diff --git a/uppbox/Unicode/Unicode.cpp b/uppbox/Unicode/Unicode.cpp
index e5eeb80a8..c6c7856a2 100644
--- a/uppbox/Unicode/Unicode.cpp
+++ b/uppbox/Unicode/Unicode.cpp
@@ -2,6 +2,18 @@
 
 using namespace Upp;
 
+void Output(const String& id, const String& data)
+{
+	VppLog() << "byte " << id << "[] = { // " << data.GetCount() << " bytes";
+	for(int i = 0; i < data.GetCount(); i++) {
+		if((i & 1023) == 0) {
+			VppLog() << "\n\t";
+		}
+		VppLog() << Format("0x%02X,", (byte)data[i]);
+	}
+	VppLog() << "\n};\n";
+}
+
 void Delta(Vector<dword>& data)
 {
 	for(int i = data.GetCount() - 1; i > 0; i--)
@@ -33,22 +45,119 @@ String Encode(const Vector<dword>& data)
 	return r;
 }
 
+Vector<dword> Decode(const String& data)
+{
+	DDUMP(data.GetCount());
+	Vector<dword> r;
+	dword prev = 0;
+	StringStream ss(data);
+	while(!ss.IsEof()) {
+		byte ch = ss.Get();
+		dword val;
+		if(ch < 255) {
+			val = prev + ch - 127;
+		}
+		else {
+			val = ss.Get();
+			val |= ss.Get() << 8;
+			val |= ss.Get() << 16;
+			val |= ss.Get() << 24;
+			val = prev + val - 127;
+		}
+		r.Add(val);
+		prev = val;
+	}
+	return r;
+}
+
+dword ScanHex(const char *s)
+{
+	return Nvl(ScanInt(s, NULL, 16));
+}
+
+String Hex(dword d)
+{
+	return FormatIntHexUpper(d);
+}
+
+enum {
+	UNICODE_CATEGORY_Cc, UNICODE_CATEGORY_Cf, UNICODE_CATEGORY_Cn, UNICODE_CATEGORY_Co,
+	UNICODE_CATEGORY_Cs, UNICODE_CATEGORY_LC, UNICODE_CATEGORY_Ll, UNICODE_CATEGORY_Lm,
+	UNICODE_CATEGORY_Lo, UNICODE_CATEGORY_Lt, UNICODE_CATEGORY_Lu, UNICODE_CATEGORY_Mc,
+	UNICODE_CATEGORY_Me, UNICODE_CATEGORY_Mn, UNICODE_CATEGORY_Nd, UNICODE_CATEGORY_Nl,
+	UNICODE_CATEGORY_No, UNICODE_CATEGORY_Pc, UNICODE_CATEGORY_Pd, UNICODE_CATEGORY_Pe,
+	UNICODE_CATEGORY_Pf, UNICODE_CATEGORY_Pi, UNICODE_CATEGORY_Po, UNICODE_CATEGORY_Ps,
+	UNICODE_CATEGORY_Sc, UNICODE_CATEGORY_Sk, UNICODE_CATEGORY_Sm, UNICODE_CATEGORY_So,
+	UNICODE_CATEGORY_Zl, UNICODE_CATEGORY_Zp, UNICODE_CATEGORY_Zs
+};
+
 CONSOLE_APP_MAIN
 {
 	FileIn in(GetDataFile("UnicodeData.txt"));
-	Vector<dword> code, comb1, comb2, er, ercode;
+	Vector<dword> code, comb[3], er, ercode;
+	Vector<dword> catcode, cat;
+	Vector<dword> lettercode, letter;
+	VectorMap<String, int> catcount;
+	Index<String> kind;
+	kind.Add("Cc");
+	kind.Add("Cf");
+	kind.Add("Cn");
+	kind.Add("Co");
+	kind.Add("Cs");
+	kind.Add("LC");
+	kind.Add("Ll");
+	kind.Add("Lm");
+	kind.Add("Lo");
+	kind.Add("Lt");
+	kind.Add("Lu");
+	kind.Add("Mc");
+	kind.Add("Me");
+	kind.Add("Mn");
+	kind.Add("Nd");
+	kind.Add("Nl");
+	kind.Add("No");
+	kind.Add("Pc");
+	kind.Add("Pd");
+	kind.Add("Pe");
+	kind.Add("Pf");
+	kind.Add("Pi");
+	kind.Add("Po");
+	kind.Add("Ps");
+	kind.Add("Sc");
+	kind.Add("Sk");
+	kind.Add("Sm");
+	kind.Add("So");
+	kind.Add("Zl");
+	kind.Add("Zp");
+	kind.Add("Zs");
 	while(!in.IsEof()) {
 		Vector<String> h = Split(in.GetLine(), ';', false);
+//		DUMP(h);
+//		DUMPC(h);
 		if(h.GetCount() > 14) {
+			dword lower = ScanHex(h[13]);
+			dword upper = ScanHex(h[14]);
 			dword cde = ScanInt(h[0], NULL, 16);
-			Vector<String> comb;
-			comb.AppendRange(FilterRange(Split(h[5], ' '), [] (const String& x) { return x[0] != '<'; }));
-			if(comb.GetCount() > 1) {
+//			kind.FindAdd(h[2]);
+			Vector<String> decomb;
+			decomb.AppendRange(FilterRange(Split(h[5], ' '), [] (const String& x) { return x[0] != '<'; }));
+			if(decomb.GetCount() > 1) {
 				// DLOG(h[0] << " -> " << comb[0] << ' ' << comb[1]);
 				code.Add(cde);
-				comb1.Add(ScanInt(comb[0], NULL, 16));
-				comb2.Add(ScanInt(comb[1], NULL, 16));
+				decomb.Add("0");
+				comb[0].Add(ScanInt(decomb[0], NULL, 16));
+				comb[1].Add(ScanInt(decomb[1], NULL, 16));
+				comb[2].Add(ScanInt(decomb[2], NULL, 16));
 			}
+			if(cde > 2048) {
+				if(*h[2] == 'L') {
+					catcode.Add(cde);
+					cat.Add(cde);
+				}
+			}
+			catcount.GetAdd(h[2], 0)++;
+			catcode.Add(cde);
+			cat.Add(max(kind.Find(h[2]), 0));
 			dword other = Nvl(ScanInt(Nvl(h[12], h[13]), NULL, 16));
 			if(other) {
 			//	DUMP(other);
@@ -56,26 +165,49 @@ CONSOLE_APP_MAIN
 				ercode.Add(cde);
 				er.Add(other);
 			}
+			
+			dword t[MAX_DECOMPOSED];
+			UnicodeDecompose(cde, t);
+			dword composed_upper = 0;
+			dword composed_lower = 0;
+			if(t[0] < 128 && t[1]) {
+				t[0] = ToUpper((char)t[0]);
+				composed_upper = UnicodeCompose(t, 2);
+				t[0] = ToLower((char)t[0]);
+				composed_lower = UnicodeCompose(t, 2);
+			}
+			if(upper)
+				if(composed_upper = upper)
+					DLOG("Composed upper " << Hex(cde) << " -> " << Hex(composed_upper));
+				else
+					DLOG("Composed upper failed " << cde);
 		}
 	}
-
+/*
 	DDUMP(code.GetCount());
 	
-	String h = Encode(code);
-	DDUMP(h.GetCount());
-//	DUMPHEX(h);
+	Vector<dword> data;
+	data.Add(code.GetCount());
+	data.Append(code);
+	data.Append(comb[0]);
+	data.Append(comb[1]);
+	DUMP(comb[0]);
+	DUMP(comb[2]);
+	data.Append(comb[2]);
 
-	h << Encode(comb1);
-	DDUMP(h.GetCount());
-//	DUMPHEX(h);
+	DDUMP(letter.GetCount());
+	DUMPC(letter);
+	DUMP(ZCompress(Encode(letter)).GetCount());
+	DUMP(ZCompress(Encode(lettercode)).GetCount());
 
-	h << Encode(comb2);
-	DDUMP(h.GetCount());
-	DUMPHEX(h);
-
-	DDUMP(ZCompress(h).GetCount());
-	DDUMP(FastCompress(h).GetCount());
+	DUMP(cat.GetCount());
+	DUMP(ZCompress(Encode(cat)).GetCount());
+	DUMP(ZCompress(Encode(catcode)).GetCount());
 	
+	Output("unicode_info__", ZCompress(Encode(data)));
+*/	
+	DUMPM(catcount);
+	DUMP(Sum(catcount));
 #if 0
 	Delta(code);
 	Delta(comb1);