yihua commented on code in PR #12866:
URL: https://github.com/apache/hudi/pull/12866#discussion_r2105491554
##########
hudi-io/src/main/java/org/apache/hudi/io/hfile/HFileBlock.java:
##########
@@ -187,34 +162,161 @@ public int getOnDiskSizeWithHeader() {
* @throws IOException upon decoding and decompression error.
*/
public void unpack() throws IOException {
- if (!isUnpacked) {
+ if (!readAttributesOpt.get().isUnpacked) {
// Should only be called for compressed blocks
CompressionCodec compression = context.getCompressionCodec();
if (compression != CompressionCodec.NONE) {
// Copy the block header which is not compressed
System.arraycopy(
- compressedByteBuff, startOffsetInCompressedBuff, byteBuff, 0,
HFILEBLOCK_HEADER_SIZE);
+ readAttributesOpt.get().compressedByteBuff,
+ readAttributesOpt.get().startOffsetInCompressedBuff,
+ readAttributesOpt.get().byteBuff,
+ 0,
+ HFILEBLOCK_HEADER_SIZE);
try (InputStream byteBuffInputStream = new ByteArrayInputStream(
- compressedByteBuff, startOffsetInCompressedBuff +
HFILEBLOCK_HEADER_SIZE, onDiskSizeWithoutHeader)) {
+ readAttributesOpt.get().compressedByteBuff,
+ readAttributesOpt.get().startOffsetInCompressedBuff +
HFILEBLOCK_HEADER_SIZE,
+ readAttributesOpt.get().onDiskSizeWithoutHeader)) {
context.getDecompressor().decompress(
byteBuffInputStream,
- byteBuff,
+ readAttributesOpt.get().byteBuff,
HFILEBLOCK_HEADER_SIZE,
- byteBuff.length - HFILEBLOCK_HEADER_SIZE);
+ readAttributesOpt.get().byteBuff.length -
HFILEBLOCK_HEADER_SIZE);
}
}
- isUnpacked = true;
+ readAttributesOpt.get().isUnpacked = true;
}
}
+ // ================ Below are for Write ================
+
/**
- * Allocates new byte buffer for the uncompressed bytes.
- *
- * @return a new byte array based on the size of uncompressed data, holding
the same header
- * bytes.
+ * Returns serialized "data" part of the block.
+ * This function should be implemented by each block type separately.
+ * By default, it does nothing.
+ */
+ public ByteBuffer getPayload() {
+ return ByteBuffer.allocate(0);
+ }
+
+ /**
+ * Return serialized block including header, data, checksum.
+ */
+ public ByteBuffer serialize() throws IOException {
+ // Block payload.
+ ByteBuffer payloadBuff = getPayload();
+ // Compress if specified.
+ ByteBuffer compressedPayload = compress(payloadBuff);
+ // Buffer for building block.
+ ByteBuffer buf = ByteBuffer.allocate(Math.max(
+ context.getBlockSize() * 2,
+ compressedPayload.limit() + HFILEBLOCK_HEADER_SIZE * 2));
+ // Block header
+ // 1. Magic is always 8 bytes.
+ buf.put(blockType.getMagic(), 0, 8);
+ // 2. onDiskSizeWithoutHeader.
+ buf.putInt(compressedPayload.limit());
+ // 3. uncompressedSizeWithoutHeader.
+ buf.putInt(payloadBuff.limit());
+ // 4. previous block offset.
+ buf.putLong(writeAttributesOpt.get().previousBlockOffset);
+ // TODO: set checksum type properly.
+ // 5. checksum type.
+ buf.put(CHECKSUM_TYPE.getCode());
+ // TODO: verify that if hudi uses 4 bytes for checksum always.
+ // 6. bytes covered per checksum.
+ buf.putInt(DEFAULT_BYTES_PER_CHECKSUM);
+ // 7. onDiskDataSizeWithHeader
+ int onDiskDataSizeWithHeader =
+ HFileBlock.HFILEBLOCK_HEADER_SIZE + payloadBuff.limit();
+ buf.putInt(onDiskDataSizeWithHeader);
+ // 8. payload.
+ buf.put(compressedPayload);
+ // 9. Checksum
+ buf.put(calcChecksumBytes(CHECKSUM_TYPE));
+
+ // Update sizes
+ buf.flip();
+ return buf;
+ }
+
+ protected ByteBuffer compress(ByteBuffer payload) throws IOException {
+ if (context.getCompressionCodec() == GZIP) {
+ byte[] temp = new byte[payload.remaining()];
+ payload.get(temp);
+ return ByteBuffer.wrap(new
HoodieAirliftGzipDecompressor().compress(temp));
+ } else {
+ return payload;
+ }
+ }
+
+ // TODO: support non-NULL checksum types.
+ /**
+ * Returns checksum bytes if checksum type is not NULL.
+ */
+ private byte[] calcChecksumBytes(ChecksumType type) {
+ if (type == ChecksumType.NULL) {
+ return EMPTY_BYTE_ARRAY;
+ } else if (type == ChecksumType.CRC32) {
+ return EMPTY_BYTE_ARRAY;
+ } else {
+ return EMPTY_BYTE_ARRAY;
+ }
+ }
+
+ /**
+ * Sets start offset of the block in the buffer.
+ */
+ public void setStartOffsetInBuff(long startOffsetInBuff) {
+ this.writeAttributesOpt.get().startOffsetInBuff = startOffsetInBuff;
+ }
+
+ /**
+ * Gets start offset of the block in the buffer.
+ */
+ public long getStartOffsetInBuff() {
+ return this.writeAttributesOpt.get().startOffsetInBuff;
+ }
+
+ /**
+ * Returns the number of bytes that should be used by checksum.
+ * @param onDiskBlockBytesWithHeaderSize
+ * @param bytesPerChecksum
+ * @return
+ */
+ private long calcNumChecksumBytes(int onDiskBlockBytesWithHeaderSize, int
bytesPerChecksum) {
+ return numBytes(onDiskBlockBytesWithHeaderSize, bytesPerChecksum);
+ }
+
+ /**
+ * Returns the number of bytes needed to store the checksums for a specified
data size
+ * @param datasize number of bytes of data
+ * @param bytesPerChecksum number of bytes in a checksum chunk
+ * @return The number of bytes needed to store the checksum values
*/
- protected byte[] allocateBufferForUnpacking() {
- int capacity = HFILEBLOCK_HEADER_SIZE + uncompressedSizeWithoutHeader +
sizeCheckSum;
- return new byte[capacity];
+ static long numBytes(long datasize, int bytesPerChecksum) {
+ return numChunks(datasize, bytesPerChecksum) * HFileBlock.CHECKSUM_SIZE;
+ }
+
+ /**
+ * Returns the number of checksum chunks needed to store the checksums for a
specified data size
+ * @param datasize number of bytes of data
+ * @param bytesPerChecksum number of bytes in a checksum chunk
+ * @return The number of checksum chunks
+ */
+ static long numChunks(long datasize, int bytesPerChecksum) {
+ long numChunks = datasize / bytesPerChecksum;
+ if (datasize % bytesPerChecksum != 0) {
+ numChunks++;
+ }
+ return numChunks;
+ }
+
+ static byte[] getVariableLengthEncodes(int length) throws IOException {
Review Comment:
This writes `uint32` not `varint` correct?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
