]> git.lyx.org Git - features.git/commitdiff
DocBook: simplify code to handle abstracts.
authorThibaut Cuvelier <tcuvelier@lyx.org>
Sun, 2 Aug 2020 18:43:39 +0000 (20:43 +0200)
committerThibaut Cuvelier <tcuvelier@lyx.org>
Mon, 3 Aug 2020 14:04:58 +0000 (16:04 +0200)
autotests/export/docbook/bibliography_precooked_aastex.xml
lib/layouts/aastex.layout
src/output_docbook.cpp

index 2c9f5bd32aab762ba31d6cab4d7a277910435370..e15bfddc51e0e445f8d9b2ce69a322af457fe44b 100644 (file)
 <affiliation role="alternate"><orgname>Patron, Alonso's Bar and Grill</orgname></affiliation></author>
 <abstract><para>
 <para>This is a preliminary report on surface photometry of the major fraction of known globular clusters, to see which of them show the signs of a collapsed core. We also explore some diversionary mathematics and recreational tables. </para>
-<!-- Output Error: Tried to close `Keywords' when tag was not open. Tag discarded. -->
+<!-- Output Error: Tried to close `keyword' when tag was not open. Tag discarded. -->
+<!-- Output Error: Tried to close `keywordset' when tag was not open. Tag discarded. -->
 
+<keywordset><keyword>clusters: globular, peanut—bosons: bozos</keyword></keywordset>
 <!-- Output Error: Closing tag `info' when other tags are open, namely: -->
 <!-- Output Error: para -->
 </para><!-- Output Error: abstract -->
index 445d25f536031198acba94b093ef7add5eb7a234..97c8b64940590c8cac1362d9b6a573e94d2eec5a 100644 (file)
@@ -282,6 +282,9 @@ Style Abstract
          Series        Bold
          Size          Normal
        EndFont
+       DocBookTag            abstract
+       DocBookItemTag        para
+       DocBookInInfo         always
 End
 
 
@@ -331,6 +334,9 @@ Style Keywords
          Shape         Italic
          Size          Normal
        EndFont
+       DocBookTag          keyword
+       DocBookWrapperTag   keywordset
+       DocBookInInfo       always
 End
 
 
index d0536eee3ad7c07080c79520e3ea1c85d48b78cb..fd38940064c744f4dcbee4d581d7e2d3aaa92640 100644 (file)
@@ -50,7 +50,7 @@ namespace lyx {
 
 namespace {
 
-std::string const fontToDocBookTag(xml::FontTypes type)
+std::string fontToDocBookTag(xml::FontTypes type)
 {
        switch (type) {
        case xml::FontTypes::FT_EMPH:
@@ -90,6 +90,7 @@ std::string const fontToDocBookTag(xml::FontTypes type)
        }
 }
 
+
 string fontToRole(xml::FontTypes type)
 {
        // Specific fonts are achieved with roles. The only common ones are "" for basic emphasis,
@@ -103,14 +104,13 @@ string fontToRole(xml::FontTypes type)
                return "";
        case xml::FontTypes::FT_BOLD:
                return "bold";
-       case xml::FontTypes::FT_NOUN:
-               return ""; // Outputs a <person>
-       case xml::FontTypes::FT_TYPE:
-               return ""; // Outputs a <code>
+       case xml::FontTypes::FT_NOUN: // Outputs a <person>
+       case xml::FontTypes::FT_TYPE: // Outputs a <code>
+               return "";
        case xml::FontTypes::FT_UBAR:
                return "underline";
 
-               // All other roles are non-standard for DocBook.
+       // All other roles are non-standard for DocBook.
 
        case xml::FontTypes::FT_WAVE:
                return "wave";
@@ -814,7 +814,21 @@ pair<ParagraphList::const_iterator, ParagraphList::const_iterator> makeAny(
 
 
 using DocBookDocumentSectioning = tuple<bool, pit_type>;
-using DocBookInfoTag = tuple<set<pit_type>, set<pit_type>, pit_type, pit_type>;
+
+
+struct DocBookInfoTag
+{
+       const set<pit_type> shouldBeInInfo;
+       const set<pit_type> mustBeInInfo;
+       const set<pit_type> abstract;
+       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) :
+                                  shouldBeInInfo(shouldBeInInfo), mustBeInInfo(mustBeInInfo), abstract(abstract),
+                                  bpit(bpit), epit(epit) {}
+};
 
 
 DocBookDocumentSectioning hasDocumentSectioning(ParagraphList const &paragraphs, pit_type bpit, pit_type const epit) {
@@ -847,6 +861,7 @@ bool hasOnlyNotes(Paragraph const & par)
 DocBookInfoTag getParagraphsWithInfo(ParagraphList const &paragraphs, pit_type bpit, pit_type const epit) {
        set<pit_type> shouldBeInInfo;
        set<pit_type> mustBeInInfo;
+       set<pit_type> abstract;
 
        // Find the first non empty paragraph by mutating bpit.
        while (bpit < epit) {
@@ -859,14 +874,18 @@ DocBookInfoTag getParagraphsWithInfo(ParagraphList const &paragraphs, pit_type b
 
        // Find the last info-like paragraph.
        pit_type cpit = bpit;
+       bool hasAbstractLayout = false;
        while (cpit < epit) {
                // Skip paragraphs only containing one note.
-               Paragraph const &par = paragraphs[cpit];
+               Paragraph const & par = paragraphs[cpit];
                if (hasOnlyNotes(par)) {
                        cpit += 1;
                        continue;
                }
 
+               if (par.layout().name() == from_ascii("Abstract"))
+                       hasAbstractLayout = true;
+
                // Based on layout information, store this paragraph in one set: should be in <info>, must be.
                Layout const &style = par.layout();
 
@@ -876,42 +895,55 @@ DocBookInfoTag getParagraphsWithInfo(ParagraphList const &paragraphs, pit_type b
                        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.
                        break;
                }
                cpit += 1;
        }
        // Now, cpit points to the last paragraph that has things that could go in <info>.
-       // bpit is still the beginning of the <info> part.
-
-       return make_tuple(shouldBeInInfo, mustBeInInfo, bpit, cpit);
-}
-
-
-bool hasAbstractBetween(ParagraphList const &paragraphs, pit_type const bpitAbstract, pit_type const epitAbstract)
-{
-       // Hypothesis: the paragraphs between bpitAbstract and epitAbstract can be considered an abstract because they
-       // are just after a document or part title.
-       if (epitAbstract - bpitAbstract <= 0)
-               return false;
-
-       // If there is something between these paragraphs, check if it's compatible with an abstract (i.e. some text).
-       pit_type bpit = bpitAbstract;
-       while (bpit < epitAbstract) {
-               const Paragraph &p = paragraphs.at(bpit);
-
-               if (p.layout().name() == from_ascii("Abstract"))
-                       return true;
+       // 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().name() == from_ascii("Abstract"))
+                               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.
+                                               }
 
-               if (!p.insetList().empty()) {
-                       for (const auto &i : p.insetList()) {
-                               if (i.inset->getText(0) != nullptr) {
-                                       return true;
+                                               abstract.emplace(pit);
+                                               break;
+                                       }
                                }
                        }
+                       pit++;
                }
-               bpit++;
        }
-       return false;
+
+       done:
+       return DocBookInfoTag(shouldBeInInfo, mustBeInInfo, abstract, bpit, cpit);
 }
 
 
@@ -947,26 +979,18 @@ void outputDocBookInfo(
                XMLStream & xs,
                OutputParams const & runparams,
                ParagraphList const & paragraphs,
-               DocBookInfoTag const & info,
-               pit_type bpitAbstract,
-               pit_type const epitAbstract)
+               DocBookInfoTag const & info)
 {
-       // Consider everything between bpitAbstract and epitAbstract (excluded) as paragraphs for the abstract.
-       // Use bpitAbstract >= epitAbstract to indicate there is no abstract.
-
-       set<pit_type> shouldBeInInfo;
-       set<pit_type> mustBeInInfo;
-       pit_type bpitInfo;
-       pit_type epitInfo;
-       tie(shouldBeInInfo, mustBeInInfo, bpitInfo, epitInfo) = info;
-
        // Perform an additional check on the abstract. Sometimes, there are many paragraphs that should go
        // into the abstract, but none generates actual content. Thus, first generate to a temporary stream,
        // then only create the <abstract> tag if these paragraphs generate some content.
        // This check must be performed *before* a decision on whether or not to output <info> is made.
-       bool hasAbstract = hasAbstractBetween(paragraphs, bpitAbstract, epitAbstract);
+       bool hasAbstract = !info.abstract.empty();
        docstring abstract;
        if (hasAbstract) {
+               pit_type bpitAbstract = *std::min_element(info.abstract.begin(), info.abstract.end());
+               pit_type epitAbstract = *std::max_element(info.abstract.begin(), info.abstract.end());
+
                odocstringstream os2;
                XMLStream xs2(os2);
                generateDocBookParagraphWithoutSectioning(text, buf, xs2, runparams, paragraphs, bpitAbstract, epitAbstract);
@@ -983,7 +1007,7 @@ void outputDocBookInfo(
        }
 
        // The abstract must go in <info>.
-       bool needInfo = !mustBeInInfo.empty() || hasAbstract;
+       bool needInfo = !info.mustBeInInfo.empty() || hasAbstract;
 
        // Start the <info> tag if required.
        if (needInfo) {
@@ -993,10 +1017,10 @@ void outputDocBookInfo(
        }
 
        // Output the elements that should go in <info>.
-       generateDocBookParagraphWithoutSectioning(text, buf, xs, runparams, paragraphs, bpitInfo, epitInfo);
+       generateDocBookParagraphWithoutSectioning(text, buf, xs, runparams, paragraphs, info.bpit, info.epit);
 
        if (hasAbstract && !abstract.empty()) { // The second test is probably superfluous.
-               string tag = paragraphs[bpitAbstract].layout().docbookforceabstracttag();
+               string tag = paragraphs[*info.abstract.begin()].layout().docbookforceabstracttag();
                if (tag == "NONE")
                        tag = "abstract";
 
@@ -1029,7 +1053,7 @@ void docbookFirstParagraphs(
        ParagraphList const &paragraphs = text.paragraphs();
        pit_type bpit = runparams.par_begin;
        DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, epit);
-       outputDocBookInfo(text, buf, xs, runparams, paragraphs, info, get<3>(info), epit);
+       outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
 }
 
 
@@ -1056,8 +1080,8 @@ void docbookSimpleAllParagraphs(
        pit_type bpit = runparams.par_begin;
        pit_type const epit = runparams.par_end;
        DocBookInfoTag info = getParagraphsWithInfo(paragraphs, bpit, epit);
-       outputDocBookInfo(text, buf, xs, runparams, paragraphs, info, 0, 0);
-       bpit = get<3>(info); // Generate the content starting from the end of the <info> part.
+       outputDocBookInfo(text, buf, xs, runparams, paragraphs, info);
+       bpit = info.bpit;
 
        // Then, the content.
        ParagraphList::const_iterator const pend =