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 attr = from_utf8("xml:id='") + bibitem->getParam("key") + from_utf8("'");
383 xs << xml::StartTag(from_utf8("bibliomixed"), attr);
385 // Generate the entry. Concatenate the different parts of the paragraph if any.
386 auto const begin = text.paragraphs().begin();
387 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(std::distance(begin, par)), 0);
388 for (auto & parXML : pars)
389 xs << XMLStream::ESCAPE_NONE << parXML;
391 // End the precooked bibliography entry.
392 xs << xml::EndTag("bibliomixed");
395 // If this is the last paragraph in a bibliography, close the bibliography tag.
396 auto const end = text.paragraphs().end();
399 bool endBibliography = nextpar == end || nextpar->layout().latextype != LATEX_BIB_ENVIRONMENT;
401 if (endBibliography) {
402 xs << xml::EndTag("bibliography");
412 OutputParams const & runparams,
413 ParagraphList::const_iterator const & par)
415 auto const begin = text.paragraphs().begin();
416 auto const end = text.paragraphs().end();
417 auto prevpar = text.paragraphs().getParagraphBefore(par);
419 // We want to open the paragraph tag if:
420 // (i) the current layout permits multiple paragraphs
421 // (ii) we are either not already inside a paragraph (HTMLIsBlock) OR
422 // we are, but this is not the first paragraph
424 // But there is also a special case, and we first see whether we are in it.
425 // We do not want to open the paragraph tag if this paragraph contains
426 // only one item, and that item is "inline", i.e., not HTMLIsBlock (such
427 // as a branch). On the other hand, if that single item has a font change
428 // applied to it, then we still do need to open the paragraph.
430 // Obviously, this is very fragile. The main reason we need to do this is
431 // because of branches, e.g., a branch that contains an entire new section.
432 // We do not really want to wrap that whole thing in a <div>...</div>.
433 bool special_case = false;
434 Inset const *specinset = par->size() == 1 ? par->getInset(0) : nullptr;
435 if (specinset && !specinset->getLayout().htmlisblock()) { // TODO: Convert htmlisblock to a DocBook parameter?
436 Layout const &style = par->layout();
437 FontInfo const first_font = style.labeltype == LABEL_MANUAL ?
438 style.labelfont : style.font;
439 FontInfo const our_font =
440 par->getFont(buf.masterBuffer()->params(), 0,
441 text.outerFont(std::distance(begin, par))).fontInfo();
443 if (first_font == our_font)
447 size_t nInsets = std::distance(par->insetList().begin(), par->insetList().end());
449 // Plain layouts must be ignored.
450 special_case |= buf.params().documentClass().isPlainLayout(par->layout()) && !runparams.docbook_force_pars;
451 // Equations do not deserve their own paragraph (DocBook allows them outside paragraphs).
452 // Exception: any case that generates an <inlineequation> must still get a paragraph to be valid.
453 special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
454 return inset.inset && inset.inset->asInsetMath() && inset.inset->asInsetMath()->getType() != hullSimple;
456 // Floats cannot be in paragraphs.
457 special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
458 return inset.inset->lyxCode() == FLOAT_CODE;
460 // Bibliographies cannot be in paragraphs. Bibitems should still be handled as paragraphs, though
461 // (see makeParagraphBibliography).
462 special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
463 return inset.inset->lyxCode() == BIBTEX_CODE;
465 // ERTs are in comments, not paragraphs.
466 special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
467 return inset.inset->lyxCode() == ERT_CODE;
469 // Listings should not get into their own paragraph.
470 special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
471 return inset.inset->lyxCode() == LISTINGS_CODE;
473 // Boxes cannot get into their own paragraph.
474 special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
475 return inset.inset->lyxCode() == BOX_CODE;
477 // Includes should not have a paragraph.
478 special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
479 return inset.inset->lyxCode() == INCLUDE_CODE;
482 bool const open_par = runparams.docbook_make_pars
483 && !runparams.docbook_in_par
486 // We want to issue the closing tag if either:
487 // (i) We opened it, and either docbook_in_par is false,
488 // or we're not in the last paragraph, anyway.
489 // (ii) We didn't open it and docbook_in_par is true,
490 // but we are in the first par, and there is a next par.
491 bool const close_par = open_par && (!runparams.docbook_in_par);
493 // Determine if this paragraph has some real content. Things like new pages are not caught
494 // by Paragraph::empty(), even though they do not generate anything useful in DocBook.
495 // Thus, remove all spaces (including new lines: \r, \n) before checking for emptiness.
496 // std::all_of allows doing this check without having to copy the string.
497 // Open and close tags around each contained paragraph.
500 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(distance(begin, par)), 0, nextpar == end, special_case);
501 for (docstring const & parXML : pars) {
502 if (xml::isNotOnlySpace(parXML)) {
504 openParTag(xs, &*par, prevpar);
506 xs << XMLStream::ESCAPE_NONE << parXML;
509 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
515 void makeEnvironment(Text const &text,
518 OutputParams const &runparams,
519 ParagraphList::const_iterator const & par)
521 auto const end = text.paragraphs().end();
525 // Special cases for listing-like environments provided in layouts. This is quite ad-hoc, but provides a useful
526 // default. This should not be used by too many environments (only LyX-Code right now).
527 // This would be much simpler if LyX-Code was implemented as InsetListings...
528 bool mimicListing = false;
529 bool ignoreFonts = false;
530 if (par->layout().docbooktag() == "programlisting") {
535 // Output the opening tag for this environment, but only if it has not been previously opened (condition
536 // implemented in openParTag).
537 auto prevpar = text.paragraphs().getParagraphBefore(par);
538 openParTag(xs, &*par, prevpar); // TODO: switch in layout for par/block?
540 // Generate the contents of this environment. There is a special case if this is like some environment.
541 Layout const & style = par->layout();
542 if (style.latextype == LATEX_COMMAND) {
543 // Nothing to do (otherwise, infinite loops).
544 } else if (style.latextype == LATEX_ENVIRONMENT) {
545 // Generate the paragraph, if need be.
546 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(std::distance(text.paragraphs().begin(), par)), 0, false, ignoreFonts);
549 auto p = pars.begin();
550 while (p != pars.end()) {
551 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(), par->layout().docbookiteminnertagtype());
552 xs << XMLStream::ESCAPE_NONE << *p;
553 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
560 for (auto const & p : pars) {
561 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(), par->layout().docbookiteminnertagtype());
562 xs << XMLStream::ESCAPE_NONE << p;
563 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
567 makeAny(text, buf, xs, runparams, par);
570 // Close the environment.
571 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr); // TODO: switch in layout for par/block?
575 ParagraphList::const_iterator findEndOfEnvironment(
576 ParagraphList::const_iterator const & pstart,
577 ParagraphList::const_iterator const & pend)
579 // Copy-paste from XHTML. Should be factored out at some point...
580 ParagraphList::const_iterator p = pstart;
581 Layout const & bstyle = p->layout();
582 size_t const depth = p->params().depth();
583 for (++p; p != pend; ++p) {
584 Layout const & style = p->layout();
585 // It shouldn't happen that e.g. a section command occurs inside
586 // a quotation environment, at a higher depth, but as of 6/2009,
587 // it can happen. We pretend that it's just at lowest depth.
588 if (style.latextype == LATEX_COMMAND)
591 // If depth is down, we're done
592 if (p->params().depth() < depth)
595 // If depth is up, we're not done
596 if (p->params().depth() > depth)
599 // FIXME I am not sure about the first check.
600 // Surely we *could* have different layouts that count as
601 // LATEX_PARAGRAPH, right?
602 if (style.latextype == LATEX_PARAGRAPH || style != bstyle)
609 ParagraphList::const_iterator makeListEnvironment(Text const &text,
612 OutputParams const &runparams,
613 ParagraphList::const_iterator const & begin)
616 auto const end = text.paragraphs().end();
617 auto const envend = findEndOfEnvironment(par, end);
619 // Output the opening tag for this environment.
620 Layout const & envstyle = par->layout();
621 openTag(xs, envstyle.docbookwrappertag(), envstyle.docbookwrapperattr(), envstyle.docbookwrappertagtype());
622 openTag(xs, envstyle.docbooktag(), envstyle.docbookattr(), envstyle.docbooktagtype());
624 // Handle the content of the list environment, item by item.
625 while (par != envend) {
626 Layout const & style = par->layout();
628 // Open the item wrapper.
629 openTag(xs, style.docbookitemwrappertag(), style.docbookitemwrapperattr(), style.docbookitemwrappertagtype());
631 // Generate the label, if need be. If it is taken from the text, sep != 0 and corresponds to the first
632 // character after the label.
634 if (style.labeltype != LABEL_NO_LABEL && style.docbookitemlabeltag() != "NONE") {
635 if (style.labeltype == LABEL_MANUAL) {
636 // Only variablelist gets here (or similar items defined as an extension in the layout).
637 openTag(xs, style.docbookitemlabeltag(), style.docbookitemlabelattr(), style.docbookitemlabeltagtype());
638 sep = 1 + par->firstWordDocBook(xs, runparams);
639 closeTag(xs, style.docbookitemlabeltag(), style.docbookitemlabeltagtype());
641 // Usual cases: maybe there is something specified at the layout level. Highly unlikely, though.
642 docstring const lbl = par->params().labelString();
645 openTag(xs, style.docbookitemlabeltag(), style.docbookitemlabelattr(), style.docbookitemlabeltagtype());
647 closeTag(xs, style.docbookitemlabeltag(), style.docbookitemlabeltagtype());
652 // Open the item (after the wrapper and the label).
653 openTag(xs, style.docbookitemtag(), style.docbookitemattr(), style.docbookitemtagtype());
655 // Generate the content of the item.
656 if (sep < par->size()) {
657 auto pars = par->simpleDocBookOnePar(buf, runparams,
658 text.outerFont(std::distance(text.paragraphs().begin(), par)), sep);
659 for (auto &p : pars) {
660 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
661 par->layout().docbookiteminnertagtype());
662 xs << XMLStream::ESCAPE_NONE << p;
663 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
666 // DocBook doesn't like emptiness.
667 compTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
668 par->layout().docbookiteminnertagtype());
671 // If the next item is deeper, it must go entirely within this item (do it recursively).
672 // By construction, with findEndOfEnvironment, depth can only stay constant or increase, never decrease.
673 depth_type currentDepth = par->getDepth();
675 while (par != envend && par->getDepth() != currentDepth)
676 par = makeAny(text, buf, xs, runparams, par);
677 // Usually, this loop only makes one iteration, except in complex scenarios, like an item with a paragraph,
678 // a list, and another paragraph; or an item with two types of list (itemise then enumerate, for instance).
681 closeTag(xs, style.docbookitemtag(), style.docbookitemtagtype());
682 closeTag(xs, style.docbookitemwrappertag(), style.docbookitemwrappertagtype());
685 // Close this environment in exactly the same way as it was opened.
686 closeTag(xs, envstyle.docbooktag(), envstyle.docbooktagtype());
687 closeTag(xs, envstyle.docbookwrappertag(), envstyle.docbookwrappertagtype());
697 OutputParams const & runparams,
698 ParagraphList::const_iterator const & par)
700 // Unlike XHTML, no need for labels, as they are handled by DocBook tags.
701 auto const begin = text.paragraphs().begin();
702 auto const end = text.paragraphs().end();
706 // Generate this command.
707 auto prevpar = text.paragraphs().getParagraphBefore(par);
708 openParTag(xs, &*par, prevpar);
710 auto pars = par->simpleDocBookOnePar(buf, runparams,text.outerFont(distance(begin, par)));
711 for (auto & parXML : pars)
712 // TODO: decide what to do with openParTag/closeParTag in new lines.
713 xs << XMLStream::ESCAPE_NONE << parXML;
715 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
719 bool isLayoutSectioning(Layout const & lay)
721 return lay.category() == from_utf8("Sectioning");
725 using DocBookDocumentSectioning = tuple<bool, pit_type>;
728 struct DocBookInfoTag
730 const set<pit_type> shouldBeInInfo;
731 const set<pit_type> mustBeInInfo; // With the notable exception of the abstract!
732 const set<pit_type> abstract;
733 const bool abstractLayout;
737 DocBookInfoTag(const set<pit_type> & shouldBeInInfo, const set<pit_type> & mustBeInInfo,
738 const set<pit_type> & abstract, bool abstractLayout, pit_type bpit, pit_type epit) :
739 shouldBeInInfo(shouldBeInInfo), mustBeInInfo(mustBeInInfo), abstract(abstract),
740 abstractLayout(abstractLayout), bpit(bpit), epit(epit) {}
744 DocBookDocumentSectioning hasDocumentSectioning(ParagraphList const ¶graphs, pit_type bpit, pit_type const epit) {
745 bool documentHasSections = false;
747 while (bpit < epit) {
748 Layout const &style = paragraphs[bpit].layout();
749 documentHasSections |= isLayoutSectioning(style);
751 if (documentHasSections)
755 // Paragraphs before the first section: [ runparams.par_begin ; eppit )
757 return make_tuple(documentHasSections, bpit);
761 bool hasOnlyNotes(Paragraph const & par)
763 // Precondition: the paragraph is not empty. Otherwise, the function will always return true...
764 for (int i = 0; i < par.size(); ++i)
765 // If you find something that is not an inset (like actual text) or an inset that is not a note,
767 if (!par.isInset(i) || par.getInset(i)->lyxCode() != NOTE_CODE)
773 DocBookInfoTag getParagraphsWithInfo(ParagraphList const ¶graphs,
774 pit_type bpit, pit_type const epit,
775 // Typically, bpit is the beginning of the document and epit the end *or* the first section.
776 bool documentHasSections) {
777 set<pit_type> shouldBeInInfo;
778 set<pit_type> mustBeInInfo;
779 set<pit_type> abstractWithLayout;
780 set<pit_type> abstractNoLayout;
782 // Find the first non empty paragraph by mutating bpit.
783 while (bpit < epit) {
784 Paragraph const &par = paragraphs[bpit];
785 if (par.empty() || hasOnlyNotes(par))
791 // Traverse everything that might belong to <info>.
792 bool hasAbstractLayout = false;
793 pit_type cpit = bpit;
794 for (; cpit < epit; ++cpit) {
795 // Skip paragraphs that don't generate anything in DocBook.
796 Paragraph const & par = paragraphs[cpit];
797 if (hasOnlyNotes(par))
800 // There should never be any section here. (Just a sanity check: if this fails, this function could end up
801 // processing the whole document.)
802 if (isLayoutSectioning(par.layout())) {
803 LYXERR0("Assertion failed: section found in potential <info> paragraphs.");
807 // If this is marked as an abstract by the layout, put it in the right set.
808 if (par.layout().docbookabstract()) {
809 hasAbstractLayout = true;
810 abstractWithLayout.emplace(cpit);
814 // Based on layout information, store this paragraph in one set: should be in <info>, must be,
815 // or abstract (either because of layout or of position).
816 Layout const &style = par.layout();
818 if (style.docbookininfo() == "always")
819 mustBeInInfo.emplace(cpit);
820 else if (style.docbookininfo() == "maybe")
821 shouldBeInInfo.emplace(cpit);
822 else if (documentHasSections && !hasAbstractLayout)
823 abstractNoLayout.emplace(cpit);
824 else // This should definitely not be in <info>.
827 // Now, cpit points to the first paragraph that no more has things that could go in <info>.
828 // bpit is the beginning of the <info> part.
830 return DocBookInfoTag(shouldBeInInfo, mustBeInInfo,
831 hasAbstractLayout ? abstractWithLayout : abstractNoLayout,
832 hasAbstractLayout, bpit, cpit);
835 } // end anonymous namespace
838 ParagraphList::const_iterator makeAny(Text const &text,
841 OutputParams const &runparams,
842 ParagraphList::const_iterator par)
844 switch (par->layout().latextype) {
846 makeCommand(text, buf, xs, runparams, par);
848 case LATEX_ENVIRONMENT:
849 makeEnvironment(text, buf, xs, runparams, par);
851 case LATEX_LIST_ENVIRONMENT:
852 case LATEX_ITEM_ENVIRONMENT:
853 // Only case when makeAny() might consume more than one paragraph.
854 return makeListEnvironment(text, buf, xs, runparams, par);
855 case LATEX_PARAGRAPH:
856 makeParagraph(text, buf, xs, runparams, par);
858 case LATEX_BIB_ENVIRONMENT:
859 makeBibliography(text, buf, xs, runparams, par);
867 xml::FontTag docbookStartFontTag(xml::FontTypes type)
869 return xml::FontTag(from_utf8(fontToDocBookTag(type)), from_utf8(fontToAttribute(type)), type);
873 xml::EndFontTag docbookEndFontTag(xml::FontTypes type)
875 return xml::EndFontTag(from_utf8(fontToDocBookTag(type)), type);
879 void outputDocBookInfo(
883 OutputParams const & runparams,
884 ParagraphList const & paragraphs,
885 DocBookInfoTag const & info)
887 // Perform an additional check on the abstract. Sometimes, there are many paragraphs that should go
888 // into the abstract, but none generates actual content. Thus, first generate to a temporary stream,
889 // then only create the <abstract> tag if these paragraphs generate some content.
890 // This check must be performed *before* a decision on whether or not to output <info> is made.
891 bool hasAbstract = !info.abstract.empty();
894 // Generate the abstract XML into a string before further checks.
895 odocstringstream os2;
897 for (auto const & p : info.abstract)
898 makeAny(text, buf, xs2, runparams, paragraphs.iterator_at(p));
900 // Actually output the abstract if there is something to do. Don't count line feeds or spaces in this,
901 // even though they must be properly output if there is some abstract.
902 abstract = os2.str();
903 docstring cleaned = abstract;
904 cleaned.erase(std::remove_if(cleaned.begin(), cleaned.end(), lyx::isSpace), cleaned.end());
906 // Nothing? Then there is no abstract!
911 // The abstract must go in <info>. Otherwise, decide whether to open <info> based on the layouts.
912 bool needInfo = !info.mustBeInInfo.empty() || hasAbstract;
914 // Start the <info> tag if required.
916 xs.startDivision(false);
917 xs << xml::StartTag("info");
921 // Output the elements that should go in <info>, before and after the abstract.
922 for (auto pit : info.shouldBeInInfo) // Typically, the title: these elements are so important and ubiquitous
923 // that mandating a wrapper like <info> would repel users. Thus, generate them first.
924 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
925 for (auto pit : info.mustBeInInfo)
926 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
928 // If there is no title, generate one (required for the document to be valid).
929 // This code is called for the main document, for table cells, etc., so be precise in this condition.
930 if (text.isMainText() && info.shouldBeInInfo.empty() && !runparams.inInclude) {
931 xs << xml::StartTag("title");
932 xs << "Untitled Document";
933 xs << xml::EndTag("title");
937 // Always output the abstract as the last item of the <info>, as it requires special treatment (especially if
938 // it contains several paragraphs that are empty).
940 if (info.abstractLayout) {
941 xs << XMLStream::ESCAPE_NONE << abstract;
944 string tag = paragraphs[*info.abstract.begin()].layout().docbookforceabstracttag();
948 if (!xs.isLastTagCR())
951 xs << xml::StartTag(tag);
953 xs << XMLStream::ESCAPE_NONE << abstract;
954 xs << xml::EndTag(tag);
959 // End the <info> tag if it was started.
961 if (!xs.isLastTagCR())
964 xs << xml::EndTag("info");
971 void docbookSimpleAllParagraphs(
975 OutputParams const & runparams)
977 // Handle the given text, supposing it has no sections (i.e. a "simple" text). The input may vary in length
978 // between a single paragraph to a whole document.
979 pit_type const bpit = runparams.par_begin;
980 pit_type const epit = runparams.par_end;
981 ParagraphList const ¶graphs = text.paragraphs();
983 // First, the <info> tag.
984 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, epit, false);
985 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
987 // Then, the content. It starts where the <info> ends.
988 auto par = paragraphs.iterator_at(info.epit);
989 auto end = paragraphs.iterator_at(epit);
991 if (!hasOnlyNotes(*par))
992 par = makeAny(text, buf, xs, runparams, par);
999 void docbookParagraphs(Text const &text,
1002 OutputParams const &runparams) {
1003 ParagraphList const ¶graphs = text.paragraphs();
1004 if (runparams.par_begin == runparams.par_end) {
1005 runparams.par_begin = 0;
1006 runparams.par_end = paragraphs.size();
1008 pit_type bpit = runparams.par_begin;
1009 pit_type const epit = runparams.par_end;
1010 LASSERT(bpit < epit,
1012 xs << XMLStream::ESCAPE_NONE << "<!-- DocBook output error! -->\n";
1016 std::stack<std::pair<int, string>> headerLevels; // Used to determine when to open/close sections: store the depth
1017 // of the section and the tag that was used to open it.
1019 // Detect whether the document contains sections. If there are no sections, treatment is largely simplified.
1020 // In particular, there can't be an abstract, unless it is manually marked.
1021 bool documentHasSections;
1023 tie(documentHasSections, eppit) = hasDocumentSectioning(paragraphs, bpit, epit);
1025 // Deal with "simple" documents, i.e. those without sections.
1026 if (!documentHasSections) {
1027 docbookSimpleAllParagraphs(text, buf, xs, runparams);
1031 // Output the first <info> tag (or just the title).
1032 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, eppit, true);
1033 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
1036 // Then, iterate through the paragraphs of this document.
1037 bool currentlyInAppendix = false;
1039 auto par = text.paragraphs().iterator_at(bpit);
1040 auto end = text.paragraphs().iterator_at(epit);
1041 while (par != end) {
1042 OutputParams ourparams = runparams;
1044 if (par->params().startOfAppendix())
1045 currentlyInAppendix = true;
1046 if (hasOnlyNotes(*par)) {
1051 Layout const &style = par->layout();
1053 // Think about adding <section> and/or </section>s.
1054 if (isLayoutSectioning(style)) {
1055 int level = style.toclevel;
1057 // Need to close a previous section if it has the same level or a higher one (close <section> if opening a
1058 // <h2> after a <h2>, <h3>, <h4>, <h5> or <h6>). More examples:
1059 // - current: h2; back: h1; do not close any <section>
1060 // - current: h1; back: h2; close two <section> (first the <h2>, then the <h1>, so a new <h1> can come)
1061 while (!headerLevels.empty() && level <= headerLevels.top().first) {
1062 // Output the tag only if it corresponds to a legit section.
1063 int stackLevel = headerLevels.top().first;
1064 if (stackLevel != Layout::NOT_IN_TOC) {
1065 xs << xml::EndTag(headerLevels.top().second);
1071 // Open the new section: first push it onto the stack, then output it in DocBook.
1072 string sectionTag = (currentlyInAppendix && style.docbooksectiontag() == "chapter") ?
1073 "appendix" : style.docbooksectiontag();
1074 headerLevels.push(std::make_pair(level, sectionTag));
1076 // Some sectioning-like elements should not be output (such as FrontMatter).
1077 if (level != Layout::NOT_IN_TOC) {
1078 // Look for a label in the title, i.e. a InsetLabel as a child.
1079 docstring id = docstring();
1080 for (pos_type i = 0; i < par->size(); ++i) {
1081 Inset const *inset = par->getInset(i);
1083 if (auto label = dynamic_cast<InsetLabel const *>(inset)) {
1084 // Generate the attributes for the section if need be.
1085 id += "xml:id=\"" + xml::cleanID(label->screenLabel()) + "\"";
1087 // Don't output the ID as a DocBook <anchor>.
1088 ourparams.docbook_anchors_to_ignore.emplace(label->screenLabel());
1090 // Cannot have multiple IDs per tag.
1096 // Write the open tag for this section.
1100 xs << xml::StartTag(sectionTag, attrs);
1105 // Close all sections before the bibliography.
1106 // 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)?
1107 if (!par->insetList().empty()) {
1108 Inset const *firstInset = par->getInset(0);
1109 if (firstInset && (firstInset->lyxCode() == BIBITEM_CODE || firstInset->lyxCode() == BIBTEX_CODE)) {
1110 while (!headerLevels.empty()) {
1111 int level = headerLevels.top().first;
1112 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1115 // Output the tag only if it corresponds to a legit section.
1116 if (level != Layout::NOT_IN_TOC) {
1117 xs << XMLStream::ESCAPE_NONE << tag;
1124 // Generate this paragraph.
1125 par = makeAny(text, buf, xs, ourparams, par);
1127 // Some special sections may require abstracts (mostly parts, in books).
1128 // 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.
1129 if (isLayoutSectioning(style) && style.docbookforceabstracttag() != "NONE") {
1130 // This abstract may be found between the next paragraph and the next title.
1131 pit_type cpit = std::distance(text.paragraphs().begin(), par);
1132 pit_type ppit = std::get<1>(hasDocumentSectioning(paragraphs, cpit, epit));
1134 // Generate this abstract (this code corresponds to parts of outputDocBookInfo).
1135 DocBookInfoTag secInfo = getParagraphsWithInfo(paragraphs, cpit, ppit, true);
1137 if (!secInfo.abstract.empty()) {
1138 xs << xml::StartTag(style.docbookforceabstracttag());
1140 for (auto const &p : secInfo.abstract)
1141 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(p));
1142 xs << xml::EndTag(style.docbookforceabstracttag());
1146 // Skip all the text that just has been generated.
1147 par = paragraphs.iterator_at(ppit);
1151 // If need be, close <section>s, but only at the end of the document (otherwise, dealt with at the beginning
1153 while (!headerLevels.empty() && headerLevels.top().first > Layout::NOT_IN_TOC) {
1154 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1156 xs << XMLStream::ESCAPE_NONE << tag;