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 void compTag(XMLStream & xs, const std::string & tag, const std::string & attr, const std::string & tagtype)
260 if (tag.empty() || tag == "NONE")
263 // Special case for <para>: always considered as a paragraph.
264 if (tag == "para" || tagtype == "paragraph" || tagtype == "block") {
265 if (!xs.isLastTagCR())
267 xs << xml::CompTag(tag, attr);
269 } else if (tagtype == "inline") {
270 xs << xml::CompTag(tag, attr);
272 xs.writeError("Unrecognised tag type '" + tagtype + "' for '" + tag + "'");
277 // Higher-level convenience functions.
279 void openParTag(XMLStream & xs, const Paragraph * par, const Paragraph * prevpar)
281 Layout const & lay = par->layout();
286 // When should the wrapper be opened here? Only if the previous paragraph has the SAME wrapper tag
287 // (usually, they won't have the same layout) and the CURRENT one allows merging.
288 // The main use case is author information in several paragraphs: if the name of the author is the
289 // first paragraph of an author, then merging with the previous tag does not make sense. Say the
290 // next paragraph is the affiliation, then it should be output in the same <author> tag (different
291 // layout, same wrapper tag).
292 bool openWrapper = lay.docbookwrappertag() != "NONE";
293 if (prevpar != nullptr) {
294 Layout const & prevlay = prevpar->layout();
295 if (prevlay.docbookwrappertag() != "NONE") {
296 if (prevlay.docbookwrappertag() == lay.docbookwrappertag() &&
297 prevlay.docbookwrapperattr() == lay.docbookwrapperattr())
298 openWrapper = !lay.docbookwrappermergewithprevious();
306 openTag(xs, lay.docbookwrappertag(), lay.docbookwrapperattr(), lay.docbookwrappertagtype());
308 const string & tag = lay.docbooktag();
310 auto xmltag = xml::ParTag(tag, lay.docbookattr());
311 if (!xs.isTagOpen(xmltag, 1)) // Don't nest a paragraph directly in a paragraph.
312 // TODO: required or not?
313 // TODO: avoid creating a ParTag object just for this query...
314 openTag(xs, lay.docbooktag(), lay.docbookattr(), lay.docbooktagtype());
317 openTag(xs, lay.docbookitemtag(), lay.docbookitemattr(), lay.docbookitemtagtype());
318 openTag(xs, lay.docbookiteminnertag(), lay.docbookiteminnerattr(), lay.docbookiteminnertagtype());
322 void closeParTag(XMLStream & xs, Paragraph const * par, Paragraph const * nextpar)
327 // See comment in openParTag.
328 Layout const & lay = par->layout();
329 bool closeWrapper = lay.docbookwrappertag() != "NONE";
330 if (nextpar != nullptr) {
331 Layout const & nextlay = nextpar->layout();
332 if (nextlay.docbookwrappertag() != "NONE") {
333 if (nextlay.docbookwrappertag() == lay.docbookwrappertag() &&
334 nextlay.docbookwrapperattr() == lay.docbookwrapperattr())
335 closeWrapper = !nextlay.docbookwrappermergewithprevious();
342 closeTag(xs, lay.docbookiteminnertag(), lay.docbookiteminnertagtype());
343 closeTag(xs, lay.docbookitemtag(), lay.docbookitemtagtype());
344 closeTag(xs, lay.docbooktag(), lay.docbooktagtype());
346 closeTag(xs, lay.docbookwrappertag(), lay.docbookwrappertagtype());
350 void openLabelTag(XMLStream & xs, Layout const & lay) // Mostly for definition lists.
352 openTag(xs, lay.docbookitemlabeltag(), lay.docbookitemlabelattr(), lay.docbookitemlabeltagtype());
356 void closeLabelTag(XMLStream & xs, Layout const & lay)
358 closeTag(xs, lay.docbookitemlabeltag(), lay.docbookitemlabeltagtype());
362 void openItemTag(XMLStream & xs, Layout const & lay)
364 openTag(xs, lay.docbookitemtag(), lay.docbookitemattr(), lay.docbookitemtagtype());
368 void closeItemTag(XMLStream & xs, Layout const & lay)
370 closeTag(xs, lay.docbookitemtag(), lay.docbookitemtagtype());
374 ParagraphList::const_iterator makeAny(Text const &,
377 OutputParams const &,
378 ParagraphList::const_iterator);
381 void makeBibliography(
385 OutputParams const & runparams,
386 ParagraphList::const_iterator const & par)
388 // If this is the first paragraph in a bibliography, open the bibliography tag.
389 auto pbegin_before = text.paragraphs().getParagraphBefore(par);
390 if (pbegin_before->layout().latextype != LATEX_BIB_ENVIRONMENT) {
391 xs << xml::StartTag("bibliography");
395 // Start the precooked bibliography entry. This is very much like opening a paragraph tag.
396 // Don't forget the citation ID!
398 for (auto i = 0; i < par->size(); ++i) {
399 Inset const *ip = par->getInset(i);
402 if (const auto * bibitem = dynamic_cast<const InsetBibitem*>(ip)) {
403 attr = from_utf8("xml:id='") + bibitem->getParam("key") + from_utf8("'");
407 xs << xml::StartTag(from_utf8("bibliomixed"), attr);
409 // Generate the entry. Concatenate the different parts of the paragraph if any.
410 auto const begin = text.paragraphs().begin();
411 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(std::distance(begin, par)), 0);
412 for (auto & parXML : pars)
413 xs << XMLStream::ESCAPE_NONE << parXML;
415 // End the precooked bibliography entry.
416 xs << xml::EndTag("bibliomixed");
419 // If this is the last paragraph in a bibliography, close the bibliography tag.
420 auto const end = text.paragraphs().end();
423 bool endBibliography = nextpar == end || nextpar->layout().latextype != LATEX_BIB_ENVIRONMENT;
425 if (endBibliography) {
426 xs << xml::EndTag("bibliography");
432 bool isNotOnlySpace(docstring const & str)
434 for (auto const & c: str) {
435 if (c != ' ' && c != '\t' && c != '\n' && c != '\v' && c != '\f' && c != '\r')
446 OutputParams const & runparams,
447 ParagraphList::const_iterator const & par)
449 auto const begin = text.paragraphs().begin();
450 auto const end = text.paragraphs().end();
451 auto prevpar = text.paragraphs().getParagraphBefore(par);
453 // We want to open the paragraph tag if:
454 // (i) the current layout permits multiple paragraphs
455 // (ii) we are either not already inside a paragraph (HTMLIsBlock) OR
456 // we are, but this is not the first paragraph
458 // But there is also a special case, and we first see whether we are in it.
459 // We do not want to open the paragraph tag if this paragraph contains
460 // only one item, and that item is "inline", i.e., not HTMLIsBlock (such
461 // as a branch). On the other hand, if that single item has a font change
462 // applied to it, then we still do need to open the paragraph.
464 // Obviously, this is very fragile. The main reason we need to do this is
465 // because of branches, e.g., a branch that contains an entire new section.
466 // We do not really want to wrap that whole thing in a <div>...</div>.
467 bool special_case = false;
468 Inset const *specinset = par->size() == 1 ? par->getInset(0) : nullptr;
469 if (specinset && !specinset->getLayout().htmlisblock()) { // TODO: Convert htmlisblock to a DocBook parameter?
470 Layout const &style = par->layout();
471 FontInfo const first_font = style.labeltype == LABEL_MANUAL ?
472 style.labelfont : style.font;
473 FontInfo const our_font =
474 par->getFont(buf.masterBuffer()->params(), 0,
475 text.outerFont(std::distance(begin, par))).fontInfo();
477 if (first_font == our_font)
481 size_t nInsets = std::distance(par->insetList().begin(), par->insetList().end());
483 // Plain layouts must be ignored.
484 special_case |= buf.params().documentClass().isPlainLayout(par->layout()) && !runparams.docbook_force_pars;
485 // Equations do not deserve their own paragraph (DocBook allows them outside paragraphs).
486 special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
487 return inset.inset && inset.inset->asInsetMath();
490 // 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.
491 if (!special_case && par->size() == 1 && par->getInset(0)) {
492 Inset const * firstInset = par->getInset(0);
494 // Floats cannot be in paragraphs.
495 special_case = to_utf8(firstInset->layoutName()).substr(0, 6) == "Float:";
497 // Bibliographies cannot be in paragraphs.
498 if (!special_case && firstInset->asInsetCommand())
499 special_case = firstInset->asInsetCommand()->params().getCmdName() == "bibtex";
501 // ERTs are in comments, not paragraphs.
502 if (!special_case && firstInset->lyxCode() == lyx::ERT_CODE)
505 // Listings should not get into their own paragraph.
506 if (!special_case && firstInset->lyxCode() == lyx::LISTINGS_CODE)
509 // Boxes cannot get into their own paragraph.
510 if (!special_case && firstInset->lyxCode() == lyx::BOX_CODE)
514 bool const open_par = runparams.docbook_make_pars
515 && !runparams.docbook_in_par
518 // We want to issue the closing tag if either:
519 // (i) We opened it, and either docbook_in_par is false,
520 // or we're not in the last paragraph, anyway.
521 // (ii) We didn't open it and docbook_in_par is true,
522 // but we are in the first par, and there is a next par.
523 bool const close_par = open_par && (!runparams.docbook_in_par);
525 // Determine if this paragraph has some real content. Things like new pages are not caught
526 // by Paragraph::empty(), even though they do not generate anything useful in DocBook.
527 // Thus, remove all spaces (including new lines: \r, \n) before checking for emptiness.
528 // std::all_of allows doing this check without having to copy the string.
529 // Open and close tags around each contained paragraph.
532 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(distance(begin, par)), 0, nextpar == end, special_case);
533 for (docstring const & parXML : pars) {
534 if (isNotOnlySpace(parXML)) {
536 openParTag(xs, &*par, prevpar);
538 xs << XMLStream::ESCAPE_NONE << parXML;
541 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
547 void makeEnvironment(Text const &text,
550 OutputParams const &runparams,
551 ParagraphList::const_iterator const & par)
553 auto const end = text.paragraphs().end();
557 // Special cases for listing-like environments provided in layouts. This is quite ad-hoc, but provides a useful
558 // default. This should not be used by too many environments (only LyX-Code right now).
559 // This would be much simpler if LyX-Code was implemented as InsetListings...
560 bool mimicListing = false;
561 bool ignoreFonts = false;
562 if (par->layout().docbooktag() == "programlisting") {
567 // Output the opening tag for this environment, but only if it has not been previously opened (condition
568 // implemented in openParTag).
569 auto prevpar = text.paragraphs().getParagraphBefore(par);
570 openParTag(xs, &*par, prevpar); // TODO: switch in layout for par/block?
572 // Generate the contents of this environment. There is a special case if this is like some environment.
573 Layout const & style = par->layout();
574 if (style.latextype == LATEX_COMMAND) {
575 // Nothing to do (otherwise, infinite loops).
576 } else if (style.latextype == LATEX_ENVIRONMENT) {
577 // Generate the paragraph, if need be.
578 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(std::distance(text.paragraphs().begin(), par)), 0, false, ignoreFonts);
581 auto p = pars.begin();
582 while (p != pars.end()) {
583 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(), par->layout().docbookiteminnertagtype());
584 xs << XMLStream::ESCAPE_NONE << *p;
585 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
592 for (auto const & p : pars) {
593 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(), par->layout().docbookiteminnertagtype());
594 xs << XMLStream::ESCAPE_NONE << p;
595 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
599 makeAny(text, buf, xs, runparams, par);
602 // Close the environment.
603 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr); // TODO: switch in layout for par/block?
607 ParagraphList::const_iterator findEndOfEnvironment(
608 ParagraphList::const_iterator const & pstart,
609 ParagraphList::const_iterator const & pend)
611 // Copy-paste from XHTML. Should be factored out at some point...
612 ParagraphList::const_iterator p = pstart;
613 Layout const & bstyle = p->layout();
614 size_t const depth = p->params().depth();
615 for (++p; p != pend; ++p) {
616 Layout const & style = p->layout();
617 // It shouldn't happen that e.g. a section command occurs inside
618 // a quotation environment, at a higher depth, but as of 6/2009,
619 // it can happen. We pretend that it's just at lowest depth.
620 if (style.latextype == LATEX_COMMAND)
623 // If depth is down, we're done
624 if (p->params().depth() < depth)
627 // If depth is up, we're not done
628 if (p->params().depth() > depth)
631 // FIXME I am not sure about the first check.
632 // Surely we *could* have different layouts that count as
633 // LATEX_PARAGRAPH, right?
634 if (style.latextype == LATEX_PARAGRAPH || style != bstyle)
641 ParagraphList::const_iterator makeListEnvironment(Text const &text,
644 OutputParams const &runparams,
645 ParagraphList::const_iterator const & begin)
648 auto const end = text.paragraphs().end();
649 auto const envend = findEndOfEnvironment(par, end);
651 // Output the opening tag for this environment.
652 Layout const & envstyle = par->layout();
653 openTag(xs, envstyle.docbookwrappertag(), envstyle.docbookwrapperattr(), envstyle.docbookwrappertagtype());
654 openTag(xs, envstyle.docbooktag(), envstyle.docbookattr(), envstyle.docbooktagtype());
656 // Handle the content of the list environment, item by item.
657 while (par != envend) {
658 Layout const & style = par->layout();
660 // Open the item wrapper.
661 openTag(xs, style.docbookitemwrappertag(), style.docbookitemwrapperattr(), style.docbookitemwrappertagtype());
663 // Generate the label, if need be. If it is taken from the text, sep != 0 and corresponds to the first
664 // character after the label.
666 if (style.labeltype != LABEL_NO_LABEL && style.docbookitemlabeltag() != "NONE") {
667 if (style.labeltype == LABEL_MANUAL) {
668 // Only variablelist gets here (or similar items defined as an extension in the layout).
669 openLabelTag(xs, style);
670 sep = 1 + par->firstWordDocBook(xs, runparams);
671 closeLabelTag(xs, style);
673 // Usual cases: maybe there is something specified at the layout level. Highly unlikely, though.
674 docstring const lbl = par->params().labelString();
677 openLabelTag(xs, style);
679 closeLabelTag(xs, style);
684 // Open the item (after the wrapper and the label).
685 openTag(xs, style.docbookitemtag(), style.docbookitemattr(), style.docbookitemtagtype());
687 // Generate the content of the item.
688 if (sep < par->size()) {
689 auto pars = par->simpleDocBookOnePar(buf, runparams,
690 text.outerFont(std::distance(text.paragraphs().begin(), par)), sep);
691 for (auto &p : pars) {
692 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
693 par->layout().docbookiteminnertagtype());
694 xs << XMLStream::ESCAPE_NONE << p;
695 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
698 // DocBook doesn't like emptiness.
699 compTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
700 par->layout().docbookiteminnertagtype());
703 // If the next item is deeper, it must go entirely within this item (do it recursively).
704 // By construction, with findEndOfEnvironment, depth can only stay constant or increase, never decrease.
705 depth_type currentDepth = par->getDepth();
707 while (par != envend && par->getDepth() != currentDepth)
708 par = makeAny(text, buf, xs, runparams, par);
709 // Usually, this loop only makes one iteration, except in complex scenarios, like an item with a paragraph,
710 // a list, and another paragraph; or an item with two types of list (itemise then enumerate, for instance).
713 closeTag(xs, style.docbookitemtag(), style.docbookitemtagtype());
714 closeTag(xs, style.docbookitemwrappertag(), style.docbookitemwrappertagtype());
717 // Close this environment in exactly the same way as it was opened.
718 closeTag(xs, envstyle.docbooktag(), envstyle.docbooktagtype());
719 closeTag(xs, envstyle.docbookwrappertag(), envstyle.docbookwrappertagtype());
729 OutputParams const & runparams,
730 ParagraphList::const_iterator const & par)
732 // Unlike XHTML, no need for labels, as they are handled by DocBook tags.
733 auto const begin = text.paragraphs().begin();
734 auto const end = text.paragraphs().end();
738 // Generate this command.
739 auto prevpar = text.paragraphs().getParagraphBefore(par);
740 openParTag(xs, &*par, prevpar);
742 auto pars = par->simpleDocBookOnePar(buf, runparams,text.outerFont(distance(begin, par)));
743 for (auto & parXML : pars)
744 // TODO: decide what to do with openParTag/closeParTag in new lines.
745 xs << XMLStream::ESCAPE_NONE << parXML;
747 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
751 ParagraphList::const_iterator makeAny(Text const &text,
754 OutputParams const &runparams,
755 ParagraphList::const_iterator par)
757 switch (par->layout().latextype) {
759 makeCommand(text, buf, xs, runparams, par);
761 case LATEX_ENVIRONMENT:
762 makeEnvironment(text, buf, xs, runparams, par);
764 case LATEX_LIST_ENVIRONMENT:
765 case LATEX_ITEM_ENVIRONMENT:
766 // Only case when makeAny() might consume more than one paragraph.
767 return makeListEnvironment(text, buf, xs, runparams, par);
768 case LATEX_PARAGRAPH:
769 makeParagraph(text, buf, xs, runparams, par);
771 case LATEX_BIB_ENVIRONMENT:
772 makeBibliography(text, buf, xs, runparams, par);
780 bool isLayoutSectioning(Layout const & lay)
782 return lay.category() == from_utf8("Sectioning");
786 using DocBookDocumentSectioning = tuple<bool, pit_type>;
789 struct DocBookInfoTag
791 const set<pit_type> shouldBeInInfo;
792 const set<pit_type> mustBeInInfo; // With the notable exception of the abstract!
793 const set<pit_type> abstract;
794 const bool abstractLayout;
798 DocBookInfoTag(const set<pit_type> & shouldBeInInfo, const set<pit_type> & mustBeInInfo,
799 const set<pit_type> & abstract, bool abstractLayout, pit_type bpit, pit_type epit) :
800 shouldBeInInfo(shouldBeInInfo), mustBeInInfo(mustBeInInfo), abstract(abstract),
801 abstractLayout(abstractLayout), bpit(bpit), epit(epit) {}
805 DocBookDocumentSectioning hasDocumentSectioning(ParagraphList const ¶graphs, pit_type bpit, pit_type const epit) {
806 bool documentHasSections = false;
808 while (bpit < epit) {
809 Layout const &style = paragraphs[bpit].layout();
810 documentHasSections |= isLayoutSectioning(style);
812 if (documentHasSections)
816 // Paragraphs before the first section: [ runparams.par_begin ; eppit )
818 return make_tuple(documentHasSections, bpit);
822 bool hasOnlyNotes(Paragraph const & par)
824 // Precondition: the paragraph is not empty. Otherwise, the function will always return true...
825 for (int i = 0; i < par.size(); ++i)
826 // If you find something that is not an inset (like actual text) or an inset that is not a note,
828 if (!par.isInset(i) || par.getInset(i)->lyxCode() != NOTE_CODE)
834 DocBookInfoTag getParagraphsWithInfo(ParagraphList const ¶graphs,
835 pit_type bpit, pit_type const epit,
836 // Typically, bpit is the beginning of the document and epit the end *or* the first section.
837 bool documentHasSections) {
838 set<pit_type> shouldBeInInfo;
839 set<pit_type> mustBeInInfo;
840 set<pit_type> abstractWithLayout;
841 set<pit_type> abstractNoLayout;
843 // Find the first non empty paragraph by mutating bpit.
844 while (bpit < epit) {
845 Paragraph const &par = paragraphs[bpit];
846 if (par.empty() || hasOnlyNotes(par))
852 // Traverse everything that might belong to <info>.
853 bool hasAbstractLayout = false;
854 pit_type cpit = bpit;
855 for (; cpit < epit; ++cpit) {
856 // Skip paragraphs that don't generate anything in DocBook.
857 Paragraph const & par = paragraphs[cpit];
858 if (hasOnlyNotes(par))
861 // There should never be any section here. (Just a sanity check: if this fails, this function could end up
862 // processing the whole document.)
863 if (isLayoutSectioning(par.layout())) {
864 LYXERR0("Assertion failed: section found in potential <info> paragraphs.");
868 // If this is marked as an abstract by the layout, put it in the right set.
869 if (par.layout().docbookabstract()) {
870 hasAbstractLayout = true;
871 abstractWithLayout.emplace(cpit);
875 // Based on layout information, store this paragraph in one set: should be in <info>, must be,
876 // or abstract (either because of layout or of position).
877 Layout const &style = par.layout();
879 if (style.docbookininfo() == "always")
880 mustBeInInfo.emplace(cpit);
881 else if (style.docbookininfo() == "maybe")
882 shouldBeInInfo.emplace(cpit);
883 else if (documentHasSections && !hasAbstractLayout)
884 abstractNoLayout.emplace(cpit);
885 else // This should definitely not be in <info>.
888 // Now, cpit points to the first paragraph that no more has things that could go in <info>.
889 // bpit is the beginning of the <info> part.
891 return DocBookInfoTag(shouldBeInInfo, mustBeInInfo,
892 hasAbstractLayout ? abstractWithLayout : abstractNoLayout,
893 hasAbstractLayout, bpit, cpit);
896 } // end anonymous namespace
899 xml::FontTag docbookStartFontTag(xml::FontTypes type)
901 return xml::FontTag(from_utf8(fontToDocBookTag(type)), from_utf8(fontToAttribute(type)), type);
905 xml::EndFontTag docbookEndFontTag(xml::FontTypes type)
907 return xml::EndFontTag(from_utf8(fontToDocBookTag(type)), type);
911 void outputDocBookInfo(
915 OutputParams const & runparams,
916 ParagraphList const & paragraphs,
917 DocBookInfoTag const & info)
919 // Perform an additional check on the abstract. Sometimes, there are many paragraphs that should go
920 // into the abstract, but none generates actual content. Thus, first generate to a temporary stream,
921 // then only create the <abstract> tag if these paragraphs generate some content.
922 // This check must be performed *before* a decision on whether or not to output <info> is made.
923 bool hasAbstract = !info.abstract.empty();
926 // Generate the abstract XML into a string before further checks.
927 odocstringstream os2;
929 for (auto const & p : info.abstract)
930 makeAny(text, buf, xs2, runparams, paragraphs.iterator_at(p));
932 // Actually output the abstract if there is something to do. Don't count line feeds or spaces in this,
933 // even though they must be properly output if there is some abstract.
934 abstract = os2.str();
935 docstring cleaned = abstract;
936 cleaned.erase(std::remove_if(cleaned.begin(), cleaned.end(), ::isspace), cleaned.end());
938 // Nothing? Then there is no abstract!
943 // The abstract must go in <info>. Otherwise, decide whether to open <info> based on the layouts.
944 bool needInfo = !info.mustBeInInfo.empty() || hasAbstract;
946 // Start the <info> tag if required.
948 xs.startDivision(false);
949 xs << xml::StartTag("info");
953 // Output the elements that should go in <info>, before and after the abstract.
954 for (auto pit : info.shouldBeInInfo) // Typically, the title: these elements are so important and ubiquitous
955 // that mandating a wrapper like <info> would repel users. Thus, generate them first.
956 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
957 for (auto pit : info.mustBeInInfo)
958 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
960 // Always output the abstract as the last item of the <info>, as it requires special treatment (especially if
961 // it contains several paragraphs that are empty).
963 if (info.abstractLayout) {
964 xs << XMLStream::ESCAPE_NONE << abstract;
967 string tag = paragraphs[*info.abstract.begin()].layout().docbookforceabstracttag();
971 if (!xs.isLastTagCR())
974 xs << xml::StartTag(tag);
976 xs << XMLStream::ESCAPE_NONE << abstract;
977 xs << xml::EndTag(tag);
982 // End the <info> tag if it was started.
984 if (!xs.isLastTagCR())
987 xs << xml::EndTag("info");
994 void docbookSimpleAllParagraphs(
998 OutputParams const & runparams)
1000 // Handle the given text, supposing it has no sections (i.e. a "simple" text). The input may vary in length
1001 // between a single paragraph to a whole document.
1002 pit_type const bpit = runparams.par_begin;
1003 pit_type const epit = runparams.par_end;
1004 ParagraphList const ¶graphs = text.paragraphs();
1006 // First, the <info> tag.
1007 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, epit, false);
1008 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
1010 // Then, the content. It starts where the <info> ends.
1011 auto par = paragraphs.iterator_at(info.epit);
1012 auto end = paragraphs.iterator_at(epit);
1013 while (par != end) {
1014 if (!hasOnlyNotes(*par))
1015 par = makeAny(text, buf, xs, runparams, par);
1022 void docbookParagraphs(Text const &text,
1025 OutputParams const &runparams) {
1026 ParagraphList const ¶graphs = text.paragraphs();
1027 if (runparams.par_begin == runparams.par_end) {
1028 runparams.par_begin = 0;
1029 runparams.par_end = paragraphs.size();
1031 pit_type bpit = runparams.par_begin;
1032 pit_type const epit = runparams.par_end;
1033 LASSERT(bpit < epit,
1035 xs << XMLStream::ESCAPE_NONE << "<!-- DocBook output error! -->\n";
1039 std::stack<std::pair<int, string>> headerLevels; // Used to determine when to open/close sections: store the depth
1040 // of the section and the tag that was used to open it.
1042 // Detect whether the document contains sections. If there are no sections, treatment is largely simplified.
1043 // In particular, there can't be an abstract, unless it is manually marked.
1044 bool documentHasSections;
1046 tie(documentHasSections, eppit) = hasDocumentSectioning(paragraphs, bpit, epit);
1048 // Deal with "simple" documents, i.e. those without sections.
1049 if (!documentHasSections) {
1050 docbookSimpleAllParagraphs(text, buf, xs, runparams);
1054 // Output the first <info> tag (or just the title).
1055 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, eppit, true);
1056 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
1059 // Then, iterate through the paragraphs of this document.
1060 bool currentlyInAppendix = false;
1062 auto par = text.paragraphs().iterator_at(bpit);
1063 auto end = text.paragraphs().iterator_at(epit);
1064 while (par != end) {
1065 OutputParams ourparams = runparams;
1067 if (par->params().startOfAppendix())
1068 currentlyInAppendix = true;
1069 if (hasOnlyNotes(*par)) {
1074 Layout const &style = par->layout();
1076 // Think about adding <section> and/or </section>s.
1077 if (isLayoutSectioning(style)) {
1078 int level = style.toclevel;
1080 // Need to close a previous section if it has the same level or a higher one (close <section> if opening a
1081 // <h2> after a <h2>, <h3>, <h4>, <h5> or <h6>). More examples:
1082 // - current: h2; back: h1; do not close any <section>
1083 // - current: h1; back: h2; close two <section> (first the <h2>, then the <h1>, so a new <h1> can come)
1084 while (!headerLevels.empty() && level <= headerLevels.top().first) {
1085 // Output the tag only if it corresponds to a legit section.
1086 int stackLevel = headerLevels.top().first;
1087 if (stackLevel != Layout::NOT_IN_TOC) {
1088 xs << xml::EndTag(headerLevels.top().second);
1094 // Open the new section: first push it onto the stack, then output it in DocBook.
1095 string sectionTag = (currentlyInAppendix && style.docbooksectiontag() == "chapter") ?
1096 "appendix" : style.docbooksectiontag();
1097 headerLevels.push(std::make_pair(level, sectionTag));
1099 // Some sectioning-like elements should not be output (such as FrontMatter).
1100 if (level != Layout::NOT_IN_TOC) {
1101 // Look for a label in the title, i.e. a InsetLabel as a child.
1102 docstring id = docstring();
1103 for (pos_type i = 0; i < par->size(); ++i) {
1104 Inset const *inset = par->getInset(i);
1106 if (auto label = dynamic_cast<InsetLabel const *>(inset)) {
1107 // Generate the attributes for the section if need be.
1108 id += "xml:id=\"" + xml::cleanID(label->screenLabel()) + "\"";
1110 // Don't output the ID as a DocBook <anchor>.
1111 ourparams.docbook_anchors_to_ignore.emplace(label->screenLabel());
1113 // Cannot have multiple IDs per tag.
1119 // Write the open tag for this section.
1123 xs << xml::StartTag(sectionTag, attrs);
1128 // Close all sections before the bibliography.
1129 // 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)?
1130 auto insetsLength = distance(par->insetList().begin(), par->insetList().end());
1131 if (insetsLength > 0) {
1132 Inset const *firstInset = par->getInset(0);
1133 if (firstInset && (firstInset->lyxCode() == BIBITEM_CODE || firstInset->lyxCode() == BIBTEX_CODE)) {
1134 while (!headerLevels.empty()) {
1135 int level = headerLevels.top().first;
1136 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1139 // Output the tag only if it corresponds to a legit section.
1140 if (level != Layout::NOT_IN_TOC) {
1141 xs << XMLStream::ESCAPE_NONE << tag;
1148 // Generate this paragraph.
1149 par = makeAny(text, buf, xs, ourparams, par);
1151 // Some special sections may require abstracts (mostly parts, in books).
1152 // 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.
1153 if (isLayoutSectioning(style) && style.docbookforceabstracttag() != "NONE") {
1154 // This abstract may be found between the next paragraph and the next title.
1155 pit_type cpit = std::distance(text.paragraphs().begin(), par);
1156 pit_type ppit = std::get<1>(hasDocumentSectioning(paragraphs, cpit, epit));
1158 // Generate this abstract (this code corresponds to parts of outputDocBookInfo).
1159 DocBookInfoTag secInfo = getParagraphsWithInfo(paragraphs, cpit, ppit, true);
1161 if (!secInfo.abstract.empty()) {
1162 xs << xml::StartTag(style.docbookforceabstracttag());
1164 for (auto const &p : secInfo.abstract)
1165 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(p));
1166 xs << xml::EndTag(style.docbookforceabstracttag());
1170 // Skip all the text that just has been generated.
1171 par = paragraphs.iterator_at(ppit);
1175 // If need be, close <section>s, but only at the end of the document (otherwise, dealt with at the beginning
1177 while (!headerLevels.empty() && headerLevels.top().first > Layout::NOT_IN_TOC) {
1178 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1180 xs << XMLStream::ESCAPE_NONE << tag;