senseidb / zoie Goto Github PK

hi ~ 各位,
最近在看zoie的一些源码的时候, 有个关于添加document到内存索引的时候, 构造UIDFilter并显示调用writer.deleteDocuments(new ConstantScoreQuery(uidFilter)); 这个api有些不太明白, 为啥不是直接调用lucene 的 update 的api, 而是先delete---commit ----add---commit呢 ？求大神指点 .... Solr 是先查出业务id对应的docId, 构造updateDocument(Term term, Iterable<? extends IndexableField> doc)中的term实现对业务唯一性的保证. 请问Zoie 使用显示删除并commit, 然后在调用add api这样的做法是否是和自己扩展实现了ZoieSegmentReader或者ZoieMultiReader有关呢 ?

i found that zoie disk index did not merge index unless use OptimizeType.FULL when DiskLuceneIndexDataLoader loadFromIndex from ramindex. but after use OptimizeType.FULL , swap SearchIndexManager staus from working to sleep, the lockAndSwapMem method did not change the _mem to new mem , because the oldDiskReader == diskIndexReader
.

Hi!

I notice you write a return statement in some test cases.

Although you return nothing, I would like to know why you write this statement since as far as I know, it is not a good practice to write a return statement in test cases.Thanks.

in Lucene 3.4 writer.addIndexes(new Directory[] { dir }) does not merge any more, so we must inoke maybeMerge after add dir to index. :)

public void loadFromIndex(BaseSearchIndex<R> index) throws IOException
{
    // hao: open readOnly ram index reader
    ZoieIndexReader<R> reader = index.openIndexReader();
    if (reader == null)
        return;

    Directory dir = reader.directory();

    // hao: delete docs in disk index
    LongSet delDocs = _delDocs;
    clearDeletes();
    deleteDocs(delDocs);

    // hao: merge the readOnly ram index with the disk index
    IndexWriter writer = null;
    try
    {
        writer = openIndexWriter(null, null);
        writer.addIndexes(new Directory[] { dir });
        writer.maybeMerge();
    } finally
    {
        closeIndexWriter();
    }
} 

Hi,
I wonder that what happen if if jvm or OS crash? The two Ram index will lost? That 's unacceptable.

Thanks & Best Regards!

//L43
_readerList = _zoieSystem.getIndexReaders();
if (readerList!=null){
_zoieSystem.returnIndexReaders(_readerList);
}

-->

_readerList = _zoieSystem.getIndexReaders();
if (readerList!=null){
_zoieSystem.returnIndexReaders(readerList);
}