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)
288 // When should the wrapper be opened here? Only if the previous paragraph has the SAME wrapper tag
289 // (usually, they won't have the same layout) and the CURRENT one allows merging.
290 // The main use case is author information in several paragraphs: if the name of the author is the
291 // first paragraph of an author, then merging with the previous tag does not make sense. Say the
292 // next paragraph is the affiliation, then it should be output in the same <author> tag (different
293 // layout, same wrapper tag).
294 Layout const & lay = par->layout();
295 bool openWrapper = lay.docbookwrappertag() != "NONE";
296 if (prevpar != nullptr) {
297 Layout const & prevlay = prevpar->layout();
298 if (prevlay.docbookwrappertag() != "NONE") {
299 if (prevlay.docbookwrappertag() == lay.docbookwrappertag() &&
300 prevlay.docbookwrapperattr() == lay.docbookwrapperattr())
301 openWrapper = !lay.docbookwrappermergewithprevious();
309 openTag(xs, lay.docbookwrappertag(), lay.docbookwrapperattr(), lay.docbookwrappertagtype());
311 const string & tag = lay.docbooktag();
313 auto xmltag = xml::ParTag(tag, lay.docbookattr());
314 if (!xs.isTagOpen(xmltag, 1)) { // Don't nest a paragraph directly in a paragraph.
315 // TODO: required or not?
316 // TODO: avoid creating a ParTag object just for this query...
317 openTag(xs, lay.docbooktag(), lay.docbookattr(), lay.docbooktagtype());
318 openTag(xs, lay.docbookinnertag(), lay.docbookinnerattr(), lay.docbookinnertagtype());
322 openTag(xs, lay.docbookitemtag(), lay.docbookitemattr(), lay.docbookitemtagtype());
323 openTag(xs, lay.docbookiteminnertag(), lay.docbookiteminnerattr(), lay.docbookiteminnertagtype());
327 void closeParTag(XMLStream & xs, Paragraph const * par, Paragraph const * nextpar)
332 // See comment in openParTag.
333 Layout const & lay = par->layout();
334 bool closeWrapper = lay.docbookwrappertag() != "NONE";
335 if (nextpar != nullptr) {
336 Layout const & nextlay = nextpar->layout();
337 if (nextlay.docbookwrappertag() != "NONE") {
338 if (nextlay.docbookwrappertag() == lay.docbookwrappertag() &&
339 nextlay.docbookwrapperattr() == lay.docbookwrapperattr())
340 closeWrapper = !nextlay.docbookwrappermergewithprevious();
347 closeTag(xs, lay.docbookiteminnertag(), lay.docbookiteminnertagtype());
348 closeTag(xs, lay.docbookitemtag(), lay.docbookitemtagtype());
349 closeTag(xs, lay.docbookinnertag(), lay.docbookinnertagtype());
350 closeTag(xs, lay.docbooktag(), lay.docbooktagtype());
352 closeTag(xs, lay.docbookwrappertag(), lay.docbookwrappertagtype());
356 void makeBibliography(
360 OutputParams const & runparams,
361 ParagraphList::const_iterator const & par)
363 // If this is the first paragraph in a bibliography, open the bibliography tag.
364 auto const * pbegin_before = text.paragraphs().getParagraphBefore(par);
365 if (pbegin_before == nullptr || (pbegin_before && pbegin_before->layout().latextype != LATEX_BIB_ENVIRONMENT)) {
366 xs << xml::StartTag("bibliography");
370 // Start the precooked bibliography entry. This is very much like opening a paragraph tag.
371 // Don't forget the citation ID!
373 for (auto i = 0; i < par->size(); ++i) {
374 Inset const *ip = par->getInset(i);
377 if (const auto * bibitem = dynamic_cast<const InsetBibitem*>(ip)) {
378 auto id = xml::cleanID(bibitem->getParam("key"));
379 attr = from_utf8("xml:id='") + id + 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());
556 // Insert a new line after each "paragraph" (i.e. line in the listing), except for the last one.
557 // Otherwise, there would one more new line in the output than in the LyX document.
562 for (auto const & p : pars) {
563 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(), par->layout().docbookiteminnertagtype());
564 xs << XMLStream::ESCAPE_NONE << p;
565 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
569 makeAny(text, buf, xs, runparams, par);
572 // Close the environment.
573 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr); // TODO: switch in layout for par/block?
577 ParagraphList::const_iterator findEndOfEnvironment(
578 ParagraphList::const_iterator const & pstart,
579 ParagraphList::const_iterator const & pend)
581 // Copy-paste from XHTML. Should be factored out at some point...
582 ParagraphList::const_iterator p = pstart;
583 Layout const & bstyle = p->layout();
584 size_t const depth = p->params().depth();
585 for (++p; p != pend; ++p) {
586 Layout const & style = p->layout();
587 // It shouldn't happen that e.g. a section command occurs inside
588 // a quotation environment, at a higher depth, but as of 6/2009,
589 // it can happen. We pretend that it's just at lowest depth.
590 if (style.latextype == LATEX_COMMAND)
593 // If depth is down, we're done
594 if (p->params().depth() < depth)
597 // If depth is up, we're not done
598 if (p->params().depth() > depth)
601 // FIXME I am not sure about the first check.
602 // Surely we *could* have different layouts that count as
603 // LATEX_PARAGRAPH, right?
604 if (style.latextype == LATEX_PARAGRAPH || style != bstyle)
611 ParagraphList::const_iterator makeListEnvironment(Text const &text,
614 OutputParams const &runparams,
615 ParagraphList::const_iterator const & begin)
618 auto const end = text.paragraphs().end();
619 auto const envend = findEndOfEnvironment(par, end);
621 // Output the opening tag for this environment.
622 Layout const & envstyle = par->layout();
623 openTag(xs, envstyle.docbookwrappertag(), envstyle.docbookwrapperattr(), envstyle.docbookwrappertagtype());
624 openTag(xs, envstyle.docbooktag(), envstyle.docbookattr(), envstyle.docbooktagtype());
626 // Handle the content of the list environment, item by item.
627 while (par != envend) {
628 // Skip this paragraph if it is both empty and the last one (otherwise, there may be deeper paragraphs after).
631 if (par->empty() && nextpar == envend)
634 // Open the item wrapper.
635 Layout const & style = par->layout();
636 openTag(xs, style.docbookitemwrappertag(), style.docbookitemwrapperattr(), style.docbookitemwrappertagtype());
638 // Generate the label, if need be. If it is taken from the text, sep != 0 and corresponds to the first
639 // character after the label.
641 if (style.labeltype != LABEL_NO_LABEL && style.docbookitemlabeltag() != "NONE") {
642 if (style.labeltype == LABEL_MANUAL) {
643 // Only variablelist gets here (or similar items defined as an extension in the layout).
644 openTag(xs, style.docbookitemlabeltag(), style.docbookitemlabelattr(), style.docbookitemlabeltagtype());
645 sep = 1 + par->firstWordDocBook(xs, runparams);
646 closeTag(xs, style.docbookitemlabeltag(), style.docbookitemlabeltagtype());
648 // Usual cases: maybe there is something specified at the layout level. Highly unlikely, though.
649 docstring const lbl = par->params().labelString();
652 openTag(xs, style.docbookitemlabeltag(), style.docbookitemlabelattr(), style.docbookitemlabeltagtype());
654 closeTag(xs, style.docbookitemlabeltag(), style.docbookitemlabeltagtype());
659 // Open the item (after the wrapper and the label).
660 openTag(xs, style.docbookitemtag(), style.docbookitemattr(), style.docbookitemtagtype());
662 // Generate the content of the item.
663 if (sep < par->size()) {
664 auto pars = par->simpleDocBookOnePar(buf, runparams,
665 text.outerFont(std::distance(text.paragraphs().begin(), par)), sep);
666 for (auto &p : pars) {
667 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
668 par->layout().docbookiteminnertagtype());
669 xs << XMLStream::ESCAPE_NONE << p;
670 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
673 // DocBook doesn't like emptiness.
674 compTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
675 par->layout().docbookiteminnertagtype());
678 // If the next item is deeper, it must go entirely within this item (do it recursively).
679 // By construction, with findEndOfEnvironment, depth can only stay constant or increase, never decrease.
680 depth_type currentDepth = par->getDepth();
682 while (par != envend && par->getDepth() != currentDepth)
683 par = makeAny(text, buf, xs, runparams, par);
684 // Usually, this loop only makes one iteration, except in complex scenarios, like an item with a paragraph,
685 // a list, and another paragraph; or an item with two types of list (itemise then enumerate, for instance).
688 closeTag(xs, style.docbookitemtag(), style.docbookitemtagtype());
689 closeTag(xs, style.docbookitemwrappertag(), style.docbookitemwrappertagtype());
692 // Close this environment in exactly the same way as it was opened.
693 closeTag(xs, envstyle.docbooktag(), envstyle.docbooktagtype());
694 closeTag(xs, envstyle.docbookwrappertag(), envstyle.docbookwrappertagtype());
704 OutputParams const & runparams,
705 ParagraphList::const_iterator const & par)
707 // Unlike XHTML, no need for labels, as they are handled by DocBook tags.
708 auto const begin = text.paragraphs().begin();
709 auto const end = text.paragraphs().end();
713 // Generate this command.
714 auto prevpar = text.paragraphs().getParagraphBefore(par);
715 openParTag(xs, &*par, prevpar);
717 auto pars = par->simpleDocBookOnePar(buf, runparams,text.outerFont(distance(begin, par)));
718 for (auto & parXML : pars)
719 // TODO: decide what to do with openParTag/closeParTag in new lines.
720 xs << XMLStream::ESCAPE_NONE << parXML;
722 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
726 bool isLayoutSectioning(Layout const & lay)
728 if (lay.docbooksection()) // Special case: some DocBook styles must be handled as sections.
730 else if (lay.category() == from_utf8("Sectioning")) // Generic case.
731 return lay.toclevel != Layout::NOT_IN_TOC;
736 using DocBookDocumentSectioning = tuple<bool, pit_type>;
739 struct DocBookInfoTag
741 const set<pit_type> shouldBeInInfo;
742 const set<pit_type> mustBeInInfo; // With the notable exception of the abstract!
743 const set<pit_type> abstract;
744 const bool abstractLayout;
748 DocBookInfoTag(const set<pit_type> & shouldBeInInfo, const set<pit_type> & mustBeInInfo,
749 const set<pit_type> & abstract, bool abstractLayout, pit_type bpit, pit_type epit) :
750 shouldBeInInfo(shouldBeInInfo), mustBeInInfo(mustBeInInfo), abstract(abstract),
751 abstractLayout(abstractLayout), bpit(bpit), epit(epit) {}
755 DocBookDocumentSectioning hasDocumentSectioning(ParagraphList const ¶graphs, pit_type bpit, pit_type const epit) {
756 bool documentHasSections = false;
758 while (bpit < epit) {
759 Layout const &style = paragraphs[bpit].layout();
760 documentHasSections |= isLayoutSectioning(style);
762 if (documentHasSections)
766 // Paragraphs before the first section: [ runparams.par_begin ; eppit )
768 return make_tuple(documentHasSections, bpit);
772 bool hasOnlyNotes(Paragraph const & par)
774 // Precondition: the paragraph is not empty. Otherwise, the function will always return true...
775 for (int i = 0; i < par.size(); ++i)
776 // If you find something that is not an inset (like actual text) or an inset that is not a note,
778 if (!par.isInset(i) || par.getInset(i)->lyxCode() != NOTE_CODE)
781 // An empty paragraph may still require some output.
782 if (par.layout().docbooksection())
785 // There should be really no content here.
790 DocBookInfoTag getParagraphsWithInfo(ParagraphList const ¶graphs,
791 pit_type bpit, pit_type const epit,
792 // Typically, bpit is the beginning of the document and epit the end of the
793 // document *or* the first section.
794 bool documentHasSections,
795 bool detectUnlayoutedAbstract
796 // Whether paragraphs with no specific layout should be detected as abstracts.
797 // For inner sections, an abstract should only be detected if it has a specific
798 // layout. For others, anything that might look like an abstract should be sought.
800 set<pit_type> shouldBeInInfo;
801 set<pit_type> mustBeInInfo;
802 set<pit_type> abstractWithLayout;
803 set<pit_type> abstractNoLayout;
805 // Find the first non empty paragraph by mutating bpit.
806 while (bpit < epit) {
807 Paragraph const &par = paragraphs[bpit];
808 if (par.empty() || hasOnlyNotes(par))
814 // Traverse everything that might belong to <info>.
815 bool hasAbstractLayout = false;
816 pit_type cpit = bpit;
817 for (; cpit < epit; ++cpit) {
818 // Skip paragraphs that don't generate anything in DocBook.
819 Paragraph const & par = paragraphs[cpit];
820 Layout const &style = par.layout();
821 if (hasOnlyNotes(par))
824 // There should never be any section here, except for the first paragraph (a title can be part of <info>).
825 // (Just a sanity check: if this fails, this function could end up processing the whole document.)
826 if (cpit != bpit && isLayoutSectioning(par.layout())) {
827 LYXERR0("Assertion failed: section found in potential <info> paragraphs.");
831 // If this is marked as an abstract by the layout, put it in the right set.
832 if (style.docbookabstract()) {
833 hasAbstractLayout = true;
834 abstractWithLayout.emplace(cpit);
838 // Based on layout information, store this paragraph in one set: should be in <info>, must be,
839 // or abstract (either because of layout or of position).
840 if (style.docbookininfo() == "always")
841 mustBeInInfo.emplace(cpit);
842 else if (style.docbookininfo() == "maybe")
843 shouldBeInInfo.emplace(cpit);
844 else if (documentHasSections && !hasAbstractLayout && detectUnlayoutedAbstract &&
845 (style.docbooktag() == "NONE" || style.docbooktag() == "para") &&
846 style.docbookwrappertag() == "NONE")
847 // In this case, it is very likely that style.docbookininfo() == "never"! Be extra careful
848 // about anything that gets caught here.
849 abstractNoLayout.emplace(cpit);
850 else // This should definitely not be in <info>.
853 // Now, cpit points to the first paragraph that no more has things that could go in <info>.
854 // bpit is the beginning of the <info> part.
856 return DocBookInfoTag(shouldBeInInfo, mustBeInInfo,
857 hasAbstractLayout ? abstractWithLayout : abstractNoLayout,
858 hasAbstractLayout, bpit, cpit);
861 } // end anonymous namespace
864 ParagraphList::const_iterator makeAny(Text const &text,
867 OutputParams const &runparams,
868 ParagraphList::const_iterator par)
870 switch (par->layout().latextype) {
872 makeCommand(text, buf, xs, runparams, par);
874 case LATEX_ENVIRONMENT:
875 makeEnvironment(text, buf, xs, runparams, par);
877 case LATEX_LIST_ENVIRONMENT:
878 case LATEX_ITEM_ENVIRONMENT:
879 // Only case when makeAny() might consume more than one paragraph.
880 return makeListEnvironment(text, buf, xs, runparams, par);
881 case LATEX_PARAGRAPH:
882 makeParagraph(text, buf, xs, runparams, par);
884 case LATEX_BIB_ENVIRONMENT:
885 makeBibliography(text, buf, xs, runparams, par);
893 xml::FontTag docbookStartFontTag(xml::FontTypes type)
895 return xml::FontTag(from_utf8(fontToDocBookTag(type)), from_utf8(fontToAttribute(type)), type);
899 xml::EndFontTag docbookEndFontTag(xml::FontTypes type)
901 return xml::EndFontTag(from_utf8(fontToDocBookTag(type)), type);
905 void outputDocBookInfo(
909 OutputParams const & runparams,
910 ParagraphList const & paragraphs,
911 DocBookInfoTag const & info)
913 // Perform an additional check on the abstract. Sometimes, there are many paragraphs that should go
914 // into the abstract, but none generates actual content. Thus, first generate to a temporary stream,
915 // then only create the <abstract> tag if these paragraphs generate some content.
916 // This check must be performed *before* a decision on whether or not to output <info> is made.
917 bool hasAbstract = !info.abstract.empty();
920 // Generate the abstract XML into a string before further checks.
921 // Usually, makeAny only generates one paragraph at a time. However, for the specific case of lists, it might
922 // generate more than one paragraph, as indicated in the return value.
923 odocstringstream os2;
926 set<pit_type> doneParas;
927 for (auto const & p : info.abstract) {
928 if (doneParas.find(p) == doneParas.end()) {
929 auto oldPar = paragraphs.iterator_at(p);
930 auto newPar = makeAny(text, buf, xs2, runparams, oldPar);
932 // Insert the indices of all the paragraphs that were just generated (typically, one).
933 // **Make the hypothesis that, when an abstract has a list, all its items are consecutive.**
935 while (oldPar != newPar) {
936 doneParas.emplace(id);
943 // Actually output the abstract if there is something to do. Don't count line feeds or spaces in this,
944 // even though they must be properly output if there is some abstract.
945 abstract = os2.str();
946 docstring cleaned = abstract;
947 cleaned.erase(std::remove_if(cleaned.begin(), cleaned.end(), lyx::isSpace), cleaned.end());
949 // Nothing? Then there is no abstract!
954 // The abstract must go in <info>. Otherwise, decide whether to open <info> based on the layouts.
955 bool needInfo = !info.mustBeInInfo.empty() || hasAbstract;
957 // Start the <info> tag if required.
959 xs.startDivision(false);
960 xs << xml::StartTag("info");
964 // Output the elements that should go in <info>, before and after the abstract.
965 for (auto pit : info.shouldBeInInfo) // Typically, the title: these elements are so important and ubiquitous
966 // that mandating a wrapper like <info> would repel users. Thus, generate them first.
967 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
968 for (auto pit : info.mustBeInInfo)
969 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
971 // If there is no title, generate one (required for the document to be valid).
972 // This code is called for the main document, for table cells, etc., so be precise in this condition.
973 if (text.isMainText() && info.shouldBeInInfo.empty() && !runparams.inInclude) {
974 xs << xml::StartTag("title");
975 xs << "Untitled Document";
976 xs << xml::EndTag("title");
980 // Always output the abstract as the last item of the <info>, as it requires special treatment (especially if
981 // it contains several paragraphs that are empty).
983 if (info.abstractLayout) {
984 xs << XMLStream::ESCAPE_NONE << abstract;
987 string tag = paragraphs[*info.abstract.begin()].layout().docbookforceabstracttag();
991 if (!xs.isLastTagCR())
994 xs << xml::StartTag(tag);
996 xs << XMLStream::ESCAPE_NONE << abstract;
997 xs << xml::EndTag(tag);
1002 // End the <info> tag if it was started.
1004 if (!xs.isLastTagCR())
1007 xs << xml::EndTag("info");
1014 void docbookSimpleAllParagraphs(
1018 OutputParams const & runparams)
1020 // Handle the given text, supposing it has no sections (i.e. a "simple" text). The input may vary in length
1021 // between a single paragraph to a whole document.
1022 pit_type const bpit = runparams.par_begin;
1023 pit_type const epit = runparams.par_end;
1024 ParagraphList const ¶graphs = text.paragraphs();
1026 // First, the <info> tag.
1027 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, epit, false, true);
1028 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
1030 // Then, the content. It starts where the <info> ends.
1031 auto par = paragraphs.iterator_at(info.epit);
1032 auto end = paragraphs.iterator_at(epit);
1033 while (par != end) {
1034 if (!hasOnlyNotes(*par))
1035 par = makeAny(text, buf, xs, runparams, par);
1042 void docbookParagraphs(Text const &text,
1045 OutputParams const &runparams) {
1046 ParagraphList const ¶graphs = text.paragraphs();
1047 if (runparams.par_begin == runparams.par_end) {
1048 runparams.par_begin = 0;
1049 runparams.par_end = paragraphs.size();
1051 pit_type bpit = runparams.par_begin;
1052 pit_type const epit = runparams.par_end;
1053 LASSERT(bpit < epit,
1055 xs << XMLStream::ESCAPE_NONE << "<!-- DocBook output error! -->\n";
1059 std::stack<std::pair<int, string>> headerLevels; // Used to determine when to open/close sections: store the depth
1060 // of the section and the tag that was used to open it.
1062 // Detect whether the document contains sections. If there are no sections, treatment is largely simplified.
1063 // In particular, there can't be an abstract, unless it is manually marked.
1064 bool documentHasSections;
1066 tie(documentHasSections, eppit) = hasDocumentSectioning(paragraphs, bpit, epit);
1068 // Deal with "simple" documents, i.e. those without sections.
1069 if (!documentHasSections) {
1070 docbookSimpleAllParagraphs(text, buf, xs, runparams);
1074 // Output the first <info> tag (or just the title).
1075 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, eppit, true, true);
1076 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
1079 // Then, iterate through the paragraphs of this document.
1080 bool currentlyInAppendix = false;
1082 auto par = text.paragraphs().iterator_at(bpit);
1083 auto end = text.paragraphs().iterator_at(epit);
1084 while (par != end) {
1085 OutputParams ourparams = runparams;
1087 if (par->params().startOfAppendix())
1088 currentlyInAppendix = true;
1089 if (hasOnlyNotes(*par)) {
1094 Layout const &style = par->layout();
1096 // Think about adding <section> and/or </section>s.
1097 if (isLayoutSectioning(style)) {
1098 int level = style.toclevel;
1100 // Need to close a previous section if it has the same level or a higher one (close <section> if opening a
1101 // <h2> after a <h2>, <h3>, <h4>, <h5> or <h6>). More examples:
1102 // - current: h2; back: h1; do not close any <section>
1103 // - current: h1; back: h2; close two <section> (first the <h2>, then the <h1>, so a new <h1> can come)
1104 while (!headerLevels.empty() && level <= headerLevels.top().first) {
1105 // Output the tag only if it corresponds to a legit section.
1106 int stackLevel = headerLevels.top().first;
1107 if (stackLevel != Layout::NOT_IN_TOC) {
1108 xs << xml::EndTag(headerLevels.top().second);
1114 // Open the new section: first push it onto the stack, then output it in DocBook.
1115 string sectionTag = (currentlyInAppendix && style.docbooksectiontag() == "chapter") ?
1116 "appendix" : style.docbooksectiontag();
1117 headerLevels.push(std::make_pair(level, sectionTag));
1119 // Some sectioning-like elements should not be output (such as FrontMatter).
1120 if (level != Layout::NOT_IN_TOC) {
1121 // Look for a label in the title, i.e. a InsetLabel as a child.
1122 docstring id = docstring();
1123 for (pos_type i = 0; i < par->size(); ++i) {
1124 Inset const *inset = par->getInset(i);
1126 if (auto label = dynamic_cast<InsetLabel const *>(inset)) {
1127 // Generate the attributes for the section if need be.
1128 id += "xml:id=\"" + xml::cleanID(label->screenLabel()) + "\"";
1130 // Don't output the ID as a DocBook <anchor>.
1131 ourparams.docbook_anchors_to_ignore.emplace(label->screenLabel());
1133 // Cannot have multiple IDs per tag. If there is another ID inset in the document, it will
1134 // be output as a DocBook anchor.
1140 // Write the open tag for this section.
1144 xs << xml::StartTag(sectionTag, attrs);
1149 // Close all sections before the bibliography.
1150 // 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)?
1151 if (!par->insetList().empty()) {
1152 Inset const *firstInset = par->getInset(0);
1153 if (firstInset && (firstInset->lyxCode() == BIBITEM_CODE || firstInset->lyxCode() == BIBTEX_CODE)) {
1154 while (!headerLevels.empty()) {
1155 int level = headerLevels.top().first;
1156 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1159 // Output the tag only if it corresponds to a legit section.
1160 if (level != Layout::NOT_IN_TOC) {
1161 xs << XMLStream::ESCAPE_NONE << tag;
1168 // Generate the <info> tag if a section was just opened.
1169 // Some sections may require abstracts (mostly parts, in books: DocBookForceAbstractTag will not be NONE),
1170 // others can still have an abstract (it must be detected so that it can be output at the right place).
1171 // 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.
1172 if (isLayoutSectioning(style)) {
1173 // This abstract may be found between the next paragraph and the next title.
1174 pit_type cpit = std::distance(text.paragraphs().begin(), par);
1175 pit_type ppit = std::get<1>(hasDocumentSectioning(paragraphs, cpit + 1L, epit));
1177 // Generate this abstract (this code corresponds to parts of outputDocBookInfo).
1178 DocBookInfoTag secInfo = getParagraphsWithInfo(paragraphs, cpit, ppit, true,
1179 style.docbookforceabstracttag() != "NONE");
1181 if (!secInfo.mustBeInInfo.empty() || !secInfo.shouldBeInInfo.empty() || !secInfo.abstract.empty()) {
1182 // Generate the <info>, if required. If DocBookForceAbstractTag != NONE, this abstract will not be in
1183 // <info>, unlike other ("standard") abstracts.
1184 bool hasStandardAbstract = !secInfo.abstract.empty() && style.docbookforceabstracttag() == "NONE";
1185 bool needInfo = !secInfo.mustBeInInfo.empty() || hasStandardAbstract;
1188 xs.startDivision(false);
1189 xs << xml::StartTag("info");
1193 // Output the elements that should go in <info>, before and after the abstract.
1194 for (auto pit : secInfo.shouldBeInInfo) // Typically, the title: these elements are so important and ubiquitous
1195 // that mandating a wrapper like <info> would repel users. Thus, generate them first.
1196 makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(pit));
1197 for (auto pit : secInfo.mustBeInInfo)
1198 makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(pit));
1200 // Deal with the abstract in <info> if it is standard (i.e. its tag is <abstract>).
1201 if (!secInfo.abstract.empty() && hasStandardAbstract) {
1202 if (!secInfo.abstractLayout) {
1203 xs << xml::StartTag("abstract");
1207 for (auto const &p : secInfo.abstract)
1208 makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(p));
1210 if (!secInfo.abstractLayout) {
1211 xs << xml::EndTag("abstract");
1216 // End the <info> tag if it was started.
1218 if (!xs.isLastTagCR())
1221 xs << xml::EndTag("info");
1226 // Deal with the abstract outside <info> if it is not standard (i.e. its tag is layout-defined).
1227 if (!secInfo.abstract.empty() && !hasStandardAbstract) {
1228 // Assert: style.docbookforceabstracttag() != NONE.
1229 xs << xml::StartTag(style.docbookforceabstracttag());
1231 for (auto const &p : secInfo.abstract)
1232 makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(p));
1233 xs << xml::EndTag(style.docbookforceabstracttag());
1237 // Skip all the text that has just been generated.
1238 par = paragraphs.iterator_at(secInfo.epit);
1240 // No <info> tag to generate, proceed as for normal paragraphs.
1241 par = makeAny(text, buf, xs, ourparams, par);
1244 // Generate this paragraph, as it has nothing special.
1245 par = makeAny(text, buf, xs, ourparams, par);
1249 // If need be, close <section>s, but only at the end of the document (otherwise, dealt with at the beginning
1251 while (!headerLevels.empty() && headerLevels.top().first > Layout::NOT_IN_TOC) {
1252 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1254 xs << XMLStream::ESCAPE_NONE << tag;