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 // 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());
349 ParagraphList::const_iterator makeAny(Text const &,
352 OutputParams const &,
353 ParagraphList::const_iterator);
356 void makeBibliography(
360 OutputParams const & runparams,
361 ParagraphList::const_iterator const & par)
363 // If this is the first paragraph in a bibliography, open the bibliography tag.
364 auto pbegin_before = text.paragraphs().getParagraphBefore(par);
365 if (pbegin_before->layout().latextype != LATEX_BIB_ENVIRONMENT) {
366 xs << xml::StartTag("bibliography");
370 // Start the precooked bibliography entry. This is very much like opening a paragraph tag.
371 // Don't forget the citation ID!
373 for (auto i = 0; i < par->size(); ++i) {
374 Inset const *ip = par->getInset(i);
377 if (const auto * bibitem = dynamic_cast<const InsetBibitem*>(ip)) {
378 attr = from_utf8("xml:id='") + bibitem->getParam("key") + from_utf8("'");
382 xs << xml::StartTag(from_utf8("bibliomixed"), attr);
384 // Generate the entry. Concatenate the different parts of the paragraph if any.
385 auto const begin = text.paragraphs().begin();
386 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(std::distance(begin, par)), 0);
387 for (auto & parXML : pars)
388 xs << XMLStream::ESCAPE_NONE << parXML;
390 // End the precooked bibliography entry.
391 xs << xml::EndTag("bibliomixed");
394 // If this is the last paragraph in a bibliography, close the bibliography tag.
395 auto const end = text.paragraphs().end();
398 bool endBibliography = nextpar == end || nextpar->layout().latextype != LATEX_BIB_ENVIRONMENT;
400 if (endBibliography) {
401 xs << xml::EndTag("bibliography");
411 OutputParams const & runparams,
412 ParagraphList::const_iterator const & par)
414 auto const begin = text.paragraphs().begin();
415 auto const end = text.paragraphs().end();
416 auto prevpar = text.paragraphs().getParagraphBefore(par);
418 // We want to open the paragraph tag if:
419 // (i) the current layout permits multiple paragraphs
420 // (ii) we are either not already inside a paragraph (HTMLIsBlock) OR
421 // we are, but this is not the first paragraph
423 // But there is also a special case, and we first see whether we are in it.
424 // We do not want to open the paragraph tag if this paragraph contains
425 // only one item, and that item is "inline", i.e., not HTMLIsBlock (such
426 // as a branch). On the other hand, if that single item has a font change
427 // applied to it, then we still do need to open the paragraph.
429 // Obviously, this is very fragile. The main reason we need to do this is
430 // because of branches, e.g., a branch that contains an entire new section.
431 // We do not really want to wrap that whole thing in a <div>...</div>.
432 bool special_case = false;
433 Inset const *specinset = par->size() == 1 ? par->getInset(0) : nullptr;
434 if (specinset && !specinset->getLayout().htmlisblock()) { // TODO: Convert htmlisblock to a DocBook parameter?
435 Layout const &style = par->layout();
436 FontInfo const first_font = style.labeltype == LABEL_MANUAL ?
437 style.labelfont : style.font;
438 FontInfo const our_font =
439 par->getFont(buf.masterBuffer()->params(), 0,
440 text.outerFont(std::distance(begin, par))).fontInfo();
442 if (first_font == our_font)
446 size_t nInsets = std::distance(par->insetList().begin(), par->insetList().end());
448 // Plain layouts must be ignored.
449 special_case |= buf.params().documentClass().isPlainLayout(par->layout()) && !runparams.docbook_force_pars;
450 // Equations do not deserve their own paragraph (DocBook allows them outside paragraphs).
451 special_case |= nInsets == par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
452 return inset.inset && inset.inset->asInsetMath();
455 // 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.
456 if (!special_case && par->size() == 1 && par->getInset(0)) {
457 Inset const * firstInset = par->getInset(0);
459 // Floats cannot be in paragraphs.
460 special_case = to_utf8(firstInset->layoutName()).substr(0, 6) == "Float:";
462 // Bibliographies cannot be in paragraphs.
463 if (!special_case && firstInset->asInsetCommand())
464 special_case = firstInset->asInsetCommand()->params().getCmdName() == "bibtex";
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 // std::all_of allows doing this check without having to copy the string.
492 // Open and close tags around each contained paragraph.
493 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(distance(begin, par)), 0);
494 for (auto & parXML : pars) {
495 if (!std::all_of(parXML.begin(), parXML.end(), ::isspace)) {
497 openParTag(xs, &*par, prevpar);
499 xs << XMLStream::ESCAPE_NONE << parXML;
502 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
508 void makeEnvironment(Text const &text,
511 OutputParams const &runparams,
512 ParagraphList::const_iterator const & par)
514 // TODO: simplify me!
515 auto const end = text.paragraphs().end();
517 // Output the opening tag for this environment, but only if it has not been previously opened (condition
518 // implemented in openParTag).
519 auto prevpar = text.paragraphs().getParagraphBefore(par);
520 openParTag(xs, &*par, prevpar); // TODO: switch in layout for par/block?
522 // Generate the contents of this environment. There is a special case if this is like some environment.
523 Layout const & style = par->layout();
524 if (style.latextype == LATEX_COMMAND) {
525 // Nothing to do (otherwise, infinite loops).
526 } else if (style.latextype == LATEX_ENVIRONMENT) {
527 // Open a wrapper tag if needed.
528 if (style.docbookitemwrappertag() != "NONE")
529 openTag(xs, style.docbookitemwrappertag(), style.docbookitemwrapperattr(), style.docbookitemwrappertagtype());
531 // Generate the label, if need be. If it is taken from the text, sep != 0 and corresponds to the first
532 // character after the label.
534 if (style.labeltype != LABEL_NO_LABEL && style.docbookitemlabeltag() != "NONE") {
535 // At least one condition must be met:
536 // - this environment is not a list
537 // - if this is a list, the label must not be manual (i.e. it must be taken from the layout)
538 if (style.latextype != LATEX_LIST_ENVIRONMENT || style.labeltype != LABEL_MANUAL) {
539 // Usual cases: maybe there is something specified at the layout level. Highly unlikely, though.
540 docstring const lbl = par->params().labelString();
543 openLabelTag(xs, style);
545 closeLabelTag(xs, style);
548 // Only variablelist gets here (or similar items defined as an extension in the layout).
549 openLabelTag(xs, style);
550 sep = par->firstWordDocBook(xs, runparams);
551 closeLabelTag(xs, style);
555 // Maybe the item is completely empty, i.e. if the first word ends at the end of the current paragraph
556 // AND if the next paragraph doesn't have the same depth (if there is such a paragraph).
557 // Common case: there is only the first word on the line, but there is a nested list instead
559 bool emptyItem = false;
560 if (sep == par->size()) { // If the separator is already at the end of this paragraph...
563 if (next_par == text.paragraphs().end()) // There is no next paragraph.
565 else // There is a next paragraph: check depth.
566 emptyItem = par->params().depth() >= next_par->params().depth();
570 // Avoid having an empty item, this is not valid DocBook. A single character is enough to force
571 // generation of a full <para>.
572 // TODO: this always worked only by magic...
575 // Generate the rest of the paragraph, if need be. Open as many inner tags as necessary.
576 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(std::distance(text.paragraphs().begin(), par)), sep);
577 auto p = pars.begin();
579 xs << XMLStream::ESCAPE_NONE << *p;
581 if (p != pars.end()) {
582 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
583 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(), par->layout().docbookiteminnertagtype());
589 makeAny(text, buf, xs, runparams, par);
592 // Close the environment.
595 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr); // TODO: switch in layout for par/block?
599 ParagraphList::const_iterator findEndOfEnvironment(
600 ParagraphList::const_iterator const & pstart,
601 ParagraphList::const_iterator const & pend)
603 // Copy-paste from XHTML. Should be factored out at some point...
604 ParagraphList::const_iterator p = pstart;
605 Layout const & bstyle = p->layout();
606 size_t const depth = p->params().depth();
607 for (++p; p != pend; ++p) {
608 Layout const & style = p->layout();
609 // It shouldn't happen that e.g. a section command occurs inside
610 // a quotation environment, at a higher depth, but as of 6/2009,
611 // it can happen. We pretend that it's just at lowest depth.
612 if (style.latextype == LATEX_COMMAND)
615 // If depth is down, we're done
616 if (p->params().depth() < depth)
619 // If depth is up, we're not done
620 if (p->params().depth() > depth)
623 // FIXME I am not sure about the first check.
624 // Surely we *could* have different layouts that count as
625 // LATEX_PARAGRAPH, right?
626 if (style.latextype == LATEX_PARAGRAPH || style != bstyle)
633 ParagraphList::const_iterator makeListEnvironment(Text const &text,
636 OutputParams const &runparams,
637 ParagraphList::const_iterator const & begin)
640 auto const end = text.paragraphs().end();
641 auto const envend = findEndOfEnvironment(par, end);
643 // Output the opening tag for this environment.
644 Layout const & envstyle = par->layout();
645 openTag(xs, envstyle.docbookwrappertag(), envstyle.docbookwrapperattr(), envstyle.docbookwrappertagtype());
646 openTag(xs, envstyle.docbooktag(), envstyle.docbookattr(), envstyle.docbooktagtype());
648 // Handle the content of the list environment, item by item.
649 while (par != envend) {
650 Layout const & style = par->layout();
652 // Open the item wrapper.
653 openTag(xs, style.docbookitemwrappertag(), style.docbookitemwrapperattr(), style.docbookitemwrappertagtype());
655 // Generate the label, if need be. If it is taken from the text, sep != 0 and corresponds to the first
656 // character after the label.
658 if (style.labeltype != LABEL_NO_LABEL && style.docbookitemlabeltag() != "NONE") {
659 if (style.labeltype == LABEL_MANUAL) {
660 // Only variablelist gets here (or similar items defined as an extension in the layout).
661 openLabelTag(xs, style);
662 sep = 1 + par->firstWordDocBook(xs, runparams);
663 closeLabelTag(xs, style);
665 // Usual cases: maybe there is something specified at the layout level. Highly unlikely, though.
666 docstring const lbl = par->params().labelString();
669 openLabelTag(xs, style);
671 closeLabelTag(xs, style);
676 // Open the item (after the wrapper and the label).
677 openTag(xs, style.docbookitemtag(), style.docbookitemattr(), style.docbookitemtagtype());
679 // Generate the content of the item.
680 if (sep < par->size()) {
681 auto pars = par->simpleDocBookOnePar(buf, runparams,
682 text.outerFont(std::distance(text.paragraphs().begin(), par)), sep);
683 for (auto &p : pars) {
684 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
685 par->layout().docbookiteminnertagtype());
686 xs << XMLStream::ESCAPE_NONE << p;
687 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
690 // DocBook doesn't like emptiness.
691 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
692 par->layout().docbookiteminnertagtype());
693 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
696 // If the next item is deeper, it must go entirely within this item (do it recursively).
697 // By construction, with findEndOfEnvironment, depth can only stay constant or increase, never decrease.
698 depth_type currentDepth = par->getDepth();
700 while (par != envend && par->getDepth() != currentDepth)
701 par = makeAny(text, buf, xs, runparams, par);
702 // Usually, this loop only makes one iteration, except in complex scenarios, like an item with a paragraph,
703 // a list, and another paragraph; or an item with two types of list (itemise then enumerate, for instance).
706 closeTag(xs, style.docbookitemtag(), style.docbookitemtagtype());
707 closeTag(xs, style.docbookitemwrappertag(), style.docbookitemwrappertagtype());
710 // Close this environment in exactly the same way as it was opened.
711 closeTag(xs, envstyle.docbooktag(), envstyle.docbooktagtype());
712 closeTag(xs, envstyle.docbookwrappertag(), envstyle.docbookwrappertagtype());
722 OutputParams const & runparams,
723 ParagraphList::const_iterator const & par)
725 // Unlike XHTML, no need for labels, as they are handled by DocBook tags.
726 auto const begin = text.paragraphs().begin();
727 auto const end = text.paragraphs().end();
731 // Generate this command.
732 auto prevpar = text.paragraphs().getParagraphBefore(par);
733 openParTag(xs, &*par, prevpar);
735 auto pars = par->simpleDocBookOnePar(buf, runparams,text.outerFont(distance(begin, par)));
736 for (auto & parXML : pars)
737 // TODO: decide what to do with openParTag/closeParTag in new lines.
738 xs << XMLStream::ESCAPE_NONE << parXML;
740 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
744 ParagraphList::const_iterator makeAny(Text const &text,
747 OutputParams const &runparams,
748 ParagraphList::const_iterator par)
750 switch (par->layout().latextype) {
752 makeCommand(text, buf, xs, runparams, par);
754 case LATEX_ENVIRONMENT:
755 makeEnvironment(text, buf, xs, runparams, par);
757 case LATEX_LIST_ENVIRONMENT:
758 case LATEX_ITEM_ENVIRONMENT:
759 // Only case when makeAny() might consume more than one paragraph.
760 return makeListEnvironment(text, buf, xs, runparams, par);
761 case LATEX_PARAGRAPH:
762 makeParagraph(text, buf, xs, runparams, par);
764 case LATEX_BIB_ENVIRONMENT:
765 makeBibliography(text, buf, xs, runparams, par);
773 bool isLayoutSectioning(Layout const & lay)
775 return lay.category() == from_utf8("Sectioning");
779 using DocBookDocumentSectioning = tuple<bool, pit_type>;
782 struct DocBookInfoTag
784 const set<pit_type> shouldBeInInfo;
785 const set<pit_type> mustBeInInfo; // With the notable exception of the abstract!
786 const set<pit_type> abstract;
787 const bool abstractLayout;
791 DocBookInfoTag(const set<pit_type> & shouldBeInInfo, const set<pit_type> & mustBeInInfo,
792 const set<pit_type> & abstract, bool abstractLayout, pit_type bpit, pit_type epit) :
793 shouldBeInInfo(shouldBeInInfo), mustBeInInfo(mustBeInInfo), abstract(abstract),
794 abstractLayout(abstractLayout), bpit(bpit), epit(epit) {}
798 DocBookDocumentSectioning hasDocumentSectioning(ParagraphList const ¶graphs, pit_type bpit, pit_type const epit) {
799 bool documentHasSections = false;
801 while (bpit < epit) {
802 Layout const &style = paragraphs[bpit].layout();
803 documentHasSections |= isLayoutSectioning(style);
805 if (documentHasSections)
809 // Paragraphs before the first section: [ runparams.par_begin ; eppit )
811 return make_tuple(documentHasSections, bpit);
815 bool hasOnlyNotes(Paragraph const & par)
817 // Precondition: the paragraph is not empty. Otherwise, the function will always return true...
818 for (int i = 0; i < par.size(); ++i)
819 // If you find something that is not an inset (like actual text) or an inset that is not a note,
821 if (!par.isInset(i) || par.getInset(i)->lyxCode() != NOTE_CODE)
827 DocBookInfoTag getParagraphsWithInfo(ParagraphList const ¶graphs,
828 pit_type bpit, pit_type const epit,
829 // Typically, bpit is the beginning of the document and epit the end *or* the first section.
830 bool documentHasSections) {
831 set<pit_type> shouldBeInInfo;
832 set<pit_type> mustBeInInfo;
833 set<pit_type> abstractWithLayout;
834 set<pit_type> abstractNoLayout;
836 // Find the first non empty paragraph by mutating bpit.
837 while (bpit < epit) {
838 Paragraph const &par = paragraphs[bpit];
839 if (par.empty() || hasOnlyNotes(par))
845 // Traverse everything that might belong to <info>.
846 bool hasAbstractLayout = false;
847 pit_type cpit = bpit;
848 for (; cpit < epit; ++cpit) {
849 // Skip paragraphs that don't generate anything in DocBook.
850 Paragraph const & par = paragraphs[cpit];
851 if (hasOnlyNotes(par))
854 // There should never be any section here. (Just a sanity check: if this fails, this function could end up
855 // processing the whole document.)
856 if (isLayoutSectioning(par.layout())) {
857 LYXERR0("Assertion failed: section found in potential <info> paragraphs.");
861 // If this is marked as an abstract by the layout, put it in the right set.
862 if (par.layout().docbookabstract()) {
863 hasAbstractLayout = true;
864 abstractWithLayout.emplace(cpit);
868 // Based on layout information, store this paragraph in one set: should be in <info>, must be,
869 // or abstract (either because of layout or of position).
870 Layout const &style = par.layout();
872 if (style.docbookininfo() == "always")
873 mustBeInInfo.emplace(cpit);
874 else if (style.docbookininfo() == "maybe")
875 shouldBeInInfo.emplace(cpit);
876 else if (documentHasSections && !hasAbstractLayout)
877 abstractNoLayout.emplace(cpit);
878 else // This should definitely not be in <info>.
881 // Now, cpit points to the first paragraph that no more has things that could go in <info>.
882 // bpit is the beginning of the <info> part.
884 return DocBookInfoTag(shouldBeInInfo, mustBeInInfo,
885 hasAbstractLayout ? abstractWithLayout : abstractNoLayout,
886 hasAbstractLayout, bpit, cpit);
889 } // end anonymous namespace
892 xml::FontTag docbookStartFontTag(xml::FontTypes type)
894 return xml::FontTag(from_utf8(fontToDocBookTag(type)), from_utf8(fontToAttribute(type)), type);
898 xml::EndFontTag docbookEndFontTag(xml::FontTypes type)
900 return xml::EndFontTag(from_utf8(fontToDocBookTag(type)), type);
904 void outputDocBookInfo(
908 OutputParams const & runparams,
909 ParagraphList const & paragraphs,
910 DocBookInfoTag const & info)
912 // Perform an additional check on the abstract. Sometimes, there are many paragraphs that should go
913 // into the abstract, but none generates actual content. Thus, first generate to a temporary stream,
914 // then only create the <abstract> tag if these paragraphs generate some content.
915 // This check must be performed *before* a decision on whether or not to output <info> is made.
916 bool hasAbstract = !info.abstract.empty();
919 // Generate the abstract XML into a string before further checks.
920 odocstringstream os2;
922 for (auto const & p : info.abstract)
923 makeAny(text, buf, xs2, runparams, paragraphs.iterator_at(p));
925 // Actually output the abstract if there is something to do. Don't count line feeds or spaces in this,
926 // even though they must be properly output if there is some abstract.
927 abstract = os2.str();
928 docstring cleaned = abstract;
929 cleaned.erase(std::remove_if(cleaned.begin(), cleaned.end(), ::isspace), cleaned.end());
931 // Nothing? Then there is no abstract!
936 // The abstract must go in <info>. Otherwise, decide whether to open <info> based on the layouts.
937 bool needInfo = !info.mustBeInInfo.empty() || hasAbstract;
939 // Start the <info> tag if required.
941 xs.startDivision(false);
942 xs << xml::StartTag("info");
946 // Output the elements that should go in <info>, before and after the abstract.
947 for (auto pit : info.shouldBeInInfo) // Typically, the title: these elements are so important and ubiquitous
948 // that mandating a wrapper like <info> would repel users. Thus, generate them first.
949 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
950 for (auto pit : info.mustBeInInfo)
951 if (info.abstract.find(pit) == info.abstract.end()) // The abstract must be in info, but is dealt with after.
952 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
954 // Always output the abstract as the last item of the <info>, as it requires special treatment (especially if
955 // it contains several paragraphs that are empty).
957 if (info.abstractLayout) {
958 xs << XMLStream::ESCAPE_NONE << abstract;
961 string tag = paragraphs[*info.abstract.begin()].layout().docbookforceabstracttag();
965 if (!xs.isLastTagCR())
968 xs << xml::StartTag(tag);
970 xs << XMLStream::ESCAPE_NONE << abstract;
971 xs << xml::EndTag(tag);
976 // End the <info> tag if it was started.
978 xs << xml::EndTag("info");
985 void docbookSimpleAllParagraphs(
989 OutputParams const & runparams)
991 // Handle the given text, supposing it has no sections (i.e. a "simple" text). The input may vary in length
992 // between a single paragraph to a whole document.
993 pit_type const bpit = runparams.par_begin;
994 pit_type const epit = runparams.par_end;
995 ParagraphList const ¶graphs = text.paragraphs();
997 // First, the <info> tag.
998 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, epit, false);
999 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
1001 // Then, the content. It starts where the <info> ends.
1002 auto par = paragraphs.iterator_at(info.epit);
1003 auto end = paragraphs.iterator_at(epit);
1004 while (par != end) {
1005 if (!hasOnlyNotes(*par))
1006 par = makeAny(text, buf, xs, runparams, par);
1013 void docbookParagraphs(Text const &text,
1016 OutputParams const &runparams) {
1017 ParagraphList const ¶graphs = text.paragraphs();
1018 if (runparams.par_begin == runparams.par_end) {
1019 runparams.par_begin = 0;
1020 runparams.par_end = paragraphs.size();
1022 pit_type bpit = runparams.par_begin;
1023 pit_type const epit = runparams.par_end;
1024 LASSERT(bpit < epit,
1026 xs << XMLStream::ESCAPE_NONE << "<!-- DocBook output error! -->\n";
1030 std::stack<std::pair<int, string>> headerLevels; // Used to determine when to open/close sections: store the depth
1031 // of the section and the tag that was used to open it.
1033 // Detect whether the document contains sections. If there are no sections, treatment is largely simplified.
1034 // In particular, there can't be an abstract, unless it is manually marked.
1035 bool documentHasSections;
1037 tie(documentHasSections, eppit) = hasDocumentSectioning(paragraphs, bpit, epit);
1039 // Deal with "simple" documents, i.e. those without sections.
1040 if (!documentHasSections) {
1041 docbookSimpleAllParagraphs(text, buf, xs, runparams);
1045 // Output the first <info> tag (or just the title).
1046 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, eppit, true);
1047 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
1050 // Then, iterate through the paragraphs of this document.
1051 bool currentlyInAppendix = false;
1053 auto par = text.paragraphs().iterator_at(bpit);
1054 auto end = text.paragraphs().iterator_at(epit);
1055 while (par != end) {
1056 OutputParams ourparams = runparams;
1058 if (par->params().startOfAppendix())
1059 currentlyInAppendix = true;
1060 if (hasOnlyNotes(*par)) {
1065 Layout const &style = par->layout();
1067 // Think about adding <section> and/or </section>s.
1068 if (isLayoutSectioning(style)) {
1069 int level = style.toclevel;
1071 // Need to close a previous section if it has the same level or a higher one (close <section> if opening a
1072 // <h2> after a <h2>, <h3>, <h4>, <h5> or <h6>). More examples:
1073 // - current: h2; back: h1; do not close any <section>
1074 // - current: h1; back: h2; close two <section> (first the <h2>, then the <h1>, so a new <h1> can come)
1075 while (!headerLevels.empty() && level <= headerLevels.top().first) {
1076 // Output the tag only if it corresponds to a legit section.
1077 int stackLevel = headerLevels.top().first;
1078 if (stackLevel != Layout::NOT_IN_TOC) {
1079 xs << xml::EndTag(headerLevels.top().second);
1085 // Open the new section: first push it onto the stack, then output it in DocBook.
1086 string sectionTag = (currentlyInAppendix && style.docbooksectiontag() == "chapter") ?
1087 "appendix" : style.docbooksectiontag();
1088 headerLevels.push(std::make_pair(level, sectionTag));
1090 // Some sectioning-like elements should not be output (such as FrontMatter).
1091 if (level != Layout::NOT_IN_TOC) {
1092 // Look for a label in the title, i.e. a InsetLabel as a child.
1093 docstring id = docstring();
1094 for (pos_type i = 0; i < par->size(); ++i) {
1095 Inset const *inset = par->getInset(i);
1097 if (auto label = dynamic_cast<InsetLabel const *>(inset)) {
1098 // Generate the attributes for the section if need be.
1099 id += "xml:id=\"" + xml::cleanID(label->screenLabel()) + "\"";
1101 // Don't output the ID as a DocBook <anchor>.
1102 ourparams.docbook_anchors_to_ignore.emplace(label->screenLabel());
1104 // Cannot have multiple IDs per tag.
1110 // Write the open tag for this section.
1114 xs << xml::StartTag(sectionTag, attrs);
1119 // Close all sections before the bibliography.
1120 // 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)?
1121 auto insetsLength = distance(par->insetList().begin(), par->insetList().end());
1122 if (insetsLength > 0) {
1123 Inset const *firstInset = par->getInset(0);
1124 if (firstInset && (firstInset->lyxCode() == BIBITEM_CODE || firstInset->lyxCode() == BIBTEX_CODE)) {
1125 while (!headerLevels.empty()) {
1126 int level = headerLevels.top().first;
1127 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1130 // Output the tag only if it corresponds to a legit section.
1131 if (level != Layout::NOT_IN_TOC) {
1132 xs << XMLStream::ESCAPE_NONE << tag;
1139 // Generate this paragraph.
1140 par = makeAny(text, buf, xs, ourparams, par);
1142 // Some special sections may require abstracts (mostly parts, in books).
1143 // 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.
1144 if (isLayoutSectioning(style) && style.docbookforceabstracttag() != "NONE") {
1145 // This abstract may be found between the next paragraph and the next title.
1146 pit_type cpit = std::distance(text.paragraphs().begin(), par);
1147 pit_type ppit = std::get<1>(hasDocumentSectioning(paragraphs, cpit, epit));
1149 // Generate this abstract (this code corresponds to parts of outputDocBookInfo).
1150 DocBookInfoTag secInfo = getParagraphsWithInfo(paragraphs, cpit, ppit, true);
1152 if (!secInfo.abstract.empty()) {
1153 xs << xml::StartTag(style.docbookforceabstracttag());
1155 for (auto const &p : secInfo.abstract)
1156 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(p));
1157 xs << xml::EndTag(style.docbookforceabstracttag());
1161 // Skip all the text that just has been generated.
1162 par = paragraphs.iterator_at(ppit);
1166 // If need be, close <section>s, but only at the end of the document (otherwise, dealt with at the beginning
1168 while (!headerLevels.empty() && headerLevels.top().first > Layout::NOT_IN_TOC) {
1169 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1171 xs << XMLStream::ESCAPE_NONE << tag;