Switch to side-by-side view

--- a/src/internfile/htmlparse.cpp
+++ b/src/internfile/htmlparse.cpp
@@ -1,11 +1,10 @@
-/* This file was copied from omega-0.8.5 and modified */
+/* This file was copied/updated from xapian-omega-1.0.1 and modified */
 
 /* htmlparse.cc: simple HTML parser for omega indexer
  *
- * ----START-LICENCE----
  * Copyright 1999,2000,2001 BrightStation PLC
  * Copyright 2001 Ananova Ltd
- * Copyright 2002 Olly Betts
+ * Copyright 2002,2006 Olly Betts
  *
  * This program is free software; you can redistribute it and/or
  * modify it under the terms of the GNU General Public License as
@@ -19,212 +18,99 @@
  *
  * You should have received a copy of the GNU General Public License
  * along with this program; if not, write to the Free Software
- * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307
+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301
  * USA
- * -----END-LICENCE-----
  */
 
 #ifndef lint
-static char rcsid[] = "@(#$Id: htmlparse.cpp,v 1.6 2006-01-30 11:15:27 dockes Exp $ ";
+static char rcsid[] = "@(#$Id: htmlparse.cpp,v 1.7 2007-06-19 10:28:40 dockes Exp $ ";
 #endif
 
-//#include <config.h>
-
 #include <algorithm>
-#ifndef NO_NAMESPACES
-using namespace std;
-//using std::find;
-//using std::find_if;
-#endif /* NO_NAMESPACES */
+using std::find;
+using std::find_if;
 #include "htmlparse.h"
 #include <stdio.h>
 #include <ctype.h>
 
-#include "transcode.h"
-
-map<string, string> HtmlParser::named_ents;
-
-inline static bool
-p_alpha(char c)
-{
-    return isalpha(c);
-}
+map<string, unsigned int> HtmlParser::named_ents;
 
 inline static bool
 p_notdigit(char c)
 {
-    return !isdigit(c);
+    return !isdigit(static_cast<unsigned char>(c));
 }
 
 inline static bool
 p_notxdigit(char c)
 {
-    return !isxdigit(c);
+    return !isxdigit(static_cast<unsigned char>(c));
 }
 
 inline static bool
 p_notalnum(char c)
 {
-    return !isalnum(c);
+    return !isalnum(static_cast<unsigned char>(c));
 }
 
 inline static bool
 p_notwhitespace(char c)
 {
-    return !isspace(c);
+    return !isspace(static_cast<unsigned char>(c));
 }
 
 inline static bool
 p_nottag(char c)
 {
-    return !isalnum(c) && c != '.' && c != '-';
+    return !isalnum(static_cast<unsigned char>(c)) &&
+	c != '.' && c != '-' && c != ':'; // ':' for XML namespaces.
 }
 
 inline static bool
 p_whitespacegt(char c)
 {
-    return isspace(c) || c == '>';
+    return isspace(static_cast<unsigned char>(c)) || c == '>';
 }
 
 inline static bool
 p_whitespaceeqgt(char c)
 {
-    return isspace(c) || c == '=' || c == '>';
-}
-
-/*
- * The following array was taken from Estraier. Estraier was
- * written by Mikio Hirabayashi. 
- *                Copyright (C) 2003-2004 Mikio Hirabayashi
- * The version where this comes from 
- * is covered by the GNU licence, as this file.*/
-static const char *epairs[] = {
-    /* basic symbols */
-    "amp", "&", "lt", "<", "gt", ">", "quot", "\"", "apos", "'",
-    /* ISO-8859-1 */
-    "nbsp", "\xc2\xa0", "iexcl", "\xc2\xa1", "cent", "\xc2\xa2",
-    "pound", "\xc2\xa3", "curren", "\xc2\xa4", "yen", "\xc2\xa5",
-    "brvbar", "\xc2\xa6", "sect", "\xc2\xa7", "uml", "\xc2\xa8",
-    "copy", "\xc2\xa9", "ordf", "\xc2\xaa", "laquo", "\xc2\xab",
-    "not", "\xc2\xac", "shy", "\xc2\xad", "reg", "\xc2\xae",
-    "macr", "\xc2\xaf", "deg", "\xc2\xb0", "plusmn", "\xc2\xb1",
-    "sup2", "\xc2\xb2", "sup3", "\xc2\xb3", "acute", "\xc2\xb4",
-    "micro", "\xc2\xb5", "para", "\xc2\xb6", "middot", "\xc2\xb7",
-    "cedil", "\xc2\xb8", "sup1", "\xc2\xb9", "ordm", "\xc2\xba",
-    "raquo", "\xc2\xbb", "frac14", "\xc2\xbc", "frac12", "\xc2\xbd",
-    "frac34", "\xc2\xbe", "iquest", "\xc2\xbf", "Agrave", "\xc3\x80",
-    "Aacute", "\xc3\x81", "Acirc", "\xc3\x82", "Atilde", "\xc3\x83",
-    "Auml", "\xc3\x84", "Aring", "\xc3\x85", "AElig", "\xc3\x86",
-    "Ccedil", "\xc3\x87", "Egrave", "\xc3\x88", "Eacute", "\xc3\x89",
-    "Ecirc", "\xc3\x8a", "Euml", "\xc3\x8b", "Igrave", "\xc3\x8c",
-    "Iacute", "\xc3\x8d", "Icirc", "\xc3\x8e", "Iuml", "\xc3\x8f",
-    "ETH", "\xc3\x90", "Ntilde", "\xc3\x91", "Ograve", "\xc3\x92",
-    "Oacute", "\xc3\x93", "Ocirc", "\xc3\x94", "Otilde", "\xc3\x95",
-    "Ouml", "\xc3\x96", "times", "\xc3\x97", "Oslash", "\xc3\x98",
-    "Ugrave", "\xc3\x99", "Uacute", "\xc3\x9a", "Ucirc", "\xc3\x9b",
-    "Uuml", "\xc3\x9c", "Yacute", "\xc3\x9d", "THORN", "\xc3\x9e",
-    "szlig", "\xc3\x9f", "agrave", "\xc3\xa0", "aacute", "\xc3\xa1",
-    "acirc", "\xc3\xa2", "atilde", "\xc3\xa3", "auml", "\xc3\xa4",
-    "aring", "\xc3\xa5", "aelig", "\xc3\xa6", "ccedil", "\xc3\xa7",
-    "egrave", "\xc3\xa8", "eacute", "\xc3\xa9", "ecirc", "\xc3\xaa",
-    "euml", "\xc3\xab", "igrave", "\xc3\xac", "iacute", "\xc3\xad",
-    "icirc", "\xc3\xae", "iuml", "\xc3\xaf", "eth", "\xc3\xb0",
-    "ntilde", "\xc3\xb1", "ograve", "\xc3\xb2", "oacute", "\xc3\xb3",
-    "ocirc", "\xc3\xb4", "otilde", "\xc3\xb5", "ouml", "\xc3\xb6",
-    "divide", "\xc3\xb7", "oslash", "\xc3\xb8", "ugrave", "\xc3\xb9",
-    "uacute", "\xc3\xba", "ucirc", "\xc3\xbb", "uuml", "\xc3\xbc",
-    "yacute", "\xc3\xbd", "thorn", "\xc3\xbe", "yuml", "\xc3\xbf",
-    /* ISO-10646 */
-    "fnof", "\xc6\x92", "Alpha", "\xce\x91", "Beta", "\xce\x92",
-    "Gamma", "\xce\x93", "Delta", "\xce\x94", "Epsilon", "\xce\x95",
-    "Zeta", "\xce\x96", "Eta", "\xce\x97", "Theta", "\xce\x98",
-    "Iota", "\xce\x99", "Kappa", "\xce\x9a", "Lambda", "\xce\x9b",
-    "Mu", "\xce\x9c", "Nu", "\xce\x9d", "Xi", "\xce\x9e",
-    "Omicron", "\xce\x9f", "Pi", "\xce\xa0", "Rho", "\xce\xa1",
-    "Sigma", "\xce\xa3", "Tau", "\xce\xa4", "Upsilon", "\xce\xa5",
-    "Phi", "\xce\xa6", "Chi", "\xce\xa7", "Psi", "\xce\xa8",
-    "Omega", "\xce\xa9", "alpha", "\xce\xb1", "beta", "\xce\xb2",
-    "gamma", "\xce\xb3", "delta", "\xce\xb4", "epsilon", "\xce\xb5",
-    "zeta", "\xce\xb6", "eta", "\xce\xb7", "theta", "\xce\xb8",
-    "iota", "\xce\xb9", "kappa", "\xce\xba", "lambda", "\xce\xbb",
-    "mu", "\xce\xbc", "nu", "\xce\xbd", "xi", "\xce\xbe",
-    "omicron", "\xce\xbf", "pi", "\xcf\x80", "rho", "\xcf\x81",
-    "sigmaf", "\xcf\x82", "sigma", "\xcf\x83", "tau", "\xcf\x84",
-    "upsilon", "\xcf\x85", "phi", "\xcf\x86", "chi", "\xcf\x87",
-    "psi", "\xcf\x88", "omega", "\xcf\x89", "thetasym", "\xcf\x91",
-    "upsih", "\xcf\x92", "piv", "\xcf\x96", "bull", "\xe2\x80\xa2",
-    "hellip", "\xe2\x80\xa6", "prime", "\xe2\x80\xb2", "Prime", "\xe2\x80\xb3",
-    "oline", "\xe2\x80\xbe", "frasl", "\xe2\x81\x84", "weierp", "\xe2\x84\x98",
-    "image", "\xe2\x84\x91", "real", "\xe2\x84\x9c", "trade", "\xe2\x84\xa2",
-    "alefsym", "\xe2\x84\xb5", "larr", "\xe2\x86\x90", "uarr", "\xe2\x86\x91",
-    "rarr", "\xe2\x86\x92", "darr", "\xe2\x86\x93", "harr", "\xe2\x86\x94",
-    "crarr", "\xe2\x86\xb5", "lArr", "\xe2\x87\x90", "uArr", "\xe2\x87\x91",
-    "rArr", "\xe2\x87\x92", "dArr", "\xe2\x87\x93", "hArr", "\xe2\x87\x94",
-    "forall", "\xe2\x88\x80", "part", "\xe2\x88\x82", "exist", "\xe2\x88\x83",
-    "empty", "\xe2\x88\x85", "nabla", "\xe2\x88\x87", "isin", "\xe2\x88\x88",
-    "notin", "\xe2\x88\x89", "ni", "\xe2\x88\x8b", "prod", "\xe2\x88\x8f",
-    "sum", "\xe2\x88\x91", "minus", "\xe2\x88\x92", "lowast", "\xe2\x88\x97",
-    "radic", "\xe2\x88\x9a", "prop", "\xe2\x88\x9d", "infin", "\xe2\x88\x9e",
-    "ang", "\xe2\x88\xa0", "and", "\xe2\x88\xa7", "or", "\xe2\x88\xa8",
-    "cap", "\xe2\x88\xa9", "cup", "\xe2\x88\xaa", "int", "\xe2\x88\xab",
-    "there4", "\xe2\x88\xb4", "sim", "\xe2\x88\xbc", "cong", "\xe2\x89\x85",
-    "asymp", "\xe2\x89\x88", "ne", "\xe2\x89\xa0", "equiv", "\xe2\x89\xa1",
-    "le", "\xe2\x89\xa4", "ge", "\xe2\x89\xa5", "sub", "\xe2\x8a\x82",
-    "sup", "\xe2\x8a\x83", "nsub", "\xe2\x8a\x84", "sube", "\xe2\x8a\x86",
-    "supe", "\xe2\x8a\x87", "oplus", "\xe2\x8a\x95", "otimes", "\xe2\x8a\x97",
-    "perp", "\xe2\x8a\xa5", "sdot", "\xe2\x8b\x85", "lceil", "\xe2\x8c\x88",
-    "rceil", "\xe2\x8c\x89", "lfloor", "\xe2\x8c\x8a", "rfloor", "\xe2\x8c\x8b",
-    "lang", "\xe2\x8c\xa9", "rang", "\xe2\x8c\xaa", "loz", "\xe2\x97\x8a",
-    "spades", "\xe2\x99\xa0", "clubs", "\xe2\x99\xa3", "hearts", "\xe2\x99\xa5",
-    "diams", "\xe2\x99\xa6", "OElig", "\xc5\x92", "oelig", "\xc5\x93",
-    "Scaron", "\xc5\xa0", "scaron", "\xc5\xa1", "Yuml", "\xc5\xb8",
-    "circ", "\xcb\x86", "tilde", "\xcb\x9c", "ensp", "\xe2\x80\x82",
-    "emsp", "\xe2\x80\x83", "thinsp", "\xe2\x80\x89", "zwnj", "\xe2\x80\x8c",
-    "zwj", "\xe2\x80\x8d", "lrm", "\xe2\x80\x8e", "rlm", "\xe2\x80\x8f",
-    "ndash", "\xe2\x80\x93", "mdash", "\xe2\x80\x94", "lsquo", "\xe2\x80\x98",
-    "rsquo", "\xe2\x80\x99", "sbquo", "\xe2\x80\x9a", "ldquo", "\xe2\x80\x9c",
-    "rdquo", "\xe2\x80\x9d", "bdquo", "\xe2\x80\x9e", "dagger", "\xe2\x80\xa0",
-    "Dagger", "\xe2\x80\xa1", "permil", "\xe2\x80\xb0", "lsaquo", "\xe2\x80\xb9",
-    "rsaquo", "\xe2\x80\xba", "euro", "\xe2\x82\xac",
-    NULL, NULL
-};
+    return isspace(static_cast<unsigned char>(c)) || c == '=' || c == '>';
+}
 
 HtmlParser::HtmlParser()
 {
+    // RECOLL: no need to initialize these entities, we use those from
+    // myhtmlparse
+#if 0
+    static const struct ent { const char *n; unsigned int v; } ents[] = {
+#include "namedentities.h"
+	{ NULL, 0 }
+    };
     if (named_ents.empty()) {
-	for (int i = 0;;) {
-	    const char *ent;
-	    const char *val;
-	    ent = epairs[i++];
-	    if (ent == 0) 
-		break;
-	    val = epairs[i++];
-	    if (val == 0) 
-		break;
-	    named_ents[string(ent)] = val;
+	const struct ent *i = ents;
+	while (i->n) {
+	    named_ents[string(i->n)] = i->v;
+	    ++i;
 	}
     }
+#endif
 }
 
 void
 HtmlParser::decode_entities(string &s)
 {
-    // This has no meaning whatsoever if the character encoding is unknown,
-    // so don't do it. If charset known, caller has converted text to utf-8, 
-    // and this is also how we translate entities
-    //    if (charset != "utf-8")
-    //    	return;
-
+    // Not used for recoll. Kept here to minimize the amount of diffs
+#if 0
     // We need a const_iterator version of s.end() - otherwise the
     // find() and find_if() templates don't work...
     string::const_iterator amp = s.begin(), s_end = s.end();
     while ((amp = find(amp, s_end, '&')) != s_end) {
 	unsigned int val = 0;
 	string::const_iterator end, p = amp + 1;
-	string subs;
 	if (p != s_end && *p == '#') {
 	    p++;
-	    if (p != s_end && tolower(*p) == 'x') {
+	    if (p != s_end && (*p == 'x' || *p == 'X')) {
 		// hex
 		p++;
 		end = find_if(p, s_end, p_notxdigit);
@@ -237,59 +123,95 @@
 	} else {
 	    end = find_if(p, s_end, p_notalnum);
 	    string code = s.substr(p - s.begin(), end - p);
-	    map<string, string>::const_iterator i;
+	    map<string, unsigned int>::const_iterator i;
 	    i = named_ents.find(code);
-	    if (i != named_ents.end()) 
-		subs = i->second;
-	}
-
-	if (end < s_end && *end == ';') 
-	    end++;
-	
+	    if (i != named_ents.end()) val = i->second;
+	}
+	if (end < s_end && *end == ';') end++;
 	if (val) {
-	    // The code is the code position for a unicode char. We need
-	    // to translate it to an utf-8 string.
-	    string utf16be;
-	    utf16be += char(val / 256);
-	    utf16be += char(val % 256);
-	    transcode(utf16be, subs, "UTF-16BE", "UTF-8");
-	} 
-
-	if (subs.length() > 0) {
 	    string::size_type amp_pos = amp - s.begin();
-	    s.replace(amp_pos, end - amp, subs);
+	    if (val < 0x80) {
+		s.replace(amp_pos, end - amp, 1u, char(val));
+	    } else {
+		// Convert unicode value val to UTF-8.
+		char seq[4];
+		unsigned len = Xapian::Unicode::nonascii_to_utf8(val, seq);
+		s.replace(amp_pos, end - amp, seq, len);
+	    }
 	    s_end = s.end();
 	    // We've modified the string, so the iterators are no longer
 	    // valid...
-	    amp = s.begin() + amp_pos + subs.length();
+	    amp = s.begin() + amp_pos + 1;
 	} else {
 	    amp = end;
 	}
     }
+#endif
 }
 
 void
 HtmlParser::parse_html(const string &body)
 {
+    in_script = false;
+
     map<string,string> Param;
     string::const_iterator start = body.begin();
 
-    while (1) {
+    while (true) {
+	// Skip through until we find an HTML tag, a comment, or the end of
+	// document.  Ignore isolated occurences of `<' which don't start
+	// a tag or comment.	
 	string::const_iterator p = start;
-
-	// Eat text until we find an HTML tag, a comment, or the end
-	// of document.  Ignore isolated occurences of `<' which don't
-	// start a tag or comment
-	while (1) {
+	while (true) {
 	    p = find(p, body.end(), '<');
 	    if (p == body.end()) break;
-	    char ch = *(p + 1);
-	    // tag, closing tag, comment (or SGML declaration), or PHP
-	    if (isalpha(ch) || ch == '/' || ch == '!' || ch == '?') break;
+	    unsigned char ch = *(p + 1);
+
+	    // Tag, closing tag, or comment (or SGML declaration).
+	    if ((!in_script && isalpha(ch)) || ch == '/' || ch == '!') break;
+	    if (ch == '?') {
+		// PHP code or XML declaration.
+		// XML declaration is only valid at the start of the first line.
+		// FIXME: need to deal with BOMs...
+		if (p != body.begin() || body.size() < 20) break;
+
+		// XML declaration looks something like this:
+		// <?xml version="1.0" encoding="UTF-8"?>
+		if (p[2] != 'x' || p[3] != 'm' || p[4] != 'l') break;
+		if (strchr(" \t\r\n", p[5]) == NULL) break;
+
+		string::const_iterator decl_end = find(p + 6, body.end(), '?');
+		if (decl_end == body.end()) break;
+
+		// Default charset for XML is UTF-8.
+		charset = "UTF-8";
+
+		string decl(p + 6, decl_end);
+		size_t enc = decl.find("encoding");
+		if (enc == string::npos) break;
+
+		enc = decl.find_first_not_of(" \t\r\n", enc + 8);
+		if (enc == string::npos || enc == decl.size()) break;
+
+		if (decl[enc] != '=') break;
+		
+		enc = decl.find_first_not_of(" \t\r\n", enc + 1);
+		if (enc == string::npos || enc == decl.size()) break;
+
+		if (decl[enc] != '"' && decl[enc] != '\'') break;
+
+		char quote = decl[enc++];
+		size_t enc_end = decl.find(quote, enc);
+
+		if (enc != string::npos)
+		    charset = decl.substr(enc, enc_end - enc);
+
+		break;
+	    }
 	    p++; 
 	}
 
-	// Process text
+	// Process text up to start of tag.
 	if (p > start || p == body.end()) {
 	    string text = body.substr(start - body.begin(), p - start);
 	    decode_entities(text);
@@ -310,19 +232,32 @@
 	    if (++start == body.end()) break;
 	    // comment or SGML declaration
 	    if (*(start - 1) == '-' && *start == '-') {
-		start = find(start + 1, body.end(), '>');
-		// unterminated comment swallows rest of document
-		// (like NS, but unlike MSIE iirc)
-		if (start == body.end()) break;
-		
-		p = start;
+		++start;
+		string::const_iterator close = find(start, body.end(), '>');
+		// An unterminated comment swallows rest of document
+		// (like Netscape, but unlike MSIE IIRC)
+		if (close == body.end()) break;
+
+		p = close;
 		// look for -->
 		while (p != body.end() && (*(p - 1) != '-' || *(p - 2) != '-'))
 		    p = find(p + 1, body.end(), '>');
 
-		// If we found --> skip to there, otherwise
-		// skip to the first > we found (as Netscape does)
-		if (p != body.end()) start = p;
+		if (p != body.end()) {
+		    // Check for htdig's "ignore this bit" comments.
+		    if (p - start == 15 && string(start, p - 2) == "htdig_noindex") {
+			string::size_type i;
+			i = body.find("<!--/htdig_noindex-->", p + 1 - body.begin());
+			if (i == string::npos) break;
+			start = body.begin() + i + 21;
+			continue;
+		    }
+		    // If we found --> skip to there.
+		    start = p;
+		} else {
+		    // Otherwise skip to the first > we found (as Netscape does).
+		    start = close;
+		}
 	    } else {
 		// just an SGML declaration, perhaps giving the DTD - ignore it
 		start = find(start - 1, body.end(), '>');
@@ -354,11 +289,12 @@
 	    start = find_if(start, body.end(), p_nottag);
 	    string tag = body.substr(p - body.begin(), start - p);
 	    // convert tagname to lowercase
-	    for (string::iterator i = tag.begin(); i != tag.end(); i++)
-		*i = tolower(*i);
+	    for (string::iterator i = tag.begin(); i != tag.end(); ++i)
+		*i = tolower(static_cast<unsigned char>(*i));
 	       
 	    if (closing) {
 		closing_tag(tag);
+		if (in_script && tag == "script") in_script = false;
 		   
 		/* ignore any bogus parameters on closing tags */
 		p = find(start, body.end(), '>');
@@ -402,8 +338,8 @@
 			if (name.size()) {
 			    // convert parameter name to lowercase
 			    string::iterator i;
-			    for (i = name.begin(); i != name.end(); i++)
-				*i = tolower(*i);
+			    for (i = name.begin(); i != name.end(); ++i)
+				*i = tolower(static_cast<unsigned char>(*i));
 			    // in case of multiple entries, use the first
 			    // (as Netscape does)
 			    if (Param.find(name) == Param.end())
@@ -414,6 +350,10 @@
 		opening_tag(tag, Param);
 		Param.clear();
 
+		// In <script> tags we ignore opening tags to avoid problems
+		// with "a<b".
+		if (tag == "script") in_script = true;
+
 		if (start != body.end() && *start == '>') ++start;
 	    }
 	}