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 // If this kind of layout should be ignored, already leave.
416 if (par->layout().docbooktag() == "IGNORE")
420 auto const begin = text.paragraphs().begin();
421 auto const end = text.paragraphs().end();
422 auto prevpar = text.paragraphs().getParagraphBefore(par);
424 // We want to open the paragraph tag if:
425 // (i) the current layout permits multiple paragraphs
426 // (ii) we are either not already inside a paragraph (HTMLIsBlock) OR
427 // we are, but this is not the first paragraph
429 // But there is also a special case, and we first see whether we are in it.
430 // We do not want to open the paragraph tag if this paragraph contains
431 // only one item, and that item is "inline", i.e., not HTMLIsBlock (such
432 // as a branch). On the other hand, if that single item has a font change
433 // applied to it, then we still do need to open the paragraph.
435 // Obviously, this is very fragile. The main reason we need to do this is
436 // because of branches, e.g., a branch that contains an entire new section.
437 // We do not really want to wrap that whole thing in a <div>...</div>.
438 bool special_case = false;
439 Inset const *specinset = par->size() == 1 ? par->getInset(0) : nullptr;
440 if (specinset && !specinset->getLayout().htmlisblock()) { // TODO: Convert htmlisblock to a DocBook parameter?
441 Layout const &style = par->layout();
442 FontInfo const first_font = style.labeltype == LABEL_MANUAL ?
443 style.labelfont : style.font;
444 FontInfo const our_font =
445 par->getFont(buf.masterBuffer()->params(), 0,
446 text.outerFont(std::distance(begin, par))).fontInfo();
448 if (first_font == our_font)
452 size_t nInsets = std::distance(par->insetList().begin(), par->insetList().end());
454 // Plain layouts must be ignored.
455 special_case |= buf.params().documentClass().isPlainLayout(par->layout()) && !runparams.docbook_force_pars;
456 // Equations do not deserve their own paragraph (DocBook allows them outside paragraphs).
457 // Exception: any case that generates an <inlineequation> must still get a paragraph to be valid.
458 special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
459 return inset.inset && inset.inset->asInsetMath() && inset.inset->asInsetMath()->getType() != hullSimple;
461 // Tables doe not deserve their own paragraphs (DocBook allows them outside paragraphs).
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() == TABULAR_CODE;
465 // Floats cannot be in 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() == FLOAT_CODE;
469 // Bibliographies cannot be in paragraphs. Bibitems should still be handled as paragraphs, though
470 // (see makeParagraphBibliography).
471 special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
472 return inset.inset->lyxCode() == BIBTEX_CODE;
474 // ERTs are in comments, not paragraphs.
475 special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
476 return inset.inset->lyxCode() == ERT_CODE;
478 // Listings should not get into their own paragraph.
479 special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
480 return inset.inset->lyxCode() == LISTINGS_CODE;
482 // Boxes cannot get into their own paragraph.
483 special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
484 return inset.inset->lyxCode() == BOX_CODE;
486 // Includes should not have a paragraph.
487 special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
488 return inset.inset->lyxCode() == INCLUDE_CODE;
491 bool const open_par = runparams.docbook_make_pars
492 && !runparams.docbook_in_par
495 // We want to issue the closing tag if either:
496 // (i) We opened it, and either docbook_in_par is false,
497 // or we're not in the last paragraph, anyway.
498 // (ii) We didn't open it and docbook_in_par is true,
499 // but we are in the first par, and there is a next par.
500 bool const close_par = open_par && (!runparams.docbook_in_par);
502 // Determine if this paragraph has some real content. Things like new pages are not caught
503 // by Paragraph::empty(), even though they do not generate anything useful in DocBook.
504 // Thus, remove all spaces (including new lines: \r, \n) before checking for emptiness.
505 // std::all_of allows doing this check without having to copy the string.
506 // Open and close tags around each contained paragraph.
509 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(distance(begin, par)), 0, nextpar == end, special_case);
510 for (docstring const & parXML : pars) {
511 if (!xml::isNotOnlySpace(parXML))
515 openParTag(xs, &*par, prevpar);
517 xs << XMLStream::ESCAPE_NONE << parXML;
520 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
525 void makeEnvironment(Text const &text,
528 OutputParams const &runparams,
529 ParagraphList::const_iterator const & par)
531 // If this kind of layout should be ignored, already leave.
532 if (par->layout().docbooktag() == "IGNORE")
536 auto const end = text.paragraphs().end();
540 // Special cases for listing-like environments provided in layouts. This is quite ad-hoc, but provides a useful
541 // default. This should not be used by too many environments (only LyX-Code right now).
542 // This would be much simpler if LyX-Code was implemented as InsetListings...
543 bool mimicListing = false;
544 bool ignoreFonts = false;
545 if (par->layout().docbooktag() == "programlisting") {
550 // Output the opening tag for this environment, but only if it has not been previously opened (condition
551 // implemented in openParTag).
552 auto prevpar = text.paragraphs().getParagraphBefore(par);
553 openParTag(xs, &*par, prevpar); // TODO: switch in layout for par/block?
555 // Generate the contents of this environment. There is a special case if this is like some environment.
556 Layout const & style = par->layout();
557 if (style.latextype == LATEX_COMMAND) {
558 // Nothing to do (otherwise, infinite loops).
559 } else if (style.latextype == LATEX_ENVIRONMENT) {
560 // Generate the paragraph, if need be.
561 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(std::distance(text.paragraphs().begin(), par)), 0, false, ignoreFonts);
564 auto p = pars.begin();
565 while (p != pars.end()) {
566 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(), par->layout().docbookiteminnertagtype());
567 xs << XMLStream::ESCAPE_NONE << *p;
568 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
571 // Insert a new line after each "paragraph" (i.e. line in the listing), except for the last one.
572 // Otherwise, there would one more new line in the output than in the LyX document.
577 for (auto const & p : pars) {
578 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(), par->layout().docbookiteminnertagtype());
579 xs << XMLStream::ESCAPE_NONE << p;
580 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
584 makeAny(text, buf, xs, runparams, par);
587 // Close the environment.
588 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr); // TODO: switch in layout for par/block?
592 ParagraphList::const_iterator findEndOfEnvironment(
593 ParagraphList::const_iterator const & pstart,
594 ParagraphList::const_iterator const & pend)
596 // Copy-paste from XHTML. Should be factored out at some point...
597 ParagraphList::const_iterator p = pstart;
598 Layout const & bstyle = p->layout();
599 size_t const depth = p->params().depth();
600 for (++p; p != pend; ++p) {
601 Layout const & style = p->layout();
602 // It shouldn't happen that e.g. a section command occurs inside
603 // a quotation environment, at a higher depth, but as of 6/2009,
604 // it can happen. We pretend that it's just at lowest depth.
605 if (style.latextype == LATEX_COMMAND)
608 // If depth is down, we're done
609 if (p->params().depth() < depth)
612 // If depth is up, we're not done
613 if (p->params().depth() > depth)
616 // FIXME I am not sure about the first check.
617 // Surely we *could* have different layouts that count as
618 // LATEX_PARAGRAPH, right?
619 if (style.latextype == LATEX_PARAGRAPH || style != bstyle)
626 ParagraphList::const_iterator makeListEnvironment(Text const &text,
629 OutputParams const &runparams,
630 ParagraphList::const_iterator const & begin)
634 auto const end = text.paragraphs().end();
635 auto const envend = findEndOfEnvironment(par, end);
637 // If this kind of layout should be ignored, already leave.
638 if (begin->layout().docbooktag() == "IGNORE") {
644 // Output the opening tag for this environment.
645 Layout const & envstyle = par->layout();
646 openTag(xs, envstyle.docbookwrappertag(), envstyle.docbookwrapperattr(), envstyle.docbookwrappertagtype());
647 openTag(xs, envstyle.docbooktag(), envstyle.docbookattr(), envstyle.docbooktagtype());
649 // Handle the content of the list environment, item by item.
650 while (par != envend) {
651 // Skip this paragraph if it is both empty and the last one (otherwise, there may be deeper paragraphs after).
654 if (par->empty() && nextpar == envend)
657 // Open the item wrapper.
658 Layout const & style = par->layout();
659 openTag(xs, style.docbookitemwrappertag(), style.docbookitemwrapperattr(), style.docbookitemwrappertagtype());
661 // Generate the label, if need be. If it is taken from the text, sep != 0 and corresponds to the first
662 // character after the label.
664 if (style.labeltype != LABEL_NO_LABEL && style.docbookitemlabeltag() != "NONE") {
665 if (style.labeltype == LABEL_MANUAL) {
666 // Only variablelist gets here (or similar items defined as an extension in the layout).
667 openTag(xs, style.docbookitemlabeltag(), style.docbookitemlabelattr(), style.docbookitemlabeltagtype());
668 sep = 1 + par->firstWordDocBook(xs, runparams);
669 closeTag(xs, style.docbookitemlabeltag(), style.docbookitemlabeltagtype());
671 // Usual cases: maybe there is something specified at the layout level. Highly unlikely, though.
672 docstring const lbl = par->params().labelString();
675 openTag(xs, style.docbookitemlabeltag(), style.docbookitemlabelattr(), style.docbookitemlabeltagtype());
677 closeTag(xs, style.docbookitemlabeltag(), style.docbookitemlabeltagtype());
682 // Open the item (after the wrapper and the label).
683 openTag(xs, style.docbookitemtag(), style.docbookitemattr(), style.docbookitemtagtype());
685 // Generate the content of the item.
686 if (sep < par->size()) {
687 auto pars = par->simpleDocBookOnePar(buf, runparams,
688 text.outerFont(std::distance(text.paragraphs().begin(), par)), sep);
689 for (auto &p : pars) {
690 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
691 par->layout().docbookiteminnertagtype());
692 xs << XMLStream::ESCAPE_NONE << p;
693 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
696 // DocBook doesn't like emptiness.
697 compTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
698 par->layout().docbookiteminnertagtype());
701 // If the next item is deeper, it must go entirely within this item (do it recursively).
702 // By construction, with findEndOfEnvironment, depth can only stay constant or increase, never decrease.
703 depth_type currentDepth = par->getDepth();
705 while (par != envend && par->getDepth() != currentDepth)
706 par = makeAny(text, buf, xs, runparams, par);
707 // Usually, this loop only makes one iteration, except in complex scenarios, like an item with a paragraph,
708 // a list, and another paragraph; or an item with two types of list (itemise then enumerate, for instance).
711 closeTag(xs, style.docbookitemtag(), style.docbookitemtagtype());
712 closeTag(xs, style.docbookitemwrappertag(), style.docbookitemwrappertagtype());
715 // Close this environment in exactly the same way as it was opened.
716 closeTag(xs, envstyle.docbooktag(), envstyle.docbooktagtype());
717 closeTag(xs, envstyle.docbookwrappertag(), envstyle.docbookwrappertagtype());
727 OutputParams const & runparams,
728 ParagraphList::const_iterator const & par)
730 // If this kind of layout should be ignored, already leave.
731 if (par->layout().docbooktag() == "IGNORE")
735 // Unlike XHTML, no need for labels, as they are handled by DocBook tags.
736 auto const begin = text.paragraphs().begin();
737 auto const end = text.paragraphs().end();
741 // Generate this command.
742 auto prevpar = text.paragraphs().getParagraphBefore(par);
743 openParTag(xs, &*par, prevpar);
745 auto pars = par->simpleDocBookOnePar(buf, runparams,text.outerFont(distance(begin, par)));
746 for (auto & parXML : pars)
747 // TODO: decide what to do with openParTag/closeParTag in new lines.
748 xs << XMLStream::ESCAPE_NONE << parXML;
750 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
754 bool isLayoutSectioning(Layout const & lay)
756 if (lay.docbooksection()) // Special case: some DocBook styles must be handled as sections.
758 else if (lay.category() == from_utf8("Sectioning") || lay.docbooktag() == "section") // Generic case.
759 return lay.toclevel != Layout::NOT_IN_TOC;
764 bool isLayoutSectioningOrSimilar(Layout const & lay)
766 return isLayoutSectioning(lay) || lay.docbooktag() == "bridgehead";
770 using DocBookDocumentSectioning = tuple<bool, pit_type>;
773 struct DocBookInfoTag
775 const set<pit_type> shouldBeInInfo;
776 const set<pit_type> mustBeInInfo; // With the notable exception of the abstract!
777 const set<pit_type> abstract;
778 const bool abstractLayout;
782 DocBookInfoTag(const set<pit_type> & shouldBeInInfo, const set<pit_type> & mustBeInInfo,
783 const set<pit_type> & abstract, bool abstractLayout, pit_type bpit, pit_type epit) :
784 shouldBeInInfo(shouldBeInInfo), mustBeInInfo(mustBeInInfo), abstract(abstract),
785 abstractLayout(abstractLayout), bpit(bpit), epit(epit) {}
789 DocBookDocumentSectioning hasDocumentSectioning(ParagraphList const ¶graphs, pit_type bpit, pit_type const epit) {
790 bool documentHasSections = false;
792 while (bpit < epit) {
793 Layout const &style = paragraphs[bpit].layout();
794 documentHasSections |= isLayoutSectioningOrSimilar(style);
796 if (documentHasSections)
800 // Paragraphs before the first section: [ runparams.par_begin ; eppit )
802 return make_tuple(documentHasSections, bpit);
806 bool hasOnlyNotes(Paragraph const & par)
808 // Precondition: the paragraph is not empty. Otherwise, the function will always return true...
809 for (int i = 0; i < par.size(); ++i)
810 // If you find something that is not an inset (like actual text) or an inset that is not a note,
812 if (!par.isInset(i) || par.getInset(i)->lyxCode() != NOTE_CODE)
815 // An empty paragraph may still require some output.
816 if (par.layout().docbooksection())
819 // There should be really no content here.
824 DocBookInfoTag getParagraphsWithInfo(ParagraphList const ¶graphs,
825 pit_type bpit, pit_type const epit,
826 // Typically, bpit is the beginning of the document and epit the end of the
827 // document *or* the first section.
828 bool documentHasSections,
829 bool detectUnlayoutedAbstract
830 // Whether paragraphs with no specific layout should be detected as abstracts.
831 // For inner sections, an abstract should only be detected if it has a specific
832 // layout. For others, anything that might look like an abstract should be sought.
834 set<pit_type> shouldBeInInfo;
835 set<pit_type> mustBeInInfo;
836 set<pit_type> abstractWithLayout;
837 set<pit_type> abstractNoLayout;
839 // Find the first non empty paragraph by mutating bpit.
840 while (bpit < epit) {
841 Paragraph const &par = paragraphs[bpit];
842 if (par.empty() || hasOnlyNotes(par))
848 // Traverse everything that might belong to <info>.
849 bool hasAbstractLayout = false;
850 pit_type cpit = bpit;
851 for (; cpit < epit; ++cpit) {
852 // Skip paragraphs that don't generate anything in DocBook.
853 Paragraph const & par = paragraphs[cpit];
854 Layout const &style = par.layout();
855 if (hasOnlyNotes(par))
858 // There should never be any section here, except for the first paragraph (a title can be part of <info>).
859 // (Just a sanity check: if this fails, this function could end up processing the whole document.)
860 if (cpit != bpit && isLayoutSectioningOrSimilar(par.layout())) {
861 LYXERR0("Assertion failed: section found in potential <info> paragraphs.");
865 // If this is marked as an abstract by the layout, put it in the right set.
866 if (style.docbookabstract()) {
867 hasAbstractLayout = true;
868 abstractWithLayout.emplace(cpit);
872 // Based on layout information, store this paragraph in one set: should be in <info>, must be,
873 // or abstract (either because of layout or of position).
874 if (style.docbookininfo() == "always")
875 mustBeInInfo.emplace(cpit);
876 else if (style.docbookininfo() == "maybe")
877 shouldBeInInfo.emplace(cpit);
878 else if (documentHasSections && !hasAbstractLayout && detectUnlayoutedAbstract &&
879 (style.docbooktag() == "NONE" || style.docbooktag() == "para") &&
880 style.docbookwrappertag() == "NONE")
881 // In this case, it is very likely that style.docbookininfo() == "never"! Be extra careful
882 // about anything that gets caught here.
883 abstractNoLayout.emplace(cpit);
884 else // This should definitely not be in <info>.
887 // Now, cpit points to the first paragraph that no more has things that could go in <info>.
888 // bpit is the beginning of the <info> part.
890 return DocBookInfoTag(shouldBeInInfo, mustBeInInfo,
891 hasAbstractLayout ? abstractWithLayout : abstractNoLayout,
892 hasAbstractLayout, bpit, cpit);
895 } // end anonymous namespace
898 ParagraphList::const_iterator makeAny(Text const &text,
901 OutputParams const &runparams,
902 ParagraphList::const_iterator par)
904 switch (par->layout().latextype) {
906 makeCommand(text, buf, xs, runparams, par);
908 case LATEX_ENVIRONMENT:
909 makeEnvironment(text, buf, xs, runparams, par);
911 case LATEX_LIST_ENVIRONMENT:
912 case LATEX_ITEM_ENVIRONMENT:
913 // Only case when makeAny() might consume more than one paragraph.
914 return makeListEnvironment(text, buf, xs, runparams, par);
915 case LATEX_PARAGRAPH:
916 makeParagraph(text, buf, xs, runparams, par);
918 case LATEX_BIB_ENVIRONMENT:
919 makeBibliography(text, buf, xs, runparams, par);
927 xml::FontTag docbookStartFontTag(xml::FontTypes type)
929 return xml::FontTag(from_utf8(fontToDocBookTag(type)), from_utf8(fontToAttribute(type)), type);
933 xml::EndFontTag docbookEndFontTag(xml::FontTypes type)
935 return xml::EndFontTag(from_utf8(fontToDocBookTag(type)), type);
939 void outputDocBookInfo(
943 OutputParams const & runparams,
944 ParagraphList const & paragraphs,
945 DocBookInfoTag const & info)
947 // Perform an additional check on the abstract. Sometimes, there are many paragraphs that should go
948 // into the abstract, but none generates actual content. Thus, first generate to a temporary stream,
949 // then only create the <abstract> tag if these paragraphs generate some content.
950 // This check must be performed *before* a decision on whether or not to output <info> is made.
951 bool hasAbstract = !info.abstract.empty();
954 // Generate the abstract XML into a string before further checks.
955 // Usually, makeAny only generates one paragraph at a time. However, for the specific case of lists, it might
956 // generate more than one paragraph, as indicated in the return value.
957 odocstringstream os2;
960 set<pit_type> doneParas;
961 for (auto const & p : info.abstract) {
962 if (doneParas.find(p) == doneParas.end()) {
963 auto oldPar = paragraphs.iterator_at(p);
964 auto newPar = makeAny(text, buf, xs2, runparams, oldPar);
966 // Insert the indices of all the paragraphs that were just generated (typically, one).
967 // **Make the hypothesis that, when an abstract has a list, all its items are consecutive.**
969 while (oldPar != newPar) {
970 doneParas.emplace(id);
977 // Actually output the abstract if there is something to do. Don't count line feeds or spaces in this,
978 // even though they must be properly output if there is some abstract.
979 abstract = os2.str();
980 docstring cleaned = abstract;
981 cleaned.erase(std::remove_if(cleaned.begin(), cleaned.end(), lyx::isSpace), cleaned.end());
983 // Nothing? Then there is no abstract!
988 // The abstract must go in <info>. Otherwise, decide whether to open <info> based on the layouts.
989 bool needInfo = !info.mustBeInInfo.empty() || hasAbstract;
991 // Start the <info> tag if required.
993 xs.startDivision(false);
994 xs << xml::StartTag("info");
998 // Output the elements that should go in <info>, before and after the abstract.
999 for (auto pit : info.shouldBeInInfo) // Typically, the title: these elements are so important and ubiquitous
1000 // that mandating a wrapper like <info> would repel users. Thus, generate them first.
1001 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
1002 for (auto pit : info.mustBeInInfo)
1003 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
1005 // If there is no title, generate one (required for the document to be valid).
1006 // This code is called for the main document, for table cells, etc., so be precise in this condition.
1007 if (text.isMainText() && info.shouldBeInInfo.empty() && !runparams.inInclude) {
1008 xs << xml::StartTag("title");
1009 xs << "Untitled Document";
1010 xs << xml::EndTag("title");
1014 // Always output the abstract as the last item of the <info>, as it requires special treatment (especially if
1015 // it contains several paragraphs that are empty).
1017 if (info.abstractLayout) {
1018 xs << XMLStream::ESCAPE_NONE << abstract;
1021 string tag = paragraphs[*info.abstract.begin()].layout().docbookforceabstracttag();
1025 if (!xs.isLastTagCR())
1028 xs << xml::StartTag(tag);
1030 xs << XMLStream::ESCAPE_NONE << abstract;
1031 xs << xml::EndTag(tag);
1036 // End the <info> tag if it was started.
1038 if (!xs.isLastTagCR())
1041 xs << xml::EndTag("info");
1048 void docbookSimpleAllParagraphs(
1052 OutputParams const & runparams)
1054 // Handle the given text, supposing it has no sections (i.e. a "simple" text). The input may vary in length
1055 // between a single paragraph to a whole document.
1056 pit_type const bpit = runparams.par_begin;
1057 pit_type const epit = runparams.par_end;
1058 ParagraphList const ¶graphs = text.paragraphs();
1060 // First, the <info> tag.
1061 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, epit, false, true);
1062 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
1064 // Then, the content. It starts where the <info> ends.
1065 auto par = paragraphs.iterator_at(info.epit);
1066 auto end = paragraphs.iterator_at(epit);
1067 while (par != end) {
1068 if (!hasOnlyNotes(*par))
1069 par = makeAny(text, buf, xs, runparams, par);
1076 void docbookParagraphs(Text const &text,
1079 OutputParams const &runparams) {
1080 ParagraphList const ¶graphs = text.paragraphs();
1081 if (runparams.par_begin == runparams.par_end) {
1082 runparams.par_begin = 0;
1083 runparams.par_end = paragraphs.size();
1085 pit_type bpit = runparams.par_begin;
1086 pit_type const epit = runparams.par_end;
1087 LASSERT(bpit < epit,
1089 xs << XMLStream::ESCAPE_NONE << "<!-- DocBook output error! -->\n";
1093 std::stack<std::pair<int, string>> headerLevels; // Used to determine when to open/close sections: store the depth
1094 // of the section and the tag that was used to open it.
1096 // Detect whether the document contains sections. If there are no sections, treatment is largely simplified.
1097 // In particular, there can't be an abstract, unless it is manually marked.
1098 bool documentHasSections;
1100 tie(documentHasSections, eppit) = hasDocumentSectioning(paragraphs, bpit, epit);
1102 // Deal with "simple" documents, i.e. those without sections.
1103 if (!documentHasSections) {
1104 docbookSimpleAllParagraphs(text, buf, xs, runparams);
1108 // Output the first <info> tag (or just the title).
1109 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, eppit, true, true);
1110 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
1113 // Then, iterate through the paragraphs of this document.
1114 bool currentlyInAppendix = false;
1116 auto par = text.paragraphs().iterator_at(bpit);
1117 auto end = text.paragraphs().iterator_at(epit);
1118 while (par != end) {
1119 OutputParams ourparams = runparams;
1121 if (par->params().startOfAppendix())
1122 currentlyInAppendix = true;
1123 if (hasOnlyNotes(*par)) {
1128 Layout const &style = par->layout();
1130 // Think about adding <section> and/or </section>s.
1131 if (isLayoutSectioning(style)) {
1132 int level = style.toclevel;
1134 // Need to close a previous section if it has the same level or a higher one (close <section> if opening a
1135 // <h2> after a <h2>, <h3>, <h4>, <h5> or <h6>). More examples:
1136 // - current: h2; back: h1; do not close any <section>
1137 // - current: h1; back: h2; close two <section> (first the <h2>, then the <h1>, so a new <h1> can come)
1138 while (!headerLevels.empty() && level <= headerLevels.top().first) {
1139 // Output the tag only if it corresponds to a legit section.
1140 int stackLevel = headerLevels.top().first;
1141 if (stackLevel != Layout::NOT_IN_TOC) {
1142 xs << xml::EndTag(headerLevels.top().second);
1148 // Open the new section: first push it onto the stack, then output it in DocBook.
1149 string sectionTag = (currentlyInAppendix && style.docbooksectiontag() == "chapter") ?
1150 "appendix" : style.docbooksectiontag();
1151 headerLevels.push(std::make_pair(level, sectionTag));
1153 // Some sectioning-like elements should not be output (such as FrontMatter).
1154 if (level != Layout::NOT_IN_TOC) {
1155 // Look for a label in the title, i.e. a InsetLabel as a child.
1156 docstring id = docstring();
1157 for (pos_type i = 0; i < par->size(); ++i) {
1158 Inset const *inset = par->getInset(i);
1160 if (auto label = dynamic_cast<InsetLabel const *>(inset)) {
1161 // Generate the attributes for the section if need be.
1162 id += "xml:id=\"" + xml::cleanID(label->screenLabel()) + "\"";
1164 // Don't output the ID as a DocBook <anchor>.
1165 ourparams.docbook_anchors_to_ignore.emplace(label->screenLabel());
1167 // Cannot have multiple IDs per tag. If there is another ID inset in the document, it will
1168 // be output as a DocBook anchor.
1174 // Write the open tag for this section.
1178 xs << xml::StartTag(sectionTag, attrs);
1183 // Close all sections before the bibliography.
1184 // 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)?
1185 if (!par->insetList().empty()) {
1186 Inset const *firstInset = par->getInset(0);
1187 if (firstInset && (firstInset->lyxCode() == BIBITEM_CODE || firstInset->lyxCode() == BIBTEX_CODE)) {
1188 while (!headerLevels.empty()) {
1189 int level = headerLevels.top().first;
1190 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1193 // Output the tag only if it corresponds to a legit section.
1194 if (level != Layout::NOT_IN_TOC) {
1195 xs << XMLStream::ESCAPE_NONE << tag;
1202 // Generate the <info> tag if a section was just opened.
1203 // Some sections may require abstracts (mostly parts, in books: DocBookForceAbstractTag will not be NONE),
1204 // others can still have an abstract (it must be detected so that it can be output at the right place).
1205 // 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.
1206 if (isLayoutSectioning(style)) {
1207 // This abstract may be found between the next paragraph and the next title.
1208 pit_type cpit = std::distance(text.paragraphs().begin(), par);
1209 pit_type ppit = std::get<1>(hasDocumentSectioning(paragraphs, cpit + 1L, epit));
1211 // Generate this abstract (this code corresponds to parts of outputDocBookInfo).
1212 DocBookInfoTag secInfo = getParagraphsWithInfo(paragraphs, cpit, ppit, true,
1213 style.docbookforceabstracttag() != "NONE");
1215 if (!secInfo.mustBeInInfo.empty() || !secInfo.shouldBeInInfo.empty() || !secInfo.abstract.empty()) {
1216 // Generate the <info>, if required. If DocBookForceAbstractTag != NONE, this abstract will not be in
1217 // <info>, unlike other ("standard") abstracts.
1218 bool hasStandardAbstract = !secInfo.abstract.empty() && style.docbookforceabstracttag() == "NONE";
1219 bool needInfo = !secInfo.mustBeInInfo.empty() || hasStandardAbstract;
1222 xs.startDivision(false);
1223 xs << xml::StartTag("info");
1227 // Output the elements that should go in <info>, before and after the abstract.
1228 for (auto pit : secInfo.shouldBeInInfo) // Typically, the title: these elements are so important and ubiquitous
1229 // that mandating a wrapper like <info> would repel users. Thus, generate them first.
1230 makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(pit));
1231 for (auto pit : secInfo.mustBeInInfo)
1232 makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(pit));
1234 // Deal with the abstract in <info> if it is standard (i.e. its tag is <abstract>).
1235 if (!secInfo.abstract.empty() && hasStandardAbstract) {
1236 if (!secInfo.abstractLayout) {
1237 xs << xml::StartTag("abstract");
1241 for (auto const &p : secInfo.abstract)
1242 makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(p));
1244 if (!secInfo.abstractLayout) {
1245 xs << xml::EndTag("abstract");
1250 // End the <info> tag if it was started.
1252 if (!xs.isLastTagCR())
1255 xs << xml::EndTag("info");
1260 // Deal with the abstract outside <info> if it is not standard (i.e. its tag is layout-defined).
1261 if (!secInfo.abstract.empty() && !hasStandardAbstract) {
1262 // Assert: style.docbookforceabstracttag() != NONE.
1263 xs << xml::StartTag(style.docbookforceabstracttag());
1265 for (auto const &p : secInfo.abstract)
1266 makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(p));
1267 xs << xml::EndTag(style.docbookforceabstracttag());
1271 // Skip all the text that has just been generated.
1272 par = paragraphs.iterator_at(secInfo.epit);
1274 // No <info> tag to generate, proceed as for normal paragraphs.
1275 par = makeAny(text, buf, xs, ourparams, par);
1278 // Generate this paragraph, as it has nothing special.
1279 par = makeAny(text, buf, xs, ourparams, par);
1283 // If need be, close <section>s, but only at the end of the document (otherwise, dealt with at the beginning
1285 while (!headerLevels.empty() && headerLevels.top().first > Layout::NOT_IN_TOC) {
1286 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1288 xs << XMLStream::ESCAPE_NONE << tag;