2 * \file output_docbook.cpp
3 * This file is part of LyX, the document processor.
4 * Licence details can be found in the file COPYING.
6 * \author Lars Gullik Bjønnes
9 * Full author contact details are available in file CREDITS.
14 #include "output_docbook.h"
17 #include "buffer_funcs.h"
18 #include "BufferParams.h"
20 #include "InsetList.h"
21 #include "Paragraph.h"
22 #include "ParagraphList.h"
23 #include "ParagraphParameters.h"
26 #include "TextClass.h"
28 #include "insets/InsetBibtex.h"
29 #include "insets/InsetBibitem.h"
30 #include "insets/InsetLabel.h"
31 #include "mathed/InsetMath.h"
32 #include "insets/InsetNote.h"
34 #include "support/debug.h"
35 #include "support/lassert.h"
36 #include "support/textutils.h"
44 using namespace lyx::support;
50 std::string fontToDocBookTag(xml::FontTypes type)
53 case xml::FontTypes::FT_EMPH:
54 case xml::FontTypes::FT_BOLD:
56 case xml::FontTypes::FT_NOUN:
58 case xml::FontTypes::FT_UBAR:
59 case xml::FontTypes::FT_WAVE:
60 case xml::FontTypes::FT_DBAR:
61 case xml::FontTypes::FT_SOUT:
62 case xml::FontTypes::FT_XOUT:
63 case xml::FontTypes::FT_ITALIC:
64 case xml::FontTypes::FT_UPRIGHT:
65 case xml::FontTypes::FT_SLANTED:
66 case xml::FontTypes::FT_SMALLCAPS:
67 case xml::FontTypes::FT_ROMAN:
68 case xml::FontTypes::FT_SANS:
70 case xml::FontTypes::FT_TYPE:
72 case xml::FontTypes::FT_SIZE_TINY:
73 case xml::FontTypes::FT_SIZE_SCRIPT:
74 case xml::FontTypes::FT_SIZE_FOOTNOTE:
75 case xml::FontTypes::FT_SIZE_SMALL:
76 case xml::FontTypes::FT_SIZE_NORMAL:
77 case xml::FontTypes::FT_SIZE_LARGE:
78 case xml::FontTypes::FT_SIZE_LARGER:
79 case xml::FontTypes::FT_SIZE_LARGEST:
80 case xml::FontTypes::FT_SIZE_HUGE:
81 case xml::FontTypes::FT_SIZE_HUGER:
82 case xml::FontTypes::FT_SIZE_INCREASE:
83 case xml::FontTypes::FT_SIZE_DECREASE:
91 string fontToRole(xml::FontTypes type)
93 // Specific fonts are achieved with roles. The only common ones are "" for basic emphasis,
94 // and "bold"/"strong" for bold. With some specific options, other roles are copied into
95 // HTML output (via the DocBook XSLT sheets); otherwise, if not recognised, they are just ignored.
96 // Hence, it is not a problem to have many roles by default here.
97 // See https://www.sourceware.org/ml/docbook/2003-05/msg00269.html
99 case xml::FontTypes::FT_ITALIC:
100 case xml::FontTypes::FT_EMPH:
102 case xml::FontTypes::FT_BOLD:
104 case xml::FontTypes::FT_NOUN: // Outputs a <person>
105 case xml::FontTypes::FT_TYPE: // Outputs a <code>
107 case xml::FontTypes::FT_UBAR:
110 // All other roles are non-standard for DocBook.
112 case xml::FontTypes::FT_WAVE:
114 case xml::FontTypes::FT_DBAR:
116 case xml::FontTypes::FT_SOUT:
118 case xml::FontTypes::FT_XOUT:
120 case xml::FontTypes::FT_UPRIGHT:
122 case xml::FontTypes::FT_SLANTED:
124 case xml::FontTypes::FT_SMALLCAPS:
126 case xml::FontTypes::FT_ROMAN:
128 case xml::FontTypes::FT_SANS:
130 case xml::FontTypes::FT_SIZE_TINY:
132 case xml::FontTypes::FT_SIZE_SCRIPT:
133 return "size_script";
134 case xml::FontTypes::FT_SIZE_FOOTNOTE:
135 return "size_footnote";
136 case xml::FontTypes::FT_SIZE_SMALL:
138 case xml::FontTypes::FT_SIZE_NORMAL:
139 return "size_normal";
140 case xml::FontTypes::FT_SIZE_LARGE:
142 case xml::FontTypes::FT_SIZE_LARGER:
143 return "size_larger";
144 case xml::FontTypes::FT_SIZE_LARGEST:
145 return "size_largest";
146 case xml::FontTypes::FT_SIZE_HUGE:
148 case xml::FontTypes::FT_SIZE_HUGER:
150 case xml::FontTypes::FT_SIZE_INCREASE:
151 return "size_increase";
152 case xml::FontTypes::FT_SIZE_DECREASE:
153 return "size_decrease";
160 string fontToAttribute(xml::FontTypes type) {
161 // If there is a role (i.e. nonstandard use of a tag), output the attribute. Otherwise, the sheer tag is sufficient
163 string role = fontToRole(type);
165 return "role='" + role + "'";
171 // Higher-level convenience functions.
173 void openParTag(XMLStream & xs, const Paragraph * par, const Paragraph * prevpar, const OutputParams & runparams)
178 // When should the wrapper be opened here? Only if the previous paragraph has the SAME wrapper tag
179 // (usually, they won't have the same layout) and the CURRENT one allows merging.
180 // The main use case is author information in several paragraphs: if the name of the author is the
181 // first paragraph of an author, then merging with the previous tag does not make sense. Say the
182 // next paragraph is the affiliation, then it should be output in the same <author> tag (different
183 // layout, same wrapper tag).
184 Layout const & lay = par->layout();
185 bool openWrapper = lay.docbookwrappertag() != "NONE" && !runparams.docbook_ignore_wrapper;
187 if (prevpar != nullptr && !runparams.docbook_ignore_wrapper) {
188 Layout const & prevlay = prevpar->layout();
189 if (prevlay.docbookwrappertag() != "NONE") {
190 if (prevlay.docbookwrappertag() == lay.docbookwrappertag() &&
191 prevlay.docbookwrapperattr() == lay.docbookwrapperattr())
192 openWrapper = !lay.docbookwrappermergewithprevious();
200 xml::openTag(xs, lay.docbookwrappertag(), lay.docbookwrapperattr(), lay.docbookwrappertagtype());
202 const string & tag = lay.docbooktag();
204 auto xmltag = xml::ParTag(tag, lay.docbookattr());
205 if (!xs.isTagOpen(xmltag, 1)) { // Don't nest a paragraph directly in a paragraph.
206 // TODO: required or not?
207 // TODO: avoid creating a ParTag object just for this query...
208 xml::openTag(xs, lay.docbooktag(), lay.docbookattr(), lay.docbooktagtype());
209 xml::openTag(xs, lay.docbookinnertag(), lay.docbookinnerattr(), lay.docbookinnertagtype());
213 xml::openTag(xs, lay.docbookitemwrappertag(), lay.docbookitemwrapperattr(), lay.docbookitemwrappertagtype());
214 xml::openTag(xs, lay.docbookitemtag(), lay.docbookitemattr(), lay.docbookitemtagtype());
215 xml::openTag(xs, lay.docbookiteminnertag(), lay.docbookiteminnerattr(), lay.docbookiteminnertagtype());
219 void closeParTag(XMLStream & xs, Paragraph const * par, Paragraph const * nextpar, const OutputParams & runparams)
224 // See comment in openParTag.
225 Layout const & lay = par->layout();
226 bool closeWrapper = lay.docbookwrappertag() != "NONE" && !runparams.docbook_ignore_wrapper;
228 if (nextpar != nullptr && !runparams.docbook_ignore_wrapper) {
229 Layout const & nextlay = nextpar->layout();
230 if (nextlay.docbookwrappertag() != "NONE") {
231 if (nextlay.docbookwrappertag() == lay.docbookwrappertag() &&
232 nextlay.docbookwrapperattr() == lay.docbookwrapperattr())
233 closeWrapper = !nextlay.docbookwrappermergewithprevious();
240 xml::closeTag(xs, lay.docbookiteminnertag(), lay.docbookiteminnertagtype());
241 xml::closeTag(xs, lay.docbookitemtag(), lay.docbookitemtagtype());
242 xml::closeTag(xs, lay.docbookitemwrappertag(), lay.docbookitemwrappertagtype());
243 xml::closeTag(xs, lay.docbookinnertag(), lay.docbookinnertagtype());
244 xml::closeTag(xs, lay.docbooktag(), lay.docbooktagtype());
246 xml::closeTag(xs, lay.docbookwrappertag(), lay.docbookwrappertagtype());
250 void makeBibliography(
254 OutputParams const & runparams,
255 ParagraphList::const_iterator const & par)
257 // If this is the first paragraph in a bibliography, open the bibliography tag.
258 auto const * pbegin_before = text.paragraphs().getParagraphBefore(par);
259 if (pbegin_before == nullptr || (pbegin_before && pbegin_before->layout().latextype != LATEX_BIB_ENVIRONMENT)) {
260 xs << xml::StartTag("bibliography");
264 // Start the precooked bibliography entry. This is very much like opening a paragraph tag.
265 // Don't forget the citation ID!
267 for (auto i = 0; i < par->size(); ++i) {
268 Inset const *ip = par->getInset(i);
271 if (const auto * bibitem = dynamic_cast<const InsetBibitem*>(ip)) {
272 auto id = xml::cleanID(bibitem->getParam("key"));
273 attr = from_utf8("xml:id='") + id + from_utf8("'");
277 xs << xml::StartTag(from_utf8("bibliomixed"), attr);
279 // Generate the entry. Concatenate the different parts of the paragraph if any.
280 auto const begin = text.paragraphs().begin();
281 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(std::distance(begin, par)), 0);
282 for (auto & parXML : pars)
283 xs << XMLStream::ESCAPE_NONE << parXML;
285 // End the precooked bibliography entry.
286 xs << xml::EndTag("bibliomixed");
289 // If this is the last paragraph in a bibliography, close the bibliography tag.
290 auto const end = text.paragraphs().end();
293 bool endBibliography = nextpar == end || nextpar->layout().latextype != LATEX_BIB_ENVIRONMENT;
295 if (endBibliography) {
296 xs << xml::EndTag("bibliography");
306 OutputParams const & runparams,
307 ParagraphList::const_iterator const & par)
310 auto const begin = text.paragraphs().begin();
311 auto const end = text.paragraphs().end();
312 auto prevpar = text.paragraphs().getParagraphBefore(par);
314 // We want to open the paragraph tag if:
315 // (i) the current layout permits multiple paragraphs
316 // (ii) we are either not already inside a paragraph (HTMLIsBlock) OR
317 // we are, but this is not the first paragraph
319 // But there is also a special case, and we first see whether we are in it.
320 // We do not want to open the paragraph tag if this paragraph contains
321 // only one item, and that item is "inline", i.e., not HTMLIsBlock (such
322 // as a branch). On the other hand, if that single item has a font change
323 // applied to it, then we still do need to open the paragraph.
325 // Obviously, this is very fragile. The main reason we need to do this is
326 // because of branches, e.g., a branch that contains an entire new section.
327 // We do not really want to wrap that whole thing in a <div>...</div>.
328 bool special_case = false;
329 Inset const *specinset = par->size() == 1 ? par->getInset(0) : nullptr;
330 if (specinset && !specinset->getLayout().htmlisblock()) { // TODO: Convert htmlisblock to a DocBook parameter?
331 Layout const &style = par->layout();
332 FontInfo const first_font = style.labeltype == LABEL_MANUAL ?
333 style.labelfont : style.font;
334 FontInfo const our_font =
335 par->getFont(buf.masterBuffer()->params(), 0,
336 text.outerFont(std::distance(begin, par))).fontInfo();
338 if (first_font == our_font)
342 size_t nInsets = std::distance(par->insetList().begin(), par->insetList().end());
343 auto parSize = (size_t) par->size();
345 // If this LyX code does not produce any output, it can be safely ignored in the following checks: if this thing
346 // is present in the paragraph, it has no impact on the definition of the special case (i.e. whether or not
347 // a <para> tag should be output).
348 auto isLyxCodeToIgnore = [](InsetCode x) { return x == TOC_CODE || x == NOTE_CODE; };
350 // TODO: if a paragraph *only* contains floats, listings, bibliographies, etc., should this be considered as a
351 // special case? If so, the code could be largely simplifies (all the calls to all_of, basically) and optimised
352 // at the compilation stage.
354 // Plain layouts must be ignored.
355 special_case |= buf.params().documentClass().isPlainLayout(par->layout()) && !runparams.docbook_force_pars;
356 // Equations do not deserve their own paragraph (DocBook allows them outside paragraphs).
357 // Exception: any case that generates an <inlineequation> must still get a paragraph to be valid.
358 special_case |= nInsets == parSize && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
359 return inset.inset && inset.inset->asInsetMath() && inset.inset->asInsetMath()->getType() != hullSimple;
361 // Tables do not deserve their own paragraphs (DocBook allows them outside paragraphs).
362 special_case |= nInsets == parSize && std::all_of(par->insetList().begin(), par->insetList().end(), [isLyxCodeToIgnore](InsetList::Element inset) {
363 return inset.inset->lyxCode() == TABULAR_CODE || isLyxCodeToIgnore(inset.inset->lyxCode());
365 // Floats cannot be in paragraphs.
366 special_case |= nInsets == parSize && std::all_of(par->insetList().begin(), par->insetList().end(), [isLyxCodeToIgnore](InsetList::Element inset) {
367 return inset.inset->lyxCode() == FLOAT_CODE || isLyxCodeToIgnore(inset.inset->lyxCode());
369 // Bibliographies cannot be in paragraphs. Bibitems should still be handled as paragraphs, though
370 // (see makeParagraphBibliography).
371 special_case |= nInsets == parSize && std::all_of(par->insetList().begin(), par->insetList().end(), [isLyxCodeToIgnore](InsetList::Element inset) {
372 return inset.inset->lyxCode() == BIBTEX_CODE || isLyxCodeToIgnore(inset.inset->lyxCode());
374 // ERTs are in comments, not paragraphs.
375 special_case |= nInsets == parSize && std::all_of(par->insetList().begin(), par->insetList().end(), [isLyxCodeToIgnore](InsetList::Element inset) {
376 return inset.inset->lyxCode() == ERT_CODE || isLyxCodeToIgnore(inset.inset->lyxCode());
378 // Listings should not get into their own paragraph.
379 special_case |= nInsets == parSize && std::all_of(par->insetList().begin(), par->insetList().end(), [isLyxCodeToIgnore](InsetList::Element inset) {
380 return inset.inset->lyxCode() == LISTINGS_CODE || isLyxCodeToIgnore(inset.inset->lyxCode());
382 // Boxes cannot get into their own paragraph.
383 special_case |= nInsets == parSize && std::all_of(par->insetList().begin(), par->insetList().end(), [isLyxCodeToIgnore](InsetList::Element inset) {
384 return inset.inset->lyxCode() == BOX_CODE || isLyxCodeToIgnore(inset.inset->lyxCode());
386 // Includes should not have a paragraph.
387 special_case |= nInsets == parSize && std::all_of(par->insetList().begin(), par->insetList().end(), [isLyxCodeToIgnore](InsetList::Element inset) {
388 return inset.inset->lyxCode() == INCLUDE_CODE || isLyxCodeToIgnore(inset.inset->lyxCode());
390 // Glossaries should not have a paragraph.
391 special_case |= nInsets == parSize && std::all_of(par->insetList().begin(), par->insetList().end(), [isLyxCodeToIgnore](InsetList::Element inset) {
392 return inset.inset->lyxCode() == NOMENCL_PRINT_CODE || isLyxCodeToIgnore(inset.inset->lyxCode());
395 bool const open_par = runparams.docbook_make_pars
396 && !runparams.docbook_in_par
399 // We want to issue the closing tag if either:
400 // (i) We opened it, and either docbook_in_par is false,
401 // or we're not in the last paragraph, anyway.
402 // (ii) We didn't open it and docbook_in_par is true,
403 // but we are in the first par, and there is a next par.
404 bool const close_par = open_par && !runparams.docbook_in_par;
406 // Determine if this paragraph has some real content. Things like new pages are not caught
407 // by Paragraph::empty(), even though they do not generate anything useful in DocBook.
408 // Thus, remove all spaces (including new lines: \r, \n) before checking for emptiness.
409 // std::all_of allows doing this check without having to copy the string.
410 // Open and close tags around each contained paragraph.
413 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(distance(begin, par)), 0, nextpar == end, special_case);
414 for (docstring const & parXML : pars) {
415 if (!xml::isNotOnlySpace(parXML))
419 openParTag(xs, &*par, prevpar, runparams);
421 xs << XMLStream::ESCAPE_NONE << parXML;
424 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr, runparams);
429 void makeEnvironment(Text const &text,
432 OutputParams const &runparams,
433 ParagraphList::const_iterator const & par)
436 auto const end = text.paragraphs().end();
440 // Special cases for listing-like environments provided in layouts. This is quite ad-hoc, but provides a useful
441 // default. This should not be used by too many environments (only LyX-Code right now).
442 // This would be much simpler if LyX-Code was implemented as InsetListings...
443 bool mimicListing = false;
444 bool ignoreFonts = false;
445 if (par->layout().docbooktag() == "programlisting") {
450 // Output the opening tag for this environment, but only if it has not been previously opened (condition
451 // implemented in openParTag).
452 auto prevpar = text.paragraphs().getParagraphBefore(par);
453 openParTag(xs, &*par, prevpar, runparams);
455 // Generate the contents of this environment. There is a special case if this is like some environment.
456 Layout const & style = par->layout();
457 if (style.latextype == LATEX_COMMAND) {
458 // Nothing to do (otherwise, infinite loops).
459 } else if (style.latextype == LATEX_ENVIRONMENT) {
460 // Generate the paragraph, if need be.
461 auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(std::distance(text.paragraphs().begin(), par)), 0, false, ignoreFonts);
464 auto p = pars.begin();
465 while (p != pars.end()) {
466 xml::openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
467 par->layout().docbookiteminnertagtype());
468 xs << XMLStream::ESCAPE_NONE << *p;
469 xml::closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
472 // Insert a new line after each "paragraph" (i.e. line in the listing), except for the last one.
473 // Otherwise, there would one more new line in the output than in the LyX document.
478 for (auto const & p : pars) {
479 xml::openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
480 par->layout().docbookiteminnertagtype());
481 xs << XMLStream::ESCAPE_NONE << p;
482 xml::closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
486 makeAny(text, buf, xs, runparams, par);
489 // Close the environment.
490 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr, runparams);
494 ParagraphList::const_iterator findEndOfEnvironment(
495 ParagraphList::const_iterator const & pstart,
496 ParagraphList::const_iterator const & pend)
498 // Copy-paste from XHTML. Should be factored out at some point...
499 ParagraphList::const_iterator p = pstart;
500 Layout const & bstyle = p->layout();
501 size_t const depth = p->params().depth();
502 for (++p; p != pend; ++p) {
503 Layout const & style = p->layout();
504 // It shouldn't happen that e.g. a section command occurs inside
505 // a quotation environment, at a higher depth, but as of 6/2009,
506 // it can happen. We pretend that it's just at lowest depth.
507 if (style.latextype == LATEX_COMMAND)
510 // If depth is down, we're done
511 if (p->params().depth() < depth)
514 // If depth is up, we're not done
515 if (p->params().depth() > depth)
518 // FIXME I am not sure about the first check.
519 // Surely we *could* have different layouts that count as
520 // LATEX_PARAGRAPH, right?
521 if (style.latextype == LATEX_PARAGRAPH || style != bstyle)
528 ParagraphList::const_iterator makeListEnvironment(Text const &text,
531 OutputParams const &runparams,
532 ParagraphList::const_iterator const & begin)
536 auto const end = text.paragraphs().end();
537 auto const envend = findEndOfEnvironment(par, end);
539 // Output the opening tag for this environment.
540 Layout const & envstyle = par->layout();
541 xml::openTag(xs, envstyle.docbookwrappertag(), envstyle.docbookwrapperattr(), envstyle.docbookwrappertagtype());
542 xml::openTag(xs, envstyle.docbooktag(), envstyle.docbookattr(), envstyle.docbooktagtype());
544 // Handle the content of the list environment, item by item.
545 while (par != envend) {
546 // Skip this paragraph if it is both empty and the last one (otherwise, there may be deeper paragraphs after).
549 if (par->empty() && nextpar == envend)
552 // Open the item wrapper.
553 Layout const & style = par->layout();
554 xml::openTag(xs, style.docbookitemwrappertag(), style.docbookitemwrapperattr(),
555 style.docbookitemwrappertagtype());
557 // Generate the label, if need be. If it is taken from the text, sep != 0 and corresponds to the first
558 // character after the label.
560 if (style.labeltype != LABEL_NO_LABEL && style.docbookitemlabeltag() != "NONE") {
561 if (style.labeltype == LABEL_MANUAL) {
562 // Only variablelist gets here (or similar items defined as an extension in the layout).
563 xml::openTag(xs, style.docbookitemlabeltag(), style.docbookitemlabelattr(),
564 style.docbookitemlabeltagtype());
565 sep = 1 + par->firstWordDocBook(xs, runparams);
566 xml::closeTag(xs, style.docbookitemlabeltag(), style.docbookitemlabeltagtype());
568 // Usual cases: maybe there is something specified at the layout level. Highly unlikely, though.
569 docstring const lbl = par->params().labelString();
572 xml::openTag(xs, style.docbookitemlabeltag(), style.docbookitemlabelattr(),
573 style.docbookitemlabeltagtype());
575 xml::closeTag(xs, style.docbookitemlabeltag(), style.docbookitemlabeltagtype());
580 // Open the item (after the wrapper and the label).
581 xml::openTag(xs, style.docbookitemtag(), style.docbookitemattr(), style.docbookitemtagtype());
583 // Generate the content of the item.
584 if (sep < par->size()) {
585 auto pars = par->simpleDocBookOnePar(buf, runparams,
586 text.outerFont(std::distance(text.paragraphs().begin(), par)), sep);
587 for (auto &p : pars) {
588 xml::openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
589 par->layout().docbookiteminnertagtype());
590 xs << XMLStream::ESCAPE_NONE << p;
591 xml::closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
594 // DocBook doesn't like emptiness.
595 xml::compTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
596 par->layout().docbookiteminnertagtype());
599 // If the next item is deeper, it must go entirely within this item (do it recursively).
600 // By construction, with findEndOfEnvironment, depth can only stay constant or increase, never decrease.
601 depth_type currentDepth = par->getDepth();
603 while (par != envend && par->getDepth() != currentDepth)
604 par = makeAny(text, buf, xs, runparams, par);
605 // Usually, this loop only makes one iteration, except in complex scenarios, like an item with a paragraph,
606 // a list, and another paragraph; or an item with two types of list (itemise then enumerate, for instance).
609 xml::closeTag(xs, style.docbookitemtag(), style.docbookitemtagtype());
610 xml::closeTag(xs, style.docbookitemwrappertag(), style.docbookitemwrappertagtype());
613 // Close this environment in exactly the same way as it was opened.
614 xml::closeTag(xs, envstyle.docbooktag(), envstyle.docbooktagtype());
615 xml::closeTag(xs, envstyle.docbookwrappertag(), envstyle.docbookwrappertagtype());
625 OutputParams const & runparams,
626 ParagraphList::const_iterator const & par)
629 // Unlike XHTML, no need for labels, as they are handled by DocBook tags.
630 auto const begin = text.paragraphs().begin();
631 auto const end = text.paragraphs().end();
635 // Generate this command.
636 auto prevpar = text.paragraphs().getParagraphBefore(par);
637 openParTag(xs, &*par, prevpar, runparams);
639 auto pars = par->simpleDocBookOnePar(buf, runparams,text.outerFont(distance(begin, par)));
640 for (auto & parXML : pars)
641 // TODO: decide what to do with openParTag/closeParTag in new lines.
642 xs << XMLStream::ESCAPE_NONE << parXML;
644 closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr, runparams);
648 bool isLayoutSectioning(Layout const & lay)
650 if (lay.docbooksection()) // Special case: some DocBook styles must be handled as sections.
652 else if (lay.category() == from_utf8("Sectioning") || lay.docbooktag() == "section") // Generic case.
653 return lay.toclevel != Layout::NOT_IN_TOC;
658 bool isLayoutSectioningOrSimilar(Layout const & lay)
660 return isLayoutSectioning(lay) || lay.docbooktag() == "bridgehead";
664 using DocBookDocumentSectioning = tuple<bool, pit_type>;
667 struct DocBookInfoTag
669 const set<pit_type> shouldBeInInfo;
670 const set<pit_type> mustBeInInfo; // With the notable exception of the abstract!
671 const set<pit_type> abstract;
672 const bool abstractLayout;
676 DocBookInfoTag(const set<pit_type> & shouldBeInInfo, const set<pit_type> & mustBeInInfo,
677 const set<pit_type> & abstract, bool abstractLayout, pit_type bpit, pit_type epit) :
678 shouldBeInInfo(shouldBeInInfo), mustBeInInfo(mustBeInInfo), abstract(abstract),
679 abstractLayout(abstractLayout), bpit(bpit), epit(epit) {}
683 DocBookDocumentSectioning hasDocumentSectioning(ParagraphList const ¶graphs, pit_type bpit, pit_type const epit) {
684 bool documentHasSections = false;
686 while (bpit < epit) {
687 Layout const &style = paragraphs[bpit].layout();
688 documentHasSections |= isLayoutSectioningOrSimilar(style);
690 if (documentHasSections)
694 // Paragraphs before the first section: [ runparams.par_begin ; eppit )
696 return make_tuple(documentHasSections, bpit);
700 bool hasOnlyNotes(Paragraph const & par)
702 // Precondition: the paragraph is not empty. Otherwise, the function will always return true...
703 for (int i = 0; i < par.size(); ++i)
704 // If you find something that is not an inset (like actual text) or an inset that is not a note,
706 if (!par.isInset(i) || par.getInset(i)->lyxCode() != NOTE_CODE)
709 // An empty paragraph may still require some output.
710 if (par.layout().docbooksection())
713 // There should be really no content here.
718 DocBookInfoTag getParagraphsWithInfo(ParagraphList const ¶graphs,
719 pit_type bpit, pit_type const epit,
720 // Typically, bpit is the beginning of the document and epit the end of the
721 // document *or* the first section.
722 bool documentHasSections,
723 bool detectUnlayoutedAbstract
724 // Whether paragraphs with no specific layout should be detected as abstracts.
725 // For inner sections, an abstract should only be detected if it has a specific
726 // layout. For others, anything that might look like an abstract should be sought.
728 set<pit_type> shouldBeInInfo;
729 set<pit_type> mustBeInInfo;
730 set<pit_type> abstractWithLayout;
731 set<pit_type> abstractNoLayout;
733 // Find the first non empty paragraph by mutating bpit.
734 while (bpit < epit) {
735 Paragraph const &par = paragraphs[bpit];
736 if (par.empty() || hasOnlyNotes(par))
742 // Traverse everything that might belong to <info>.
743 bool hasAbstractLayout = false;
744 static depth_type INVALID_DEPTH = 100000;
745 depth_type abstractDepth = INVALID_DEPTH;
746 pit_type cpit = bpit;
747 for (; cpit < epit; ++cpit) {
748 // Skip paragraphs that don't generate anything in DocBook.
749 Paragraph const & par = paragraphs[cpit];
750 Layout const &style = par.layout();
751 if (hasOnlyNotes(par))
754 // There should never be any section here, except for the first paragraph (a title can be part of <info>).
755 // (Just a sanity check: if this fails, this function could end up processing the whole document.)
756 if (cpit != bpit && isLayoutSectioningOrSimilar(par.layout())) {
757 LYXERR0("Assertion failed: section found in potential <info> paragraphs.");
761 // If this is marked as an abstract by the layout, put it in the right set.
762 if (style.docbookabstract()) {
763 hasAbstractLayout = true;
764 abstractDepth = par.getDepth();
765 abstractWithLayout.emplace(cpit);
769 // Deeper paragraphs following the abstract must still be considered as part of the abstract.
770 // For instance, this includes lists. There should not be any other kind of paragraph in between.
771 if (abstractDepth != INVALID_DEPTH && style.docbookininfo() == "never") {
772 if (par.getDepth() > abstractDepth) {
773 abstractWithLayout.emplace(cpit);
776 if (par.getDepth() == abstractDepth) {
777 // This is not an abstract paragraph and it should not either be considered as part
778 // of it. It breaks the rule that abstract paragraphs must follow each other.
779 abstractDepth = INVALID_DEPTH;
784 // Based on layout information, store this paragraph in one set: should be in <info>, must be,
785 // or abstract (either because of layout or of position).
786 if (style.docbookininfo() == "always")
787 mustBeInInfo.emplace(cpit);
788 else if (style.docbookininfo() == "maybe")
789 shouldBeInInfo.emplace(cpit);
790 else if (documentHasSections && !hasAbstractLayout && detectUnlayoutedAbstract &&
791 (style.docbooktag() == "NONE" || style.docbooktag() == "para") &&
792 style.docbookwrappertag() == "NONE")
793 // In this case, it is very likely that style.docbookininfo() == "never"! Be extra careful
794 // about anything that gets caught here. For instance, don't ake into account
795 abstractNoLayout.emplace(cpit);
796 else // This should definitely not be in <info>.
799 // Now, cpit points to the first paragraph that no more has things that could go in <info>.
800 // bpit is the beginning of the <info> part.
802 return DocBookInfoTag(shouldBeInInfo, mustBeInInfo,
803 hasAbstractLayout ? abstractWithLayout : abstractNoLayout,
804 hasAbstractLayout, bpit, cpit);
807 } // end anonymous namespace
810 std::set<const Inset *> gatherInfo(ParagraphList::const_iterator par)
812 // This function has a structure highly similar to makeAny and its friends. It's only made to be called on what
813 // should become the document's <abstract>.
814 std::set<const Inset *> values;
816 // If this kind of layout should be ignored, already leave.
817 if (par->layout().docbooktag() == "IGNORE")
820 // If this should go in info, mark it as such. Dive deep into the abstract, as it may hide many things that
821 // DocBook doesn't want to be inside the abstract.
822 for (pos_type i = 0; i < par->size(); ++i) {
823 if (par->getInset(i) && par->getInset(i)->asInsetText()) {
824 InsetText const *inset = par->getInset(i)->asInsetText();
826 if (inset->getLayout().docbookininfo() != "never") {
827 values.insert(inset);
829 auto subpar = inset->paragraphs().begin();
830 while (subpar != inset->paragraphs().end()) {
831 auto subinfos = gatherInfo(subpar);
832 for (auto & subinfo: subinfos)
833 values.insert(subinfo);
844 ParagraphList::const_iterator makeAny(Text const &text,
847 OutputParams const &runparams,
848 ParagraphList::const_iterator par)
850 bool ignoreParagraph = false;
852 // If this kind of layout should be ignored, already leave.
853 ignoreParagraph |= par->layout().docbooktag() == "IGNORE";
855 // For things that should go into <info>, check the variable rp.docbook_generate_info. This does not apply to the
857 bool isAbstract = par->layout().docbookabstract() || par->layout().docbooktag() == "abstract";
858 ignoreParagraph |= !isAbstract && par->layout().docbookininfo() != "never" && !runparams.docbook_generate_info;
860 // Switch on the type of paragraph to call the right handler.
861 if (!ignoreParagraph) {
862 switch (par->layout().latextype) {
864 makeCommand(text, buf, xs, runparams, par);
866 case LATEX_ENVIRONMENT:
867 makeEnvironment(text, buf, xs, runparams, par);
869 case LATEX_LIST_ENVIRONMENT:
870 case LATEX_ITEM_ENVIRONMENT:
871 // Only case when makeAny() might consume more than one paragraph.
872 return makeListEnvironment(text, buf, xs, runparams, par);
873 case LATEX_PARAGRAPH:
874 makeParagraph(text, buf, xs, runparams, par);
876 case LATEX_BIB_ENVIRONMENT:
877 makeBibliography(text, buf, xs, runparams, par);
882 // For cases that are not lists, the next paragraph to handle is the next one.
888 xml::FontTag docbookStartFontTag(xml::FontTypes type)
890 return xml::FontTag(from_utf8(fontToDocBookTag(type)), from_utf8(fontToAttribute(type)), type);
894 xml::EndFontTag docbookEndFontTag(xml::FontTypes type)
896 return xml::EndFontTag(from_utf8(fontToDocBookTag(type)), type);
900 void outputDocBookInfo(
904 OutputParams const & runparams,
905 ParagraphList const & paragraphs,
906 DocBookInfoTag const & info)
908 // Perform an additional check on the abstract. Sometimes, there are many paragraphs that should go
909 // into the abstract, but none generates actual content. Thus, first generate to a temporary stream,
910 // then only create the <abstract> tag if these paragraphs generate some content.
911 // This check must be performed *before* a decision on whether or not to output <info> is made.
912 bool hasAbstract = !info.abstract.empty();
914 set<const Inset *> infoInsets; // Paragraphs that should go into <info>, but are hidden in an <abstract>
915 // paragraph. (This happens for quite a few layouts, unfortunately.)
918 // Generate the abstract XML into a string before further checks.
919 // Usually, makeAny only generates one paragraph at a time. However, for the specific case of lists, it might
920 // generate more than one paragraph, as indicated in the return value.
921 odocstringstream os2;
925 rp.docbook_generate_info = false;
926 rp.docbook_ignore_wrapper = true;
928 set<pit_type> doneParas; // Paragraphs that have already been converted (mostly to deal with lists).
929 for (auto const & p : info.abstract) {
930 if (doneParas.find(p) == doneParas.end()) {
931 auto oldPar = paragraphs.iterator_at(p);
932 auto newPar = makeAny(text, buf, xs2, rp, oldPar);
934 // Find insets that should go outside the abstract.
935 auto subinfos = gatherInfo(oldPar);
936 for (auto & subinfo: subinfos)
937 infoInsets.insert(subinfo);
939 // Insert the indices of all the paragraphs that were just generated (typically, one).
940 // **Make the hypothesis that, when an abstract has a list, all its items are consecutive.**
941 // Otherwise, makeAny and makeListEnvironment would have to be adapted too.
943 while (oldPar != newPar) {
944 doneParas.emplace(id);
951 // Actually output the abstract if there is something to do. Don't count line feeds or spaces in this,
952 // even though they must be properly output if there is some abstract.
953 abstract = os2.str();
954 docstring cleaned = abstract;
955 cleaned.erase(std::remove_if(cleaned.begin(), cleaned.end(), lyx::isSpace), cleaned.end());
957 // Nothing? Then there is no abstract!
962 // The abstract must go in <info>. Otherwise, decide whether to open <info> based on the layouts.
963 bool needInfo = !info.mustBeInInfo.empty() || hasAbstract;
965 // Start the <info> tag if required.
967 xs.startDivision(false);
968 xs << xml::StartTag("info");
972 // Output the elements that should go in <info>.
973 // - First, the title.
974 for (auto pit : info.shouldBeInInfo) // Typically, the title: these elements are so important and ubiquitous
975 // that mandating a wrapper like <info> would repel users. Thus, generate them first.
976 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
977 // If there is no title, generate one (required for the document to be valid).
978 // This code is called for the main document, for table cells, etc., so be precise in this condition.
979 if (text.isMainText() && info.shouldBeInInfo.empty() && !runparams.inInclude) {
980 xs << xml::StartTag("title");
981 xs << "Untitled Document";
982 xs << xml::EndTag("title");
986 // - Then, other metadata.
987 for (auto pit : info.mustBeInInfo)
988 makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
989 for (auto const * inset : infoInsets)
990 inset->docbook(xs, runparams);
992 // - Finally, always output the abstract as the last item of the <info>, as it requires special treatment
993 // (especially if it contains several paragraphs that are empty).
995 string tag = paragraphs[*info.abstract.begin()].layout().docbookforceabstracttag();
999 if (!xs.isLastTagCR())
1002 xs << xml::StartTag(tag);
1004 xs << XMLStream::ESCAPE_NONE << abstract;
1005 xs << xml::EndTag(tag);
1009 // End the <info> tag if it was started.
1011 if (!xs.isLastTagCR())
1014 xs << xml::EndTag("info");
1021 void docbookSimpleAllParagraphs(
1025 OutputParams const & runparams)
1027 // Handle the given text, supposing it has no sections (i.e. a "simple" text). The input may vary in length
1028 // between a single paragraph to a whole document.
1029 pit_type const bpit = runparams.par_begin;
1030 pit_type const epit = runparams.par_end;
1031 ParagraphList const ¶graphs = text.paragraphs();
1033 // First, the <info> tag.
1034 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, epit, false, true);
1035 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
1037 // Then, the content. It starts where the <info> ends.
1038 auto par = paragraphs.iterator_at(info.epit);
1039 auto end = paragraphs.iterator_at(epit);
1040 while (par != end) {
1041 if (!hasOnlyNotes(*par))
1042 par = makeAny(text, buf, xs, runparams, par);
1049 void docbookParagraphs(Text const &text,
1052 OutputParams const &runparams) {
1053 ParagraphList const ¶graphs = text.paragraphs();
1054 if (runparams.par_begin == runparams.par_end) {
1055 runparams.par_begin = 0;
1056 runparams.par_end = paragraphs.size();
1058 pit_type bpit = runparams.par_begin;
1059 pit_type const epit = runparams.par_end;
1060 LASSERT(bpit < epit,
1062 xs << XMLStream::ESCAPE_NONE << "<!-- DocBook output error! -->\n";
1066 std::stack<std::pair<int, string>> headerLevels; // Used to determine when to open/close sections: store the depth
1067 // of the section and the tag that was used to open it.
1069 // Detect whether the document contains sections. If there are no sections, treatment is largely simplified.
1070 // In particular, there can't be an abstract, unless it is manually marked.
1071 bool documentHasSections;
1073 tie(documentHasSections, eppit) = hasDocumentSectioning(paragraphs, bpit, epit);
1075 // Deal with "simple" documents, i.e. those without sections.
1076 if (!documentHasSections) {
1077 docbookSimpleAllParagraphs(text, buf, xs, runparams);
1081 // Output the first <info> tag (or just the title).
1082 DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, eppit, true, true);
1083 outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
1086 // Then, iterate through the paragraphs of this document.
1087 bool currentlyInAppendix = false;
1089 auto par = text.paragraphs().iterator_at(bpit);
1090 auto end = text.paragraphs().iterator_at(epit);
1091 while (par != end) {
1092 OutputParams ourparams = runparams;
1094 if (par->params().startOfAppendix())
1095 currentlyInAppendix = true;
1096 if (hasOnlyNotes(*par)) {
1101 Layout const &style = par->layout();
1103 // Think about adding <section> and/or </section>s.
1104 if (isLayoutSectioning(style)) {
1105 int level = style.toclevel;
1107 // Need to close a previous section if it has the same level or a higher one (close <section> if opening a
1108 // <h2> after a <h2>, <h3>, <h4>, <h5> or <h6>). More examples:
1109 // - current: h2; back: h1; do not close any <section>
1110 // - current: h1; back: h2; close two <section> (first the <h2>, then the <h1>, so a new <h1> can come)
1111 // Some layouts require that Layout::NOT_IN_TOC sections still cause closing of previous sections. This is
1112 // mostly to ensure that the section is positioned at a DocBook-compatible level (acknowledgements: cannot
1113 // be under a section!).
1114 while (!headerLevels.empty() && level <= headerLevels.top().first) {
1115 // Output the tag only if it corresponds to a legit section.
1116 int stackLevel = headerLevels.top().first;
1117 if (stackLevel != Layout::NOT_IN_TOC) {
1118 xs << xml::EndTag(headerLevels.top().second);
1124 // Open the new section: first push it onto the stack, then output it in DocBook.
1125 string sectionTag = (currentlyInAppendix && style.docbooksectiontag() == "chapter") ?
1126 "appendix" : style.docbooksectiontag();
1127 headerLevels.push(std::make_pair(level, sectionTag));
1129 // Some sectioning-like elements should not be output (such as FrontMatter).
1130 if (level != Layout::NOT_IN_TOC) {
1131 // Look for a label in the title, i.e. a InsetLabel as a child.
1132 docstring id = docstring();
1133 for (pos_type i = 0; i < par->size(); ++i) {
1134 Inset const *inset = par->getInset(i);
1136 if (auto label = dynamic_cast<InsetLabel const *>(inset)) {
1137 // Generate the attributes for the section if need be.
1138 id += "xml:id=\"" + xml::cleanID(label->screenLabel()) + "\"";
1140 // Don't output the ID as a DocBook <anchor>.
1141 ourparams.docbook_anchors_to_ignore.emplace(label->screenLabel());
1143 // Cannot have multiple IDs per tag. If there is another ID inset in the document, it will
1144 // be output as a DocBook anchor.
1150 // Write the open tag for this section.
1154 xs << xml::StartTag(sectionTag, attrs);
1159 // Close all sections before the bibliography.
1160 // 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)?
1161 if (!par->insetList().empty()) {
1162 Inset const *firstInset = par->getInset(0);
1163 if (firstInset && (firstInset->lyxCode() == BIBITEM_CODE || firstInset->lyxCode() == BIBTEX_CODE)) {
1164 while (!headerLevels.empty()) {
1165 int level = headerLevels.top().first;
1166 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1169 // Output the tag only if it corresponds to a legit section.
1170 if (level != Layout::NOT_IN_TOC) {
1171 xs << XMLStream::ESCAPE_NONE << tag;
1178 // Generate the <info> tag if a section was just opened.
1179 // Some sections may require abstracts (mostly parts, in books: DocBookForceAbstractTag will not be NONE),
1180 // others can still have an abstract (it must be detected so that it can be output at the right place).
1181 // 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.
1182 if (isLayoutSectioning(style)) {
1183 // This abstract may be found between the next paragraph and the next title.
1184 pit_type cpit = std::distance(text.paragraphs().begin(), par);
1185 pit_type ppit = std::get<1>(hasDocumentSectioning(paragraphs, cpit + 1L, epit));
1187 // Generate this abstract (this code corresponds to parts of outputDocBookInfo).
1188 DocBookInfoTag secInfo = getParagraphsWithInfo(paragraphs, cpit, ppit, true,
1189 style.docbookforceabstracttag() != "NONE");
1191 if (!secInfo.mustBeInInfo.empty() || !secInfo.shouldBeInInfo.empty() || !secInfo.abstract.empty()) {
1192 // Generate the <info>, if required. If DocBookForceAbstractTag != NONE, this abstract will not be in
1193 // <info>, unlike other ("standard") abstracts.
1194 bool hasStandardAbstract = !secInfo.abstract.empty() && style.docbookforceabstracttag() == "NONE";
1195 bool needInfo = !secInfo.mustBeInInfo.empty() || hasStandardAbstract;
1198 xs.startDivision(false);
1199 xs << xml::StartTag("info");
1203 // Output the elements that should go in <info>, before and after the abstract.
1204 for (auto pit : secInfo.shouldBeInInfo) // Typically, the title: these elements are so important and ubiquitous
1205 // that mandating a wrapper like <info> would repel users. Thus, generate them first.
1206 makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(pit));
1207 for (auto pit : secInfo.mustBeInInfo)
1208 makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(pit));
1210 // Deal with the abstract in <info> if it is standard (i.e. its tag is <abstract>).
1211 if (!secInfo.abstract.empty() && hasStandardAbstract) {
1212 if (!secInfo.abstractLayout) {
1213 xs << xml::StartTag("abstract");
1217 for (auto const &p : secInfo.abstract)
1218 makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(p));
1220 if (!secInfo.abstractLayout) {
1221 xs << xml::EndTag("abstract");
1226 // End the <info> tag if it was started.
1228 if (!xs.isLastTagCR())
1231 xs << xml::EndTag("info");
1236 // Deal with the abstract outside <info> if it is not standard (i.e. its tag is layout-defined).
1237 if (!secInfo.abstract.empty() && !hasStandardAbstract) {
1238 // Assert: style.docbookforceabstracttag() != NONE.
1239 xs << xml::StartTag(style.docbookforceabstracttag());
1241 for (auto const &p : secInfo.abstract)
1242 makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(p));
1243 xs << xml::EndTag(style.docbookforceabstracttag());
1247 // Skip all the text that has just been generated.
1248 par = paragraphs.iterator_at(secInfo.epit);
1250 // No <info> tag to generate, proceed as for normal paragraphs.
1251 par = makeAny(text, buf, xs, ourparams, par);
1254 // Generate this paragraph, as it has nothing special.
1255 par = makeAny(text, buf, xs, ourparams, par);
1259 // If need be, close <section>s, but only at the end of the document (otherwise, dealt with at the beginning
1261 while (!headerLevels.empty() && headerLevels.top().first > Layout::NOT_IN_TOC) {
1262 docstring tag = from_utf8("</" + headerLevels.top().second + ">");
1264 xs << XMLStream::ESCAPE_NONE << tag;