[ http://issues.apache.org/jira/browse/LUCENE-739?page=comments#action_12456505 ] Yonik Seeley commented on LUCENE-739: -------------------------------------
+1, looks great Michael! > Performance improvement for SegmentMerger.mergeNorms() > ------------------------------------------------------ > > Key: LUCENE-739 > URL: http://issues.apache.org/jira/browse/LUCENE-739 > Project: Lucene - Java > Issue Type: Improvement > Components: Index > Reporter: Michael Busch > Assigned To: Michael Busch > Priority: Minor > Attachments: mergeNorms.patch > > > This patch makes two improvements to SegmentMerger.mergeNorms(): > 1) When the SegmentMerger merges the norm values it allocates a new byte > array to buffer the values for every field of every segment. The size of such > an array equals the size of the corresponding segment, so if large segments > are being merged, those arrays become very large, too. > We can easily reduce the number of array allocations by reusing a byte array > to buffer the norm values that only grows, if a segment is larger than the > previous one. > 2) Before a norm value is written it is checked if the corresponding document > is deleted. If not, the norm is written using IndexOutput.writeByte(byte[]). > This patch introduces an optimized case for segments that do not have any > deleted docs. In this case the frequent call of IndexReader.isDeleted(int) > can be avoided and the more efficient method IndexOutput.writeBytes(byte[], > int) can be used. > This patch only changes the method SegmentMerger.mergeNorms(). All unit tests > pass. -- This message is automatically generated by JIRA. - If you think it was sent incorrectly contact one of the administrators: http://issues.apache.org/jira/secure/Administrators.jspa - For more information on JIRA, see: http://www.atlassian.com/software/jira --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
