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 if (prevlay.docbookwrappertag() == lay.docbookwrappertag())
278 openWrapper = !lay.docbookwrappermergewithprevious();
286 openTag(xs, lay.docbookwrappertag(), lay.docbookwrapperattr(), lay.docbookwrappertagtype());
288 const string & tag = lay.docbooktag();
290 auto xmltag = xml::ParTag(tag, lay.docbookattr());
291 if (!xs.isTagOpen(xmltag, 1)) // Don't nest a paragraph directly in a paragraph.
292 // TODO: required or not?
293 // TODO: avoid creating a ParTag object just for this query...
294 openTag(xs, lay.docbooktag(), lay.docbookattr(), lay.docbooktagtype());
297 openTag(xs, lay.docbookitemtag(), lay.docbookitemattr(), lay.docbookitemtagtype());
298 openTag(xs, lay.docbookiteminnertag(), lay.docbookiteminnerattr(), lay.docbookiteminnertagtype());
302 void closeParTag(XMLStream & xs, Paragraph const * par, Paragraph const * nextpar)
307 // See comment in openParTag.
308 Layout const & lay = par->layout();
309 bool closeWrapper = lay.docbookwrappertag() != "NONE";
310 if (nextpar != nullptr) {
311 Layout const & nextlay = nextpar->layout();
312 if (nextlay.docbookwrappertag() != "NONE") {
313 if (nextlay.docbookwrappertag() == lay.docbookwrappertag())
314 closeWrapper = !nextlay.docbookwrappermergewithprevious();
321 closeTag(xs, lay.docbookiteminnertag(), lay.docbookiteminnertagtype());
322 closeTag(xs, lay.docbookitemtag(), lay.docbookitemtagtype());
323 closeTag(xs, lay.docbooktag(), lay.docbooktagtype());
325 closeTag(xs, lay.docbookwrappertag(), lay.docbookwrappertagtype());
329 void openLabelTag(XMLStream & xs, Layout const & lay) // Mostly for definition lists.
331 openTag(xs, lay.docbookitemlabeltag(), lay.docbookitemlabelattr(), lay.docbookitemlabeltagtype());
335 void closeLabelTag(XMLStream & xs, Layout const & lay)
337 closeTag(xs, lay.docbookitemlabeltag(), lay.docbookitemlabeltagtype());
341 void openItemTag(XMLStream & xs, Layout const & lay)
343 openTag(xs, lay.docbookitemtag(), lay.docbookitemattr(), lay.docbookitemtagtype());
347 void closeItemTag(XMLStream & xs, Layout const & lay)
349 closeTag(xs, lay.docbookitemtag(), lay.docbookitemtagtype());
353 ParagraphList::const_iterator makeAny(Text const &,
356 OutputParams const &,
357 ParagraphList::const_iterator);
360 void makeBibliography(
364 OutputParams const & runparams,
365 ParagraphList::const_iterator const & par)
367 // If this is the first paragraph in a bibliography, open the bibliography tag.
368 auto pbegin_before = text.paragraphs().getParagraphBefore(par);
369 if (pbegin_before->layout().latextype != LATEX_BIB_ENVIRONMENT) {
370 xs << xml::StartTag("bibliography");
374 // Start the precooked bibliography entry. This is very much like opening a paragraph tag.
375 // Don't forget the citation ID!
377 for (auto i = 0; i < par->size(); ++i) {
378 Inset const *ip = par->getInset(i);
381 if (const auto * bibitem = dynamic_cast<const InsetBibitem*>(ip)) {
382 attr = from_utf8("xml:id='") + bibitem->getParam("key") + from_utf8("'");
386 xs << xml::StartTag(from_utf8("bibliomixed"), attr);
388 // Generate the entry. Concatenate the different parts of the paragraph if any.
389 auto const begin = text.paragraphs().begin();
390 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(std::distance(begin, par)), 0);
391 for (auto & parXML : pars)
392 xs << XMLStream::ESCAPE_NONE << parXML;
394 // End the precooked bibliography entry.
395 xs << xml::EndTag("bibliomixed");
398 // If this is the last paragraph in a bibliography, close the bibliography tag.
399 auto const end = text.paragraphs().end();
402 bool endBibliography = nextpar == end || nextpar->layout().latextype != LATEX_BIB_ENVIRONMENT;
404 if (endBibliography) {
405 xs << xml::EndTag("bibliography");
415 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 size_t nInsets = std::distance(par->insetList().begin(), par->insetList().end());
452 // Plain layouts must be ignored.
453 special_case |= buf.params().documentClass().isPlainLayout(par->layout()) && !runparams.docbook_force_pars;
454 // Equations do not deserve their own paragraph (DocBook allows them outside paragraphs).
455 special_case |= nInsets == par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
456 return inset.inset && inset.inset->asInsetMath();
459 // 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.
460 if (!special_case && par->size() == 1 && par->getInset(0)) {
461 Inset const * firstInset = par->getInset(0);
463 // Floats cannot be in paragraphs.
464 special_case = to_utf8(firstInset->layoutName()).substr(0, 6) == "Float:";
466 // Bibliographies cannot be in paragraphs.
467 if (!special_case && firstInset->asInsetCommand())
468 special_case = firstInset->asInsetCommand()->params().getCmdName() == "bibtex";
470 // ERTs are in comments, not paragraphs.
471 if (!special_case && firstInset->lyxCode() == lyx::ERT_CODE)
474 // Listings should not get into their own paragraph.
475 if (!special_case && firstInset->lyxCode() == lyx::LISTINGS_CODE)
479 bool const open_par = runparams.docbook_make_pars
480 && !runparams.docbook_in_par
483 // We want to issue the closing tag if either:
484 // (i) We opened it, and either docbook_in_par is false,
485 // or we're not in the last paragraph, anyway.
486 // (ii) We didn't open it and docbook_in_par is true,
487 // but we are in the first par, and there is a next par.
490 bool const close_par = open_par && (!runparams.docbook_in_par);
492 // Determine if this paragraph has some real content. Things like new pages are not caught
493 // by Paragraph::empty(), even though they do not generate anything useful in DocBook.
494 // Thus, remove all spaces (including new lines: \r, \n) before checking for emptiness.
495 // std::all_of allows doing this check without having to copy the string.
496 // Open and close tags around each contained paragraph.
497 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(distance(begin, par)), 0);
498 for (auto & parXML : pars) {
499 if (!std::all_of(parXML.begin(), parXML.end(), ::isspace)) {
501 openParTag(xs, &*par, prevpar);
503 xs << XMLStream::ESCAPE_NONE << parXML;
506 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
512 void makeEnvironment(Text const &text,
515 OutputParams const &runparams,
516 ParagraphList::const_iterator const & par)
518 // TODO: simplify me!
519 auto const end = text.paragraphs().end();
521 // Output the opening tag for this environment, but only if it has not been previously opened (condition
522 // implemented in openParTag).
523 auto prevpar = text.paragraphs().getParagraphBefore(par);
524 openParTag(xs, &*par, prevpar); // TODO: switch in layout for par/block?
526 // Generate the contents of this environment. There is a special case if this is like some environment.
527 Layout const & style = par->layout();
528 if (style.latextype == LATEX_COMMAND) {
529 // Nothing to do (otherwise, infinite loops).
530 } else if (style.latextype == LATEX_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();
547 openLabelTag(xs, style);
549 closeLabelTag(xs, style);
552 // Only variablelist gets here (or similar items defined as an extension in the layout).
553 openLabelTag(xs, style);
554 sep = par->firstWordDocBook(xs, runparams);
555 closeLabelTag(xs, style);
559 // Maybe the item is completely empty, i.e. if the first word ends at the end of the current paragraph
560 // AND if the next paragraph doesn't have the same depth (if there is such a paragraph).
561 // Common case: there is only the first word on the line, but there is a nested list instead
563 bool emptyItem = false;
564 if (sep == par->size()) { // If the separator is already at the end of this paragraph...
567 if (next_par == text.paragraphs().end()) // There is no next paragraph.
569 else // There is a next paragraph: check depth.
570 emptyItem = par->params().depth() >= next_par->params().depth();
574 // Avoid having an empty item, this is not valid DocBook. A single character is enough to force
575 // generation of a full <para>.
576 // TODO: this always worked only by magic...
579 // Generate the rest of the paragraph, if need be. Open as many inner tags as necessary.
580 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(std::distance(text.paragraphs().begin(), par)), sep);
581 auto p = pars.begin();
583 xs << XMLStream::ESCAPE_NONE << *p;
585 if (p != pars.end()) {
586 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
587 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(), par->layout().docbookiteminnertagtype());
593 makeAny(text, buf, xs, runparams, par);
596 // Close the environment.
599 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr); // TODO: switch in layout for par/block?
603 ParagraphList::const_iterator findEndOfEnvironment(
604 ParagraphList::const_iterator const & pstart,
605 ParagraphList::const_iterator const & pend)
607 // Copy-paste from XHTML. Should be factored out at some point...
608 ParagraphList::const_iterator p = pstart;
609 Layout const & bstyle = p->layout();
610 size_t const depth = p->params().depth();
611 for (++p; p != pend; ++p) {
612 Layout const & style = p->layout();
613 // It shouldn't happen that e.g. a section command occurs inside
614 // a quotation environment, at a higher depth, but as of 6/2009,
615 // it can happen. We pretend that it's just at lowest depth.
616 if (style.latextype == LATEX_COMMAND)
619 // If depth is down, we're done
620 if (p->params().depth() < depth)
623 // If depth is up, we're not done
624 if (p->params().depth() > depth)
627 // FIXME I am not sure about the first check.
628 // Surely we *could* have different layouts that count as
629 // LATEX_PARAGRAPH, right?
630 if (style.latextype == LATEX_PARAGRAPH || style != bstyle)
637 ParagraphList::const_iterator makeListEnvironment(Text const &text,
640 OutputParams const &runparams,
641 ParagraphList::const_iterator const & begin)
644 auto const end = text.paragraphs().end();
645 auto const envend = findEndOfEnvironment(par, end);
647 // Output the opening tag for this environment.
648 Layout const & envstyle = par->layout();
649 openTag(xs, envstyle.docbookwrappertag(), envstyle.docbookwrapperattr(), envstyle.docbookwrappertagtype());
650 openTag(xs, envstyle.docbooktag(), envstyle.docbookattr(), envstyle.docbooktagtype());
652 // Handle the content of the list environment, item by item.
653 while (par != envend) {
654 Layout const & style = par->layout();
656 // Open the item wrapper.
657 openTag(xs, style.docbookitemwrappertag(), style.docbookitemwrapperattr(), style.docbookitemwrappertagtype());
659 // Generate the label, if need be. If it is taken from the text, sep != 0 and corresponds to the first
660 // character after the label.
662 if (style.labeltype != LABEL_NO_LABEL && style.docbookitemlabeltag() != "NONE") {
663 if (style.labeltype == LABEL_MANUAL) {
664 // Only variablelist gets here (or similar items defined as an extension in the layout).
665 openLabelTag(xs, style);
666 sep = 1 + par->firstWordDocBook(xs, runparams);
667 closeLabelTag(xs, style);
669 // Usual cases: maybe there is something specified at the layout level. Highly unlikely, though.
670 docstring const lbl = par->params().labelString();
673 openLabelTag(xs, style);
675 closeLabelTag(xs, style);
680 // Open the item (after the wrapper and the label).
681 openTag(xs, style.docbookitemtag(), style.docbookitemattr(), style.docbookitemtagtype());
683 // Generate the content of the item.
684 if (sep < par->size()) {
685 auto pars = par->simpleDocBookOnePar(buf, runparams,
686 text.outerFont(std::distance(text.paragraphs().begin(), par)), sep);
687 for (auto &p : pars) {
688 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
689 par->layout().docbookiteminnertagtype());
690 xs << XMLStream::ESCAPE_NONE << p;
691 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
694 // DocBook doesn't like emptiness.
695 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
696 par->layout().docbookiteminnertagtype());
697 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
700 // If the next item is deeper, it must go entirely within this item (do it recursively).
701 // By construction, with findEndOfEnvironment, depth can only stay constant or increase, never decrease.
702 depth_type currentDepth = par->getDepth();
704 while (par != envend && par->getDepth() != currentDepth)
705 par = makeAny(text, buf, xs, runparams, par);
706 // Usually, this loop only makes one iteration, except in complex scenarios, like an item with a paragraph,
707 // a list, and another paragraph; or an item with two types of list (itemise then enumerate, for instance).
710 closeTag(xs, style.docbookitemtag(), style.docbookitemtagtype());
711 closeTag(xs, style.docbookitemwrappertag(), style.docbookitemwrappertagtype());
714 // Close this environment in exactly the same way as it was opened.
715 closeTag(xs, envstyle.docbooktag(), envstyle.docbooktagtype());
716 closeTag(xs, envstyle.docbookwrappertag(), envstyle.docbookwrappertagtype());
726 OutputParams const & runparams,
727 ParagraphList::const_iterator const & par)
729 // Unlike XHTML, no need for labels, as they are handled by DocBook tags.
730 auto const begin = text.paragraphs().begin();
731 auto const end = text.paragraphs().end();
735 // Generate this command.
736 auto prevpar = text.paragraphs().getParagraphBefore(par);
737 openParTag(xs, &*par, prevpar);
739 auto pars = par->simpleDocBookOnePar(buf, runparams,text.outerFont(distance(begin, par)));
740 for (auto & parXML : pars)
741 // TODO: decide what to do with openParTag/closeParTag in new lines.
742 xs << XMLStream::ESCAPE_NONE << parXML;
744 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
748 ParagraphList::const_iterator makeAny(Text const &text,
751 OutputParams const &runparams,
752 ParagraphList::const_iterator par)
754 switch (par->layout().latextype) {
756 makeCommand(text, buf, xs, runparams, par);
758 case LATEX_ENVIRONMENT:
759 makeEnvironment(text, buf, xs, runparams, par);
761 case LATEX_LIST_ENVIRONMENT:
762 case LATEX_ITEM_ENVIRONMENT:
763 // Only case when makeAny() might consume more than one paragraph.
764 return makeListEnvironment(text, buf, xs, runparams, par);
765 case LATEX_PARAGRAPH:
766 makeParagraph(text, buf, xs, runparams, par);
768 case LATEX_BIB_ENVIRONMENT:
769 makeBibliography(text, buf, xs, runparams, par);
777 bool isLayoutSectioning(Layout const & lay)
779 return lay.category() == from_utf8("Sectioning");
783 using DocBookDocumentSectioning = tuple<bool, pit_type>;
786 struct DocBookInfoTag
788 const set<pit_type> shouldBeInInfo;
789 const set<pit_type> mustBeInInfo; // With the notable exception of the abstract!
790 const set<pit_type> abstract;
791 const bool abstractLayout;
795 DocBookInfoTag(const set<pit_type> & shouldBeInInfo, const set<pit_type> & mustBeInInfo,
796 const set<pit_type> & abstract, bool abstractLayout, pit_type bpit, pit_type epit) :
797 shouldBeInInfo(shouldBeInInfo), mustBeInInfo(mustBeInInfo), abstract(abstract),
798 abstractLayout(abstractLayout), bpit(bpit), epit(epit) {}
802 DocBookDocumentSectioning hasDocumentSectioning(ParagraphList const ¶graphs, pit_type bpit, pit_type const epit) {
803 bool documentHasSections = false;
805 while (bpit < epit) {
806 Layout const &style = paragraphs[bpit].layout();
807 documentHasSections |= isLayoutSectioning(style);
809 if (documentHasSections)
813 // Paragraphs before the first section: [ runparams.par_begin ; eppit )
815 return make_tuple(documentHasSections, bpit);
819 bool hasOnlyNotes(Paragraph const & par)
821 // Precondition: the paragraph is not empty. Otherwise, the function will always return true...
822 for (int i = 0; i < par.size(); ++i)
823 // If you find something that is not an inset (like actual text) or an inset that is not a note,
825 if (!par.isInset(i) || par.getInset(i)->lyxCode() != NOTE_CODE)
831 DocBookInfoTag getParagraphsWithInfo(ParagraphList const ¶graphs,
832 pit_type bpit, pit_type const epit,
833 // Typically, bpit is the beginning of the document and epit the end *or* the first section.
834 bool documentHasSections) {
835 set<pit_type> shouldBeInInfo;
836 set<pit_type> mustBeInInfo;
837 set<pit_type> abstractWithLayout;
838 set<pit_type> abstractNoLayout;
840 // Find the first non empty paragraph by mutating bpit.
841 while (bpit < epit) {
842 Paragraph const &par = paragraphs[bpit];
843 if (par.empty() || hasOnlyNotes(par))
849 // Traverse everything that might belong to <info>.
850 bool hasAbstractLayout = false;
851 pit_type cpit = bpit;
852 for (; cpit < epit; ++cpit) {
853 // Skip paragraphs that don't generate anything in DocBook.
854 Paragraph const & par = paragraphs[cpit];
855 if (hasOnlyNotes(par))
858 // There should never be any section here. (Just a sanity check: if this fails, this function could end up
859 // processing the whole document.)
860 if (isLayoutSectioning(par.layout())) {
861 LYXERR0("Assertion failed: section found in potential <info> paragraphs.");
865 // If this is marked as an abstract by the layout, put it in the right set.
866 if (par.layout().docbookabstract()) {
867 hasAbstractLayout = true;
868 abstractWithLayout.emplace(cpit);
872 // Based on layout information, store this paragraph in one set: should be in <info>, must be,
873 // or abstract (either because of layout or of position).
874 Layout const &style = par.layout();
876 if (style.docbookininfo() == "always")
877 mustBeInInfo.emplace(cpit);
878 else if (style.docbookininfo() == "maybe")
879 shouldBeInInfo.emplace(cpit);
880 else if (documentHasSections && !hasAbstractLayout)
881 abstractNoLayout.emplace(cpit);
882 else // This should definitely not be in <info>.
885 // Now, cpit points to the first paragraph that no more has things that could go in <info>.
886 // bpit is the beginning of the <info> part.
888 return DocBookInfoTag(shouldBeInInfo, mustBeInInfo,
889 hasAbstractLayout ? abstractWithLayout : abstractNoLayout,
890 hasAbstractLayout, bpit, cpit);
893 } // end anonymous namespace
896 xml::FontTag docbookStartFontTag(xml::FontTypes type)
898 return xml::FontTag(from_utf8(fontToDocBookTag(type)), from_utf8(fontToAttribute(type)), type);
902 xml::EndFontTag docbookEndFontTag(xml::FontTypes type)
904 return xml::EndFontTag(from_utf8(fontToDocBookTag(type)), type);
908 void outputDocBookInfo(
912 OutputParams const & runparams,
913 ParagraphList const & paragraphs,
914 DocBookInfoTag const & info)
916 // Perform an additional check on the abstract. Sometimes, there are many paragraphs that should go
917 // into the abstract, but none generates actual content. Thus, first generate to a temporary stream,
918 // then only create the <abstract> tag if these paragraphs generate some content.
919 // This check must be performed *before* a decision on whether or not to output <info> is made.
920 bool hasAbstract = !info.abstract.empty();
923 // Generate the abstract XML into a string before further checks.
924 odocstringstream os2;
926 for (auto const & p : info.abstract)
927 makeAny(text, buf, xs2, runparams, paragraphs.iterator_at(p));
929 // Actually output the abstract if there is something to do. Don't count line feeds or spaces in this,
930 // even though they must be properly output if there is some abstract.
931 abstract = os2.str();
932 docstring cleaned = abstract;
933 cleaned.erase(std::remove_if(cleaned.begin(), cleaned.end(), ::isspace), cleaned.end());
935 // Nothing? Then there is no abstract!
940 // The abstract must go in <info>. Otherwise, decide whether to open <info> based on the layouts.
941 bool needInfo = !info.mustBeInInfo.empty() || hasAbstract;
943 // Start the <info> tag if required.
945 xs.startDivision(false);
946 xs << xml::StartTag("info");
950 // Output the elements that should go in <info>, before and after the abstract.
951 for (auto pit : info.shouldBeInInfo) // Typically, the title: these elements are so important and ubiquitous
952 // that mandating a wrapper like <info> would repel users. Thus, generate them first.
953 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
954 for (auto pit : info.mustBeInInfo)
955 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
957 // Always output the abstract as the last item of the <info>, as it requires special treatment (especially if
958 // it contains several paragraphs that are empty).
960 if (info.abstractLayout) {
961 xs << XMLStream::ESCAPE_NONE << abstract;
964 string tag = paragraphs[*info.abstract.begin()].layout().docbookforceabstracttag();
968 if (!xs.isLastTagCR())
971 xs << xml::StartTag(tag);
973 xs << XMLStream::ESCAPE_NONE << abstract;
974 xs << xml::EndTag(tag);
979 // End the <info> tag if it was started.
981 xs << xml::EndTag("info");
988 void docbookSimpleAllParagraphs(
992 OutputParams const & runparams)
994 // Handle the given text, supposing it has no sections (i.e. a "simple" text). The input may vary in length
995 // between a single paragraph to a whole document.
996 pit_type const bpit = runparams.par_begin;
997 pit_type const epit = runparams.par_end;
998 ParagraphList const ¶graphs = text.paragraphs();
1000 // First, the <info> tag.
1001 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, epit, false);
1002 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
1004 // Then, the content. It starts where the <info> ends.
1005 auto par = paragraphs.iterator_at(info.epit);
1006 auto end = paragraphs.iterator_at(epit);
1007 while (par != end) {
1008 if (!hasOnlyNotes(*par))
1009 par = makeAny(text, buf, xs, runparams, par);
1016 void docbookParagraphs(Text const &text,
1019 OutputParams const &runparams) {
1020 ParagraphList const ¶graphs = text.paragraphs();
1021 if (runparams.par_begin == runparams.par_end) {
1022 runparams.par_begin = 0;
1023 runparams.par_end = paragraphs.size();
1025 pit_type bpit = runparams.par_begin;
1026 pit_type const epit = runparams.par_end;
1027 LASSERT(bpit < epit,
1029 xs << XMLStream::ESCAPE_NONE << "<!-- DocBook output error! -->\n";
1033 std::stack<std::pair<int, string>> headerLevels; // Used to determine when to open/close sections: store the depth
1034 // of the section and the tag that was used to open it.
1036 // Detect whether the document contains sections. If there are no sections, treatment is largely simplified.
1037 // In particular, there can't be an abstract, unless it is manually marked.
1038 bool documentHasSections;
1040 tie(documentHasSections, eppit) = hasDocumentSectioning(paragraphs, bpit, epit);
1042 // Deal with "simple" documents, i.e. those without sections.
1043 if (!documentHasSections) {
1044 docbookSimpleAllParagraphs(text, buf, xs, runparams);
1048 // Output the first <info> tag (or just the title).
1049 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, eppit, true);
1050 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
1053 // Then, iterate through the paragraphs of this document.
1054 bool currentlyInAppendix = false;
1056 auto par = text.paragraphs().iterator_at(bpit);
1057 auto end = text.paragraphs().iterator_at(epit);
1058 while (par != end) {
1059 OutputParams ourparams = runparams;
1061 if (par->params().startOfAppendix())
1062 currentlyInAppendix = true;
1063 if (hasOnlyNotes(*par)) {
1068 Layout const &style = par->layout();
1070 // Think about adding <section> and/or </section>s.
1071 if (isLayoutSectioning(style)) {
1072 int level = style.toclevel;
1074 // Need to close a previous section if it has the same level or a higher one (close <section> if opening a
1075 // <h2> after a <h2>, <h3>, <h4>, <h5> or <h6>). More examples:
1076 // - current: h2; back: h1; do not close any <section>
1077 // - current: h1; back: h2; close two <section> (first the <h2>, then the <h1>, so a new <h1> can come)
1078 while (!headerLevels.empty() && level <= headerLevels.top().first) {
1079 // Output the tag only if it corresponds to a legit section.
1080 int stackLevel = headerLevels.top().first;
1081 if (stackLevel != Layout::NOT_IN_TOC) {
1082 xs << xml::EndTag(headerLevels.top().second);
1088 // Open the new section: first push it onto the stack, then output it in DocBook.
1089 string sectionTag = (currentlyInAppendix && style.docbooksectiontag() == "chapter") ?
1090 "appendix" : style.docbooksectiontag();
1091 headerLevels.push(std::make_pair(level, sectionTag));
1093 // Some sectioning-like elements should not be output (such as FrontMatter).
1094 if (level != Layout::NOT_IN_TOC) {
1095 // Look for a label in the title, i.e. a InsetLabel as a child.
1096 docstring id = docstring();
1097 for (pos_type i = 0; i < par->size(); ++i) {
1098 Inset const *inset = par->getInset(i);
1100 if (auto label = dynamic_cast<InsetLabel const *>(inset)) {
1101 // Generate the attributes for the section if need be.
1102 id += "xml:id=\"" + xml::cleanID(label->screenLabel()) + "\"";
1104 // Don't output the ID as a DocBook <anchor>.
1105 ourparams.docbook_anchors_to_ignore.emplace(label->screenLabel());
1107 // Cannot have multiple IDs per tag.
1113 // Write the open tag for this section.
1117 xs << xml::StartTag(sectionTag, attrs);
1122 // Close all sections before the bibliography.
1123 // 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)?
1124 auto insetsLength = distance(par->insetList().begin(), par->insetList().end());
1125 if (insetsLength > 0) {
1126 Inset const *firstInset = par->getInset(0);
1127 if (firstInset && (firstInset->lyxCode() == BIBITEM_CODE || firstInset->lyxCode() == BIBTEX_CODE)) {
1128 while (!headerLevels.empty()) {
1129 int level = headerLevels.top().first;
1130 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1133 // Output the tag only if it corresponds to a legit section.
1134 if (level != Layout::NOT_IN_TOC) {
1135 xs << XMLStream::ESCAPE_NONE << tag;
1142 // Generate this paragraph.
1143 par = makeAny(text, buf, xs, ourparams, par);
1145 // Some special sections may require abstracts (mostly parts, in books).
1146 // 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.
1147 if (isLayoutSectioning(style) && style.docbookforceabstracttag() != "NONE") {
1148 // This abstract may be found between the next paragraph and the next title.
1149 pit_type cpit = std::distance(text.paragraphs().begin(), par);
1150 pit_type ppit = std::get<1>(hasDocumentSectioning(paragraphs, cpit, epit));
1152 // Generate this abstract (this code corresponds to parts of outputDocBookInfo).
1153 DocBookInfoTag secInfo = getParagraphsWithInfo(paragraphs, cpit, ppit, true);
1155 if (!secInfo.abstract.empty()) {
1156 xs << xml::StartTag(style.docbookforceabstracttag());
1158 for (auto const &p : secInfo.abstract)
1159 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(p));
1160 xs << xml::EndTag(style.docbookforceabstracttag());
1164 // Skip all the text that just has been generated.
1165 par = paragraphs.iterator_at(ppit);
1169 // If need be, close <section>s, but only at the end of the document (otherwise, dealt with at the beginning
1171 while (!headerLevels.empty() && headerLevels.top().first > Layout::NOT_IN_TOC) {
1172 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1174 xs << XMLStream::ESCAPE_NONE << tag;