/// <summary>
/// 建索引主體
/// </summary>
/// <returns></returns>
public bool MakeIndex()
{
try
{
// 1. 解文
ParseSourceDoc();
// 2. 建索引
ChineseSegmentor segmentor = new ChineseSegmentor();
var server = MongoDbLib.GetServerConnection(_indexer.MongoDbServer);
var database = server.GetDatabase(_indexer.DbName);
var tblSourceText = database.GetCollection <SourceText>(Constants.TblSourceText);
// 斷詞，處理每個 Token
var sourcees = from s in tblSourceText.AsQueryable <SourceText>()
orderby s.DocId, s.ParaId
select s;
Dictionary <String, InvertedIndex> fullIndexes = new Dictionary <String, InvertedIndex>();
InvertedIndex aIndex = null;
foreach (var aSourceText in sourcees)
{
List <Pair <String, Int32> > result = segmentor.SegWords(aSourceText.Para);
foreach (var aToken in result)
{
if (fullIndexes.ContainsKey(aToken.First))
{
aIndex = fullIndexes[aToken.First];
}
else
{
aIndex = new InvertedIndex();
aIndex.Word = aToken.First;
}
aIndex.Indexes.Add(new IndexElement()
{
DocId = aSourceText.DocId,
ParaId = aSourceText.ParaId,
Offset = aToken.Second
});
fullIndexes[aToken.First] = aIndex;
}
}
// 在 Storage 存入 Word List
var wordListCollection = database.GetCollection(Constants.TblWordList);
List <BsonDocument> batch = new List <BsonDocument>();
List <String> wordList = fullIndexes.Keys.ToList();
for (int wordId = 0; wordId < fullIndexes.Count; wordId++)
{
aIndex = fullIndexes[wordList[wordId]];
aIndex.WordId = wordId;
batch.Add(new BsonDocument()
{
{ "Word", wordList[wordId] },
{ "WordId", wordId }
});
}
wordListCollection.InsertBatch(batch);
// 儲存全文索引
var tblFullText = database.GetCollection(Constants.TblFullText);
List <InvertedIndex> fullText = new List <InvertedIndex>();
tblFullText.InsertBatch <InvertedIndex>(fullIndexes.Values.ToList());
return(true);
}
catch (Exception e)
{
Console.WriteLine(e.StackTrace);
return(false);
}
}
public SearchResult Search(Repository repo, String keyword, int startPos, int fetchSize)
{
try
{
SearchResult result = new SearchResult();
// 計時器
Stopwatch sw = new Stopwatch();
sw.Reset();
sw.Start();
// MongoDb 初始化
var server = MongoDbLib.GetServerConnection(MongoDbServer);
var database = server.GetDatabase(DbName);
var tblWordList = database.GetCollection <WordItem>(Constants.TblWordList);
var tblFullText = database.GetCollection <InvertedIndex>(Constants.TblFullText);
// 針對搜尋關鍵字斷詞
List <Pair <String, Int32> > keywordTokens = _segmentor.SegWords(keyword);
// 自索引中取出對應的 word list
var buf = (from t in keywordTokens select t.First).ToList();
var query = from w in tblWordList.AsQueryable <WordItem>()
where w.Word.In(buf)
select new { w.WordId };
List <Int32> wordIdList = new List <Int32>();
foreach (var aWord in query)
{
wordIdList.Add(aWord.WordId);
}
// word id 為 0 筆，表示搜尋結果為 0
if (wordIdList.Count == 0)
{
sw.Stop();
result.SearchTime = sw.ElapsedMilliseconds / 1000.0;
return(result);
}
// 自全文索引中，取出對應的記錄
var indexes = from i in tblFullText.AsQueryable <InvertedIndex>()
where i.WordId.In(wordIdList)
select i;
if (indexes.Count() != wordIdList.Count)
{
return(null);
}
// 將每個 keyword token 對應回相對應的 index
List <List <IndexElement> > checkedIndex = new List <List <IndexElement> >();
foreach (var aToken in keywordTokens)
{
checkedIndex.Add(indexes.Where(t => t.Word == aToken.First).First().Indexes);
}
// 檢查各文件是否為符合的文件
var firstTokenIndex = checkedIndex[0];
Dictionary <Int32, Int32> hittedDocs = new Dictionary <Int32, Int32>();
foreach (var currentIndex in firstTokenIndex)
{
if (keywordTokens.Count == 1 || CheckDocumentIsHitted(keywordTokens, 1, checkedIndex, currentIndex))
{
if (hittedDocs.ContainsKey(currentIndex.DocId))
{
hittedDocs[currentIndex.DocId]++;
}
else
{
hittedDocs[currentIndex.DocId] = 1;
}
}
}
// 文件照分數排序，取出指定區間的 doc id 列表
var sortedDocIds = (from entry in hittedDocs orderby entry.Value descending select entry.Key).Skip(startPos).Take(fetchSize).ToList();
// 結果儲存
result.Matches = hittedDocs.Count;
sw.Stop();
result.SearchTime = sw.ElapsedMilliseconds / 1000.0;
for (int i = 0; i < fetchSize && i < sortedDocIds.Count; i++)
{
String rawText = this.GetDocumentById(sortedDocIds[i]);
result.Results.Add(new ResultItem()
{
Rank = startPos + 1 + i,
Score = hittedDocs[sortedDocIds[i]],
HitField = rawText.Replace(keyword, "<<" + keyword + ">>")
});
}
return(result);
}
catch (Exception e)
{
return(null);
}
}