Examples with ValueStreamSource com.facebook.presto.orc.stream.ValueStreamSource used on opensource projects

Example 1 with ValueStreamSource

use of com.facebook.presto.orc.stream.ValueStreamSource in project presto by prestodb.

the class StripeReader method readStripe.

public Stripe readStripe(StripeInformation stripe, AggregatedMemoryContext systemMemoryUsage) throws IOException {
    // read the stripe footer
    StripeFooter stripeFooter = readStripeFooter(stripe, systemMemoryUsage);
    List<ColumnEncoding> columnEncodings = stripeFooter.getColumnEncodings();
    // get streams for selected columns
    Map<StreamId, Stream> streams = new HashMap<>();
    boolean hasRowGroupDictionary = false;
    for (Stream stream : stripeFooter.getStreams()) {
        if (includedOrcColumns.contains(stream.getColumn())) {
            streams.put(new StreamId(stream), stream);
            ColumnEncodingKind columnEncoding = columnEncodings.get(stream.getColumn()).getColumnEncodingKind();
            if (columnEncoding == DICTIONARY && stream.getStreamKind() == StreamKind.IN_DICTIONARY) {
                hasRowGroupDictionary = true;
            }
        }
    }
    // handle stripes with more than one row group or a dictionary
    if ((stripe.getNumberOfRows() > rowsInRowGroup) || hasRowGroupDictionary) {
        // determine ranges of the stripe to read
        Map<StreamId, DiskRange> diskRanges = getDiskRanges(stripeFooter.getStreams());
        diskRanges = Maps.filterKeys(diskRanges, Predicates.in(streams.keySet()));
        // read the file regions
        Map<StreamId, OrcInputStream> streamsData = readDiskRanges(stripe.getOffset(), diskRanges, systemMemoryUsage);
        // read the bloom filter for each column
        Map<Integer, List<HiveBloomFilter>> bloomFilterIndexes = readBloomFilterIndexes(streams, streamsData);
        // read the row index for each column
        Map<Integer, List<RowGroupIndex>> columnIndexes = readColumnIndexes(streams, streamsData, bloomFilterIndexes);
        // select the row groups matching the tuple domain
        Set<Integer> selectedRowGroups = selectRowGroups(stripe, columnIndexes);
        // if all row groups are skipped, return null
        if (selectedRowGroups.isEmpty()) {
            // set accounted memory usage to zero
            systemMemoryUsage.close();
            return null;
        }
        // value streams
        Map<StreamId, ValueStream<?>> valueStreams = createValueStreams(streams, streamsData, columnEncodings);
        // build the dictionary streams
        StreamSources dictionaryStreamSources = createDictionaryStreamSources(streams, valueStreams, columnEncodings);
        // build the row groups
        try {
            List<RowGroup> rowGroups = createRowGroups(stripe.getNumberOfRows(), streams, valueStreams, columnIndexes, selectedRowGroups, columnEncodings);
            return new Stripe(stripe.getNumberOfRows(), columnEncodings, rowGroups, dictionaryStreamSources);
        } catch (InvalidCheckpointException e) {
            // we must fail because the length of the row group dictionary is contained in the checkpoint stream.
            if (hasRowGroupDictionary) {
                throw new OrcCorruptionException(e, "ORC file %s has corrupt checkpoints", orcDataSource);
            }
        }
    }
    // stripe only has one row group and no dictionary
    ImmutableMap.Builder<StreamId, DiskRange> diskRangesBuilder = ImmutableMap.builder();
    for (Entry<StreamId, DiskRange> entry : getDiskRanges(stripeFooter.getStreams()).entrySet()) {
        StreamId streamId = entry.getKey();
        if (streamId.getStreamKind() != ROW_INDEX && streams.keySet().contains(streamId)) {
            diskRangesBuilder.put(entry);
        }
    }
    ImmutableMap<StreamId, DiskRange> diskRanges = diskRangesBuilder.build();
    // read the file regions
    Map<StreamId, OrcInputStream> streamsData = readDiskRanges(stripe.getOffset(), diskRanges, systemMemoryUsage);
    // value streams
    Map<StreamId, ValueStream<?>> valueStreams = createValueStreams(streams, streamsData, columnEncodings);
    // build the dictionary streams
    StreamSources dictionaryStreamSources = createDictionaryStreamSources(streams, valueStreams, columnEncodings);
    // build the row group
    ImmutableMap.Builder<StreamId, StreamSource<?>> builder = ImmutableMap.builder();
    for (Entry<StreamId, ValueStream<?>> entry : valueStreams.entrySet()) {
        builder.put(entry.getKey(), new ValueStreamSource<>(entry.getValue()));
    }
    RowGroup rowGroup = new RowGroup(0, 0, stripe.getNumberOfRows(), new StreamSources(builder.build()));
    return new Stripe(stripe.getNumberOfRows(), columnEncodings, ImmutableList.of(rowGroup), dictionaryStreamSources);
}