X-Git-Url: http://git.indexdata.com/?a=blobdiff_plain;f=src%2Fhtml_parser.cpp;h=1a403d46b113b7c223c6136927791d7dcd52e6e0;hb=73c1b7597e636f6bd66d7fbcdd2f8e8d0ae342f5;hp=fba55951b2a9e8d7c8f57a0e3bb89db675bac858;hpb=2ee83c18eef23c98fd4e97d581401be193fdf41f;p=metaproxy-moved-to-github.git

diff --git a/src/html_parser.cpp b/src/html_parser.cpp
index fba5595..1a403d4 100644
--- a/src/html_parser.cpp
+++ b/src/html_parser.cpp
@@ -24,9 +24,11 @@ Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
 #include <stdlib.h>
 #include <ctype.h>
 #include <stdio.h>
+#include <yaz/matchstr.h>
 
 #define SPACECHR " \t\r\n\f"
 
+// http://www.whatwg.org/specs/web-apps/current-work/multipage/parsing.html
 
 namespace metaproxy_1 {
     class HTMLParser::Rep {
@@ -46,6 +48,7 @@ namespace metaproxy_1 {
         Rep();
         ~Rep();
         int m_verbose;
+        bool nest;
     };
 }
 
@@ -54,6 +57,7 @@ namespace mp = metaproxy_1;
 mp::HTMLParser::Rep::Rep()
 {
     m_verbose = 0;
+    nest = true;
 }
 
 mp::HTMLParser::Rep::~Rep()
@@ -95,7 +99,7 @@ static int skipSpace(const char *cp)
 static int skipName(const char *cp)
 {
     int i;
-    for (i = 0; cp[i] && !strchr(SPACECHR "/>=<", cp[i]); i++)
+    for (i = 0; cp[i] && !strchr(SPACECHR "/><=", cp[i]); i++)
         ;
     return i;
 }
@@ -146,7 +150,7 @@ int mp::HTMLParser::Rep::tagAttrs(HTMLParserEvent &event,
                                   const char *cp)
 {
     int i = skipSpace(cp);
-    while (cp[i] && cp[i] != '>' && cp[i] != '/' && cp[i] != '<')
+    while (cp[i] && !strchr("/><", cp[i]))
     {
         const char *attr_name = cp + i;
         int attr_len;
@@ -174,7 +178,7 @@ int mp::HTMLParser::Rep::tagEnd(HTMLParserEvent &event,
 {
     int i = 0;
     int close_it = 0;
-    for (; cp[i] && cp[i] != '/' && cp[i] != '>' && cp[i] != '<'; i++)
+    for (; cp[i] && !strchr("/><", cp[i]); i++)
         ;
     if (i > 0)
     {
@@ -218,7 +222,7 @@ void mp::HTMLParser::Rep::parse_str(HTMLParserEvent &event, const char *cp)
         if (*cp++ != '<')
             continue;
 
-        if (*cp == '!')
+        if (nest && *cp == '!')
         {
             int i;
             tagText(event, text_start, cp - 1);
@@ -244,7 +248,7 @@ void mp::HTMLParser::Rep::parse_str(HTMLParserEvent &event, const char *cp)
             cp += i;
             text_start = cp;
         }
-        else if (*cp == '?')
+        else if (nest && *cp == '?')
         {
             int i;
             tagText(event, text_start, cp - 1);
@@ -260,9 +264,19 @@ void mp::HTMLParser::Rep::parse_str(HTMLParserEvent &event, const char *cp)
         else if (*cp == '/' && isAlpha(cp[1]))
         {
             int i;
-            tagText(event, text_start, cp - 1);
-
+ 
             i = skipName(++cp);
+
+            if (!nest)
+            {
+                if (i == 6 && !yaz_strncasecmp(cp, "script", i))
+                    nest = true;
+                else
+                {
+                    continue;
+                }
+            }
+            tagText(event, text_start, cp - 2);
             event.closeTag(cp, i);
             if (m_verbose)
                 printf("------ tag close %.*s\n", i, cp);
@@ -270,7 +284,7 @@ void mp::HTMLParser::Rep::parse_str(HTMLParserEvent &event, const char *cp)
             cp += i;
             text_start = cp;
         }
-        else if (isAlpha(*cp))
+        else if (nest && isAlpha(*cp))
         {
             int i, j;
             tagText(event, text_start, cp - 1);
@@ -280,6 +294,10 @@ void mp::HTMLParser::Rep::parse_str(HTMLParserEvent &event, const char *cp)
                 printf("------ tag open %.*s\n", i, cp);
             j = tagAttrs(event, cp, i, cp + i);
             j += tagEnd(event, cp, i, cp + i + j);
+
+            if (i == 6 && !yaz_strncasecmp(cp, "script", i))
+                nest = false;
+
             cp += i + j;
             text_start = cp;
         }