2 * \file output_docbook.cpp
3 * This file is part of LyX, the document processor.
4 * Licence details can be found in the file COPYING.
6 * \author Lars Gullik Bjønnes
9 * Full author contact details are available in file CREDITS.
15 #include "buffer_funcs.h"
16 #include "BufferParams.h"
18 #include "InsetList.h"
19 #include "Paragraph.h"
20 #include "ParagraphList.h"
21 #include "ParagraphParameters.h"
24 #include "TextClass.h"
26 #include "insets/InsetBibtex.h"
27 #include "insets/InsetBibitem.h"
28 #include "insets/InsetLabel.h"
29 #include "insets/InsetNote.h"
31 #include "support/lassert.h"
39 using namespace lyx::support;
45 std::string fontToDocBookTag(xml::FontTypes type)
48 case xml::FontTypes::FT_EMPH:
49 case xml::FontTypes::FT_BOLD:
51 case xml::FontTypes::FT_NOUN:
53 case xml::FontTypes::FT_UBAR:
54 case xml::FontTypes::FT_WAVE:
55 case xml::FontTypes::FT_DBAR:
56 case xml::FontTypes::FT_SOUT:
57 case xml::FontTypes::FT_XOUT:
58 case xml::FontTypes::FT_ITALIC:
59 case xml::FontTypes::FT_UPRIGHT:
60 case xml::FontTypes::FT_SLANTED:
61 case xml::FontTypes::FT_SMALLCAPS:
62 case xml::FontTypes::FT_ROMAN:
63 case xml::FontTypes::FT_SANS:
65 case xml::FontTypes::FT_TYPE:
67 case xml::FontTypes::FT_SIZE_TINY:
68 case xml::FontTypes::FT_SIZE_SCRIPT:
69 case xml::FontTypes::FT_SIZE_FOOTNOTE:
70 case xml::FontTypes::FT_SIZE_SMALL:
71 case xml::FontTypes::FT_SIZE_NORMAL:
72 case xml::FontTypes::FT_SIZE_LARGE:
73 case xml::FontTypes::FT_SIZE_LARGER:
74 case xml::FontTypes::FT_SIZE_LARGEST:
75 case xml::FontTypes::FT_SIZE_HUGE:
76 case xml::FontTypes::FT_SIZE_HUGER:
77 case xml::FontTypes::FT_SIZE_INCREASE:
78 case xml::FontTypes::FT_SIZE_DECREASE:
86 string fontToRole(xml::FontTypes type)
88 // Specific fonts are achieved with roles. The only common ones are "" for basic emphasis,
89 // and "bold"/"strong" for bold. With some specific options, other roles are copied into
90 // HTML output (via the DocBook XSLT sheets); otherwise, if not recognised, they are just ignored.
91 // Hence, it is not a problem to have many roles by default here.
92 // See https://www.sourceware.org/ml/docbook/2003-05/msg00269.html
94 case xml::FontTypes::FT_ITALIC:
95 case xml::FontTypes::FT_EMPH:
97 case xml::FontTypes::FT_BOLD:
99 case xml::FontTypes::FT_NOUN: // Outputs a <person>
100 case xml::FontTypes::FT_TYPE: // Outputs a <code>
102 case xml::FontTypes::FT_UBAR:
105 // All other roles are non-standard for DocBook.
107 case xml::FontTypes::FT_WAVE:
109 case xml::FontTypes::FT_DBAR:
111 case xml::FontTypes::FT_SOUT:
113 case xml::FontTypes::FT_XOUT:
115 case xml::FontTypes::FT_UPRIGHT:
117 case xml::FontTypes::FT_SLANTED:
119 case xml::FontTypes::FT_SMALLCAPS:
121 case xml::FontTypes::FT_ROMAN:
123 case xml::FontTypes::FT_SANS:
125 case xml::FontTypes::FT_SIZE_TINY:
127 case xml::FontTypes::FT_SIZE_SCRIPT:
128 return "size_script";
129 case xml::FontTypes::FT_SIZE_FOOTNOTE:
130 return "size_footnote";
131 case xml::FontTypes::FT_SIZE_SMALL:
133 case xml::FontTypes::FT_SIZE_NORMAL:
134 return "size_normal";
135 case xml::FontTypes::FT_SIZE_LARGE:
137 case xml::FontTypes::FT_SIZE_LARGER:
138 return "size_larger";
139 case xml::FontTypes::FT_SIZE_LARGEST:
140 return "size_largest";
141 case xml::FontTypes::FT_SIZE_HUGE:
143 case xml::FontTypes::FT_SIZE_HUGER:
145 case xml::FontTypes::FT_SIZE_INCREASE:
146 return "size_increase";
147 case xml::FontTypes::FT_SIZE_DECREASE:
148 return "size_decrease";
155 string fontToAttribute(xml::FontTypes type) {
156 // If there is a role (i.e. nonstandard use of a tag), output the attribute. Otherwise, the sheer tag is sufficient
158 string role = fontToRole(type);
160 return "role='" + role + "'";
167 // Convenience functions to open and close tags. First, very low-level ones to ensure a consistent new-line behaviour.
171 // Contents of the block.
176 // <paratag>Contents of the paragraph.</paratag>
179 // Content before<inlinetag>Contents of the paragraph.</inlinetag>Content after
181 void openInlineTag(XMLStream & xs, const std::string & tag, const std::string & attr)
183 xs << xml::StartTag(tag, attr);
187 void closeInlineTag(XMLStream & xs, const std::string & tag)
189 xs << xml::EndTag(tag);
193 void openParTag(XMLStream & xs, const std::string & tag, const std::string & attr)
195 if (!xs.isLastTagCR())
197 xs << xml::StartTag(tag, attr);
201 void closeParTag(XMLStream & xs, const std::string & tag)
203 xs << xml::EndTag(tag);
208 void openBlockTag(XMLStream & xs, const std::string & tag, const std::string & attr)
210 if (!xs.isLastTagCR())
212 xs << xml::StartTag(tag, attr);
217 void closeBlockTag(XMLStream & xs, const std::string & tag)
219 if (!xs.isLastTagCR())
221 xs << xml::EndTag(tag);
226 void openTag(XMLStream & xs, const std::string & tag, const std::string & attr, const std::string & tagtype)
228 if (tag.empty() || tag == "NONE") // Common check to be performed elsewhere, if it was not here.
231 if (tag == "para" || tagtype == "paragraph") // Special case for <para>: always considered as a paragraph.
232 openParTag(xs, tag, attr);
233 else if (tagtype == "block")
234 openBlockTag(xs, tag, attr);
235 else if (tagtype == "inline")
236 openInlineTag(xs, tag, attr);
238 xs.writeError("Unrecognised tag type '" + tagtype + "' for '" + tag + " " + attr + "'");
242 void closeTag(XMLStream & xs, const std::string & tag, const std::string & tagtype)
244 if (tag.empty() || tag == "NONE")
247 if (tag == "para" || tagtype == "paragraph") // Special case for <para>: always considered as a paragraph.
248 closeParTag(xs, tag);
249 else if (tagtype == "block")
250 closeBlockTag(xs, tag);
251 else if (tagtype == "inline")
252 closeInlineTag(xs, tag);
254 xs.writeError("Unrecognised tag type '" + tagtype + "' for '" + tag + "'");
258 void compTag(XMLStream & xs, const std::string & tag, const std::string & attr, const std::string & tagtype)
260 if (tag.empty() || tag == "NONE")
263 // Special case for <para>: always considered as a paragraph.
264 if (tag == "para" || tagtype == "paragraph" || tagtype == "block") {
265 if (!xs.isLastTagCR())
267 xs << xml::CompTag(tag, attr);
269 } else if (tagtype == "inline") {
270 xs << xml::CompTag(tag, attr);
272 xs.writeError("Unrecognised tag type '" + tagtype + "' for '" + tag + "'");
277 // Higher-level convenience functions.
279 void openParTag(XMLStream & xs, const Paragraph * par, const Paragraph * prevpar)
281 Layout const & lay = par->layout();
286 // When should the wrapper be opened here? Only if the previous paragraph has the SAME wrapper tag
287 // (usually, they won't have the same layout) and the CURRENT one allows merging.
288 // The main use case is author information in several paragraphs: if the name of the author is the
289 // first paragraph of an author, then merging with the previous tag does not make sense. Say the
290 // next paragraph is the affiliation, then it should be output in the same <author> tag (different
291 // layout, same wrapper tag).
292 bool openWrapper = lay.docbookwrappertag() != "NONE";
293 if (prevpar != nullptr) {
294 Layout const & prevlay = prevpar->layout();
295 if (prevlay.docbookwrappertag() != "NONE") {
296 if (prevlay.docbookwrappertag() == lay.docbookwrappertag())
297 openWrapper = !lay.docbookwrappermergewithprevious();
305 openTag(xs, lay.docbookwrappertag(), lay.docbookwrapperattr(), lay.docbookwrappertagtype());
307 const string & tag = lay.docbooktag();
309 auto xmltag = xml::ParTag(tag, lay.docbookattr());
310 if (!xs.isTagOpen(xmltag, 1)) // Don't nest a paragraph directly in a paragraph.
311 // TODO: required or not?
312 // TODO: avoid creating a ParTag object just for this query...
313 openTag(xs, lay.docbooktag(), lay.docbookattr(), lay.docbooktagtype());
316 openTag(xs, lay.docbookitemtag(), lay.docbookitemattr(), lay.docbookitemtagtype());
317 openTag(xs, lay.docbookiteminnertag(), lay.docbookiteminnerattr(), lay.docbookiteminnertagtype());
321 void closeParTag(XMLStream & xs, Paragraph const * par, Paragraph const * nextpar)
326 // See comment in openParTag.
327 Layout const & lay = par->layout();
328 bool closeWrapper = lay.docbookwrappertag() != "NONE";
329 if (nextpar != nullptr) {
330 Layout const & nextlay = nextpar->layout();
331 if (nextlay.docbookwrappertag() != "NONE") {
332 if (nextlay.docbookwrappertag() == lay.docbookwrappertag())
333 closeWrapper = !nextlay.docbookwrappermergewithprevious();
340 closeTag(xs, lay.docbookiteminnertag(), lay.docbookiteminnertagtype());
341 closeTag(xs, lay.docbookitemtag(), lay.docbookitemtagtype());
342 closeTag(xs, lay.docbooktag(), lay.docbooktagtype());
344 closeTag(xs, lay.docbookwrappertag(), lay.docbookwrappertagtype());
348 void openLabelTag(XMLStream & xs, Layout const & lay) // Mostly for definition lists.
350 openTag(xs, lay.docbookitemlabeltag(), lay.docbookitemlabelattr(), lay.docbookitemlabeltagtype());
354 void closeLabelTag(XMLStream & xs, Layout const & lay)
356 closeTag(xs, lay.docbookitemlabeltag(), lay.docbookitemlabeltagtype());
360 void openItemTag(XMLStream & xs, Layout const & lay)
362 openTag(xs, lay.docbookitemtag(), lay.docbookitemattr(), lay.docbookitemtagtype());
366 void closeItemTag(XMLStream & xs, Layout const & lay)
368 closeTag(xs, lay.docbookitemtag(), lay.docbookitemtagtype());
372 ParagraphList::const_iterator makeAny(Text const &,
375 OutputParams const &,
376 ParagraphList::const_iterator);
379 void makeBibliography(
383 OutputParams const & runparams,
384 ParagraphList::const_iterator const & par)
386 // If this is the first paragraph in a bibliography, open the bibliography tag.
387 auto pbegin_before = text.paragraphs().getParagraphBefore(par);
388 if (pbegin_before->layout().latextype != LATEX_BIB_ENVIRONMENT) {
389 xs << xml::StartTag("bibliography");
393 // Start the precooked bibliography entry. This is very much like opening a paragraph tag.
394 // Don't forget the citation ID!
396 for (auto i = 0; i < par->size(); ++i) {
397 Inset const *ip = par->getInset(i);
400 if (const auto * bibitem = dynamic_cast<const InsetBibitem*>(ip)) {
401 attr = from_utf8("xml:id='") + bibitem->getParam("key") + from_utf8("'");
405 xs << xml::StartTag(from_utf8("bibliomixed"), attr);
407 // Generate the entry. Concatenate the different parts of the paragraph if any.
408 auto const begin = text.paragraphs().begin();
409 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(std::distance(begin, par)), 0);
410 for (auto & parXML : pars)
411 xs << XMLStream::ESCAPE_NONE << parXML;
413 // End the precooked bibliography entry.
414 xs << xml::EndTag("bibliomixed");
417 // If this is the last paragraph in a bibliography, close the bibliography tag.
418 auto const end = text.paragraphs().end();
421 bool endBibliography = nextpar == end || nextpar->layout().latextype != LATEX_BIB_ENVIRONMENT;
423 if (endBibliography) {
424 xs << xml::EndTag("bibliography");
434 OutputParams const & runparams,
435 ParagraphList::const_iterator const & par)
437 auto const begin = text.paragraphs().begin();
438 auto const end = text.paragraphs().end();
439 auto prevpar = text.paragraphs().getParagraphBefore(par);
441 // We want to open the paragraph tag if:
442 // (i) the current layout permits multiple paragraphs
443 // (ii) we are either not already inside a paragraph (HTMLIsBlock) OR
444 // we are, but this is not the first paragraph
446 // But there is also a special case, and we first see whether we are in it.
447 // We do not want to open the paragraph tag if this paragraph contains
448 // only one item, and that item is "inline", i.e., not HTMLIsBlock (such
449 // as a branch). On the other hand, if that single item has a font change
450 // applied to it, then we still do need to open the paragraph.
452 // Obviously, this is very fragile. The main reason we need to do this is
453 // because of branches, e.g., a branch that contains an entire new section.
454 // We do not really want to wrap that whole thing in a <div>...</div>.
455 bool special_case = false;
456 Inset const *specinset = par->size() == 1 ? par->getInset(0) : nullptr;
457 if (specinset && !specinset->getLayout().htmlisblock()) { // TODO: Convert htmlisblock to a DocBook parameter?
458 Layout const &style = par->layout();
459 FontInfo const first_font = style.labeltype == LABEL_MANUAL ?
460 style.labelfont : style.font;
461 FontInfo const our_font =
462 par->getFont(buf.masterBuffer()->params(), 0,
463 text.outerFont(std::distance(begin, par))).fontInfo();
465 if (first_font == our_font)
469 size_t nInsets = std::distance(par->insetList().begin(), par->insetList().end());
471 // Plain layouts must be ignored.
472 special_case |= buf.params().documentClass().isPlainLayout(par->layout()) && !runparams.docbook_force_pars;
473 // Equations do not deserve their own paragraph (DocBook allows them outside paragraphs).
474 special_case |= nInsets == par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
475 return inset.inset && inset.inset->asInsetMath();
478 // TODO: Could get rid of this with a DocBook equivalent to htmlisblock? Not for all cases, unfortunately... See above for those that have been determined not to be allowable for this potential refactoring.
479 if (!special_case && par->size() == 1 && par->getInset(0)) {
480 Inset const * firstInset = par->getInset(0);
482 // Floats cannot be in paragraphs.
483 special_case = to_utf8(firstInset->layoutName()).substr(0, 6) == "Float:";
485 // Bibliographies cannot be in paragraphs.
486 if (!special_case && firstInset->asInsetCommand())
487 special_case = firstInset->asInsetCommand()->params().getCmdName() == "bibtex";
489 // ERTs are in comments, not paragraphs.
490 if (!special_case && firstInset->lyxCode() == lyx::ERT_CODE)
493 // Listings should not get into their own paragraph.
494 if (!special_case && firstInset->lyxCode() == lyx::LISTINGS_CODE)
498 bool const open_par = runparams.docbook_make_pars
499 && !runparams.docbook_in_par
502 // We want to issue the closing tag if either:
503 // (i) We opened it, and either docbook_in_par is false,
504 // or we're not in the last paragraph, anyway.
505 // (ii) We didn't open it and docbook_in_par is true,
506 // but we are in the first par, and there is a next par.
507 bool const close_par = open_par && (!runparams.docbook_in_par);
509 // Determine if this paragraph has some real content. Things like new pages are not caught
510 // by Paragraph::empty(), even though they do not generate anything useful in DocBook.
511 // Thus, remove all spaces (including new lines: \r, \n) before checking for emptiness.
512 // std::all_of allows doing this check without having to copy the string.
513 // Open and close tags around each contained paragraph.
516 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(distance(begin, par)), 0, nextpar == end);
517 for (auto & parXML : pars) {
518 if (!std::all_of(parXML.begin(), parXML.end(), ::isspace)) {
520 openParTag(xs, &*par, prevpar);
522 xs << XMLStream::ESCAPE_NONE << parXML;
525 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
531 void makeEnvironment(Text const &text,
534 OutputParams const &runparams,
535 ParagraphList::const_iterator const & par)
537 // TODO: simplify me!
538 auto const end = text.paragraphs().end();
540 // Output the opening tag for this environment, but only if it has not been previously opened (condition
541 // implemented in openParTag).
542 auto prevpar = text.paragraphs().getParagraphBefore(par);
543 openParTag(xs, &*par, prevpar); // TODO: switch in layout for par/block?
545 // Generate the contents of this environment. There is a special case if this is like some environment.
546 Layout const & style = par->layout();
547 if (style.latextype == LATEX_COMMAND) {
548 // Nothing to do (otherwise, infinite loops).
549 } else if (style.latextype == LATEX_ENVIRONMENT) {
550 // Open a wrapper tag if needed.
551 if (style.docbookitemwrappertag() != "NONE")
552 openTag(xs, style.docbookitemwrappertag(), style.docbookitemwrapperattr(), style.docbookitemwrappertagtype());
554 // Generate the label, if need be. If it is taken from the text, sep != 0 and corresponds to the first
555 // character after the label.
557 if (style.labeltype != LABEL_NO_LABEL && style.docbookitemlabeltag() != "NONE") {
558 // At least one condition must be met:
559 // - this environment is not a list
560 // - if this is a list, the label must not be manual (i.e. it must be taken from the layout)
561 if (style.latextype != LATEX_LIST_ENVIRONMENT || style.labeltype != LABEL_MANUAL) {
562 // Usual cases: maybe there is something specified at the layout level. Highly unlikely, though.
563 docstring const lbl = par->params().labelString();
566 openLabelTag(xs, style);
568 closeLabelTag(xs, style);
571 // Only variablelist gets here (or similar items defined as an extension in the layout).
572 openLabelTag(xs, style);
573 sep = par->firstWordDocBook(xs, runparams);
574 closeLabelTag(xs, style);
578 // Maybe the item is completely empty, i.e. if the first word ends at the end of the current paragraph
579 // AND if the next paragraph doesn't have the same depth (if there is such a paragraph).
580 // Common case: there is only the first word on the line, but there is a nested list instead
582 bool emptyItem = false;
583 if (sep == par->size()) { // If the separator is already at the end of this paragraph...
586 if (next_par == text.paragraphs().end()) // There is no next paragraph.
588 else // There is a next paragraph: check depth.
589 emptyItem = par->params().depth() >= next_par->params().depth();
593 // Avoid having an empty item, this is not valid DocBook. A single character is enough to force
594 // generation of a full <para>.
595 // TODO: this always worked only by magic...
598 // Generate the rest of the paragraph, if need be. Open as many inner tags as necessary.
599 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(std::distance(text.paragraphs().begin(), par)), sep);
600 auto p = pars.begin();
602 xs << XMLStream::ESCAPE_NONE << *p;
604 if (p != pars.end()) {
605 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
606 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(), par->layout().docbookiteminnertagtype());
612 makeAny(text, buf, xs, runparams, par);
615 // Close the environment.
618 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr); // TODO: switch in layout for par/block?
622 ParagraphList::const_iterator findEndOfEnvironment(
623 ParagraphList::const_iterator const & pstart,
624 ParagraphList::const_iterator const & pend)
626 // Copy-paste from XHTML. Should be factored out at some point...
627 ParagraphList::const_iterator p = pstart;
628 Layout const & bstyle = p->layout();
629 size_t const depth = p->params().depth();
630 for (++p; p != pend; ++p) {
631 Layout const & style = p->layout();
632 // It shouldn't happen that e.g. a section command occurs inside
633 // a quotation environment, at a higher depth, but as of 6/2009,
634 // it can happen. We pretend that it's just at lowest depth.
635 if (style.latextype == LATEX_COMMAND)
638 // If depth is down, we're done
639 if (p->params().depth() < depth)
642 // If depth is up, we're not done
643 if (p->params().depth() > depth)
646 // FIXME I am not sure about the first check.
647 // Surely we *could* have different layouts that count as
648 // LATEX_PARAGRAPH, right?
649 if (style.latextype == LATEX_PARAGRAPH || style != bstyle)
656 ParagraphList::const_iterator makeListEnvironment(Text const &text,
659 OutputParams const &runparams,
660 ParagraphList::const_iterator const & begin)
663 auto const end = text.paragraphs().end();
664 auto const envend = findEndOfEnvironment(par, end);
666 // Output the opening tag for this environment.
667 Layout const & envstyle = par->layout();
668 openTag(xs, envstyle.docbookwrappertag(), envstyle.docbookwrapperattr(), envstyle.docbookwrappertagtype());
669 openTag(xs, envstyle.docbooktag(), envstyle.docbookattr(), envstyle.docbooktagtype());
671 // Handle the content of the list environment, item by item.
672 while (par != envend) {
673 Layout const & style = par->layout();
675 // Open the item wrapper.
676 openTag(xs, style.docbookitemwrappertag(), style.docbookitemwrapperattr(), style.docbookitemwrappertagtype());
678 // Generate the label, if need be. If it is taken from the text, sep != 0 and corresponds to the first
679 // character after the label.
681 if (style.labeltype != LABEL_NO_LABEL && style.docbookitemlabeltag() != "NONE") {
682 if (style.labeltype == LABEL_MANUAL) {
683 // Only variablelist gets here (or similar items defined as an extension in the layout).
684 openLabelTag(xs, style);
685 sep = 1 + par->firstWordDocBook(xs, runparams);
686 closeLabelTag(xs, style);
688 // Usual cases: maybe there is something specified at the layout level. Highly unlikely, though.
689 docstring const lbl = par->params().labelString();
692 openLabelTag(xs, style);
694 closeLabelTag(xs, style);
699 // Open the item (after the wrapper and the label).
700 openTag(xs, style.docbookitemtag(), style.docbookitemattr(), style.docbookitemtagtype());
702 // Generate the content of the item.
703 if (sep < par->size()) {
704 auto pars = par->simpleDocBookOnePar(buf, runparams,
705 text.outerFont(std::distance(text.paragraphs().begin(), par)), sep);
706 for (auto &p : pars) {
707 openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
708 par->layout().docbookiteminnertagtype());
709 xs << XMLStream::ESCAPE_NONE << p;
710 closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
713 // DocBook doesn't like emptiness.
714 compTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
715 par->layout().docbookiteminnertagtype());
718 // If the next item is deeper, it must go entirely within this item (do it recursively).
719 // By construction, with findEndOfEnvironment, depth can only stay constant or increase, never decrease.
720 depth_type currentDepth = par->getDepth();
722 while (par != envend && par->getDepth() != currentDepth)
723 par = makeAny(text, buf, xs, runparams, par);
724 // Usually, this loop only makes one iteration, except in complex scenarios, like an item with a paragraph,
725 // a list, and another paragraph; or an item with two types of list (itemise then enumerate, for instance).
728 closeTag(xs, style.docbookitemtag(), style.docbookitemtagtype());
729 closeTag(xs, style.docbookitemwrappertag(), style.docbookitemwrappertagtype());
732 // Close this environment in exactly the same way as it was opened.
733 closeTag(xs, envstyle.docbooktag(), envstyle.docbooktagtype());
734 closeTag(xs, envstyle.docbookwrappertag(), envstyle.docbookwrappertagtype());
744 OutputParams const & runparams,
745 ParagraphList::const_iterator const & par)
747 // Unlike XHTML, no need for labels, as they are handled by DocBook tags.
748 auto const begin = text.paragraphs().begin();
749 auto const end = text.paragraphs().end();
753 // Generate this command.
754 auto prevpar = text.paragraphs().getParagraphBefore(par);
755 openParTag(xs, &*par, prevpar);
757 auto pars = par->simpleDocBookOnePar(buf, runparams,text.outerFont(distance(begin, par)));
758 for (auto & parXML : pars)
759 // TODO: decide what to do with openParTag/closeParTag in new lines.
760 xs << XMLStream::ESCAPE_NONE << parXML;
762 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
766 ParagraphList::const_iterator makeAny(Text const &text,
769 OutputParams const &runparams,
770 ParagraphList::const_iterator par)
772 switch (par->layout().latextype) {
774 makeCommand(text, buf, xs, runparams, par);
776 case LATEX_ENVIRONMENT:
777 makeEnvironment(text, buf, xs, runparams, par);
779 case LATEX_LIST_ENVIRONMENT:
780 case LATEX_ITEM_ENVIRONMENT:
781 // Only case when makeAny() might consume more than one paragraph.
782 return makeListEnvironment(text, buf, xs, runparams, par);
783 case LATEX_PARAGRAPH:
784 makeParagraph(text, buf, xs, runparams, par);
786 case LATEX_BIB_ENVIRONMENT:
787 makeBibliography(text, buf, xs, runparams, par);
795 bool isLayoutSectioning(Layout const & lay)
797 return lay.category() == from_utf8("Sectioning");
801 using DocBookDocumentSectioning = tuple<bool, pit_type>;
804 struct DocBookInfoTag
806 const set<pit_type> shouldBeInInfo;
807 const set<pit_type> mustBeInInfo; // With the notable exception of the abstract!
808 const set<pit_type> abstract;
809 const bool abstractLayout;
813 DocBookInfoTag(const set<pit_type> & shouldBeInInfo, const set<pit_type> & mustBeInInfo,
814 const set<pit_type> & abstract, bool abstractLayout, pit_type bpit, pit_type epit) :
815 shouldBeInInfo(shouldBeInInfo), mustBeInInfo(mustBeInInfo), abstract(abstract),
816 abstractLayout(abstractLayout), bpit(bpit), epit(epit) {}
820 DocBookDocumentSectioning hasDocumentSectioning(ParagraphList const ¶graphs, pit_type bpit, pit_type const epit) {
821 bool documentHasSections = false;
823 while (bpit < epit) {
824 Layout const &style = paragraphs[bpit].layout();
825 documentHasSections |= isLayoutSectioning(style);
827 if (documentHasSections)
831 // Paragraphs before the first section: [ runparams.par_begin ; eppit )
833 return make_tuple(documentHasSections, bpit);
837 bool hasOnlyNotes(Paragraph const & par)
839 // Precondition: the paragraph is not empty. Otherwise, the function will always return true...
840 for (int i = 0; i < par.size(); ++i)
841 // If you find something that is not an inset (like actual text) or an inset that is not a note,
843 if (!par.isInset(i) || par.getInset(i)->lyxCode() != NOTE_CODE)
849 DocBookInfoTag getParagraphsWithInfo(ParagraphList const ¶graphs,
850 pit_type bpit, pit_type const epit,
851 // Typically, bpit is the beginning of the document and epit the end *or* the first section.
852 bool documentHasSections) {
853 set<pit_type> shouldBeInInfo;
854 set<pit_type> mustBeInInfo;
855 set<pit_type> abstractWithLayout;
856 set<pit_type> abstractNoLayout;
858 // Find the first non empty paragraph by mutating bpit.
859 while (bpit < epit) {
860 Paragraph const &par = paragraphs[bpit];
861 if (par.empty() || hasOnlyNotes(par))
867 // Traverse everything that might belong to <info>.
868 bool hasAbstractLayout = false;
869 pit_type cpit = bpit;
870 for (; cpit < epit; ++cpit) {
871 // Skip paragraphs that don't generate anything in DocBook.
872 Paragraph const & par = paragraphs[cpit];
873 if (hasOnlyNotes(par))
876 // There should never be any section here. (Just a sanity check: if this fails, this function could end up
877 // processing the whole document.)
878 if (isLayoutSectioning(par.layout())) {
879 LYXERR0("Assertion failed: section found in potential <info> paragraphs.");
883 // If this is marked as an abstract by the layout, put it in the right set.
884 if (par.layout().docbookabstract()) {
885 hasAbstractLayout = true;
886 abstractWithLayout.emplace(cpit);
890 // Based on layout information, store this paragraph in one set: should be in <info>, must be,
891 // or abstract (either because of layout or of position).
892 Layout const &style = par.layout();
894 if (style.docbookininfo() == "always")
895 mustBeInInfo.emplace(cpit);
896 else if (style.docbookininfo() == "maybe")
897 shouldBeInInfo.emplace(cpit);
898 else if (documentHasSections && !hasAbstractLayout)
899 abstractNoLayout.emplace(cpit);
900 else // This should definitely not be in <info>.
903 // Now, cpit points to the first paragraph that no more has things that could go in <info>.
904 // bpit is the beginning of the <info> part.
906 return DocBookInfoTag(shouldBeInInfo, mustBeInInfo,
907 hasAbstractLayout ? abstractWithLayout : abstractNoLayout,
908 hasAbstractLayout, bpit, cpit);
911 } // end anonymous namespace
914 xml::FontTag docbookStartFontTag(xml::FontTypes type)
916 return xml::FontTag(from_utf8(fontToDocBookTag(type)), from_utf8(fontToAttribute(type)), type);
920 xml::EndFontTag docbookEndFontTag(xml::FontTypes type)
922 return xml::EndFontTag(from_utf8(fontToDocBookTag(type)), type);
926 void outputDocBookInfo(
930 OutputParams const & runparams,
931 ParagraphList const & paragraphs,
932 DocBookInfoTag const & info)
934 // Perform an additional check on the abstract. Sometimes, there are many paragraphs that should go
935 // into the abstract, but none generates actual content. Thus, first generate to a temporary stream,
936 // then only create the <abstract> tag if these paragraphs generate some content.
937 // This check must be performed *before* a decision on whether or not to output <info> is made.
938 bool hasAbstract = !info.abstract.empty();
941 // Generate the abstract XML into a string before further checks.
942 odocstringstream os2;
944 for (auto const & p : info.abstract)
945 makeAny(text, buf, xs2, runparams, paragraphs.iterator_at(p));
947 // Actually output the abstract if there is something to do. Don't count line feeds or spaces in this,
948 // even though they must be properly output if there is some abstract.
949 abstract = os2.str();
950 docstring cleaned = abstract;
951 cleaned.erase(std::remove_if(cleaned.begin(), cleaned.end(), ::isspace), cleaned.end());
953 // Nothing? Then there is no abstract!
958 // The abstract must go in <info>. Otherwise, decide whether to open <info> based on the layouts.
959 bool needInfo = !info.mustBeInInfo.empty() || hasAbstract;
961 // Start the <info> tag if required.
963 xs.startDivision(false);
964 xs << xml::StartTag("info");
968 // Output the elements that should go in <info>, before and after the abstract.
969 for (auto pit : info.shouldBeInInfo) // Typically, the title: these elements are so important and ubiquitous
970 // that mandating a wrapper like <info> would repel users. Thus, generate them first.
971 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
972 for (auto pit : info.mustBeInInfo)
973 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
975 // Always output the abstract as the last item of the <info>, as it requires special treatment (especially if
976 // it contains several paragraphs that are empty).
978 if (info.abstractLayout) {
979 xs << XMLStream::ESCAPE_NONE << abstract;
982 string tag = paragraphs[*info.abstract.begin()].layout().docbookforceabstracttag();
986 if (!xs.isLastTagCR())
989 xs << xml::StartTag(tag);
991 xs << XMLStream::ESCAPE_NONE << abstract;
992 xs << xml::EndTag(tag);
997 // End the <info> tag if it was started.
999 if (!xs.isLastTagCR())
1002 xs << xml::EndTag("info");
1009 void docbookSimpleAllParagraphs(
1013 OutputParams const & runparams)
1015 // Handle the given text, supposing it has no sections (i.e. a "simple" text). The input may vary in length
1016 // between a single paragraph to a whole document.
1017 pit_type const bpit = runparams.par_begin;
1018 pit_type const epit = runparams.par_end;
1019 ParagraphList const ¶graphs = text.paragraphs();
1021 // First, the <info> tag.
1022 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, epit, false);
1023 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
1025 // Then, the content. It starts where the <info> ends.
1026 auto par = paragraphs.iterator_at(info.epit);
1027 auto end = paragraphs.iterator_at(epit);
1028 while (par != end) {
1029 if (!hasOnlyNotes(*par))
1030 par = makeAny(text, buf, xs, runparams, par);
1037 void docbookParagraphs(Text const &text,
1040 OutputParams const &runparams) {
1041 ParagraphList const ¶graphs = text.paragraphs();
1042 if (runparams.par_begin == runparams.par_end) {
1043 runparams.par_begin = 0;
1044 runparams.par_end = paragraphs.size();
1046 pit_type bpit = runparams.par_begin;
1047 pit_type const epit = runparams.par_end;
1048 LASSERT(bpit < epit,
1050 xs << XMLStream::ESCAPE_NONE << "<!-- DocBook output error! -->\n";
1054 std::stack<std::pair<int, string>> headerLevels; // Used to determine when to open/close sections: store the depth
1055 // of the section and the tag that was used to open it.
1057 // Detect whether the document contains sections. If there are no sections, treatment is largely simplified.
1058 // In particular, there can't be an abstract, unless it is manually marked.
1059 bool documentHasSections;
1061 tie(documentHasSections, eppit) = hasDocumentSectioning(paragraphs, bpit, epit);
1063 // Deal with "simple" documents, i.e. those without sections.
1064 if (!documentHasSections) {
1065 docbookSimpleAllParagraphs(text, buf, xs, runparams);
1069 // Output the first <info> tag (or just the title).
1070 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, eppit, true);
1071 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
1074 // Then, iterate through the paragraphs of this document.
1075 bool currentlyInAppendix = false;
1077 auto par = text.paragraphs().iterator_at(bpit);
1078 auto end = text.paragraphs().iterator_at(epit);
1079 while (par != end) {
1080 OutputParams ourparams = runparams;
1082 if (par->params().startOfAppendix())
1083 currentlyInAppendix = true;
1084 if (hasOnlyNotes(*par)) {
1089 Layout const &style = par->layout();
1091 // Think about adding <section> and/or </section>s.
1092 if (isLayoutSectioning(style)) {
1093 int level = style.toclevel;
1095 // Need to close a previous section if it has the same level or a higher one (close <section> if opening a
1096 // <h2> after a <h2>, <h3>, <h4>, <h5> or <h6>). More examples:
1097 // - current: h2; back: h1; do not close any <section>
1098 // - current: h1; back: h2; close two <section> (first the <h2>, then the <h1>, so a new <h1> can come)
1099 while (!headerLevels.empty() && level <= headerLevels.top().first) {
1100 // Output the tag only if it corresponds to a legit section.
1101 int stackLevel = headerLevels.top().first;
1102 if (stackLevel != Layout::NOT_IN_TOC) {
1103 xs << xml::EndTag(headerLevels.top().second);
1109 // Open the new section: first push it onto the stack, then output it in DocBook.
1110 string sectionTag = (currentlyInAppendix && style.docbooksectiontag() == "chapter") ?
1111 "appendix" : style.docbooksectiontag();
1112 headerLevels.push(std::make_pair(level, sectionTag));
1114 // Some sectioning-like elements should not be output (such as FrontMatter).
1115 if (level != Layout::NOT_IN_TOC) {
1116 // Look for a label in the title, i.e. a InsetLabel as a child.
1117 docstring id = docstring();
1118 for (pos_type i = 0; i < par->size(); ++i) {
1119 Inset const *inset = par->getInset(i);
1121 if (auto label = dynamic_cast<InsetLabel const *>(inset)) {
1122 // Generate the attributes for the section if need be.
1123 id += "xml:id=\"" + xml::cleanID(label->screenLabel()) + "\"";
1125 // Don't output the ID as a DocBook <anchor>.
1126 ourparams.docbook_anchors_to_ignore.emplace(label->screenLabel());
1128 // Cannot have multiple IDs per tag.
1134 // Write the open tag for this section.
1138 xs << xml::StartTag(sectionTag, attrs);
1143 // Close all sections before the bibliography.
1144 // 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)?
1145 auto insetsLength = distance(par->insetList().begin(), par->insetList().end());
1146 if (insetsLength > 0) {
1147 Inset const *firstInset = par->getInset(0);
1148 if (firstInset && (firstInset->lyxCode() == BIBITEM_CODE || firstInset->lyxCode() == BIBTEX_CODE)) {
1149 while (!headerLevels.empty()) {
1150 int level = headerLevels.top().first;
1151 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1154 // Output the tag only if it corresponds to a legit section.
1155 if (level != Layout::NOT_IN_TOC) {
1156 xs << XMLStream::ESCAPE_NONE << tag;
1163 // Generate this paragraph.
1164 par = makeAny(text, buf, xs, ourparams, par);
1166 // Some special sections may require abstracts (mostly parts, in books).
1167 // 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.
1168 if (isLayoutSectioning(style) && style.docbookforceabstracttag() != "NONE") {
1169 // This abstract may be found between the next paragraph and the next title.
1170 pit_type cpit = std::distance(text.paragraphs().begin(), par);
1171 pit_type ppit = std::get<1>(hasDocumentSectioning(paragraphs, cpit, epit));
1173 // Generate this abstract (this code corresponds to parts of outputDocBookInfo).
1174 DocBookInfoTag secInfo = getParagraphsWithInfo(paragraphs, cpit, ppit, true);
1176 if (!secInfo.abstract.empty()) {
1177 xs << xml::StartTag(style.docbookforceabstracttag());
1179 for (auto const &p : secInfo.abstract)
1180 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(p));
1181 xs << xml::EndTag(style.docbookforceabstracttag());
1185 // Skip all the text that just has been generated.
1186 par = paragraphs.iterator_at(ppit);
1190 // If need be, close <section>s, but only at the end of the document (otherwise, dealt with at the beginning
1192 while (!headerLevels.empty() && headerLevels.top().first > Layout::NOT_IN_TOC) {
1193 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1195 xs << XMLStream::ESCAPE_NONE << tag;