Lucene4.3进阶开发之潇湘夜雨（十七）

Lucene4.3进阶开发之潇湘夜雨（十七）
当所有的文本被语汇成单元后，lucene将会通过一种链式方式，在后台层层把这些信息添加到lucene的各个索引文件里。默认的索引链在DocumentsWriterPerThread中有初始化 

1. static final IndexingChain defaultIndexingChain = new IndexingChain() {  
2.   
3.     @Override  
4.     DocConsumer getChain(DocumentsWriterPerThread documentsWriterPerThread) {  
5.       /* 
6.       This is the current indexing chain: 
7.  
8.       DocConsumer / DocConsumerPerThread 
9.         --> code: DocFieldProcessor 
10.           --> DocFieldConsumer / DocFieldConsumerPerField 
11.             --> code: DocFieldConsumers / DocFieldConsumersPerField 
12.               --> code: DocInverter / DocInverterPerField 
13.                 --> InvertedDocConsumer / InvertedDocConsumerPerField 
14.                   --> code: TermsHash / TermsHashPerField 
15.                     --> TermsHashConsumer / TermsHashConsumerPerField 
16.                       --> code: FreqProxTermsWriter / FreqProxTermsWriterPerField 
17.                       --> code: TermVectorsTermsWriter / TermVectorsTermsWriterPerField 
18.                 --> InvertedDocEndConsumer / InvertedDocConsumerPerField 
19.                   --> code: NormsConsumer / NormsConsumerPerField 
20.           --> StoredFieldsConsumer 
21.             --> TwoStoredFieldConsumers 
22.               -> code: StoredFieldsProcessor 
23.               -> code: DocValuesProcessor 
24.     */  
25.   
26.     // Build up indexing chain:  
27.   
28.       final TermsHashConsumer termVectorsWriter = new TermVectorsConsumer(documentsWriterPerThread);  
29.       final TermsHashConsumer freqProxWriter = new FreqProxTermsWriter();  
30.   
31.       final InvertedDocConsumer termsHash = new TermsHash(documentsWriterPerThread, freqProxWriter, true,  
32.                                                           new TermsHash(documentsWriterPerThread, termVectorsWriter, false, null));  
33.       final NormsConsumer normsWriter = new NormsConsumer();  
34.       final DocInverter docInverter = new DocInverter(documentsWriterPerThread.docState, termsHash, normsWriter);  
35.       final StoredFieldsConsumer storedFields = new TwoStoredFieldsConsumers(  
36.                                                       new StoredFieldsProcessor(documentsWriterPerThread),  
37.                                                       new DocValuesProcessor(documentsWriterPerThread.bytesUsed));  
38.       return new DocFieldProcessor(documentsWriterPerThread, docInverter, storedFields);  
39.     }  
40.   };

经过一系列链式传递后，索引并不会马上写入磁盘中，而是先写入内存中，然后到一定时机会flush到磁盘上，这实际上也是一种优化策略，从而避免了频繁访问磁盘带来的IO开销，当然我们也可以显式得调用commit方法，将其刷新到磁盘上。 
Please read full article from Lucene4.3进阶开发之潇湘夜雨（十七）