the class org.apache.lucene.search.spell.SpellChecker uses the following hard-coded values in its method
indexDictionary:
writer.setMergeFactor(300);
writer.setMaxBufferedDocs(150);
this poses problems when the spellcheck index is created on systems with certain limits, i.e. in unix
environments where the ulimit settings are restricted for the user (http://www.gossamer-threads.com/lists/lucene/java-dev/47428#47428).
there are several ways to circumvent this:
1. add another indexDictionary method with additional parameters:
public void indexDictionary (Dictionary dict, int mergeFactor, int maxBufferedDocs) throws IOException
the class org.apache.lucene.search.spell.SpellChecker uses the following hard-coded values in its method
indexDictionary:
writer.setMergeFactor(300);
writer.setMaxBufferedDocs(150);
this poses problems when the spellcheck index is created on systems with certain limits, i.e. in unix
environments where the ulimit settings are restricted for the user (http://www.gossamer-threads.com/lists/lucene/java-dev/47428#47428).
there are several ways to circumvent this:
1. add another indexDictionary method with additional parameters:
public void indexDictionary (Dictionary dict, int mergeFactor, int maxBufferedDocs) throws IOException
2. add setter methods for mergeFactor and maxBufferedDocs
(see code in http://www.gossamer-threads.com/lists/lucene/java-dev/47428#47428 )
3. Make SpellChecker subclassing easier as suggested by Chris Hostetter
(see reply http://www.gossamer-threads.com/lists/lucene/java-dev/47463#47463)
thanx,
karin
Improvement
Resolved
Minor
spellchecker: make hard-coded values configurable
