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());
454 // Plain layouts must be ignored.
455 special_case |= buf.params().documentClass().isPlainLayout(par->layout()) && !runparams.docbook_force_pars;
456 // Equations do not deserve their own paragraph (DocBook allows them outside paragraphs).
457 // Exception: any case that generates an <inlineequation> must still get a paragraph to be valid.
458 special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
459 return inset.inset && inset.inset->asInsetMath() && inset.inset->asInsetMath()->getType() != hullSimple;
461 // Tables do not deserve their own paragraphs (DocBook allows them outside paragraphs).
462 special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
463 return inset.inset->lyxCode() == TABULAR_CODE;
465 // Floats cannot be in paragraphs.
466 special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
467 return inset.inset->lyxCode() == FLOAT_CODE;
469 // Bibliographies cannot be in paragraphs. Bibitems should still be handled as paragraphs, though
470 // (see makeParagraphBibliography).
471 special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
472 return inset.inset->lyxCode() == BIBTEX_CODE;
474 // ERTs are in comments, not paragraphs.
475 special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
476 return inset.inset->lyxCode() == ERT_CODE;
478 // Listings should not get into their own paragraph.
479 special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
480 return inset.inset->lyxCode() == LISTINGS_CODE;
482 // Boxes cannot get into their own paragraph.
483 special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
484 return inset.inset->lyxCode() == BOX_CODE;
486 // Includes should not have a paragraph.
487 special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
488 return inset.inset->lyxCode() == INCLUDE_CODE;
490 // Glossaries should not have a paragraph.
491 special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
492 return inset.inset->lyxCode() == NOMENCL_PRINT_CODE;
495 bool const open_par = runparams.docbook_make_pars
496 && !runparams.docbook_in_par
499 // We want to issue the closing tag if either:
500 // (i) We opened it, and either docbook_in_par is false,
501 // or we're not in the last paragraph, anyway.
502 // (ii) We didn't open it and docbook_in_par is true,
503 // but we are in the first par, and there is a next par.
504 bool const close_par = open_par && (!runparams.docbook_in_par);
506 // Determine if this paragraph has some real content. Things like new pages are not caught
507 // by Paragraph::empty(), even though they do not generate anything useful in DocBook.
508 // Thus, remove all spaces (including new lines: \r, \n) before checking for emptiness.
509 // std::all_of allows doing this check without having to copy the string.
510 // Open and close tags around each contained paragraph.
513 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(distance(begin, par)), 0, nextpar == end, special_case);
514 for (docstring const & parXML : pars) {
515 if (!xml::isNotOnlySpace(parXML))
519 openParTag(xs, &*par, prevpar);
521 xs << XMLStream::ESCAPE_NONE << parXML;
524 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
529 void makeEnvironment(Text const &text,
532 OutputParams const &runparams,
533 ParagraphList::const_iterator const & par)
535 // If this kind of layout should be ignored, already leave.
536 if (par->layout().docbooktag() == "IGNORE")
540 auto const end = text.paragraphs().end();
544 // Special cases for listing-like environments provided in layouts. This is quite ad-hoc, but provides a useful
545 // default. This should not be used by too many environments (only LyX-Code right now).
546 // This would be much simpler if LyX-Code was implemented as InsetListings...
547 bool mimicListing = false;
548 bool ignoreFonts = false;
549 if (par->layout().docbooktag() == "programlisting") {
554 // Output the opening tag for this environment, but only if it has not been previously opened (condition
555 // implemented in openParTag).
556 auto prevpar = text.paragraphs().getParagraphBefore(par);
557 openParTag(xs, &*par, prevpar); // TODO: switch in layout for par/block?
559 // Generate the contents of this environment. There is a special case if this is like some environment.
560 Layout const & style = par->layout();
561 if (style.latextype == LATEX_COMMAND) {
562 // Nothing to do (otherwise, infinite loops).
563 } else if (style.latextype == LATEX_ENVIRONMENT) {
564 // Generate the paragraph, if need be.
565 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(std::distance(text.paragraphs().begin(), par)), 0, false, ignoreFonts);
568 auto p = pars.begin();
569 while (p != pars.end()) {
570 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(), par->layout().docbookiteminnertagtype());
571 xs << XMLStream::ESCAPE_NONE << *p;
572 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
575 // Insert a new line after each "paragraph" (i.e. line in the listing), except for the last one.
576 // Otherwise, there would one more new line in the output than in the LyX document.
581 for (auto const & p : pars) {
582 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(), par->layout().docbookiteminnertagtype());
583 xs << XMLStream::ESCAPE_NONE << p;
584 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
588 makeAny(text, buf, xs, runparams, par);
591 // Close the environment.
592 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr); // TODO: switch in layout for par/block?
596 ParagraphList::const_iterator findEndOfEnvironment(
597 ParagraphList::const_iterator const & pstart,
598 ParagraphList::const_iterator const & pend)
600 // Copy-paste from XHTML. Should be factored out at some point...
601 ParagraphList::const_iterator p = pstart;
602 Layout const & bstyle = p->layout();
603 size_t const depth = p->params().depth();
604 for (++p; p != pend; ++p) {
605 Layout const & style = p->layout();
606 // It shouldn't happen that e.g. a section command occurs inside
607 // a quotation environment, at a higher depth, but as of 6/2009,
608 // it can happen. We pretend that it's just at lowest depth.
609 if (style.latextype == LATEX_COMMAND)
612 // If depth is down, we're done
613 if (p->params().depth() < depth)
616 // If depth is up, we're not done
617 if (p->params().depth() > depth)
620 // FIXME I am not sure about the first check.
621 // Surely we *could* have different layouts that count as
622 // LATEX_PARAGRAPH, right?
623 if (style.latextype == LATEX_PARAGRAPH || style != bstyle)
630 ParagraphList::const_iterator makeListEnvironment(Text const &text,
633 OutputParams const &runparams,
634 ParagraphList::const_iterator const & begin)
638 auto const end = text.paragraphs().end();
639 auto const envend = findEndOfEnvironment(par, end);
641 // If this kind of layout should be ignored, already leave.
642 if (begin->layout().docbooktag() == "IGNORE") {
648 // Output the opening tag for this environment.
649 Layout const & envstyle = par->layout();
650 openTag(xs, envstyle.docbookwrappertag(), envstyle.docbookwrapperattr(), envstyle.docbookwrappertagtype());
651 openTag(xs, envstyle.docbooktag(), envstyle.docbookattr(), envstyle.docbooktagtype());
653 // Handle the content of the list environment, item by item.
654 while (par != envend) {
655 // Skip this paragraph if it is both empty and the last one (otherwise, there may be deeper paragraphs after).
658 if (par->empty() && nextpar == envend)
661 // Open the item wrapper.
662 Layout const & style = par->layout();
663 openTag(xs, style.docbookitemwrappertag(), style.docbookitemwrapperattr(), style.docbookitemwrappertagtype());
665 // Generate the label, if need be. If it is taken from the text, sep != 0 and corresponds to the first
666 // character after the label.
668 if (style.labeltype != LABEL_NO_LABEL && style.docbookitemlabeltag() != "NONE") {
669 if (style.labeltype == LABEL_MANUAL) {
670 // Only variablelist gets here (or similar items defined as an extension in the layout).
671 openTag(xs, style.docbookitemlabeltag(), style.docbookitemlabelattr(), style.docbookitemlabeltagtype());
672 sep = 1 + par->firstWordDocBook(xs, runparams);
673 closeTag(xs, style.docbookitemlabeltag(), style.docbookitemlabeltagtype());
675 // Usual cases: maybe there is something specified at the layout level. Highly unlikely, though.
676 docstring const lbl = par->params().labelString();
679 openTag(xs, style.docbookitemlabeltag(), style.docbookitemlabelattr(), style.docbookitemlabeltagtype());
681 closeTag(xs, style.docbookitemlabeltag(), style.docbookitemlabeltagtype());
686 // Open the item (after the wrapper and the label).
687 openTag(xs, style.docbookitemtag(), style.docbookitemattr(), style.docbookitemtagtype());
689 // Generate the content of the item.
690 if (sep < par->size()) {
691 auto pars = par->simpleDocBookOnePar(buf, runparams,
692 text.outerFont(std::distance(text.paragraphs().begin(), par)), sep);
693 for (auto &p : pars) {
694 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
695 par->layout().docbookiteminnertagtype());
696 xs << XMLStream::ESCAPE_NONE << p;
697 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
700 // DocBook doesn't like emptiness.
701 compTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
702 par->layout().docbookiteminnertagtype());
705 // If the next item is deeper, it must go entirely within this item (do it recursively).
706 // By construction, with findEndOfEnvironment, depth can only stay constant or increase, never decrease.
707 depth_type currentDepth = par->getDepth();
709 while (par != envend && par->getDepth() != currentDepth)
710 par = makeAny(text, buf, xs, runparams, par);
711 // Usually, this loop only makes one iteration, except in complex scenarios, like an item with a paragraph,
712 // a list, and another paragraph; or an item with two types of list (itemise then enumerate, for instance).
715 closeTag(xs, style.docbookitemtag(), style.docbookitemtagtype());
716 closeTag(xs, style.docbookitemwrappertag(), style.docbookitemwrappertagtype());
719 // Close this environment in exactly the same way as it was opened.
720 closeTag(xs, envstyle.docbooktag(), envstyle.docbooktagtype());
721 closeTag(xs, envstyle.docbookwrappertag(), envstyle.docbookwrappertagtype());
731 OutputParams const & runparams,
732 ParagraphList::const_iterator const & par)
734 // If this kind of layout should be ignored, already leave.
735 if (par->layout().docbooktag() == "IGNORE")
739 // Unlike XHTML, no need for labels, as they are handled by DocBook tags.
740 auto const begin = text.paragraphs().begin();
741 auto const end = text.paragraphs().end();
745 // Generate this command.
746 auto prevpar = text.paragraphs().getParagraphBefore(par);
747 openParTag(xs, &*par, prevpar);
749 auto pars = par->simpleDocBookOnePar(buf, runparams,text.outerFont(distance(begin, par)));
750 for (auto & parXML : pars)
751 // TODO: decide what to do with openParTag/closeParTag in new lines.
752 xs << XMLStream::ESCAPE_NONE << parXML;
754 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
758 bool isLayoutSectioning(Layout const & lay)
760 if (lay.docbooksection()) // Special case: some DocBook styles must be handled as sections.
762 else if (lay.category() == from_utf8("Sectioning") || lay.docbooktag() == "section") // Generic case.
763 return lay.toclevel != Layout::NOT_IN_TOC;
768 bool isLayoutSectioningOrSimilar(Layout const & lay)
770 return isLayoutSectioning(lay) || lay.docbooktag() == "bridgehead";
774 using DocBookDocumentSectioning = tuple<bool, pit_type>;
777 struct DocBookInfoTag
779 const set<pit_type> shouldBeInInfo;
780 const set<pit_type> mustBeInInfo; // With the notable exception of the abstract!
781 const set<pit_type> abstract;
782 const bool abstractLayout;
786 DocBookInfoTag(const set<pit_type> & shouldBeInInfo, const set<pit_type> & mustBeInInfo,
787 const set<pit_type> & abstract, bool abstractLayout, pit_type bpit, pit_type epit) :
788 shouldBeInInfo(shouldBeInInfo), mustBeInInfo(mustBeInInfo), abstract(abstract),
789 abstractLayout(abstractLayout), bpit(bpit), epit(epit) {}
793 DocBookDocumentSectioning hasDocumentSectioning(ParagraphList const ¶graphs, pit_type bpit, pit_type const epit) {
794 bool documentHasSections = false;
796 while (bpit < epit) {
797 Layout const &style = paragraphs[bpit].layout();
798 documentHasSections |= isLayoutSectioningOrSimilar(style);
800 if (documentHasSections)
804 // Paragraphs before the first section: [ runparams.par_begin ; eppit )
806 return make_tuple(documentHasSections, bpit);
810 bool hasOnlyNotes(Paragraph const & par)
812 // Precondition: the paragraph is not empty. Otherwise, the function will always return true...
813 for (int i = 0; i < par.size(); ++i)
814 // If you find something that is not an inset (like actual text) or an inset that is not a note,
816 if (!par.isInset(i) || par.getInset(i)->lyxCode() != NOTE_CODE)
819 // An empty paragraph may still require some output.
820 if (par.layout().docbooksection())
823 // There should be really no content here.
828 DocBookInfoTag getParagraphsWithInfo(ParagraphList const ¶graphs,
829 pit_type bpit, pit_type const epit,
830 // Typically, bpit is the beginning of the document and epit the end of the
831 // document *or* the first section.
832 bool documentHasSections,
833 bool detectUnlayoutedAbstract
834 // Whether paragraphs with no specific layout should be detected as abstracts.
835 // For inner sections, an abstract should only be detected if it has a specific
836 // layout. For others, anything that might look like an abstract should be sought.
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 Layout const &style = par.layout();
859 if (hasOnlyNotes(par))
862 // There should never be any section here, except for the first paragraph (a title can be part of <info>).
863 // (Just a sanity check: if this fails, this function could end up processing the whole document.)
864 if (cpit != bpit && isLayoutSectioningOrSimilar(par.layout())) {
865 LYXERR0("Assertion failed: section found in potential <info> paragraphs.");
869 // If this is marked as an abstract by the layout, put it in the right set.
870 if (style.docbookabstract()) {
871 hasAbstractLayout = true;
872 abstractWithLayout.emplace(cpit);
876 // Based on layout information, store this paragraph in one set: should be in <info>, must be,
877 // or abstract (either because of layout or of position).
878 if (style.docbookininfo() == "always")
879 mustBeInInfo.emplace(cpit);
880 else if (style.docbookininfo() == "maybe")
881 shouldBeInInfo.emplace(cpit);
882 else if (documentHasSections && !hasAbstractLayout && detectUnlayoutedAbstract &&
883 (style.docbooktag() == "NONE" || style.docbooktag() == "para") &&
884 style.docbookwrappertag() == "NONE")
885 // In this case, it is very likely that style.docbookininfo() == "never"! Be extra careful
886 // about anything that gets caught here.
887 abstractNoLayout.emplace(cpit);
888 else // This should definitely not be in <info>.
891 // Now, cpit points to the first paragraph that no more has things that could go in <info>.
892 // bpit is the beginning of the <info> part.
894 return DocBookInfoTag(shouldBeInInfo, mustBeInInfo,
895 hasAbstractLayout ? abstractWithLayout : abstractNoLayout,
896 hasAbstractLayout, bpit, cpit);
899 } // end anonymous namespace
902 ParagraphList::const_iterator makeAny(Text const &text,
905 OutputParams const &runparams,
906 ParagraphList::const_iterator par)
908 switch (par->layout().latextype) {
910 makeCommand(text, buf, xs, runparams, par);
912 case LATEX_ENVIRONMENT:
913 makeEnvironment(text, buf, xs, runparams, par);
915 case LATEX_LIST_ENVIRONMENT:
916 case LATEX_ITEM_ENVIRONMENT:
917 // Only case when makeAny() might consume more than one paragraph.
918 return makeListEnvironment(text, buf, xs, runparams, par);
919 case LATEX_PARAGRAPH:
920 makeParagraph(text, buf, xs, runparams, par);
922 case LATEX_BIB_ENVIRONMENT:
923 makeBibliography(text, buf, xs, runparams, par);
931 xml::FontTag docbookStartFontTag(xml::FontTypes type)
933 return xml::FontTag(from_utf8(fontToDocBookTag(type)), from_utf8(fontToAttribute(type)), type);
937 xml::EndFontTag docbookEndFontTag(xml::FontTypes type)
939 return xml::EndFontTag(from_utf8(fontToDocBookTag(type)), type);
943 void outputDocBookInfo(
947 OutputParams const & runparams,
948 ParagraphList const & paragraphs,
949 DocBookInfoTag const & info)
951 // Perform an additional check on the abstract. Sometimes, there are many paragraphs that should go
952 // into the abstract, but none generates actual content. Thus, first generate to a temporary stream,
953 // then only create the <abstract> tag if these paragraphs generate some content.
954 // This check must be performed *before* a decision on whether or not to output <info> is made.
955 bool hasAbstract = !info.abstract.empty();
958 // Generate the abstract XML into a string before further checks.
959 // Usually, makeAny only generates one paragraph at a time. However, for the specific case of lists, it might
960 // generate more than one paragraph, as indicated in the return value.
961 odocstringstream os2;
964 set<pit_type> doneParas;
965 for (auto const & p : info.abstract) {
966 if (doneParas.find(p) == doneParas.end()) {
967 auto oldPar = paragraphs.iterator_at(p);
968 auto newPar = makeAny(text, buf, xs2, runparams, oldPar);
970 // Insert the indices of all the paragraphs that were just generated (typically, one).
971 // **Make the hypothesis that, when an abstract has a list, all its items are consecutive.**
973 while (oldPar != newPar) {
974 doneParas.emplace(id);
981 // Actually output the abstract if there is something to do. Don't count line feeds or spaces in this,
982 // even though they must be properly output if there is some abstract.
983 abstract = os2.str();
984 docstring cleaned = abstract;
985 cleaned.erase(std::remove_if(cleaned.begin(), cleaned.end(), lyx::isSpace), cleaned.end());
987 // Nothing? Then there is no abstract!
992 // The abstract must go in <info>. Otherwise, decide whether to open <info> based on the layouts.
993 bool needInfo = !info.mustBeInInfo.empty() || hasAbstract;
995 // Start the <info> tag if required.
997 xs.startDivision(false);
998 xs << xml::StartTag("info");
1002 // Output the elements that should go in <info>, before and after the abstract.
1003 for (auto pit : info.shouldBeInInfo) // Typically, the title: these elements are so important and ubiquitous
1004 // that mandating a wrapper like <info> would repel users. Thus, generate them first.
1005 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
1006 for (auto pit : info.mustBeInInfo)
1007 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
1009 // If there is no title, generate one (required for the document to be valid).
1010 // This code is called for the main document, for table cells, etc., so be precise in this condition.
1011 if (text.isMainText() && info.shouldBeInInfo.empty() && !runparams.inInclude) {
1012 xs << xml::StartTag("title");
1013 xs << "Untitled Document";
1014 xs << xml::EndTag("title");
1018 // Always output the abstract as the last item of the <info>, as it requires special treatment (especially if
1019 // it contains several paragraphs that are empty).
1021 if (info.abstractLayout) {
1022 xs << XMLStream::ESCAPE_NONE << abstract;
1025 string tag = paragraphs[*info.abstract.begin()].layout().docbookforceabstracttag();
1029 if (!xs.isLastTagCR())
1032 xs << xml::StartTag(tag);
1034 xs << XMLStream::ESCAPE_NONE << abstract;
1035 xs << xml::EndTag(tag);
1040 // End the <info> tag if it was started.
1042 if (!xs.isLastTagCR())
1045 xs << xml::EndTag("info");
1052 void docbookSimpleAllParagraphs(
1056 OutputParams const & runparams)
1058 // Handle the given text, supposing it has no sections (i.e. a "simple" text). The input may vary in length
1059 // between a single paragraph to a whole document.
1060 pit_type const bpit = runparams.par_begin;
1061 pit_type const epit = runparams.par_end;
1062 ParagraphList const ¶graphs = text.paragraphs();
1064 // First, the <info> tag.
1065 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, epit, false, true);
1066 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
1068 // Then, the content. It starts where the <info> ends.
1069 auto par = paragraphs.iterator_at(info.epit);
1070 auto end = paragraphs.iterator_at(epit);
1071 while (par != end) {
1072 if (!hasOnlyNotes(*par))
1073 par = makeAny(text, buf, xs, runparams, par);
1080 void docbookParagraphs(Text const &text,
1083 OutputParams const &runparams) {
1084 ParagraphList const ¶graphs = text.paragraphs();
1085 if (runparams.par_begin == runparams.par_end) {
1086 runparams.par_begin = 0;
1087 runparams.par_end = paragraphs.size();
1089 pit_type bpit = runparams.par_begin;
1090 pit_type const epit = runparams.par_end;
1091 LASSERT(bpit < epit,
1093 xs << XMLStream::ESCAPE_NONE << "<!-- DocBook output error! -->\n";
1097 std::stack<std::pair<int, string>> headerLevels; // Used to determine when to open/close sections: store the depth
1098 // of the section and the tag that was used to open it.
1100 // Detect whether the document contains sections. If there are no sections, treatment is largely simplified.
1101 // In particular, there can't be an abstract, unless it is manually marked.
1102 bool documentHasSections;
1104 tie(documentHasSections, eppit) = hasDocumentSectioning(paragraphs, bpit, epit);
1106 // Deal with "simple" documents, i.e. those without sections.
1107 if (!documentHasSections) {
1108 docbookSimpleAllParagraphs(text, buf, xs, runparams);
1112 // Output the first <info> tag (or just the title).
1113 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, eppit, true, true);
1114 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
1117 // Then, iterate through the paragraphs of this document.
1118 bool currentlyInAppendix = false;
1120 auto par = text.paragraphs().iterator_at(bpit);
1121 auto end = text.paragraphs().iterator_at(epit);
1122 while (par != end) {
1123 OutputParams ourparams = runparams;
1125 if (par->params().startOfAppendix())
1126 currentlyInAppendix = true;
1127 if (hasOnlyNotes(*par)) {
1132 Layout const &style = par->layout();
1134 // Think about adding <section> and/or </section>s.
1135 if (isLayoutSectioning(style)) {
1136 int level = style.toclevel;
1138 // Need to close a previous section if it has the same level or a higher one (close <section> if opening a
1139 // <h2> after a <h2>, <h3>, <h4>, <h5> or <h6>). More examples:
1140 // - current: h2; back: h1; do not close any <section>
1141 // - current: h1; back: h2; close two <section> (first the <h2>, then the <h1>, so a new <h1> can come)
1142 while (!headerLevels.empty() && level <= headerLevels.top().first) {
1143 // Output the tag only if it corresponds to a legit section.
1144 int stackLevel = headerLevels.top().first;
1145 if (stackLevel != Layout::NOT_IN_TOC) {
1146 xs << xml::EndTag(headerLevels.top().second);
1152 // Open the new section: first push it onto the stack, then output it in DocBook.
1153 string sectionTag = (currentlyInAppendix && style.docbooksectiontag() == "chapter") ?
1154 "appendix" : style.docbooksectiontag();
1155 headerLevels.push(std::make_pair(level, sectionTag));
1157 // Some sectioning-like elements should not be output (such as FrontMatter).
1158 if (level != Layout::NOT_IN_TOC) {
1159 // Look for a label in the title, i.e. a InsetLabel as a child.
1160 docstring id = docstring();
1161 for (pos_type i = 0; i < par->size(); ++i) {
1162 Inset const *inset = par->getInset(i);
1164 if (auto label = dynamic_cast<InsetLabel const *>(inset)) {
1165 // Generate the attributes for the section if need be.
1166 id += "xml:id=\"" + xml::cleanID(label->screenLabel()) + "\"";
1168 // Don't output the ID as a DocBook <anchor>.
1169 ourparams.docbook_anchors_to_ignore.emplace(label->screenLabel());
1171 // Cannot have multiple IDs per tag. If there is another ID inset in the document, it will
1172 // be output as a DocBook anchor.
1178 // Write the open tag for this section.
1182 xs << xml::StartTag(sectionTag, attrs);
1187 // Close all sections before the bibliography.
1188 // 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)?
1189 if (!par->insetList().empty()) {
1190 Inset const *firstInset = par->getInset(0);
1191 if (firstInset && (firstInset->lyxCode() == BIBITEM_CODE || firstInset->lyxCode() == BIBTEX_CODE)) {
1192 while (!headerLevels.empty()) {
1193 int level = headerLevels.top().first;
1194 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1197 // Output the tag only if it corresponds to a legit section.
1198 if (level != Layout::NOT_IN_TOC) {
1199 xs << XMLStream::ESCAPE_NONE << tag;
1206 // Generate the <info> tag if a section was just opened.
1207 // Some sections may require abstracts (mostly parts, in books: DocBookForceAbstractTag will not be NONE),
1208 // others can still have an abstract (it must be detected so that it can be output at the right place).
1209 // 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.
1210 if (isLayoutSectioning(style)) {
1211 // This abstract may be found between the next paragraph and the next title.
1212 pit_type cpit = std::distance(text.paragraphs().begin(), par);
1213 pit_type ppit = std::get<1>(hasDocumentSectioning(paragraphs, cpit + 1L, epit));
1215 // Generate this abstract (this code corresponds to parts of outputDocBookInfo).
1216 DocBookInfoTag secInfo = getParagraphsWithInfo(paragraphs, cpit, ppit, true,
1217 style.docbookforceabstracttag() != "NONE");
1219 if (!secInfo.mustBeInInfo.empty() || !secInfo.shouldBeInInfo.empty() || !secInfo.abstract.empty()) {
1220 // Generate the <info>, if required. If DocBookForceAbstractTag != NONE, this abstract will not be in
1221 // <info>, unlike other ("standard") abstracts.
1222 bool hasStandardAbstract = !secInfo.abstract.empty() && style.docbookforceabstracttag() == "NONE";
1223 bool needInfo = !secInfo.mustBeInInfo.empty() || hasStandardAbstract;
1226 xs.startDivision(false);
1227 xs << xml::StartTag("info");
1231 // Output the elements that should go in <info>, before and after the abstract.
1232 for (auto pit : secInfo.shouldBeInInfo) // Typically, the title: these elements are so important and ubiquitous
1233 // that mandating a wrapper like <info> would repel users. Thus, generate them first.
1234 makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(pit));
1235 for (auto pit : secInfo.mustBeInInfo)
1236 makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(pit));
1238 // Deal with the abstract in <info> if it is standard (i.e. its tag is <abstract>).
1239 if (!secInfo.abstract.empty() && hasStandardAbstract) {
1240 if (!secInfo.abstractLayout) {
1241 xs << xml::StartTag("abstract");
1245 for (auto const &p : secInfo.abstract)
1246 makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(p));
1248 if (!secInfo.abstractLayout) {
1249 xs << xml::EndTag("abstract");
1254 // End the <info> tag if it was started.
1256 if (!xs.isLastTagCR())
1259 xs << xml::EndTag("info");
1264 // Deal with the abstract outside <info> if it is not standard (i.e. its tag is layout-defined).
1265 if (!secInfo.abstract.empty() && !hasStandardAbstract) {
1266 // Assert: style.docbookforceabstracttag() != NONE.
1267 xs << xml::StartTag(style.docbookforceabstracttag());
1269 for (auto const &p : secInfo.abstract)
1270 makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(p));
1271 xs << xml::EndTag(style.docbookforceabstracttag());
1275 // Skip all the text that has just been generated.
1276 par = paragraphs.iterator_at(secInfo.epit);
1278 // No <info> tag to generate, proceed as for normal paragraphs.
1279 par = makeAny(text, buf, xs, ourparams, par);
1282 // Generate this paragraph, as it has nothing special.
1283 par = makeAny(text, buf, xs, ourparams, par);
1287 // If need be, close <section>s, but only at the end of the document (otherwise, dealt with at the beginning
1289 while (!headerLevels.empty() && headerLevels.top().first > Layout::NOT_IN_TOC) {
1290 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1292 xs << XMLStream::ESCAPE_NONE << tag;