搜索引擎之全文搜索算法功能實現（基於Lucene）

時間 2019-12-13

原文原文鏈接

以前作去轉盤網的時候，我已經公開了非全文搜索的代碼，須要的朋友但願可以前去閱讀個人博客。本文主要討論如何進行全文搜索，因爲本人花了很長時間設計了新做：觀點，觀點對全文搜索的要求仍是很高的，因此我又花了很多時間研究全文搜索，你能夠先體驗下：點我搜索。廢話也很少說了，直接上代碼：html

public Map<String,Object>  articleSearchAlgorithms(SearchCondition condition,IndexSearcher searcher) throws ParseException, IOException{
         
            Map<String,Object> map =new HashMap<String,Object>();
             String[] filedsList=condition.getFiledsList();
             String keyWord=condition.getKeyWord();
             int currentPage=condition.getCurrentPage();
             int pageSize=condition.getPageSize();
             String sortField=condition.getSortField();
             boolean isASC=condition.isDESC();
             String sDate=condition.getsDate();
            String eDate=condition.geteDate();
            String classify=condition.getClassify();
             
            
            //過濾終結字符
            keyWord=escapeExprSpecialWord(keyWord);
            
            BooleanQuery q1 = new BooleanQuery();
            BooleanQuery q2 = new BooleanQuery();
             BooleanQuery booleanQuery = new BooleanQuery(); //boolean查詢
             
             if(classify!=null&&(classify.equals("guanzhi")||classify.equals("opinion")||classify.equals("write"))){
                 String typeId="1";//默認言論
                 if(classify.equals("guanzhi")){
                     typeId="2";
                 }
                 if(classify.equals("opinion")){
                     typeId="3";
                 }
                 Query termQuery = new TermQuery(new Term("typeId",typeId)); 
                 q1.add(termQuery,BooleanClause.Occur.MUST);
             }

             if(sDate!=null&&eDate!=null){//是否範圍查詢由這兩個參數決定
                Query rangeQuery = new TermRangeQuery("writingTime", new BytesRef(sDate), new BytesRef(eDate),true, true);
                q1.add(rangeQuery,BooleanClause.Occur.MUST);
             }

            Sort sort = new Sort(); // 排序
            sort.setSort(SortField.FIELD_SCORE);
            if(sortField!=null){
                sort.setSort(new SortField(sortField, SortField.Type.STRING, isASC));
            }
            
            int start = (currentPage - 1) * pageSize;
            int hm = start + pageSize;
            
            TopFieldCollector res = TopFieldCollector.create(sort,hm,false, false, false, false);

            //徹底匹配查詢
            Term t0=new Term(filedsList[1],keyWord);
            TermQuery termQuery = new TermQuery(t0);//兩種高度匹配的查詢
            q2.add(termQuery,BooleanClause.Occur.SHOULD);
            
            //前綴匹配
            Term t1=new Term(filedsList[1],keyWord);
            PrefixQuery prefixQuery=new PrefixQuery(t1);
            q2.add(prefixQuery,BooleanClause.Occur.SHOULD);
            
            //短語，類似度匹配，適用於分詞的內容
            for(int i=0;i<filedsList.length;i++){ //多字段term查詢算法
                if(i!=1){
                    PhraseQuery phraseQuery=new PhraseQuery();
                    Term ts0=new Term(filedsList[i],keyWord);
                    phraseQuery.add(ts0);
                    
                    FuzzyQuery fQuery=new FuzzyQuery(new Term(filedsList[i],keyWord),2);//最後類似度查詢
                    
                    q2.add(phraseQuery,BooleanClause.Occur.SHOULD);
                    q2.add(fQuery,BooleanClause.Occur.SHOULD);//後綴類似的拿出來
                }
            }

            MultiFieldQueryParser  queryParser = new MultiFieldQueryParser(Version.LUCENE_47,filedsList,analyzer);
            queryParser.setDefaultOperator(QueryParser.AND_OPERATOR);
            Query query = queryParser.parse(keyWord);

            q2.add(query,BooleanClause.Occur.SHOULD);
            
            //必須加邏輯判斷，不然結果是不一樣的
            if(q1!=null && q1.toString().length()>0){
                booleanQuery.add(q1,BooleanClause.Occur.MUST);
            }
            if(q2!=null && q2.toString().length()>0){
                 booleanQuery.add(q2,BooleanClause.Occur.MUST);
            }
            
            searcher.search(booleanQuery, res);
            long amount = res.getTotalHits(); 
            TopDocs tds = res.topDocs(start, pageSize);
            map.put("amount",amount);
            map.put("tds",tds);
            map.put("query",booleanQuery);
            return map;
    }

注意下：上面代碼的搜索條件（SearchCondition ）是觀點網的具體需求，您能夠按照您本身的搜索條件作改動，這裏也很難適配全部讀者。算法

public Map<String, Object> searchArticle(SearchCondition condition) throws Exception{
            
        Map<String,Object> map =new HashMap<String,Object>();
        List<Write> list=new ArrayList<Write>();
        
         DirectoryReader reader=condition.getReader();
         String URL=condition.getURL();
         boolean isHighligth=condition.isHighlight();
         String keyWord=condition.getKeyWord();
         IndexSearcher searcher=getSearcher(reader,URL);
        
        try{
            Map<String,Object> output=articleSearchAlgorithms(condition,searcher);
            if(output==null){
                map.put("amount",0L);
                map.put("source",null);
                return map;
            }
            
            map.put("amount", output.get("amount"));
            TopDocs tds = (TopDocs) output.get("tds");
            ScoreDoc[] sd = tds.scoreDocs;
            Query query =(Query) output.get("query");
            
            for (int i = 0; i < sd.length; i++) {
                
                Document doc = searcher.doc(sd[i].doc);

                String id = doc.get("id");
                /**********************start*************************須要處理的放一起********************/
                String temp=doc.get("title");
                String title =temp; //默認不高亮
                if(isHighligth){
                    //高亮文章標題
                    Highlighter highlighterTitle = new Highlighter(simpleHTMLFormatter, new QueryScorer(query));
                    highlighterTitle.setTextFragmenter(new SimpleFragmenter(40)); // 字長度
                    TokenStream ts = analyzer.tokenStream("title", new StringReader(temp));
                    title= highlighterTitle.getBestFragment(ts,temp); 
                    if(title==null){
                        title=temp.replace(keyWord,"<span style='color:red'>"+keyWord+"</span>");//高亮處理插件bug，加這句話避免
                    }
                }
                
                String temp1=HtmlEnDecode.htmlEncode(doc.get("content"));
                String content=temp1;//使用本身封裝的方法來轉義
                
                if(isHighligth){
                    //作高亮處理,content
                    Highlighter highlighterContent = new Highlighter(simpleHTMLFormatter, new QueryScorer(query));
                    highlighterContent.setTextFragmenter(new SimpleFragmenter(Constant.HIGHLIGHT_CONTENT_LENGTH)); // 字長度
                    //temp1=StringEscapeUtils.escapeHtml(temp1);//將漢字轉義致使高亮失效
                    TokenStream ts1 = analyzer.tokenStream("content", new StringReader(temp1));
                    content = highlighterContent.getBestFragment(ts1,temp1);
                    
                    if(content==null){
                        content=temp1.replace(keyWord,"<span style='color:red'>"+keyWord+"</span>");//高亮處理插件bug，加這句話避免
                        
                        //假設趕上這種狀況作處理，其餘的高亮器會自動截圖
                        content=subContent(content);//截取處理
                        content=HtmlEnDecode.htmldecode(content);//html解碼
                        content=SubStringHTML.sub(content,Constant.HIGHLIGHT_CONTENT_LENGTH);
                    }
                }
                /*---------------------------------------不斷變更的數據放一起----------------------------*/
                
                Write write=writeDao.getArticle(Long.parseLong(id));
                if(write!=null){
                    write.setTitle(title);
                    write.setContent(content);
                    
                    Date writingTime=write.getWritingTime();
                    String timeGap=DateUtil.dateGap(writingTime);//timeGap
                    write.setTimeGap(timeGap);
                    
                    list.add(write);
                }
            }
            
        }catch(Exception e){
            e.printStackTrace();
        }
        map.put("source",list);
        return map;
    }

注意上面，這是具體的搜索代碼，不一樣的應用場景有不一樣的需求，請您按照本身的需求封裝對象，查詢數據庫等，代碼毫無保留，絕對可用。數據庫

若是有什麼疑問能夠加qq羣：284205104 若是羣滿了就麻煩去趟去轉盤找下最新的羣加了便可，謝謝您的閱讀。spa