From d69411e9988cbc17eeba6ddb558e5903aa01744f Mon Sep 17 00:00:00 2001
From: gwdwyy <wuyanyi09@gmail.com>
Date: Mon, 22 Jul 2013 14:49:00 +0800
Subject: [PATCH] rewriting trie.cpp/h

---
 src/Segment.cpp | 16 +++---------
 src/Trie.cpp    | 66 +++++++++++++++++++++++--------------------------
 src/Trie.h      |  4 +--
 src/globals.h   |  1 -
 4 files changed, 37 insertions(+), 50 deletions(-)

diff --git a/src/Segment.cpp b/src/Segment.cpp
index 7e8f74b..0baf1d9 100644
--- a/src/Segment.cpp
+++ b/src/Segment.cpp
@@ -16,8 +16,7 @@ namespace CppJieba
 
 	bool Segment::init()
 	{
-		bool retFlag;
-		retFlag = _trie.init();
+		bool retFlag = _trie.init();
 		if(!retFlag)
 		{
 			LogError("_trie.init failed.");
@@ -28,9 +27,8 @@ namespace CppJieba
 	
 	bool Segment::loadSegDict(const string& filePath)
 	{
-		bool retFlag;
-		retFlag = _trie.loadDict(filePath);
 		LogInfo(string_format("_trie.loadDict(%s) start...", filePath.c_str()));
+		bool retFlag = _trie.loadDict(filePath);
 		LogInfo("_trie.loadDict end.");
 		return retFlag;
 	}
@@ -48,10 +46,10 @@ namespace CppJieba
 		string uniStr = gEncoding.decode(str);
 		if(uniStr.empty())
 		{
-			LogError("_utf8ToUni failed.");
+			LogError("gEncoding.decode failed.");
 			return false;
 		}
-
+		
 		//calc DAG
 		vector<vector<uint> > dag;
 		retFlag = _calcDAG(uniStr, dag);
@@ -181,12 +179,8 @@ using namespace CppJieba;
 
 int main()
 {
-	/*
-	cout<<__FILE__<<__LINE__<<endl;
 	Segment segment;
-	cout<<__FILE__<<__LINE__<<endl;
 	segment.init();
-	cout<<__FILE__<<__LINE__<<endl;
 	if(!segment.loadSegDict("../dicts/segdict.utf8.v2.1"))
 	{
 		cerr<<"1"<<endl;
@@ -199,14 +193,12 @@ int main()
 	while(getline(ifile, line))
 	{
 		res.clear();
-		cout<<__FILE__<<__LINE__<<endl;
 		segment.cutDAG(line, res);
 		PRINT_VECTOR(res);
 		getchar();
 	}
 
 	segment.dispose();
-	*/
 	return 0;
 }
 
diff --git a/src/Trie.cpp b/src/Trie.cpp
index a659f9f..c913d7f 100644
--- a/src/Trie.cpp
+++ b/src/Trie.cpp
@@ -114,13 +114,7 @@ namespace CppJieba
 			//insert node
 			TrieNodeInfo nodeInfo;
 			nodeInfo.word = chWord;
-			size_t wLen = getUtf8WordLen(chWord);
-			if(0 == wLen)
-			{
-				LogFatal(string_format("getUtf8WordLen(%s) return 0", chWord.c_str()));
-				return false;
-			}
-			nodeInfo.wLen = wLen;
+			nodeInfo.wLen = 0;
 			nodeInfo.count = count;
 			nodeInfo.tag = tag;
 
@@ -162,28 +156,21 @@ namespace CppJieba
 			LogFatal("trie not initted!");
 			return NULL;
 		}
-		if(str.empty())
-		{
-			LogError("str is empty");
-			return NULL;
-		}
-		string uniStr = gEncoding.decode(str);
-		if(uniStr.empty())
+		vector<uint16_t> unicode;
+		
+		bool retFlag = gEncoding.decode(str, unicode);
+		if(retFlag)
 		{
 			LogError("gEncoding.decode failed.");
 			return NULL;
 		}
-		if(uniStr.size() % 2)
-		{
-			LogError("utf8ToUnicode return uniStr illegal");
-			return NULL;
-		}
+
 		//find
 		TrieNode* p = _root;
 		TrieNodeInfo * res = NULL;
-		for(uint i = 0; i < uniStr.size(); i+=2)
+		for(uint i = 0; i < unicode.size(); i++)
 		{
-			ChUnicode chUni = twocharToUint16(uniStr[0], uniStr[i+1]);
+			uint16_t chUni = unicode[i];
 			if(p->isLeaf)
 			{
 				uint pos = p->nodeInfoVecPos;
@@ -212,25 +199,32 @@ namespace CppJieba
 
 	const TrieNodeInfo* Trie::find(const string& str)
 	{
-		string uniStr = gEncoding.decode(str);
-		return _findUniStr(uniStr);
+		vector<uint16_t> unicode;
+		bool retFlag = gEncoding.decode(str, unicode);
+		if(!retFlag)
+		{
+			return NULL;
+		}
+		return find(unicode);
 	}
 
-	const TrieNodeInfo* Trie::_findUniStr(const string& uniStr)
+	const TrieNodeInfo* Trie::find(const vector<uint16_t>& unicode)
 	{
+		
 		if(!_getInitFlag())
 		{
 			LogFatal("trie not initted!");
 			return NULL;
 		}
-		if(uniStr.empty() || uniStr.size() % 2)
+		if(unicode.empty())
 		{
-			LogError("uniStr illegal");
+			LogError("unicode empty");
+			return NULL;
 		}
 		TrieNode* p = _root;
-		for(uint i = 0; i < uniStr.size(); i+=2)
+		for(uint i = 0; i < unicode.size(); i++)
 		{
-			ChUnicode chUni = twocharToUint16(uniStr[i], uniStr[i+1]);
+			uint16_t chUni = unicode[i];
 			if(p->hmap.find(chUni) == p-> hmap.end())
 			{
 				return NULL;
@@ -258,8 +252,10 @@ namespace CppJieba
 
 	double Trie::getWeight(const string& str)
 	{
-		string uniStr = gEncoding.decode(str);
-		const TrieNodeInfo * p = _findUniStr(uniStr);
+
+		vector<uint16_t> unicode;
+		gEncoding.decode(str, unicode);
+		const TrieNodeInfo * p = find(unicode);
 		if(NULL != p)
 		{
 			return p->weight;
@@ -303,17 +299,18 @@ namespace CppJieba
 
 		const string& word = nodeInfo.word;
 		
-		string uniStr = gEncoding.decode(word);
-		if(uniStr.empty())
+		vector<uint16_t> unicode;
+		bool retFlag = gEncoding.decode(word, unicode);
+		if(!retFlag)
 		{
 			LogError("gEncoding.decode error.");
 			return false;
 		}
 		
         TrieNode* p = _root;
-        for(uint i = 0; i < uniStr.size(); i+=2)
+        for(uint i = 0; i < unicode.size(); i++)
         {
-			ChUnicode cu = twocharToUint16(uniStr[i], uniStr[i+1]);
+			uint16_t cu = unicode[i];
 			if(NULL == p)
 			{
 				return false;
@@ -398,7 +395,6 @@ namespace CppJieba
 using namespace CppJieba;
 int main()
 {
-	cout<<__FILE__<<__FILE__<<endl;
     Trie trie;
     trie.init();
 	trie.loadDict("../dicts/segdict.utf8.v2.1");
diff --git a/src/Trie.h b/src/Trie.h
index 5d8a217..8163f33 100644
--- a/src/Trie.h
+++ b/src/Trie.h
@@ -26,7 +26,7 @@ namespace CppJieba
     using namespace CPPCOMMON;
     using namespace std;
     //using __gnu_cxx::hash_map;
-	typedef map<ChUnicode, struct TrieNode*> TrieNodeMap;
+	typedef map<uint16_t, struct TrieNode*> TrieNodeMap;
 
 	struct TrieNodeInfo
 	{
@@ -88,6 +88,7 @@ namespace CppJieba
 
 		public:
 			const TrieNodeInfo* find(const string& str);
+			const TrieNodeInfo* find(const vector<uint16_t>& unicode);
 			const TrieNodeInfo* findPrefix(const string& str);
 
 		public:
@@ -102,7 +103,6 @@ namespace CppJieba
 			bool _buildTree(const string& filePath);
 			bool _countWeight();
             bool _deleteNode(TrieNode* node);
-			const TrieNodeInfo* _findUniStr(const string& uniStr);
 
     };
 }
diff --git a/src/globals.h b/src/globals.h
index 1c73c45..4d89362 100644
--- a/src/globals.h
+++ b/src/globals.h
@@ -13,7 +13,6 @@ namespace CppJieba
 {
 
 	//typedefs
-	typedef uint16_t ChUnicode;
 	typedef unsigned int uint;
 	typedef std::vector<std::string>::iterator VSI;