#include <config.h>
+#include "output_docbook.h"
+
#include "Buffer.h"
#include "buffer_funcs.h"
#include "BufferParams.h"
#include "insets/InsetBibtex.h"
#include "insets/InsetBibitem.h"
#include "insets/InsetLabel.h"
+#include "mathed/InsetMath.h"
#include "insets/InsetNote.h"
#include "support/lassert.h"
+#include "support/textutils.h"
#include <stack>
#include <iostream>
case xml::FontTypes::FT_BOLD:
return "emphasis";
case xml::FontTypes::FT_NOUN:
- return "person";
+ return "personname";
case xml::FontTypes::FT_UBAR:
case xml::FontTypes::FT_WAVE:
case xml::FontTypes::FT_DBAR:
void openTag(XMLStream & xs, const std::string & tag, const std::string & attr, const std::string & tagtype)
{
- if (tag.empty() || tag == "NONE")
+ if (tag.empty() || tag == "NONE") // Common check to be performed elsewhere, if it was not here.
return;
if (tag == "para" || tagtype == "paragraph") // Special case for <para>: always considered as a paragraph.
}
+void compTag(XMLStream & xs, const std::string & tag, const std::string & attr, const std::string & tagtype)
+{
+ if (tag.empty() || tag == "NONE")
+ return;
+
+ // Special case for <para>: always considered as a paragraph.
+ if (tag == "para" || tagtype == "paragraph" || tagtype == "block") {
+ if (!xs.isLastTagCR())
+ xs << xml::CR();
+ xs << xml::CompTag(tag, attr);
+ xs << xml::CR();
+ } else if (tagtype == "inline") {
+ xs << xml::CompTag(tag, attr);
+ } else {
+ xs.writeError("Unrecognised tag type '" + tagtype + "' for '" + tag + "'");
+ }
+}
+
+
// Higher-level convenience functions.
void openParTag(XMLStream & xs, const Paragraph * par, const Paragraph * prevpar)
if (prevpar != nullptr) {
Layout const & prevlay = prevpar->layout();
if (prevlay.docbookwrappertag() != "NONE") {
- openWrapper = prevlay.docbookwrappertag() == lay.docbookwrappertag()
- && !lay.docbookwrappermergewithprevious();
+ if (prevlay.docbookwrappertag() == lay.docbookwrappertag() &&
+ prevlay.docbookwrapperattr() == lay.docbookwrapperattr())
+ openWrapper = !lay.docbookwrappermergewithprevious();
+ else
+ openWrapper = true;
}
}
const string & tag = lay.docbooktag();
if (tag != "NONE") {
auto xmltag = xml::ParTag(tag, lay.docbookattr());
- if (!xs.isTagOpen(xmltag, 1)) // Don't nest a paragraph directly in a paragraph.
+ if (!xs.isTagOpen(xmltag, 1)) { // Don't nest a paragraph directly in a paragraph.
// TODO: required or not?
// TODO: avoid creating a ParTag object just for this query...
openTag(xs, lay.docbooktag(), lay.docbookattr(), lay.docbooktagtype());
+ openTag(xs, lay.docbookinnertag(), lay.docbookinnerattr(), lay.docbookinnertagtype());
+ }
}
openTag(xs, lay.docbookitemtag(), lay.docbookitemattr(), lay.docbookitemtagtype());
if (nextpar != nullptr) {
Layout const & nextlay = nextpar->layout();
if (nextlay.docbookwrappertag() != "NONE") {
- closeWrapper = nextlay.docbookwrappertag() == lay.docbookwrappertag()
- && !nextlay.docbookwrappermergewithprevious();
+ if (nextlay.docbookwrappertag() == lay.docbookwrappertag() &&
+ nextlay.docbookwrapperattr() == lay.docbookwrapperattr())
+ closeWrapper = !nextlay.docbookwrappermergewithprevious();
+ else
+ closeWrapper = true;
}
}
// Main logic.
closeTag(xs, lay.docbookiteminnertag(), lay.docbookiteminnertagtype());
closeTag(xs, lay.docbookitemtag(), lay.docbookitemtagtype());
+ closeTag(xs, lay.docbookinnertag(), lay.docbookinnertagtype());
closeTag(xs, lay.docbooktag(), lay.docbooktagtype());
if (closeWrapper)
closeTag(xs, lay.docbookwrappertag(), lay.docbookwrappertagtype());
}
-void openLabelTag(XMLStream & xs, Layout const & lay) // Mostly for definition lists.
-{
- openTag(xs, lay.docbookitemlabeltag(), lay.docbookitemlabelattr(), lay.docbookitemlabeltagtype());
-}
-
-
-void closeLabelTag(XMLStream & xs, Layout const & lay)
-{
- closeTag(xs, lay.docbookitemlabeltag(), lay.docbookitemlabeltagtype());
-}
-
-
-void openItemTag(XMLStream & xs, Layout const & lay)
-{
- openTag(xs, lay.docbookitemtag(), lay.docbookitemattr(), lay.docbookitemtagtype());
-}
-
-
-void closeItemTag(XMLStream & xs, Layout const & lay)
-{
- closeTag(xs, lay.docbookitemtag(), lay.docbookitemtagtype());
-}
-
-
-void makeAny(
- Text const &,
- Buffer const &,
- XMLStream &,
- OutputParams const &,
- ParagraphList::const_iterator);
-
-
-void makeParagraphBibliography(
+void makeBibliography(
+ Text const & text,
Buffer const & buf,
XMLStream & xs,
OutputParams const & runparams,
- Text const & text,
ParagraphList::const_iterator const & par)
{
// If this is the first paragraph in a bibliography, open the bibliography tag.
- auto pbegin_before = text.paragraphs().getParagraphBefore(par);
- if (pbegin_before->layout().latextype != LATEX_BIB_ENVIRONMENT) {
+ auto const * pbegin_before = text.paragraphs().getParagraphBefore(par);
+ if (pbegin_before == nullptr || (pbegin_before && pbegin_before->layout().latextype != LATEX_BIB_ENVIRONMENT)) {
xs << xml::StartTag("bibliography");
xs << xml::CR();
}
if (!ip)
continue;
if (const auto * bibitem = dynamic_cast<const InsetBibitem*>(ip)) {
- attr = from_utf8("xml:id='") + bibitem->getParam("key") + from_utf8("'");
+ auto id = xml::cleanID(bibitem->getParam("key"));
+ attr = from_utf8("xml:id='") + id + from_utf8("'");
break;
}
}
xs << xml::StartTag(from_utf8("bibliomixed"), attr);
- // Generate the entry.
+ // Generate the entry. Concatenate the different parts of the paragraph if any.
auto const begin = text.paragraphs().begin();
- par->simpleDocBookOnePar(buf, xs, runparams, text.outerFont(std::distance(begin, par)), true, true, 0);
+ auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(std::distance(begin, par)), 0);
+ for (auto & parXML : pars)
+ xs << XMLStream::ESCAPE_NONE << parXML;
// End the precooked bibliography entry.
xs << xml::EndTag("bibliomixed");
// If this is the last paragraph in a bibliography, close the bibliography tag.
auto const end = text.paragraphs().end();
- bool endBibliography = par == end;
- if (!endBibliography) {
- auto nextpar = par;
- ++nextpar;
- endBibliography = par->layout().latextype != LATEX_BIB_ENVIRONMENT;
- }
+ auto nextpar = par;
+ ++nextpar;
+ bool endBibliography = nextpar == end || nextpar->layout().latextype != LATEX_BIB_ENVIRONMENT;
if (endBibliography) {
xs << xml::EndTag("bibliography");
void makeParagraph(
+ Text const & text,
Buffer const & buf,
XMLStream & xs,
OutputParams const & runparams,
- Text const & text,
ParagraphList::const_iterator const & par)
{
auto const begin = text.paragraphs().begin();
special_case = true;
}
- // Plain layouts must be ignored.
- if (!special_case && buf.params().documentClass().isPlainLayout(par->layout()) && !runparams.docbook_force_pars)
- special_case = true;
- // TODO: Could get rid of this with a DocBook equivalent to htmlisblock?
- if (!special_case && par->size() == 1 && par->getInset(0)) {
- Inset const * firstInset = par->getInset(0);
+ size_t nInsets = std::distance(par->insetList().begin(), par->insetList().end());
- // Floats cannot be in paragraphs.
- special_case = to_utf8(firstInset->layoutName()).substr(0, 6) == "Float:";
-
- // Bibliographies cannot be in paragraphs.
- if (!special_case && firstInset->asInsetCommand())
- special_case = firstInset->asInsetCommand()->params().getCmdName() == "bibtex";
-
- // Equations do not deserve their own paragraph (DocBook allows them outside paragraphs).
- if (!special_case && firstInset->asInsetMath())
- special_case = true;
-
- // ERTs are in comments, not paragraphs.
- if (!special_case && firstInset->lyxCode() == lyx::ERT_CODE)
- special_case = true;
-
- // Listings should not get into their own paragraph.
- if (!special_case && firstInset->lyxCode() == lyx::LISTINGS_CODE)
- special_case = true;
- }
+ // Plain layouts must be ignored.
+ special_case |= buf.params().documentClass().isPlainLayout(par->layout()) && !runparams.docbook_force_pars;
+ // Equations do not deserve their own paragraph (DocBook allows them outside paragraphs).
+ // Exception: any case that generates an <inlineequation> must still get a paragraph to be valid.
+ special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
+ return inset.inset && inset.inset->asInsetMath() && inset.inset->asInsetMath()->getType() != hullSimple;
+ });
+ // Floats cannot be in paragraphs.
+ special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
+ return inset.inset->lyxCode() == FLOAT_CODE;
+ });
+ // Bibliographies cannot be in paragraphs. Bibitems should still be handled as paragraphs, though
+ // (see makeParagraphBibliography).
+ special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
+ return inset.inset->lyxCode() == BIBTEX_CODE;
+ });
+ // ERTs are in comments, not paragraphs.
+ special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
+ return inset.inset->lyxCode() == ERT_CODE;
+ });
+ // Listings should not get into their own paragraph.
+ special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
+ return inset.inset->lyxCode() == LISTINGS_CODE;
+ });
+ // Boxes cannot get into their own paragraph.
+ special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
+ return inset.inset->lyxCode() == BOX_CODE;
+ });
+ // Includes should not have a paragraph.
+ special_case |= nInsets == (size_t) par->size() && std::all_of(par->insetList().begin(), par->insetList().end(), [](InsetList::Element inset) {
+ return inset.inset->lyxCode() == INCLUDE_CODE;
+ });
bool const open_par = runparams.docbook_make_pars
&& !runparams.docbook_in_par
// or we're not in the last paragraph, anyway.
// (ii) We didn't open it and docbook_in_par is true,
// but we are in the first par, and there is a next par.
- auto nextpar = par;
- ++nextpar;
bool const close_par = open_par && (!runparams.docbook_in_par);
// Determine if this paragraph has some real content. Things like new pages are not caught
// by Paragraph::empty(), even though they do not generate anything useful in DocBook.
// Thus, remove all spaces (including new lines: \r, \n) before checking for emptiness.
- odocstringstream os2;
- XMLStream xs2(os2);
- par->simpleDocBookOnePar(buf, xs2, runparams, text.outerFont(distance(begin, par)), open_par, close_par, 0);
-
- docstring cleaned = os2.str();
- cleaned.erase(std::remove_if(cleaned.begin(), cleaned.end(), ::isspace), cleaned.end());
-
- if (!cleaned.empty()) {
- if (open_par)
- openParTag(xs, &*par, prevpar);
+ // std::all_of allows doing this check without having to copy the string.
+ // Open and close tags around each contained paragraph.
+ auto nextpar = par;
+ ++nextpar;
+ auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(distance(begin, par)), 0, nextpar == end, special_case);
+ for (docstring const & parXML : pars) {
+ if (xml::isNotOnlySpace(parXML)) {
+ if (open_par)
+ openParTag(xs, &*par, prevpar);
- xs << XMLStream::ESCAPE_NONE << os2.str();
+ xs << XMLStream::ESCAPE_NONE << parXML;
- if (close_par)
- closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
+ if (close_par)
+ closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
+ }
}
}
-void makeEnvironment(
- Buffer const &buf,
- XMLStream &xs,
- OutputParams const &runparams,
- Text const &text,
- ParagraphList::const_iterator const & par)
+void makeEnvironment(Text const &text,
+ Buffer const &buf,
+ XMLStream &xs,
+ OutputParams const &runparams,
+ ParagraphList::const_iterator const & par)
{
auto const end = text.paragraphs().end();
+ auto nextpar = par;
+ ++nextpar;
+
+ // Special cases for listing-like environments provided in layouts. This is quite ad-hoc, but provides a useful
+ // default. This should not be used by too many environments (only LyX-Code right now).
+ // This would be much simpler if LyX-Code was implemented as InsetListings...
+ bool mimicListing = false;
+ bool ignoreFonts = false;
+ if (par->layout().docbooktag() == "programlisting") {
+ mimicListing = true;
+ ignoreFonts = true;
+ }
// Output the opening tag for this environment, but only if it has not been previously opened (condition
// implemented in openParTag).
Layout const & style = par->layout();
if (style.latextype == LATEX_COMMAND) {
// Nothing to do (otherwise, infinite loops).
- } else if (style.latextype == LATEX_ENVIRONMENT ||
- style.latextype == LATEX_LIST_ENVIRONMENT ||
- style.latextype == LATEX_ITEM_ENVIRONMENT) {
- // Open a wrapper tag if needed.
- if (style.docbookitemwrappertag() != "NONE") {
- xs << xml::StartTag(style.docbookitemwrappertag(), style.docbookitemwrapperattr());
- xs << xml::CR();
+ } else if (style.latextype == LATEX_ENVIRONMENT) {
+ // Generate the paragraph, if need be.
+ auto pars = par->simpleDocBookOnePar(buf, runparams, text.outerFont(std::distance(text.paragraphs().begin(), par)), 0, false, ignoreFonts);
+
+ if (mimicListing) {
+ auto p = pars.begin();
+ while (p != pars.end()) {
+ openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(), par->layout().docbookiteminnertagtype());
+ xs << XMLStream::ESCAPE_NONE << *p;
+ closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
+ ++p;
+
+ // Insert a new line after each "paragraph" (i.e. line in the listing), except for the last one.
+ // Otherwise, there would one more new line in the output than in the LyX document.
+ if (p != pars.end())
+ xs << xml::CR();
+ }
+ } else {
+ for (auto const & p : pars) {
+ openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(), par->layout().docbookiteminnertagtype());
+ xs << XMLStream::ESCAPE_NONE << p;
+ closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
+ }
}
+ } else {
+ makeAny(text, buf, xs, runparams, par);
+ }
+
+ // Close the environment.
+ closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr); // TODO: switch in layout for par/block?
+}
+
+
+ParagraphList::const_iterator findEndOfEnvironment(
+ ParagraphList::const_iterator const & pstart,
+ ParagraphList::const_iterator const & pend)
+{
+ // Copy-paste from XHTML. Should be factored out at some point...
+ ParagraphList::const_iterator p = pstart;
+ Layout const & bstyle = p->layout();
+ size_t const depth = p->params().depth();
+ for (++p; p != pend; ++p) {
+ Layout const & style = p->layout();
+ // It shouldn't happen that e.g. a section command occurs inside
+ // a quotation environment, at a higher depth, but as of 6/2009,
+ // it can happen. We pretend that it's just at lowest depth.
+ if (style.latextype == LATEX_COMMAND)
+ return p;
+
+ // If depth is down, we're done
+ if (p->params().depth() < depth)
+ return p;
+
+ // If depth is up, we're not done
+ if (p->params().depth() > depth)
+ continue;
+
+ // FIXME I am not sure about the first check.
+ // Surely we *could* have different layouts that count as
+ // LATEX_PARAGRAPH, right?
+ if (style.latextype == LATEX_PARAGRAPH || style != bstyle)
+ return p;
+ }
+ return pend;
+}
+
+
+ParagraphList::const_iterator makeListEnvironment(Text const &text,
+ Buffer const &buf,
+ XMLStream &xs,
+ OutputParams const &runparams,
+ ParagraphList::const_iterator const & begin)
+{
+ auto par = begin;
+ auto const end = text.paragraphs().end();
+ auto const envend = findEndOfEnvironment(par, end);
+
+ // Output the opening tag for this environment.
+ Layout const & envstyle = par->layout();
+ openTag(xs, envstyle.docbookwrappertag(), envstyle.docbookwrapperattr(), envstyle.docbookwrappertagtype());
+ openTag(xs, envstyle.docbooktag(), envstyle.docbookattr(), envstyle.docbooktagtype());
+
+ // Handle the content of the list environment, item by item.
+ while (par != envend) {
+ // Skip this paragraph if it is both empty and the last one (otherwise, there may be deeper paragraphs after).
+ auto nextpar = par;
+ ++nextpar;
+ if (par->empty() && nextpar == envend)
+ break;
+
+ // Open the item wrapper.
+ Layout const & style = par->layout();
+ openTag(xs, style.docbookitemwrappertag(), style.docbookitemwrapperattr(), style.docbookitemwrappertagtype());
// Generate the label, if need be. If it is taken from the text, sep != 0 and corresponds to the first
// character after the label.
pos_type sep = 0;
if (style.labeltype != LABEL_NO_LABEL && style.docbookitemlabeltag() != "NONE") {
- // At least one condition must be met:
- // - this environment is not a list
- // - if this is a list, the label must not be manual (i.e. it must be taken from the layout)
- if (style.latextype != LATEX_LIST_ENVIRONMENT || style.labeltype != LABEL_MANUAL) {
+ if (style.labeltype == LABEL_MANUAL) {
+ // Only variablelist gets here (or similar items defined as an extension in the layout).
+ openTag(xs, style.docbookitemlabeltag(), style.docbookitemlabelattr(), style.docbookitemlabeltagtype());
+ sep = 1 + par->firstWordDocBook(xs, runparams);
+ closeTag(xs, style.docbookitemlabeltag(), style.docbookitemlabeltagtype());
+ } else {
// Usual cases: maybe there is something specified at the layout level. Highly unlikely, though.
docstring const lbl = par->params().labelString();
- if (lbl.empty()) {
- xs << xml::CR();
- } else {
- openLabelTag(xs, style);
+ if (!lbl.empty()) {
+ openTag(xs, style.docbookitemlabeltag(), style.docbookitemlabelattr(), style.docbookitemlabeltagtype());
xs << lbl;
- closeLabelTag(xs, style);
+ closeTag(xs, style.docbookitemlabeltag(), style.docbookitemlabeltagtype());
}
- } else {
- // Only variablelist gets here (or similar items defined as an extension in the layout).
- openLabelTag(xs, style);
- sep = par->firstWordDocBook(xs, runparams);
- closeLabelTag(xs, style);
}
}
- // Maybe the item is completely empty, i.e. if the first word ends at the end of the current paragraph
- // AND if the next paragraph doesn't have the same depth (if there is such a paragraph).
- // Common case: there is only the first word on the line, but there is a nested list instead
- // of more text.
- bool emptyItem = false;
- if (sep == par->size()) { // If the separator is already at the end of this paragraph...
- auto next_par = par;
- ++next_par;
- if (next_par == text.paragraphs().end()) // There is no next paragraph.
- emptyItem = true;
- else // There is a next paragraph: check depth.
- emptyItem = par->params().depth() >= next_par->params().depth();
- }
-
- if (emptyItem) {
- // Avoid having an empty item, this is not valid DocBook. A single character is enough to force
- // generation of a full <para>.
- // TODO: this always worked only by magic...
- xs << ' ';
+ // Open the item (after the wrapper and the label).
+ openTag(xs, style.docbookitemtag(), style.docbookitemattr(), style.docbookitemtagtype());
+
+ // Generate the content of the item.
+ if (sep < par->size()) {
+ auto pars = par->simpleDocBookOnePar(buf, runparams,
+ text.outerFont(std::distance(text.paragraphs().begin(), par)), sep);
+ for (auto &p : pars) {
+ openTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
+ par->layout().docbookiteminnertagtype());
+ xs << XMLStream::ESCAPE_NONE << p;
+ closeTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnertagtype());
+ }
} else {
- // Generate the rest of the paragraph, if need be.
- par->simpleDocBookOnePar(buf, xs, runparams, text.outerFont(std::distance(text.paragraphs().begin(), par)),
- true, true, sep);
+ // DocBook doesn't like emptiness.
+ compTag(xs, par->layout().docbookiteminnertag(), par->layout().docbookiteminnerattr(),
+ par->layout().docbookiteminnertagtype());
}
- } else {
- makeAny(text, buf, xs, runparams, par);
+
+ // If the next item is deeper, it must go entirely within this item (do it recursively).
+ // By construction, with findEndOfEnvironment, depth can only stay constant or increase, never decrease.
+ depth_type currentDepth = par->getDepth();
+ ++par;
+ while (par != envend && par->getDepth() != currentDepth)
+ par = makeAny(text, buf, xs, runparams, par);
+ // Usually, this loop only makes one iteration, except in complex scenarios, like an item with a paragraph,
+ // a list, and another paragraph; or an item with two types of list (itemise then enumerate, for instance).
+
+ // Close the item.
+ closeTag(xs, style.docbookitemtag(), style.docbookitemtagtype());
+ closeTag(xs, style.docbookitemwrappertag(), style.docbookitemwrappertagtype());
}
- // Close the environment.
- auto nextpar = par;
- ++nextpar;
- closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr); // TODO: switch in layout for par/block?
+ // Close this environment in exactly the same way as it was opened.
+ closeTag(xs, envstyle.docbooktag(), envstyle.docbooktagtype());
+ closeTag(xs, envstyle.docbookwrappertag(), envstyle.docbookwrappertagtype());
+
+ return envend;
}
void makeCommand(
+ Text const & text,
Buffer const & buf,
XMLStream & xs,
OutputParams const & runparams,
- Text const & text,
ParagraphList::const_iterator const & par)
{
// Unlike XHTML, no need for labels, as they are handled by DocBook tags.
auto prevpar = text.paragraphs().getParagraphBefore(par);
openParTag(xs, &*par, prevpar);
- par->simpleDocBookOnePar(buf, xs, runparams,
- text.outerFont(distance(begin, par)));
+ auto pars = par->simpleDocBookOnePar(buf, runparams,text.outerFont(distance(begin, par)));
+ for (auto & parXML : pars)
+ // TODO: decide what to do with openParTag/closeParTag in new lines.
+ xs << XMLStream::ESCAPE_NONE << parXML;
closeParTag(xs, &*par, (nextpar != end) ? &*nextpar : nullptr);
}
-void makeAny(
- Text const &text,
- Buffer const &buf,
- XMLStream &xs,
- OutputParams const &ourparams,
- ParagraphList::const_iterator par)
+bool isLayoutSectioning(Layout const & lay)
{
- switch (par->layout().latextype) {
- case LATEX_COMMAND:
- makeCommand(buf, xs, ourparams, text, par);
- break;
- case LATEX_ENVIRONMENT:
- case LATEX_LIST_ENVIRONMENT:
- case LATEX_ITEM_ENVIRONMENT:
- makeEnvironment(buf, xs, ourparams, text, par);
- break;
- case LATEX_PARAGRAPH:
- makeParagraph(buf, xs, ourparams, text, par);
- break;
- case LATEX_BIB_ENVIRONMENT:
- makeParagraphBibliography(buf, xs, ourparams, text, par);
- break;
- }
+ if (lay.docbooksection()) // Special case: some DocBook styles must be handled as sections.
+ return true;
+ else if (lay.category() == from_utf8("Sectioning")) // Generic case.
+ return lay.toclevel != Layout::NOT_IN_TOC;
+ return false;
}
struct DocBookInfoTag
{
const set<pit_type> shouldBeInInfo;
- const set<pit_type> mustBeInInfo;
+ const set<pit_type> mustBeInInfo; // With the notable exception of the abstract!
const set<pit_type> abstract;
+ const bool abstractLayout;
pit_type bpit;
pit_type epit;
DocBookInfoTag(const set<pit_type> & shouldBeInInfo, const set<pit_type> & mustBeInInfo,
- const set<pit_type> & abstract, pit_type bpit, pit_type epit) :
+ const set<pit_type> & abstract, bool abstractLayout, pit_type bpit, pit_type epit) :
shouldBeInInfo(shouldBeInInfo), mustBeInInfo(mustBeInInfo), abstract(abstract),
- bpit(bpit), epit(epit) {}
+ abstractLayout(abstractLayout), bpit(bpit), epit(epit) {}
};
while (bpit < epit) {
Layout const &style = paragraphs[bpit].layout();
- documentHasSections |= style.category() == from_utf8("Sectioning");
+ documentHasSections |= isLayoutSectioning(style);
if (documentHasSections)
break;
for (int i = 0; i < par.size(); ++i)
// If you find something that is not an inset (like actual text) or an inset that is not a note,
// return false.
- if (!par.isInset(i) || !dynamic_cast<InsetNote *>(par.insetList().get(i)))
+ if (!par.isInset(i) || par.getInset(i)->lyxCode() != NOTE_CODE)
return false;
+
+ // An empty paragraph may still require some output.
+ if (par.layout().docbooksection())
+ return false;
+
+ // There should be really no content here.
return true;
}
-DocBookInfoTag getParagraphsWithInfo(ParagraphList const ¶graphs, pit_type bpit, pit_type const epit) {
+DocBookInfoTag getParagraphsWithInfo(ParagraphList const ¶graphs,
+ pit_type bpit, pit_type const epit,
+ // Typically, bpit is the beginning of the document and epit the end of the
+ // document *or* the first section.
+ bool documentHasSections,
+ bool detectUnlayoutedAbstract
+ // Whether paragraphs with no specific layout should be detected as abstracts.
+ // For inner sections, an abstract should only be detected if it has a specific
+ // layout. For others, anything that might look like an abstract should be sought.
+ ) {
set<pit_type> shouldBeInInfo;
set<pit_type> mustBeInInfo;
- set<pit_type> abstract;
+ set<pit_type> abstractWithLayout;
+ set<pit_type> abstractNoLayout;
// Find the first non empty paragraph by mutating bpit.
while (bpit < epit) {
break;
}
- // Find the last info-like paragraph.
- pit_type cpit = bpit;
+ // Traverse everything that might belong to <info>.
bool hasAbstractLayout = false;
- while (cpit < epit) {
- // Skip paragraphs only containing one note.
+ pit_type cpit = bpit;
+ for (; cpit < epit; ++cpit) {
+ // Skip paragraphs that don't generate anything in DocBook.
Paragraph const & par = paragraphs[cpit];
- if (hasOnlyNotes(par)) {
- cpit += 1;
+ Layout const &style = par.layout();
+ if (hasOnlyNotes(par))
continue;
+
+ // There should never be any section here, except for the first paragraph (a title can be part of <info>).
+ // (Just a sanity check: if this fails, this function could end up processing the whole document.)
+ if (cpit != bpit && isLayoutSectioning(par.layout())) {
+ LYXERR0("Assertion failed: section found in potential <info> paragraphs.");
+ break;
}
- if (par.layout().docbookabstract())
+ // If this is marked as an abstract by the layout, put it in the right set.
+ if (style.docbookabstract()) {
hasAbstractLayout = true;
+ abstractWithLayout.emplace(cpit);
+ continue;
+ }
- // Based on layout information, store this paragraph in one set: should be in <info>, must be.
- Layout const &style = par.layout();
-
- if (style.docbookininfo() == "always") {
+ // Based on layout information, store this paragraph in one set: should be in <info>, must be,
+ // or abstract (either because of layout or of position).
+ if (style.docbookininfo() == "always")
mustBeInInfo.emplace(cpit);
- } else if (style.docbookininfo() == "maybe") {
+ else if (style.docbookininfo() == "maybe")
shouldBeInInfo.emplace(cpit);
- } else {
- // Hypothesis: the <info> parts should be grouped together near the beginning bpit.
- // There may be notes in between, but nothing else.
+ else if (documentHasSections && !hasAbstractLayout && detectUnlayoutedAbstract)
+ abstractNoLayout.emplace(cpit);
+ else // This should definitely not be in <info>.
break;
- }
- cpit += 1;
}
- // Now, cpit points to the last paragraph that has things that could go in <info>.
+ // Now, cpit points to the first paragraph that no more has things that could go in <info>.
// bpit is the beginning of the <info> part.
- // Go once again through the list of paragraphs to find the abstract. If there is an abstract
- // layout, only consider it. Otherwise, an abstract is just a sequence of paragraphs with text.
- if (hasAbstractLayout) {
- pit_type pit = bpit;
- while (pit < cpit) { // Don't overshoot the <info> part.
- if (paragraphs[pit].layout().docbookabstract())
- abstract.emplace(pit);
- pit++;
- }
- } else {
- pit_type lastAbstract = epit + 1; // A nonsensical value.
- docstring lastAbstractLayout;
-
- pit_type pit = bpit;
- while (pit < cpit) { // Don't overshoot the <info> part.
- const Paragraph & par = paragraphs.at(pit);
- if (!par.insetList().empty()) {
- for (const auto &i : par.insetList()) {
- if (i.inset->getText(0) != nullptr) {
- if (lastAbstract == epit + 1) {
- // First paragraph that matches the heuristic definition of abstract.
- lastAbstract = pit;
- lastAbstractLayout = par.layout().name();
- } else if (pit > lastAbstract + 1 || par.layout().name() != lastAbstractLayout) {
- // This is either too far from the last abstract paragraph or doesn't
- // have the right layout name, BUT there has already been an abstract
- // in this document: done with detecting the abstract.
- goto done; // Easier to get out of two nested loops.
- }
-
- abstract.emplace(pit);
- break;
- }
- }
- }
- pit++;
- }
- }
-
- done:
- return DocBookInfoTag(shouldBeInInfo, mustBeInInfo, abstract, bpit, cpit);
+ return DocBookInfoTag(shouldBeInInfo, mustBeInInfo,
+ hasAbstractLayout ? abstractWithLayout : abstractNoLayout,
+ hasAbstractLayout, bpit, cpit);
}
} // end anonymous namespace
+ParagraphList::const_iterator makeAny(Text const &text,
+ Buffer const &buf,
+ XMLStream &xs,
+ OutputParams const &runparams,
+ ParagraphList::const_iterator par)
+{
+ switch (par->layout().latextype) {
+ case LATEX_COMMAND:
+ makeCommand(text, buf, xs, runparams, par);
+ break;
+ case LATEX_ENVIRONMENT:
+ makeEnvironment(text, buf, xs, runparams, par);
+ break;
+ case LATEX_LIST_ENVIRONMENT:
+ case LATEX_ITEM_ENVIRONMENT:
+ // Only case when makeAny() might consume more than one paragraph.
+ return makeListEnvironment(text, buf, xs, runparams, par);
+ case LATEX_PARAGRAPH:
+ makeParagraph(text, buf, xs, runparams, par);
+ break;
+ case LATEX_BIB_ENVIRONMENT:
+ makeBibliography(text, buf, xs, runparams, par);
+ break;
+ }
+ ++par;
+ return par;
+}
+
+
xml::FontTag docbookStartFontTag(xml::FontTypes type)
{
return xml::FontTag(from_utf8(fontToDocBookTag(type)), from_utf8(fontToAttribute(type)), type);
docstring abstract;
if (hasAbstract) {
// Generate the abstract XML into a string before further checks.
+ // Usually, makeAny only generates one paragraph at a time. However, for the specific case of lists, it might
+ // generate more than one paragraph, as indicated in the return value.
odocstringstream os2;
- {
- XMLStream xs2(os2);
- auto bpit = *std::min_element(info.abstract.begin(), info.abstract.end());
- auto epit = 1 + *std::max_element(info.abstract.begin(), info.abstract.end());
- // info.abstract is inclusive, epit is exclusive, hence +1 for looping.
-
- while (bpit < epit) {
- makeAny(text, buf, xs2, runparams, paragraphs.iterator_at(bpit));
- bpit += 1;
+ XMLStream xs2(os2);
+
+ set<pit_type> doneParas;
+ for (auto const & p : info.abstract) {
+ if (doneParas.find(p) == doneParas.end()) {
+ auto oldPar = paragraphs.iterator_at(p);
+ auto newPar = makeAny(text, buf, xs2, runparams, oldPar);
+
+ // Insert the indices of all the paragraphs that were just generated (typically, one).
+ // **Make the hypothesis that, when an abstract has a list, all its items are consecutive.**
+ pit_type id = p;
+ while (oldPar != newPar) {
+ doneParas.emplace(id);
+ ++oldPar;
+ ++id;
+ }
}
}
// even though they must be properly output if there is some abstract.
abstract = os2.str();
docstring cleaned = abstract;
- cleaned.erase(std::remove_if(cleaned.begin(), cleaned.end(), ::isspace), cleaned.end());
+ cleaned.erase(std::remove_if(cleaned.begin(), cleaned.end(), lyx::isSpace), cleaned.end());
// Nothing? Then there is no abstract!
if (cleaned.empty())
}
// Output the elements that should go in <info>, before and after the abstract.
- for (auto pit : info.shouldBeInInfo) { // Typically, the title: these elements are so important and ubiquitous
+ for (auto pit : info.shouldBeInInfo) // Typically, the title: these elements are so important and ubiquitous
// that mandating a wrapper like <info> would repel users. Thus, generate them first.
makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
- }
- for (auto pit : info.mustBeInInfo) {
- if (info.abstract.find(pit) == info.abstract.end()) // The abstract must be in info, but is dealt with after.
- makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
+ for (auto pit : info.mustBeInInfo)
+ makeAny(text, buf, xs, runparams, paragraphs.iterator_at(pit));
+
+ // If there is no title, generate one (required for the document to be valid).
+ // This code is called for the main document, for table cells, etc., so be precise in this condition.
+ if (text.isMainText() && info.shouldBeInInfo.empty() && !runparams.inInclude) {
+ xs << xml::StartTag("title");
+ xs << "Untitled Document";
+ xs << xml::EndTag("title");
+ xs << xml::CR();
}
// Always output the abstract as the last item of the <info>, as it requires special treatment (especially if
// it contains several paragraphs that are empty).
if (hasAbstract) {
-// string tag = paragraphs[*info.abstract.begin()].layout().docbookforceabstracttag();
-// if (tag == "NONE")
-// tag = "abstract";
-//
-// xs << xml::StartTag(tag);
-// xs << xml::CR();
- xs << XMLStream::ESCAPE_NONE << abstract;
-// xs << xml::EndTag(tag);
-// xs << xml::CR();
+ if (info.abstractLayout) {
+ xs << XMLStream::ESCAPE_NONE << abstract;
+ xs << xml::CR();
+ } else {
+ string tag = paragraphs[*info.abstract.begin()].layout().docbookforceabstracttag();
+ if (tag == "NONE")
+ tag = "abstract";
+
+ if (!xs.isLastTagCR())
+ xs << xml::CR();
+
+ xs << xml::StartTag(tag);
+ xs << xml::CR();
+ xs << XMLStream::ESCAPE_NONE << abstract;
+ xs << xml::EndTag(tag);
+ xs << xml::CR();
+ }
}
// End the <info> tag if it was started.
if (needInfo) {
+ if (!xs.isLastTagCR())
+ xs << xml::CR();
+
xs << xml::EndTag("info");
xs << xml::CR();
xs.endDivision();
}
-void docbookFirstParagraphs(
- Text const &text,
- Buffer const &buf,
- XMLStream &xs,
- OutputParams const &runparams,
- pit_type epit)
-{
- // Handle the beginning of the document, supposing it has sections.
- // Major role: output the first <info> tag.
-
- ParagraphList const ¶graphs = text.paragraphs();
- pit_type bpit = runparams.par_begin;
- DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, epit);
- outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
-}
-
-
void docbookSimpleAllParagraphs(
Text const & text,
Buffer const & buf,
{
// Handle the given text, supposing it has no sections (i.e. a "simple" text). The input may vary in length
// between a single paragraph to a whole document.
+ pit_type const bpit = runparams.par_begin;
+ pit_type const epit = runparams.par_end;
+ ParagraphList const ¶graphs = text.paragraphs();
// First, the <info> tag.
- ParagraphList const ¶graphs = text.paragraphs();
- pit_type bpit = runparams.par_begin;
- pit_type const epit = runparams.par_end;
- DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, epit);
+ DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, epit, false, true);
outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
// Then, the content. It starts where the <info> ends.
- bpit = info.epit;
- while (bpit < epit) {
- auto par = paragraphs.iterator_at(bpit);
+ auto par = paragraphs.iterator_at(info.epit);
+ auto end = paragraphs.iterator_at(epit);
+ while (par != end) {
if (!hasOnlyNotes(*par))
- makeAny(text, buf, xs, runparams, par);
- bpit += 1;
+ par = makeAny(text, buf, xs, runparams, par);
+ else
+ ++par;
}
}
std::stack<std::pair<int, string>> headerLevels; // Used to determine when to open/close sections: store the depth
// of the section and the tag that was used to open it.
- // Detect whether the document contains sections. If there are no sections, there can be no automatically
- // discovered abstract.
+ // Detect whether the document contains sections. If there are no sections, treatment is largely simplified.
+ // In particular, there can't be an abstract, unless it is manually marked.
bool documentHasSections;
pit_type eppit;
tie(documentHasSections, eppit) = hasDocumentSectioning(paragraphs, bpit, epit);
- if (documentHasSections) {
- docbookFirstParagraphs(text, buf, xs, runparams, eppit);
- bpit = eppit;
- } else {
+ // Deal with "simple" documents, i.e. those without sections.
+ if (!documentHasSections) {
docbookSimpleAllParagraphs(text, buf, xs, runparams);
return;
}
+ // Output the first <info> tag (or just the title).
+ DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, eppit, true, true);
+ outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
+ bpit = info.epit;
+
+ // Then, iterate through the paragraphs of this document.
bool currentlyInAppendix = false;
- while (bpit < epit) {
+ auto par = text.paragraphs().iterator_at(bpit);
+ auto end = text.paragraphs().iterator_at(epit);
+ while (par != end) {
OutputParams ourparams = runparams;
- auto par = paragraphs.iterator_at(bpit);
if (par->params().startOfAppendix())
currentlyInAppendix = true;
- Layout const &style = par->layout();
- ParagraphList::const_iterator const lastStartedPar = par;
- ParagraphList::const_iterator send;
-
if (hasOnlyNotes(*par)) {
- bpit += 1;
+ ++par;
continue;
}
+ Layout const &style = par->layout();
+
// Think about adding <section> and/or </section>s.
- const bool isLayoutSectioning = style.category() == from_utf8("Sectioning");
- if (isLayoutSectioning) {
+ if (isLayoutSectioning(style)) {
int level = style.toclevel;
- // Need to close a previous section if it has the same level or a higher one (close <section> if opening a <h2>
- // after a <h2>, <h3>, <h4>, <h5> or <h6>). More examples:
+ // Need to close a previous section if it has the same level or a higher one (close <section> if opening a
+ // <h2> after a <h2>, <h3>, <h4>, <h5> or <h6>). More examples:
// - current: h2; back: h1; do not close any <section>
// - current: h1; back: h2; close two <section> (first the <h2>, then the <h1>, so a new <h1> can come)
while (!headerLevels.empty() && level <= headerLevels.top().first) {
+ // Output the tag only if it corresponds to a legit section.
int stackLevel = headerLevels.top().first;
- docstring stackTag = from_utf8("</" + headerLevels.top().second + ">");
+ if (stackLevel != Layout::NOT_IN_TOC) {
+ xs << xml::EndTag(headerLevels.top().second);
+ xs << xml::CR();
+ }
headerLevels.pop();
-
- // Output the tag only if it corresponds to a legit section.
- if (stackLevel != Layout::NOT_IN_TOC)
- xs << XMLStream::ESCAPE_NONE << stackTag << xml::CR();
}
// Open the new section: first push it onto the stack, then output it in DocBook.
// Don't output the ID as a DocBook <anchor>.
ourparams.docbook_anchors_to_ignore.emplace(label->screenLabel());
- // Cannot have multiple IDs per tag.
+ // Cannot have multiple IDs per tag. If there is another ID inset in the document, it will
+ // be output as a DocBook anchor.
break;
}
}
}
// Write the open tag for this section.
- docstring tag = from_utf8("<" + sectionTag);
+ docstring attrs;
if (!id.empty())
- tag += from_utf8(" ") + id;
- tag += from_utf8(">");
- xs << XMLStream::ESCAPE_NONE << tag;
+ attrs = id;
+ xs << xml::StartTag(sectionTag, attrs);
xs << xml::CR();
}
}
// Close all sections before the bibliography.
// 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)?
- auto insetsLength = distance(par->insetList().begin(), par->insetList().end());
- if (insetsLength > 0) {
+ if (!par->insetList().empty()) {
Inset const *firstInset = par->getInset(0);
- if (firstInset && dynamic_cast<InsetBibtex const *>(firstInset)) {
+ if (firstInset && (firstInset->lyxCode() == BIBITEM_CODE || firstInset->lyxCode() == BIBTEX_CODE)) {
while (!headerLevels.empty()) {
int level = headerLevels.top().first;
docstring tag = from_utf8("</" + headerLevels.top().second + ">");
}
}
- // Generate this paragraph.
- makeAny(text, buf, xs, ourparams, par);
- bpit += 1;
+ // Generate the <info> tag if a section was just opened.
+ // Some sections may require abstracts (mostly parts, in books: DocBookForceAbstractTag will not be NONE),
+ // others can still have an abstract (it must be detected so that it can be output at the right place).
+ // 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.
+ if (isLayoutSectioning(style)) {
+ // This abstract may be found between the next paragraph and the next title.
+ pit_type cpit = std::distance(text.paragraphs().begin(), par);
+ pit_type ppit = std::get<1>(hasDocumentSectioning(paragraphs, cpit + 1L, epit));
+
+ // Generate this abstract (this code corresponds to parts of outputDocBookInfo).
+ DocBookInfoTag secInfo = getParagraphsWithInfo(paragraphs, cpit, ppit, true,
+ style.docbookforceabstracttag() != "NONE");
+
+ if (!secInfo.mustBeInInfo.empty() || !secInfo.shouldBeInInfo.empty() || !secInfo.abstract.empty()) {
+ // Generate the <info>, if required. If DocBookForceAbstractTag != NONE, this abstract will not be in
+ // <info>, unlike other ("standard") abstracts.
+ bool hasStandardAbstract = !secInfo.abstract.empty() && style.docbookforceabstracttag() == "NONE";
+ bool needInfo = !secInfo.mustBeInInfo.empty() || hasStandardAbstract;
+
+ if (needInfo) {
+ xs.startDivision(false);
+ xs << xml::StartTag("info");
+ xs << xml::CR();
+ }
+
+ // Output the elements that should go in <info>, before and after the abstract.
+ for (auto pit : secInfo.shouldBeInInfo) // Typically, the title: these elements are so important and ubiquitous
+ // that mandating a wrapper like <info> would repel users. Thus, generate them first.
+ makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(pit));
+ for (auto pit : secInfo.mustBeInInfo)
+ makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(pit));
+
+ // Deal with the abstract in <info> if it is standard (i.e. its tag is <abstract>).
+ if (!secInfo.abstract.empty() && hasStandardAbstract) {
+ if (!secInfo.abstractLayout) {
+ xs << xml::StartTag("abstract");
+ xs << xml::CR();
+ }
+
+ for (auto const &p : secInfo.abstract)
+ makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(p));
+
+ if (!secInfo.abstractLayout) {
+ xs << xml::EndTag("abstract");
+ xs << xml::CR();
+ }
+ }
+
+ // End the <info> tag if it was started.
+ if (needInfo) {
+ if (!xs.isLastTagCR())
+ xs << xml::CR();
+
+ xs << xml::EndTag("info");
+ xs << xml::CR();
+ xs.endDivision();
+ }
+
+ // Deal with the abstract outside <info> if it is not standard (i.e. its tag is layout-defined).
+ if (!secInfo.abstract.empty() && !hasStandardAbstract) {
+ // Assert: style.docbookforceabstracttag() != NONE.
+ xs << xml::StartTag(style.docbookforceabstracttag());
+ xs << xml::CR();
+ for (auto const &p : secInfo.abstract)
+ makeAny(text, buf, xs, ourparams, paragraphs.iterator_at(p));
+ xs << xml::EndTag(style.docbookforceabstracttag());
+ xs << xml::CR();
+ }
+
+ // Skip all the text that has just been generated.
+ par = paragraphs.iterator_at(secInfo.epit);
+ } else {
+ // No <info> tag to generate, proceed as for normal paragraphs.
+ par = makeAny(text, buf, xs, ourparams, par);
+ }
+ } else {
+ // Generate this paragraph, as it has nothing special.
+ par = makeAny(text, buf, xs, ourparams, par);
+ }
}
// If need be, close <section>s, but only at the end of the document (otherwise, dealt with at the beginning
projects / features.git / blobdiff