/// <summary> Add a complete document specified by all its term vectors. If document has no /// term vectors, add value for tvx. /// /// </summary> /// <param name="vectors"> /// </param> /// <throws> IOException </throws> public void AddAllDocVectors(ITermFreqVector[] vectors) { tvx.WriteLong(tvd.FilePointer); tvx.WriteLong(tvf.FilePointer); if (vectors != null) { int numFields = vectors.Length; tvd.WriteVInt(numFields); var fieldPointers = new long[numFields]; for (int i = 0; i < numFields; i++) { fieldPointers[i] = tvf.FilePointer; int fieldNumber = fieldInfos.FieldNumber(vectors[i].Field); // 1st pass: write field numbers to tvd tvd.WriteVInt(fieldNumber); int numTerms = vectors[i].Size; tvf.WriteVInt(numTerms); TermPositionVector tpVector; byte bits; bool storePositions; bool storeOffsets; if (vectors[i] is TermPositionVector) { // May have positions & offsets tpVector = (TermPositionVector)vectors[i]; storePositions = tpVector.Size > 0 && tpVector.GetTermPositions(0) != null; storeOffsets = tpVector.Size > 0 && tpVector.GetOffsets(0) != null; bits = (byte)((storePositions?TermVectorsReader.STORE_POSITIONS_WITH_TERMVECTOR: (byte)0) + (storeOffsets?TermVectorsReader.STORE_OFFSET_WITH_TERMVECTOR: (byte)0)); } else { tpVector = null; bits = 0; storePositions = false; storeOffsets = false; } tvf.WriteVInt(bits); System.String[] terms = vectors[i].GetTerms(); int[] freqs = vectors[i].GetTermFrequencies(); int utf8Upto = 0; utf8Results[1].length = 0; for (int j = 0; j < numTerms; j++) { UnicodeUtil.UTF16toUTF8(terms[j], 0, terms[j].Length, utf8Results[utf8Upto]); int start = StringHelper.BytesDifference(utf8Results[1 - utf8Upto].result, utf8Results[1 - utf8Upto].length, utf8Results[utf8Upto].result, utf8Results[utf8Upto].length); int length = utf8Results[utf8Upto].length - start; tvf.WriteVInt(start); // write shared prefix length tvf.WriteVInt(length); // write delta length tvf.WriteBytes(utf8Results[utf8Upto].result, start, length); // write delta bytes utf8Upto = 1 - utf8Upto; int termFreq = freqs[j]; tvf.WriteVInt(termFreq); if (storePositions) { int[] positions = tpVector.GetTermPositions(j); if (positions == null) { throw new System.SystemException("Trying to write positions that are null!"); } System.Diagnostics.Debug.Assert(positions.Length == termFreq); // use delta encoding for positions int lastPosition = 0; foreach (int position in positions) { tvf.WriteVInt(position - lastPosition); lastPosition = position; } } if (storeOffsets) { TermVectorOffsetInfo[] offsets = tpVector.GetOffsets(j); if (offsets == null) { throw new System.SystemException("Trying to write offsets that are null!"); } System.Diagnostics.Debug.Assert(offsets.Length == termFreq); // use delta encoding for offsets int lastEndOffset = 0; foreach (TermVectorOffsetInfo t in offsets) { int startOffset = t.StartOffset; int endOffset = t.EndOffset; tvf.WriteVInt(startOffset - lastEndOffset); tvf.WriteVInt(endOffset - startOffset); lastEndOffset = endOffset; } } } } // 2nd pass: write field pointers to tvd if (numFields > 1) { long lastFieldPointer = fieldPointers[0]; for (int i = 1; i < numFields; i++) { long fieldPointer = fieldPointers[i]; tvd.WriteVLong(fieldPointer - lastFieldPointer); lastFieldPointer = fieldPointer; } } } else { tvd.WriteVInt(0); } }
internal void Add(Term term, TermInfo ti) { UnicodeUtil.UTF16toUTF8(term.Text, 0, term.Text.Length, utf8Result); Add(fieldInfos.FieldNumber(term.Field), utf8Result.result, utf8Result.length, ti); }
public virtual void Get(int docNum, System.String field, TermVectorMapper mapper) { if (tvx != null) { int fieldNumber = fieldInfos.FieldNumber(field); //We need to account for the FORMAT_SIZE at when seeking in the tvx //We don't need to do this in other seeks because we already have the // file pointer //that was written in another file SeekTvx(docNum); //System.out.println("TVX Pointer: " + tvx.getFilePointer()); long tvdPosition = tvx.ReadLong(); tvd.Seek(tvdPosition); int fieldCount = tvd.ReadVInt(); //System.out.println("Num Fields: " + fieldCount); // There are only a few fields per document. We opt for a full scan // rather then requiring that they be ordered. We need to read through // all of the fields anyway to get to the tvf pointers. int number = 0; int found = -1; for (int i = 0; i < fieldCount; i++) { if (format >= FORMAT_VERSION) { number = tvd.ReadVInt(); } else { number += tvd.ReadVInt(); } if (number == fieldNumber) { found = i; } } // This field, although valid in the segment, was not found in this // document if (found != -1) { // Compute position in the tvf file long position; if (format >= FORMAT_VERSION2) { position = tvx.ReadLong(); } else { position = tvd.ReadVLong(); } for (int i = 1; i <= found; i++) { position += tvd.ReadVLong(); } mapper.SetDocumentNumber(docNum); ReadTermVector(field, position, mapper); } else { //System.out.println("Fieldable not found"); } } else { //System.out.println("No tvx file"); } }