FW: Thans for your help with my Lucene problem

[Uwe Schindler]

Forwarded personal mail:

Hi Uwe,

Many thanks for your excellent help with my problem in Lucene (below), 
I really appreciate such a prompt and professional support 
that you guys provide.

I'm sending this email directly to you because I don't know how to reply 
to your reply on the Lucene Mailing List Archives page.

Greetings, best wishes and God's blessings to everyone on the Lucene team!

Slav

---------------------------------------------------------------

Hi Slavomir,

 

The problem in your code is the HitCollector, it does not respect the
docBase or IndexReader passed in by setNextReader:

 

Your index contains more than one segment, so the docId passed to collect is
only relative to the current segment not to the global searcher.
IndexSearcher runs the query not on the top-level IndexReader, it runs the
search on each index segment separately (as your index contains of several
sub-indexes). When you call optimize, your index will only contain one
segment at the end, so this fixes your bug. Also when you not commit before,
this will also happen for this small index size you have (merging does not
occur at all, so after closing you have only one sub-index). With larger
indexes, more than one segment will created in all cases after some time,
with commit you just enforce that.

 

You can fix this in two ways:

 

1. The naïve (looks ugly) approach (not recommended). Please note, this one
uses IndexSearcher to retrieve the document, which is stupid for collectors:

 

private class HitCollector extends Collector {
  private final IndexSearcher searcher;

  private int docBase = 0;
 
  public HitCollector(IndexSearcher searcher) {
   this.searcher = searcher;
  }
 
  public boolean acceptsDocsOutOfOrder() {
   return true;
  }

 

  public void collect(int docIndex) throws IOException {
   Document document = searcher.doc(docBase + docIndex);
   System.out.println(format(new
Date(Long.parseLong(document.get("startdatetime")))));
  }

 

  public void setNextReader(IndexReader reader, int docBase) throws
IOException {

    this.docBase = docBase;

  }
  

  public void setScorer(org.apache.lucene.search.Scorer scorer) throws
IOException {}
}



 

2. In my opinion the better approach: Remove IndexSearcher completely from
the collector, you don’t need it:

 

private class HitCollector extends Collector {
  private IndexReader currentReader = null;
 
  public HitCollector() {}
 
  public boolean acceptsDocsOutOfOrder() {
   return true;
  }

 

  public void collect(int docIndex) throws IOException {
   Document document = currentReader.doc(docIndex);
   System.out.println(format(new
Date(Long.parseLong(document.get("startdatetime")))));
  }

 

  public void setNextReader(IndexReader reader, int docBase) throws
IOException {

   this.currentReader = reader;

  }


  public void setScorer(org.apache.lucene.search.Scorer scorer) throws
IOException {}
}