Search in sources :

Example 11 with RunFileWriter

use of org.apache.hyracks.dataflow.common.io.RunFileWriter in project asterixdb by apache.

the class AbstractExternalSortRunMerger method process.

public void process() throws HyracksDataException {
    IFrameWriter finalWriter = null;
    try {
        if (runs.isEmpty()) {
            finalWriter = prepareSkipMergingFinalResultWriter(writer);
            finalWriter.open();
            if (sorter != null) {
                try {
                    if (sorter.hasRemaining()) {
                        sorter.flush(finalWriter);
                    }
                } finally {
                    sorter.close();
                }
            }
        } else {
            /** recycle sort buffer */
            if (sorter != null) {
                sorter.close();
            }
            finalWriter = prepareFinalMergeResultWriter(writer);
            finalWriter.open();
            int maxMergeWidth = framesLimit - 1;
            inFrames = new ArrayList<>(maxMergeWidth);
            outputFrame = new VSizeFrame(ctx);
            List<GeneratedRunFileReader> partialRuns = new ArrayList<>(maxMergeWidth);
            int stop = runs.size();
            currentGenerationRunAvailable.set(0, stop);
            while (true) {
                int unUsed = selectPartialRuns(maxMergeWidth * ctx.getInitialFrameSize(), runs, partialRuns, currentGenerationRunAvailable, stop);
                prepareFrames(unUsed, inFrames, partialRuns);
                if (!currentGenerationRunAvailable.isEmpty() || stop < runs.size()) {
                    GeneratedRunFileReader reader;
                    if (partialRuns.size() == 1) {
                        if (!currentGenerationRunAvailable.isEmpty()) {
                            throw new HyracksDataException("The record is too big to put into the merging frame, please" + " allocate more sorting memory");
                        } else {
                            reader = partialRuns.get(0);
                        }
                    } else {
                        RunFileWriter mergeFileWriter = prepareIntermediateMergeRunFile();
                        IFrameWriter mergeResultWriter = prepareIntermediateMergeResultWriter(mergeFileWriter);
                        try {
                            mergeResultWriter.open();
                            merge(mergeResultWriter, partialRuns);
                        } catch (Throwable t) {
                            mergeResultWriter.fail();
                            throw t;
                        } finally {
                            mergeResultWriter.close();
                        }
                        reader = mergeFileWriter.createReader();
                    }
                    runs.add(reader);
                    if (currentGenerationRunAvailable.isEmpty()) {
                        if (LOGGER.isLoggable(Level.FINE)) {
                            LOGGER.fine("generated runs:" + stop);
                        }
                        runs.subList(0, stop).clear();
                        currentGenerationRunAvailable.clear();
                        currentGenerationRunAvailable.set(0, runs.size());
                        stop = runs.size();
                    }
                } else {
                    if (LOGGER.isLoggable(Level.FINE)) {
                        LOGGER.fine("final runs:" + stop);
                    }
                    merge(finalWriter, partialRuns);
                    break;
                }
            }
        }
    } catch (Exception e) {
        if (finalWriter != null) {
            finalWriter.fail();
        }
        throw HyracksDataException.create(e);
    } finally {
        try {
            if (finalWriter != null) {
                finalWriter.close();
            }
        } finally {
            for (RunFileReader reader : runs) {
                try {
                    // close is idempotent.
                    reader.close();
                } catch (Exception e) {
                    if (LOGGER.isLoggable(Level.WARNING)) {
                        LOGGER.log(Level.WARNING, e.getMessage(), e);
                    }
                }
            }
        }
    }
}
Also used : IFrameWriter(org.apache.hyracks.api.comm.IFrameWriter) ArrayList(java.util.ArrayList) GeneratedRunFileReader(org.apache.hyracks.dataflow.common.io.GeneratedRunFileReader) RunFileReader(org.apache.hyracks.dataflow.common.io.RunFileReader) VSizeFrame(org.apache.hyracks.api.comm.VSizeFrame) GroupVSizeFrame(org.apache.hyracks.dataflow.std.sort.util.GroupVSizeFrame) HyracksDataException(org.apache.hyracks.api.exceptions.HyracksDataException) HyracksDataException(org.apache.hyracks.api.exceptions.HyracksDataException) GeneratedRunFileReader(org.apache.hyracks.dataflow.common.io.GeneratedRunFileReader) RunFileWriter(org.apache.hyracks.dataflow.common.io.RunFileWriter)

Example 12 with RunFileWriter

use of org.apache.hyracks.dataflow.common.io.RunFileWriter in project asterixdb by apache.

the class OptimizedHybridHashJoin method closeAllSpilledPartitions.

private void closeAllSpilledPartitions(SIDE whichSide) throws HyracksDataException {
    RunFileWriter[] runFileWriters = null;
    switch(whichSide) {
        case BUILD:
            runFileWriters = buildRFWriters;
            break;
        case PROBE:
            runFileWriters = probeRFWriters;
            break;
    }
    try {
        for (int pid = spilledStatus.nextSetBit(0); pid >= 0 && pid < numOfPartitions; pid = spilledStatus.nextSetBit(pid + 1)) {
            if (bufferManager.getNumTuples(pid) > 0) {
                bufferManager.flushPartition(pid, getSpillWriterOrCreateNewOneIfNotExist(pid, whichSide));
                bufferManager.clearPartition(pid);
            }
        }
    } finally {
        // Force to close all run file writers.
        if (runFileWriters != null) {
            for (RunFileWriter runFileWriter : runFileWriters) {
                if (runFileWriter != null) {
                    runFileWriter.close();
                }
            }
        }
    }
}
Also used : RunFileWriter(org.apache.hyracks.dataflow.common.io.RunFileWriter)

Example 13 with RunFileWriter

use of org.apache.hyracks.dataflow.common.io.RunFileWriter in project asterixdb by apache.

the class OptimizedHybridHashJoin method flushBigProbeObjectToDisk.

private void flushBigProbeObjectToDisk(int pid, FrameTupleAccessor accessorProbe, int i) throws HyracksDataException {
    if (bigProbeFrameAppender == null) {
        bigProbeFrameAppender = new FrameTupleAppender(new VSizeFrame(ctx));
    }
    RunFileWriter runFileWriter = getSpillWriterOrCreateNewOneIfNotExist(pid, SIDE.PROBE);
    if (!bigProbeFrameAppender.append(accessorProbe, i)) {
        throw new HyracksDataException("The given tuple is too big");
    }
    bigProbeFrameAppender.write(runFileWriter, true);
}
Also used : FrameTupleAppender(org.apache.hyracks.dataflow.common.comm.io.FrameTupleAppender) VSizeFrame(org.apache.hyracks.api.comm.VSizeFrame) HyracksDataException(org.apache.hyracks.api.exceptions.HyracksDataException) RunFileWriter(org.apache.hyracks.dataflow.common.io.RunFileWriter)

Aggregations

RunFileWriter (org.apache.hyracks.dataflow.common.io.RunFileWriter)13 HyracksDataException (org.apache.hyracks.api.exceptions.HyracksDataException)4 ISpillableTable (org.apache.hyracks.dataflow.std.group.ISpillableTable)3 IFrameWriter (org.apache.hyracks.api.comm.IFrameWriter)2 VSizeFrame (org.apache.hyracks.api.comm.VSizeFrame)2 FileReference (org.apache.hyracks.api.io.FileReference)2 ArrayList (java.util.ArrayList)1 FrameTupleAppender (org.apache.hyracks.dataflow.common.comm.io.FrameTupleAppender)1 GeneratedRunFileReader (org.apache.hyracks.dataflow.common.io.GeneratedRunFileReader)1 RunFileReader (org.apache.hyracks.dataflow.common.io.RunFileReader)1 GroupVSizeFrame (org.apache.hyracks.dataflow.std.sort.util.GroupVSizeFrame)1