hasOnlyNotes(*par))
981 par = makeAny(text, buf, xs, runparams, par);
988 void docbookParagraphs(Text const &text,
991 OutputParams const &runparams) {
992 ParagraphList const ¶graphs = text.paragraphs();
993 if (runparams.par_begin == runparams.par_end) {
994 runparams.par_begin = 0;
995 runparams.par_end = paragraphs.size();
997 pit_type bpit = runparams.par_begin;
998 pit_type const epit = runparams.par_end;
1001 xs << XMLStream::ESCAPE_NONE << "<!-- DocBook output error! -->\n";
1005 std::stack<std::pair<int, string>> headerLevels; // Used to determine when to open/close sections: store the depth
1006 // of the section and the tag that was used to open it.
1008 // Detect whether the document contains sections. If there are no sections, treatment is largely simplified.
1009 // In particular, there can't be an abstract, unless it is manually marked.
1010 bool documentHasSections;
1012 tie(documentHasSections, eppit) = hasDocumentSectioning(paragraphs, bpit, epit);
1014 // Deal with "simple" documents, i.e. those without sections.
1015 if (!documentHasSections) {
1016 docbookSimpleAllParagraphs(text, buf, xs, runparams);
1020 // Output the first <info> tag (or just the title).
1021 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, eppit, true);
1022 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
1025 // Then, iterate through the paragraphs of this document.
1026 bool currentlyInAppendix = false;
1028 auto par = text.paragraphs().iterator_at(bpit);
1029 auto end = text.paragraphs().iterator_at(epit);
1030 while (par != end) {
1031 OutputParams ourparams = runparams;
1033 if (par->params().startOfAppendix())
1034 currentlyInAppendix = true;
1035 if (hasOnlyNotes(*par)) {
1040 Layout const &style = par->layout();
1042 // Think about adding <section> and/or </section>s.
1043 if (isLayoutSectioning(style)) {
1044 int level = style.toclevel;
1046 // Need to close a previous section if it has the same level or a higher one (close <section> if opening a
1047 // <h2> after a <h2>, <h3>, <h4>, <h5> or <h6>). More examples:
1048 // - current: h2; back: h1; do not close any <section>
1049 // - current: h1; back: h2; close two <section> (first the <h2>, then the <h1>, so a new <h1> can come)
1050 while (!headerLevels.empty() && level <= headerLevels.top().first) {
1051 // Output the tag only if it corresponds to a legit section.
1052 int stackLevel = headerLevels.top().first;
1053 if (stackLevel != Layout::NOT_IN_TOC) {
1054 xs << xml::EndTag(headerLevels.top().second);
1060 // Open the new section: first push it onto the stack, then output it in DocBook.
1061 string sectionTag = (currentlyInAppendix && style.docbooksectiontag() == "chapter") ?
1062 "appendix" : style.docbooksectiontag();
1063 headerLevels.push(std::make_pair(level, sectionTag));
1065 // Some sectioning-like elements should not be output (such as FrontMatter).
1066 if (level != Layout::NOT_IN_TOC) {
1067 // Look for a label in the title, i.e. a InsetLabel as a child.
1068 docstring id = docstring();
1069 for (pos_type i = 0; i < par->size(); ++i) {
1070 Inset const *inset = par->getInset(i);
1072 if (auto label = dynamic_cast<InsetLabel const *>(inset)) {
1073 // Generate the attributes for the section if need be.
1074 id += "xml:id=\"" + xml::cleanID(label->screenLabel()) + "\"";
1076 // Don't output the ID as a DocBook <anchor>.
1077 ourparams.docbook_anchors_to_ignore.emplace(label->screenLabel());
1079 // Cannot have multiple IDs per tag.
1085 // Write the open tag for this section.
1089 xs << xml::StartTag(sectionTag, attrs);
1094 // Close all sections before the bibliography.
1095 // 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)?
1096 auto insetsLength = distance(par->insetList().begin(), par->insetList().end());
1097 if (insetsLength > 0) {
1098 Inset const *firstInset = par->getInset(0);
1099 if (firstInset && (firstInset->lyxCode() == BIBITEM_CODE || firstInset->lyxCode() == BIBTEX_CODE)) {
1100 while (!headerLevels.empty()) {
1101 int level = headerLevels.top().first;
1102 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1105 // Output the tag only if it corresponds to a legit section.
1106 if (level != Layout::NOT_IN_TOC) {
1107 xs << XMLStream::ESCAPE_NONE << tag;
1114 // Generate this paragraph.
1115 par = makeAny(text, buf, xs, ourparams, par);
1117 // Some special sections may require abstracts (mostly parts, in books).
1118 // TODO: docbookforceabstracttag is a bit contrived here, but it does the job. Having another field just for this would be cleaner, but that's just for <part> and <partintro>, so it's probably not worth the effort.
1119 if (isLayoutSectioning(style) && style.docbookforceabstracttag() != "NONE") {
1120 // This abstract may be found between the next paragraph and the next title.
1121 pit_type cpit = std::distance(text.paragraphs().begin(), par);
1122 pit_type ppit = std::get<1>(hasDocumentSectioning(paragraphs, cpit, epit));
1124 // Generate this abstract (this code corresponds to parts of outputDocBookInfo).
1125 DocBookInfoTag secInfo = getParagraphsWithInfo(paragraphs, cpit, ppit, true);
1127 if (!secInfo.abstract.empty()) {
1128 xs << xml::StartTag(style.docbookforceabstracttag());
1130 for (auto const &p : secInfo.abstract)
1131 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(p));
1132 xs << xml::EndTag(style.docbookforceabstracttag());
1136 // Skip all the text that just has been generated.
1137 par = paragraphs.iterator_at(ppit);
1141 // If need be, close <section>s, but only at the end of the document (otherwise, dealt with at the beginning
1143 while (!headerLevels.empty() && headerLevels.top().first > Layout::NOT_IN_TOC) {
1144 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1146 xs << XMLStream::ESCAPE_NONE << tag;