Examples with IOContext org.apache.lucene.store.IOContext used on opensource projects

Example 6 with IOContext

use of org.apache.lucene.store.IOContext in project lucene-solr by apache.

the class TermVectorsConsumer method initTermVectorsWriter.

void initTermVectorsWriter() throws IOException {
    if (writer == null) {
        IOContext context = new IOContext(new FlushInfo(docWriter.getNumDocsInRAM(), docWriter.bytesUsed()));
        writer = docWriter.codec.termVectorsFormat().vectorsWriter(docWriter.directory, docWriter.getSegmentInfo(), context);
        lastDocID = 0;
    }
}

Example 7 with IOContext

use of org.apache.lucene.store.IOContext in project lucene-solr by apache.

the class TestDoc method merge.

private SegmentCommitInfo merge(Directory dir, SegmentCommitInfo si1, SegmentCommitInfo si2, String merged, boolean useCompoundFile) throws Exception {
    IOContext context = newIOContext(random(), new IOContext(new MergeInfo(-1, -1, false, -1)));
    SegmentReader r1 = new SegmentReader(si1, Version.LATEST.major, context);
    SegmentReader r2 = new SegmentReader(si2, Version.LATEST.major, context);
    final Codec codec = Codec.getDefault();
    TrackingDirectoryWrapper trackingDir = new TrackingDirectoryWrapper(si1.info.dir);
    final SegmentInfo si = new SegmentInfo(si1.info.dir, Version.LATEST, null, merged, -1, false, codec, Collections.emptyMap(), StringHelper.randomId(), new HashMap<>(), null);
    SegmentMerger merger = new SegmentMerger(Arrays.<CodecReader>asList(r1, r2), si, InfoStream.getDefault(), trackingDir, new FieldInfos.FieldNumbers(), context);
    MergeState mergeState = merger.merge();
    r1.close();
    r2.close();
    ;
    si.setFiles(new HashSet<>(trackingDir.getCreatedFiles()));
    if (useCompoundFile) {
        Collection<String> filesToDelete = si.files();
        codec.compoundFormat().write(dir, si, context);
        si.setUseCompoundFile(true);
        for (String name : filesToDelete) {
            si1.info.dir.deleteFile(name);
        }
    }
    return new SegmentCommitInfo(si, 0, -1L, -1L, -1L);
}

Example 8 with IOContext

use of org.apache.lucene.store.IOContext in project lucene-solr by apache.

the class IndexWriter method mergeMiddle.

/** Does the actual (time-consuming) work of the merge,
   *  but without holding synchronized lock on IndexWriter
   *  instance */
private int mergeMiddle(MergePolicy.OneMerge merge, MergePolicy mergePolicy) throws IOException {
    merge.checkAborted();
    Directory mergeDirectory = config.getMergeScheduler().wrapForMerge(merge, directory);
    List<SegmentCommitInfo> sourceSegments = merge.segments;
    IOContext context = new IOContext(merge.getStoreMergeInfo());
    final TrackingDirectoryWrapper dirWrapper = new TrackingDirectoryWrapper(mergeDirectory);
    if (infoStream.isEnabled("IW")) {
        infoStream.message("IW", "merging " + segString(merge.segments));
    }
    merge.readers = new ArrayList<>(sourceSegments.size());
    // This is try/finally to make sure merger's readers are
    // closed:
    boolean success = false;
    try {
        int segUpto = 0;
        while (segUpto < sourceSegments.size()) {
            final SegmentCommitInfo info = sourceSegments.get(segUpto);
            // Hold onto the "live" reader; we will use this to
            // commit merged deletes
            final ReadersAndUpdates rld = readerPool.get(info, true);
            // Carefully pull the most recent live docs and reader
            SegmentReader reader;
            final Bits liveDocs;
            final int delCount;
            synchronized (this) {
                // Must sync to ensure BufferedDeletesStream cannot change liveDocs,
                // pendingDeleteCount and field updates while we pull a copy:
                reader = rld.getReaderForMerge(context);
                liveDocs = rld.getReadOnlyLiveDocs();
                delCount = rld.getPendingDeleteCount() + info.getDelCount();
                assert reader != null;
                assert rld.verifyDocCounts();
                if (infoStream.isEnabled("IW")) {
                    if (rld.getPendingDeleteCount() != 0) {
                        infoStream.message("IW", "seg=" + segString(info) + " delCount=" + info.getDelCount() + " pendingDelCount=" + rld.getPendingDeleteCount());
                    } else if (info.getDelCount() != 0) {
                        infoStream.message("IW", "seg=" + segString(info) + " delCount=" + info.getDelCount());
                    } else {
                        infoStream.message("IW", "seg=" + segString(info) + " no deletes");
                    }
                }
            }
            // make a new reader with updated live docs and del count.
            if (reader.numDeletedDocs() != delCount) {
                // beware of zombies
                assert delCount > reader.numDeletedDocs();
                SegmentReader newReader;
                synchronized (this) {
                    // We must also sync on IW here, because another thread could be writing
                    // new DV updates / remove old gen field infos files causing FNFE:
                    newReader = new SegmentReader(info, reader, liveDocs, info.info.maxDoc() - delCount);
                }
                boolean released = false;
                try {
                    rld.release(reader);
                    released = true;
                } finally {
                    if (!released) {
                        newReader.decRef();
                    }
                }
                reader = newReader;
            }
            merge.readers.add(reader);
            assert delCount <= info.info.maxDoc() : "delCount=" + delCount + " info.maxDoc=" + info.info.maxDoc() + " rld.pendingDeleteCount=" + rld.getPendingDeleteCount() + " info.getDelCount()=" + info.getDelCount();
            segUpto++;
        }
        //      System.out.println("[" + Thread.currentThread().getName() + "] IW.mergeMiddle: merging " + merge.getMergeReaders());
        // Let the merge wrap readers
        List<CodecReader> mergeReaders = new ArrayList<>();
        for (SegmentReader reader : merge.readers) {
            CodecReader wrappedReader = merge.wrapForMerge(reader);
            validateMergeReader(wrappedReader);
            mergeReaders.add(wrappedReader);
        }
        final SegmentMerger merger = new SegmentMerger(mergeReaders, merge.info.info, infoStream, dirWrapper, globalFieldNumberMap, context);
        merge.checkAborted();
        merge.mergeStartNS = System.nanoTime();
        // This is where all the work happens:
        if (merger.shouldMerge()) {
            merger.merge();
        }
        MergeState mergeState = merger.mergeState;
        assert mergeState.segmentInfo == merge.info.info;
        merge.info.info.setFiles(new HashSet<>(dirWrapper.getCreatedFiles()));
        if (infoStream.isEnabled("IW")) {
            if (merger.shouldMerge()) {
                String pauseInfo = merge.getMergeProgress().getPauseTimes().entrySet().stream().filter((e) -> e.getValue() > 0).map((e) -> String.format(Locale.ROOT, "%.1f sec %s", e.getValue() / 1000000000., e.getKey().name().toLowerCase(Locale.ROOT))).collect(Collectors.joining(", "));
                if (!pauseInfo.isEmpty()) {
                    pauseInfo = " (" + pauseInfo + ")";
                }
                long t1 = System.nanoTime();
                double sec = (t1 - merge.mergeStartNS) / 1000000000.;
                double segmentMB = (merge.info.sizeInBytes() / 1024. / 1024.);
                infoStream.message("IW", "merge codec=" + codec + " maxDoc=" + merge.info.info.maxDoc() + "; merged segment has " + (mergeState.mergeFieldInfos.hasVectors() ? "vectors" : "no vectors") + "; " + (mergeState.mergeFieldInfos.hasNorms() ? "norms" : "no norms") + "; " + (mergeState.mergeFieldInfos.hasDocValues() ? "docValues" : "no docValues") + "; " + (mergeState.mergeFieldInfos.hasProx() ? "prox" : "no prox") + "; " + (mergeState.mergeFieldInfos.hasProx() ? "freqs" : "no freqs") + "; " + (mergeState.mergeFieldInfos.hasPointValues() ? "points" : "no points") + "; " + String.format(Locale.ROOT, "%.1f sec%s to merge segment [%.2f MB, %.2f MB/sec]", sec, pauseInfo, segmentMB, segmentMB / sec));
            } else {
                infoStream.message("IW", "skip merging fully deleted segments");
            }
        }
        if (merger.shouldMerge() == false) {
            // Merge would produce a 0-doc segment, so we do nothing except commit the merge to remove all the 0-doc segments that we "merged":
            assert merge.info.info.maxDoc() == 0;
            commitMerge(merge, mergeState);
            return 0;
        }
        assert merge.info.info.maxDoc() > 0;
        // Very important to do this before opening the reader
        // because codec must know if prox was written for
        // this segment:
        //System.out.println("merger set hasProx=" + merger.hasProx() + " seg=" + merge.info.name);
        boolean useCompoundFile;
        synchronized (this) {
            // Guard segmentInfos
            useCompoundFile = mergePolicy.useCompoundFile(segmentInfos, merge.info, this);
        }
        if (useCompoundFile) {
            success = false;
            Collection<String> filesToRemove = merge.info.files();
            TrackingDirectoryWrapper trackingCFSDir = new TrackingDirectoryWrapper(mergeDirectory);
            try {
                createCompoundFile(infoStream, trackingCFSDir, merge.info.info, context);
                success = true;
            } catch (Throwable t) {
                synchronized (this) {
                    if (merge.isAborted()) {
                        // merged files:
                        if (infoStream.isEnabled("IW")) {
                            infoStream.message("IW", "hit merge abort exception creating compound file during merge");
                        }
                        return 0;
                    } else {
                        handleMergeException(t, merge);
                    }
                }
            } finally {
                if (success == false) {
                    if (infoStream.isEnabled("IW")) {
                        infoStream.message("IW", "hit exception creating compound file during merge");
                    }
                    // Safe: these files must exist
                    deleteNewFiles(merge.info.files());
                }
            }
            // So that, if we hit exc in deleteNewFiles (next)
            // or in commitMerge (later), we close the
            // per-segment readers in the finally clause below:
            success = false;
            synchronized (this) {
                // delete new non cfs files directly: they were never
                // registered with IFD
                deleteNewFiles(filesToRemove);
                if (merge.isAborted()) {
                    if (infoStream.isEnabled("IW")) {
                        infoStream.message("IW", "abort merge after building CFS");
                    }
                    // Safe: these files must exist
                    deleteNewFiles(merge.info.files());
                    return 0;
                }
            }
            merge.info.info.setUseCompoundFile(true);
        } else {
            // So that, if we hit exc in commitMerge (later),
            // we close the per-segment readers in the finally
            // clause below:
            success = false;
        }
        // Have codec write SegmentInfo.  Must do this after
        // creating CFS so that 1) .si isn't slurped into CFS,
        // and 2) .si reflects useCompoundFile=true change
        // above:
        boolean success2 = false;
        try {
            codec.segmentInfoFormat().write(directory, merge.info.info, context);
            success2 = true;
        } finally {
            if (!success2) {
                // Safe: these files must exist
                deleteNewFiles(merge.info.files());
            }
        }
        if (infoStream.isEnabled("IW")) {
            infoStream.message("IW", String.format(Locale.ROOT, "merged segment size=%.3f MB vs estimate=%.3f MB", merge.info.sizeInBytes() / 1024. / 1024., merge.estimatedMergeBytes / 1024 / 1024.));
        }
        final IndexReaderWarmer mergedSegmentWarmer = config.getMergedSegmentWarmer();
        if (poolReaders && mergedSegmentWarmer != null) {
            final ReadersAndUpdates rld = readerPool.get(merge.info, true);
            final SegmentReader sr = rld.getReader(IOContext.READ);
            try {
                mergedSegmentWarmer.warm(sr);
            } finally {
                synchronized (this) {
                    rld.release(sr);
                    readerPool.release(rld);
                }
            }
        }
        if (!commitMerge(merge, mergeState)) {
            // aborted
            return 0;
        }
        success = true;
    } finally {
        // an exc:
        if (success == false) {
            closeMergeReaders(merge, true);
        }
    }
    return merge.info.info.maxDoc();
}

Example 9 with IOContext

use of org.apache.lucene.store.IOContext in project lucene-solr by apache.

the class IndexWriter method addIndexes.

/**
   * Adds all segments from an array of indexes into this index.
   *
   * <p>This may be used to parallelize batch indexing. A large document
   * collection can be broken into sub-collections. Each sub-collection can be
   * indexed in parallel, on a different thread, process or machine. The
   * complete index can then be created by merging sub-collection indexes
   * with this method.
   *
   * <p>
   * <b>NOTE:</b> this method acquires the write lock in
   * each directory, to ensure that no {@code IndexWriter}
   * is currently open or tries to open while this is
   * running.
   *
   * <p>This method is transactional in how Exceptions are
   * handled: it does not commit a new segments_N file until
   * all indexes are added.  This means if an Exception
   * occurs (for example disk full), then either no indexes
   * will have been added or they all will have been.
   *
   * <p>Note that this requires temporary free space in the
   * {@link Directory} up to 2X the sum of all input indexes
   * (including the starting index). If readers/searchers
   * are open against the starting index, then temporary
   * free space required will be higher by the size of the
   * starting index (see {@link #forceMerge(int)} for details).
   *
   * <p>This requires this index not be among those to be added.
   *
   * <p>All added indexes must have been created by the same
   * Lucene version as this index.
   *
   * @return The <a href="#sequence_number">sequence number</a>
   * for this operation
   *
   * @throws CorruptIndexException if the index is corrupt
   * @throws IOException if there is a low-level IO error
   * @throws IllegalArgumentException if addIndexes would cause
   *   the index to exceed {@link #MAX_DOCS}, or if the indoming
   *   index sort does not match this index's index sort
   */
public long addIndexes(Directory... dirs) throws IOException {
    ensureOpen();
    noDupDirs(dirs);
    List<Lock> locks = acquireWriteLocks(dirs);
    Sort indexSort = config.getIndexSort();
    boolean successTop = false;
    long seqNo;
    try {
        if (infoStream.isEnabled("IW")) {
            infoStream.message("IW", "flush at addIndexes(Directory...)");
        }
        flush(false, true);
        List<SegmentCommitInfo> infos = new ArrayList<>();
        // long so we can detect int overflow:
        long totalMaxDoc = 0;
        List<SegmentInfos> commits = new ArrayList<>(dirs.length);
        for (Directory dir : dirs) {
            if (infoStream.isEnabled("IW")) {
                infoStream.message("IW", "addIndexes: process directory " + dir);
            }
            // read infos from dir
            SegmentInfos sis = SegmentInfos.readLatestCommit(dir);
            if (segmentInfos.getIndexCreatedVersionMajor() != sis.getIndexCreatedVersionMajor()) {
                throw new IllegalArgumentException("Cannot use addIndexes(Directory) with indexes that have been created " + "by a different Lucene version. The current index was generated by Lucene " + segmentInfos.getIndexCreatedVersionMajor() + " while one of the directories contains an index that was generated with Lucene " + sis.getIndexCreatedVersionMajor());
            }
            totalMaxDoc += sis.totalMaxDoc();
            commits.add(sis);
        }
        // Best-effort up front check:
        testReserveDocs(totalMaxDoc);
        boolean success = false;
        try {
            for (SegmentInfos sis : commits) {
                for (SegmentCommitInfo info : sis) {
                    assert !infos.contains(info) : "dup info dir=" + info.info.dir + " name=" + info.info.name;
                    Sort segmentIndexSort = info.info.getIndexSort();
                    if (indexSort != null && segmentIndexSort != null && indexSort.equals(segmentIndexSort) == false) {
                        // TODO: we could make this smarter, e.g. if the incoming indexSort is congruent with our sort ("starts with") then it's OK
                        throw new IllegalArgumentException("cannot change index sort from " + segmentIndexSort + " to " + indexSort);
                    }
                    String newSegName = newSegmentName();
                    if (infoStream.isEnabled("IW")) {
                        infoStream.message("IW", "addIndexes: process segment origName=" + info.info.name + " newName=" + newSegName + " info=" + info);
                    }
                    IOContext context = new IOContext(new FlushInfo(info.info.maxDoc(), info.sizeInBytes()));
                    FieldInfos fis = readFieldInfos(info);
                    for (FieldInfo fi : fis) {
                        // This will throw exceptions if any of the incoming fields have an illegal schema change:
                        globalFieldNumberMap.addOrGet(fi.name, fi.number, fi.getDocValuesType(), fi.getPointDimensionCount(), fi.getPointNumBytes());
                    }
                    infos.add(copySegmentAsIs(info, newSegName, context));
                }
            }
            success = true;
        } finally {
            if (!success) {
                for (SegmentCommitInfo sipc : infos) {
                    // Safe: these files must exist
                    deleteNewFiles(sipc.files());
                }
            }
        }
        synchronized (this) {
            success = false;
            try {
                ensureOpen();
                // Now reserve the docs, just before we update SIS:
                reserveDocs(totalMaxDoc);
                seqNo = docWriter.deleteQueue.getNextSequenceNumber();
                success = true;
            } finally {
                if (!success) {
                    for (SegmentCommitInfo sipc : infos) {
                        // Safe: these files must exist
                        deleteNewFiles(sipc.files());
                    }
                }
            }
            segmentInfos.addAll(infos);
            checkpoint();
        }
        successTop = true;
    } catch (VirtualMachineError tragedy) {
        tragicEvent(tragedy, "addIndexes(Directory...)");
        // dead code but javac disagrees:
        seqNo = -1;
    } finally {
        if (successTop) {
            IOUtils.close(locks);
        } else {
            IOUtils.closeWhileHandlingException(locks);
        }
    }
    maybeMerge();
    return seqNo;
}

Example 10 with IOContext

use of org.apache.lucene.store.IOContext in project lucene-solr by apache.

the class ReadersAndUpdates method handleNumericDVUpdates.

@SuppressWarnings("synthetic-access")
private void handleNumericDVUpdates(FieldInfos infos, Map<String, NumericDocValuesFieldUpdates> updates, Directory dir, DocValuesFormat dvFormat, final SegmentReader reader, Map<Integer, Set<String>> fieldFiles) throws IOException {
    for (Entry<String, NumericDocValuesFieldUpdates> e : updates.entrySet()) {
        final String field = e.getKey();
        final NumericDocValuesFieldUpdates fieldUpdates = e.getValue();
        final long nextDocValuesGen = info.getNextDocValuesGen();
        final String segmentSuffix = Long.toString(nextDocValuesGen, Character.MAX_RADIX);
        final long estUpdatesSize = fieldUpdates.ramBytesPerDoc() * info.info.maxDoc();
        final IOContext updatesContext = new IOContext(new FlushInfo(info.info.maxDoc(), estUpdatesSize));
        final FieldInfo fieldInfo = infos.fieldInfo(field);
        assert fieldInfo != null;
        fieldInfo.setDocValuesGen(nextDocValuesGen);
        final FieldInfos fieldInfos = new FieldInfos(new FieldInfo[] { fieldInfo });
        // separately also track which files were created for this gen
        final TrackingDirectoryWrapper trackingDir = new TrackingDirectoryWrapper(dir);
        final SegmentWriteState state = new SegmentWriteState(null, trackingDir, info.info, fieldInfos, null, updatesContext, segmentSuffix);
        try (final DocValuesConsumer fieldsConsumer = dvFormat.fieldsConsumer(state)) {
            // write the numeric updates to a new gen'd docvalues file
            fieldsConsumer.addNumericField(fieldInfo, new EmptyDocValuesProducer() {

                @Override
                public NumericDocValues getNumeric(FieldInfo fieldInfoIn) throws IOException {
                    if (fieldInfoIn != fieldInfo) {
                        throw new IllegalArgumentException("wrong fieldInfo");
                    }
                    final int maxDoc = reader.maxDoc();
                    final NumericDocValuesFieldUpdates.Iterator updatesIter = fieldUpdates.iterator();
                    final NumericDocValues currentValues = reader.getNumericDocValues(field);
                    updatesIter.reset();
                    // Merge sort of the original doc values with updated doc values:
                    return new NumericDocValues() {

                        // merged docID
                        private int docIDOut = -1;

                        // docID from our original doc values
                        private int docIDIn = -1;

                        // docID from our updates
                        private int updateDocID = -1;

                        private long value;

                        @Override
                        public int docID() {
                            return docIDOut;
                        }

                        @Override
                        public int advance(int target) {
                            throw new UnsupportedOperationException();
                        }

                        @Override
                        public boolean advanceExact(int target) throws IOException {
                            throw new UnsupportedOperationException();
                        }

                        @Override
                        public long cost() {
                            // TODO
                            return 0;
                        }

                        @Override
                        public long longValue() {
                            return value;
                        }

                        @Override
                        public int nextDoc() throws IOException {
                            if (docIDIn == docIDOut) {
                                if (currentValues == null) {
                                    docIDIn = NO_MORE_DOCS;
                                } else {
                                    docIDIn = currentValues.nextDoc();
                                }
                            }
                            if (updateDocID == docIDOut) {
                                updateDocID = updatesIter.nextDoc();
                            }
                            if (docIDIn < updateDocID) {
                                // no update to this doc
                                docIDOut = docIDIn;
                                value = currentValues.longValue();
                            } else {
                                docIDOut = updateDocID;
                                if (docIDOut != NO_MORE_DOCS) {
                                    value = updatesIter.value();
                                }
                            }
                            return docIDOut;
                        }
                    };
                }
            });
        }
        info.advanceDocValuesGen();
        assert !fieldFiles.containsKey(fieldInfo.number);
        fieldFiles.put(fieldInfo.number, trackingDir.getCreatedFiles());
    }
}