[ lucene扩展 ] 自定义Collector实现统计功能

[ lucene扩展 ] 自定义Collector实现统计功能

在search中我们来看看collector是怎么收集结果的！

public void search(Weight weight, Filter filter, Collector collector)             throws IOException {           // TODO: should we make this         // threaded...? the Collector could be sync'd?           // always use single thread:         for (int i = 0; i < subReaders.length; i++) { // 检索每个子索引             collector.setNextReader(subReaders[i], docBase + docStarts[i]);             final Scorer scorer = (filter == null) ? weight.scorer(                     subReaders[i], !collector.acceptsDocsOutOfOrder(), true)                     : FilteredQuery.getFilteredScorer(subReaders[i],                             getSimilarity(), weight, weight, filter);//构建打分器             if (scorer != null) {                 scorer.score(collector);//打分             }         }     }

scorer.score(collector)的过程如下：

public void score(Collector collector) throws IOException {     collector.setScorer(this);     int doc;     while ((doc = nextDoc()) != NO_MORE_DOCS) {       collector.collect(doc);//搜集结果     }   }

collector.collect(doc)的过程如下：

@Override public void collect(int doc) throws IOException {   float score = scorer.score();     // This collector cannot handle these scores:   assert score != Float.NEGATIVE_INFINITY;   assert !Float.isNaN(score);     totalHits++;   if (score <= pqTop.score) {     // 以下的实现使用了优先级队列，如果当前分值小于队列中pqTop.score则直接pass！     return;   }   pqTop.doc = doc + docBase;   pqTop.score = score;   pqTop = pq.updateTop(); }

public class GroupCollectorDemo extends Collector {

 

    private GF gf = new GF();// 保存分组统计结果

    private String[] fc;// fieldCache

    private String f;// 统计字段

    String spliter;

    int length;

 

    public void setFc(String[] fc) {

        this.fc = fc;

    }

 

    @Override

    public void setScorer(Scorer scorer) throws IOException {

    }

 

    @Override

    public void setNextReader(IndexReader reader, int docBase)

            throws IOException {

        fc = FieldCache.DEFAULT.getStrings(reader, f);

    }

 

    @Override

    public void collect(int doc) throws IOException {

        // 添加的GroupField中，由GroupField负责统计每个不同值的数目

        gf.addValue(fc[doc]);

    }

 

    @Override

    public boolean acceptsDocsOutOfOrder() {

        return true;

    }

 

    public GF getGroupField() {

        return gf;

    }

 

    public void setSpliter(String spliter) {

        this.spliter = spliter;

    }

 

    public void setLength(int length) {

        this.length = length;

    }

 

    public void setF(String f) {

        this.f = f;

    }

}

 

class GF {

    // 所有可能的分组字段值，排序按每个字段值的文档个数大小排序

    private List<String> values = new ArrayList<String>();

    // 保存字段值和文档个数的对应关系

    private Map<String, Integer> countMap = new HashMap<String, Integer>();

 

    public Map<String, Integer> getCountMap() {

        return countMap;

    }

 

    public void setCountMap(Map<String, Integer> countMap) {

        this.countMap = countMap;

    }

 

    public List<String> getValues() {

        Collections.sort(values, new ValueComparator());

        return values;

    }

 

    public void setValues(List<String> values) {

        this.values = values;

    }

 

    public void addValue(String value) {

        if (value == null || "".equals(value))

            return;

        if (countMap.get(value) == null) {

            countMap.put(value, 1);

            values.add(value);

        } else {

            countMap.put(value, countMap.get(value) + 1);

        }

    }

 

    class ValueComparator implements Comparator<String> {

        public int compare(String value0, String value1) {

            if (countMap.get(value0) > countMap.get(value1)) {

                return -1;

            } else if (countMap.get(value0) < countMap.get(value1)) {

                return 1;

            }

            return 0;

        }

    }

}

Please read full article from [ lucene扩展 ] 自定义Collector实现统计功能