Examples with FileInfo org.apache.hadoop.hbase.io.hfile.HFile.FileInfo used on opensource projects

Example 1 with FileInfo

use of org.apache.hadoop.hbase.io.hfile.HFile.FileInfo in project hbase by apache.

the class TestHFileWriterV3 method writeDataAndReadFromHFile.

private void writeDataAndReadFromHFile(Path hfilePath, Algorithm compressAlgo, int entryCount, boolean findMidKey, boolean useTags) throws IOException {
    HFileContext context = new HFileContextBuilder().withBlockSize(4096).withIncludesTags(useTags).withCompression(compressAlgo).build();
    HFile.Writer writer = new HFile.WriterFactory(conf, new CacheConfig(conf)).withPath(fs, hfilePath).withFileContext(context).withComparator(CellComparator.COMPARATOR).create();
    // Just a fixed seed.
    Random rand = new Random(9713312);
    List<KeyValue> keyValues = new ArrayList<>(entryCount);
    for (int i = 0; i < entryCount; ++i) {
        byte[] keyBytes = RandomKeyValueUtil.randomOrderedKey(rand, i);
        // A random-length random value.
        byte[] valueBytes = RandomKeyValueUtil.randomValue(rand);
        KeyValue keyValue = null;
        if (useTags) {
            ArrayList<Tag> tags = new ArrayList<>();
            for (int j = 0; j < 1 + rand.nextInt(4); j++) {
                byte[] tagBytes = new byte[16];
                rand.nextBytes(tagBytes);
                tags.add(new ArrayBackedTag((byte) 1, tagBytes));
            }
            keyValue = new KeyValue(keyBytes, null, null, HConstants.LATEST_TIMESTAMP, valueBytes, tags);
        } else {
            keyValue = new KeyValue(keyBytes, null, null, HConstants.LATEST_TIMESTAMP, valueBytes);
        }
        writer.append(keyValue);
        keyValues.add(keyValue);
    }
    // Add in an arbitrary order. They will be sorted lexicographically by
    // the key.
    writer.appendMetaBlock("CAPITAL_OF_USA", new Text("Washington, D.C."));
    writer.appendMetaBlock("CAPITAL_OF_RUSSIA", new Text("Moscow"));
    writer.appendMetaBlock("CAPITAL_OF_FRANCE", new Text("Paris"));
    writer.close();
    FSDataInputStream fsdis = fs.open(hfilePath);
    long fileSize = fs.getFileStatus(hfilePath).getLen();
    FixedFileTrailer trailer = FixedFileTrailer.readFromStream(fsdis, fileSize);
    assertEquals(3, trailer.getMajorVersion());
    assertEquals(entryCount, trailer.getEntryCount());
    HFileContext meta = new HFileContextBuilder().withCompression(compressAlgo).withIncludesMvcc(false).withIncludesTags(useTags).withHBaseCheckSum(true).build();
    HFileBlock.FSReader blockReader = new HFileBlock.FSReaderImpl(fsdis, fileSize, meta);
    // Comparator class name is stored in the trailer in version 3.
    CellComparator comparator = trailer.createComparator();
    HFileBlockIndex.BlockIndexReader dataBlockIndexReader = new HFileBlockIndex.CellBasedKeyBlockIndexReader(comparator, trailer.getNumDataIndexLevels());
    HFileBlockIndex.BlockIndexReader metaBlockIndexReader = new HFileBlockIndex.ByteArrayKeyBlockIndexReader(1);
    HFileBlock.BlockIterator blockIter = blockReader.blockRange(trailer.getLoadOnOpenDataOffset(), fileSize - trailer.getTrailerSize());
    // Data index. We also read statistics about the block index written after
    // the root level.
    dataBlockIndexReader.readMultiLevelIndexRoot(blockIter.nextBlockWithBlockType(BlockType.ROOT_INDEX), trailer.getDataIndexCount());
    if (findMidKey) {
        Cell midkey = dataBlockIndexReader.midkey();
        assertNotNull("Midkey should not be null", midkey);
    }
    // Meta index.
    metaBlockIndexReader.readRootIndex(blockIter.nextBlockWithBlockType(BlockType.ROOT_INDEX).getByteStream(), trailer.getMetaIndexCount());
    // File info
    FileInfo fileInfo = new FileInfo();
    fileInfo.read(blockIter.nextBlockWithBlockType(BlockType.FILE_INFO).getByteStream());
    byte[] keyValueFormatVersion = fileInfo.get(HFileWriterImpl.KEY_VALUE_VERSION);
    boolean includeMemstoreTS = keyValueFormatVersion != null && Bytes.toInt(keyValueFormatVersion) > 0;
    // Counters for the number of key/value pairs and the number of blocks
    int entriesRead = 0;
    int blocksRead = 0;
    long memstoreTS = 0;
    // Scan blocks the way the reader would scan them
    fsdis.seek(0);
    long curBlockPos = 0;
    while (curBlockPos <= trailer.getLastDataBlockOffset()) {
        HFileBlock block = blockReader.readBlockData(curBlockPos, -1, false).unpack(context, blockReader);
        assertEquals(BlockType.DATA, block.getBlockType());
        ByteBuff buf = block.getBufferWithoutHeader();
        int keyLen = -1;
        while (buf.hasRemaining()) {
            keyLen = buf.getInt();
            int valueLen = buf.getInt();
            byte[] key = new byte[keyLen];
            buf.get(key);
            byte[] value = new byte[valueLen];
            buf.get(value);
            byte[] tagValue = null;
            if (useTags) {
                int tagLen = ((buf.get() & 0xff) << 8) ^ (buf.get() & 0xff);
                tagValue = new byte[tagLen];
                buf.get(tagValue);
            }
            if (includeMemstoreTS) {
                ByteArrayInputStream byte_input = new ByteArrayInputStream(buf.array(), buf.arrayOffset() + buf.position(), buf.remaining());
                DataInputStream data_input = new DataInputStream(byte_input);
                memstoreTS = WritableUtils.readVLong(data_input);
                buf.position(buf.position() + WritableUtils.getVIntSize(memstoreTS));
            }
            // A brute-force check to see that all keys and values are correct.
            KeyValue kv = keyValues.get(entriesRead);
            assertTrue(Bytes.compareTo(key, kv.getKey()) == 0);
            assertTrue(Bytes.compareTo(value, 0, value.length, kv.getValueArray(), kv.getValueOffset(), kv.getValueLength()) == 0);
            if (useTags) {
                assertNotNull(tagValue);
                KeyValue tkv = kv;
                assertEquals(tagValue.length, tkv.getTagsLength());
                assertTrue(Bytes.compareTo(tagValue, 0, tagValue.length, tkv.getTagsArray(), tkv.getTagsOffset(), tkv.getTagsLength()) == 0);
            }
            ++entriesRead;
        }
        ++blocksRead;
        curBlockPos += block.getOnDiskSizeWithHeader();
    }
    LOG.info("Finished reading: entries=" + entriesRead + ", blocksRead=" + blocksRead);
    assertEquals(entryCount, entriesRead);
    // Meta blocks. We can scan until the load-on-open data offset (which is
    // the root block index offset in version 2) because we are not testing
    // intermediate-level index blocks here.
    int metaCounter = 0;
    while (fsdis.getPos() < trailer.getLoadOnOpenDataOffset()) {
        LOG.info("Current offset: " + fsdis.getPos() + ", scanning until " + trailer.getLoadOnOpenDataOffset());
        HFileBlock block = blockReader.readBlockData(curBlockPos, -1, false).unpack(context, blockReader);
        assertEquals(BlockType.META, block.getBlockType());
        Text t = new Text();
        ByteBuff buf = block.getBufferWithoutHeader();
        if (Writables.getWritable(buf.array(), buf.arrayOffset(), buf.limit(), t) == null) {
            throw new IOException("Failed to deserialize block " + this + " into a " + t.getClass().getSimpleName());
        }
        Text expectedText = (metaCounter == 0 ? new Text("Paris") : metaCounter == 1 ? new Text("Moscow") : new Text("Washington, D.C."));
        assertEquals(expectedText, t);
        LOG.info("Read meta block data: " + t);
        ++metaCounter;
        curBlockPos += block.getOnDiskSizeWithHeader();
    }
    fsdis.close();
}