Changeset 13456

Timestamp:

2006-12-07T17:32:49+13:00 (17 years ago)

Author:

kjdon

Message:

cgi_safe was sometimes getting passed in unicode, sometimes utf-8. for one of these, it needs to encode chars about 127, for the other, chars above 255. The change I made last time, 255->127 mucked up the e arg for search terms e.g with diacritics. that change had been to make cgisafe work for metadata with diacritics in format statements. so now there are two cgi-safe methods, one for utf8, one for unicode. minus_safe also takes an arg for utf8 or not. Also, removed + from the list of chars to output directly. a space gets output as a +. so we need to encode + - this will enable a + to be used in a search string. hope it doesn't break anything else...

Location:

trunk/gsdl/src/recpt

Files:

• cgiutils.cpp (modified) (5 diffs)
• cgiutils.h (modified) (1 diff)

trunk/gsdl/src/recpt/cgiutils.cpp

@@ -340 +340 @@
     // convert %xx and + to their appropriate equivalents
     decode_cgi_arg (value);
-
     value.setencoding(1); // other encoding
     // store this key=value pair
@@ -416 +415 @@
 }
 
-text_t minus_safe (const text_t &intext) {
+// set utf8 to true if input is in utf-8, otherwise expects input in unicode
+text_t minus_safe (const text_t &intext, bool utf8) {
 
   text_t outtext;
@@ -428 +428 @@
     ++here;
   }
-  outtext = cgi_safe (outtext);
+  if (utf8) {
+    outtext = cgi_safe_utf8 (outtext);
+  } else {
+    outtext = cgi_safe_unicode (outtext);
+  }
   return outtext;
 }
 
-text_t cgi_safe (const text_t &intext) {
+// takes utf-8 input
+text_t cgi_safe_utf8 (const text_t &intext) {
   text_t outtext;
   
@@ -445 +450 @@
     ((c >= 'A') && (c <= 'Z')) ||
     ((c >= '0') && (c <= '9')) ||
-    (c == '+') || (c == '%') || (c == '-')) {
+    (c == '%') || (c == '-')) {
+      // alphanumeric character
+      outtext.push_back(c);
+    } else if (c == ' ') {
+      // space
+      outtext.push_back('+');
+    } else if (c > 255) { // not utf-8 character
+      cerr << "WARNING: expected utf-8 char, but got unicode!!\n";
+    } else {
+      // everything else
+      outtext.push_back('%');
+      c2hex(c, ttmp);
+      outtext += ttmp;
+    }
+    
+    ++here;
+  }
+  
+  return outtext;
+}
+// takes unicode input
+text_t cgi_safe_unicode (const text_t &intext) {
+  text_t outtext;
+  
+  text_t::const_iterator here = intext.begin ();
+  text_t::const_iterator end = intext.end ();
+  unsigned short c;
+  text_t ttmp;
+  
+  while (here != end) {
+    c = *here;
+    if (((c >= 'a') && (c <= 'z')) ||
+    ((c >= 'A') && (c <= 'Z')) ||
+    ((c >= '0') && (c <= '9')) ||
+    (c == '%') || (c == '-')) {
       // alphanumeric character
       outtext.push_back(c);
@@ -738 +777 @@
       // multiple character argument -- sort out any '-' chars
       if (args["w"]=="utf-16be") // browsers don't like \0 in urls...
-        compressed_args += minus_safe (args[argname]);
+        compressed_args += minus_safe (args[argname], false);
       else
-        compressed_args += minus_safe (outconvert.convert(args[argname]));
-
+        compressed_args += minus_safe (outconvert.convert(args[argname]), true);
+      
       if (saveconfhere != saveconfend) compressed_args.push_back ('-');
 

trunk/gsdl/src/recpt/cgiutils.h

@@ -47 +47 @@
 
 // turns any '-' in a cgi argument into "(-)"
-text_t minus_safe (const text_t &intext);
+// set utf8 to true if input is in utf-8, otherwise expects input in unicode
+text_t minus_safe (const text_t &intext, bool utf8);
 
 // returns the encoded version of a cgi argument
-text_t cgi_safe (const text_t &intext);
+// the original text should be in utf8
+text_t cgi_safe_utf8 (const text_t &intext);
+// the original text should be in unicode
+text_t cgi_safe_unicode (const text_t &intext);
 
 // check_save_conf_str checks the configuration string for