2 * \file output_docbook.cpp
3 * This file is part of LyX, the document processor.
4 * Licence details can be found in the file COPYING.
6 * \author Lars Gullik Bjønnes
9 * Full author contact details are available in file CREDITS.
14 #include "output_docbook.h"
17 #include "buffer_funcs.h"
18 #include "BufferParams.h"
20 #include "InsetList.h"
21 #include "Paragraph.h"
22 #include "ParagraphList.h"
23 #include "ParagraphParameters.h"
26 #include "TextClass.h"
28 #include "insets/InsetBibtex.h"
29 #include "insets/InsetBibitem.h"
30 #include "insets/InsetLabel.h"
31 #include "mathed/InsetMath.h"
32 #include "insets/InsetNote.h"
34 #include "support/debug.h"
35 #include "support/lassert.h"
36 #include "support/textutils.h"
44 using namespace lyx::support;
50 std::string fontToDocBookTag(xml::FontTypes type)
53 case xml::FontTypes::FT_EMPH:
54 case xml::FontTypes::FT_BOLD:
56 case xml::FontTypes::FT_NOUN:
58 case xml::FontTypes::FT_UBAR:
59 case xml::FontTypes::FT_WAVE:
60 case xml::FontTypes::FT_DBAR:
61 case xml::FontTypes::FT_SOUT:
62 case xml::FontTypes::FT_XOUT:
63 case xml::FontTypes::FT_ITALIC:
64 case xml::FontTypes::FT_UPRIGHT:
65 case xml::FontTypes::FT_SLANTED:
66 case xml::FontTypes::FT_SMALLCAPS:
67 case xml::FontTypes::FT_ROMAN:
68 case xml::FontTypes::FT_SANS:
70 case xml::FontTypes::FT_TYPE:
72 case xml::FontTypes::FT_SIZE_TINY:
73 case xml::FontTypes::FT_SIZE_SCRIPT:
74 case xml::FontTypes::FT_SIZE_FOOTNOTE:
75 case xml::FontTypes::FT_SIZE_SMALL:
76 case xml::FontTypes::FT_SIZE_NORMAL:
77 case xml::FontTypes::FT_SIZE_LARGE:
78 case xml::FontTypes::FT_SIZE_LARGER:
79 case xml::FontTypes::FT_SIZE_LARGEST:
80 case xml::FontTypes::FT_SIZE_HUGE:
81 case xml::FontTypes::FT_SIZE_HUGER:
82 case xml::FontTypes::FT_SIZE_INCREASE:
83 case xml::FontTypes::FT_SIZE_DECREASE:
91 string fontToRole(xml::FontTypes type)
93 // Specific fonts are achieved with roles. The only common ones are "" for basic emphasis,
94 // and "bold"/"strong" for bold. With some specific options, other roles are copied into
95 // HTML output (via the DocBook XSLT sheets); otherwise, if not recognised, they are just ignored.
96 // Hence, it is not a problem to have many roles by default here.
97 // See https://www.sourceware.org/ml/docbook/2003-05/msg00269.html
99 case xml::FontTypes::FT_ITALIC:
100 case xml::FontTypes::FT_EMPH:
102 case xml::FontTypes::FT_BOLD:
104 case xml::FontTypes::FT_NOUN: // Outputs a <person>
105 case xml::FontTypes::FT_TYPE: // Outputs a <code>
107 case xml::FontTypes::FT_UBAR:
110 // All other roles are non-standard for DocBook.
112 case xml::FontTypes::FT_WAVE:
114 case xml::FontTypes::FT_DBAR:
116 case xml::FontTypes::FT_SOUT:
118 case xml::FontTypes::FT_XOUT:
120 case xml::FontTypes::FT_UPRIGHT:
122 case xml::FontTypes::FT_SLANTED:
124 case xml::FontTypes::FT_SMALLCAPS:
126 case xml::FontTypes::FT_ROMAN:
128 case xml::FontTypes::FT_SANS:
130 case xml::FontTypes::FT_SIZE_TINY:
132 case xml::FontTypes::FT_SIZE_SCRIPT:
133 return "size_script";
134 case xml::FontTypes::FT_SIZE_FOOTNOTE:
135 return "size_footnote";
136 case xml::FontTypes::FT_SIZE_SMALL:
138 case xml::FontTypes::FT_SIZE_NORMAL:
139 return "size_normal";
140 case xml::FontTypes::FT_SIZE_LARGE:
142 case xml::FontTypes::FT_SIZE_LARGER:
143 return "size_larger";
144 case xml::FontTypes::FT_SIZE_LARGEST:
145 return "size_largest";
146 case xml::FontTypes::FT_SIZE_HUGE:
148 case xml::FontTypes::FT_SIZE_HUGER:
150 case xml::FontTypes::FT_SIZE_INCREASE:
151 return "size_increase";
152 case xml::FontTypes::FT_SIZE_DECREASE:
153 return "size_decrease";
160 string fontToAttribute(xml::FontTypes type) {
161 // If there is a role (i.e. nonstandard use of a tag), output the attribute. Otherwise, the sheer tag is sufficient
163 string role = fontToRole(type);
165 return "role='" + role + "'";
171 // Higher-level convenience functions.
173 void openParTag(XMLStream & xs, const Paragraph * par, const Paragraph * prevpar, const OutputParams & runparams)
178 // If the previous paragraph is empty, don't consider it when opening wrappers.
179 if (prevpar && prevpar->empty() && !prevpar->allowEmpty())
182 // When should the wrapper be opened here? Only if the previous paragraph has the SAME wrapper tag
183 // (usually, they won't have the same layout) and the CURRENT one allows merging.
184 // The main use case is author information in several paragraphs: if the name of the author is the
185 // first paragraph of an author, then merging with the previous tag does not make sense. Say the
186 // next paragraph is the affiliation, then it should be output in the same <author> tag (different
187 // layout, same wrapper tag).
188 Layout const & lay = par->layout();
189 bool openWrapper = lay.docbookwrappertag() != "NONE" && !runparams.docbook_ignore_wrapper;
191 if (prevpar != nullptr && !runparams.docbook_ignore_wrapper) {
192 Layout const & prevlay = prevpar->layout();
193 if (prevlay.docbookwrappertag() != "NONE") {
194 if (prevlay.docbookwrappertag() == lay.docbookwrappertag() &&
195 prevlay.docbookwrapperattr() == lay.docbookwrapperattr())
196 openWrapper = !lay.docbookwrappermergewithprevious();
204 xml::openTag(xs, lay.docbookwrappertag(), lay.docbookwrapperattr(), lay.docbookwrappertagtype());
206 const string & tag = lay.docbooktag();
208 auto xmltag = xml::ParTag(tag, lay.docbookattr());
209 if (!xs.isTagOpen(xmltag, 1)) { // Don't nest a paragraph directly in a paragraph.
210 // TODO: required or not?
211 // TODO: avoid creating a ParTag object just for this query...
212 xml::openTag(xs, lay.docbooktag(), lay.docbookattr(), lay.docbooktagtype());
213 xml::openTag(xs, lay.docbookinnertag(), lay.docbookinnerattr(), lay.docbookinnertagtype());
217 xml::openTag(xs, lay.docbookitemwrappertag(), lay.docbookitemwrapperattr(), lay.docbookitemwrappertagtype());
218 xml::openTag(xs, lay.docbookitemtag(), lay.docbookitemattr(), lay.docbookitemtagtype());
219 xml::openTag(xs, lay.docbookiteminnertag(), lay.docbookiteminnerattr(), lay.docbookiteminnertagtype());
223 void closeParTag(XMLStream & xs, Paragraph const * par, Paragraph const * nextpar, const OutputParams & runparams)
228 // If the next paragraph is empty, don't consider it when closing wrappers.
229 if (nextpar && nextpar->empty() && !nextpar->allowEmpty())
232 // See comment in openParTag.
233 Layout const & lay = par->layout();
234 bool closeWrapper = lay.docbookwrappertag() != "NONE" && !runparams.docbook_ignore_wrapper;
236 if (nextpar != nullptr && !runparams.docbook_ignore_wrapper) {
237 Layout const & nextlay = nextpar->layout();
238 if (nextlay.docbookwrappertag() != "NONE") {
239 if (nextlay.docbookwrappertag() == lay.docbookwrappertag() &&
240 nextlay.docbookwrapperattr() == lay.docbookwrapperattr())
241 closeWrapper = !nextlay.docbookwrappermergewithprevious();
248 xml::closeTag(xs, lay.docbookiteminnertag(), lay.docbookiteminnertagtype());
249 xml::closeTag(xs, lay.docbookitemtag(), lay.docbookitemtagtype());
250 xml::closeTag(xs, lay.docbookitemwrappertag(), lay.docbookitemwrappertagtype());
251 xml::closeTag(xs, lay.docbookinnertag(), lay.docbookinnertagtype());
252 xml::closeTag(xs, lay.docbooktag(), lay.docbooktagtype());
254 xml::closeTag(xs, lay.docbookwrappertag(), lay.docbookwrappertagtype());
258 void makeBibliography(
262 OutputParams const & runparams,
263 ParagraphList::const_iterator const & par)
265 // If this is the first paragraph in a bibliography, open the bibliography tag.
266 auto const * pbegin_before = text.paragraphs().getParagraphBefore(par);
267 if (pbegin_before == nullptr || (pbegin_before && pbegin_before->layout().latextype != LATEX_BIB_ENVIRONMENT)) {
268 xs << xml::StartTag("bibliography");
272 // Start the precooked bibliography entry. This is very much like opening a paragraph tag.
273 // Don't forget the citation ID!
275 for (auto i = 0; i < par->size(); ++i) {
276 Inset const *ip = par->getInset(i);
279 if (const auto * bibitem = dynamic_cast<const InsetBibitem*>(ip)) {
280 auto id = xml::cleanID(bibitem->getParam("key"));
281 attr = from_utf8("xml:id='") + id + from_utf8("'");
285 xs << xml::StartTag(from_utf8("bibliomixed"), attr);
287 // Generate the entry. Concatenate the different parts of the paragraph if any.
288 auto const begin = text.paragraphs().begin();
289 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(std::distance(begin, par)), 0);
290 for (auto & parXML : pars)
291 xs << XMLStream::ESCAPE_NONE << parXML;
293 // End the precooked bibliography entry.
294 xs << xml::EndTag("bibliomixed");
297 // If this is the last paragraph in a bibliography, close the bibliography tag.
298 auto const end = text.paragraphs().end();
301 bool endBibliography = nextpar == end || nextpar->layout().latextype != LATEX_BIB_ENVIRONMENT;
303 if (endBibliography) {
304 xs << xml::EndTag("bibliography");
314 OutputParams const & runparams,
315 ParagraphList::const_iterator const & par)
318 auto const begin = text.paragraphs().begin();
319 auto const end = text.paragraphs().end();
320 auto prevpar = text.paragraphs().getParagraphBefore(par);
322 // We want to open the paragraph tag if:
323 // (i) the current layout permits multiple paragraphs
324 // (ii) we are either not already inside a paragraph (HTMLIsBlock) OR
325 // we are, but this is not the first paragraph
327 // But there is also a special case, and we first see whether we are in it.
328 // We do not want to open the paragraph tag if this paragraph contains
329 // only one item, and that item is "inline", i.e., not HTMLIsBlock (such
330 // as a branch). On the other hand, if that single item has a font change
331 // applied to it, then we still do need to open the paragraph.
333 // Obviously, this is very fragile. The main reason we need to do this is
334 // because of branches, e.g., a branch that contains an entire new section.
335 // We do not really want to wrap that whole thing in a <div>...</div>.
336 bool special_case = false;
337 Inset const *specinset = par->size() == 1 ? par->getInset(0) : nullptr;
338 if (specinset && !specinset->getLayout().htmlisblock()) { // TODO: Convert htmlisblock to a DocBook parameter?
339 Layout const &style = par->layout();
340 FontInfo const first_font = style.labeltype == LABEL_MANUAL ?
341 style.labelfont : style.font;
342 FontInfo const our_font =
343 par->getFont(buf.masterBuffer()->params(), 0,
344 text.outerFont(std::distance(begin, par))).fontInfo();
346 if (first_font == our_font)
350 auto nInsets = std::distance(par->insetList().begin(), par->insetList().end());
351 auto parSize = (size_t) par->size();
353 // Plain layouts must be ignored.
354 special_case |= buf.params().documentClass().isPlainLayout(par->layout()) && !runparams.docbook_force_pars;
355 // Equations do not deserve their own paragraph (DocBook allows them outside paragraphs).
356 // Exception: any case that generates an <inlineequation> must still get a paragraph to be valid.
357 special_case |= nInsets == parSize && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
358 return inset.inset && inset.inset->asInsetMath() && inset.inset->asInsetMath()->getType() != hullSimple;
361 // Things that should not get into their own paragraph. (Only valid for DocBook.)
362 static std::set<InsetCode> lyxCodeSpecialCases = {
365 BIBTEX_CODE, // Bibliographies cannot be in paragraphs. Bibitems should still be handled as paragraphs,
366 // though (see makeParagraphBibliography).
367 ERT_CODE, // ERTs are in comments, not paragraphs.
372 TOC_CODE, // To be ignored in DocBook, the processor afterwards should deal with ToCs.
373 NOTE_CODE // Notes do not produce any output.
375 auto isLyxCodeSpecialCase = [](InsetList::Element inset) {
376 return lyxCodeSpecialCases.find(inset.inset->lyxCode()) != lyxCodeSpecialCases.end();
378 special_case |= nInsets == parSize && std::all_of(par->insetList().begin(), par->insetList().end(), isLyxCodeSpecialCase);
380 // Flex elements (InsetLayout) have their own parameter to control the special case.
381 auto isFlexSpecialCase = [](InsetList::Element inset) {
382 if (inset.inset->lyxCode() != FLEX_CODE)
384 // Standard condition: check the parameter.
385 if (inset.inset->getLayout().docbooknotinpara())
388 // If the parameter is not set, maybe the flex inset only contains things that should match the standard
389 // condition. In this case, isLyxCodeSpecialCase must also check for bibitems...
390 auto isLyxCodeSpecialCase = [](InsetList::Element inset) {
391 return lyxCodeSpecialCases.find(inset.inset->lyxCode()) != lyxCodeSpecialCases.end() ||
392 inset.inset->lyxCode() == BIBITEM_CODE;
394 if (InsetText * text = inset.inset->asInsetText()) {
395 for (auto const & par : text->paragraphs()) {
396 auto nInsets = std::distance(par.insetList().begin(), par.insetList().end());
397 auto parSize = (size_t) par.size();
399 if (nInsets == 1 && par.insetList().begin()->inset->lyxCode() == BIBITEM_CODE)
401 if (nInsets != parSize)
403 if (!std::all_of(par.insetList().begin(), par.insetList().end(), isLyxCodeSpecialCase))
409 // No case matched: give up.
412 special_case |= nInsets == parSize && std::all_of(par->insetList().begin(), par->insetList().end(), isFlexSpecialCase);
414 // Open a paragraph if it is allowed, we are not already within a paragraph, and the insets in the paragraph do
415 // not forbid paragraphs (aka special cases).
416 bool const open_par = runparams.docbook_make_pars
417 && !runparams.docbook_in_par
420 // We want to issue the closing tag if either:
421 // (i) We opened it, and either docbook_in_par is false,
422 // or we're not in the last paragraph, anyway.
423 // (ii) We didn't open it and docbook_in_par is true,
424 // but we are in the first par, and there is a next par.
425 bool const close_par = open_par && !runparams.docbook_in_par;
427 // Determine if this paragraph has some real content. Things like new pages are not caught
428 // by Paragraph::empty(), even though they do not generate anything useful in DocBook.
429 // Thus, remove all spaces (including new lines: \r, \n) before checking for emptiness.
430 // std::all_of allows doing this check without having to copy the string.
431 // Open and close tags around each contained paragraph.
434 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(distance(begin, par)), 0, nextpar == end, special_case);
435 for (docstring const & parXML : pars) {
436 if (!xml::isNotOnlySpace(parXML))
440 openParTag(xs, &*par, prevpar, runparams);
442 xs << XMLStream::ESCAPE_NONE << parXML;
445 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr, runparams);
450 void makeEnvironment(Text const &text,
453 OutputParams const &runparams,
454 ParagraphList::const_iterator const & par)
457 auto const end = text.paragraphs().end();
461 // Special cases for listing-like environments provided in layouts. This is quite ad-hoc, but provides a useful
462 // default. This should not be used by too many environments (only LyX-Code right now).
463 // This would be much simpler if LyX-Code was implemented as InsetListings...
464 bool mimicListing = false;
465 bool ignoreFonts = false;
466 if (par->layout().docbooktag() == "programlisting") {
471 // Output the opening tag for this environment, but only if it has not been previously opened (condition
472 // implemented in openParTag).
473 auto prevpar = text.paragraphs().getParagraphBefore(par);
474 openParTag(xs, &*par, prevpar, runparams);
476 // Generate the contents of this environment. There is a special case if this is like some environment.
477 Layout const & style = par->layout();
478 if (style.latextype == LATEX_COMMAND) {
479 // Nothing to do (otherwise, infinite loops).
480 } else if (style.latextype == LATEX_ENVIRONMENT) {
481 // Generate the paragraph, if need be.
482 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(std::distance(text.paragraphs().begin(), par)), 0, false, ignoreFonts);
485 auto p = pars.begin();
486 while (p != pars.end()) {
487 xml::openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
488 par->layout().docbookiteminnertagtype());
489 xs << XMLStream::ESCAPE_NONE << *p;
490 xml::closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
493 // Insert a new line after each "paragraph" (i.e. line in the listing), except for the last one.
494 // Otherwise, there would one more new line in the output than in the LyX document.
499 for (auto const & p : pars) {
500 xml::openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
501 par->layout().docbookiteminnertagtype());
502 xs << XMLStream::ESCAPE_NONE << p;
503 xml::closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
507 makeAny(text, buf, xs, runparams, par);
510 // Close the environment.
511 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr, runparams);
515 ParagraphList::const_iterator findEndOfEnvironment(
516 ParagraphList::const_iterator const & pstart,
517 ParagraphList::const_iterator const & pend)
519 // Copy-paste from XHTML. Should be factored out at some point...
520 ParagraphList::const_iterator p = pstart;
521 Layout const & bstyle = p->layout();
522 size_t const depth = p->params().depth();
523 for (++p; p != pend; ++p) {
524 Layout const & style = p->layout();
525 // It shouldn't happen that e.g. a section command occurs inside
526 // a quotation environment, at a higher depth, but as of 6/2009,
527 // it can happen. We pretend that it's just at lowest depth.
528 if (style.latextype == LATEX_COMMAND)
531 // If depth is down, we're done
532 if (p->params().depth() < depth)
535 // If depth is up, we're not done
536 if (p->params().depth() > depth)
539 // FIXME I am not sure about the first check.
540 // Surely we *could* have different layouts that count as
541 // LATEX_PARAGRAPH, right?
542 if (style.latextype == LATEX_PARAGRAPH || style != bstyle)
549 ParagraphList::const_iterator makeListEnvironment(Text const &text,
552 OutputParams const &runparams,
553 ParagraphList::const_iterator const & begin)
557 auto const end = text.paragraphs().end();
558 auto const envend = findEndOfEnvironment(par, end);
560 // Output the opening tag for this environment.
561 Layout const & envstyle = par->layout();
562 xml::openTag(xs, envstyle.docbookwrappertag(), envstyle.docbookwrapperattr(), envstyle.docbookwrappertagtype());
563 xml::openTag(xs, envstyle.docbooktag(), envstyle.docbookattr(), envstyle.docbooktagtype());
565 // Handle the content of the list environment, item by item.
566 while (par != envend) {
567 // Skip this paragraph if it is both empty and the last one (otherwise, there may be deeper paragraphs after).
570 if (par->empty() && nextpar == envend)
573 // Open the item wrapper.
574 Layout const & style = par->layout();
575 xml::openTag(xs, style.docbookitemwrappertag(), style.docbookitemwrapperattr(),
576 style.docbookitemwrappertagtype());
578 // Generate the label, if need be. If it is taken from the text, sep != 0 and corresponds to the first
579 // character after the label.
581 if (style.labeltype != LABEL_NO_LABEL && style.docbookitemlabeltag() != "NONE") {
582 if (style.labeltype == LABEL_MANUAL) {
583 // Only variablelist gets here (or similar items defined as an extension in the layout).
584 xml::openTag(xs, style.docbookitemlabeltag(), style.docbookitemlabelattr(),
585 style.docbookitemlabeltagtype());
586 sep = 1 + par->firstWordDocBook(xs, runparams);
587 xml::closeTag(xs, style.docbookitemlabeltag(), style.docbookitemlabeltagtype());
589 // Usual cases: maybe there is something specified at the layout level. Highly unlikely, though.
590 docstring const lbl = par->params().labelString();
593 xml::openTag(xs, style.docbookitemlabeltag(), style.docbookitemlabelattr(),
594 style.docbookitemlabeltagtype());
596 xml::closeTag(xs, style.docbookitemlabeltag(), style.docbookitemlabeltagtype());
601 // Open the item (after the wrapper and the label).
602 xml::openTag(xs, style.docbookitemtag(), style.docbookitemattr(), style.docbookitemtagtype());
604 // Generate the content of the item.
605 if (sep < par->size()) {
606 auto pars = par->simpleDocBookOnePar(buf, runparams,
607 text.outerFont(std::distance(text.paragraphs().begin(), par)), sep);
608 for (auto &p : pars) {
609 xml::openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
610 par->layout().docbookiteminnertagtype());
611 xs << XMLStream::ESCAPE_NONE << p;
612 xml::closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
615 // DocBook doesn't like emptiness.
616 xml::compTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
617 par->layout().docbookiteminnertagtype());
620 // If the next item is deeper, it must go entirely within this item (do it recursively).
621 // By construction, with findEndOfEnvironment, depth can only stay constant or increase, never decrease.
622 depth_type currentDepth = par->getDepth();
624 while (par != envend && par->getDepth() != currentDepth)
625 par = makeAny(text, buf, xs, runparams, par);
626 // Usually, this loop only makes one iteration, except in complex scenarios, like an item with a paragraph,
627 // a list, and another paragraph; or an item with two types of list (itemise then enumerate, for instance).
630 xml::closeTag(xs, style.docbookitemtag(), style.docbookitemtagtype());
631 xml::closeTag(xs, style.docbookitemwrappertag(), style.docbookitemwrappertagtype());
634 // Close this environment in exactly the same way as it was opened.
635 xml::closeTag(xs, envstyle.docbooktag(), envstyle.docbooktagtype());
636 xml::closeTag(xs, envstyle.docbookwrappertag(), envstyle.docbookwrappertagtype());
646 OutputParams const & runparams,
647 ParagraphList::const_iterator const & par)
650 // Unlike XHTML, no need for labels, as they are handled by DocBook tags.
651 auto const begin = text.paragraphs().begin();
652 auto const end = text.paragraphs().end();
656 // Generate this command.
657 auto prevpar = text.paragraphs().getParagraphBefore(par);
658 openParTag(xs, &*par, prevpar, runparams);
660 auto pars = par->simpleDocBookOnePar(buf, runparams,text.outerFont(distance(begin, par)));
661 for (auto & parXML : pars)
662 // TODO: decide what to do with openParTag/closeParTag in new lines.
663 xs << XMLStream::ESCAPE_NONE << parXML;
665 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr, runparams);
669 bool isLayoutSectioning(Layout const & lay)
671 if (lay.docbooksection()) // Special case: some DocBook styles must be handled as sections.
673 else if (lay.category() == from_utf8("Sectioning") || lay.docbooktag() == "section") // Generic case.
674 return lay.toclevel != Layout::NOT_IN_TOC;
679 bool isLayoutSectioningOrSimilar(Layout const & lay)
681 return isLayoutSectioning(lay) || lay.docbooktag() == "bridgehead";
685 using DocBookDocumentSectioning = tuple<bool, pit_type>;
688 struct DocBookInfoTag
690 const set<pit_type> shouldBeInInfo;
691 const set<pit_type> mustBeInInfo; // With the notable exception of the abstract!
692 const set<pit_type> abstract;
693 const bool abstractLayout;
697 DocBookInfoTag(const set<pit_type> & shouldBeInInfo, const set<pit_type> & mustBeInInfo,
698 const set<pit_type> & abstract, bool abstractLayout, pit_type bpit, pit_type epit) :
699 shouldBeInInfo(shouldBeInInfo), mustBeInInfo(mustBeInInfo), abstract(abstract),
700 abstractLayout(abstractLayout), bpit(bpit), epit(epit) {}
704 DocBookDocumentSectioning hasDocumentSectioning(ParagraphList const ¶graphs, pit_type bpit, pit_type const epit) {
705 bool documentHasSections = false;
707 while (bpit < epit) {
708 Layout const &style = paragraphs[bpit].layout();
709 documentHasSections |= isLayoutSectioningOrSimilar(style);
711 if (documentHasSections)
715 // Paragraphs before the first section: [ runparams.par_begin ; eppit )
717 return make_tuple(documentHasSections, bpit);
721 bool hasOnlyNotes(Paragraph const & par)
723 // Precondition: the paragraph is not empty. Otherwise, the function will always return true...
724 for (int i = 0; i < par.size(); ++i)
725 // If you find something that is not an inset (like actual text) or an inset that is not a note,
727 if (!par.isInset(i) || par.getInset(i)->lyxCode() != NOTE_CODE)
730 // An empty paragraph may still require some output.
731 if (par.layout().docbooksection())
734 // There should be really no content here.
739 DocBookInfoTag getParagraphsWithInfo(ParagraphList const ¶graphs,
740 pit_type bpit, pit_type const epit,
741 // Typically, bpit is the beginning of the document and epit the end of the
742 // document *or* the first section.
743 bool documentHasSections,
744 bool detectUnlayoutedAbstract
745 // Whether paragraphs with no specific layout should be detected as abstracts.
746 // For inner sections, an abstract should only be detected if it has a specific
747 // layout. For others, anything that might look like an abstract should be sought.
749 set<pit_type> shouldBeInInfo;
750 set<pit_type> mustBeInInfo;
751 set<pit_type> abstractWithLayout;
752 set<pit_type> abstractNoLayout;
754 // Find the first non empty paragraph by mutating bpit.
755 while (bpit < epit) {
756 Paragraph const &par = paragraphs[bpit];
757 if (par.empty() || hasOnlyNotes(par))
763 // Traverse everything that might belong to <info>.
764 bool hasAbstractLayout = false;
765 static depth_type INVALID_DEPTH = 100000;
766 depth_type abstractDepth = INVALID_DEPTH;
767 pit_type cpit = bpit;
768 for (; cpit < epit; ++cpit) {
769 // Skip paragraphs that don't generate anything in DocBook.
770 Paragraph const & par = paragraphs[cpit];
771 Layout const &style = par.layout();
772 if (hasOnlyNotes(par))
775 // There should never be any section here, except for the first paragraph (a title can be part of <info>).
776 // (Just a sanity check: if this fails, this function could end up processing the whole document.)
777 if (cpit != bpit && isLayoutSectioningOrSimilar(par.layout())) {
778 LYXERR0("Assertion failed: section found in potential <info> paragraphs.");
782 // If this is marked as an abstract by the layout, put it in the right set.
783 if (style.docbookabstract()) {
784 hasAbstractLayout = true;
785 abstractDepth = par.getDepth();
786 abstractWithLayout.emplace(cpit);
790 // Deeper paragraphs following the abstract must still be considered as part of the abstract.
791 // For instance, this includes lists. There should not be any other kind of paragraph in between.
792 if (abstractDepth != INVALID_DEPTH && style.docbookininfo() == "never") {
793 if (par.getDepth() > abstractDepth) {
794 abstractWithLayout.emplace(cpit);
797 if (par.getDepth() == abstractDepth) {
798 // This is not an abstract paragraph and it should not either be considered as part
799 // of it. It breaks the rule that abstract paragraphs must follow each other.
800 abstractDepth = INVALID_DEPTH;
805 // Based on layout information, store this paragraph in one set: should be in <info>, must be,
806 // or abstract (either because of layout or of position).
807 if (style.docbookininfo() == "always")
808 mustBeInInfo.emplace(cpit);
809 else if (style.docbookininfo() == "maybe")
810 shouldBeInInfo.emplace(cpit);
811 else if (documentHasSections && !hasAbstractLayout && detectUnlayoutedAbstract &&
812 (style.docbooktag() == "NONE" || style.docbooktag() == "para") &&
813 style.docbookwrappertag() == "NONE")
814 // In this case, it is very likely that style.docbookininfo() == "never"! Be extra careful
815 // about anything that gets caught here. For instance, don't ake into account
816 abstractNoLayout.emplace(cpit);
817 else // This should definitely not be in <info>.
820 // Now, cpit points to the first paragraph that no more has things that could go in <info>.
821 // bpit is the beginning of the <info> part.
823 return DocBookInfoTag(shouldBeInInfo, mustBeInInfo,
824 hasAbstractLayout ? abstractWithLayout : abstractNoLayout,
825 hasAbstractLayout, bpit, cpit);
828 } // end anonymous namespace
831 std::set<const Inset *> gatherInfo(ParagraphList::const_iterator par)
833 // This function has a structure highly similar to makeAny and its friends. It's only made to be called on what
834 // should become the document's <abstract>.
835 std::set<const Inset *> values;
837 // If this kind of layout should be ignored, already leave.
838 if (par->layout().docbooktag() == "IGNORE")
841 // If this should go in info, mark it as such. Dive deep into the abstract, as it may hide many things that
842 // DocBook doesn't want to be inside the abstract.
843 for (pos_type i = 0; i < par->size(); ++i) {
844 if (par->getInset(i) && par->getInset(i)->asInsetText()) {
845 InsetText const *inset = par->getInset(i)->asInsetText();
847 if (inset->getLayout().docbookininfo() != "never") {
848 values.insert(inset);
850 auto subpar = inset->paragraphs().begin();
851 while (subpar != inset->paragraphs().end()) {
852 auto subinfos = gatherInfo(subpar);
853 for (auto & subinfo: subinfos)
854 values.insert(subinfo);
865 ParagraphList::const_iterator makeAny(Text const &text,
868 OutputParams const &runparams,
869 ParagraphList::const_iterator par)
871 bool ignoreParagraph = false;
873 // If this kind of layout should be ignored, already leave.
874 ignoreParagraph |= par->layout().docbooktag() == "IGNORE";
876 // For things that should go into <info>, check the variable rp.docbook_generate_info. This does not apply to the
878 bool isAbstract = par->layout().docbookabstract() || par->layout().docbooktag() == "abstract";
879 ignoreParagraph |= !isAbstract && par->layout().docbookininfo() != "never" && !runparams.docbook_generate_info;
881 // Switch on the type of paragraph to call the right handler.
882 if (!ignoreParagraph) {
883 switch (par->layout().latextype) {
885 makeCommand(text, buf, xs, runparams, par);
887 case LATEX_ENVIRONMENT:
888 makeEnvironment(text, buf, xs, runparams, par);
890 case LATEX_LIST_ENVIRONMENT:
891 case LATEX_ITEM_ENVIRONMENT:
892 // Only case when makeAny() might consume more than one paragraph.
893 return makeListEnvironment(text, buf, xs, runparams, par);
894 case LATEX_PARAGRAPH:
895 makeParagraph(text, buf, xs, runparams, par);
897 case LATEX_BIB_ENVIRONMENT:
898 makeBibliography(text, buf, xs, runparams, par);
903 // For cases that are not lists, the next paragraph to handle is the next one.
909 xml::FontTag docbookStartFontTag(xml::FontTypes type)
911 return xml::FontTag(from_utf8(fontToDocBookTag(type)), from_utf8(fontToAttribute(type)), type);
915 xml::EndFontTag docbookEndFontTag(xml::FontTypes type)
917 return xml::EndFontTag(from_utf8(fontToDocBookTag(type)), type);
921 void outputDocBookInfo(
925 OutputParams const & runparams,
926 ParagraphList const & paragraphs,
927 DocBookInfoTag const & info)
929 // Perform an additional check on the abstract. Sometimes, there are many paragraphs that should go
930 // into the abstract, but none generates actual content. Thus, first generate to a temporary stream,
931 // then only create the <abstract> tag if these paragraphs generate some content.
932 // This check must be performed *before* a decision on whether or not to output <info> is made.
933 bool hasAbstract = !info.abstract.empty();
935 set<const Inset *> infoInsets; // Paragraphs that should go into <info>, but are hidden in an <abstract>
936 // paragraph. (This happens for quite a few layouts, unfortunately.)
939 // Generate the abstract XML into a string before further checks.
940 // Usually, makeAny only generates one paragraph at a time. However, for the specific case of lists, it might
941 // generate more than one paragraph, as indicated in the return value.
942 odocstringstream os2;
946 rp.docbook_generate_info = false;
947 rp.docbook_ignore_wrapper = true;
949 set<pit_type> doneParas; // Paragraphs that have already been converted (mostly to deal with lists).
950 for (auto const & p : info.abstract) {
951 if (doneParas.find(p) == doneParas.end()) {
952 auto oldPar = paragraphs.iterator_at(p);
953 auto newPar = makeAny(text, buf, xs2, rp, oldPar);
955 // Find insets that should go outside the abstract.
956 auto subinfos = gatherInfo(oldPar);
957 for (auto & subinfo: subinfos)
958 infoInsets.insert(subinfo);
960 // Insert the indices of all the paragraphs that were just generated (typically, one).
961 // **Make the hypothesis that, when an abstract has a list, all its items are consecutive.**
962 // Otherwise, makeAny and makeListEnvironment would have to be adapted too.
964 while (oldPar != newPar) {
965 doneParas.emplace(id);
972 // Actually output the abstract if there is something to do. Don't count line feeds or spaces in this,
973 // even though they must be properly output if there is some abstract.
974 abstract = os2.str();
975 docstring cleaned = abstract;
976 cleaned.erase(std::remove_if(cleaned.begin(), cleaned.end(), lyx::isSpace), cleaned.end());
978 // Nothing? Then there is no abstract!
983 // The abstract must go in <info>. Otherwise, decide whether to open <info> based on the layouts.
984 bool needInfo = !info.mustBeInInfo.empty() || hasAbstract;
986 // Start the <info> tag if required.
988 xs.startDivision(false);
989 xs << xml::StartTag("info");
993 // Output the elements that should go in <info>.
994 // - First, the title.
995 for (auto pit : info.shouldBeInInfo) // Typically, the title: these elements are so important and ubiquitous
996 // that mandating a wrapper like <info> would repel users. Thus, generate them first.
997 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
998 // If there is no title, generate one (required for the document to be valid).
999 // This code is called for the main document, for table cells, etc., so be precise in this condition.
1000 if (text.isMainText() && info.shouldBeInInfo.empty() && !runparams.inInclude) {
1001 xs << xml::StartTag("title");
1002 xs << "Untitled Document";
1003 xs << xml::EndTag("title");
1007 // - Then, other metadata.
1008 for (auto pit : info.mustBeInInfo)
1009 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
1010 for (auto const * inset : infoInsets)
1011 inset->docbook(xs, runparams);
1013 // - Finally, always output the abstract as the last item of the <info>, as it requires special treatment
1014 // (especially if it contains several paragraphs that are empty).
1016 string tag = paragraphs[*info.abstract.begin()].layout().docbookforceabstracttag();
1020 if (!xs.isLastTagCR())
1023 xs << xml::StartTag(tag);
1025 xs << XMLStream::ESCAPE_NONE << abstract;
1026 xs << xml::EndTag(tag);
1030 // End the <info> tag if it was started.
1032 if (!xs.isLastTagCR())
1035 xs << xml::EndTag("info");
1042 void docbookSimpleAllParagraphs(
1046 OutputParams const & runparams)
1048 // Handle the given text, supposing it has no sections (i.e. a "simple" text). The input may vary in length
1049 // between a single paragraph to a whole document.
1050 pit_type const bpit = runparams.par_begin;
1051 pit_type const epit = runparams.par_end;
1052 ParagraphList const ¶graphs = text.paragraphs();
1054 // First, the <info> tag.
1055 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, epit, false, true);
1056 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
1058 // Then, the content. It starts where the <info> ends.
1059 auto par = paragraphs.iterator_at(info.epit);
1060 auto end = paragraphs.iterator_at(epit);
1061 while (par != end) {
1062 if (!hasOnlyNotes(*par))
1063 par = makeAny(text, buf, xs, runparams, par);
1070 void docbookParagraphs(Text const &text,
1073 OutputParams const &runparams) {
1074 ParagraphList const ¶graphs = text.paragraphs();
1075 if (runparams.par_begin == runparams.par_end) {
1076 runparams.par_begin = 0;
1077 runparams.par_end = paragraphs.size();
1079 pit_type bpit = runparams.par_begin;
1080 pit_type const epit = runparams.par_end;
1081 LASSERT(bpit < epit,
1083 xs << XMLStream::ESCAPE_NONE << "<!-- DocBook output error! -->\n";
1087 // Detect whether the document contains sections. If there are no sections, treatment is largely simplified.
1088 // In particular, there can't be an abstract, unless it is manually marked.
1089 bool documentHasSections;
1091 tie(documentHasSections, eppit) = hasDocumentSectioning(paragraphs, bpit, epit);
1093 // Deal with "simple" documents, i.e. those without sections.
1094 if (!documentHasSections) {
1095 docbookSimpleAllParagraphs(text, buf, xs, runparams);
1099 // Output the first <info> tag (or just the title).
1100 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, eppit, true, true);
1101 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
1104 std::stack<std::pair<int, string>> headerLevels; // Used to determine when to open/close sections: store the depth
1105 // of the section and the tag that was used to open it.
1107 // Then, iterate through the paragraphs of this document.
1108 auto par = text.paragraphs().iterator_at(bpit);
1109 auto end = text.paragraphs().iterator_at(epit);
1110 while (par != end) {
1111 // Skip paragraphs not producing any output.
1112 if (hasOnlyNotes(*par)) {
1117 OutputParams ourparams = runparams;
1118 Layout const &style = par->layout();
1120 // Think about adding <section> and/or </section>s.
1121 if (isLayoutSectioning(style) || par->params().startOfAppendix()) {
1122 int level = style.toclevel;
1124 // Need to close a previous section if it has the same level or a higher one (close <section> if opening a
1125 // <h2> after a <h2>, <h3>, <h4>, <h5> or <h6>). More examples:
1126 // - current: h2; back: h1; do not close any <section>
1127 // - current: h1; back: h2; close two <section> (first the <h2>, then the <h1>, so a new <h1> can come)
1128 // Some layouts require that Layout::NOT_IN_TOC sections still cause closing of previous sections. This is
1129 // mostly to ensure that the section is positioned at a DocBook-compatible level (acknowledgements: cannot
1130 // be under a section!).
1131 while (!headerLevels.empty() && level <= headerLevels.top().first) {
1132 // Output the tag only if it corresponds to a legit section.
1133 int stackLevel = headerLevels.top().first;
1134 if (stackLevel != Layout::NOT_IN_TOC) {
1135 xs << xml::EndTag(headerLevels.top().second);
1141 // Open the new section: first push it onto the stack, then output it in DocBook.
1142 string sectionTag = (par->params().startOfAppendix()) ? "appendix" : style.docbooksectiontag();
1143 headerLevels.push(std::make_pair(level, sectionTag));
1145 // Some sectioning-like elements should not be output (such as FrontMatter).
1146 if (level != Layout::NOT_IN_TOC) {
1147 // Look for a label in the title, i.e. a InsetLabel as a child.
1148 docstring id = docstring();
1149 for (pos_type i = 0; i < par->size(); ++i) {
1150 Inset const *inset = par->getInset(i);
1152 if (auto label = dynamic_cast<InsetLabel const *>(inset)) {
1153 // Generate the attributes for the section if need be.
1154 id += "xml:id=\"" + xml::cleanID(label->screenLabel()) + "\"";
1156 // Don't output the ID as a DocBook <anchor>.
1157 ourparams.docbook_anchors_to_ignore.emplace(label->screenLabel());
1159 // Cannot have multiple IDs per tag. If there is another ID inset in the document, it will
1160 // be output as a DocBook anchor.
1166 // Write the open tag for this section.
1170 xs << xml::StartTag(sectionTag, attrs);
1175 // Close all sections before the bibliography.
1176 // 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)?
1177 if (!par->insetList().empty()) {
1178 Inset const *firstInset = par->getInset(0);
1179 if (firstInset && (firstInset->lyxCode() == BIBITEM_CODE || firstInset->lyxCode() == BIBTEX_CODE)) {
1180 while (!headerLevels.empty()) {
1181 // Don't close appendices before bibliographies.
1182 if (headerLevels.top().second == "appendix")
1185 // Pop the section from the stack.
1186 int level = headerLevels.top().first;
1187 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1190 // Output the tag only if it corresponds to a legit section, as the rest of the code.
1191 if (level != Layout::NOT_IN_TOC) {
1192 xs << XMLStream::ESCAPE_NONE << tag;
1199 // Generate the <info> tag if a section was just opened.
1200 // Some sections may require abstracts (mostly parts, in books: DocBookForceAbstractTag will not be NONE),
1201 // others can still have an abstract (it must be detected so that it can be output at the right place).
1202 // 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.
1203 if (isLayoutSectioning(style)) {
1204 // This abstract may be found between the next paragraph and the next title.
1205 pit_type cpit = std::distance(text.paragraphs().begin(), par);
1206 pit_type ppit = std::get<1>(hasDocumentSectioning(paragraphs, cpit + 1L, epit));
1208 // Generate this abstract (this code corresponds to parts of outputDocBookInfo).
1209 DocBookInfoTag secInfo = getParagraphsWithInfo(paragraphs, cpit, ppit, true,
1210 style.docbookforceabstracttag() != "NONE");
1212 if (!secInfo.mustBeInInfo.empty() || !secInfo.shouldBeInInfo.empty() || !secInfo.abstract.empty()) {
1213 // Generate the <info>, if required. If DocBookForceAbstractTag != NONE, this abstract will not be in
1214 // <info>, unlike other ("standard") abstracts.
1215 bool hasStandardAbstract = !secInfo.abstract.empty() && style.docbookforceabstracttag() == "NONE";
1216 bool needInfo = !secInfo.mustBeInInfo.empty() || hasStandardAbstract;
1219 xs.startDivision(false);
1220 xs << xml::StartTag("info");
1224 // Output the elements that should go in <info>, before and after the abstract.
1225 for (auto pit : secInfo.shouldBeInInfo) // Typically, the title: these elements are so important and ubiquitous
1226 // that mandating a wrapper like <info> would repel users. Thus, generate them first.
1227 makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(pit));
1228 for (auto pit : secInfo.mustBeInInfo)
1229 makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(pit));
1231 // Deal with the abstract in <info> if it is standard (i.e. its tag is <abstract>).
1232 if (!secInfo.abstract.empty() && hasStandardAbstract) {
1233 if (!secInfo.abstractLayout) {
1234 xs << xml::StartTag("abstract");
1238 for (auto const &p : secInfo.abstract)
1239 makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(p));
1241 if (!secInfo.abstractLayout) {
1242 xs << xml::EndTag("abstract");
1247 // End the <info> tag if it was started.
1249 if (!xs.isLastTagCR())
1252 xs << xml::EndTag("info");
1257 // Deal with the abstract outside <info> if it is not standard (i.e. its tag is layout-defined).
1258 if (!secInfo.abstract.empty() && !hasStandardAbstract) {
1259 // Assert: style.docbookforceabstracttag() != NONE.
1260 xs << xml::StartTag(style.docbookforceabstracttag());
1262 for (auto const &p : secInfo.abstract)
1263 makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(p));
1264 xs << xml::EndTag(style.docbookforceabstracttag());
1268 // Skip all the text that has just been generated.
1269 par = paragraphs.iterator_at(secInfo.epit);
1271 // No <info> tag to generate, proceed as for normal paragraphs.
1272 par = makeAny(text, buf, xs, ourparams, par);
1275 // Generate this paragraph, as it has nothing special.
1276 par = makeAny(text, buf, xs, ourparams, par);
1280 // If need be, close <section>s, but only at the end of the document (otherwise, dealt with at the beginning
1282 while (!headerLevels.empty() && headerLevels.top().first > Layout::NOT_IN_TOC) {
1283 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1285 xs << XMLStream::ESCAPE_NONE << tag;