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 // Tables doe not deserve their own paragraphs (DocBook allows them outside 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() == TABULAR_CODE;
460 // Floats cannot be in paragraphs.
461 special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
462 return inset.inset->lyxCode() == FLOAT_CODE;
464 // Bibliographies cannot be in paragraphs. Bibitems should still be handled as paragraphs, though
465 // (see makeParagraphBibliography).
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() == BIBTEX_CODE;
469 // ERTs are in comments, not paragraphs.
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() == ERT_CODE;
473 // Listings should not 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() == LISTINGS_CODE;
477 // Boxes cannot get into their own 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() == BOX_CODE;
481 // Includes should not have a paragraph.
482 special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
483 return inset.inset->lyxCode() == INCLUDE_CODE;
486 bool const open_par = runparams.docbook_make_pars
487 && !runparams.docbook_in_par
490 // We want to issue the closing tag if either:
491 // (i) We opened it, and either docbook_in_par is false,
492 // or we're not in the last paragraph, anyway.
493 // (ii) We didn't open it and docbook_in_par is true,
494 // but we are in the first par, and there is a next par.
495 bool const close_par = open_par && (!runparams.docbook_in_par);
497 // Determine if this paragraph has some real content. Things like new pages are not caught
498 // by Paragraph::empty(), even though they do not generate anything useful in DocBook.
499 // Thus, remove all spaces (including new lines: \r, \n) before checking for emptiness.
500 // std::all_of allows doing this check without having to copy the string.
501 // Open and close tags around each contained paragraph.
504 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(distance(begin, par)), 0, nextpar == end, special_case);
505 for (docstring const & parXML : pars) {
506 if (xml::isNotOnlySpace(parXML)) {
508 openParTag(xs, &*par, prevpar);
510 xs << XMLStream::ESCAPE_NONE << parXML;
513 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
519 void makeEnvironment(Text const &text,
522 OutputParams const &runparams,
523 ParagraphList::const_iterator const & par)
525 auto const end = text.paragraphs().end();
529 // Special cases for listing-like environments provided in layouts. This is quite ad-hoc, but provides a useful
530 // default. This should not be used by too many environments (only LyX-Code right now).
531 // This would be much simpler if LyX-Code was implemented as InsetListings...
532 bool mimicListing = false;
533 bool ignoreFonts = false;
534 if (par->layout().docbooktag() == "programlisting") {
539 // Output the opening tag for this environment, but only if it has not been previously opened (condition
540 // implemented in openParTag).
541 auto prevpar = text.paragraphs().getParagraphBefore(par);
542 openParTag(xs, &*par, prevpar); // TODO: switch in layout for par/block?
544 // Generate the contents of this environment. There is a special case if this is like some environment.
545 Layout const & style = par->layout();
546 if (style.latextype == LATEX_COMMAND) {
547 // Nothing to do (otherwise, infinite loops).
548 } else if (style.latextype == LATEX_ENVIRONMENT) {
549 // Generate the paragraph, if need be.
550 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(std::distance(text.paragraphs().begin(), par)), 0, false, ignoreFonts);
553 auto p = pars.begin();
554 while (p != pars.end()) {
555 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(), par->layout().docbookiteminnertagtype());
556 xs << XMLStream::ESCAPE_NONE << *p;
557 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
560 // Insert a new line after each "paragraph" (i.e. line in the listing), except for the last one.
561 // Otherwise, there would one more new line in the output than in the LyX document.
566 for (auto const & p : pars) {
567 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(), par->layout().docbookiteminnertagtype());
568 xs << XMLStream::ESCAPE_NONE << p;
569 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
573 makeAny(text, buf, xs, runparams, par);
576 // Close the environment.
577 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr); // TODO: switch in layout for par/block?
581 ParagraphList::const_iterator findEndOfEnvironment(
582 ParagraphList::const_iterator const & pstart,
583 ParagraphList::const_iterator const & pend)
585 // Copy-paste from XHTML. Should be factored out at some point...
586 ParagraphList::const_iterator p = pstart;
587 Layout const & bstyle = p->layout();
588 size_t const depth = p->params().depth();
589 for (++p; p != pend; ++p) {
590 Layout const & style = p->layout();
591 // It shouldn't happen that e.g. a section command occurs inside
592 // a quotation environment, at a higher depth, but as of 6/2009,
593 // it can happen. We pretend that it's just at lowest depth.
594 if (style.latextype == LATEX_COMMAND)
597 // If depth is down, we're done
598 if (p->params().depth() < depth)
601 // If depth is up, we're not done
602 if (p->params().depth() > depth)
605 // FIXME I am not sure about the first check.
606 // Surely we *could* have different layouts that count as
607 // LATEX_PARAGRAPH, right?
608 if (style.latextype == LATEX_PARAGRAPH || style != bstyle)
615 ParagraphList::const_iterator makeListEnvironment(Text const &text,
618 OutputParams const &runparams,
619 ParagraphList::const_iterator const & begin)
622 auto const end = text.paragraphs().end();
623 auto const envend = findEndOfEnvironment(par, end);
625 // Output the opening tag for this environment.
626 Layout const & envstyle = par->layout();
627 openTag(xs, envstyle.docbookwrappertag(), envstyle.docbookwrapperattr(), envstyle.docbookwrappertagtype());
628 openTag(xs, envstyle.docbooktag(), envstyle.docbookattr(), envstyle.docbooktagtype());
630 // Handle the content of the list environment, item by item.
631 while (par != envend) {
632 // Skip this paragraph if it is both empty and the last one (otherwise, there may be deeper paragraphs after).
635 if (par->empty() && nextpar == envend)
638 // Open the item wrapper.
639 Layout const & style = par->layout();
640 openTag(xs, style.docbookitemwrappertag(), style.docbookitemwrapperattr(), style.docbookitemwrappertagtype());
642 // Generate the label, if need be. If it is taken from the text, sep != 0 and corresponds to the first
643 // character after the label.
645 if (style.labeltype != LABEL_NO_LABEL && style.docbookitemlabeltag() != "NONE") {
646 if (style.labeltype == LABEL_MANUAL) {
647 // Only variablelist gets here (or similar items defined as an extension in the layout).
648 openTag(xs, style.docbookitemlabeltag(), style.docbookitemlabelattr(), style.docbookitemlabeltagtype());
649 sep = 1 + par->firstWordDocBook(xs, runparams);
650 closeTag(xs, style.docbookitemlabeltag(), style.docbookitemlabeltagtype());
652 // Usual cases: maybe there is something specified at the layout level. Highly unlikely, though.
653 docstring const lbl = par->params().labelString();
656 openTag(xs, style.docbookitemlabeltag(), style.docbookitemlabelattr(), style.docbookitemlabeltagtype());
658 closeTag(xs, style.docbookitemlabeltag(), style.docbookitemlabeltagtype());
663 // Open the item (after the wrapper and the label).
664 openTag(xs, style.docbookitemtag(), style.docbookitemattr(), style.docbookitemtagtype());
666 // Generate the content of the item.
667 if (sep < par->size()) {
668 auto pars = par->simpleDocBookOnePar(buf, runparams,
669 text.outerFont(std::distance(text.paragraphs().begin(), par)), sep);
670 for (auto &p : pars) {
671 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
672 par->layout().docbookiteminnertagtype());
673 xs << XMLStream::ESCAPE_NONE << p;
674 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
677 // DocBook doesn't like emptiness.
678 compTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
679 par->layout().docbookiteminnertagtype());
682 // If the next item is deeper, it must go entirely within this item (do it recursively).
683 // By construction, with findEndOfEnvironment, depth can only stay constant or increase, never decrease.
684 depth_type currentDepth = par->getDepth();
686 while (par != envend && par->getDepth() != currentDepth)
687 par = makeAny(text, buf, xs, runparams, par);
688 // Usually, this loop only makes one iteration, except in complex scenarios, like an item with a paragraph,
689 // a list, and another paragraph; or an item with two types of list (itemise then enumerate, for instance).
692 closeTag(xs, style.docbookitemtag(), style.docbookitemtagtype());
693 closeTag(xs, style.docbookitemwrappertag(), style.docbookitemwrappertagtype());
696 // Close this environment in exactly the same way as it was opened.
697 closeTag(xs, envstyle.docbooktag(), envstyle.docbooktagtype());
698 closeTag(xs, envstyle.docbookwrappertag(), envstyle.docbookwrappertagtype());
708 OutputParams const & runparams,
709 ParagraphList::const_iterator const & par)
711 // Unlike XHTML, no need for labels, as they are handled by DocBook tags.
712 auto const begin = text.paragraphs().begin();
713 auto const end = text.paragraphs().end();
717 // Generate this command.
718 auto prevpar = text.paragraphs().getParagraphBefore(par);
719 openParTag(xs, &*par, prevpar);
721 auto pars = par->simpleDocBookOnePar(buf, runparams,text.outerFont(distance(begin, par)));
722 for (auto & parXML : pars)
723 // TODO: decide what to do with openParTag/closeParTag in new lines.
724 xs << XMLStream::ESCAPE_NONE << parXML;
726 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
730 bool isLayoutSectioning(Layout const & lay)
732 if (lay.docbooksection()) // Special case: some DocBook styles must be handled as sections.
734 else if (lay.category() == from_utf8("Sectioning") || lay.docbooktag() == "section") // Generic case.
735 return lay.toclevel != Layout::NOT_IN_TOC;
740 bool isLayoutSectioningOrSimilar(Layout const & lay)
742 return isLayoutSectioning(lay) || lay.docbooktag() == "bridgehead";
746 using DocBookDocumentSectioning = tuple<bool, pit_type>;
749 struct DocBookInfoTag
751 const set<pit_type> shouldBeInInfo;
752 const set<pit_type> mustBeInInfo; // With the notable exception of the abstract!
753 const set<pit_type> abstract;
754 const bool abstractLayout;
758 DocBookInfoTag(const set<pit_type> & shouldBeInInfo, const set<pit_type> & mustBeInInfo,
759 const set<pit_type> & abstract, bool abstractLayout, pit_type bpit, pit_type epit) :
760 shouldBeInInfo(shouldBeInInfo), mustBeInInfo(mustBeInInfo), abstract(abstract),
761 abstractLayout(abstractLayout), bpit(bpit), epit(epit) {}
765 DocBookDocumentSectioning hasDocumentSectioning(ParagraphList const ¶graphs, pit_type bpit, pit_type const epit) {
766 bool documentHasSections = false;
768 while (bpit < epit) {
769 Layout const &style = paragraphs[bpit].layout();
770 documentHasSections |= isLayoutSectioningOrSimilar(style);
772 if (documentHasSections)
776 // Paragraphs before the first section: [ runparams.par_begin ; eppit )
778 return make_tuple(documentHasSections, bpit);
782 bool hasOnlyNotes(Paragraph const & par)
784 // Precondition: the paragraph is not empty. Otherwise, the function will always return true...
785 for (int i = 0; i < par.size(); ++i)
786 // If you find something that is not an inset (like actual text) or an inset that is not a note,
788 if (!par.isInset(i) || par.getInset(i)->lyxCode() != NOTE_CODE)
791 // An empty paragraph may still require some output.
792 if (par.layout().docbooksection())
795 // There should be really no content here.
800 DocBookInfoTag getParagraphsWithInfo(ParagraphList const ¶graphs,
801 pit_type bpit, pit_type const epit,
802 // Typically, bpit is the beginning of the document and epit the end of the
803 // document *or* the first section.
804 bool documentHasSections,
805 bool detectUnlayoutedAbstract
806 // Whether paragraphs with no specific layout should be detected as abstracts.
807 // For inner sections, an abstract should only be detected if it has a specific
808 // layout. For others, anything that might look like an abstract should be sought.
810 set<pit_type> shouldBeInInfo;
811 set<pit_type> mustBeInInfo;
812 set<pit_type> abstractWithLayout;
813 set<pit_type> abstractNoLayout;
815 // Find the first non empty paragraph by mutating bpit.
816 while (bpit < epit) {
817 Paragraph const &par = paragraphs[bpit];
818 if (par.empty() || hasOnlyNotes(par))
824 // Traverse everything that might belong to <info>.
825 bool hasAbstractLayout = false;
826 pit_type cpit = bpit;
827 for (; cpit < epit; ++cpit) {
828 // Skip paragraphs that don't generate anything in DocBook.
829 Paragraph const & par = paragraphs[cpit];
830 Layout const &style = par.layout();
831 if (hasOnlyNotes(par))
834 // There should never be any section here, except for the first paragraph (a title can be part of <info>).
835 // (Just a sanity check: if this fails, this function could end up processing the whole document.)
836 if (cpit != bpit && isLayoutSectioningOrSimilar(par.layout())) {
837 LYXERR0("Assertion failed: section found in potential <info> paragraphs.");
841 // If this is marked as an abstract by the layout, put it in the right set.
842 if (style.docbookabstract()) {
843 hasAbstractLayout = true;
844 abstractWithLayout.emplace(cpit);
848 // Based on layout information, store this paragraph in one set: should be in <info>, must be,
849 // or abstract (either because of layout or of position).
850 if (style.docbookininfo() == "always")
851 mustBeInInfo.emplace(cpit);
852 else if (style.docbookininfo() == "maybe")
853 shouldBeInInfo.emplace(cpit);
854 else if (documentHasSections && !hasAbstractLayout && detectUnlayoutedAbstract &&
855 (style.docbooktag() == "NONE" || style.docbooktag() == "para") &&
856 style.docbookwrappertag() == "NONE")
857 // In this case, it is very likely that style.docbookininfo() == "never"! Be extra careful
858 // about anything that gets caught here.
859 abstractNoLayout.emplace(cpit);
860 else // This should definitely not be in <info>.
863 // Now, cpit points to the first paragraph that no more has things that could go in <info>.
864 // bpit is the beginning of the <info> part.
866 return DocBookInfoTag(shouldBeInInfo, mustBeInInfo,
867 hasAbstractLayout ? abstractWithLayout : abstractNoLayout,
868 hasAbstractLayout, bpit, cpit);
871 } // end anonymous namespace
874 ParagraphList::const_iterator makeAny(Text const &text,
877 OutputParams const &runparams,
878 ParagraphList::const_iterator par)
880 switch (par->layout().latextype) {
882 makeCommand(text, buf, xs, runparams, par);
884 case LATEX_ENVIRONMENT:
885 makeEnvironment(text, buf, xs, runparams, par);
887 case LATEX_LIST_ENVIRONMENT:
888 case LATEX_ITEM_ENVIRONMENT:
889 // Only case when makeAny() might consume more than one paragraph.
890 return makeListEnvironment(text, buf, xs, runparams, par);
891 case LATEX_PARAGRAPH:
892 makeParagraph(text, buf, xs, runparams, par);
894 case LATEX_BIB_ENVIRONMENT:
895 makeBibliography(text, buf, xs, runparams, par);
903 xml::FontTag docbookStartFontTag(xml::FontTypes type)
905 return xml::FontTag(from_utf8(fontToDocBookTag(type)), from_utf8(fontToAttribute(type)), type);
909 xml::EndFontTag docbookEndFontTag(xml::FontTypes type)
911 return xml::EndFontTag(from_utf8(fontToDocBookTag(type)), type);
915 void outputDocBookInfo(
919 OutputParams const & runparams,
920 ParagraphList const & paragraphs,
921 DocBookInfoTag const & info)
923 // Perform an additional check on the abstract. Sometimes, there are many paragraphs that should go
924 // into the abstract, but none generates actual content. Thus, first generate to a temporary stream,
925 // then only create the <abstract> tag if these paragraphs generate some content.
926 // This check must be performed *before* a decision on whether or not to output <info> is made.
927 bool hasAbstract = !info.abstract.empty();
930 // Generate the abstract XML into a string before further checks.
931 // Usually, makeAny only generates one paragraph at a time. However, for the specific case of lists, it might
932 // generate more than one paragraph, as indicated in the return value.
933 odocstringstream os2;
936 set<pit_type> doneParas;
937 for (auto const & p : info.abstract) {
938 if (doneParas.find(p) == doneParas.end()) {
939 auto oldPar = paragraphs.iterator_at(p);
940 auto newPar = makeAny(text, buf, xs2, runparams, oldPar);
942 // Insert the indices of all the paragraphs that were just generated (typically, one).
943 // **Make the hypothesis that, when an abstract has a list, all its items are consecutive.**
945 while (oldPar != newPar) {
946 doneParas.emplace(id);
953 // Actually output the abstract if there is something to do. Don't count line feeds or spaces in this,
954 // even though they must be properly output if there is some abstract.
955 abstract = os2.str();
956 docstring cleaned = abstract;
957 cleaned.erase(std::remove_if(cleaned.begin(), cleaned.end(), lyx::isSpace), cleaned.end());
959 // Nothing? Then there is no abstract!
964 // The abstract must go in <info>. Otherwise, decide whether to open <info> based on the layouts.
965 bool needInfo = !info.mustBeInInfo.empty() || hasAbstract;
967 // Start the <info> tag if required.
969 xs.startDivision(false);
970 xs << xml::StartTag("info");
974 // Output the elements that should go in <info>, before and after the abstract.
975 for (auto pit : info.shouldBeInInfo) // Typically, the title: these elements are so important and ubiquitous
976 // that mandating a wrapper like <info> would repel users. Thus, generate them first.
977 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
978 for (auto pit : info.mustBeInInfo)
979 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
981 // If there is no title, generate one (required for the document to be valid).
982 // This code is called for the main document, for table cells, etc., so be precise in this condition.
983 if (text.isMainText() && info.shouldBeInInfo.empty() && !runparams.inInclude) {
984 xs << xml::StartTag("title");
985 xs << "Untitled Document";
986 xs << xml::EndTag("title");
990 // Always output the abstract as the last item of the <info>, as it requires special treatment (especially if
991 // it contains several paragraphs that are empty).
993 if (info.abstractLayout) {
994 xs << XMLStream::ESCAPE_NONE << abstract;
997 string tag = paragraphs[*info.abstract.begin()].layout().docbookforceabstracttag();
1001 if (!xs.isLastTagCR())
1004 xs << xml::StartTag(tag);
1006 xs << XMLStream::ESCAPE_NONE << abstract;
1007 xs << xml::EndTag(tag);
1012 // End the <info> tag if it was started.
1014 if (!xs.isLastTagCR())
1017 xs << xml::EndTag("info");
1024 void docbookSimpleAllParagraphs(
1028 OutputParams const & runparams)
1030 // Handle the given text, supposing it has no sections (i.e. a "simple" text). The input may vary in length
1031 // between a single paragraph to a whole document.
1032 pit_type const bpit = runparams.par_begin;
1033 pit_type const epit = runparams.par_end;
1034 ParagraphList const ¶graphs = text.paragraphs();
1036 // First, the <info> tag.
1037 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, epit, false, true);
1038 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
1040 // Then, the content. It starts where the <info> ends.
1041 auto par = paragraphs.iterator_at(info.epit);
1042 auto end = paragraphs.iterator_at(epit);
1043 while (par != end) {
1044 if (!hasOnlyNotes(*par))
1045 par = makeAny(text, buf, xs, runparams, par);
1052 void docbookParagraphs(Text const &text,
1055 OutputParams const &runparams) {
1056 ParagraphList const ¶graphs = text.paragraphs();
1057 if (runparams.par_begin == runparams.par_end) {
1058 runparams.par_begin = 0;
1059 runparams.par_end = paragraphs.size();
1061 pit_type bpit = runparams.par_begin;
1062 pit_type const epit = runparams.par_end;
1063 LASSERT(bpit < epit,
1065 xs << XMLStream::ESCAPE_NONE << "<!-- DocBook output error! -->\n";
1069 std::stack<std::pair<int, string>> headerLevels; // Used to determine when to open/close sections: store the depth
1070 // of the section and the tag that was used to open it.
1072 // Detect whether the document contains sections. If there are no sections, treatment is largely simplified.
1073 // In particular, there can't be an abstract, unless it is manually marked.
1074 bool documentHasSections;
1076 tie(documentHasSections, eppit) = hasDocumentSectioning(paragraphs, bpit, epit);
1078 // Deal with "simple" documents, i.e. those without sections.
1079 if (!documentHasSections) {
1080 docbookSimpleAllParagraphs(text, buf, xs, runparams);
1084 // Output the first <info> tag (or just the title).
1085 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, eppit, true, true);
1086 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
1089 // Then, iterate through the paragraphs of this document.
1090 bool currentlyInAppendix = false;
1092 auto par = text.paragraphs().iterator_at(bpit);
1093 auto end = text.paragraphs().iterator_at(epit);
1094 while (par != end) {
1095 OutputParams ourparams = runparams;
1097 if (par->params().startOfAppendix())
1098 currentlyInAppendix = true;
1099 if (hasOnlyNotes(*par)) {
1104 Layout const &style = par->layout();
1106 // Think about adding <section> and/or </section>s.
1107 if (isLayoutSectioning(style)) {
1108 int level = style.toclevel;
1110 // Need to close a previous section if it has the same level or a higher one (close <section> if opening a
1111 // <h2> after a <h2>, <h3>, <h4>, <h5> or <h6>). More examples:
1112 // - current: h2; back: h1; do not close any <section>
1113 // - current: h1; back: h2; close two <section> (first the <h2>, then the <h1>, so a new <h1> can come)
1114 while (!headerLevels.empty() && level <= headerLevels.top().first) {
1115 // Output the tag only if it corresponds to a legit section.
1116 int stackLevel = headerLevels.top().first;
1117 if (stackLevel != Layout::NOT_IN_TOC) {
1118 xs << xml::EndTag(headerLevels.top().second);
1124 // Open the new section: first push it onto the stack, then output it in DocBook.
1125 string sectionTag = (currentlyInAppendix && style.docbooksectiontag() == "chapter") ?
1126 "appendix" : style.docbooksectiontag();
1127 headerLevels.push(std::make_pair(level, sectionTag));
1129 // Some sectioning-like elements should not be output (such as FrontMatter).
1130 if (level != Layout::NOT_IN_TOC) {
1131 // Look for a label in the title, i.e. a InsetLabel as a child.
1132 docstring id = docstring();
1133 for (pos_type i = 0; i < par->size(); ++i) {
1134 Inset const *inset = par->getInset(i);
1136 if (auto label = dynamic_cast<InsetLabel const *>(inset)) {
1137 // Generate the attributes for the section if need be.
1138 id += "xml:id=\"" + xml::cleanID(label->screenLabel()) + "\"";
1140 // Don't output the ID as a DocBook <anchor>.
1141 ourparams.docbook_anchors_to_ignore.emplace(label->screenLabel());
1143 // Cannot have multiple IDs per tag. If there is another ID inset in the document, it will
1144 // be output as a DocBook anchor.
1150 // Write the open tag for this section.
1154 xs << xml::StartTag(sectionTag, attrs);
1159 // Close all sections before the bibliography.
1160 // 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)?
1161 if (!par->insetList().empty()) {
1162 Inset const *firstInset = par->getInset(0);
1163 if (firstInset && (firstInset->lyxCode() == BIBITEM_CODE || firstInset->lyxCode() == BIBTEX_CODE)) {
1164 while (!headerLevels.empty()) {
1165 int level = headerLevels.top().first;
1166 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1169 // Output the tag only if it corresponds to a legit section.
1170 if (level != Layout::NOT_IN_TOC) {
1171 xs << XMLStream::ESCAPE_NONE << tag;
1178 // Generate the <info> tag if a section was just opened.
1179 // Some sections may require abstracts (mostly parts, in books: DocBookForceAbstractTag will not be NONE),
1180 // others can still have an abstract (it must be detected so that it can be output at the right place).
1181 // 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.
1182 if (isLayoutSectioning(style)) {
1183 // This abstract may be found between the next paragraph and the next title.
1184 pit_type cpit = std::distance(text.paragraphs().begin(), par);
1185 pit_type ppit = std::get<1>(hasDocumentSectioning(paragraphs, cpit + 1L, epit));
1187 // Generate this abstract (this code corresponds to parts of outputDocBookInfo).
1188 DocBookInfoTag secInfo = getParagraphsWithInfo(paragraphs, cpit, ppit, true,
1189 style.docbookforceabstracttag() != "NONE");
1191 if (!secInfo.mustBeInInfo.empty() || !secInfo.shouldBeInInfo.empty() || !secInfo.abstract.empty()) {
1192 // Generate the <info>, if required. If DocBookForceAbstractTag != NONE, this abstract will not be in
1193 // <info>, unlike other ("standard") abstracts.
1194 bool hasStandardAbstract = !secInfo.abstract.empty() && style.docbookforceabstracttag() == "NONE";
1195 bool needInfo = !secInfo.mustBeInInfo.empty() || hasStandardAbstract;
1198 xs.startDivision(false);
1199 xs << xml::StartTag("info");
1203 // Output the elements that should go in <info>, before and after the abstract.
1204 for (auto pit : secInfo.shouldBeInInfo) // Typically, the title: these elements are so important and ubiquitous
1205 // that mandating a wrapper like <info> would repel users. Thus, generate them first.
1206 makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(pit));
1207 for (auto pit : secInfo.mustBeInInfo)
1208 makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(pit));
1210 // Deal with the abstract in <info> if it is standard (i.e. its tag is <abstract>).
1211 if (!secInfo.abstract.empty() && hasStandardAbstract) {
1212 if (!secInfo.abstractLayout) {
1213 xs << xml::StartTag("abstract");
1217 for (auto const &p : secInfo.abstract)
1218 makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(p));
1220 if (!secInfo.abstractLayout) {
1221 xs << xml::EndTag("abstract");
1226 // End the <info> tag if it was started.
1228 if (!xs.isLastTagCR())
1231 xs << xml::EndTag("info");
1236 // Deal with the abstract outside <info> if it is not standard (i.e. its tag is layout-defined).
1237 if (!secInfo.abstract.empty() && !hasStandardAbstract) {
1238 // Assert: style.docbookforceabstracttag() != NONE.
1239 xs << xml::StartTag(style.docbookforceabstracttag());
1241 for (auto const &p : secInfo.abstract)
1242 makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(p));
1243 xs << xml::EndTag(style.docbookforceabstracttag());
1247 // Skip all the text that has just been generated.
1248 par = paragraphs.iterator_at(secInfo.epit);
1250 // No <info> tag to generate, proceed as for normal paragraphs.
1251 par = makeAny(text, buf, xs, ourparams, par);
1254 // Generate this paragraph, as it has nothing special.
1255 par = makeAny(text, buf, xs, ourparams, par);
1259 // If need be, close <section>s, but only at the end of the document (otherwise, dealt with at the beginning
1261 while (!headerLevels.empty() && headerLevels.top().first > Layout::NOT_IN_TOC) {
1262 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1264 xs << XMLStream::ESCAPE_NONE << tag;