对于满足全文检索的需求来说,Lucene.Net无疑是一个很好的选择。它引入了增量索引的策略,解决了在数据频繁改动时重建索引的问题,这对于提高web的性能至关重要(其他相关特性大家可以参看官方文档)。Lucene.Net是基于文档性的全文搜索,所以使用Lucene.Net时要把数据库中的数据先导出来,这也是一个建立索引的过程。代码如下:
1 /// <summary> 2 /// Add Data into Indexes 3 /// </summary> 4 /// <param name="models">Data collection</param> 5 /// <param name="optimize">Whether to optimize the indexes after adding new indexes</param> 6 public void AddToSearchIndex(IEnumerable<T> models, bool optimize = false) 7 { 8 var analyzer = new StandardAnalyzer(Version.LUCENE_30); 9 using (var writer = new IndexWriter(_directory,analyzer,IndexWriter.MaxFieldLength.UNLIMITED)) 10 { 11 foreach (var model in models) 12 { 13 //remove older index entry 14 var searchQuery = new TermQuery(new Term("Id", (model as dynamic).ID.ToString())); 16 writer.DeleteDocuments(searchQuery); 17 18 var doc = new Document(); 19 foreach (var prop in Props) 20 { 21 var value = prop.GetValue(model); 22 if (value == null) 23 { 24 continue; 25 }
26 //only store ID,we use it to retrieve model data from DB 27 doc.Add(new Field(prop.Name, value.ToString(), 28 prop.Name == "ID" ? Field.Store.YES : Field.Store.NO, 29 Field.Index.ANALYZED)); 30 } 31 writer.AddDocument(doc); 32 } 33 if (optimize) 34 { 35 writer.Optimize(); 36 } 37 } 38 }
上述函数用于把到处的数据添加到索引文件中,我们可以指定是否在完成插入后优化索引。优化索引可以提高检索速度,但会消耗Cpu资源,不建议经常优化它。另外,我们在插入索引时会先检测时更新还是添加,这用于完成对旧数据的更新。那么,如果当数据库移除了一条记录,对于索引文件我们又该如何做呢?
和数据库操作类似,当从数据库移除记录时,从所以文件中移除相应记录即可,代码如下:
/// <summary> /// Remove specfied index record /// </summary> /// <param name="record_id">the record's ID</param> public void ClearSearchIndex(int record_id) { var analyzer = new StandardAnalyzer(Version.LUCENE_30); using (var writer = new IndexWriter(_directory, analyzer, IndexWriter.MaxFieldLength.UNLIMITED)) { // remove older index entry var searchQuery = new TermQuery(new Term("ID", record_id.ToString())); writer.DeleteDocuments(searchQuery); writer.Commit(); } analyzer.Dispose(); }
同样,我们可以删除所有的索引记录
/// <summary> /// Remove all index records /// </summary> /// <returns>whether operation success or not</returns> public bool ClearAllSearchIndex() { StandardAnalyzer analyzer = null; try { analyzer = new StandardAnalyzer(Version.LUCENE_30); using (var writer = new IndexWriter(_directory, analyzer, true,
IndexWriter.MaxFieldLength.UNLIMITED)) { //remove older index entries writer.DeleteAll(); writer.Commit(); } analyzer.Dispose(); } catch (Exception) { analyzer.Dispose(); return false; } return true; }
下面该主角登场了,看看如何检索记录吧:
/// <summary> /// Searching specfied value in all fields,or you can specfied a field to search in. /// </summary> /// <param name="querystring">value to search</param> /// <param name="fieldname">field to search, search all fieds at default</param> /// <returns>realted records' ID sequence</returns> public IEnumerable<int> Search(string querystring, string fieldname = "") { IEnumerable<int> result = new List<int>(); if (string.IsNullOrEmpty(querystring)) { return new List<int>(); } //remove invalid characters querystring = ParseSearchString(querystring); // validation if (string.IsNullOrEmpty(querystring.Replace("*", "").Replace("?", ""))) { return new List<int>(); } using (var searcher = new IndexSearcher(_directory, true)) { ScoreDoc[] hits = null; //the max hited racord count var hits_limit = 1000; var analyzer = new StandardAnalyzer(Version.LUCENE_30); //used to separate the querystring to match records in indexes QueryParser parser = null; Query query = null; if (!string.IsNullOrEmpty(fieldname)) { //create a QueryParser instance in the specified field parser = new QueryParser(Version.LUCENE_30, fieldname, analyzer); } else { string[] fields = Props.Select(p => p.Name).ToArray<string>(); //create a QueryParser instance in the all fields parser = new MultiFieldQueryParser(Version.LUCENE_30, fields, analyzer); } //create a query instance from QueryParser and querystring query = ParseQuery(querystring, parser); //get the hited record hits = searcher.Search(query, hits_limit).ScoreDocs; var resultDocs = hits.Select(hit => searcher.Doc(hit.Doc)); //transmit the index record's ID to the DB record's ID result = resultDocs.
Select(doc => ((SpecEquipmentID)int.Parse(doc.Get("ID"))).CurrentID).
ToList(); analyzer.Dispose(); } return result; }
从上述可以看出,我们可以指定在若干字段间搜索,这些字段间的检索同样可采用模糊检索的模式:
public IEnumerable<int> MultiFieldsSearch(Dictionary<string, string> multiFieldsDict) { IEnumerable<int> result = new List<int>(); if (multiFieldsDict.Count == 0) { return result; } using (var searcher = new IndexSearcher(_directory, true)) { ScoreDoc[] hits = null; var hits_limit = 1000; var analyzer = new StandardAnalyzer(Version.LUCENE_30); var occurs = (from field in multiFieldsDict.Keys select Occur.MUST).ToArray(); var queries = (from key in multiFieldsDict.Keys select multiFieldsDict[key]).ToArray(); Query query = MultiFieldQueryParser.Parse(Version.LUCENE_30, queries,
multiFieldsDict.Keys.ToArray(), occurs, analyzer); hits = searcher.Search(query, hits_limit).ScoreDocs; var resultDocs = hits.Select(hit => searcher.Doc(hit.Doc)); result = resultDocs.
Select(doc => ((SpecEquipmentID)int.Parse(doc.Get("ID"))).CurrentID).
Distinct().ToList(); analyzer.Dispose(); } return result; }
在这里解释下:为什么用QueryParser生成Query的实例?
使用QueryParser可以让我们在指定的字段间使用模糊查询,也就是说,只要相应的记录之中包含检索值,都会被命中,这也正是全文搜索所必需的。如果不采用以上方式,可以使用BooleanQuery结合TermQuery在指定字段间搜索,但这样以来,只有同值记录(精确查询)会被命中。这些搜索条件间同样可以像数据库查询那样采用‘与或非’的形式。
最后说明一下:对于数值类型和日期类型的处理比较特殊,如果采用像字符串那样的处理方式,结果的精确性就会下降,至于如何处理针对数值类型和日期类型的数据检索,大家可以参考Lucene的官方文档。提及一下我的解决方案:我们可以采用常规数据库与Lucene结合的方式,让Lucene处理字符串类型的检索,常规数据库处理日期及数值类型的检索,各抒其长。