lucene-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From robert engels <reng...@ix.netcom.com>
Subject Re: [jira] Commented: (LUCENE-1043) Speedup merging of stored fields when field mapping "matches"
Date Fri, 02 Nov 2007 21:49:12 GMT
Here is the fixed code - need special handling when determining the  
length of the document in a segment.

FieldsReader.java

   final int docs(byte[] buffer,int[] lengths,int start,int max)  
throws IOException {
       final IndexInput indexStream = (IndexInput) indexLocal.get();
       indexStream.seek(start * 8L);
       long startoffset = indexStream.readLong();
       long lastoffset = startoffset;
       int count=0;
       while(count<max) {
	  long offset = (start + count + 1 < size) ? indexStream.readLong 
() : fieldsStream.length();
	  if(offset-startoffset>buffer.length)
	      break;
	  lengths[count]=(int) (offset-lastoffset);
	  lastoffset = offset;
	  count++;
       }
       if(count==0)
	  throw new IllegalStateException("buffer too small for a single  
document");
	
       final IndexInput fieldsStream = (IndexInput) fieldsLocal.get();
       fieldsStream.seek(startoffset);
       fieldsStream.readBytes(buffer,0,(int) (lastoffset-startoffset));

       return count;
   }


On Nov 2, 2007, at 4:33 PM, robert engels wrote:

> Oops. Be careful with that code, merging works but indexing  
> fails... I have a bug lurking...
>
> On Nov 2, 2007, at 4:28 PM, robert engels wrote:
>
>> As expected it did not make much difference. Here is the code:
>>
>> FieldsReader.java
>>   final int docs(byte[] buffer,int[] lengths,int start,int max)  
>> throws IOException {
>>       final IndexInput indexStream = (IndexInput) indexLocal.get();
>>       indexStream.seek(start * 8L);
>>       long startoffset = indexStream.readLong();
>>       long lastoffset = startoffset;
>>       int count=0;
>>       while(count<max) {
>> 	  long offset = indexStream.readLong();
>> 	  if(offset-startoffset>buffer.length)
>> 	      break;
>> 	  lengths[count]=(int) (offset-lastoffset);
>> 	  lastoffset = offset;
>> 	  count++;
>>       }
>>       if(count==0)
>> 	  throw new IllegalStateException("buffer too small for a single  
>> document");
>> 	
>>       final IndexInput fieldsStream = (IndexInput) fieldsLocal.get();
>>       fieldsStream.seek(startoffset);
>>       fieldsStream.readBytes(buffer,0,(int) (lastoffset- 
>> startoffset));
>>
>>       return count;
>>   }
>>
>> FieldsWriter.java
>>     final void addDocuments(byte[] buffer, int[] lengths, int  
>> ndocs) throws IOException {
>> 	long position = fieldsStream.getFilePointer();
>> 	long start = position;
>> 	for(int i=0;i<ndocs;i++) {
>> 	    indexStream.writeLong(position);
>> 	    position+=lengths[i];
>> 	}
>>         fieldsStream.writeBytes(buffer,(int) (position-start));
>>     }
>>
>> SegmentReader.java
>>   public int documents(byte[] buffer,int[] lengths,int start,int  
>> max) throws IOException {
>>       return fieldsReader.docs(buffer,lengths,start,max);
>>   }
>>
>> SegmentMerger.java
>>   private final int mergeFields() throws IOException {
>> 	fieldInfos = new FieldInfos(); // merge field names
>> 	int docCount = 0;
>> 	for (int i = 0; i < readers.size(); i++) {
>> 	    IndexReader reader = (IndexReader) readers.elementAt(i);
>> 	    if (reader instanceof SegmentReader) {
>> 		SegmentReader sreader = (SegmentReader) reader;
>> 		for (int j = 0; j < sreader.getFieldInfos().size(); j++) {
>> 		    FieldInfo fi = sreader.getFieldInfos().fieldInfo(j);
>> 		    fieldInfos.add(fi.name, fi.isIndexed, fi.storeTermVector,  
>> fi.storePositionWithTermVector, fi.storeOffsetWithTermVector, ! 
>> reader.hasNorms(fi.name));
>> 		}
>> 	    } else {
>> 		addIndexed(reader, fieldInfos, reader.getFieldNames 
>> (IndexReader.FieldOption.TERMVECTOR_WITH_POSITION_OFFSET), true,  
>> true, true);
>> 		addIndexed(reader, fieldInfos, reader.getFieldNames 
>> (IndexReader.FieldOption.TERMVECTOR_WITH_POSITION), true, true,  
>> false);
>> 		addIndexed(reader, fieldInfos, reader.getFieldNames 
>> (IndexReader.FieldOption.TERMVECTOR_WITH_OFFSET), true, false, true);
>> 		addIndexed(reader, fieldInfos, reader.getFieldNames 
>> (IndexReader.FieldOption.TERMVECTOR), true, false, false);
>> 		addIndexed(reader, fieldInfos, reader.getFieldNames 
>> (IndexReader.FieldOption.INDEXED), false, false, false);
>> 		fieldInfos.add(reader.getFieldNames 
>> (IndexReader.FieldOption.UNINDEXED), false);
>> 	    }
>> 	}
>> 	fieldInfos.write(directory, segment + ".fnm");
>>
>> 	SegmentReader[] sreaders = new SegmentReader[readers.size()];
>> 	for (int i = 0; i < readers.size(); i++) {
>> 	    IndexReader reader = (IndexReader) readers.elementAt(i);
>> 	    boolean same = reader.getFieldNames().size() ==  
>> fieldInfos.size() && reader instanceof SegmentReader;
>> 	    if(same) {
>> 		SegmentReader sreader = (SegmentReader) reader;
>> 		for (int j = 0; same && j < fieldInfos.size(); j++) {
>> 		    same = fieldInfos.fieldName(j).equals(sreader.getFieldInfos 
>> ().fieldName(j));
>> 		}
>> 		if(same)
>> 		    sreaders[i] = sreader;
>> 	    }
>> 	}
>> 	
>> 	byte[] buffer = new byte[512*1024];
>> 	int[] lengths = new int[10000];
>>
>> 	// merge field values
>> 	FieldsWriter fieldsWriter = new FieldsWriter(directory, segment,  
>> fieldInfos);
>> 	
>> 	try {
>> 	    for (int i = 0; i < readers.size(); i++) {
>> 		IndexReader reader = (IndexReader) readers.elementAt(i);
>> 		SegmentReader sreader = sreaders[i];
>> 		int maxDoc = reader.maxDoc();
>> 		for (int j = 0; j < maxDoc;)
>> 		    if (!reader.isDeleted(j)) { // skip deleted docs
>> 			if (sreader!=null) {
>> 			    int start=j;
>> 			    int ndocs=1;
>> 			    for(j++;j<maxDoc && !reader.isDeleted(j) && ndocs<10000;j+

>> +,ndocs++);
>> 			    ndocs = sreader.documents(buffer,lengths,start,j);
>> 			    fieldsWriter.addDocuments(buffer,lengths,ndocs);
>> 			} else {
>> 			    fieldsWriter.addDocument(reader.document(j));
>> 			    j++;
>> 			}
>> 			docCount++;
>> 		    }
>> 	    }
>> 	} finally {
>> 	    fieldsWriter.close();
>> 	}
>> 	return docCount;
>>     }
>>
>>
>>
>>
>> On Nov 2, 2007, at 3:49 PM, robert engels wrote:
>>
>>> I am working on a bulk document copy right now - I will let you  
>>> know if it improves things much.
>>>
>>> I doubt it, because I already configure the streams to use fairly  
>>> large input and output buffers during a merge - but the memory  
>>> index merge may see additional benefits due to lower CPU calls.
>>>
>>>
>>> On Nov 2, 2007, at 3:39 PM, robert engels (JIRA) wrote:
>>>
>>>>
>>>>     [ https://issues.apache.org/jira/browse/LUCENE-1043? 
>>>> page=com.atlassian.jira.plugin.system.issuetabpanels:comment- 
>>>> tabpanel#action_12539696 ]
>>>>
>>>> robert engels commented on LUCENE-1043:
>>>> ---------------------------------------
>>>>
>>>> When bulk copying the documents, I think you need to:
>>>>
>>>> read array of long from index (8 * (ndocs+1)) in long[ndocs+1]  
>>>> offsets;
>>>> calculate length = offset[ndocs]-offset[0];
>>>> read bytes of length from document file
>>>> startoffset = current output document stream position
>>>> write bytes to output document
>>>> modify offset[] adding startoffset-offset[0] to each entry
>>>> write offset[] in bulk to index output
>>>>
>>>>> Speedup merging of stored fields when field mapping "matches"
>>>>> -------------------------------------------------------------
>>>>>
>>>>>                 Key: LUCENE-1043
>>>>>                 URL: https://issues.apache.org/jira/browse/ 
>>>>> LUCENE-1043
>>>>>             Project: Lucene - Java
>>>>>          Issue Type: Improvement
>>>>>          Components: Index
>>>>>    Affects Versions: 2.2
>>>>>            Reporter: Michael McCandless
>>>>>            Assignee: Michael McCandless
>>>>>            Priority: Minor
>>>>>             Fix For: 2.3
>>>>>
>>>>>         Attachments: LUCENE-1043.patch
>>>>>
>>>>>
>>>>> Robert Engels suggested the following idea, here:
>>>>>   http://www.gossamer-threads.com/lists/lucene/java-dev/54217
>>>>> When merging in the stored fields from a segment, if the field  
>>>>> name ->
>>>>> number mapping is identical then we can simply bulk copy the  
>>>>> entire
>>>>> entry for the document rather than re-interpreting and then re- 
>>>>> writing
>>>>> the actual stored fields.
>>>>> I've pulled the code from the above thread and got it working  
>>>>> on the
>>>>> current trunk.
>>>>
>>>> -- 
>>>> This message is automatically generated by JIRA.
>>>> -
>>>> You can reply to this email to add a comment to the issue online.
>>>>
>>>>
>>>> ------------------------------------------------------------------- 
>>>> --
>>>> To unsubscribe, e-mail: java-dev-unsubscribe@lucene.apache.org
>>>> For additional commands, e-mail: java-dev-help@lucene.apache.org
>>>>
>>>
>>>
>>> -------------------------------------------------------------------- 
>>> -
>>> To unsubscribe, e-mail: java-dev-unsubscribe@lucene.apache.org
>>> For additional commands, e-mail: java-dev-help@lucene.apache.org
>>>
>>
>


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message