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 // Plain layouts must be ignored.
450 if (!special_case && buf.params().documentClass().isPlainLayout(par->layout()) && !runparams.docbook_force_pars)
452 // TODO: Could get rid of this with a DocBook equivalent to htmlisblock?
453 if (!special_case && par->size() == 1 && par->getInset(0)) {
454 Inset const * firstInset = par->getInset(0);
456 // Floats cannot be in paragraphs.
457 special_case = to_utf8(firstInset->layoutName()).substr(0, 6) == "Float:";
459 // Bibliographies cannot be in paragraphs.
460 if (!special_case && firstInset->asInsetCommand())
461 special_case = firstInset->asInsetCommand()->params().getCmdName() == "bibtex";
463 // Equations do not deserve their own paragraph (DocBook allows them outside paragraphs).
464 if (!special_case && firstInset->asInsetMath())
467 // ERTs are in comments, not paragraphs.
468 if (!special_case && firstInset->lyxCode() == lyx::ERT_CODE)
471 // Listings should not get into their own paragraph.
472 if (!special_case && firstInset->lyxCode() == lyx::LISTINGS_CODE)
476 bool const open_par = runparams.docbook_make_pars
477 && !runparams.docbook_in_par
480 // We want to issue the closing tag if either:
481 // (i) We opened it, and either docbook_in_par is false,
482 // or we're not in the last paragraph, anyway.
483 // (ii) We didn't open it and docbook_in_par is true,
484 // but we are in the first par, and there is a next par.
487 bool const close_par = open_par && (!runparams.docbook_in_par);
489 // Determine if this paragraph has some real content. Things like new pages are not caught
490 // by Paragraph::empty(), even though they do not generate anything useful in DocBook.
491 // Thus, remove all spaces (including new lines: \r, \n) before checking for emptiness.
492 // std::all_of allows doing this check without having to copy the string.
493 // Open and close tags around each contained paragraph.
494 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(distance(begin, par)), 0);
495 for (auto & parXML : pars) {
496 if (!std::all_of(parXML.begin(), parXML.end(), ::isspace)) {
498 openParTag(xs, &*par, prevpar);
500 xs << XMLStream::ESCAPE_NONE << parXML;
503 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
509 void makeEnvironment(Text const &text,
512 OutputParams const &runparams,
513 ParagraphList::const_iterator const & par)
515 // TODO: simplify me!
516 auto const end = text.paragraphs().end();
518 // Output the opening tag for this environment, but only if it has not been previously opened (condition
519 // implemented in openParTag).
520 auto prevpar = text.paragraphs().getParagraphBefore(par);
521 openParTag(xs, &*par, prevpar); // TODO: switch in layout for par/block?
523 // Generate the contents of this environment. There is a special case if this is like some environment.
524 Layout const & style = par->layout();
525 if (style.latextype == LATEX_COMMAND) {
526 // Nothing to do (otherwise, infinite loops).
527 } else if (style.latextype == LATEX_ENVIRONMENT) {
528 // Open a wrapper tag if needed.
529 if (style.docbookitemwrappertag() != "NONE")
530 openTag(xs, style.docbookitemwrappertag(), style.docbookitemwrapperattr(), style.docbookitemwrappertagtype());
532 // Generate the label, if need be. If it is taken from the text, sep != 0 and corresponds to the first
533 // character after the label.
535 if (style.labeltype != LABEL_NO_LABEL && style.docbookitemlabeltag() != "NONE") {
536 // At least one condition must be met:
537 // - this environment is not a list
538 // - if this is a list, the label must not be manual (i.e. it must be taken from the layout)
539 if (style.latextype != LATEX_LIST_ENVIRONMENT || style.labeltype != LABEL_MANUAL) {
540 // Usual cases: maybe there is something specified at the layout level. Highly unlikely, though.
541 docstring const lbl = par->params().labelString();
544 openLabelTag(xs, style);
546 closeLabelTag(xs, style);
549 // Only variablelist gets here (or similar items defined as an extension in the layout).
550 openLabelTag(xs, style);
551 sep = par->firstWordDocBook(xs, runparams);
552 closeLabelTag(xs, style);
556 // Maybe the item is completely empty, i.e. if the first word ends at the end of the current paragraph
557 // AND if the next paragraph doesn't have the same depth (if there is such a paragraph).
558 // Common case: there is only the first word on the line, but there is a nested list instead
560 bool emptyItem = false;
561 if (sep == par->size()) { // If the separator is already at the end of this paragraph...
564 if (next_par == text.paragraphs().end()) // There is no next paragraph.
566 else // There is a next paragraph: check depth.
567 emptyItem = par->params().depth() >= next_par->params().depth();
571 // Avoid having an empty item, this is not valid DocBook. A single character is enough to force
572 // generation of a full <para>.
573 // TODO: this always worked only by magic...
576 // Generate the rest of the paragraph, if need be. Open as many inner tags as necessary.
577 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(std::distance(text.paragraphs().begin(), par)), sep);
578 auto p = pars.begin();
580 xs << XMLStream::ESCAPE_NONE << *p;
582 if (p != pars.end()) {
583 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
584 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(), par->layout().docbookiteminnertagtype());
590 makeAny(text, buf, xs, runparams, par);
593 // Close the environment.
596 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr); // TODO: switch in layout for par/block?
600 ParagraphList::const_iterator findEndOfEnvironment(
601 ParagraphList::const_iterator const & pstart,
602 ParagraphList::const_iterator const & pend)
604 // Copy-paste from XHTML. Should be factored out at some point...
605 ParagraphList::const_iterator p = pstart;
606 Layout const & bstyle = p->layout();
607 size_t const depth = p->params().depth();
608 for (++p; p != pend; ++p) {
609 Layout const & style = p->layout();
610 // It shouldn't happen that e.g. a section command occurs inside
611 // a quotation environment, at a higher depth, but as of 6/2009,
612 // it can happen. We pretend that it's just at lowest depth.
613 if (style.latextype == LATEX_COMMAND)
616 // If depth is down, we're done
617 if (p->params().depth() < depth)
620 // If depth is up, we're not done
621 if (p->params().depth() > depth)
624 // FIXME I am not sure about the first check.
625 // Surely we *could* have different layouts that count as
626 // LATEX_PARAGRAPH, right?
627 if (style.latextype == LATEX_PARAGRAPH || style != bstyle)
634 ParagraphList::const_iterator makeListEnvironment(Text const &text,
637 OutputParams const &runparams,
638 ParagraphList::const_iterator const & begin)
641 auto const end = text.paragraphs().end();
642 auto const envend = findEndOfEnvironment(par, end);
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 Layout const & style = par->layout();
653 // Open the item wrapper.
654 openTag(xs, style.docbookitemwrappertag(), style.docbookitemwrapperattr(), style.docbookitemwrappertagtype());
656 // Generate the label, if need be. If it is taken from the text, sep != 0 and corresponds to the first
657 // character after the label.
659 if (style.labeltype != LABEL_NO_LABEL && style.docbookitemlabeltag() != "NONE") {
660 if (style.labeltype == LABEL_MANUAL) {
661 // Only variablelist gets here (or similar items defined as an extension in the layout).
662 openLabelTag(xs, style);
663 sep = 1 + par->firstWordDocBook(xs, runparams);
664 closeLabelTag(xs, style);
666 // Usual cases: maybe there is something specified at the layout level. Highly unlikely, though.
667 docstring const lbl = par->params().labelString();
670 openLabelTag(xs, style);
672 closeLabelTag(xs, style);
677 // Open the item (after the wrapper and the label).
678 openTag(xs, style.docbookitemtag(), style.docbookitemattr(), style.docbookitemtagtype());
680 // Generate the content of the item.
681 if (sep < par->size()) {
682 auto pars = par->simpleDocBookOnePar(buf, runparams,
683 text.outerFont(std::distance(text.paragraphs().begin(), par)), sep);
684 for (auto &p : pars) {
685 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
686 par->layout().docbookiteminnertagtype());
687 xs << XMLStream::ESCAPE_NONE << p;
688 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
691 // DocBook doesn't like emptiness.
692 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
693 par->layout().docbookiteminnertagtype());
694 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
697 // If the next item is deeper, it must go entirely within this item (do it recursively).
698 // By construction, with findEndOfEnvironment, depth can only stay constant or increase, never decrease.
699 depth_type currentDepth = par->getDepth();
701 while (par != envend && par->getDepth() != currentDepth)
702 par = makeAny(text, buf, xs, runparams, par);
703 // Usually, this loop only makes one iteration, except in complex scenarios, like an item with a paragraph,
704 // a list, and another paragraph; or an item with two types of list (itemise then enumerate, for instance).
707 closeTag(xs, style.docbookitemtag(), style.docbookitemtagtype());
708 closeTag(xs, style.docbookitemwrappertag(), style.docbookitemwrappertagtype());
711 // Close this environment in exactly the same way as it was opened.
712 closeTag(xs, envstyle.docbooktag(), envstyle.docbooktagtype());
713 closeTag(xs, envstyle.docbookwrappertag(), envstyle.docbookwrappertagtype());
723 OutputParams const & runparams,
724 ParagraphList::const_iterator const & par)
726 // Unlike XHTML, no need for labels, as they are handled by DocBook tags.
727 auto const begin = text.paragraphs().begin();
728 auto const end = text.paragraphs().end();
732 // Generate this command.
733 auto prevpar = text.paragraphs().getParagraphBefore(par);
734 openParTag(xs, &*par, prevpar);
736 auto pars = par->simpleDocBookOnePar(buf, runparams,text.outerFont(distance(begin, par)));
737 for (auto & parXML : pars)
738 // TODO: decide what to do with openParTag/closeParTag in new lines.
739 xs << XMLStream::ESCAPE_NONE << parXML;
741 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
745 ParagraphList::const_iterator makeAny(Text const &text,
748 OutputParams const &runparams,
749 ParagraphList::const_iterator par)
751 switch (par->layout().latextype) {
753 makeCommand(text, buf, xs, runparams, par);
755 case LATEX_ENVIRONMENT:
756 makeEnvironment(text, buf, xs, runparams, par);
758 case LATEX_LIST_ENVIRONMENT:
759 case LATEX_ITEM_ENVIRONMENT:
760 // Only case when makeAny() might consume more than one paragraph.
761 return makeListEnvironment(text, buf, xs, runparams, par);
762 case LATEX_PARAGRAPH:
763 makeParagraph(text, buf, xs, runparams, par);
765 case LATEX_BIB_ENVIRONMENT:
766 makeBibliography(text, buf, xs, runparams, par);
774 using DocBookDocumentSectioning = tuple<bool, pit_type>;
777 struct DocBookInfoTag
779 const set<pit_type> shouldBeInInfo;
780 const set<pit_type> mustBeInInfo;
781 const set<pit_type> abstract;
785 DocBookInfoTag(const set<pit_type> & shouldBeInInfo, const set<pit_type> & mustBeInInfo,
786 const set<pit_type> & abstract, pit_type bpit, pit_type epit) :
787 shouldBeInInfo(shouldBeInInfo), mustBeInInfo(mustBeInInfo), abstract(abstract),
788 bpit(bpit), epit(epit) {}
792 DocBookDocumentSectioning hasDocumentSectioning(ParagraphList const ¶graphs, pit_type bpit, pit_type const epit) {
793 bool documentHasSections = false;
795 while (bpit < epit) {
796 Layout const &style = paragraphs[bpit].layout();
797 documentHasSections |= style.category() == from_utf8("Sectioning");
799 if (documentHasSections)
803 // Paragraphs before the first section: [ runparams.par_begin ; eppit )
805 return make_tuple(documentHasSections, bpit);
809 bool hasOnlyNotes(Paragraph const & par)
811 // Precondition: the paragraph is not empty. Otherwise, the function will always return true...
812 for (int i = 0; i < par.size(); ++i)
813 // If you find something that is not an inset (like actual text) or an inset that is not a note,
815 if (!par.isInset(i) || !dynamic_cast<InsetNote *>(par.insetList().get(i)))
821 DocBookInfoTag getParagraphsWithInfo(ParagraphList const ¶graphs, pit_type bpit, pit_type const epit) {
822 set<pit_type> shouldBeInInfo;
823 set<pit_type> mustBeInInfo;
824 set<pit_type> abstract;
826 // Find the first non empty paragraph by mutating bpit.
827 while (bpit < epit) {
828 Paragraph const &par = paragraphs[bpit];
829 if (par.empty() || hasOnlyNotes(par))
835 // Find the last info-like paragraph.
836 pit_type cpit = bpit;
837 bool hasAbstractLayout = false;
838 while (cpit < epit) {
839 // Skip paragraphs only containing one note.
840 Paragraph const & par = paragraphs[cpit];
841 if (hasOnlyNotes(par)) {
846 if (par.layout().docbookabstract())
847 hasAbstractLayout = true;
849 // Based on layout information, store this paragraph in one set: should be in <info>, must be.
850 Layout const &style = par.layout();
852 if (style.docbookininfo() == "always") {
853 mustBeInInfo.emplace(cpit);
854 } else if (style.docbookininfo() == "maybe") {
855 shouldBeInInfo.emplace(cpit);
857 // Hypothesis: the <info> parts should be grouped together near the beginning bpit.
858 // There may be notes in between, but nothing else.
863 // Now, cpit points to the last paragraph that has things that could go in <info>.
864 // bpit is the beginning of the <info> part.
866 // Go once again through the list of paragraphs to find the abstract. If there is an abstract
867 // layout, only consider it. Otherwise, an abstract is just a sequence of paragraphs with text.
868 if (hasAbstractLayout) {
870 while (pit < cpit) { // Don't overshoot the <info> part.
871 if (paragraphs[pit].layout().docbookabstract())
872 abstract.emplace(pit);
876 pit_type lastAbstract = epit + 1; // A nonsensical value.
877 docstring lastAbstractLayout;
880 while (pit < cpit) { // Don't overshoot the <info> part.
881 const Paragraph & par = paragraphs.at(pit);
882 if (!par.insetList().empty()) {
883 for (const auto &i : par.insetList()) {
884 if (i.inset->getText(0) != nullptr) {
885 if (lastAbstract == epit + 1) {
886 // First paragraph that matches the heuristic definition of abstract.
888 lastAbstractLayout = par.layout().name();
889 } else if (pit > lastAbstract + 1 || par.layout().name() != lastAbstractLayout) {
890 // This is either too far from the last abstract paragraph or doesn't
891 // have the right layout name, BUT there has already been an abstract
892 // in this document: done with detecting the abstract.
893 goto done; // Easier to get out of two nested loops.
896 abstract.emplace(pit);
906 return DocBookInfoTag(shouldBeInInfo, mustBeInInfo, abstract, bpit, cpit);
909 } // end anonymous namespace
912 xml::FontTag docbookStartFontTag(xml::FontTypes type)
914 return xml::FontTag(from_utf8(fontToDocBookTag(type)), from_utf8(fontToAttribute(type)), type);
918 xml::EndFontTag docbookEndFontTag(xml::FontTypes type)
920 return xml::EndFontTag(from_utf8(fontToDocBookTag(type)), type);
924 void outputDocBookInfo(
928 OutputParams const & runparams,
929 ParagraphList const & paragraphs,
930 DocBookInfoTag const & info)
932 // Perform an additional check on the abstract. Sometimes, there are many paragraphs that should go
933 // into the abstract, but none generates actual content. Thus, first generate to a temporary stream,
934 // then only create the <abstract> tag if these paragraphs generate some content.
935 // This check must be performed *before* a decision on whether or not to output <info> is made.
936 bool hasAbstract = !info.abstract.empty();
939 // Generate the abstract XML into a string before further checks.
940 odocstringstream os2;
943 auto bpit = *std::min_element(info.abstract.begin(), info.abstract.end());
944 auto epit = 1 + *std::max_element(info.abstract.begin(), info.abstract.end());
945 // info.abstract is inclusive, epit is exclusive, hence +1 for looping.
947 while (bpit < epit) {
948 makeAny(text, buf, xs2, runparams, paragraphs.iterator_at(bpit));
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(), ::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));
979 for (auto pit : info.mustBeInInfo) {
980 if (info.abstract.find(pit) == info.abstract.end()) // The abstract must be in info, but is dealt with after.
981 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
984 // Always output the abstract as the last item of the <info>, as it requires special treatment (especially if
985 // it contains several paragraphs that are empty).
987 // string tag = paragraphs[*info.abstract.begin()].layout().docbookforceabstracttag();
988 // if (tag == "NONE")
991 // xs << xml::StartTag(tag);
993 xs << XMLStream::ESCAPE_NONE << abstract;
994 // xs << xml::EndTag(tag);
998 // End the <info> tag if it was started.
1000 xs << xml::EndTag("info");
1007 void docbookFirstParagraphs(
1011 OutputParams const &runparams,
1014 // Handle the beginning of the document, supposing it has sections.
1015 // Major role: output the first <info> tag.
1017 ParagraphList const ¶graphs = text.paragraphs();
1018 pit_type bpit = runparams.par_begin;
1019 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, epit);
1020 outputDocBookInfo(text, buf, xs, runparams, paragraphs, 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.
1033 // First, the <info> tag.
1034 ParagraphList const ¶graphs = text.paragraphs();
1035 pit_type bpit = runparams.par_begin;
1036 pit_type const epit = runparams.par_end;
1037 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, epit);
1038 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
1040 // Then, the content. It starts where the <info> ends.
1041 auto par = text.paragraphs().iterator_at(info.epit);
1042 auto end = text.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, there can be no automatically
1073 // discovered abstract.
1074 bool documentHasSections;
1076 tie(documentHasSections, eppit) = hasDocumentSectioning(paragraphs, bpit, epit);
1078 if (documentHasSections) {
1079 docbookFirstParagraphs(text, buf, xs, runparams, eppit);
1082 docbookSimpleAllParagraphs(text, buf, xs, runparams);
1086 bool currentlyInAppendix = false;
1088 auto par = text.paragraphs().iterator_at(bpit);
1089 auto end = text.paragraphs().iterator_at(epit);
1090 while (par != end) {
1091 OutputParams ourparams = runparams;
1093 if (par->params().startOfAppendix())
1094 currentlyInAppendix = true;
1095 if (hasOnlyNotes(*par)) {
1100 Layout const &style = par->layout();
1102 // Think about adding <section> and/or </section>s.
1103 const bool isLayoutSectioning = style.category() == from_utf8("Sectioning");
1104 if (isLayoutSectioning) {
1105 int level = style.toclevel;
1107 // Need to close a previous section if it has the same level or a higher one (close <section> if opening a <h2>
1108 // after a <h2>, <h3>, <h4>, <h5> or <h6>). More examples:
1109 // - current: h2; back: h1; do not close any <section>
1110 // - current: h1; back: h2; close two <section> (first the <h2>, then the <h1>, so a new <h1> can come)
1111 while (!headerLevels.empty() && level <= headerLevels.top().first) {
1112 int stackLevel = headerLevels.top().first;
1113 docstring stackTag = from_utf8("</" + headerLevels.top().second + ">");
1116 // Output the tag only if it corresponds to a legit section.
1117 if (stackLevel != Layout::NOT_IN_TOC)
1118 xs << XMLStream::ESCAPE_NONE << stackTag << xml::CR();
1121 // Open the new section: first push it onto the stack, then output it in DocBook.
1122 string sectionTag = (currentlyInAppendix && style.docbooksectiontag() == "chapter") ?
1123 "appendix" : style.docbooksectiontag();
1124 headerLevels.push(std::make_pair(level, sectionTag));
1126 // Some sectioning-like elements should not be output (such as FrontMatter).
1127 if (level != Layout::NOT_IN_TOC) {
1128 // Look for a label in the title, i.e. a InsetLabel as a child.
1129 docstring id = docstring();
1130 for (pos_type i = 0; i < par->size(); ++i) {
1131 Inset const *inset = par->getInset(i);
1133 if (auto label = dynamic_cast<InsetLabel const *>(inset)) {
1134 // Generate the attributes for the section if need be.
1135 id += "xml:id=\"" + xml::cleanID(label->screenLabel()) + "\"";
1137 // Don't output the ID as a DocBook <anchor>.
1138 ourparams.docbook_anchors_to_ignore.emplace(label->screenLabel());
1140 // Cannot have multiple IDs per tag.
1146 // Write the open tag for this section.
1147 docstring tag = from_utf8("<" + sectionTag);
1149 tag += from_utf8(" ") + id;
1150 tag += from_utf8(">");
1151 xs << XMLStream::ESCAPE_NONE << tag;
1156 // Close all sections before the bibliography.
1157 // 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)?
1158 auto insetsLength = distance(par->insetList().begin(), par->insetList().end());
1159 if (insetsLength > 0) {
1160 Inset const *firstInset = par->getInset(0);
1161 if (firstInset && dynamic_cast<InsetBibtex const *>(firstInset)) {
1162 while (!headerLevels.empty()) {
1163 int level = headerLevels.top().first;
1164 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1167 // Output the tag only if it corresponds to a legit section.
1168 if (level != Layout::NOT_IN_TOC) {
1169 xs << XMLStream::ESCAPE_NONE << tag;
1176 // Generate this paragraph.
1177 par = makeAny(text, buf, xs, ourparams, par);
1180 // If need be, close <section>s, but only at the end of the document (otherwise, dealt with at the beginning
1182 while (!headerLevels.empty() && headerLevels.top().first > Layout::NOT_IN_TOC) {
1183 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1185 xs << XMLStream::ESCAPE_NONE << tag;