public static void addRgFilteredStreamToRanges(OrcProto.Stream stream,
bool[] includedRowGroups, bool isCompressed, OrcProto.RowIndex index,
OrcProto.ColumnEncoding encoding, OrcProto.Type type, int compressionSize, bool hasNull,
long offset, long length, DiskRangeList.CreateHelper list, bool doMergeBuffers)
{
for (int group = 0; group < includedRowGroups.Length; ++group)
{
if (!includedRowGroups[group])
{
continue;
}
int posn = getIndexPosition(
encoding.Kind, type.Kind, stream.Kind, isCompressed, hasNull);
long start = (long)index.EntryList[group].PositionsList[posn];
long nextGroupOffset;
bool isLast = group == (includedRowGroups.Length - 1);
nextGroupOffset = isLast ? length : (int)index.EntryList[group + 1].PositionsList[posn];
start += offset;
long end = offset + estimateRgEndOffset(
isCompressed, isLast, nextGroupOffset, length, compressionSize);
list.addOrMerge(start, end, doMergeBuffers, true);
}
}
public RecordReaderImpl.Index readRowIndex(StripeInformation stripe,
OrcProto.StripeFooter footer, bool[] included, OrcProto.RowIndex[] indexes,
bool[] sargColumns, OrcProto.BloomFilterIndex[] bloomFilterIndices)
{
if (footer == null)
{
footer = readStripeFooter(stripe);
}
if (indexes == null)
{
indexes = new OrcProto.RowIndex[typeCount];
}
if (bloomFilterIndices == null)
{
bloomFilterIndices = new OrcProto.BloomFilterIndex[typeCount];
}
long offset = stripe.getOffset();
IList <OrcProto.Stream> streams = footer.StreamsList;
for (int i = 0; i < streams.Count; i++)
{
OrcProto.Stream stream = streams[i];
OrcProto.Stream nextStream = null;
if (i < streams.Count - 1)
{
nextStream = streams[i + 1];
}
int col = (int)stream.Column;
int len = (int)stream.Length;
// row index stream and bloom filter are interlaced, check if the sarg column contains bloom
// filter and combine the io to read row index and bloom filters for that column together
if (stream.HasKind && (stream.Kind == OrcProto.Stream.Types.Kind.ROW_INDEX))
{
bool readBloomFilter = false;
if (sargColumns != null && sargColumns[col] &&
nextStream.Kind == OrcProto.Stream.Types.Kind.BLOOM_FILTER)
{
len += (int)nextStream.Length;
i += 1;
readBloomFilter = true;
}
if ((included == null || included[col]) && indexes[col] == null)
{
byte[] buffer = new byte[len];
file.readFully(offset, buffer, 0, buffer.Length);
ByteBuffer bb = ByteBuffer.wrap(buffer);
indexes[col] = OrcProto.RowIndex.ParseFrom(InStream.create(null, "index",
new List <DiskRange> {
new RecordReaderImpl.BufferChunk(bb, 0)
},
(long)stream.Length, codec, bufferSize));
if (readBloomFilter)
{
bb.position((int)stream.Length);
bloomFilterIndices[col] = OrcProto.BloomFilterIndex.ParseFrom(InStream.create(
null, "bloom_filter", new List <DiskRange> {
new RecordReaderImpl.BufferChunk(bb, 0)
},
(long)nextStream.Length, codec, bufferSize));
}
}
}
offset += len;
}
RecordReaderImpl.Index index = new RecordReaderImpl.Index(indexes, bloomFilterIndices);
return(index);
}
