2 * \file output_docbook.cpp
3 * This file is part of LyX, the document processor.
4 * Licence details can be found in the file COPYING.
6 * \author Lars Gullik Bjønnes
9 * Full author contact details are available in file CREDITS.
14 #include "output_docbook.h"
17 #include "buffer_funcs.h"
18 #include "BufferParams.h"
20 #include "InsetList.h"
21 #include "Paragraph.h"
22 #include "ParagraphList.h"
23 #include "ParagraphParameters.h"
26 #include "TextClass.h"
28 #include "insets/InsetBibtex.h"
29 #include "insets/InsetBibitem.h"
30 #include "insets/InsetLabel.h"
31 #include "mathed/InsetMath.h"
32 #include "insets/InsetNote.h"
34 #include "support/lassert.h"
35 #include "support/textutils.h"
43 using namespace lyx::support;
49 std::string fontToDocBookTag(xml::FontTypes type)
52 case xml::FontTypes::FT_EMPH:
53 case xml::FontTypes::FT_BOLD:
55 case xml::FontTypes::FT_NOUN:
57 case xml::FontTypes::FT_UBAR:
58 case xml::FontTypes::FT_WAVE:
59 case xml::FontTypes::FT_DBAR:
60 case xml::FontTypes::FT_SOUT:
61 case xml::FontTypes::FT_XOUT:
62 case xml::FontTypes::FT_ITALIC:
63 case xml::FontTypes::FT_UPRIGHT:
64 case xml::FontTypes::FT_SLANTED:
65 case xml::FontTypes::FT_SMALLCAPS:
66 case xml::FontTypes::FT_ROMAN:
67 case xml::FontTypes::FT_SANS:
69 case xml::FontTypes::FT_TYPE:
71 case xml::FontTypes::FT_SIZE_TINY:
72 case xml::FontTypes::FT_SIZE_SCRIPT:
73 case xml::FontTypes::FT_SIZE_FOOTNOTE:
74 case xml::FontTypes::FT_SIZE_SMALL:
75 case xml::FontTypes::FT_SIZE_NORMAL:
76 case xml::FontTypes::FT_SIZE_LARGE:
77 case xml::FontTypes::FT_SIZE_LARGER:
78 case xml::FontTypes::FT_SIZE_LARGEST:
79 case xml::FontTypes::FT_SIZE_HUGE:
80 case xml::FontTypes::FT_SIZE_HUGER:
81 case xml::FontTypes::FT_SIZE_INCREASE:
82 case xml::FontTypes::FT_SIZE_DECREASE:
90 string fontToRole(xml::FontTypes type)
92 // Specific fonts are achieved with roles. The only common ones are "" for basic emphasis,
93 // and "bold"/"strong" for bold. With some specific options, other roles are copied into
94 // HTML output (via the DocBook XSLT sheets); otherwise, if not recognised, they are just ignored.
95 // Hence, it is not a problem to have many roles by default here.
96 // See https://www.sourceware.org/ml/docbook/2003-05/msg00269.html
98 case xml::FontTypes::FT_ITALIC:
99 case xml::FontTypes::FT_EMPH:
101 case xml::FontTypes::FT_BOLD:
103 case xml::FontTypes::FT_NOUN: // Outputs a <person>
104 case xml::FontTypes::FT_TYPE: // Outputs a <code>
106 case xml::FontTypes::FT_UBAR:
109 // All other roles are non-standard for DocBook.
111 case xml::FontTypes::FT_WAVE:
113 case xml::FontTypes::FT_DBAR:
115 case xml::FontTypes::FT_SOUT:
117 case xml::FontTypes::FT_XOUT:
119 case xml::FontTypes::FT_UPRIGHT:
121 case xml::FontTypes::FT_SLANTED:
123 case xml::FontTypes::FT_SMALLCAPS:
125 case xml::FontTypes::FT_ROMAN:
127 case xml::FontTypes::FT_SANS:
129 case xml::FontTypes::FT_SIZE_TINY:
131 case xml::FontTypes::FT_SIZE_SCRIPT:
132 return "size_script";
133 case xml::FontTypes::FT_SIZE_FOOTNOTE:
134 return "size_footnote";
135 case xml::FontTypes::FT_SIZE_SMALL:
137 case xml::FontTypes::FT_SIZE_NORMAL:
138 return "size_normal";
139 case xml::FontTypes::FT_SIZE_LARGE:
141 case xml::FontTypes::FT_SIZE_LARGER:
142 return "size_larger";
143 case xml::FontTypes::FT_SIZE_LARGEST:
144 return "size_largest";
145 case xml::FontTypes::FT_SIZE_HUGE:
147 case xml::FontTypes::FT_SIZE_HUGER:
149 case xml::FontTypes::FT_SIZE_INCREASE:
150 return "size_increase";
151 case xml::FontTypes::FT_SIZE_DECREASE:
152 return "size_decrease";
159 string fontToAttribute(xml::FontTypes type) {
160 // If there is a role (i.e. nonstandard use of a tag), output the attribute. Otherwise, the sheer tag is sufficient
162 string role = fontToRole(type);
164 return "role='" + role + "'";
171 // Convenience functions to open and close tags. First, very low-level ones to ensure a consistent new-line behaviour.
175 // Contents of the block.
180 // <paratag>Contents of the paragraph.</paratag>
183 // Content before<inlinetag>Contents of the paragraph.</inlinetag>Content after
185 void openInlineTag(XMLStream & xs, const std::string & tag, const std::string & attr)
187 xs << xml::StartTag(tag, attr);
191 void closeInlineTag(XMLStream & xs, const std::string & tag)
193 xs << xml::EndTag(tag);
197 void openParTag(XMLStream & xs, const std::string & tag, const std::string & attr)
199 if (!xs.isLastTagCR())
201 xs << xml::StartTag(tag, attr);
205 void closeParTag(XMLStream & xs, const std::string & tag)
207 xs << xml::EndTag(tag);
212 void openBlockTag(XMLStream & xs, const std::string & tag, const std::string & attr)
214 if (!xs.isLastTagCR())
216 xs << xml::StartTag(tag, attr);
221 void closeBlockTag(XMLStream & xs, const std::string & tag)
223 if (!xs.isLastTagCR())
225 xs << xml::EndTag(tag);
230 void openTag(XMLStream & xs, const std::string & tag, const std::string & attr, const std::string & tagtype)
232 if (tag.empty() || tag == "NONE") // Common check to be performed elsewhere, if it was not here.
235 if (tag == "para" || tagtype == "paragraph") // Special case for <para>: always considered as a paragraph.
236 openParTag(xs, tag, attr);
237 else if (tagtype == "block")
238 openBlockTag(xs, tag, attr);
239 else if (tagtype == "inline")
240 openInlineTag(xs, tag, attr);
242 xs.writeError("Unrecognised tag type '" + tagtype + "' for '" + tag + " " + attr + "'");
246 void closeTag(XMLStream & xs, const std::string & tag, const std::string & tagtype)
248 if (tag.empty() || tag == "NONE")
251 if (tag == "para" || tagtype == "paragraph") // Special case for <para>: always considered as a paragraph.
252 closeParTag(xs, tag);
253 else if (tagtype == "block")
254 closeBlockTag(xs, tag);
255 else if (tagtype == "inline")
256 closeInlineTag(xs, tag);
258 xs.writeError("Unrecognised tag type '" + tagtype + "' for '" + tag + "'");
262 void compTag(XMLStream & xs, const std::string & tag, const std::string & attr, const std::string & tagtype)
264 if (tag.empty() || tag == "NONE")
267 // Special case for <para>: always considered as a paragraph.
268 if (tag == "para" || tagtype == "paragraph" || tagtype == "block") {
269 if (!xs.isLastTagCR())
271 xs << xml::CompTag(tag, attr);
273 } else if (tagtype == "inline") {
274 xs << xml::CompTag(tag, attr);
276 xs.writeError("Unrecognised tag type '" + tagtype + "' for '" + tag + "'");
281 // Higher-level convenience functions.
283 void openParTag(XMLStream & xs, const Paragraph * par, const Paragraph * prevpar)
288 // When should the wrapper be opened here? Only if the previous paragraph has the SAME wrapper tag
289 // (usually, they won't have the same layout) and the CURRENT one allows merging.
290 // The main use case is author information in several paragraphs: if the name of the author is the
291 // first paragraph of an author, then merging with the previous tag does not make sense. Say the
292 // next paragraph is the affiliation, then it should be output in the same <author> tag (different
293 // layout, same wrapper tag).
294 Layout const & lay = par->layout();
295 bool openWrapper = lay.docbookwrappertag() != "NONE";
296 if (prevpar != nullptr) {
297 Layout const & prevlay = prevpar->layout();
298 if (prevlay.docbookwrappertag() != "NONE") {
299 if (prevlay.docbookwrappertag() == lay.docbookwrappertag() &&
300 prevlay.docbookwrapperattr() == lay.docbookwrapperattr())
301 openWrapper = !lay.docbookwrappermergewithprevious();
309 openTag(xs, lay.docbookwrappertag(), lay.docbookwrapperattr(), lay.docbookwrappertagtype());
311 const string & tag = lay.docbooktag();
313 auto xmltag = xml::ParTag(tag, lay.docbookattr());
314 if (!xs.isTagOpen(xmltag, 1)) { // Don't nest a paragraph directly in a paragraph.
315 // TODO: required or not?
316 // TODO: avoid creating a ParTag object just for this query...
317 openTag(xs, lay.docbooktag(), lay.docbookattr(), lay.docbooktagtype());
318 openTag(xs, lay.docbookinnertag(), lay.docbookinnerattr(), lay.docbookinnertagtype());
322 openTag(xs, lay.docbookitemtag(), lay.docbookitemattr(), lay.docbookitemtagtype());
323 openTag(xs, lay.docbookiteminnertag(), lay.docbookiteminnerattr(), lay.docbookiteminnertagtype());
327 void closeParTag(XMLStream & xs, Paragraph const * par, Paragraph const * nextpar)
332 // See comment in openParTag.
333 Layout const & lay = par->layout();
334 bool closeWrapper = lay.docbookwrappertag() != "NONE";
335 if (nextpar != nullptr) {
336 Layout const & nextlay = nextpar->layout();
337 if (nextlay.docbookwrappertag() != "NONE") {
338 if (nextlay.docbookwrappertag() == lay.docbookwrappertag() &&
339 nextlay.docbookwrapperattr() == lay.docbookwrapperattr())
340 closeWrapper = !nextlay.docbookwrappermergewithprevious();
347 closeTag(xs, lay.docbookiteminnertag(), lay.docbookiteminnertagtype());
348 closeTag(xs, lay.docbookitemtag(), lay.docbookitemtagtype());
349 closeTag(xs, lay.docbookinnertag(), lay.docbookinnertagtype());
350 closeTag(xs, lay.docbooktag(), lay.docbooktagtype());
352 closeTag(xs, lay.docbookwrappertag(), lay.docbookwrappertagtype());
356 void makeBibliography(
360 OutputParams const & runparams,
361 ParagraphList::const_iterator const & par)
363 // If this is the first paragraph in a bibliography, open the bibliography tag.
364 auto const * pbegin_before = text.paragraphs().getParagraphBefore(par);
365 if (pbegin_before == nullptr || (pbegin_before && pbegin_before->layout().latextype != LATEX_BIB_ENVIRONMENT)) {
366 xs << xml::StartTag("bibliography");
370 // Start the precooked bibliography entry. This is very much like opening a paragraph tag.
371 // Don't forget the citation ID!
373 for (auto i = 0; i < par->size(); ++i) {
374 Inset const *ip = par->getInset(i);
377 if (const auto * bibitem = dynamic_cast<const InsetBibitem*>(ip)) {
378 auto id = xml::cleanID(bibitem->getParam("key"));
379 attr = from_utf8("xml:id='") + id + from_utf8("'");
383 xs << xml::StartTag(from_utf8("bibliomixed"), attr);
385 // Generate the entry. Concatenate the different parts of the paragraph if any.
386 auto const begin = text.paragraphs().begin();
387 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(std::distance(begin, par)), 0);
388 for (auto & parXML : pars)
389 xs << XMLStream::ESCAPE_NONE << parXML;
391 // End the precooked bibliography entry.
392 xs << xml::EndTag("bibliomixed");
395 // If this is the last paragraph in a bibliography, close the bibliography tag.
396 auto const end = text.paragraphs().end();
399 bool endBibliography = nextpar == end || nextpar->layout().latextype != LATEX_BIB_ENVIRONMENT;
401 if (endBibliography) {
402 xs << xml::EndTag("bibliography");
412 OutputParams const & runparams,
413 ParagraphList::const_iterator const & par)
415 // If this kind of layout should be ignored, already leave.
416 if (par->layout().docbooktag() == "IGNORE")
420 auto const begin = text.paragraphs().begin();
421 auto const end = text.paragraphs().end();
422 auto prevpar = text.paragraphs().getParagraphBefore(par);
424 // We want to open the paragraph tag if:
425 // (i) the current layout permits multiple paragraphs
426 // (ii) we are either not already inside a paragraph (HTMLIsBlock) OR
427 // we are, but this is not the first paragraph
429 // But there is also a special case, and we first see whether we are in it.
430 // We do not want to open the paragraph tag if this paragraph contains
431 // only one item, and that item is "inline", i.e., not HTMLIsBlock (such
432 // as a branch). On the other hand, if that single item has a font change
433 // applied to it, then we still do need to open the paragraph.
435 // Obviously, this is very fragile. The main reason we need to do this is
436 // because of branches, e.g., a branch that contains an entire new section.
437 // We do not really want to wrap that whole thing in a <div>...</div>.
438 bool special_case = false;
439 Inset const *specinset = par->size() == 1 ? par->getInset(0) : nullptr;
440 if (specinset && !specinset->getLayout().htmlisblock()) { // TODO: Convert htmlisblock to a DocBook parameter?
441 Layout const &style = par->layout();
442 FontInfo const first_font = style.labeltype == LABEL_MANUAL ?
443 style.labelfont : style.font;
444 FontInfo const our_font =
445 par->getFont(buf.masterBuffer()->params(), 0,
446 text.outerFont(std::distance(begin, par))).fontInfo();
448 if (first_font == our_font)
452 size_t nInsets = std::distance(par->insetList().begin(), par->insetList().end());
453 auto parSize = (size_t) par->size();
455 auto isLyxCodeToIgnore = [](InsetCode x) { return x == TOC_CODE; }; // If this LyX code does not produce any output,
456 // it can be safely ignored in the following checks: if this thing is present in the paragraph, it has no impact
457 // on the definition of the special case (i.e. whether or not a <para> tag should be output).
459 // TODO: if a paragraph *only* contains floats, listings, bibliographies, etc., should this be considered as a
460 // special case? If so, the code could be largely simplifies (all the calls to all_of, basically) and optimised
461 // at the compilation stage.
463 // Plain layouts must be ignored.
464 special_case |= buf.params().documentClass().isPlainLayout(par->layout()) && !runparams.docbook_force_pars;
465 // Equations do not deserve their own paragraph (DocBook allows them outside paragraphs).
466 // Exception: any case that generates an <inlineequation> must still get a paragraph to be valid.
467 special_case |= nInsets == parSize && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
468 return inset.inset && inset.inset->asInsetMath() && inset.inset->asInsetMath()->getType() != hullSimple;
470 // Tables do not deserve their own paragraphs (DocBook allows them outside paragraphs).
471 special_case |= nInsets == parSize && std::all_of(par->insetList().begin(), par->insetList().end(), [isLyxCodeToIgnore](InsetList::Element inset) {
472 return inset.inset->lyxCode() == TABULAR_CODE || isLyxCodeToIgnore(inset.inset->lyxCode());
474 // Floats cannot be in paragraphs.
475 special_case |= nInsets == parSize && std::all_of(par->insetList().begin(), par->insetList().end(), [isLyxCodeToIgnore](InsetList::Element inset) {
476 return inset.inset->lyxCode() == FLOAT_CODE || isLyxCodeToIgnore(inset.inset->lyxCode());
478 // Bibliographies cannot be in paragraphs. Bibitems should still be handled as paragraphs, though
479 // (see makeParagraphBibliography).
480 special_case |= nInsets == parSize && std::all_of(par->insetList().begin(), par->insetList().end(), [isLyxCodeToIgnore](InsetList::Element inset) {
481 return inset.inset->lyxCode() == BIBTEX_CODE || isLyxCodeToIgnore(inset.inset->lyxCode());
483 // ERTs are in comments, not paragraphs.
484 special_case |= nInsets == parSize && std::all_of(par->insetList().begin(), par->insetList().end(), [isLyxCodeToIgnore](InsetList::Element inset) {
485 return inset.inset->lyxCode() == ERT_CODE || isLyxCodeToIgnore(inset.inset->lyxCode());
487 // Listings should not get into their own paragraph.
488 special_case |= nInsets == parSize && std::all_of(par->insetList().begin(), par->insetList().end(), [isLyxCodeToIgnore](InsetList::Element inset) {
489 return inset.inset->lyxCode() == LISTINGS_CODE || isLyxCodeToIgnore(inset.inset->lyxCode());
491 // Boxes cannot get into their own paragraph.
492 special_case |= nInsets == parSize && std::all_of(par->insetList().begin(), par->insetList().end(), [isLyxCodeToIgnore](InsetList::Element inset) {
493 return inset.inset->lyxCode() == BOX_CODE || isLyxCodeToIgnore(inset.inset->lyxCode());
495 // Includes should not have a paragraph.
496 special_case |= nInsets == parSize && std::all_of(par->insetList().begin(), par->insetList().end(), [isLyxCodeToIgnore](InsetList::Element inset) {
497 return inset.inset->lyxCode() == INCLUDE_CODE || isLyxCodeToIgnore(inset.inset->lyxCode());
499 // Glossaries should not have a paragraph.
500 special_case |= nInsets == parSize && std::all_of(par->insetList().begin(), par->insetList().end(), [isLyxCodeToIgnore](InsetList::Element inset) {
501 return inset.inset->lyxCode() == NOMENCL_PRINT_CODE || isLyxCodeToIgnore(inset.inset->lyxCode());
504 bool const open_par = runparams.docbook_make_pars
505 && !runparams.docbook_in_par
508 // We want to issue the closing tag if either:
509 // (i) We opened it, and either docbook_in_par is false,
510 // or we're not in the last paragraph, anyway.
511 // (ii) We didn't open it and docbook_in_par is true,
512 // but we are in the first par, and there is a next par.
513 bool const close_par = open_par && (!runparams.docbook_in_par);
515 // Determine if this paragraph has some real content. Things like new pages are not caught
516 // by Paragraph::empty(), even though they do not generate anything useful in DocBook.
517 // Thus, remove all spaces (including new lines: \r, \n) before checking for emptiness.
518 // std::all_of allows doing this check without having to copy the string.
519 // Open and close tags around each contained paragraph.
522 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(distance(begin, par)), 0, nextpar == end, special_case);
523 for (docstring const & parXML : pars) {
524 if (!xml::isNotOnlySpace(parXML))
528 openParTag(xs, &*par, prevpar);
530 xs << XMLStream::ESCAPE_NONE << parXML;
533 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
538 void makeEnvironment(Text const &text,
541 OutputParams const &runparams,
542 ParagraphList::const_iterator const & par)
544 // If this kind of layout should be ignored, already leave.
545 if (par->layout().docbooktag() == "IGNORE")
549 auto const end = text.paragraphs().end();
553 // Special cases for listing-like environments provided in layouts. This is quite ad-hoc, but provides a useful
554 // default. This should not be used by too many environments (only LyX-Code right now).
555 // This would be much simpler if LyX-Code was implemented as InsetListings...
556 bool mimicListing = false;
557 bool ignoreFonts = false;
558 if (par->layout().docbooktag() == "programlisting") {
563 // Output the opening tag for this environment, but only if it has not been previously opened (condition
564 // implemented in openParTag).
565 auto prevpar = text.paragraphs().getParagraphBefore(par);
566 openParTag(xs, &*par, prevpar); // TODO: switch in layout for par/block?
568 // Generate the contents of this environment. There is a special case if this is like some environment.
569 Layout const & style = par->layout();
570 if (style.latextype == LATEX_COMMAND) {
571 // Nothing to do (otherwise, infinite loops).
572 } else if (style.latextype == LATEX_ENVIRONMENT) {
573 // Generate the paragraph, if need be.
574 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(std::distance(text.paragraphs().begin(), par)), 0, false, ignoreFonts);
577 auto p = pars.begin();
578 while (p != pars.end()) {
579 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(), par->layout().docbookiteminnertagtype());
580 xs << XMLStream::ESCAPE_NONE << *p;
581 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
584 // Insert a new line after each "paragraph" (i.e. line in the listing), except for the last one.
585 // Otherwise, there would one more new line in the output than in the LyX document.
590 for (auto const & p : pars) {
591 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(), par->layout().docbookiteminnertagtype());
592 xs << XMLStream::ESCAPE_NONE << p;
593 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
597 makeAny(text, buf, xs, runparams, par);
600 // Close the environment.
601 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr); // TODO: switch in layout for par/block?
605 ParagraphList::const_iterator findEndOfEnvironment(
606 ParagraphList::const_iterator const & pstart,
607 ParagraphList::const_iterator const & pend)
609 // Copy-paste from XHTML. Should be factored out at some point...
610 ParagraphList::const_iterator p = pstart;
611 Layout const & bstyle = p->layout();
612 size_t const depth = p->params().depth();
613 for (++p; p != pend; ++p) {
614 Layout const & style = p->layout();
615 // It shouldn't happen that e.g. a section command occurs inside
616 // a quotation environment, at a higher depth, but as of 6/2009,
617 // it can happen. We pretend that it's just at lowest depth.
618 if (style.latextype == LATEX_COMMAND)
621 // If depth is down, we're done
622 if (p->params().depth() < depth)
625 // If depth is up, we're not done
626 if (p->params().depth() > depth)
629 // FIXME I am not sure about the first check.
630 // Surely we *could* have different layouts that count as
631 // LATEX_PARAGRAPH, right?
632 if (style.latextype == LATEX_PARAGRAPH || style != bstyle)
639 ParagraphList::const_iterator makeListEnvironment(Text const &text,
642 OutputParams const &runparams,
643 ParagraphList::const_iterator const & begin)
647 auto const end = text.paragraphs().end();
648 auto const envend = findEndOfEnvironment(par, end);
650 // If this kind of layout should be ignored, already leave.
651 if (begin->layout().docbooktag() == "IGNORE") {
657 // Output the opening tag for this environment.
658 Layout const & envstyle = par->layout();
659 openTag(xs, envstyle.docbookwrappertag(), envstyle.docbookwrapperattr(), envstyle.docbookwrappertagtype());
660 openTag(xs, envstyle.docbooktag(), envstyle.docbookattr(), envstyle.docbooktagtype());
662 // Handle the content of the list environment, item by item.
663 while (par != envend) {
664 // Skip this paragraph if it is both empty and the last one (otherwise, there may be deeper paragraphs after).
667 if (par->empty() && nextpar == envend)
670 // Open the item wrapper.
671 Layout const & style = par->layout();
672 openTag(xs, style.docbookitemwrappertag(), style.docbookitemwrapperattr(), style.docbookitemwrappertagtype());
674 // Generate the label, if need be. If it is taken from the text, sep != 0 and corresponds to the first
675 // character after the label.
677 if (style.labeltype != LABEL_NO_LABEL && style.docbookitemlabeltag() != "NONE") {
678 if (style.labeltype == LABEL_MANUAL) {
679 // Only variablelist gets here (or similar items defined as an extension in the layout).
680 openTag(xs, style.docbookitemlabeltag(), style.docbookitemlabelattr(), style.docbookitemlabeltagtype());
681 sep = 1 + par->firstWordDocBook(xs, runparams);
682 closeTag(xs, style.docbookitemlabeltag(), style.docbookitemlabeltagtype());
684 // Usual cases: maybe there is something specified at the layout level. Highly unlikely, though.
685 docstring const lbl = par->params().labelString();
688 openTag(xs, style.docbookitemlabeltag(), style.docbookitemlabelattr(), style.docbookitemlabeltagtype());
690 closeTag(xs, style.docbookitemlabeltag(), style.docbookitemlabeltagtype());
695 // Open the item (after the wrapper and the label).
696 openTag(xs, style.docbookitemtag(), style.docbookitemattr(), style.docbookitemtagtype());
698 // Generate the content of the item.
699 if (sep < par->size()) {
700 auto pars = par->simpleDocBookOnePar(buf, runparams,
701 text.outerFont(std::distance(text.paragraphs().begin(), par)), sep);
702 for (auto &p : pars) {
703 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
704 par->layout().docbookiteminnertagtype());
705 xs << XMLStream::ESCAPE_NONE << p;
706 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
709 // DocBook doesn't like emptiness.
710 compTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
711 par->layout().docbookiteminnertagtype());
714 // If the next item is deeper, it must go entirely within this item (do it recursively).
715 // By construction, with findEndOfEnvironment, depth can only stay constant or increase, never decrease.
716 depth_type currentDepth = par->getDepth();
718 while (par != envend && par->getDepth() != currentDepth)
719 par = makeAny(text, buf, xs, runparams, par);
720 // Usually, this loop only makes one iteration, except in complex scenarios, like an item with a paragraph,
721 // a list, and another paragraph; or an item with two types of list (itemise then enumerate, for instance).
724 closeTag(xs, style.docbookitemtag(), style.docbookitemtagtype());
725 closeTag(xs, style.docbookitemwrappertag(), style.docbookitemwrappertagtype());
728 // Close this environment in exactly the same way as it was opened.
729 closeTag(xs, envstyle.docbooktag(), envstyle.docbooktagtype());
730 closeTag(xs, envstyle.docbookwrappertag(), envstyle.docbookwrappertagtype());
740 OutputParams const & runparams,
741 ParagraphList::const_iterator const & par)
743 // If this kind of layout should be ignored, already leave.
744 if (par->layout().docbooktag() == "IGNORE")
748 // Unlike XHTML, no need for labels, as they are handled by DocBook tags.
749 auto const begin = text.paragraphs().begin();
750 auto const end = text.paragraphs().end();
754 // Generate this command.
755 auto prevpar = text.paragraphs().getParagraphBefore(par);
756 openParTag(xs, &*par, prevpar);
758 auto pars = par->simpleDocBookOnePar(buf, runparams,text.outerFont(distance(begin, par)));
759 for (auto & parXML : pars)
760 // TODO: decide what to do with openParTag/closeParTag in new lines.
761 xs << XMLStream::ESCAPE_NONE << parXML;
763 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
767 bool isLayoutSectioning(Layout const & lay)
769 if (lay.docbooksection()) // Special case: some DocBook styles must be handled as sections.
771 else if (lay.category() == from_utf8("Sectioning") || lay.docbooktag() == "section") // Generic case.
772 return lay.toclevel != Layout::NOT_IN_TOC;
777 bool isLayoutSectioningOrSimilar(Layout const & lay)
779 return isLayoutSectioning(lay) || lay.docbooktag() == "bridgehead";
783 using DocBookDocumentSectioning = tuple<bool, pit_type>;
786 struct DocBookInfoTag
788 const set<pit_type> shouldBeInInfo;
789 const set<pit_type> mustBeInInfo; // With the notable exception of the abstract!
790 const set<pit_type> abstract;
791 const bool abstractLayout;
795 DocBookInfoTag(const set<pit_type> & shouldBeInInfo, const set<pit_type> & mustBeInInfo,
796 const set<pit_type> & abstract, bool abstractLayout, pit_type bpit, pit_type epit) :
797 shouldBeInInfo(shouldBeInInfo), mustBeInInfo(mustBeInInfo), abstract(abstract),
798 abstractLayout(abstractLayout), bpit(bpit), epit(epit) {}
802 DocBookDocumentSectioning hasDocumentSectioning(ParagraphList const ¶graphs, pit_type bpit, pit_type const epit) {
803 bool documentHasSections = false;
805 while (bpit < epit) {
806 Layout const &style = paragraphs[bpit].layout();
807 documentHasSections |= isLayoutSectioningOrSimilar(style);
809 if (documentHasSections)
813 // Paragraphs before the first section: [ runparams.par_begin ; eppit )
815 return make_tuple(documentHasSections, bpit);
819 bool hasOnlyNotes(Paragraph const & par)
821 // Precondition: the paragraph is not empty. Otherwise, the function will always return true...
822 for (int i = 0; i < par.size(); ++i)
823 // If you find something that is not an inset (like actual text) or an inset that is not a note,
825 if (!par.isInset(i) || par.getInset(i)->lyxCode() != NOTE_CODE)
828 // An empty paragraph may still require some output.
829 if (par.layout().docbooksection())
832 // There should be really no content here.
837 DocBookInfoTag getParagraphsWithInfo(ParagraphList const ¶graphs,
838 pit_type bpit, pit_type const epit,
839 // Typically, bpit is the beginning of the document and epit the end of the
840 // document *or* the first section.
841 bool documentHasSections,
842 bool detectUnlayoutedAbstract
843 // Whether paragraphs with no specific layout should be detected as abstracts.
844 // For inner sections, an abstract should only be detected if it has a specific
845 // layout. For others, anything that might look like an abstract should be sought.
847 set<pit_type> shouldBeInInfo;
848 set<pit_type> mustBeInInfo;
849 set<pit_type> abstractWithLayout;
850 set<pit_type> abstractNoLayout;
852 // Find the first non empty paragraph by mutating bpit.
853 while (bpit < epit) {
854 Paragraph const &par = paragraphs[bpit];
855 if (par.empty() || hasOnlyNotes(par))
861 // Traverse everything that might belong to <info>.
862 bool hasAbstractLayout = false;
863 pit_type cpit = bpit;
864 for (; cpit < epit; ++cpit) {
865 // Skip paragraphs that don't generate anything in DocBook.
866 Paragraph const & par = paragraphs[cpit];
867 Layout const &style = par.layout();
868 if (hasOnlyNotes(par))
871 // There should never be any section here, except for the first paragraph (a title can be part of <info>).
872 // (Just a sanity check: if this fails, this function could end up processing the whole document.)
873 if (cpit != bpit && isLayoutSectioningOrSimilar(par.layout())) {
874 LYXERR0("Assertion failed: section found in potential <info> paragraphs.");
878 // If this is marked as an abstract by the layout, put it in the right set.
879 if (style.docbookabstract()) {
880 hasAbstractLayout = true;
881 abstractWithLayout.emplace(cpit);
885 // Based on layout information, store this paragraph in one set: should be in <info>, must be,
886 // or abstract (either because of layout or of position).
887 if (style.docbookininfo() == "always")
888 mustBeInInfo.emplace(cpit);
889 else if (style.docbookininfo() == "maybe")
890 shouldBeInInfo.emplace(cpit);
891 else if (documentHasSections && !hasAbstractLayout && detectUnlayoutedAbstract &&
892 (style.docbooktag() == "NONE" || style.docbooktag() == "para") &&
893 style.docbookwrappertag() == "NONE")
894 // In this case, it is very likely that style.docbookininfo() == "never"! Be extra careful
895 // about anything that gets caught here.
896 abstractNoLayout.emplace(cpit);
897 else // This should definitely not be in <info>.
900 // Now, cpit points to the first paragraph that no more has things that could go in <info>.
901 // bpit is the beginning of the <info> part.
903 return DocBookInfoTag(shouldBeInInfo, mustBeInInfo,
904 hasAbstractLayout ? abstractWithLayout : abstractNoLayout,
905 hasAbstractLayout, bpit, cpit);
908 } // end anonymous namespace
911 ParagraphList::const_iterator makeAny(Text const &text,
914 OutputParams const &runparams,
915 ParagraphList::const_iterator par)
917 switch (par->layout().latextype) {
919 makeCommand(text, buf, xs, runparams, par);
921 case LATEX_ENVIRONMENT:
922 makeEnvironment(text, buf, xs, runparams, par);
924 case LATEX_LIST_ENVIRONMENT:
925 case LATEX_ITEM_ENVIRONMENT:
926 // Only case when makeAny() might consume more than one paragraph.
927 return makeListEnvironment(text, buf, xs, runparams, par);
928 case LATEX_PARAGRAPH:
929 makeParagraph(text, buf, xs, runparams, par);
931 case LATEX_BIB_ENVIRONMENT:
932 makeBibliography(text, buf, xs, runparams, par);
940 xml::FontTag docbookStartFontTag(xml::FontTypes type)
942 return xml::FontTag(from_utf8(fontToDocBookTag(type)), from_utf8(fontToAttribute(type)), type);
946 xml::EndFontTag docbookEndFontTag(xml::FontTypes type)
948 return xml::EndFontTag(from_utf8(fontToDocBookTag(type)), type);
952 void outputDocBookInfo(
956 OutputParams const & runparams,
957 ParagraphList const & paragraphs,
958 DocBookInfoTag const & info)
960 // Perform an additional check on the abstract. Sometimes, there are many paragraphs that should go
961 // into the abstract, but none generates actual content. Thus, first generate to a temporary stream,
962 // then only create the <abstract> tag if these paragraphs generate some content.
963 // This check must be performed *before* a decision on whether or not to output <info> is made.
964 bool hasAbstract = !info.abstract.empty();
967 // Generate the abstract XML into a string before further checks.
968 // Usually, makeAny only generates one paragraph at a time. However, for the specific case of lists, it might
969 // generate more than one paragraph, as indicated in the return value.
970 odocstringstream os2;
973 set<pit_type> doneParas;
974 for (auto const & p : info.abstract) {
975 if (doneParas.find(p) == doneParas.end()) {
976 auto oldPar = paragraphs.iterator_at(p);
977 auto newPar = makeAny(text, buf, xs2, runparams, oldPar);
979 // Insert the indices of all the paragraphs that were just generated (typically, one).
980 // **Make the hypothesis that, when an abstract has a list, all its items are consecutive.**
982 while (oldPar != newPar) {
983 doneParas.emplace(id);
990 // Actually output the abstract if there is something to do. Don't count line feeds or spaces in this,
991 // even though they must be properly output if there is some abstract.
992 abstract = os2.str();
993 docstring cleaned = abstract;
994 cleaned.erase(std::remove_if(cleaned.begin(), cleaned.end(), lyx::isSpace), cleaned.end());
996 // Nothing? Then there is no abstract!
1001 // The abstract must go in <info>. Otherwise, decide whether to open <info> based on the layouts.
1002 bool needInfo = !info.mustBeInInfo.empty() || hasAbstract;
1004 // Start the <info> tag if required.
1006 xs.startDivision(false);
1007 xs << xml::StartTag("info");
1011 // Output the elements that should go in <info>, before and after the abstract.
1012 for (auto pit : info.shouldBeInInfo) // Typically, the title: these elements are so important and ubiquitous
1013 // that mandating a wrapper like <info> would repel users. Thus, generate them first.
1014 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
1015 for (auto pit : info.mustBeInInfo)
1016 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
1018 // If there is no title, generate one (required for the document to be valid).
1019 // This code is called for the main document, for table cells, etc., so be precise in this condition.
1020 if (text.isMainText() && info.shouldBeInInfo.empty() && !runparams.inInclude) {
1021 xs << xml::StartTag("title");
1022 xs << "Untitled Document";
1023 xs << xml::EndTag("title");
1027 // Always output the abstract as the last item of the <info>, as it requires special treatment (especially if
1028 // it contains several paragraphs that are empty).
1030 if (info.abstractLayout) {
1031 xs << XMLStream::ESCAPE_NONE << abstract;
1034 string tag = paragraphs[*info.abstract.begin()].layout().docbookforceabstracttag();
1038 if (!xs.isLastTagCR())
1041 xs << xml::StartTag(tag);
1043 xs << XMLStream::ESCAPE_NONE << abstract;
1044 xs << xml::EndTag(tag);
1049 // End the <info> tag if it was started.
1051 if (!xs.isLastTagCR())
1054 xs << xml::EndTag("info");
1061 void docbookSimpleAllParagraphs(
1065 OutputParams const & runparams)
1067 // Handle the given text, supposing it has no sections (i.e. a "simple" text). The input may vary in length
1068 // between a single paragraph to a whole document.
1069 pit_type const bpit = runparams.par_begin;
1070 pit_type const epit = runparams.par_end;
1071 ParagraphList const ¶graphs = text.paragraphs();
1073 // First, the <info> tag.
1074 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, epit, false, true);
1075 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
1077 // Then, the content. It starts where the <info> ends.
1078 auto par = paragraphs.iterator_at(info.epit);
1079 auto end = paragraphs.iterator_at(epit);
1080 while (par != end) {
1081 if (!hasOnlyNotes(*par))
1082 par = makeAny(text, buf, xs, runparams, par);
1089 void docbookParagraphs(Text const &text,
1092 OutputParams const &runparams) {
1093 ParagraphList const ¶graphs = text.paragraphs();
1094 if (runparams.par_begin == runparams.par_end) {
1095 runparams.par_begin = 0;
1096 runparams.par_end = paragraphs.size();
1098 pit_type bpit = runparams.par_begin;
1099 pit_type const epit = runparams.par_end;
1100 LASSERT(bpit < epit,
1102 xs << XMLStream::ESCAPE_NONE << "<!-- DocBook output error! -->\n";
1106 std::stack<std::pair<int, string>> headerLevels; // Used to determine when to open/close sections: store the depth
1107 // of the section and the tag that was used to open it.
1109 // Detect whether the document contains sections. If there are no sections, treatment is largely simplified.
1110 // In particular, there can't be an abstract, unless it is manually marked.
1111 bool documentHasSections;
1113 tie(documentHasSections, eppit) = hasDocumentSectioning(paragraphs, bpit, epit);
1115 // Deal with "simple" documents, i.e. those without sections.
1116 if (!documentHasSections) {
1117 docbookSimpleAllParagraphs(text, buf, xs, runparams);
1121 // Output the first <info> tag (or just the title).
1122 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, eppit, true, true);
1123 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
1126 // Then, iterate through the paragraphs of this document.
1127 bool currentlyInAppendix = false;
1129 auto par = text.paragraphs().iterator_at(bpit);
1130 auto end = text.paragraphs().iterator_at(epit);
1131 while (par != end) {
1132 OutputParams ourparams = runparams;
1134 if (par->params().startOfAppendix())
1135 currentlyInAppendix = true;
1136 if (hasOnlyNotes(*par)) {
1141 Layout const &style = par->layout();
1143 // Think about adding <section> and/or </section>s.
1144 if (isLayoutSectioning(style)) {
1145 int level = style.toclevel;
1147 // Need to close a previous section if it has the same level or a higher one (close <section> if opening a
1148 // <h2> after a <h2>, <h3>, <h4>, <h5> or <h6>). More examples:
1149 // - current: h2; back: h1; do not close any <section>
1150 // - current: h1; back: h2; close two <section> (first the <h2>, then the <h1>, so a new <h1> can come)
1151 while (!headerLevels.empty() && level <= headerLevels.top().first) {
1152 // Output the tag only if it corresponds to a legit section.
1153 int stackLevel = headerLevels.top().first;
1154 if (stackLevel != Layout::NOT_IN_TOC) {
1155 xs << xml::EndTag(headerLevels.top().second);
1161 // Open the new section: first push it onto the stack, then output it in DocBook.
1162 string sectionTag = (currentlyInAppendix && style.docbooksectiontag() == "chapter") ?
1163 "appendix" : style.docbooksectiontag();
1164 headerLevels.push(std::make_pair(level, sectionTag));
1166 // Some sectioning-like elements should not be output (such as FrontMatter).
1167 if (level != Layout::NOT_IN_TOC) {
1168 // Look for a label in the title, i.e. a InsetLabel as a child.
1169 docstring id = docstring();
1170 for (pos_type i = 0; i < par->size(); ++i) {
1171 Inset const *inset = par->getInset(i);
1173 if (auto label = dynamic_cast<InsetLabel const *>(inset)) {
1174 // Generate the attributes for the section if need be.
1175 id += "xml:id=\"" + xml::cleanID(label->screenLabel()) + "\"";
1177 // Don't output the ID as a DocBook <anchor>.
1178 ourparams.docbook_anchors_to_ignore.emplace(label->screenLabel());
1180 // Cannot have multiple IDs per tag. If there is another ID inset in the document, it will
1181 // be output as a DocBook anchor.
1187 // Write the open tag for this section.
1191 xs << xml::StartTag(sectionTag, attrs);
1196 // Close all sections before the bibliography.
1197 // 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)?
1198 if (!par->insetList().empty()) {
1199 Inset const *firstInset = par->getInset(0);
1200 if (firstInset && (firstInset->lyxCode() == BIBITEM_CODE || firstInset->lyxCode() == BIBTEX_CODE)) {
1201 while (!headerLevels.empty()) {
1202 int level = headerLevels.top().first;
1203 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1206 // Output the tag only if it corresponds to a legit section.
1207 if (level != Layout::NOT_IN_TOC) {
1208 xs << XMLStream::ESCAPE_NONE << tag;
1215 // Generate the <info> tag if a section was just opened.
1216 // Some sections may require abstracts (mostly parts, in books: DocBookForceAbstractTag will not be NONE),
1217 // others can still have an abstract (it must be detected so that it can be output at the right place).
1218 // 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.
1219 if (isLayoutSectioning(style)) {
1220 // This abstract may be found between the next paragraph and the next title.
1221 pit_type cpit = std::distance(text.paragraphs().begin(), par);
1222 pit_type ppit = std::get<1>(hasDocumentSectioning(paragraphs, cpit + 1L, epit));
1224 // Generate this abstract (this code corresponds to parts of outputDocBookInfo).
1225 DocBookInfoTag secInfo = getParagraphsWithInfo(paragraphs, cpit, ppit, true,
1226 style.docbookforceabstracttag() != "NONE");
1228 if (!secInfo.mustBeInInfo.empty() || !secInfo.shouldBeInInfo.empty() || !secInfo.abstract.empty()) {
1229 // Generate the <info>, if required. If DocBookForceAbstractTag != NONE, this abstract will not be in
1230 // <info>, unlike other ("standard") abstracts.
1231 bool hasStandardAbstract = !secInfo.abstract.empty() && style.docbookforceabstracttag() == "NONE";
1232 bool needInfo = !secInfo.mustBeInInfo.empty() || hasStandardAbstract;
1235 xs.startDivision(false);
1236 xs << xml::StartTag("info");
1240 // Output the elements that should go in <info>, before and after the abstract.
1241 for (auto pit : secInfo.shouldBeInInfo) // Typically, the title: these elements are so important and ubiquitous
1242 // that mandating a wrapper like <info> would repel users. Thus, generate them first.
1243 makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(pit));
1244 for (auto pit : secInfo.mustBeInInfo)
1245 makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(pit));
1247 // Deal with the abstract in <info> if it is standard (i.e. its tag is <abstract>).
1248 if (!secInfo.abstract.empty() && hasStandardAbstract) {
1249 if (!secInfo.abstractLayout) {
1250 xs << xml::StartTag("abstract");
1254 for (auto const &p : secInfo.abstract)
1255 makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(p));
1257 if (!secInfo.abstractLayout) {
1258 xs << xml::EndTag("abstract");
1263 // End the <info> tag if it was started.
1265 if (!xs.isLastTagCR())
1268 xs << xml::EndTag("info");
1273 // Deal with the abstract outside <info> if it is not standard (i.e. its tag is layout-defined).
1274 if (!secInfo.abstract.empty() && !hasStandardAbstract) {
1275 // Assert: style.docbookforceabstracttag() != NONE.
1276 xs << xml::StartTag(style.docbookforceabstracttag());
1278 for (auto const &p : secInfo.abstract)
1279 makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(p));
1280 xs << xml::EndTag(style.docbookforceabstracttag());
1284 // Skip all the text that has just been generated.
1285 par = paragraphs.iterator_at(secInfo.epit);
1287 // No <info> tag to generate, proceed as for normal paragraphs.
1288 par = makeAny(text, buf, xs, ourparams, par);
1291 // Generate this paragraph, as it has nothing special.
1292 par = makeAny(text, buf, xs, ourparams, par);
1296 // If need be, close <section>s, but only at the end of the document (otherwise, dealt with at the beginning
1298 while (!headerLevels.empty() && headerLevels.top().first > Layout::NOT_IN_TOC) {
1299 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1301 xs << XMLStream::ESCAPE_NONE << tag;