[LyX/master] Revert "DocBook: simplify code to handle abstracts."

Thibaut Cuvelier tcuvelier at lyx.org
Mon Aug 3 20:33:02 UTC 2020


commit 42eca3d8d68a2e18701466b8669e45f722c4ea87
Author: Thibaut Cuvelier <tcuvelier at lyx.org>
Date:   Mon Aug 3 22:56:06 2020 +0200

    Revert "DocBook: simplify code to handle abstracts."
    
    This reverts commit 0199e048ffed026b909aa0e2d8161efb4f29a4a6.
---
 .../docbook/bibliography_precooked_aastex.xml      |    4 +-
 lib/layouts/aastex.layout                          |    6 -
 src/output_docbook.cpp                             |  130 ++++++++------------
 3 files changed, 54 insertions(+), 86 deletions(-)

diff --git a/autotests/export/docbook/bibliography_precooked_aastex.xml b/autotests/export/docbook/bibliography_precooked_aastex.xml
index e15bfdd..2c9f5bd 100644
--- a/autotests/export/docbook/bibliography_precooked_aastex.xml
+++ b/autotests/export/docbook/bibliography_precooked_aastex.xml
@@ -19,10 +19,8 @@
 <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 `keyword' when tag was not open. Tag discarded. -->
-<!-- Output Error: Tried to close `keywordset' when tag was not open. Tag discarded. -->
+<!-- Output Error: Tried to close `Keywords' 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 -->
diff --git a/lib/layouts/aastex.layout b/lib/layouts/aastex.layout
index 97c8b64..445d25f 100644
--- a/lib/layouts/aastex.layout
+++ b/lib/layouts/aastex.layout
@@ -282,9 +282,6 @@ Style Abstract
 	  Series	Bold
 	  Size		Normal
 	EndFont
-	DocBookTag            abstract
-	DocBookItemTag        para
-	DocBookInInfo         always
 End
 
 
@@ -334,9 +331,6 @@ Style Keywords
 	  Shape		Italic
 	  Size		Normal
 	EndFont
-	DocBookTag          keyword
-	DocBookWrapperTag   keywordset
-	DocBookInInfo       always
 End
 
 
diff --git a/src/output_docbook.cpp b/src/output_docbook.cpp
index fd38940..d0536ee 100644
--- a/src/output_docbook.cpp
+++ b/src/output_docbook.cpp
@@ -50,7 +50,7 @@ namespace lyx {
 
 namespace {
 
-std::string fontToDocBookTag(xml::FontTypes type)
+std::string const fontToDocBookTag(xml::FontTypes type)
 {
 	switch (type) {
 	case xml::FontTypes::FT_EMPH:
@@ -90,7 +90,6 @@ std::string fontToDocBookTag(xml::FontTypes type)
 	}
 }
 
-
 string fontToRole(xml::FontTypes type)
 {
 	// Specific fonts are achieved with roles. The only common ones are "" for basic emphasis,
@@ -104,13 +103,14 @@ string fontToRole(xml::FontTypes type)
 		return "";
 	case xml::FontTypes::FT_BOLD:
 		return "bold";
-	case xml::FontTypes::FT_NOUN: // Outputs a <person>
-	case xml::FontTypes::FT_TYPE: // Outputs a <code>
-		return "";
+	case xml::FontTypes::FT_NOUN:
+		return ""; // Outputs a <person>
+	case xml::FontTypes::FT_TYPE:
+		return ""; // Outputs a <code>
 	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";
@@ -814,21 +814,7 @@ pair<ParagraphList::const_iterator, ParagraphList::const_iterator> makeAny(
 
 
 using DocBookDocumentSectioning = tuple<bool, 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) {}
-};
+using DocBookInfoTag = tuple<set<pit_type>, set<pit_type>, pit_type, pit_type>;
 
 
 DocBookDocumentSectioning hasDocumentSectioning(ParagraphList const &paragraphs, pit_type bpit, pit_type const epit) {
@@ -861,7 +847,6 @@ 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) {
@@ -874,18 +859,14 @@ 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();
 
@@ -895,55 +876,42 @@ 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 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.
-						}
+	// bpit is still the beginning of the <info> part.
 
-						abstract.emplace(pit);
-						break;
-					}
+	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;
+
+		if (!p.insetList().empty()) {
+			for (const auto &i : p.insetList()) {
+				if (i.inset->getText(0) != nullptr) {
+					return true;
 				}
 			}
-			pit++;
 		}
+		bpit++;
 	}
-
-	done:
-	return DocBookInfoTag(shouldBeInInfo, mustBeInInfo, abstract, bpit, cpit);
+	return false;
 }
 
 
@@ -979,18 +947,26 @@ void outputDocBookInfo(
 		XMLStream & xs,
 		OutputParams const & runparams,
 		ParagraphList const & paragraphs,
-		DocBookInfoTag const & info)
+		DocBookInfoTag const & info,
+		pit_type bpitAbstract,
+		pit_type const epitAbstract)
 {
+	// 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 = !info.abstract.empty();
+	bool hasAbstract = hasAbstractBetween(paragraphs, bpitAbstract, epitAbstract);
 	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);
@@ -1007,7 +983,7 @@ void outputDocBookInfo(
 	}
 
 	// The abstract must go in <info>.
-	bool needInfo = !info.mustBeInInfo.empty() || hasAbstract;
+	bool needInfo = !mustBeInInfo.empty() || hasAbstract;
 
 	// Start the <info> tag if required.
 	if (needInfo) {
@@ -1017,10 +993,10 @@ void outputDocBookInfo(
 	}
 
 	// Output the elements that should go in <info>.
-	generateDocBookParagraphWithoutSectioning(text, buf, xs, runparams, paragraphs, info.bpit, info.epit);
+	generateDocBookParagraphWithoutSectioning(text, buf, xs, runparams, paragraphs, bpitInfo, epitInfo);
 
 	if (hasAbstract && !abstract.empty()) { // The second test is probably superfluous.
-		string tag = paragraphs[*info.abstract.begin()].layout().docbookforceabstracttag();
+		string tag = paragraphs[bpitAbstract].layout().docbookforceabstracttag();
 		if (tag == "NONE")
 			tag = "abstract";
 
@@ -1053,7 +1029,7 @@ 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);
+	outputDocBookInfo(text, buf, xs, runparams, paragraphs, info, get<3>(info), epit);
 }
 
 
@@ -1080,8 +1056,8 @@ 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);
-	bpit = info.bpit;
+	outputDocBookInfo(text, buf, xs, runparams, paragraphs, info, 0, 0);
+	bpit = get<3>(info); // Generate the content starting from the end of the <info> part.
 
 	// Then, the content.
 	ParagraphList::const_iterator const pend =


More information about the lyx-cvs mailing list