add NFC_Compose mapping, which is generated arithmetically from NFC_Decompose

server/nfc.cc

@@ -371,3 +371,19 @@ std::string toNFC(std::string s)
 	// TODO:
 	throw std::logic_error("NFC normalization is necessary, but unimplemented. Sorry.");
 }
+
+
+// used only to initialize the NFC Compose mapping:
+std::map< std::pair<unsigned, unsigned>, unsigned> generate_nfc_compose()
+{
+	std::map< std::pair<unsigned, unsigned>, unsigned> m;
+	for(const auto& decomp : NFC_Decompose)
+	{
+		if(decomp.second.second >= 0) // skip singleton decompositions
+		{
+			m[ decomp.second ] = decomp.first;
+		}
+	}
+	
+	return m;
+}

server/nfc_sets.cc

@@ -3013,3 +3013,5 @@ const std::map<unsigned, std::pair<int,int>> NFC_Decompose = {
 {0x2FA1D, {0x2A600, -1}},
 };
 
+std::map< std::pair<unsigned, unsigned>, unsigned> generate_nfc_compose();
+const std::map< std::pair<unsigned, unsigned>, unsigned> NFC_Compose = generate_nfc_compose();

server/nfc_sets.hh

@@ -8,6 +8,8 @@
 // from Unicode's DerivedNormalizationProps.txt and UnicodeData.txt.
 // see scripts/ subdirectory
 
+// TODO: (maybe) Replace them by flat_map or sorted arrays, because these might be faster. But make benchmarks first!
+
 // Contains all codepoints with NFC_No property.
 extern const std::set<unsigned> NFC_No;
 
@@ -20,4 +22,7 @@ extern const std::map<unsigned, unsigned char> NFC_CombiningClass;
 // Contains the canonical decomposing pairs. second member might be -1 for single decompositions.
 extern const std::map<unsigned, std::pair<int,int>> NFC_Decompose;
 
+// canonical composing mapping, except excluded ones according to Unicode TR-15
+extern const std::map< std::pair<unsigned, unsigned>, unsigned> NFC_Compose;
+
 #endif // NFC_SETS_HH

server/scripts/gen_sets.sh

@@ -93,4 +93,7 @@ cat /usr/share/unicode/UnicodeData.txt | cut -d';' -f 1,6 | grep -v '<' | \
 
 echo -en '};\n\n'
 
+echo 'std::map< std::pair<unsigned, unsigned>, unsigned> generate_nfc_compose();'
+echo -en 'const std::map< std::pair<unsigned, unsigned>, unsigned> NFC_Compose = generate_nfc_compose();\n\n'
+
 # end of file