public void InitReader(ByteSliceReader reader, RawPostingList p, int stream)
{
System.Diagnostics.Debug.Assert(stream < streamCount);
int[] ints = intPool.buffers[p.intStart >> DocumentsWriter.INT_BLOCK_SHIFT];
int upto = p.intStart & DocumentsWriter.INT_BLOCK_MASK;
reader.Init(bytePool, p.byteStart + stream * ByteBlockPool.FIRST_LEVEL_SIZE, ints[upto + stream]);
}
/* Walk through all unique text tokens (Posting
* instances) found in this field and serialize them
* into a single RAM segment. */
internal void AppendPostings(FreqProxTermsWriterPerField[] fields, FormatPostingsFieldsConsumer consumer)
{
int numFields = fields.Length;
FreqProxFieldMergeState[] mergeStates = new FreqProxFieldMergeState[numFields];
for (int i = 0; i < numFields; i++)
{
FreqProxFieldMergeState fms = mergeStates[i] = new FreqProxFieldMergeState(fields[i]);
System.Diagnostics.Debug.Assert(fms.field.fieldInfo == fields [0].fieldInfo);
// Should always be true
bool result = fms.NextTerm();
System.Diagnostics.Debug.Assert(result);
}
FormatPostingsTermsConsumer termsConsumer = consumer.AddField(fields[0].fieldInfo);
FreqProxFieldMergeState[] termStates = new FreqProxFieldMergeState[numFields];
bool currentFieldOmitTermFreqAndPositions = fields[0].fieldInfo.omitTermFreqAndPositions;
while (numFields > 0)
{
// Get the next term to merge
termStates[0] = mergeStates[0];
int numToMerge = 1;
for (int i = 1; i < numFields; i++)
{
char[] text = mergeStates[i].text;
int textOffset = mergeStates[i].textOffset;
int cmp = compareText(text, textOffset, termStates[0].text, termStates[0].textOffset);
if (cmp < 0)
{
termStates[0] = mergeStates[i];
numToMerge = 1;
}
else if (cmp == 0)
{
termStates[numToMerge++] = mergeStates[i];
}
}
FormatPostingsDocsConsumer docConsumer = termsConsumer.AddTerm(termStates[0].text, termStates[0].textOffset);
// Now termStates has numToMerge FieldMergeStates
// which all share the same term. Now we must
// interleave the docID streams.
while (numToMerge > 0)
{
FreqProxFieldMergeState minState = termStates[0];
for (int i = 1; i < numToMerge; i++)
{
if (termStates[i].docID < minState.docID)
{
minState = termStates[i];
}
}
int termDocFreq = minState.termFreq;
FormatPostingsPositionsConsumer posConsumer = docConsumer.AddDoc(minState.docID, termDocFreq);
ByteSliceReader prox = minState.prox;
// Carefully copy over the prox + payload info,
// changing the format to match Lucene's segment
// format.
if (!currentFieldOmitTermFreqAndPositions)
{
// omitTermFreqAndPositions == false so we do write positions &
// payload
int position = 0;
for (int j = 0; j < termDocFreq; j++)
{
int code = prox.ReadVInt();
position += (code >> 1);
int payloadLength;
if ((code & 1) != 0)
{
// This position has a payload
payloadLength = prox.ReadVInt();
if (payloadBuffer == null || payloadBuffer.Length < payloadLength)
{
payloadBuffer = new byte[payloadLength];
}
prox.ReadBytes(payloadBuffer, 0, payloadLength);
}
else
{
payloadLength = 0;
}
posConsumer.AddPosition(position, payloadBuffer, 0, payloadLength);
} //End for
posConsumer.Finish();
}
if (!minState.NextDoc())
{
// Remove from termStates
int upto = 0;
for (int i = 0; i < numToMerge; i++)
{
if (termStates[i] != minState)
{
termStates[upto++] = termStates[i];
}
}
numToMerge--;
System.Diagnostics.Debug.Assert(upto == numToMerge);
// Advance this state to the next term
if (!minState.NextTerm())
{
// OK, no more terms, so remove from mergeStates
// as well
upto = 0;
for (int i = 0; i < numFields; i++)
{
if (mergeStates[i] != minState)
{
mergeStates[upto++] = mergeStates[i];
}
}
numFields--;
System.Diagnostics.Debug.Assert(upto == numFields);
}
}
}
docConsumer.Finish();
}
termsConsumer.Finish();
}
public void InitReader(ByteSliceReader reader, RawPostingList p, int stream)
{
System.Diagnostics.Debug.Assert(stream < streamCount);
int[] ints = intPool.buffers[p.intStart >> DocumentsWriter.INT_BLOCK_SHIFT];
int upto = p.intStart & DocumentsWriter.INT_BLOCK_MASK;
reader.Init(bytePool, p.byteStart + stream * ByteBlockPool.FIRST_LEVEL_SIZE, ints[upto + stream]);
}
/// <summary>Called once per field per document if term vectors
/// are enabled, to write the vectors to
/// RAMOutputStream, which is then quickly flushed to
/// the real term vectors files in the Directory.
/// </summary>
internal override void Finish()
{
System.Diagnostics.Debug.Assert(docState.TestPoint("TermVectorsTermsWriterPerField.finish start"));
int numPostings = termsHashPerField.numPostings;
System.Diagnostics.Debug.Assert(numPostings >= 0);
if (!doVectors || numPostings == 0)
{
return;
}
if (numPostings > maxNumPostings)
{
maxNumPostings = numPostings;
}
IndexOutput tvf = perThread.doc.perDocTvf;
// This is called once, after inverting all occurences
// of a given field in the doc. At this point we flush
// our hash into the DocWriter.
System.Diagnostics.Debug.Assert(fieldInfo.storeTermVector);
System.Diagnostics.Debug.Assert(perThread.VectorFieldsInOrder(fieldInfo));
perThread.doc.AddField(termsHashPerField.fieldInfo.number);
RawPostingList[] postings = termsHashPerField.SortPostings();
tvf.WriteVInt(numPostings);
byte bits = (byte)(0x0);
if (doVectorPositions)
{
bits |= TermVectorsReader.STORE_POSITIONS_WITH_TERMVECTOR;
}
if (doVectorOffsets)
{
bits |= TermVectorsReader.STORE_OFFSET_WITH_TERMVECTOR;
}
tvf.WriteByte(bits);
int encoderUpto = 0;
int lastTermBytesCount = 0;
ByteSliceReader reader = perThread.vectorSliceReader;
char[][] charBuffers = perThread.termsHashPerThread.charPool.buffers;
for (int j = 0; j < numPostings; j++)
{
TermVectorsTermsWriter.PostingList posting = (TermVectorsTermsWriter.PostingList)postings[j];
int freq = posting.freq;
char[] text2 = charBuffers[posting.textStart >> DocumentsWriter.CHAR_BLOCK_SHIFT];
int start2 = posting.textStart & DocumentsWriter.CHAR_BLOCK_MASK;
// We swap between two encoders to save copying
// last Term's byte array
UnicodeUtil.UTF8Result utf8Result = perThread.utf8Results[encoderUpto];
// TODO: we could do this incrementally
UnicodeUtil.UTF16toUTF8(text2, start2, utf8Result);
int termBytesCount = utf8Result.length;
// TODO: UTF16toUTF8 could tell us this prefix
// Compute common prefix between last term and
// this term
int prefix = 0;
if (j > 0)
{
byte[] lastTermBytes = perThread.utf8Results[1 - encoderUpto].result;
byte[] termBytes = perThread.utf8Results[encoderUpto].result;
while (prefix < lastTermBytesCount && prefix < termBytesCount)
{
if (lastTermBytes[prefix] != termBytes[prefix])
{
break;
}
prefix++;
}
}
encoderUpto = 1 - encoderUpto;
lastTermBytesCount = termBytesCount;
int suffix = termBytesCount - prefix;
tvf.WriteVInt(prefix);
tvf.WriteVInt(suffix);
tvf.WriteBytes(utf8Result.result, prefix, suffix);
tvf.WriteVInt(freq);
if (doVectorPositions)
{
termsHashPerField.InitReader(reader, posting, 0);
reader.WriteTo(tvf);
}
if (doVectorOffsets)
{
termsHashPerField.InitReader(reader, posting, 1);
reader.WriteTo(tvf);
}
}
termsHashPerField.Reset();
// NOTE: we clear, per-field, at the thread level,
// because term vectors fully write themselves on each
// field; this saves RAM (eg if large doc has two large
// fields w/ term vectors on) because we recycle/reuse
// all RAM after each field:
perThread.termsHashPerThread.Reset(false);
}
