Examples of org.apache.hadoop.hbase.io.hfile.HFileBlock

Package org.apache.hadoop.hbase.io.hfile

Examples of org.apache.hadoop.hbase.io.hfile.HFileBlock

org.apache.hadoop.hbase.io.hfile.HFileBlock
Reading {@link HFile} version 1 and 2 blocks, and writing version 2 blocks.
- In version 1 all blocks are always compressed or uncompressed, as specified by the {@link HFile}'s compression algorithm, with a type-specific magic record stored in the beginning of the compressed data (i.e. one needs to uncompress the compressed block to determine the block type). There is only a single compression algorithm setting for all blocks. Offset and size information from the block index are required to read a block.
- In version 2 a block is structured as follows:
  - Magic record identifying the block type (8 bytes)
  - Compressed block size, header not included (4 bytes)
  - Uncompressed block size, header not included (4 bytes)
  - The offset of the previous block of the same type (8 bytes). This is used to be able to navigate to the previous block without going to the block
  - For minorVersions >=1, there is an additional 4 byte field bytesPerChecksum that records the number of bytes in a checksum chunk.
  - For minorVersions >=1, there is a 4 byte value to store the size of data on disk (excluding the checksums)
  - For minorVersions >=1, a series of 4 byte checksums, one each for the number of bytes specified by bytesPerChecksum. index.
  - Compressed data (or uncompressed data if compression is disabled). The compression algorithm is the same for all the blocks in the {@link HFile}, similarly to what was done in version 1.
The version 2 block representation in the block cache is the same as above, except that the data section is always uncompressed in the cache.

      // Open a scanner with (on read) caching disabled
      HFileScanner scanner = reader.getScanner(false, false);
      assertTrue(testDescription, scanner.seekTo());
      // Cribbed from io.hfile.TestCacheOnWrite
      long offset = 0;
      HFileBlock prevBlock = null;
      while (offset < reader.getTrailer().getLoadOnOpenDataOffset()) {
        long onDiskSize = -1;
        if (prevBlock != null) {
          onDiskSize = prevBlock.getNextBlockOnDiskSizeWithHeader();
        }
        // Flags: don't cache the block, use pread, this is not a compaction.
        // Also, pass null for expected block type to avoid checking it.
        HFileBlock block = reader.readBlock(offset, onDiskSize, false, true,
          false, null);
        BlockCacheKey blockCacheKey = new BlockCacheKey(reader.getName(),
          offset);
        boolean isCached = cache.getBlock(blockCacheKey, true, false) != null;
        boolean shouldBeCached = cowType.shouldBeCached(block.getBlockType());
        if (shouldBeCached != isCached) {
          throw new AssertionError(
            "shouldBeCached: " + shouldBeCached+ "\n" +
            "isCached: " + isCached + "\n" +
            "Test description: " + testDescription + "\n" +
            "block: " + block + "\n" +
            "blockCacheKey: " + blockCacheKey);
        }
        prevBlock = block;
        offset += block.getOnDiskSizeWithHeader();
      }
    } finally {
      reader.close();
    }
  }

View Full Code Here

      // Open a scanner with (on read) caching disabled
      HFileScanner scanner = reader.getScanner(false, false);
      assertTrue(testDescription, scanner.seekTo());
      // Cribbed from io.hfile.TestCacheOnWrite
      long offset = 0;
      HFileBlock prevBlock = null;
      while (offset < reader.getTrailer().getLoadOnOpenDataOffset()) {
        long onDiskSize = -1;
        if (prevBlock != null) {
          onDiskSize = prevBlock.getNextBlockOnDiskSizeWithHeader();
        }
        // Flags: don't cache the block, use pread, this is not a compaction.
        // Also, pass null for expected block type to avoid checking it.
        HFileBlock block = reader.readBlock(offset, onDiskSize, false, true,
          false, null);
        BlockCacheKey blockCacheKey = new BlockCacheKey(reader.getName(),
          offset);
        boolean isCached = cache.getBlock(blockCacheKey, true, false) != null;
        boolean shouldBeCached = cowType.shouldBeCached(block.getBlockType());
        if (shouldBeCached != isCached) {
          throw new AssertionError(
            "shouldBeCached: " + shouldBeCached+ "\n" +
            "isCached: " + isCached + "\n" +
            "Test description: " + testDescription + "\n" +
            "block: " + block + "\n" +
            "blockCacheKey: " + blockCacheKey);
        }
        prevBlock = block;
        offset += block.getOnDiskSizeWithHeader();
      }
    } finally {
      reader.close();
    }
  }

View Full Code Here

      // Open a scanner with (on read) caching disabled
      HFileScanner scanner = reader.getScanner(false, false);
      assertTrue(testDescription, scanner.seekTo());
      // Cribbed from io.hfile.TestCacheOnWrite
      long offset = 0;
      HFileBlock prevBlock = null;
      while (offset < reader.getTrailer().getLoadOnOpenDataOffset()) {
        long onDiskSize = -1;
        if (prevBlock != null) {
          onDiskSize = prevBlock.getNextBlockOnDiskSizeWithHeader();
        }
        // Flags: don't cache the block, use pread, this is not a compaction.
        // Also, pass null for expected block type to avoid checking it.
        HFileBlock block = reader.readBlock(offset, onDiskSize, false, true,
          false, null);
        BlockCacheKey blockCacheKey = new BlockCacheKey(reader.getName(),
          offset);
        boolean isCached = cache.getBlock(blockCacheKey, true, false) != null;
        boolean shouldBeCached = cowType.shouldBeCached(block.getBlockType());
        if (shouldBeCached != isCached) {
          throw new AssertionError(
            "shouldBeCached: " + shouldBeCached+ "\n" +
            "isCached: " + isCached + "\n" +
            "Test description: " + testDescription + "\n" +
            "block: " + block + "\n" +
            "blockCacheKey: " + blockCacheKey);
        }
        prevBlock = block;
        offset += block.getOnDiskSizeWithHeader();
      }
    } finally {
      reader.close();
    }
  }

View Full Code Here

      // Open a scanner with (on read) caching disabled
      HFileScanner scanner = reader.getScanner(false, false);
      assertTrue(testDescription, scanner.seekTo());
      // Cribbed from io.hfile.TestCacheOnWrite
      long offset = 0;
      HFileBlock prevBlock = null;
      while (offset < reader.getTrailer().getLoadOnOpenDataOffset()) {
        long onDiskSize = -1;
        if (prevBlock != null) {
          onDiskSize = prevBlock.getNextBlockOnDiskSizeWithHeader();
        }
        // Flags: don't cache the block, use pread, this is not a compaction.
        // Also, pass null for expected block type to avoid checking it.
        HFileBlock block = reader.readBlock(offset, onDiskSize, false, true,
          false, null);
        BlockCacheKey blockCacheKey = new BlockCacheKey(reader.getName(),
          offset);
        boolean isCached = cache.getBlock(blockCacheKey, true, false) != null;
        boolean shouldBeCached = cowType.shouldBeCached(block.getBlockType());
        if (shouldBeCached != isCached) {
          throw new AssertionError(
            "shouldBeCached: " + shouldBeCached+ "\n" +
            "isCached: " + isCached + "\n" +
            "Test description: " + testDescription + "\n" +
            "block: " + block + "\n" +
            "blockCacheKey: " + blockCacheKey);
        }
        prevBlock = block;
        offset += block.getOnDiskSizeWithHeader();
      }
    } finally {
      reader.close();
    }
  }

View Full Code Here


    int block = index.rootBlockContainingKey(key, keyOffset, keyLength);
    if (block < 0) {
      result = false; // This key is not in the file.
    } else {
      HFileBlock bloomBlock;
      try {
        // We cache the block and use a positional read.
        bloomBlock = reader.readBlock(index.getRootBlockOffset(block),
            index.getRootBlockDataSize(block), true, true, false,
            BlockType.BLOOM_CHUNK);
      } catch (IOException ex) {
        // The Bloom filter is broken, turn it off.
        throw new IllegalArgumentException(
            "Failed to load Bloom block for key "
                + Bytes.toStringBinary(key, keyOffset, keyLength), ex);
      }


      ByteBuffer bloomBuf = bloomBlock.getBufferReadOnly();
      result = ByteBloomFilter.contains(key, keyOffset, keyLength,
          bloomBuf.array(), bloomBuf.arrayOffset() + bloomBlock.headerSize(),
          bloomBlock.getUncompressedSizeWithoutHeader(), hash, hashCount);
    }


    if (numQueriesPerChunk != null && block >= 0) {
      // Update statistics. Only used in unit tests.
      ++numQueriesPerChunk[block];

View Full Code Here


    int block = index.rootBlockContainingKey(key, keyOffset, keyLength);
    if (block < 0) {
      result = false; // This key is not in the file.
    } else {
      HFileBlock bloomBlock;
      try {
        // We cache the block and use a positional read.
        bloomBlock = reader.readBlock(index.getRootBlockOffset(block),
            index.getRootBlockDataSize(block), true, true, false);
      } catch (IOException ex) {
        // The Bloom filter is broken, turn it off.
        throw new IllegalArgumentException(
            "Failed to load Bloom block for key "
                + Bytes.toStringBinary(key, keyOffset, keyLength), ex);
      }


      ByteBuffer bloomBuf = bloomBlock.getBufferReadOnly();
      result = ByteBloomFilter.contains(key, keyOffset, keyLength,
          bloomBuf.array(), bloomBuf.arrayOffset() + HFileBlock.HEADER_SIZE,
          bloomBlock.getUncompressedSizeWithoutHeader(), hash, hashCount);
    }


    if (numQueriesPerChunk != null && block >= 0) {
      // Update statistics. Only used in unit tests.
      ++numQueriesPerChunk[block];

View Full Code Here

      // Open a scanner with (on read) caching disabled
      HFileScanner scanner = reader.getScanner(false, false);
      assertTrue(testDescription, scanner.seekTo());
      // Cribbed from io.hfile.TestCacheOnWrite
      long offset = 0;
      HFileBlock prevBlock = null;
      while (offset < reader.getTrailer().getLoadOnOpenDataOffset()) {
        long onDiskSize = -1;
        if (prevBlock != null) {
          onDiskSize = prevBlock.getNextBlockOnDiskSizeWithHeader();
        }
        // Flags: don't cache the block, use pread, this is not a compaction.
        // Also, pass null for expected block type to avoid checking it.
        HFileBlock block = reader.readBlock(offset, onDiskSize, false, true,
          false, null);
        BlockCacheKey blockCacheKey = new BlockCacheKey(reader.getName(),
          offset);
        boolean isCached = cache.getBlock(blockCacheKey, true, false) != null;
        boolean shouldBeCached = cowType.shouldBeCached(block.getBlockType());
        if (shouldBeCached != isCached) {
          throw new AssertionError(
            "shouldBeCached: " + shouldBeCached+ "\n" +
            "isCached: " + isCached + "\n" +
            "Test description: " + testDescription + "\n" +
            "block: " + block + "\n" +
            "blockCacheKey: " + blockCacheKey);
        }
        prevBlock = block;
        offset += block.getOnDiskSizeWithHeader();
      }
    } finally {
      reader.close();
    }
  }

View Full Code Here


    int block = index.rootBlockContainingKey(key, keyOffset, keyLength);
    if (block < 0) {
      result = false; // This key is not in the file.
    } else {
      HFileBlock bloomBlock;
      try {
        // We cache the block and use a positional read.
        bloomBlock = reader.readBlock(index.getRootBlockOffset(block),
            index.getRootBlockDataSize(block), true, true, false, true,
            BlockType.BLOOM_CHUNK);
      } catch (IOException ex) {
        // The Bloom filter is broken, turn it off.
        throw new IllegalArgumentException(
            "Failed to load Bloom block for key "
                + Bytes.toStringBinary(key, keyOffset, keyLength), ex);
      }


      ByteBuffer bloomBuf = bloomBlock.getBufferReadOnly();
      result = ByteBloomFilter.contains(key, keyOffset, keyLength,
          bloomBuf.array(), bloomBuf.arrayOffset() + bloomBlock.headerSize(),
          bloomBlock.getUncompressedSizeWithoutHeader(), hash, hashCount);
    }


    if (numQueriesPerChunk != null && block >= 0) {
      // Update statistics. Only used in unit tests.
      ++numQueriesPerChunk[block];

View Full Code Here

    int block = index.rootBlockContainingKey(key, keyOffset,
        keyLength);
    if (block < 0) {
      result = false; // This key is not in the file.
    } else {
      HFileBlock bloomBlock;
      try {
        // We cache the block and use a positional read.
        bloomBlock = reader.readBlock(index.getRootBlockOffset(block),
            index.getRootBlockDataSize(block), true, true, false, true,
            BlockType.BLOOM_CHUNK, null);
      } catch (IOException ex) {
        // The Bloom filter is broken, turn it off.
        throw new IllegalArgumentException(
            "Failed to load Bloom block for key "
                + Bytes.toStringBinary(key, keyOffset, keyLength), ex);
      }


      ByteBuffer bloomBuf = bloomBlock.getBufferReadOnly();
      result = ByteBloomFilter.contains(key, keyOffset, keyLength,
          bloomBuf.array(), bloomBuf.arrayOffset() + bloomBlock.headerSize(),
          bloomBlock.getUncompressedSizeWithoutHeader(), hash, hashCount);
    }


    if (numQueriesPerChunk != null && block >= 0) {
      // Update statistics. Only used in unit tests.
      ++numQueriesPerChunk[block];

View Full Code Here

      // Open a scanner with (on read) caching disabled
      HFileScanner scanner = reader.getScanner(false, false);
      assertTrue(testDescription, scanner.seekTo());
      // Cribbed from io.hfile.TestCacheOnWrite
      long offset = 0;
      HFileBlock prevBlock = null;
      while (offset < reader.getTrailer().getLoadOnOpenDataOffset()) {
        long onDiskSize = -1;
        if (prevBlock != null) {
          onDiskSize = prevBlock.getNextBlockOnDiskSizeWithHeader();
        }
        // Flags: don't cache the block, use pread, this is not a compaction.
        // Also, pass null for expected block type to avoid checking it.
        HFileBlock block = reader.readBlock(offset, onDiskSize, false, true,
          false, true, null, DataBlockEncoding.NONE);
        BlockCacheKey blockCacheKey = new BlockCacheKey(reader.getName(),
          offset);
        boolean isCached = cache.getBlock(blockCacheKey, true, false, true) != null;
        boolean shouldBeCached = cowType.shouldBeCached(block.getBlockType());
        if (shouldBeCached != isCached) {
          throw new AssertionError(
            "shouldBeCached: " + shouldBeCached+ "\n" +
            "isCached: " + isCached + "\n" +
            "Test description: " + testDescription + "\n" +
            "block: " + block + "\n" +
            "blockCacheKey: " + blockCacheKey);
        }
        prevBlock = block;
        offset += block.getOnDiskSizeWithHeader();
      }
    } finally {
      reader.close();
    }
  }

View Full Code Here

0 1 2

TOP

Related Classes of org.apache.hadoop.hbase.io.hfile.HFileBlock

org.apache.hadoop.fs.FSDataInputStream

org.apache.hadoop.hbase.io.ByteBufferInputStream

org.apache.hadoop.hbase.io.compress.Compression.Algorithm

org.apache.hadoop.hbase.io.DoubleOutputStream

org.apache.hadoop.hbase.io.encoding.HFileBlockDecodingContext

org.apache.hadoop.hbase.io.encoding.HFileBlockDefaultDecodingContext

org.apache.hadoop.hbase.io.encoding.HFileBlockDefaultEncodingContext

org.apache.hadoop.hbase.io.FSDataInputStreamWrapper

org.apache.hadoop.hbase.io.hfile.bucket.BucketCache$RAMQueueEntry

org.apache.hadoop.hbase.io.hfile.HFileDataBlockEncoderImpl

All source code are property of their respective owners. Java is a trademark of Sun Microsystems, Inc and owned by ORACLE Inc. Contact coftware#gmail.com.