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.
15 #include "buffer_funcs.h"
16 #include "BufferParams.h"
18 #include "InsetList.h"
19 #include "Paragraph.h"
20 #include "ParagraphList.h"
21 #include "ParagraphParameters.h"
24 #include "TextClass.h"
26 #include "insets/InsetBibtex.h"
27 #include "insets/InsetBibitem.h"
28 #include "insets/InsetLabel.h"
29 #include "insets/InsetNote.h"
31 #include "support/lassert.h"
39 using namespace lyx::support;
45 std::string fontToDocBookTag(xml::FontTypes type)
48 case xml::FontTypes::FT_EMPH:
49 case xml::FontTypes::FT_BOLD:
51 case xml::FontTypes::FT_NOUN:
53 case xml::FontTypes::FT_UBAR:
54 case xml::FontTypes::FT_WAVE:
55 case xml::FontTypes::FT_DBAR:
56 case xml::FontTypes::FT_SOUT:
57 case xml::FontTypes::FT_XOUT:
58 case xml::FontTypes::FT_ITALIC:
59 case xml::FontTypes::FT_UPRIGHT:
60 case xml::FontTypes::FT_SLANTED:
61 case xml::FontTypes::FT_SMALLCAPS:
62 case xml::FontTypes::FT_ROMAN:
63 case xml::FontTypes::FT_SANS:
65 case xml::FontTypes::FT_TYPE:
67 case xml::FontTypes::FT_SIZE_TINY:
68 case xml::FontTypes::FT_SIZE_SCRIPT:
69 case xml::FontTypes::FT_SIZE_FOOTNOTE:
70 case xml::FontTypes::FT_SIZE_SMALL:
71 case xml::FontTypes::FT_SIZE_NORMAL:
72 case xml::FontTypes::FT_SIZE_LARGE:
73 case xml::FontTypes::FT_SIZE_LARGER:
74 case xml::FontTypes::FT_SIZE_LARGEST:
75 case xml::FontTypes::FT_SIZE_HUGE:
76 case xml::FontTypes::FT_SIZE_HUGER:
77 case xml::FontTypes::FT_SIZE_INCREASE:
78 case xml::FontTypes::FT_SIZE_DECREASE:
86 string fontToRole(xml::FontTypes type)
88 // Specific fonts are achieved with roles. The only common ones are "" for basic emphasis,
89 // and "bold"/"strong" for bold. With some specific options, other roles are copied into
90 // HTML output (via the DocBook XSLT sheets); otherwise, if not recognised, they are just ignored.
91 // Hence, it is not a problem to have many roles by default here.
92 // See https://www.sourceware.org/ml/docbook/2003-05/msg00269.html
94 case xml::FontTypes::FT_ITALIC:
95 case xml::FontTypes::FT_EMPH:
97 case xml::FontTypes::FT_BOLD:
99 case xml::FontTypes::FT_NOUN: // Outputs a <person>
100 case xml::FontTypes::FT_TYPE: // Outputs a <code>
102 case xml::FontTypes::FT_UBAR:
105 // All other roles are non-standard for DocBook.
107 case xml::FontTypes::FT_WAVE:
109 case xml::FontTypes::FT_DBAR:
111 case xml::FontTypes::FT_SOUT:
113 case xml::FontTypes::FT_XOUT:
115 case xml::FontTypes::FT_UPRIGHT:
117 case xml::FontTypes::FT_SLANTED:
119 case xml::FontTypes::FT_SMALLCAPS:
121 case xml::FontTypes::FT_ROMAN:
123 case xml::FontTypes::FT_SANS:
125 case xml::FontTypes::FT_SIZE_TINY:
127 case xml::FontTypes::FT_SIZE_SCRIPT:
128 return "size_script";
129 case xml::FontTypes::FT_SIZE_FOOTNOTE:
130 return "size_footnote";
131 case xml::FontTypes::FT_SIZE_SMALL:
133 case xml::FontTypes::FT_SIZE_NORMAL:
134 return "size_normal";
135 case xml::FontTypes::FT_SIZE_LARGE:
137 case xml::FontTypes::FT_SIZE_LARGER:
138 return "size_larger";
139 case xml::FontTypes::FT_SIZE_LARGEST:
140 return "size_largest";
141 case xml::FontTypes::FT_SIZE_HUGE:
143 case xml::FontTypes::FT_SIZE_HUGER:
145 case xml::FontTypes::FT_SIZE_INCREASE:
146 return "size_increase";
147 case xml::FontTypes::FT_SIZE_DECREASE:
148 return "size_decrease";
155 string fontToAttribute(xml::FontTypes type) {
156 // If there is a role (i.e. nonstandard use of a tag), output the attribute. Otherwise, the sheer tag is sufficient
158 string role = fontToRole(type);
160 return "role='" + role + "'";
167 // Convenience functions to open and close tags. First, very low-level ones to ensure a consistent new-line behaviour.
171 // Contents of the block.
176 // <paratag>Contents of the paragraph.</paratag>
179 // Content before<inlinetag>Contents of the paragraph.</inlinetag>Content after
181 void openInlineTag(XMLStream & xs, const std::string & tag, const std::string & attr)
183 xs << xml::StartTag(tag, attr);
187 void closeInlineTag(XMLStream & xs, const std::string & tag)
189 xs << xml::EndTag(tag);
193 void openParTag(XMLStream & xs, const std::string & tag, const std::string & attr)
195 if (!xs.isLastTagCR())
197 xs << xml::StartTag(tag, attr);
201 void closeParTag(XMLStream & xs, const std::string & tag)
203 xs << xml::EndTag(tag);
208 void openBlockTag(XMLStream & xs, const std::string & tag, const std::string & attr)
210 if (!xs.isLastTagCR())
212 xs << xml::StartTag(tag, attr);
217 void closeBlockTag(XMLStream & xs, const std::string & tag)
219 if (!xs.isLastTagCR())
221 xs << xml::EndTag(tag);
226 void openTag(XMLStream & xs, const std::string & tag, const std::string & attr, const std::string & tagtype)
228 if (tag.empty() || tag == "NONE") // Common check to be performed elsewhere, if it was not here.
231 if (tag == "para" || tagtype == "paragraph") // Special case for <para>: always considered as a paragraph.
232 openParTag(xs, tag, attr);
233 else if (tagtype == "block")
234 openBlockTag(xs, tag, attr);
235 else if (tagtype == "inline")
236 openInlineTag(xs, tag, attr);
238 xs.writeError("Unrecognised tag type '" + tagtype + "' for '" + tag + " " + attr + "'");
242 void closeTag(XMLStream & xs, const std::string & tag, const std::string & tagtype)
244 if (tag.empty() || tag == "NONE")
247 if (tag == "para" || tagtype == "paragraph") // Special case for <para>: always considered as a paragraph.
248 closeParTag(xs, tag);
249 else if (tagtype == "block")
250 closeBlockTag(xs, tag);
251 else if (tagtype == "inline")
252 closeInlineTag(xs, tag);
254 xs.writeError("Unrecognised tag type '" + tagtype + "' for '" + tag + "'");
258 // Higher-level convenience functions.
260 void openParTag(XMLStream & xs, const Paragraph * par, const Paragraph * prevpar)
262 Layout const & lay = par->layout();
267 // When should the wrapper be opened here? Only if the previous paragraph has the SAME wrapper tag
268 // (usually, they won't have the same layout) and the CURRENT one allows merging.
269 // The main use case is author information in several paragraphs: if the name of the author is the
270 // first paragraph of an author, then merging with the previous tag does not make sense. Say the
271 // next paragraph is the affiliation, then it should be output in the same <author> tag (different
272 // layout, same wrapper tag).
273 bool openWrapper = lay.docbookwrappertag() != "NONE";
274 if (prevpar != nullptr) {
275 Layout const & prevlay = prevpar->layout();
276 if (prevlay.docbookwrappertag() != "NONE") {
277 openWrapper = prevlay.docbookwrappertag() == lay.docbookwrappertag()
278 && !lay.docbookwrappermergewithprevious();
284 openTag(xs, lay.docbookwrappertag(), lay.docbookwrapperattr(), lay.docbookwrappertagtype());
286 const string & tag = lay.docbooktag();
288 auto xmltag = xml::ParTag(tag, lay.docbookattr());
289 if (!xs.isTagOpen(xmltag, 1)) // Don't nest a paragraph directly in a paragraph.
290 // TODO: required or not?
291 // TODO: avoid creating a ParTag object just for this query...
292 openTag(xs, lay.docbooktag(), lay.docbookattr(), lay.docbooktagtype());
295 openTag(xs, lay.docbookitemtag(), lay.docbookitemattr(), lay.docbookitemtagtype());
296 openTag(xs, lay.docbookiteminnertag(), lay.docbookiteminnerattr(), lay.docbookiteminnertagtype());
300 void closeParTag(XMLStream & xs, Paragraph const * par, Paragraph const * nextpar)
305 // See comment in openParTag.
306 Layout const & lay = par->layout();
307 bool closeWrapper = lay.docbookwrappertag() != "NONE";
308 if (nextpar != nullptr) {
309 Layout const & nextlay = nextpar->layout();
310 if (nextlay.docbookwrappertag() != "NONE") {
311 closeWrapper = nextlay.docbookwrappertag() == lay.docbookwrappertag()
312 && !nextlay.docbookwrappermergewithprevious();
317 closeTag(xs, lay.docbookiteminnertag(), lay.docbookiteminnertagtype());
318 closeTag(xs, lay.docbookitemtag(), lay.docbookitemtagtype());
319 closeTag(xs, lay.docbooktag(), lay.docbooktagtype());
321 closeTag(xs, lay.docbookwrappertag(), lay.docbookwrappertagtype());
325 void openLabelTag(XMLStream & xs, Layout const & lay) // Mostly for definition lists.
327 openTag(xs, lay.docbookitemlabeltag(), lay.docbookitemlabelattr(), lay.docbookitemlabeltagtype());
331 void closeLabelTag(XMLStream & xs, Layout const & lay)
333 closeTag(xs, lay.docbookitemlabeltag(), lay.docbookitemlabeltagtype());
337 void openItemTag(XMLStream & xs, Layout const & lay)
339 openTag(xs, lay.docbookitemtag(), lay.docbookitemattr(), lay.docbookitemtagtype());
343 void closeItemTag(XMLStream & xs, Layout const & lay)
345 closeTag(xs, lay.docbookitemtag(), lay.docbookitemtagtype());
349 ParagraphList::const_iterator makeAny(Text const &,
352 OutputParams const &,
353 ParagraphList::const_iterator);
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 pbegin_before = text.paragraphs().getParagraphBefore(par);
365 if (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 attr = from_utf8("xml:id='") + bibitem->getParam("key") + from_utf8("'");
382 xs << xml::StartTag(from_utf8("bibliomixed"), attr);
384 // Generate the entry. Concatenate the different parts of the paragraph if any.
385 auto const begin = text.paragraphs().begin();
386 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(std::distance(begin, par)), 0);
387 for (auto & parXML : pars)
388 xs << XMLStream::ESCAPE_NONE << parXML;
390 // End the precooked bibliography entry.
391 xs << xml::EndTag("bibliomixed");
394 // If this is the last paragraph in a bibliography, close the bibliography tag.
395 auto const end = text.paragraphs().end();
396 bool endBibliography = par == end;
397 if (!endBibliography) {
400 endBibliography = par->layout().latextype != LATEX_BIB_ENVIRONMENT;
403 if (endBibliography) {
404 xs << xml::EndTag("bibliography");
414 OutputParams const & runparams,
415 ParagraphList::const_iterator const & par)
417 auto const begin = text.paragraphs().begin();
418 auto const end = text.paragraphs().end();
419 auto prevpar = text.paragraphs().getParagraphBefore(par);
421 // We want to open the paragraph tag if:
422 // (i) the current layout permits multiple paragraphs
423 // (ii) we are either not already inside a paragraph (HTMLIsBlock) OR
424 // we are, but this is not the first paragraph
426 // But there is also a special case, and we first see whether we are in it.
427 // We do not want to open the paragraph tag if this paragraph contains
428 // only one item, and that item is "inline", i.e., not HTMLIsBlock (such
429 // as a branch). On the other hand, if that single item has a font change
430 // applied to it, then we still do need to open the paragraph.
432 // Obviously, this is very fragile. The main reason we need to do this is
433 // because of branches, e.g., a branch that contains an entire new section.
434 // We do not really want to wrap that whole thing in a <div>...</div>.
435 bool special_case = false;
436 Inset const *specinset = par->size() == 1 ? par->getInset(0) : nullptr;
437 if (specinset && !specinset->getLayout().htmlisblock()) { // TODO: Convert htmlisblock to a DocBook parameter?
438 Layout const &style = par->layout();
439 FontInfo const first_font = style.labeltype == LABEL_MANUAL ?
440 style.labelfont : style.font;
441 FontInfo const our_font =
442 par->getFont(buf.masterBuffer()->params(), 0,
443 text.outerFont(std::distance(begin, par))).fontInfo();
445 if (first_font == our_font)
449 size_t nInsets = std::distance(par->insetList().begin(), par->insetList().end());
451 // Plain layouts must be ignored.
452 special_case |= buf.params().documentClass().isPlainLayout(par->layout()) && !runparams.docbook_force_pars;
453 // Equations do not deserve their own paragraph (DocBook allows them outside paragraphs).
454 special_case |= nInsets == par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
455 return inset.inset && inset.inset->asInsetMath();
458 // TODO: Could get rid of this with a DocBook equivalent to htmlisblock? Not for all cases, unfortunately... See above for those that have been determined not to be allowable for this potential refactoring.
459 if (!special_case && par->size() == 1 && par->getInset(0)) {
460 Inset const * firstInset = par->getInset(0);
462 // Floats cannot be in paragraphs.
463 special_case = to_utf8(firstInset->layoutName()).substr(0, 6) == "Float:";
465 // Bibliographies cannot be in paragraphs.
466 if (!special_case && firstInset->asInsetCommand())
467 special_case = firstInset->asInsetCommand()->params().getCmdName() == "bibtex";
469 // ERTs are in comments, not paragraphs.
470 if (!special_case && firstInset->lyxCode() == lyx::ERT_CODE)
473 // Listings should not get into their own paragraph.
474 if (!special_case && firstInset->lyxCode() == lyx::LISTINGS_CODE)
478 bool const open_par = runparams.docbook_make_pars
479 && !runparams.docbook_in_par
482 // We want to issue the closing tag if either:
483 // (i) We opened it, and either docbook_in_par is false,
484 // or we're not in the last paragraph, anyway.
485 // (ii) We didn't open it and docbook_in_par is true,
486 // but we are in the first par, and there is a next par.
489 bool const close_par = open_par && (!runparams.docbook_in_par);
491 // Determine if this paragraph has some real content. Things like new pages are not caught
492 // by Paragraph::empty(), even though they do not generate anything useful in DocBook.
493 // Thus, remove all spaces (including new lines: \r, \n) before checking for emptiness.
494 // std::all_of allows doing this check without having to copy the string.
495 // Open and close tags around each contained paragraph.
496 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(distance(begin, par)), 0);
497 for (auto & parXML : pars) {
498 if (!std::all_of(parXML.begin(), parXML.end(), ::isspace)) {
500 openParTag(xs, &*par, prevpar);
502 xs << XMLStream::ESCAPE_NONE << parXML;
505 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
511 void makeEnvironment(Text const &text,
514 OutputParams const &runparams,
515 ParagraphList::const_iterator const & par)
517 // TODO: simplify me!
518 auto const end = text.paragraphs().end();
520 // Output the opening tag for this environment, but only if it has not been previously opened (condition
521 // implemented in openParTag).
522 auto prevpar = text.paragraphs().getParagraphBefore(par);
523 openParTag(xs, &*par, prevpar); // TODO: switch in layout for par/block?
525 // Generate the contents of this environment. There is a special case if this is like some environment.
526 Layout const & style = par->layout();
527 if (style.latextype == LATEX_COMMAND) {
528 // Nothing to do (otherwise, infinite loops).
529 } else if (style.latextype == LATEX_ENVIRONMENT) {
530 // Open a wrapper tag if needed.
531 if (style.docbookitemwrappertag() != "NONE")
532 openTag(xs, style.docbookitemwrappertag(), style.docbookitemwrapperattr(), style.docbookitemwrappertagtype());
534 // Generate the label, if need be. If it is taken from the text, sep != 0 and corresponds to the first
535 // character after the label.
537 if (style.labeltype != LABEL_NO_LABEL && style.docbookitemlabeltag() != "NONE") {
538 // At least one condition must be met:
539 // - this environment is not a list
540 // - if this is a list, the label must not be manual (i.e. it must be taken from the layout)
541 if (style.latextype != LATEX_LIST_ENVIRONMENT || style.labeltype != LABEL_MANUAL) {
542 // Usual cases: maybe there is something specified at the layout level. Highly unlikely, though.
543 docstring const lbl = par->params().labelString();
546 openLabelTag(xs, style);
548 closeLabelTag(xs, style);
551 // Only variablelist gets here (or similar items defined as an extension in the layout).
552 openLabelTag(xs, style);
553 sep = par->firstWordDocBook(xs, runparams);
554 closeLabelTag(xs, style);
558 // Maybe the item is completely empty, i.e. if the first word ends at the end of the current paragraph
559 // AND if the next paragraph doesn't have the same depth (if there is such a paragraph).
560 // Common case: there is only the first word on the line, but there is a nested list instead
562 bool emptyItem = false;
563 if (sep == par->size()) { // If the separator is already at the end of this paragraph...
566 if (next_par == text.paragraphs().end()) // There is no next paragraph.
568 else // There is a next paragraph: check depth.
569 emptyItem = par->params().depth() >= next_par->params().depth();
573 // Avoid having an empty item, this is not valid DocBook. A single character is enough to force
574 // generation of a full <para>.
575 // TODO: this always worked only by magic...
578 // Generate the rest of the paragraph, if need be. Open as many inner tags as necessary.
579 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(std::distance(text.paragraphs().begin(), par)), sep);
580 auto p = pars.begin();
582 xs << XMLStream::ESCAPE_NONE << *p;
584 if (p != pars.end()) {
585 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
586 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(), par->layout().docbookiteminnertagtype());
592 makeAny(text, buf, xs, runparams, par);
595 // Close the environment.
598 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr); // TODO: switch in layout for par/block?
602 ParagraphList::const_iterator findEndOfEnvironment(
603 ParagraphList::const_iterator const & pstart,
604 ParagraphList::const_iterator const & pend)
606 // Copy-paste from XHTML. Should be factored out at some point...
607 ParagraphList::const_iterator p = pstart;
608 Layout const & bstyle = p->layout();
609 size_t const depth = p->params().depth();
610 for (++p; p != pend; ++p) {
611 Layout const & style = p->layout();
612 // It shouldn't happen that e.g. a section command occurs inside
613 // a quotation environment, at a higher depth, but as of 6/2009,
614 // it can happen. We pretend that it's just at lowest depth.
615 if (style.latextype == LATEX_COMMAND)
618 // If depth is down, we're done
619 if (p->params().depth() < depth)
622 // If depth is up, we're not done
623 if (p->params().depth() > depth)
626 // FIXME I am not sure about the first check.
627 // Surely we *could* have different layouts that count as
628 // LATEX_PARAGRAPH, right?
629 if (style.latextype == LATEX_PARAGRAPH || style != bstyle)
636 ParagraphList::const_iterator makeListEnvironment(Text const &text,
639 OutputParams const &runparams,
640 ParagraphList::const_iterator const & begin)
643 auto const end = text.paragraphs().end();
644 auto const envend = findEndOfEnvironment(par, end);
646 // Output the opening tag for this environment.
647 Layout const & envstyle = par->layout();
648 openTag(xs, envstyle.docbookwrappertag(), envstyle.docbookwrapperattr(), envstyle.docbookwrappertagtype());
649 openTag(xs, envstyle.docbooktag(), envstyle.docbookattr(), envstyle.docbooktagtype());
651 // Handle the content of the list environment, item by item.
652 while (par != envend) {
653 Layout const & style = par->layout();
655 // Open the item wrapper.
656 openTag(xs, style.docbookitemwrappertag(), style.docbookitemwrapperattr(), style.docbookitemwrappertagtype());
658 // Generate the label, if need be. If it is taken from the text, sep != 0 and corresponds to the first
659 // character after the label.
661 if (style.labeltype != LABEL_NO_LABEL && style.docbookitemlabeltag() != "NONE") {
662 if (style.labeltype == LABEL_MANUAL) {
663 // Only variablelist gets here (or similar items defined as an extension in the layout).
664 openLabelTag(xs, style);
665 sep = 1 + par->firstWordDocBook(xs, runparams);
666 closeLabelTag(xs, style);
668 // Usual cases: maybe there is something specified at the layout level. Highly unlikely, though.
669 docstring const lbl = par->params().labelString();
672 openLabelTag(xs, style);
674 closeLabelTag(xs, style);
679 // Open the item (after the wrapper and the label).
680 openTag(xs, style.docbookitemtag(), style.docbookitemattr(), style.docbookitemtagtype());
682 // Generate the content of the item.
683 if (sep < par->size()) {
684 auto pars = par->simpleDocBookOnePar(buf, runparams,
685 text.outerFont(std::distance(text.paragraphs().begin(), par)), sep);
686 for (auto &p : pars) {
687 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
688 par->layout().docbookiteminnertagtype());
689 xs << XMLStream::ESCAPE_NONE << p;
690 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
693 // DocBook doesn't like emptiness.
694 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
695 par->layout().docbookiteminnertagtype());
696 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
699 // If the next item is deeper, it must go entirely within this item (do it recursively).
700 // By construction, with findEndOfEnvironment, depth can only stay constant or increase, never decrease.
701 depth_type currentDepth = par->getDepth();
703 while (par != envend && par->getDepth() != currentDepth)
704 par = makeAny(text, buf, xs, runparams, par);
705 // Usually, this loop only makes one iteration, except in complex scenarios, like an item with a paragraph,
706 // a list, and another paragraph; or an item with two types of list (itemise then enumerate, for instance).
709 closeTag(xs, style.docbookitemtag(), style.docbookitemtagtype());
710 closeTag(xs, style.docbookitemwrappertag(), style.docbookitemwrappertagtype());
713 // Close this environment in exactly the same way as it was opened.
714 closeTag(xs, envstyle.docbooktag(), envstyle.docbooktagtype());
715 closeTag(xs, envstyle.docbookwrappertag(), envstyle.docbookwrappertagtype());
725 OutputParams const & runparams,
726 ParagraphList::const_iterator const & par)
728 // Unlike XHTML, no need for labels, as they are handled by DocBook tags.
729 auto const begin = text.paragraphs().begin();
730 auto const end = text.paragraphs().end();
734 // Generate this command.
735 auto prevpar = text.paragraphs().getParagraphBefore(par);
736 openParTag(xs, &*par, prevpar);
738 auto pars = par->simpleDocBookOnePar(buf, runparams,text.outerFont(distance(begin, par)));
739 for (auto & parXML : pars)
740 // TODO: decide what to do with openParTag/closeParTag in new lines.
741 xs << XMLStream::ESCAPE_NONE << parXML;
743 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
747 ParagraphList::const_iterator makeAny(Text const &text,
750 OutputParams const &runparams,
751 ParagraphList::const_iterator par)
753 switch (par->layout().latextype) {
755 makeCommand(text, buf, xs, runparams, par);
757 case LATEX_ENVIRONMENT:
758 makeEnvironment(text, buf, xs, runparams, par);
760 case LATEX_LIST_ENVIRONMENT:
761 case LATEX_ITEM_ENVIRONMENT:
762 // Only case when makeAny() might consume more than one paragraph.
763 return makeListEnvironment(text, buf, xs, runparams, par);
764 case LATEX_PARAGRAPH:
765 makeParagraph(text, buf, xs, runparams, par);
767 case LATEX_BIB_ENVIRONMENT:
768 makeBibliography(text, buf, xs, runparams, par);
776 using DocBookDocumentSectioning = tuple<bool, pit_type>;
779 struct DocBookInfoTag
781 const set<pit_type> shouldBeInInfo;
782 const set<pit_type> mustBeInInfo;
783 const set<pit_type> abstract;
787 DocBookInfoTag(const set<pit_type> & shouldBeInInfo, const set<pit_type> & mustBeInInfo,
788 const set<pit_type> & abstract, pit_type bpit, pit_type epit) :
789 shouldBeInInfo(shouldBeInInfo), mustBeInInfo(mustBeInInfo), abstract(abstract),
790 bpit(bpit), epit(epit) {}
794 DocBookDocumentSectioning hasDocumentSectioning(ParagraphList const ¶graphs, pit_type bpit, pit_type const epit) {
795 bool documentHasSections = false;
797 while (bpit < epit) {
798 Layout const &style = paragraphs[bpit].layout();
799 documentHasSections |= style.category() == from_utf8("Sectioning");
801 if (documentHasSections)
805 // Paragraphs before the first section: [ runparams.par_begin ; eppit )
807 return make_tuple(documentHasSections, bpit);
811 bool hasOnlyNotes(Paragraph const & par)
813 // Precondition: the paragraph is not empty. Otherwise, the function will always return true...
814 for (int i = 0; i < par.size(); ++i)
815 // If you find something that is not an inset (like actual text) or an inset that is not a note,
817 if (!par.isInset(i) || !dynamic_cast<InsetNote *>(par.insetList().get(i)))
823 DocBookInfoTag getParagraphsWithInfo(ParagraphList const ¶graphs, pit_type bpit, pit_type const epit) {
824 set<pit_type> shouldBeInInfo;
825 set<pit_type> mustBeInInfo;
826 set<pit_type> abstract;
828 // Find the first non empty paragraph by mutating bpit.
829 while (bpit < epit) {
830 Paragraph const &par = paragraphs[bpit];
831 if (par.empty() || hasOnlyNotes(par))
837 // Find the last info-like paragraph.
838 pit_type cpit = bpit;
839 bool hasAbstractLayout = false;
840 while (cpit < epit) {
841 // Skip paragraphs only containing one note.
842 Paragraph const & par = paragraphs[cpit];
843 if (hasOnlyNotes(par)) {
848 if (par.layout().docbookabstract())
849 hasAbstractLayout = true;
851 // Based on layout information, store this paragraph in one set: should be in <info>, must be.
852 Layout const &style = par.layout();
854 if (style.docbookininfo() == "always") {
855 mustBeInInfo.emplace(cpit);
856 } else if (style.docbookininfo() == "maybe") {
857 shouldBeInInfo.emplace(cpit);
859 // Hypothesis: the <info> parts should be grouped together near the beginning bpit.
860 // There may be notes in between, but nothing else.
865 // Now, cpit points to the last paragraph that has things that could go in <info>.
866 // bpit is the beginning of the <info> part.
868 // Go once again through the list of paragraphs to find the abstract. If there is an abstract
869 // layout, only consider it. Otherwise, an abstract is just a sequence of paragraphs with text.
870 if (hasAbstractLayout) {
872 while (pit < cpit) { // Don't overshoot the <info> part.
873 if (paragraphs[pit].layout().docbookabstract())
874 abstract.emplace(pit);
878 pit_type lastAbstract = epit + 1; // A nonsensical value.
879 docstring lastAbstractLayout;
882 while (pit < cpit) { // Don't overshoot the <info> part.
883 const Paragraph & par = paragraphs.at(pit);
884 if (!par.insetList().empty()) {
885 for (const auto &i : par.insetList()) {
886 if (i.inset->getText(0) != nullptr) {
887 if (lastAbstract == epit + 1) {
888 // First paragraph that matches the heuristic definition of abstract.
890 lastAbstractLayout = par.layout().name();
891 } else if (pit > lastAbstract + 1 || par.layout().name() != lastAbstractLayout) {
892 // This is either too far from the last abstract paragraph or doesn't
893 // have the right layout name, BUT there has already been an abstract
894 // in this document: done with detecting the abstract.
895 goto done; // Easier to get out of two nested loops.
898 abstract.emplace(pit);
908 return DocBookInfoTag(shouldBeInInfo, mustBeInInfo, abstract, bpit, cpit);
911 } // end anonymous namespace
914 xml::FontTag docbookStartFontTag(xml::FontTypes type)
916 return xml::FontTag(from_utf8(fontToDocBookTag(type)), from_utf8(fontToAttribute(type)), type);
920 xml::EndFontTag docbookEndFontTag(xml::FontTypes type)
922 return xml::EndFontTag(from_utf8(fontToDocBookTag(type)), type);
926 void outputDocBookInfo(
930 OutputParams const & runparams,
931 ParagraphList const & paragraphs,
932 DocBookInfoTag const & info)
934 // Perform an additional check on the abstract. Sometimes, there are many paragraphs that should go
935 // into the abstract, but none generates actual content. Thus, first generate to a temporary stream,
936 // then only create the <abstract> tag if these paragraphs generate some content.
937 // This check must be performed *before* a decision on whether or not to output <info> is made.
938 bool hasAbstract = !info.abstract.empty();
941 // Generate the abstract XML into a string before further checks.
942 odocstringstream os2;
945 auto bpit = *std::min_element(info.abstract.begin(), info.abstract.end());
946 auto epit = 1 + *std::max_element(info.abstract.begin(), info.abstract.end());
947 // info.abstract is inclusive, epit is exclusive, hence +1 for looping.
949 while (bpit < epit) {
950 makeAny(text, buf, xs2, runparams, paragraphs.iterator_at(bpit));
955 // Actually output the abstract if there is something to do. Don't count line feeds or spaces in this,
956 // even though they must be properly output if there is some abstract.
957 abstract = os2.str();
958 docstring cleaned = abstract;
959 cleaned.erase(std::remove_if(cleaned.begin(), cleaned.end(), ::isspace), cleaned.end());
961 // Nothing? Then there is no abstract!
966 // The abstract must go in <info>. Otherwise, decide whether to open <info> based on the layouts.
967 bool needInfo = !info.mustBeInInfo.empty() || hasAbstract;
969 // Start the <info> tag if required.
971 xs.startDivision(false);
972 xs << xml::StartTag("info");
976 // Output the elements that should go in <info>, before and after the abstract.
977 for (auto pit : info.shouldBeInInfo) { // Typically, the title: these elements are so important and ubiquitous
978 // that mandating a wrapper like <info> would repel users. Thus, generate them first.
979 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
981 for (auto pit : info.mustBeInInfo) {
982 if (info.abstract.find(pit) == info.abstract.end()) // The abstract must be in info, but is dealt with after.
983 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
986 // Always output the abstract as the last item of the <info>, as it requires special treatment (especially if
987 // it contains several paragraphs that are empty).
989 // string tag = paragraphs[*info.abstract.begin()].layout().docbookforceabstracttag();
990 // if (tag == "NONE")
993 // xs << xml::StartTag(tag);
995 xs << XMLStream::ESCAPE_NONE << abstract;
996 // xs << xml::EndTag(tag);
1000 // End the <info> tag if it was started.
1002 xs << xml::EndTag("info");
1009 void docbookFirstParagraphs(
1013 OutputParams const &runparams,
1016 // Handle the beginning of the document, supposing it has sections.
1017 // Major role: output the first <info> tag.
1019 ParagraphList const ¶graphs = text.paragraphs();
1020 pit_type bpit = runparams.par_begin;
1021 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, epit);
1022 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
1026 void docbookSimpleAllParagraphs(
1030 OutputParams const & runparams)
1032 // Handle the given text, supposing it has no sections (i.e. a "simple" text). The input may vary in length
1033 // between a single paragraph to a whole document.
1035 // First, the <info> tag.
1036 ParagraphList const ¶graphs = text.paragraphs();
1037 pit_type bpit = runparams.par_begin;
1038 pit_type const epit = runparams.par_end;
1039 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, epit);
1040 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
1042 // Then, the content. It starts where the <info> ends.
1043 auto par = text.paragraphs().iterator_at(info.epit);
1044 auto end = text.paragraphs().iterator_at(epit);
1045 while (par != end) {
1046 if (!hasOnlyNotes(*par))
1047 par = makeAny(text, buf, xs, runparams, par);
1054 void docbookParagraphs(Text const &text,
1057 OutputParams const &runparams) {
1058 ParagraphList const ¶graphs = text.paragraphs();
1059 if (runparams.par_begin == runparams.par_end) {
1060 runparams.par_begin = 0;
1061 runparams.par_end = paragraphs.size();
1063 pit_type bpit = runparams.par_begin;
1064 pit_type const epit = runparams.par_end;
1065 LASSERT(bpit < epit,
1067 xs << XMLStream::ESCAPE_NONE << "<!-- DocBook output error! -->\n";
1071 std::stack<std::pair<int, string>> headerLevels; // Used to determine when to open/close sections: store the depth
1072 // of the section and the tag that was used to open it.
1074 // Detect whether the document contains sections. If there are no sections, there can be no automatically
1075 // discovered abstract.
1076 bool documentHasSections;
1078 tie(documentHasSections, eppit) = hasDocumentSectioning(paragraphs, bpit, epit);
1080 if (documentHasSections) {
1081 docbookFirstParagraphs(text, buf, xs, runparams, eppit);
1084 docbookSimpleAllParagraphs(text, buf, xs, runparams);
1088 bool currentlyInAppendix = false;
1090 auto par = text.paragraphs().iterator_at(bpit);
1091 auto end = text.paragraphs().iterator_at(epit);
1092 while (par != end) {
1093 OutputParams ourparams = runparams;
1095 if (par->params().startOfAppendix())
1096 currentlyInAppendix = true;
1097 if (hasOnlyNotes(*par)) {
1102 Layout const &style = par->layout();
1104 // Think about adding <section> and/or </section>s.
1105 const bool isLayoutSectioning = style.category() == from_utf8("Sectioning");
1106 if (isLayoutSectioning) {
1107 int level = style.toclevel;
1109 // Need to close a previous section if it has the same level or a higher one (close <section> if opening a <h2>
1110 // after a <h2>, <h3>, <h4>, <h5> or <h6>). More examples:
1111 // - current: h2; back: h1; do not close any <section>
1112 // - current: h1; back: h2; close two <section> (first the <h2>, then the <h1>, so a new <h1> can come)
1113 while (!headerLevels.empty() && level <= headerLevels.top().first) {
1114 int stackLevel = headerLevels.top().first;
1115 docstring stackTag = from_utf8("</" + headerLevels.top().second + ">");
1118 // Output the tag only if it corresponds to a legit section.
1119 if (stackLevel != Layout::NOT_IN_TOC)
1120 xs << XMLStream::ESCAPE_NONE << stackTag << xml::CR();
1123 // Open the new section: first push it onto the stack, then output it in DocBook.
1124 string sectionTag = (currentlyInAppendix && style.docbooksectiontag() == "chapter") ?
1125 "appendix" : style.docbooksectiontag();
1126 headerLevels.push(std::make_pair(level, sectionTag));
1128 // Some sectioning-like elements should not be output (such as FrontMatter).
1129 if (level != Layout::NOT_IN_TOC) {
1130 // Look for a label in the title, i.e. a InsetLabel as a child.
1131 docstring id = docstring();
1132 for (pos_type i = 0; i < par->size(); ++i) {
1133 Inset const *inset = par->getInset(i);
1135 if (auto label = dynamic_cast<InsetLabel const *>(inset)) {
1136 // Generate the attributes for the section if need be.
1137 id += "xml:id=\"" + xml::cleanID(label->screenLabel()) + "\"";
1139 // Don't output the ID as a DocBook <anchor>.
1140 ourparams.docbook_anchors_to_ignore.emplace(label->screenLabel());
1142 // Cannot have multiple IDs per tag.
1148 // Write the open tag for this section.
1149 docstring tag = from_utf8("<" + sectionTag);
1151 tag += from_utf8(" ") + id;
1152 tag += from_utf8(">");
1153 xs << XMLStream::ESCAPE_NONE << tag;
1158 // Close all sections before the bibliography.
1159 // 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)?
1160 auto insetsLength = distance(par->insetList().begin(), par->insetList().end());
1161 if (insetsLength > 0) {
1162 Inset const *firstInset = par->getInset(0);
1163 if (firstInset && dynamic_cast<InsetBibtex const *>(firstInset)) {
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 this paragraph.
1179 par = makeAny(text, buf, xs, ourparams, par);
1182 // If need be, close <section>s, but only at the end of the document (otherwise, dealt with at the beginning
1184 while (!headerLevels.empty() && headerLevels.top().first > Layout::NOT_IN_TOC) {
1185 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1187 xs << XMLStream::ESCAPE_NONE << tag;