issue #6734 parsing performance worsened

9 files changed, 88 insertions, 45 deletions

diff --git a/src/commentscan.h b/src/commentscan.h
index d324969..75cd99f 100644
--- a/src/commentscan.h
+++ b/src/commentscan.h
@@ -24,6 +24,19 @@ class ParserInterface;
 /** @file
  *  @brief Interface for the comment block parser */
 
+/** Invokes the comment block parser with the request to preprocess a 
+ *  single comment block.
+ *  @param[in] comment A string representing the actual comment block.
+ *         Note that leading *'s are already stripped from the comment block.
+ *  @param[in] fileName The name of the file in which the comment is found.
+ *         Mainly used for producing warnings.
+ *  @param[in] lineNr The line number at which the comment block was found.
+ *  @returns The prepocessed comment block
+ */
+QCString preprocessCommentBlock(const QCString &comment,
+                                const QCString &fileName,
+                                int lineNr);
+
 /** Invokes the comment block parser with the request to parse a 
  *  single comment block.
  *  @param[in] parser The language parse that invoked this function.
diff --git a/src/commentscan.l b/src/commentscan.l
index 235cd7c..f9cc3b1 100644
--- a/src/commentscan.l
+++ b/src/commentscan.l
@@ -3074,6 +3074,33 @@ static void checkFormula()
 
 //----------------------------------------------------------------------------
 
+QCString preprocessCommentBlock(const QCString &comment,
+                                const QCString &fileName,
+                                int lineNr)
+{
+  if (!comment.isEmpty() && Doxygen::markdownSupport)
+  {
+    QCString result = processMarkdown(fileName,lineNr,0,comment);
+    const char *p = result.data();
+    if (p)
+    {
+      while (*p==' ')  p++; // skip over spaces
+      while (*p=='\n') p++; // skip over newlines
+      if (qstrncmp(p,"<br>",4)==0) p+=4; // skip over <br>
+    }
+    if (p>result.data())
+    {
+      // strip part of the input
+      result = result.mid(p-result.data());
+    }
+    return result;
+  }
+  else
+  {
+    return comment;
+  }
+}
+
 bool parseCommentBlock(/* in */     ParserInterface *parser,
                        /* in */     Entry *curEntry,
                        /* in */     const QCString &comment,
@@ -3096,26 +3123,7 @@ bool parseCommentBlock(/* in */     ParserInterface *parser,
   langParser     = parser;
   current        = curEntry;
   if (comment.isEmpty()) return FALSE; // avoid empty strings
-  if (Doxygen::markdownSupport)
-  {
-    inputString = processMarkdown(fileName,lineNr,NULL,comment);
-    const char *p = inputString.data();
-    if (p)
-    {
-      while (*p==' ')  p++; // skip over spaces
-      while (*p=='\n') p++; // skip over newlines
-      if (qstrncmp(p,"<br>",4)==0) p+=4; // skip over <br>
-    }
-    if (p>inputString.data())
-    {
-      // strip part of the input
-      inputString = inputString.mid(p-inputString.data());
-    }
-  }
-  else
-  {
-    inputString    = comment;
-  }
+  inputString    = comment;
   inputString.append(" ");
   inputPosition  = position;
   yyLineNr       = lineNr;
diff --git a/src/fortranscanner.l b/src/fortranscanner.l
index 5f10669..1f0c356 100644
--- a/src/fortranscanner.l
+++ b/src/fortranscanner.l
@@ -2411,9 +2411,7 @@ static void startCommentBlock(bool brief)
 
 static void handleCommentBlock(const QCString &doc,bool brief)
 {
-  bool needsEntry = FALSE;
   static bool hideInBodyDocs = Config_getBool(HIDE_IN_BODY_DOCS);
-  int position=0;
   if (docBlockInBody && hideInBodyDocs)
   {
     docBlockInBody = FALSE;
@@ -2421,10 +2419,13 @@ static void handleCommentBlock(const QCString &doc,bool brief)
   }
   DBG_CTX((stderr,"call parseCommentBlock [%s]\n",doc.data()));
   int lineNr = brief ? current->briefLine : current->docLine;
+  int position=0;
+  bool needsEntry = FALSE;
+  QCString processedDoc = preprocessCommentBlock(doc,yyFileName,lineNr);
   while (parseCommentBlock(
 	g_thisParser,
 	docBlockInBody ? subrCurrent.getFirst() : current,
-	doc,        // text
+	processedDoc, // text
 	yyFileName, // file
 	lineNr,
 	docBlockInBody ? FALSE : brief, 
diff --git a/src/growbuf.h b/src/growbuf.h
index bc8e4b5..4c49dce 100644
--- a/src/growbuf.h
+++ b/src/growbuf.h
@@ -11,11 +11,21 @@ class GrowBuf
 {
   public:
     GrowBuf() : str(0), pos(0), len(0) {}
+    GrowBuf(int initialSize) : pos(0), len(initialSize) { str=(char*)malloc(len); }
    ~GrowBuf()         { free(str); str=0; pos=0; len=0; }
     void clear()      { pos=0; }
     void addChar(char c)  { if (pos>=len) { len+=GROW_AMOUNT; str = (char*)realloc(str,len); } 
                         str[pos++]=c; 
                       }
+    void addStr(const QCString &s) {
+                        if (!s.isEmpty())
+                        {
+                          int l=s.length();
+                          if (pos+l>=len) { len+=l+GROW_AMOUNT; str = (char*)realloc(str,len); }
+                          strcpy(&str[pos],s.data());
+                          pos+=l;
+                        }
+                      }
     void addStr(const char *s) {
                         if (s)
                         {
diff --git a/src/markdown.cpp b/src/markdown.cpp
index b00070e..5253333 100644
--- a/src/markdown.cpp
+++ b/src/markdown.cpp
@@ -2451,8 +2451,8 @@ static QCString extractPageTitle(QCString &docs,QCString &id)
 static QCString detab(const QCString &s,int &refIndent)
 {
   static int tabSize = Config_getInt(TAB_SIZE);
-  GrowBuf out;
   int size = s.length();
+  GrowBuf out(size);
   const char *data = s.data();
   int i=0;
   int col=0;
@@ -2601,19 +2601,20 @@ void MarkdownFileParser::parseInput(const char *fileName,
     }
   }
   int lineNr=1;
-  int position=0;
 
   // even without markdown support enabled, we still 
   // parse markdown files as such
   bool markdownEnabled = Doxygen::markdownSupport;
   Doxygen::markdownSupport = TRUE;
 
-  bool needsEntry = FALSE;
   Protection prot=Public;
+  bool needsEntry = FALSE;
+  int position=0;
+  QCString processedDocs = preprocessCommentBlock(docs,fileName,lineNr);
   while (parseCommentBlock(
         this,
         current,
-        docs,
+        processedDocs,
         fileName,
         lineNr,
         FALSE,     // isBrief
diff --git a/src/pyscanner.l b/src/pyscanner.l
index 1596b9d..655b6e2 100644
--- a/src/pyscanner.l
+++ b/src/pyscanner.l
@@ -324,11 +324,12 @@ static void handleCommentBlock(const QCString &doc,bool brief)
   int position = 0;
   bool needsEntry;
   int lineNr = brief ? current->briefLine : current->docLine;
+  QCString processedDoc = preprocessCommentBlock(doc,yyFileName,lineNr);
   while (parseCommentBlock(
 	g_thisParser,
 	(docBlockInBody && previous) ? previous : current,
-	doc,     // text
-	yyFileName, // file
+	processedDoc, // text
+	yyFileName,   // file
 	lineNr,
 	docBlockInBody ? FALSE : brief, 
     docBlockJavaStyle, // javadoc style // or FALSE,
diff --git a/src/scanner.l b/src/scanner.l
index 7559e0b..86a83cb 100644
--- a/src/scanner.l
+++ b/src/scanner.l
@@ -6908,12 +6908,10 @@ static void newEntry()
 static void handleCommentBlock(const QCString &doc,bool brief)
 {
   static bool hideInBodyDocs = Config_getBool(HIDE_IN_BODY_DOCS);
-  int position=0;
-  bool needsEntry=FALSE;
   if (docBlockInBody && hideInBodyDocs) return;
   //printf("parseCommentBlock [%s] brief=%d\n",doc.data(),brief);
   int lineNr = brief ? current->briefLine : current->docLine;   // line of block start
-  
+
   // fill in inbodyFile && inbodyLine the first time, see bug 633891
   Entry *docEntry = docBlockInBody && previous ? previous : current;
   if (docBlockInBody && docEntry && docEntry->inbodyLine==-1)
@@ -6922,12 +6920,15 @@ static void handleCommentBlock(const QCString &doc,bool brief)
     docEntry->inbodyLine = lineNr;
   }
 
+  int position=0;
+  bool needsEntry=FALSE;
+  QCString processedDoc = preprocessCommentBlock(stripIndentation(doc),yyFileName,lineNr);
   while (parseCommentBlock(
 	g_thisParser,
 	docBlockInBody && previous ? previous : current,
-	stripIndentation(doc),        // text
-	yyFileName, // file
-	lineNr,     // line of block start
+	processedDoc, // text
+	yyFileName,   // file
+	lineNr,       // line of block start
 	docBlockInBody ? FALSE : brief,               // isBrief
 	docBlockInBody ? FALSE : docBlockAutoBrief,   // isJavaDocStyle
 	docBlockInBody,                               // isInBody
@@ -6935,7 +6936,7 @@ static void handleCommentBlock(const QCString &doc,bool brief)
         position,
         needsEntry
         )
-     ) 
+     )
   {
     //printf("parseCommentBlock position=%d [%s]\n",position,doc.data()+position);
     if (needsEntry) 
diff --git a/src/tclscanner.l b/src/tclscanner.l
index ca5294b..88714ce 100644
--- a/src/tclscanner.l
+++ b/src/tclscanner.l
@@ -1507,6 +1507,7 @@ tcl_inf("-> %s\n",(const char *)tcl.string_comment);
       myO.shrink(myO.curPos()-2);
       myO.addChar('\0');
       QCString myDoc = myO.data();
+      QCString processedDoc;
       if (what==99)
       { // inbody comment file or namespace or class or proc/method
         int myPos0;
@@ -1526,14 +1527,16 @@ tcl_inf("-> %s\n",(const char *)tcl.string_comment);
 
         myPos0=myPos;
         myLine0=myLine;
-        while (parseCommentBlock(tcl.this_parser, &myEntry0, myDoc, tcl.file_name,
+        processedDoc = preprocessCommentBlock(myDoc,tcl.file_name,myLine);
+        while (parseCommentBlock(tcl.this_parser, &myEntry0, processedDoc, tcl.file_name,
           myLine, FALSE, tcl.config_autobrief, FALSE, myProt, myPos, myNew))
         {
           if (myNew)
           { // we need a new entry in this case
             myNew=0;
             myEntry = tcl_entry_new();
-            parseCommentBlock(tcl.this_parser, myEntry, myDoc, tcl.file_name,
+            processedDoc = preprocessCommentBlock(myDoc,tcl.file_name,myLine0);
+            parseCommentBlock(tcl.this_parser, myEntry, processedDoc, tcl.file_name,
               myLine0, FALSE, tcl.config_autobrief, FALSE, myProt, myPos0, myNew);
             tcl.entry_inside->addSubEntry(myEntry);
           }
@@ -1543,7 +1546,8 @@ tcl_inf("-> %s\n",(const char *)tcl.string_comment);
             {
               myEntry1=tcl_entry_namespace(tcl.scan.at(0)->ns);
             }
-            parseCommentBlock(tcl.this_parser, myEntry1, myDoc, tcl.file_name,
+            processedDoc = preprocessCommentBlock(myDoc,tcl.file_name,myLine0);
+            parseCommentBlock(tcl.this_parser, myEntry1, processedDoc, tcl.file_name,
               myLine0, FALSE, tcl.config_autobrief, FALSE, myProt, myPos0, myNew);
           }
           myPos0=myPos;
@@ -1553,7 +1557,8 @@ tcl_inf("-> %s\n",(const char *)tcl.string_comment);
         { // we need a new entry
           myNew=0;
           myEntry = tcl_entry_new();
-          parseCommentBlock(tcl.this_parser, myEntry, myDoc, tcl.file_name,
+          processedDoc = preprocessCommentBlock(myDoc,tcl.file_name,myLine0);
+          parseCommentBlock(tcl.this_parser, myEntry, processedDoc, tcl.file_name,
             myLine0, FALSE, tcl.config_autobrief, FALSE, myProt, myPos0, myNew);
           tcl.entry_inside->addSubEntry(myEntry);
         }
@@ -1563,14 +1568,16 @@ tcl_inf("-> %s\n",(const char *)tcl.string_comment);
           {
             myEntry1=tcl_entry_namespace(tcl.scan.at(0)->ns);
           }
-          parseCommentBlock(tcl.this_parser, myEntry1, myDoc, tcl.file_name,
+          processedDoc = preprocessCommentBlock(myDoc,tcl.file_name,myLine0);
+          parseCommentBlock(tcl.this_parser, myEntry1, processedDoc, tcl.file_name,
             myLine0, FALSE, tcl.config_autobrief, FALSE, myProt, myPos0, myNew);
         }
       }
       else
       { // new entry
         tcl.entry_current = tcl_entry_new();
-        while (parseCommentBlock(tcl.this_parser, tcl.entry_current, myDoc,
+        processedDoc = preprocessCommentBlock(myDoc,tcl.file_name,myLine);
+        while (parseCommentBlock(tcl.this_parser, tcl.entry_current, processedDoc,
           tcl.file_name, myLine, FALSE, tcl.config_autobrief, FALSE,
           myProt, myPos, myNew))
         {
diff --git a/src/vhdljjparser.cpp b/src/vhdljjparser.cpp
index 81a7ca1..4a312bd 100644
--- a/src/vhdljjparser.cpp
+++ b/src/vhdljjparser.cpp
@@ -244,7 +244,6 @@ void VhdlParser::handleFlowComment(const char* doc)
 
 void VhdlParser::handleCommentBlock(const char* doc1,bool brief)
 {
-  int position=0;
   QCString doc;
   doc.append(doc1);
  // fprintf(stderr,"\n %s",doc.data());
@@ -258,7 +257,6 @@ void VhdlParser::handleCommentBlock(const char* doc1,bool brief)
 
   VhdlDocGen::prepareComment(doc);
 
-  bool needsEntry=FALSE;
   Protection protection=Public;
 
   if (oldEntry==current)
@@ -290,10 +288,13 @@ void VhdlParser::handleCommentBlock(const char* doc1,bool brief)
     current->stat=true;
   }
 
+  int position=0;
+  bool needsEntry=FALSE;
+  QCString processedDoc = preprocessCommentBlock(doc,yyFileName,iDocLine);
   while (parseCommentBlock(
         g_thisParser,
         current,
-        doc,        // text
+        processedDoc, // text
         yyFileName, // file
         iDocLine,   // line of block start
         brief,