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.
14 #include "output_docbook.h"
17 #include "buffer_funcs.h"
18 #include "BufferParams.h"
20 #include "InsetList.h"
21 #include "Paragraph.h"
22 #include "ParagraphList.h"
23 #include "ParagraphParameters.h"
26 #include "TextClass.h"
28 #include "insets/InsetBibtex.h"
29 #include "insets/InsetBibitem.h"
30 #include "insets/InsetLabel.h"
31 #include "mathed/InsetMath.h"
32 #include "insets/InsetNote.h"
34 #include "support/lassert.h"
35 #include "support/textutils.h"
43 using namespace lyx::support;
49 std::string fontToDocBookTag(xml::FontTypes type)
52 case xml::FontTypes::FT_EMPH:
53 case xml::FontTypes::FT_BOLD:
55 case xml::FontTypes::FT_NOUN:
57 case xml::FontTypes::FT_UBAR:
58 case xml::FontTypes::FT_WAVE:
59 case xml::FontTypes::FT_DBAR:
60 case xml::FontTypes::FT_SOUT:
61 case xml::FontTypes::FT_XOUT:
62 case xml::FontTypes::FT_ITALIC:
63 case xml::FontTypes::FT_UPRIGHT:
64 case xml::FontTypes::FT_SLANTED:
65 case xml::FontTypes::FT_SMALLCAPS:
66 case xml::FontTypes::FT_ROMAN:
67 case xml::FontTypes::FT_SANS:
69 case xml::FontTypes::FT_TYPE:
71 case xml::FontTypes::FT_SIZE_TINY:
72 case xml::FontTypes::FT_SIZE_SCRIPT:
73 case xml::FontTypes::FT_SIZE_FOOTNOTE:
74 case xml::FontTypes::FT_SIZE_SMALL:
75 case xml::FontTypes::FT_SIZE_NORMAL:
76 case xml::FontTypes::FT_SIZE_LARGE:
77 case xml::FontTypes::FT_SIZE_LARGER:
78 case xml::FontTypes::FT_SIZE_LARGEST:
79 case xml::FontTypes::FT_SIZE_HUGE:
80 case xml::FontTypes::FT_SIZE_HUGER:
81 case xml::FontTypes::FT_SIZE_INCREASE:
82 case xml::FontTypes::FT_SIZE_DECREASE:
90 string fontToRole(xml::FontTypes type)
92 // Specific fonts are achieved with roles. The only common ones are "" for basic emphasis,
93 // and "bold"/"strong" for bold. With some specific options, other roles are copied into
94 // HTML output (via the DocBook XSLT sheets); otherwise, if not recognised, they are just ignored.
95 // Hence, it is not a problem to have many roles by default here.
96 // See https://www.sourceware.org/ml/docbook/2003-05/msg00269.html
98 case xml::FontTypes::FT_ITALIC:
99 case xml::FontTypes::FT_EMPH:
101 case xml::FontTypes::FT_BOLD:
103 case xml::FontTypes::FT_NOUN: // Outputs a <person>
104 case xml::FontTypes::FT_TYPE: // Outputs a <code>
106 case xml::FontTypes::FT_UBAR:
109 // All other roles are non-standard for DocBook.
111 case xml::FontTypes::FT_WAVE:
113 case xml::FontTypes::FT_DBAR:
115 case xml::FontTypes::FT_SOUT:
117 case xml::FontTypes::FT_XOUT:
119 case xml::FontTypes::FT_UPRIGHT:
121 case xml::FontTypes::FT_SLANTED:
123 case xml::FontTypes::FT_SMALLCAPS:
125 case xml::FontTypes::FT_ROMAN:
127 case xml::FontTypes::FT_SANS:
129 case xml::FontTypes::FT_SIZE_TINY:
131 case xml::FontTypes::FT_SIZE_SCRIPT:
132 return "size_script";
133 case xml::FontTypes::FT_SIZE_FOOTNOTE:
134 return "size_footnote";
135 case xml::FontTypes::FT_SIZE_SMALL:
137 case xml::FontTypes::FT_SIZE_NORMAL:
138 return "size_normal";
139 case xml::FontTypes::FT_SIZE_LARGE:
141 case xml::FontTypes::FT_SIZE_LARGER:
142 return "size_larger";
143 case xml::FontTypes::FT_SIZE_LARGEST:
144 return "size_largest";
145 case xml::FontTypes::FT_SIZE_HUGE:
147 case xml::FontTypes::FT_SIZE_HUGER:
149 case xml::FontTypes::FT_SIZE_INCREASE:
150 return "size_increase";
151 case xml::FontTypes::FT_SIZE_DECREASE:
152 return "size_decrease";
159 string fontToAttribute(xml::FontTypes type) {
160 // If there is a role (i.e. nonstandard use of a tag), output the attribute. Otherwise, the sheer tag is sufficient
162 string role = fontToRole(type);
164 return "role='" + role + "'";
171 // Convenience functions to open and close tags. First, very low-level ones to ensure a consistent new-line behaviour.
175 // Contents of the block.
180 // <paratag>Contents of the paragraph.</paratag>
183 // Content before<inlinetag>Contents of the paragraph.</inlinetag>Content after
185 void openInlineTag(XMLStream & xs, const std::string & tag, const std::string & attr)
187 xs << xml::StartTag(tag, attr);
191 void closeInlineTag(XMLStream & xs, const std::string & tag)
193 xs << xml::EndTag(tag);
197 void openParTag(XMLStream & xs, const std::string & tag, const std::string & attr)
199 if (!xs.isLastTagCR())
201 xs << xml::StartTag(tag, attr);
205 void closeParTag(XMLStream & xs, const std::string & tag)
207 xs << xml::EndTag(tag);
212 void openBlockTag(XMLStream & xs, const std::string & tag, const std::string & attr)
214 if (!xs.isLastTagCR())
216 xs << xml::StartTag(tag, attr);
221 void closeBlockTag(XMLStream & xs, const std::string & tag)
223 if (!xs.isLastTagCR())
225 xs << xml::EndTag(tag);
230 void openTag(XMLStream & xs, const std::string & tag, const std::string & attr, const std::string & tagtype)
232 if (tag.empty() || tag == "NONE") // Common check to be performed elsewhere, if it was not here.
235 if (tag == "para" || tagtype == "paragraph") // Special case for <para>: always considered as a paragraph.
236 openParTag(xs, tag, attr);
237 else if (tagtype == "block")
238 openBlockTag(xs, tag, attr);
239 else if (tagtype == "inline")
240 openInlineTag(xs, tag, attr);
242 xs.writeError("Unrecognised tag type '" + tagtype + "' for '" + tag + " " + attr + "'");
246 void closeTag(XMLStream & xs, const std::string & tag, const std::string & tagtype)
248 if (tag.empty() || tag == "NONE")
251 if (tag == "para" || tagtype == "paragraph") // Special case for <para>: always considered as a paragraph.
252 closeParTag(xs, tag);
253 else if (tagtype == "block")
254 closeBlockTag(xs, tag);
255 else if (tagtype == "inline")
256 closeInlineTag(xs, tag);
258 xs.writeError("Unrecognised tag type '" + tagtype + "' for '" + tag + "'");
262 void compTag(XMLStream & xs, const std::string & tag, const std::string & attr, const std::string & tagtype)
264 if (tag.empty() || tag == "NONE")
267 // Special case for <para>: always considered as a paragraph.
268 if (tag == "para" || tagtype == "paragraph" || tagtype == "block") {
269 if (!xs.isLastTagCR())
271 xs << xml::CompTag(tag, attr);
273 } else if (tagtype == "inline") {
274 xs << xml::CompTag(tag, attr);
276 xs.writeError("Unrecognised tag type '" + tagtype + "' for '" + tag + "'");
281 // Higher-level convenience functions.
283 void openParTag(XMLStream & xs, const Paragraph * par, const Paragraph * prevpar)
285 Layout const & lay = par->layout();
290 // When should the wrapper be opened here? Only if the previous paragraph has the SAME wrapper tag
291 // (usually, they won't have the same layout) and the CURRENT one allows merging.
292 // The main use case is author information in several paragraphs: if the name of the author is the
293 // first paragraph of an author, then merging with the previous tag does not make sense. Say the
294 // next paragraph is the affiliation, then it should be output in the same <author> tag (different
295 // layout, same wrapper tag).
296 bool openWrapper = lay.docbookwrappertag() != "NONE";
297 if (prevpar != nullptr) {
298 Layout const & prevlay = prevpar->layout();
299 if (prevlay.docbookwrappertag() != "NONE") {
300 if (prevlay.docbookwrappertag() == lay.docbookwrappertag() &&
301 prevlay.docbookwrapperattr() == lay.docbookwrapperattr())
302 openWrapper = !lay.docbookwrappermergewithprevious();
310 openTag(xs, lay.docbookwrappertag(), lay.docbookwrapperattr(), lay.docbookwrappertagtype());
312 const string & tag = lay.docbooktag();
314 auto xmltag = xml::ParTag(tag, lay.docbookattr());
315 if (!xs.isTagOpen(xmltag, 1)) { // Don't nest a paragraph directly in a paragraph.
316 // TODO: required or not?
317 // TODO: avoid creating a ParTag object just for this query...
318 openTag(xs, lay.docbooktag(), lay.docbookattr(), lay.docbooktagtype());
319 openTag(xs, lay.docbookinnertag(), lay.docbookinnerattr(), lay.docbookinnertagtype());
323 openTag(xs, lay.docbookitemtag(), lay.docbookitemattr(), lay.docbookitemtagtype());
324 openTag(xs, lay.docbookiteminnertag(), lay.docbookiteminnerattr(), lay.docbookiteminnertagtype());
328 void closeParTag(XMLStream & xs, Paragraph const * par, Paragraph const * nextpar)
333 // See comment in openParTag.
334 Layout const & lay = par->layout();
335 bool closeWrapper = lay.docbookwrappertag() != "NONE";
336 if (nextpar != nullptr) {
337 Layout const & nextlay = nextpar->layout();
338 if (nextlay.docbookwrappertag() != "NONE") {
339 if (nextlay.docbookwrappertag() == lay.docbookwrappertag() &&
340 nextlay.docbookwrapperattr() == lay.docbookwrapperattr())
341 closeWrapper = !nextlay.docbookwrappermergewithprevious();
348 closeTag(xs, lay.docbookiteminnertag(), lay.docbookiteminnertagtype());
349 closeTag(xs, lay.docbookitemtag(), lay.docbookitemtagtype());
350 closeTag(xs, lay.docbookinnertag(), lay.docbookinnertagtype());
351 closeTag(xs, lay.docbooktag(), lay.docbooktagtype());
353 closeTag(xs, lay.docbookwrappertag(), lay.docbookwrappertagtype());
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 auto id = xml::cleanID(bibitem->getParam("key"));
380 attr = from_utf8("xml:id='") + id + from_utf8("'");
384 xs << xml::StartTag(from_utf8("bibliomixed"), attr);
386 // Generate the entry. Concatenate the different parts of the paragraph if any.
387 auto const begin = text.paragraphs().begin();
388 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(std::distance(begin, par)), 0);
389 for (auto & parXML : pars)
390 xs << XMLStream::ESCAPE_NONE << parXML;
392 // End the precooked bibliography entry.
393 xs << xml::EndTag("bibliomixed");
396 // If this is the last paragraph in a bibliography, close the bibliography tag.
397 auto const end = text.paragraphs().end();
400 bool endBibliography = nextpar == end || nextpar->layout().latextype != LATEX_BIB_ENVIRONMENT;
402 if (endBibliography) {
403 xs << xml::EndTag("bibliography");
413 OutputParams const & runparams,
414 ParagraphList::const_iterator const & par)
416 auto const begin = text.paragraphs().begin();
417 auto const end = text.paragraphs().end();
418 auto prevpar = text.paragraphs().getParagraphBefore(par);
420 // We want to open the paragraph tag if:
421 // (i) the current layout permits multiple paragraphs
422 // (ii) we are either not already inside a paragraph (HTMLIsBlock) OR
423 // we are, but this is not the first paragraph
425 // But there is also a special case, and we first see whether we are in it.
426 // We do not want to open the paragraph tag if this paragraph contains
427 // only one item, and that item is "inline", i.e., not HTMLIsBlock (such
428 // as a branch). On the other hand, if that single item has a font change
429 // applied to it, then we still do need to open the paragraph.
431 // Obviously, this is very fragile. The main reason we need to do this is
432 // because of branches, e.g., a branch that contains an entire new section.
433 // We do not really want to wrap that whole thing in a <div>...</div>.
434 bool special_case = false;
435 Inset const *specinset = par->size() == 1 ? par->getInset(0) : nullptr;
436 if (specinset && !specinset->getLayout().htmlisblock()) { // TODO: Convert htmlisblock to a DocBook parameter?
437 Layout const &style = par->layout();
438 FontInfo const first_font = style.labeltype == LABEL_MANUAL ?
439 style.labelfont : style.font;
440 FontInfo const our_font =
441 par->getFont(buf.masterBuffer()->params(), 0,
442 text.outerFont(std::distance(begin, par))).fontInfo();
444 if (first_font == our_font)
448 size_t nInsets = std::distance(par->insetList().begin(), par->insetList().end());
450 // Plain layouts must be ignored.
451 special_case |= buf.params().documentClass().isPlainLayout(par->layout()) && !runparams.docbook_force_pars;
452 // Equations do not deserve their own paragraph (DocBook allows them outside paragraphs).
453 // Exception: any case that generates an <inlineequation> must still get a paragraph to be valid.
454 special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
455 return inset.inset && inset.inset->asInsetMath() && inset.inset->asInsetMath()->getType() != hullSimple;
457 // Floats cannot be in paragraphs.
458 special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
459 return inset.inset->lyxCode() == FLOAT_CODE;
461 // Bibliographies cannot be in paragraphs. Bibitems should still be handled as paragraphs, though
462 // (see makeParagraphBibliography).
463 special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
464 return inset.inset->lyxCode() == BIBTEX_CODE;
466 // ERTs are in comments, not paragraphs.
467 special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
468 return inset.inset->lyxCode() == ERT_CODE;
470 // Listings should not get into their own paragraph.
471 special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
472 return inset.inset->lyxCode() == LISTINGS_CODE;
474 // Boxes cannot get into their own paragraph.
475 special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
476 return inset.inset->lyxCode() == BOX_CODE;
478 // Includes should not have a paragraph.
479 special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
480 return inset.inset->lyxCode() == INCLUDE_CODE;
483 bool const open_par = runparams.docbook_make_pars
484 && !runparams.docbook_in_par
487 // We want to issue the closing tag if either:
488 // (i) We opened it, and either docbook_in_par is false,
489 // or we're not in the last paragraph, anyway.
490 // (ii) We didn't open it and docbook_in_par is true,
491 // but we are in the first par, and there is a next par.
492 bool const close_par = open_par && (!runparams.docbook_in_par);
494 // Determine if this paragraph has some real content. Things like new pages are not caught
495 // by Paragraph::empty(), even though they do not generate anything useful in DocBook.
496 // Thus, remove all spaces (including new lines: \r, \n) before checking for emptiness.
497 // std::all_of allows doing this check without having to copy the string.
498 // Open and close tags around each contained paragraph.
501 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(distance(begin, par)), 0, nextpar == end, special_case);
502 for (docstring const & parXML : pars) {
503 if (xml::isNotOnlySpace(parXML)) {
505 openParTag(xs, &*par, prevpar);
507 xs << XMLStream::ESCAPE_NONE << parXML;
510 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
516 void makeEnvironment(Text const &text,
519 OutputParams const &runparams,
520 ParagraphList::const_iterator const & par)
522 auto const end = text.paragraphs().end();
526 // Special cases for listing-like environments provided in layouts. This is quite ad-hoc, but provides a useful
527 // default. This should not be used by too many environments (only LyX-Code right now).
528 // This would be much simpler if LyX-Code was implemented as InsetListings...
529 bool mimicListing = false;
530 bool ignoreFonts = false;
531 if (par->layout().docbooktag() == "programlisting") {
536 // Output the opening tag for this environment, but only if it has not been previously opened (condition
537 // implemented in openParTag).
538 auto prevpar = text.paragraphs().getParagraphBefore(par);
539 openParTag(xs, &*par, prevpar); // TODO: switch in layout for par/block?
541 // Generate the contents of this environment. There is a special case if this is like some environment.
542 Layout const & style = par->layout();
543 if (style.latextype == LATEX_COMMAND) {
544 // Nothing to do (otherwise, infinite loops).
545 } else if (style.latextype == LATEX_ENVIRONMENT) {
546 // Generate the paragraph, if need be.
547 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(std::distance(text.paragraphs().begin(), par)), 0, false, ignoreFonts);
550 auto p = pars.begin();
551 while (p != pars.end()) {
552 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(), par->layout().docbookiteminnertagtype());
553 xs << XMLStream::ESCAPE_NONE << *p;
554 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
557 // Insert a new line after each "paragraph" (i.e. line in the listing), except for the last one.
558 // Otherwise, there would one more new line in the output than in the LyX document.
563 for (auto const & p : pars) {
564 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(), par->layout().docbookiteminnertagtype());
565 xs << XMLStream::ESCAPE_NONE << p;
566 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
570 makeAny(text, buf, xs, runparams, par);
573 // Close the environment.
574 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr); // TODO: switch in layout for par/block?
578 ParagraphList::const_iterator findEndOfEnvironment(
579 ParagraphList::const_iterator const & pstart,
580 ParagraphList::const_iterator const & pend)
582 // Copy-paste from XHTML. Should be factored out at some point...
583 ParagraphList::const_iterator p = pstart;
584 Layout const & bstyle = p->layout();
585 size_t const depth = p->params().depth();
586 for (++p; p != pend; ++p) {
587 Layout const & style = p->layout();
588 // It shouldn't happen that e.g. a section command occurs inside
589 // a quotation environment, at a higher depth, but as of 6/2009,
590 // it can happen. We pretend that it's just at lowest depth.
591 if (style.latextype == LATEX_COMMAND)
594 // If depth is down, we're done
595 if (p->params().depth() < depth)
598 // If depth is up, we're not done
599 if (p->params().depth() > depth)
602 // FIXME I am not sure about the first check.
603 // Surely we *could* have different layouts that count as
604 // LATEX_PARAGRAPH, right?
605 if (style.latextype == LATEX_PARAGRAPH || style != bstyle)
612 ParagraphList::const_iterator makeListEnvironment(Text const &text,
615 OutputParams const &runparams,
616 ParagraphList::const_iterator const & begin)
619 auto const end = text.paragraphs().end();
620 auto const envend = findEndOfEnvironment(par, end);
622 // Output the opening tag for this environment.
623 Layout const & envstyle = par->layout();
624 openTag(xs, envstyle.docbookwrappertag(), envstyle.docbookwrapperattr(), envstyle.docbookwrappertagtype());
625 openTag(xs, envstyle.docbooktag(), envstyle.docbookattr(), envstyle.docbooktagtype());
627 // Handle the content of the list environment, item by item.
628 while (par != envend) {
629 // Skip this paragraph if it is both empty and the last one (otherwise, there may be deeper paragraphs after).
632 if (par->empty() && nextpar == envend)
635 // Open the item wrapper.
636 Layout const & style = par->layout();
637 openTag(xs, style.docbookitemwrappertag(), style.docbookitemwrapperattr(), style.docbookitemwrappertagtype());
639 // Generate the label, if need be. If it is taken from the text, sep != 0 and corresponds to the first
640 // character after the label.
642 if (style.labeltype != LABEL_NO_LABEL && style.docbookitemlabeltag() != "NONE") {
643 if (style.labeltype == LABEL_MANUAL) {
644 // Only variablelist gets here (or similar items defined as an extension in the layout).
645 openTag(xs, style.docbookitemlabeltag(), style.docbookitemlabelattr(), style.docbookitemlabeltagtype());
646 sep = 1 + par->firstWordDocBook(xs, runparams);
647 closeTag(xs, style.docbookitemlabeltag(), style.docbookitemlabeltagtype());
649 // Usual cases: maybe there is something specified at the layout level. Highly unlikely, though.
650 docstring const lbl = par->params().labelString();
653 openTag(xs, style.docbookitemlabeltag(), style.docbookitemlabelattr(), style.docbookitemlabeltagtype());
655 closeTag(xs, style.docbookitemlabeltag(), style.docbookitemlabeltagtype());
660 // Open the item (after the wrapper and the label).
661 openTag(xs, style.docbookitemtag(), style.docbookitemattr(), style.docbookitemtagtype());
663 // Generate the content of the item.
664 if (sep < par->size()) {
665 auto pars = par->simpleDocBookOnePar(buf, runparams,
666 text.outerFont(std::distance(text.paragraphs().begin(), par)), sep);
667 for (auto &p : pars) {
668 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
669 par->layout().docbookiteminnertagtype());
670 xs << XMLStream::ESCAPE_NONE << p;
671 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
674 // DocBook doesn't like emptiness.
675 compTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
676 par->layout().docbookiteminnertagtype());
679 // If the next item is deeper, it must go entirely within this item (do it recursively).
680 // By construction, with findEndOfEnvironment, depth can only stay constant or increase, never decrease.
681 depth_type currentDepth = par->getDepth();
683 while (par != envend && par->getDepth() != currentDepth)
684 par = makeAny(text, buf, xs, runparams, par);
685 // Usually, this loop only makes one iteration, except in complex scenarios, like an item with a paragraph,
686 // a list, and another paragraph; or an item with two types of list (itemise then enumerate, for instance).
689 closeTag(xs, style.docbookitemtag(), style.docbookitemtagtype());
690 closeTag(xs, style.docbookitemwrappertag(), style.docbookitemwrappertagtype());
693 // Close this environment in exactly the same way as it was opened.
694 closeTag(xs, envstyle.docbooktag(), envstyle.docbooktagtype());
695 closeTag(xs, envstyle.docbookwrappertag(), envstyle.docbookwrappertagtype());
705 OutputParams const & runparams,
706 ParagraphList::const_iterator const & par)
708 // Unlike XHTML, no need for labels, as they are handled by DocBook tags.
709 auto const begin = text.paragraphs().begin();
710 auto const end = text.paragraphs().end();
714 // Generate this command.
715 auto prevpar = text.paragraphs().getParagraphBefore(par);
716 openParTag(xs, &*par, prevpar);
718 auto pars = par->simpleDocBookOnePar(buf, runparams,text.outerFont(distance(begin, par)));
719 for (auto & parXML : pars)
720 // TODO: decide what to do with openParTag/closeParTag in new lines.
721 xs << XMLStream::ESCAPE_NONE << parXML;
723 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
727 bool isLayoutSectioning(Layout const & lay)
729 if (lay.docbooksection()) // Special case: some DocBook styles must be handled as sections.
731 else if (lay.category() == from_utf8("Sectioning")) // Generic case.
732 return lay.toclevel != Layout::NOT_IN_TOC;
737 using DocBookDocumentSectioning = tuple<bool, pit_type>;
740 struct DocBookInfoTag
742 const set<pit_type> shouldBeInInfo;
743 const set<pit_type> mustBeInInfo; // With the notable exception of the abstract!
744 const set<pit_type> abstract;
745 const bool abstractLayout;
749 DocBookInfoTag(const set<pit_type> & shouldBeInInfo, const set<pit_type> & mustBeInInfo,
750 const set<pit_type> & abstract, bool abstractLayout, pit_type bpit, pit_type epit) :
751 shouldBeInInfo(shouldBeInInfo), mustBeInInfo(mustBeInInfo), abstract(abstract),
752 abstractLayout(abstractLayout), bpit(bpit), epit(epit) {}
756 DocBookDocumentSectioning hasDocumentSectioning(ParagraphList const ¶graphs, pit_type bpit, pit_type const epit) {
757 bool documentHasSections = false;
759 while (bpit < epit) {
760 Layout const &style = paragraphs[bpit].layout();
761 documentHasSections |= isLayoutSectioning(style);
763 if (documentHasSections)
767 // Paragraphs before the first section: [ runparams.par_begin ; eppit )
769 return make_tuple(documentHasSections, bpit);
773 bool hasOnlyNotes(Paragraph const & par)
775 // Precondition: the paragraph is not empty. Otherwise, the function will always return true...
776 for (int i = 0; i < par.size(); ++i)
777 // If you find something that is not an inset (like actual text) or an inset that is not a note,
779 if (!par.isInset(i) || par.getInset(i)->lyxCode() != NOTE_CODE)
782 // An empty paragraph may still require some output.
783 if (par.layout().docbooksection())
786 // There should be really no content here.
791 DocBookInfoTag getParagraphsWithInfo(ParagraphList const ¶graphs,
792 pit_type bpit, pit_type const epit,
793 // Typically, bpit is the beginning of the document and epit the end of the
794 // document *or* the first section.
795 bool documentHasSections,
796 bool detectUnlayoutedAbstract
797 // Whether paragraphs with no specific layout should be detected as abstracts.
798 // For inner sections, an abstract should only be detected if it has a specific
799 // layout. For others, anything that might look like an abstract should be sought.
801 set<pit_type> shouldBeInInfo;
802 set<pit_type> mustBeInInfo;
803 set<pit_type> abstractWithLayout;
804 set<pit_type> abstractNoLayout;
806 // Find the first non empty paragraph by mutating bpit.
807 while (bpit < epit) {
808 Paragraph const &par = paragraphs[bpit];
809 if (par.empty() || hasOnlyNotes(par))
815 // Traverse everything that might belong to <info>.
816 bool hasAbstractLayout = false;
817 pit_type cpit = bpit;
818 for (; cpit < epit; ++cpit) {
819 // Skip paragraphs that don't generate anything in DocBook.
820 Paragraph const & par = paragraphs[cpit];
821 Layout const &style = par.layout();
822 if (hasOnlyNotes(par))
825 // There should never be any section here, except for the first paragraph (a title can be part of <info>).
826 // (Just a sanity check: if this fails, this function could end up processing the whole document.)
827 if (cpit != bpit && isLayoutSectioning(par.layout())) {
828 LYXERR0("Assertion failed: section found in potential <info> paragraphs.");
832 // If this is marked as an abstract by the layout, put it in the right set.
833 if (style.docbookabstract()) {
834 hasAbstractLayout = true;
835 abstractWithLayout.emplace(cpit);
839 // Based on layout information, store this paragraph in one set: should be in <info>, must be,
840 // or abstract (either because of layout or of position).
841 if (style.docbookininfo() == "always")
842 mustBeInInfo.emplace(cpit);
843 else if (style.docbookininfo() == "maybe")
844 shouldBeInInfo.emplace(cpit);
845 else if (documentHasSections && !hasAbstractLayout && detectUnlayoutedAbstract)
846 abstractNoLayout.emplace(cpit);
847 else // This should definitely not be in <info>.
850 // Now, cpit points to the first paragraph that no more has things that could go in <info>.
851 // bpit is the beginning of the <info> part.
853 return DocBookInfoTag(shouldBeInInfo, mustBeInInfo,
854 hasAbstractLayout ? abstractWithLayout : abstractNoLayout,
855 hasAbstractLayout, bpit, cpit);
858 } // end anonymous namespace
861 ParagraphList::const_iterator makeAny(Text const &text,
864 OutputParams const &runparams,
865 ParagraphList::const_iterator par)
867 switch (par->layout().latextype) {
869 makeCommand(text, buf, xs, runparams, par);
871 case LATEX_ENVIRONMENT:
872 makeEnvironment(text, buf, xs, runparams, par);
874 case LATEX_LIST_ENVIRONMENT:
875 case LATEX_ITEM_ENVIRONMENT:
876 // Only case when makeAny() might consume more than one paragraph.
877 return makeListEnvironment(text, buf, xs, runparams, par);
878 case LATEX_PARAGRAPH:
879 makeParagraph(text, buf, xs, runparams, par);
881 case LATEX_BIB_ENVIRONMENT:
882 makeBibliography(text, buf, xs, runparams, par);
890 xml::FontTag docbookStartFontTag(xml::FontTypes type)
892 return xml::FontTag(from_utf8(fontToDocBookTag(type)), from_utf8(fontToAttribute(type)), type);
896 xml::EndFontTag docbookEndFontTag(xml::FontTypes type)
898 return xml::EndFontTag(from_utf8(fontToDocBookTag(type)), type);
902 void outputDocBookInfo(
906 OutputParams const & runparams,
907 ParagraphList const & paragraphs,
908 DocBookInfoTag const & info)
910 // Perform an additional check on the abstract. Sometimes, there are many paragraphs that should go
911 // into the abstract, but none generates actual content. Thus, first generate to a temporary stream,
912 // then only create the <abstract> tag if these paragraphs generate some content.
913 // This check must be performed *before* a decision on whether or not to output <info> is made.
914 bool hasAbstract = !info.abstract.empty();
917 // Generate the abstract XML into a string before further checks.
918 // Usually, makeAny only generates one paragraph at a time. However, for the specific case of lists, it might
919 // generate more than one paragraph, as indicated in the return value.
920 odocstringstream os2;
923 set<pit_type> doneParas;
924 for (auto const & p : info.abstract) {
925 if (doneParas.find(p) == doneParas.end()) {
926 auto oldPar = paragraphs.iterator_at(p);
927 auto newPar = makeAny(text, buf, xs2, runparams, oldPar);
929 // Insert the indices of all the paragraphs that were just generated (typically, one).
930 // **Make the hypothesis that, when an abstract has a list, all its items are consecutive.**
932 while (oldPar != newPar) {
933 doneParas.emplace(id);
940 // Actually output the abstract if there is something to do. Don't count line feeds or spaces in this,
941 // even though they must be properly output if there is some abstract.
942 abstract = os2.str();
943 docstring cleaned = abstract;
944 cleaned.erase(std::remove_if(cleaned.begin(), cleaned.end(), lyx::isSpace), cleaned.end());
946 // Nothing? Then there is no abstract!
951 // The abstract must go in <info>. Otherwise, decide whether to open <info> based on the layouts.
952 bool needInfo = !info.mustBeInInfo.empty() || hasAbstract;
954 // Start the <info> tag if required.
956 xs.startDivision(false);
957 xs << xml::StartTag("info");
961 // Output the elements that should go in <info>, before and after the abstract.
962 for (auto pit : info.shouldBeInInfo) // Typically, the title: these elements are so important and ubiquitous
963 // that mandating a wrapper like <info> would repel users. Thus, generate them first.
964 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
965 for (auto pit : info.mustBeInInfo)
966 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
968 // If there is no title, generate one (required for the document to be valid).
969 // This code is called for the main document, for table cells, etc., so be precise in this condition.
970 if (text.isMainText() && info.shouldBeInInfo.empty() && !runparams.inInclude) {
971 xs << xml::StartTag("title");
972 xs << "Untitled Document";
973 xs << xml::EndTag("title");
977 // Always output the abstract as the last item of the <info>, as it requires special treatment (especially if
978 // it contains several paragraphs that are empty).
980 if (info.abstractLayout) {
981 xs << XMLStream::ESCAPE_NONE << abstract;
984 string tag = paragraphs[*info.abstract.begin()].layout().docbookforceabstracttag();
988 if (!xs.isLastTagCR())
991 xs << xml::StartTag(tag);
993 xs << XMLStream::ESCAPE_NONE << abstract;
994 xs << xml::EndTag(tag);
999 // End the <info> tag if it was started.
1001 if (!xs.isLastTagCR())
1004 xs << xml::EndTag("info");
1011 void docbookSimpleAllParagraphs(
1015 OutputParams const & runparams)
1017 // Handle the given text, supposing it has no sections (i.e. a "simple" text). The input may vary in length
1018 // between a single paragraph to a whole document.
1019 pit_type const bpit = runparams.par_begin;
1020 pit_type const epit = runparams.par_end;
1021 ParagraphList const ¶graphs = text.paragraphs();
1023 // First, the <info> tag.
1024 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, epit, false, true);
1025 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
1027 // Then, the content. It starts where the <info> ends.
1028 auto par = paragraphs.iterator_at(info.epit);
1029 auto end = paragraphs.iterator_at(epit);
1030 while (par != end) {
1031 if (!hasOnlyNotes(*par))
1032 par = makeAny(text, buf, xs, runparams, par);
1039 void docbookParagraphs(Text const &text,
1042 OutputParams const &runparams) {
1043 ParagraphList const ¶graphs = text.paragraphs();
1044 if (runparams.par_begin == runparams.par_end) {
1045 runparams.par_begin = 0;
1046 runparams.par_end = paragraphs.size();
1048 pit_type bpit = runparams.par_begin;
1049 pit_type const epit = runparams.par_end;
1050 LASSERT(bpit < epit,
1052 xs << XMLStream::ESCAPE_NONE << "<!-- DocBook output error! -->\n";
1056 std::stack<std::pair<int, string>> headerLevels; // Used to determine when to open/close sections: store the depth
1057 // of the section and the tag that was used to open it.
1059 // Detect whether the document contains sections. If there are no sections, treatment is largely simplified.
1060 // In particular, there can't be an abstract, unless it is manually marked.
1061 bool documentHasSections;
1063 tie(documentHasSections, eppit) = hasDocumentSectioning(paragraphs, bpit, epit);
1065 // Deal with "simple" documents, i.e. those without sections.
1066 if (!documentHasSections) {
1067 docbookSimpleAllParagraphs(text, buf, xs, runparams);
1071 // Output the first <info> tag (or just the title).
1072 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, eppit, true, true);
1073 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
1076 // Then, iterate through the paragraphs of this document.
1077 bool currentlyInAppendix = false;
1079 auto par = text.paragraphs().iterator_at(bpit);
1080 auto end = text.paragraphs().iterator_at(epit);
1081 while (par != end) {
1082 OutputParams ourparams = runparams;
1084 if (par->params().startOfAppendix())
1085 currentlyInAppendix = true;
1086 if (hasOnlyNotes(*par)) {
1091 Layout const &style = par->layout();
1093 // Think about adding <section> and/or </section>s.
1094 if (isLayoutSectioning(style)) {
1095 int level = style.toclevel;
1097 // Need to close a previous section if it has the same level or a higher one (close <section> if opening a
1098 // <h2> after a <h2>, <h3>, <h4>, <h5> or <h6>). More examples:
1099 // - current: h2; back: h1; do not close any <section>
1100 // - current: h1; back: h2; close two <section> (first the <h2>, then the <h1>, so a new <h1> can come)
1101 while (!headerLevels.empty() && level <= headerLevels.top().first) {
1102 // Output the tag only if it corresponds to a legit section.
1103 int stackLevel = headerLevels.top().first;
1104 if (stackLevel != Layout::NOT_IN_TOC) {
1105 xs << xml::EndTag(headerLevels.top().second);
1111 // Open the new section: first push it onto the stack, then output it in DocBook.
1112 string sectionTag = (currentlyInAppendix && style.docbooksectiontag() == "chapter") ?
1113 "appendix" : style.docbooksectiontag();
1114 headerLevels.push(std::make_pair(level, sectionTag));
1116 // Some sectioning-like elements should not be output (such as FrontMatter).
1117 if (level != Layout::NOT_IN_TOC) {
1118 // Look for a label in the title, i.e. a InsetLabel as a child.
1119 docstring id = docstring();
1120 for (pos_type i = 0; i < par->size(); ++i) {
1121 Inset const *inset = par->getInset(i);
1123 if (auto label = dynamic_cast<InsetLabel const *>(inset)) {
1124 // Generate the attributes for the section if need be.
1125 id += "xml:id=\"" + xml::cleanID(label->screenLabel()) + "\"";
1127 // Don't output the ID as a DocBook <anchor>.
1128 ourparams.docbook_anchors_to_ignore.emplace(label->screenLabel());
1130 // Cannot have multiple IDs per tag. If there is another ID inset in the document, it will
1131 // be output as a DocBook anchor.
1137 // Write the open tag for this section.
1141 xs << xml::StartTag(sectionTag, attrs);
1146 // Close all sections before the bibliography.
1147 // 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)?
1148 if (!par->insetList().empty()) {
1149 Inset const *firstInset = par->getInset(0);
1150 if (firstInset && (firstInset->lyxCode() == BIBITEM_CODE || firstInset->lyxCode() == BIBTEX_CODE)) {
1151 while (!headerLevels.empty()) {
1152 int level = headerLevels.top().first;
1153 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1156 // Output the tag only if it corresponds to a legit section.
1157 if (level != Layout::NOT_IN_TOC) {
1158 xs << XMLStream::ESCAPE_NONE << tag;
1165 // Generate the <info> tag if a section was just opened.
1166 // Some sections may require abstracts (mostly parts, in books: DocBookForceAbstractTag will not be NONE),
1167 // others can still have an abstract (it must be detected so that it can be output at the right place).
1168 // 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.
1169 if (isLayoutSectioning(style)) {
1170 // This abstract may be found between the next paragraph and the next title.
1171 pit_type cpit = std::distance(text.paragraphs().begin(), par);
1172 pit_type ppit = std::get<1>(hasDocumentSectioning(paragraphs, cpit + 1L, epit));
1174 // Generate this abstract (this code corresponds to parts of outputDocBookInfo).
1175 DocBookInfoTag secInfo = getParagraphsWithInfo(paragraphs, cpit, ppit, true,
1176 style.docbookforceabstracttag() != "NONE");
1178 if (!secInfo.mustBeInInfo.empty() || !secInfo.shouldBeInInfo.empty() || !secInfo.abstract.empty()) {
1179 // Generate the <info>, if required. If DocBookForceAbstractTag != NONE, this abstract will not be in
1180 // <info>, unlike other ("standard") abstracts.
1181 bool hasStandardAbstract = !secInfo.abstract.empty() && style.docbookforceabstracttag() == "NONE";
1182 bool needInfo = !secInfo.mustBeInInfo.empty() || hasStandardAbstract;
1185 xs.startDivision(false);
1186 xs << xml::StartTag("info");
1190 // Output the elements that should go in <info>, before and after the abstract.
1191 for (auto pit : secInfo.shouldBeInInfo) // Typically, the title: these elements are so important and ubiquitous
1192 // that mandating a wrapper like <info> would repel users. Thus, generate them first.
1193 makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(pit));
1194 for (auto pit : secInfo.mustBeInInfo)
1195 makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(pit));
1197 // Deal with the abstract in <info> if it is standard (i.e. its tag is <abstract>).
1198 if (!secInfo.abstract.empty() && hasStandardAbstract) {
1199 if (!secInfo.abstractLayout) {
1200 xs << xml::StartTag("abstract");
1204 for (auto const &p : secInfo.abstract)
1205 makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(p));
1207 if (!secInfo.abstractLayout) {
1208 xs << xml::EndTag("abstract");
1213 // End the <info> tag if it was started.
1215 if (!xs.isLastTagCR())
1218 xs << xml::EndTag("info");
1223 // Deal with the abstract outside <info> if it is not standard (i.e. its tag is layout-defined).
1224 if (!secInfo.abstract.empty() && !hasStandardAbstract) {
1225 // Assert: style.docbookforceabstracttag() != NONE.
1226 xs << xml::StartTag(style.docbookforceabstracttag());
1228 for (auto const &p : secInfo.abstract)
1229 makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(p));
1230 xs << xml::EndTag(style.docbookforceabstracttag());
1234 // Skip all the text that has just been generated.
1235 par = paragraphs.iterator_at(secInfo.epit);
1237 // No <info> tag to generate, proceed as for normal paragraphs.
1238 par = makeAny(text, buf, xs, ourparams, par);
1241 // Generate this paragraph, as it has nothing special.
1242 par = makeAny(text, buf, xs, ourparams, par);
1246 // If need be, close <section>s, but only at the end of the document (otherwise, dealt with at the beginning
1248 while (!headerLevels.empty() && headerLevels.top().first > Layout::NOT_IN_TOC) {
1249 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1251 xs << XMLStream::ESCAPE_NONE << tag;