http://git-wip-us.apache.org/repos/asf/hbase-site/blob/c9d35424/devapidocs/src-html/org/apache/hadoop/hbase/io/hfile/HFileBlock.Writer.html ---------------------------------------------------------------------- diff --git a/devapidocs/src-html/org/apache/hadoop/hbase/io/hfile/HFileBlock.Writer.html b/devapidocs/src-html/org/apache/hadoop/hbase/io/hfile/HFileBlock.Writer.html index 6414009..2ad3a12 100644 --- a/devapidocs/src-html/org/apache/hadoop/hbase/io/hfile/HFileBlock.Writer.html +++ b/devapidocs/src-html/org/apache/hadoop/hbase/io/hfile/HFileBlock.Writer.html @@ -25,42 +25,42 @@ <span class="sourceLineNo">017</span> */<a name="line.17"></a> <span class="sourceLineNo">018</span>package org.apache.hadoop.hbase.io.hfile;<a name="line.18"></a> <span class="sourceLineNo">019</span><a name="line.19"></a> -<span class="sourceLineNo">020</span>import com.google.common.annotations.VisibleForTesting;<a name="line.20"></a> -<span class="sourceLineNo">021</span>import com.google.common.base.Preconditions;<a name="line.21"></a> -<span class="sourceLineNo">022</span><a name="line.22"></a> -<span class="sourceLineNo">023</span>import java.io.DataInputStream;<a name="line.23"></a> -<span class="sourceLineNo">024</span>import java.io.DataOutput;<a name="line.24"></a> -<span class="sourceLineNo">025</span>import java.io.DataOutputStream;<a name="line.25"></a> -<span class="sourceLineNo">026</span>import java.io.IOException;<a name="line.26"></a> -<span class="sourceLineNo">027</span>import java.io.InputStream;<a name="line.27"></a> -<span class="sourceLineNo">028</span>import java.nio.ByteBuffer;<a name="line.28"></a> -<span class="sourceLineNo">029</span>import java.util.concurrent.atomic.AtomicReference;<a name="line.29"></a> -<span class="sourceLineNo">030</span><a name="line.30"></a> -<span class="sourceLineNo">031</span>import org.apache.commons.logging.Log;<a name="line.31"></a> -<span class="sourceLineNo">032</span>import org.apache.commons.logging.LogFactory;<a name="line.32"></a> -<span class="sourceLineNo">033</span>import org.apache.hadoop.fs.FSDataInputStream;<a name="line.33"></a> -<span class="sourceLineNo">034</span>import org.apache.hadoop.fs.FSDataOutputStream;<a name="line.34"></a> -<span class="sourceLineNo">035</span>import org.apache.hadoop.fs.Path;<a name="line.35"></a> -<span class="sourceLineNo">036</span>import org.apache.hadoop.hbase.Cell;<a name="line.36"></a> -<span class="sourceLineNo">037</span>import org.apache.hadoop.hbase.HConstants;<a name="line.37"></a> -<span class="sourceLineNo">038</span>import org.apache.hadoop.hbase.classification.InterfaceAudience;<a name="line.38"></a> -<span class="sourceLineNo">039</span>import org.apache.hadoop.hbase.fs.HFileSystem;<a name="line.39"></a> -<span class="sourceLineNo">040</span>import org.apache.hadoop.hbase.io.ByteArrayOutputStream;<a name="line.40"></a> -<span class="sourceLineNo">041</span>import org.apache.hadoop.hbase.io.ByteBuffInputStream;<a name="line.41"></a> -<span class="sourceLineNo">042</span>import org.apache.hadoop.hbase.io.ByteBufferWriterDataOutputStream;<a name="line.42"></a> -<span class="sourceLineNo">043</span>import org.apache.hadoop.hbase.io.FSDataInputStreamWrapper;<a name="line.43"></a> -<span class="sourceLineNo">044</span>import org.apache.hadoop.hbase.io.encoding.DataBlockEncoding;<a name="line.44"></a> -<span class="sourceLineNo">045</span>import org.apache.hadoop.hbase.io.encoding.HFileBlockDecodingContext;<a name="line.45"></a> -<span class="sourceLineNo">046</span>import org.apache.hadoop.hbase.io.encoding.HFileBlockDefaultDecodingContext;<a name="line.46"></a> -<span class="sourceLineNo">047</span>import org.apache.hadoop.hbase.io.encoding.HFileBlockDefaultEncodingContext;<a name="line.47"></a> -<span class="sourceLineNo">048</span>import org.apache.hadoop.hbase.io.encoding.HFileBlockEncodingContext;<a name="line.48"></a> -<span class="sourceLineNo">049</span>import org.apache.hadoop.hbase.nio.ByteBuff;<a name="line.49"></a> -<span class="sourceLineNo">050</span>import org.apache.hadoop.hbase.nio.MultiByteBuff;<a name="line.50"></a> -<span class="sourceLineNo">051</span>import org.apache.hadoop.hbase.nio.SingleByteBuff;<a name="line.51"></a> -<span class="sourceLineNo">052</span>import org.apache.hadoop.hbase.util.Bytes;<a name="line.52"></a> -<span class="sourceLineNo">053</span>import org.apache.hadoop.hbase.util.ChecksumType;<a name="line.53"></a> -<span class="sourceLineNo">054</span>import org.apache.hadoop.hbase.util.ClassSize;<a name="line.54"></a> -<span class="sourceLineNo">055</span>import org.apache.hadoop.io.IOUtils;<a name="line.55"></a> +<span class="sourceLineNo">020</span>import java.io.DataInputStream;<a name="line.20"></a> +<span class="sourceLineNo">021</span>import java.io.DataOutput;<a name="line.21"></a> +<span class="sourceLineNo">022</span>import java.io.DataOutputStream;<a name="line.22"></a> +<span class="sourceLineNo">023</span>import java.io.IOException;<a name="line.23"></a> +<span class="sourceLineNo">024</span>import java.io.InputStream;<a name="line.24"></a> +<span class="sourceLineNo">025</span>import java.nio.ByteBuffer;<a name="line.25"></a> +<span class="sourceLineNo">026</span>import java.util.concurrent.atomic.AtomicReference;<a name="line.26"></a> +<span class="sourceLineNo">027</span><a name="line.27"></a> +<span class="sourceLineNo">028</span>import org.apache.commons.logging.Log;<a name="line.28"></a> +<span class="sourceLineNo">029</span>import org.apache.commons.logging.LogFactory;<a name="line.29"></a> +<span class="sourceLineNo">030</span>import org.apache.hadoop.fs.FSDataInputStream;<a name="line.30"></a> +<span class="sourceLineNo">031</span>import org.apache.hadoop.fs.FSDataOutputStream;<a name="line.31"></a> +<span class="sourceLineNo">032</span>import org.apache.hadoop.fs.Path;<a name="line.32"></a> +<span class="sourceLineNo">033</span>import org.apache.hadoop.hbase.Cell;<a name="line.33"></a> +<span class="sourceLineNo">034</span>import org.apache.hadoop.hbase.HConstants;<a name="line.34"></a> +<span class="sourceLineNo">035</span>import org.apache.hadoop.hbase.classification.InterfaceAudience;<a name="line.35"></a> +<span class="sourceLineNo">036</span>import org.apache.hadoop.hbase.fs.HFileSystem;<a name="line.36"></a> +<span class="sourceLineNo">037</span>import org.apache.hadoop.hbase.io.ByteArrayOutputStream;<a name="line.37"></a> +<span class="sourceLineNo">038</span>import org.apache.hadoop.hbase.io.ByteBuffInputStream;<a name="line.38"></a> +<span class="sourceLineNo">039</span>import org.apache.hadoop.hbase.io.ByteBufferWriterDataOutputStream;<a name="line.39"></a> +<span class="sourceLineNo">040</span>import org.apache.hadoop.hbase.io.FSDataInputStreamWrapper;<a name="line.40"></a> +<span class="sourceLineNo">041</span>import org.apache.hadoop.hbase.io.encoding.DataBlockEncoding;<a name="line.41"></a> +<span class="sourceLineNo">042</span>import org.apache.hadoop.hbase.io.encoding.HFileBlockDecodingContext;<a name="line.42"></a> +<span class="sourceLineNo">043</span>import org.apache.hadoop.hbase.io.encoding.HFileBlockDefaultDecodingContext;<a name="line.43"></a> +<span class="sourceLineNo">044</span>import org.apache.hadoop.hbase.io.encoding.HFileBlockDefaultEncodingContext;<a name="line.44"></a> +<span class="sourceLineNo">045</span>import org.apache.hadoop.hbase.io.encoding.HFileBlockEncodingContext;<a name="line.45"></a> +<span class="sourceLineNo">046</span>import org.apache.hadoop.hbase.nio.ByteBuff;<a name="line.46"></a> +<span class="sourceLineNo">047</span>import org.apache.hadoop.hbase.nio.MultiByteBuff;<a name="line.47"></a> +<span class="sourceLineNo">048</span>import org.apache.hadoop.hbase.nio.SingleByteBuff;<a name="line.48"></a> +<span class="sourceLineNo">049</span>import org.apache.hadoop.hbase.util.Bytes;<a name="line.49"></a> +<span class="sourceLineNo">050</span>import org.apache.hadoop.hbase.util.ChecksumType;<a name="line.50"></a> +<span class="sourceLineNo">051</span>import org.apache.hadoop.hbase.util.ClassSize;<a name="line.51"></a> +<span class="sourceLineNo">052</span>import org.apache.hadoop.io.IOUtils;<a name="line.52"></a> +<span class="sourceLineNo">053</span><a name="line.53"></a> +<span class="sourceLineNo">054</span>import com.google.common.annotations.VisibleForTesting;<a name="line.54"></a> +<span class="sourceLineNo">055</span>import com.google.common.base.Preconditions;<a name="line.55"></a> <span class="sourceLineNo">056</span><a name="line.56"></a> <span class="sourceLineNo">057</span>/**<a name="line.57"></a> <span class="sourceLineNo">058</span> * Reads {@link HFile} version 2 blocks to HFiles and via {@link Cacheable} Interface to caches.<a name="line.58"></a> @@ -443,1645 +443,1656 @@ <span class="sourceLineNo">435</span> return nextBlockOnDiskSize;<a name="line.435"></a> <span class="sourceLineNo">436</span> }<a name="line.436"></a> <span class="sourceLineNo">437</span><a name="line.437"></a> -<span class="sourceLineNo">438</span> public BlockType getBlockType() {<a name="line.438"></a> -<span class="sourceLineNo">439</span> return blockType;<a name="line.439"></a> -<span class="sourceLineNo">440</span> }<a name="line.440"></a> -<span class="sourceLineNo">441</span><a name="line.441"></a> -<span class="sourceLineNo">442</span> /** @return get data block encoding id that was used to encode this block */<a name="line.442"></a> -<span class="sourceLineNo">443</span> public short getDataBlockEncodingId() {<a name="line.443"></a> -<span class="sourceLineNo">444</span> if (blockType != BlockType.ENCODED_DATA) {<a name="line.444"></a> -<span class="sourceLineNo">445</span> throw new IllegalArgumentException("Querying encoder ID of a block " +<a name="line.445"></a> -<span class="sourceLineNo">446</span> "of type other than " + BlockType.ENCODED_DATA + ": " + blockType);<a name="line.446"></a> -<span class="sourceLineNo">447</span> }<a name="line.447"></a> -<span class="sourceLineNo">448</span> return buf.getShort(headerSize());<a name="line.448"></a> -<span class="sourceLineNo">449</span> }<a name="line.449"></a> -<span class="sourceLineNo">450</span><a name="line.450"></a> -<span class="sourceLineNo">451</span> /**<a name="line.451"></a> -<span class="sourceLineNo">452</span> * @return the on-disk size of header + data part + checksum.<a name="line.452"></a> -<span class="sourceLineNo">453</span> */<a name="line.453"></a> -<span class="sourceLineNo">454</span> public int getOnDiskSizeWithHeader() {<a name="line.454"></a> -<span class="sourceLineNo">455</span> return onDiskSizeWithoutHeader + headerSize();<a name="line.455"></a> -<span class="sourceLineNo">456</span> }<a name="line.456"></a> -<span class="sourceLineNo">457</span><a name="line.457"></a> -<span class="sourceLineNo">458</span> /**<a name="line.458"></a> -<span class="sourceLineNo">459</span> * @return the on-disk size of the data part + checksum (header excluded).<a name="line.459"></a> -<span class="sourceLineNo">460</span> */<a name="line.460"></a> -<span class="sourceLineNo">461</span> int getOnDiskSizeWithoutHeader() {<a name="line.461"></a> -<span class="sourceLineNo">462</span> return onDiskSizeWithoutHeader;<a name="line.462"></a> -<span class="sourceLineNo">463</span> }<a name="line.463"></a> -<span class="sourceLineNo">464</span><a name="line.464"></a> -<span class="sourceLineNo">465</span> /**<a name="line.465"></a> -<span class="sourceLineNo">466</span> * @return the uncompressed size of data part (header and checksum excluded).<a name="line.466"></a> -<span class="sourceLineNo">467</span> */<a name="line.467"></a> -<span class="sourceLineNo">468</span> int getUncompressedSizeWithoutHeader() {<a name="line.468"></a> -<span class="sourceLineNo">469</span> return uncompressedSizeWithoutHeader;<a name="line.469"></a> -<span class="sourceLineNo">470</span> }<a name="line.470"></a> -<span class="sourceLineNo">471</span><a name="line.471"></a> -<span class="sourceLineNo">472</span> /**<a name="line.472"></a> -<span class="sourceLineNo">473</span> * @return the offset of the previous block of the same type in the file, or<a name="line.473"></a> -<span class="sourceLineNo">474</span> * -1 if unknown<a name="line.474"></a> -<span class="sourceLineNo">475</span> */<a name="line.475"></a> -<span class="sourceLineNo">476</span> long getPrevBlockOffset() {<a name="line.476"></a> -<span class="sourceLineNo">477</span> return prevBlockOffset;<a name="line.477"></a> -<span class="sourceLineNo">478</span> }<a name="line.478"></a> -<span class="sourceLineNo">479</span><a name="line.479"></a> -<span class="sourceLineNo">480</span> /**<a name="line.480"></a> -<span class="sourceLineNo">481</span> * Rewinds {@code buf} and writes first 4 header fields. {@code buf} position<a name="line.481"></a> -<span class="sourceLineNo">482</span> * is modified as side-effect.<a name="line.482"></a> -<span class="sourceLineNo">483</span> */<a name="line.483"></a> -<span class="sourceLineNo">484</span> private void overwriteHeader() {<a name="line.484"></a> -<span class="sourceLineNo">485</span> buf.rewind();<a name="line.485"></a> -<span class="sourceLineNo">486</span> blockType.write(buf);<a name="line.486"></a> -<span class="sourceLineNo">487</span> buf.putInt(onDiskSizeWithoutHeader);<a name="line.487"></a> -<span class="sourceLineNo">488</span> buf.putInt(uncompressedSizeWithoutHeader);<a name="line.488"></a> -<span class="sourceLineNo">489</span> buf.putLong(prevBlockOffset);<a name="line.489"></a> -<span class="sourceLineNo">490</span> if (this.fileContext.isUseHBaseChecksum()) {<a name="line.490"></a> -<span class="sourceLineNo">491</span> buf.put(fileContext.getChecksumType().getCode());<a name="line.491"></a> -<span class="sourceLineNo">492</span> buf.putInt(fileContext.getBytesPerChecksum());<a name="line.492"></a> -<span class="sourceLineNo">493</span> buf.putInt(onDiskDataSizeWithHeader);<a name="line.493"></a> -<span class="sourceLineNo">494</span> }<a name="line.494"></a> -<span class="sourceLineNo">495</span> }<a name="line.495"></a> -<span class="sourceLineNo">496</span><a name="line.496"></a> -<span class="sourceLineNo">497</span> /**<a name="line.497"></a> -<span class="sourceLineNo">498</span> * Returns a buffer that does not include the header or checksum.<a name="line.498"></a> -<span class="sourceLineNo">499</span> *<a name="line.499"></a> -<span class="sourceLineNo">500</span> * @return the buffer with header skipped and checksum omitted.<a name="line.500"></a> -<span class="sourceLineNo">501</span> */<a name="line.501"></a> -<span class="sourceLineNo">502</span> public ByteBuff getBufferWithoutHeader() {<a name="line.502"></a> -<span class="sourceLineNo">503</span> ByteBuff dup = getBufferReadOnly();<a name="line.503"></a> -<span class="sourceLineNo">504</span> // Now set it up so Buffer spans content only -- no header or no checksums.<a name="line.504"></a> -<span class="sourceLineNo">505</span> return dup.position(headerSize()).limit(buf.limit() - totalChecksumBytes()).slice();<a name="line.505"></a> -<span class="sourceLineNo">506</span> }<a name="line.506"></a> -<span class="sourceLineNo">507</span><a name="line.507"></a> -<span class="sourceLineNo">508</span> /**<a name="line.508"></a> -<span class="sourceLineNo">509</span> * Returns a read-only duplicate of the buffer this block stores internally ready to be read.<a name="line.509"></a> -<span class="sourceLineNo">510</span> * Clients must not modify the buffer object though they may set position and limit on the<a name="line.510"></a> -<span class="sourceLineNo">511</span> * returned buffer since we pass back a duplicate. This method has to be public because it is used<a name="line.511"></a> -<span class="sourceLineNo">512</span> * in {@link CompoundBloomFilter} to avoid object creation on every Bloom<a name="line.512"></a> -<span class="sourceLineNo">513</span> * filter lookup, but has to be used with caution. Buffer holds header, block content,<a name="line.513"></a> -<span class="sourceLineNo">514</span> * and any follow-on checksums if present.<a name="line.514"></a> -<span class="sourceLineNo">515</span> *<a name="line.515"></a> -<span class="sourceLineNo">516</span> * @return the buffer of this block for read-only operations<a name="line.516"></a> -<span class="sourceLineNo">517</span> */<a name="line.517"></a> -<span class="sourceLineNo">518</span> public ByteBuff getBufferReadOnly() {<a name="line.518"></a> -<span class="sourceLineNo">519</span> // TODO: ByteBuf does not support asReadOnlyBuffer(). Fix.<a name="line.519"></a> -<span class="sourceLineNo">520</span> ByteBuff dup = this.buf.duplicate();<a name="line.520"></a> -<span class="sourceLineNo">521</span> assert dup.position() == 0;<a name="line.521"></a> -<span class="sourceLineNo">522</span> return dup;<a name="line.522"></a> -<span class="sourceLineNo">523</span> }<a name="line.523"></a> -<span class="sourceLineNo">524</span><a name="line.524"></a> -<span class="sourceLineNo">525</span> private void sanityCheckAssertion(long valueFromBuf, long valueFromField,<a name="line.525"></a> -<span class="sourceLineNo">526</span> String fieldName) throws IOException {<a name="line.526"></a> -<span class="sourceLineNo">527</span> if (valueFromBuf != valueFromField) {<a name="line.527"></a> -<span class="sourceLineNo">528</span> throw new AssertionError(fieldName + " in the buffer (" + valueFromBuf<a name="line.528"></a> -<span class="sourceLineNo">529</span> + ") is different from that in the field (" + valueFromField + ")");<a name="line.529"></a> -<span class="sourceLineNo">530</span> }<a name="line.530"></a> -<span class="sourceLineNo">531</span> }<a name="line.531"></a> -<span class="sourceLineNo">532</span><a name="line.532"></a> -<span class="sourceLineNo">533</span> private void sanityCheckAssertion(BlockType valueFromBuf, BlockType valueFromField)<a name="line.533"></a> -<span class="sourceLineNo">534</span> throws IOException {<a name="line.534"></a> -<span class="sourceLineNo">535</span> if (valueFromBuf != valueFromField) {<a name="line.535"></a> -<span class="sourceLineNo">536</span> throw new IOException("Block type stored in the buffer: " +<a name="line.536"></a> -<span class="sourceLineNo">537</span> valueFromBuf + ", block type field: " + valueFromField);<a name="line.537"></a> -<span class="sourceLineNo">538</span> }<a name="line.538"></a> -<span class="sourceLineNo">539</span> }<a name="line.539"></a> -<span class="sourceLineNo">540</span><a name="line.540"></a> -<span class="sourceLineNo">541</span> /**<a name="line.541"></a> -<span class="sourceLineNo">542</span> * Checks if the block is internally consistent, i.e. the first<a name="line.542"></a> -<span class="sourceLineNo">543</span> * {@link HConstants#HFILEBLOCK_HEADER_SIZE} bytes of the buffer contain a<a name="line.543"></a> -<span class="sourceLineNo">544</span> * valid header consistent with the fields. Assumes a packed block structure.<a name="line.544"></a> -<span class="sourceLineNo">545</span> * This function is primary for testing and debugging, and is not<a name="line.545"></a> -<span class="sourceLineNo">546</span> * thread-safe, because it alters the internal buffer pointer.<a name="line.546"></a> -<span class="sourceLineNo">547</span> * Used by tests only.<a name="line.547"></a> -<span class="sourceLineNo">548</span> */<a name="line.548"></a> -<span class="sourceLineNo">549</span> @VisibleForTesting<a name="line.549"></a> -<span class="sourceLineNo">550</span> void sanityCheck() throws IOException {<a name="line.550"></a> -<span class="sourceLineNo">551</span> // Duplicate so no side-effects<a name="line.551"></a> -<span class="sourceLineNo">552</span> ByteBuff dup = this.buf.duplicate().rewind();<a name="line.552"></a> -<span class="sourceLineNo">553</span> sanityCheckAssertion(BlockType.read(dup), blockType);<a name="line.553"></a> -<span class="sourceLineNo">554</span><a name="line.554"></a> -<span class="sourceLineNo">555</span> sanityCheckAssertion(dup.getInt(), onDiskSizeWithoutHeader, "onDiskSizeWithoutHeader");<a name="line.555"></a> -<span class="sourceLineNo">556</span><a name="line.556"></a> -<span class="sourceLineNo">557</span> sanityCheckAssertion(dup.getInt(), uncompressedSizeWithoutHeader,<a name="line.557"></a> -<span class="sourceLineNo">558</span> "uncompressedSizeWithoutHeader");<a name="line.558"></a> -<span class="sourceLineNo">559</span><a name="line.559"></a> -<span class="sourceLineNo">560</span> sanityCheckAssertion(dup.getLong(), prevBlockOffset, "prevBlockOffset");<a name="line.560"></a> -<span class="sourceLineNo">561</span> if (this.fileContext.isUseHBaseChecksum()) {<a name="line.561"></a> -<span class="sourceLineNo">562</span> sanityCheckAssertion(dup.get(), this.fileContext.getChecksumType().getCode(), "checksumType");<a name="line.562"></a> -<span class="sourceLineNo">563</span> sanityCheckAssertion(dup.getInt(), this.fileContext.getBytesPerChecksum(),<a name="line.563"></a> -<span class="sourceLineNo">564</span> "bytesPerChecksum");<a name="line.564"></a> -<span class="sourceLineNo">565</span> sanityCheckAssertion(dup.getInt(), onDiskDataSizeWithHeader, "onDiskDataSizeWithHeader");<a name="line.565"></a> -<span class="sourceLineNo">566</span> }<a name="line.566"></a> -<span class="sourceLineNo">567</span><a name="line.567"></a> -<span class="sourceLineNo">568</span> int cksumBytes = totalChecksumBytes();<a name="line.568"></a> -<span class="sourceLineNo">569</span> int expectedBufLimit = onDiskDataSizeWithHeader + cksumBytes;<a name="line.569"></a> -<span class="sourceLineNo">570</span> if (dup.limit() != expectedBufLimit) {<a name="line.570"></a> -<span class="sourceLineNo">571</span> throw new AssertionError("Expected limit " + expectedBufLimit + ", got " + dup.limit());<a name="line.571"></a> -<span class="sourceLineNo">572</span> }<a name="line.572"></a> -<span class="sourceLineNo">573</span><a name="line.573"></a> -<span class="sourceLineNo">574</span> // We might optionally allocate HFILEBLOCK_HEADER_SIZE more bytes to read the next<a name="line.574"></a> -<span class="sourceLineNo">575</span> // block's header, so there are two sensible values for buffer capacity.<a name="line.575"></a> -<span class="sourceLineNo">576</span> int hdrSize = headerSize();<a name="line.576"></a> -<span class="sourceLineNo">577</span> if (dup.capacity() != expectedBufLimit && dup.capacity() != expectedBufLimit + hdrSize) {<a name="line.577"></a> -<span class="sourceLineNo">578</span> throw new AssertionError("Invalid buffer capacity: " + dup.capacity() +<a name="line.578"></a> -<span class="sourceLineNo">579</span> ", expected " + expectedBufLimit + " or " + (expectedBufLimit + hdrSize));<a name="line.579"></a> -<span class="sourceLineNo">580</span> }<a name="line.580"></a> -<span class="sourceLineNo">581</span> }<a name="line.581"></a> -<span class="sourceLineNo">582</span><a name="line.582"></a> -<span class="sourceLineNo">583</span> @Override<a name="line.583"></a> -<span class="sourceLineNo">584</span> public String toString() {<a name="line.584"></a> -<span class="sourceLineNo">585</span> StringBuilder sb = new StringBuilder()<a name="line.585"></a> -<span class="sourceLineNo">586</span> .append("[")<a name="line.586"></a> -<span class="sourceLineNo">587</span> .append("blockType=").append(blockType)<a name="line.587"></a> -<span class="sourceLineNo">588</span> .append(", fileOffset=").append(offset)<a name="line.588"></a> -<span class="sourceLineNo">589</span> .append(", headerSize=").append(headerSize())<a name="line.589"></a> -<span class="sourceLineNo">590</span> .append(", onDiskSizeWithoutHeader=").append(onDiskSizeWithoutHeader)<a name="line.590"></a> -<span class="sourceLineNo">591</span> .append(", uncompressedSizeWithoutHeader=").append(uncompressedSizeWithoutHeader)<a name="line.591"></a> -<span class="sourceLineNo">592</span> .append(", prevBlockOffset=").append(prevBlockOffset)<a name="line.592"></a> -<span class="sourceLineNo">593</span> .append(", isUseHBaseChecksum=").append(fileContext.isUseHBaseChecksum());<a name="line.593"></a> -<span class="sourceLineNo">594</span> if (fileContext.isUseHBaseChecksum()) {<a name="line.594"></a> -<span class="sourceLineNo">595</span> sb.append(", checksumType=").append(ChecksumType.codeToType(this.buf.get(24)))<a name="line.595"></a> -<span class="sourceLineNo">596</span> .append(", bytesPerChecksum=").append(this.buf.getInt(24 + 1))<a name="line.596"></a> -<span class="sourceLineNo">597</span> .append(", onDiskDataSizeWithHeader=").append(onDiskDataSizeWithHeader);<a name="line.597"></a> -<span class="sourceLineNo">598</span> } else {<a name="line.598"></a> -<span class="sourceLineNo">599</span> sb.append(", onDiskDataSizeWithHeader=").append(onDiskDataSizeWithHeader)<a name="line.599"></a> -<span class="sourceLineNo">600</span> .append("(").append(onDiskSizeWithoutHeader)<a name="line.600"></a> -<span class="sourceLineNo">601</span> .append("+").append(HConstants.HFILEBLOCK_HEADER_SIZE_NO_CHECKSUM).append(")");<a name="line.601"></a> -<span class="sourceLineNo">602</span> }<a name="line.602"></a> -<span class="sourceLineNo">603</span> String dataBegin = null;<a name="line.603"></a> -<span class="sourceLineNo">604</span> if (buf.hasArray()) {<a name="line.604"></a> -<span class="sourceLineNo">605</span> dataBegin = Bytes.toStringBinary(buf.array(), buf.arrayOffset() + headerSize(),<a name="line.605"></a> -<span class="sourceLineNo">606</span> Math.min(32, buf.limit() - buf.arrayOffset() - headerSize()));<a name="line.606"></a> -<span class="sourceLineNo">607</span> } else {<a name="line.607"></a> -<span class="sourceLineNo">608</span> ByteBuff bufWithoutHeader = getBufferWithoutHeader();<a name="line.608"></a> -<span class="sourceLineNo">609</span> byte[] dataBeginBytes = new byte[Math.min(32,<a name="line.609"></a> -<span class="sourceLineNo">610</span> bufWithoutHeader.limit() - bufWithoutHeader.position())];<a name="line.610"></a> -<span class="sourceLineNo">611</span> bufWithoutHeader.get(dataBeginBytes);<a name="line.611"></a> -<span class="sourceLineNo">612</span> dataBegin = Bytes.toStringBinary(dataBeginBytes);<a name="line.612"></a> -<span class="sourceLineNo">613</span> }<a name="line.613"></a> -<span class="sourceLineNo">614</span> sb.append(", getOnDiskSizeWithHeader=").append(getOnDiskSizeWithHeader())<a name="line.614"></a> -<span class="sourceLineNo">615</span> .append(", totalChecksumBytes=").append(totalChecksumBytes())<a name="line.615"></a> -<span class="sourceLineNo">616</span> .append(", isUnpacked=").append(isUnpacked())<a name="line.616"></a> -<span class="sourceLineNo">617</span> .append(", buf=[").append(buf).append("]")<a name="line.617"></a> -<span class="sourceLineNo">618</span> .append(", dataBeginsWith=").append(dataBegin)<a name="line.618"></a> -<span class="sourceLineNo">619</span> .append(", fileContext=").append(fileContext)<a name="line.619"></a> -<span class="sourceLineNo">620</span> .append("]");<a name="line.620"></a> -<span class="sourceLineNo">621</span> return sb.toString();<a name="line.621"></a> -<span class="sourceLineNo">622</span> }<a name="line.622"></a> -<span class="sourceLineNo">623</span><a name="line.623"></a> -<span class="sourceLineNo">624</span> /**<a name="line.624"></a> -<span class="sourceLineNo">625</span> * Retrieves the decompressed/decrypted view of this block. An encoded block remains in its<a name="line.625"></a> -<span class="sourceLineNo">626</span> * encoded structure. Internal structures are shared between instances where applicable.<a name="line.626"></a> -<span class="sourceLineNo">627</span> */<a name="line.627"></a> -<span class="sourceLineNo">628</span> HFileBlock unpack(HFileContext fileContext, FSReader reader) throws IOException {<a name="line.628"></a> -<span class="sourceLineNo">629</span> if (!fileContext.isCompressedOrEncrypted()) {<a name="line.629"></a> -<span class="sourceLineNo">630</span> // TODO: cannot use our own fileContext here because HFileBlock(ByteBuffer, boolean),<a name="line.630"></a> -<span class="sourceLineNo">631</span> // which is used for block serialization to L2 cache, does not preserve encoding and<a name="line.631"></a> -<span class="sourceLineNo">632</span> // encryption details.<a name="line.632"></a> -<span class="sourceLineNo">633</span> return this;<a name="line.633"></a> -<span class="sourceLineNo">634</span> }<a name="line.634"></a> -<span class="sourceLineNo">635</span><a name="line.635"></a> -<span class="sourceLineNo">636</span> HFileBlock unpacked = new HFileBlock(this);<a name="line.636"></a> -<span class="sourceLineNo">637</span> unpacked.allocateBuffer(); // allocates space for the decompressed block<a name="line.637"></a> -<span class="sourceLineNo">638</span><a name="line.638"></a> -<span class="sourceLineNo">639</span> HFileBlockDecodingContext ctx = blockType == BlockType.ENCODED_DATA ?<a name="line.639"></a> -<span class="sourceLineNo">640</span> reader.getBlockDecodingContext() : reader.getDefaultBlockDecodingContext();<a name="line.640"></a> -<span class="sourceLineNo">641</span><a name="line.641"></a> -<span class="sourceLineNo">642</span> ByteBuff dup = this.buf.duplicate();<a name="line.642"></a> -<span class="sourceLineNo">643</span> dup.position(this.headerSize());<a name="line.643"></a> -<span class="sourceLineNo">644</span> dup = dup.slice();<a name="line.644"></a> -<span class="sourceLineNo">645</span> ctx.prepareDecoding(unpacked.getOnDiskSizeWithoutHeader(),<a name="line.645"></a> -<span class="sourceLineNo">646</span> unpacked.getUncompressedSizeWithoutHeader(), unpacked.getBufferWithoutHeader(),<a name="line.646"></a> -<span class="sourceLineNo">647</span> dup);<a name="line.647"></a> -<span class="sourceLineNo">648</span> return unpacked;<a name="line.648"></a> -<span class="sourceLineNo">649</span> }<a name="line.649"></a> -<span class="sourceLineNo">650</span><a name="line.650"></a> -<span class="sourceLineNo">651</span> /**<a name="line.651"></a> -<span class="sourceLineNo">652</span> * Always allocates a new buffer of the correct size. Copies header bytes<a name="line.652"></a> -<span class="sourceLineNo">653</span> * from the existing buffer. Does not change header fields.<a name="line.653"></a> -<span class="sourceLineNo">654</span> * Reserve room to keep checksum bytes too.<a name="line.654"></a> -<span class="sourceLineNo">655</span> */<a name="line.655"></a> -<span class="sourceLineNo">656</span> private void allocateBuffer() {<a name="line.656"></a> -<span class="sourceLineNo">657</span> int cksumBytes = totalChecksumBytes();<a name="line.657"></a> -<span class="sourceLineNo">658</span> int headerSize = headerSize();<a name="line.658"></a> -<span class="sourceLineNo">659</span> int capacityNeeded = headerSize + uncompressedSizeWithoutHeader + cksumBytes;<a name="line.659"></a> -<span class="sourceLineNo">660</span><a name="line.660"></a> -<span class="sourceLineNo">661</span> // TODO we need consider allocating offheap here?<a name="line.661"></a> -<span class="sourceLineNo">662</span> ByteBuffer newBuf = ByteBuffer.allocate(capacityNeeded);<a name="line.662"></a> -<span class="sourceLineNo">663</span><a name="line.663"></a> -<span class="sourceLineNo">664</span> // Copy header bytes into newBuf.<a name="line.664"></a> -<span class="sourceLineNo">665</span> // newBuf is HBB so no issue in calling array()<a name="line.665"></a> -<span class="sourceLineNo">666</span> buf.position(0);<a name="line.666"></a> -<span class="sourceLineNo">667</span> buf.get(newBuf.array(), newBuf.arrayOffset(), headerSize);<a name="line.667"></a> -<span class="sourceLineNo">668</span><a name="line.668"></a> -<span class="sourceLineNo">669</span> buf = new SingleByteBuff(newBuf);<a name="line.669"></a> -<span class="sourceLineNo">670</span> // set limit to exclude next block's header<a name="line.670"></a> -<span class="sourceLineNo">671</span> buf.limit(headerSize + uncompressedSizeWithoutHeader + cksumBytes);<a name="line.671"></a> -<span class="sourceLineNo">672</span> }<a name="line.672"></a> -<span class="sourceLineNo">673</span><a name="line.673"></a> -<span class="sourceLineNo">674</span> /**<a name="line.674"></a> -<span class="sourceLineNo">675</span> * Return true when this block's buffer has been unpacked, false otherwise. Note this is a<a name="line.675"></a> -<span class="sourceLineNo">676</span> * calculated heuristic, not tracked attribute of the block.<a name="line.676"></a> -<span class="sourceLineNo">677</span> */<a name="line.677"></a> -<span class="sourceLineNo">678</span> public boolean isUnpacked() {<a name="line.678"></a> -<span class="sourceLineNo">679</span> final int cksumBytes = totalChecksumBytes();<a name="line.679"></a> -<span class="sourceLineNo">680</span> final int headerSize = headerSize();<a name="line.680"></a> -<span class="sourceLineNo">681</span> final int expectedCapacity = headerSize + uncompressedSizeWithoutHeader + cksumBytes;<a name="line.681"></a> -<span class="sourceLineNo">682</span> final int bufCapacity = buf.capacity();<a name="line.682"></a> -<span class="sourceLineNo">683</span> return bufCapacity == expectedCapacity || bufCapacity == expectedCapacity + headerSize;<a name="line.683"></a> -<span class="sourceLineNo">684</span> }<a name="line.684"></a> -<span class="sourceLineNo">685</span><a name="line.685"></a> -<span class="sourceLineNo">686</span> /** An additional sanity-check in case no compression or encryption is being used. */<a name="line.686"></a> -<span class="sourceLineNo">687</span> public void sanityCheckUncompressedSize() throws IOException {<a name="line.687"></a> -<span class="sourceLineNo">688</span> if (onDiskSizeWithoutHeader != uncompressedSizeWithoutHeader + totalChecksumBytes()) {<a name="line.688"></a> -<span class="sourceLineNo">689</span> throw new IOException("Using no compression but "<a name="line.689"></a> -<span class="sourceLineNo">690</span> + "onDiskSizeWithoutHeader=" + onDiskSizeWithoutHeader + ", "<a name="line.690"></a> -<span class="sourceLineNo">691</span> + "uncompressedSizeWithoutHeader=" + uncompressedSizeWithoutHeader<a name="line.691"></a> -<span class="sourceLineNo">692</span> + ", numChecksumbytes=" + totalChecksumBytes());<a name="line.692"></a> -<span class="sourceLineNo">693</span> }<a name="line.693"></a> -<span class="sourceLineNo">694</span> }<a name="line.694"></a> -<span class="sourceLineNo">695</span><a name="line.695"></a> -<span class="sourceLineNo">696</span> /**<a name="line.696"></a> -<span class="sourceLineNo">697</span> * Cannot be {@link #UNSET}. Must be a legitimate value. Used re-making the {@link BlockCacheKey} when<a name="line.697"></a> -<span class="sourceLineNo">698</span> * block is returned to the cache.<a name="line.698"></a> -<span class="sourceLineNo">699</span> * @return the offset of this block in the file it was read from<a name="line.699"></a> -<span class="sourceLineNo">700</span> */<a name="line.700"></a> -<span class="sourceLineNo">701</span> long getOffset() {<a name="line.701"></a> -<span class="sourceLineNo">702</span> if (offset < 0) {<a name="line.702"></a> -<span class="sourceLineNo">703</span> throw new IllegalStateException("HFile block offset not initialized properly");<a name="line.703"></a> -<span class="sourceLineNo">704</span> }<a name="line.704"></a> -<span class="sourceLineNo">705</span> return offset;<a name="line.705"></a> -<span class="sourceLineNo">706</span> }<a name="line.706"></a> -<span class="sourceLineNo">707</span><a name="line.707"></a> -<span class="sourceLineNo">708</span> /**<a name="line.708"></a> -<span class="sourceLineNo">709</span> * @return a byte stream reading the data + checksum of this block<a name="line.709"></a> -<span class="sourceLineNo">710</span> */<a name="line.710"></a> -<span class="sourceLineNo">711</span> DataInputStream getByteStream() {<a name="line.711"></a> -<span class="sourceLineNo">712</span> ByteBuff dup = this.buf.duplicate();<a name="line.712"></a> -<span class="sourceLineNo">713</span> dup.position(this.headerSize());<a name="line.713"></a> -<span class="sourceLineNo">714</span> return new DataInputStream(new ByteBuffInputStream(dup));<a name="line.714"></a> -<span class="sourceLineNo">715</span> }<a name="line.715"></a> -<span class="sourceLineNo">716</span><a name="line.716"></a> -<span class="sourceLineNo">717</span> @Override<a name="line.717"></a> -<span class="sourceLineNo">718</span> public long heapSize() {<a name="line.718"></a> -<span class="sourceLineNo">719</span> long size = ClassSize.align(<a name="line.719"></a> -<span class="sourceLineNo">720</span> ClassSize.OBJECT +<a name="line.720"></a> -<span class="sourceLineNo">721</span> // Block type, multi byte buffer, MemoryType and meta references<a name="line.721"></a> -<span class="sourceLineNo">722</span> 4 * ClassSize.REFERENCE +<a name="line.722"></a> -<span class="sourceLineNo">723</span> // On-disk size, uncompressed size, and next block's on-disk size<a name="line.723"></a> -<span class="sourceLineNo">724</span> // bytePerChecksum and onDiskDataSize<a name="line.724"></a> -<span class="sourceLineNo">725</span> 4 * Bytes.SIZEOF_INT +<a name="line.725"></a> -<span class="sourceLineNo">726</span> // This and previous block offset<a name="line.726"></a> -<span class="sourceLineNo">727</span> 2 * Bytes.SIZEOF_LONG +<a name="line.727"></a> -<span class="sourceLineNo">728</span> // Heap size of the meta object. meta will be always not null.<a name="line.728"></a> -<span class="sourceLineNo">729</span> fileContext.heapSize()<a name="line.729"></a> -<span class="sourceLineNo">730</span> );<a name="line.730"></a> -<span class="sourceLineNo">731</span><a name="line.731"></a> -<span class="sourceLineNo">732</span> if (buf != null) {<a name="line.732"></a> -<span class="sourceLineNo">733</span> // Deep overhead of the byte buffer. Needs to be aligned separately.<a name="line.733"></a> -<span class="sourceLineNo">734</span> size += ClassSize.align(buf.capacity() + MULTI_BYTE_BUFFER_HEAP_SIZE);<a name="line.734"></a> -<span class="sourceLineNo">735</span> }<a name="line.735"></a> -<span class="sourceLineNo">736</span><a name="line.736"></a> -<span class="sourceLineNo">737</span> return ClassSize.align(size);<a name="line.737"></a> -<span class="sourceLineNo">738</span> }<a name="line.738"></a> -<span class="sourceLineNo">739</span><a name="line.739"></a> -<span class="sourceLineNo">740</span> /**<a name="line.740"></a> -<span class="sourceLineNo">741</span> * Read from an input stream at least <code>necessaryLen</code> and if possible,<a name="line.741"></a> -<span class="sourceLineNo">742</span> * <code>extraLen</code> also if available. Analogous to<a name="line.742"></a> -<span class="sourceLineNo">743</span> * {@link IOUtils#readFully(InputStream, byte[], int, int)}, but specifies a<a name="line.743"></a> -<span class="sourceLineNo">744</span> * number of "extra" bytes to also optionally read.<a name="line.744"></a> -<span class="sourceLineNo">745</span> *<a name="line.745"></a> -<span class="sourceLineNo">746</span> * @param in the input stream to read from<a name="line.746"></a> -<span class="sourceLineNo">747</span> * @param buf the buffer to read into<a name="line.747"></a> -<span class="sourceLineNo">748</span> * @param bufOffset the destination offset in the buffer<a name="line.748"></a> -<span class="sourceLineNo">749</span> * @param necessaryLen the number of bytes that are absolutely necessary to read<a name="line.749"></a> -<span class="sourceLineNo">750</span> * @param extraLen the number of extra bytes that would be nice to read<a name="line.750"></a> -<span class="sourceLineNo">751</span> * @return true if succeeded reading the extra bytes<a name="line.751"></a> -<span class="sourceLineNo">752</span> * @throws IOException if failed to read the necessary bytes<a name="line.752"></a> -<span class="sourceLineNo">753</span> */<a name="line.753"></a> -<span class="sourceLineNo">754</span> static boolean readWithExtra(InputStream in, byte[] buf,<a name="line.754"></a> -<span class="sourceLineNo">755</span> int bufOffset, int necessaryLen, int extraLen) throws IOException {<a name="line.755"></a> -<span class="sourceLineNo">756</span> int bytesRemaining = necessaryLen + extraLen;<a name="line.756"></a> -<span class="sourceLineNo">757</span> while (bytesRemaining > 0) {<a name="line.757"></a> -<span class="sourceLineNo">758</span> int ret = in.read(buf, bufOffset, bytesRemaining);<a name="line.758"></a> -<span class="sourceLineNo">759</span> if (ret == -1 && bytesRemaining <= extraLen) {<a name="line.759"></a> -<span class="sourceLineNo">760</span> // We could not read the "extra data", but that is OK.<a name="line.760"></a> -<span class="sourceLineNo">761</span> break;<a name="line.761"></a> -<span class="sourceLineNo">762</span> }<a name="line.762"></a> -<span class="sourceLineNo">763</span> if (ret < 0) {<a name="line.763"></a> -<span class="sourceLineNo">764</span> throw new IOException("Premature EOF from inputStream (read "<a name="line.764"></a> -<span class="sourceLineNo">765</span> + "returned " + ret + ", was trying to read " + necessaryLen<a name="line.765"></a> -<span class="sourceLineNo">766</span> + " necessary bytes and " + extraLen + " extra bytes, "<a name="line.766"></a> -<span class="sourceLineNo">767</span> + "successfully read "<a name="line.767"></a> -<span class="sourceLineNo">768</span> + (necessaryLen + extraLen - bytesRemaining));<a name="line.768"></a> -<span class="sourceLineNo">769</span> }<a name="line.769"></a> -<span class="sourceLineNo">770</span> bufOffset += ret;<a name="line.770"></a> -<span class="sourceLineNo">771</span> bytesRemaining -= ret;<a name="line.771"></a> -<span class="sourceLineNo">772</span> }<a name="line.772"></a> -<span class="sourceLineNo">773</span> return bytesRemaining <= 0;<a name="line.773"></a> -<span class="sourceLineNo">774</span> }<a name="line.774"></a> -<span class="sourceLineNo">775</span><a name="line.775"></a> -<span class="sourceLineNo">776</span> /**<a name="line.776"></a> -<span class="sourceLineNo">777</span> * Read from an input stream at least <code>necessaryLen</code> and if possible,<a name="line.777"></a> -<span class="sourceLineNo">778</span> * <code>extraLen</code> also if available. Analogous to<a name="line.778"></a> -<span class="sourceLineNo">779</span> * {@link IOUtils#readFully(InputStream, byte[], int, int)}, but uses<a name="line.779"></a> -<span class="sourceLineNo">780</span> * positional read and specifies a number of "extra" bytes that would be<a name="line.780"></a> -<span class="sourceLineNo">781</span> * desirable but not absolutely necessary to read.<a name="line.781"></a> -<span class="sourceLineNo">782</span> *<a name="line.782"></a> -<span class="sourceLineNo">783</span> * @param in the input stream to read from<a name="line.783"></a> -<span class="sourceLineNo">784</span> * @param position the position within the stream from which to start reading<a name="line.784"></a> -<span class="sourceLineNo">785</span> * @param buf the buffer to read into<a name="line.785"></a> -<span class="sourceLineNo">786</span> * @param bufOffset the destination offset in the buffer<a name="line.786"></a> -<span class="sourceLineNo">787</span> * @param necessaryLen the number of bytes that are absolutely necessary to<a name="line.787"></a> -<span class="sourceLineNo">788</span> * read<a name="line.788"></a> -<span class="sourceLineNo">789</span> * @param extraLen the number of extra bytes that would be nice to read<a name="line.789"></a> -<span class="sourceLineNo">790</span> * @return true if and only if extraLen is > 0 and reading those extra bytes<a name="line.790"></a> -<span class="sourceLineNo">791</span> * was successful<a name="line.791"></a> -<span class="sourceLineNo">792</span> * @throws IOException if failed to read the necessary bytes<a name="line.792"></a> -<span class="sourceLineNo">793</span> */<a name="line.793"></a> -<span class="sourceLineNo">794</span> @VisibleForTesting<a name="line.794"></a> -<span class="sourceLineNo">795</span> static boolean positionalReadWithExtra(FSDataInputStream in,<a name="line.795"></a> -<span class="sourceLineNo">796</span> long position, byte[] buf, int bufOffset, int necessaryLen, int extraLen)<a name="line.796"></a> -<span class="sourceLineNo">797</span> throws IOException {<a name="line.797"></a> -<span class="sourceLineNo">798</span> int bytesRemaining = necessaryLen + extraLen;<a name="line.798"></a> -<span class="sourceLineNo">799</span> int bytesRead = 0;<a name="line.799"></a> -<span class="sourceLineNo">800</span> while (bytesRead < necessaryLen) {<a name="line.800"></a> -<span class="sourceLineNo">801</span> int ret = in.read(position, buf, bufOffset, bytesRemaining);<a name="line.801"></a> -<span class="sourceLineNo">802</span> if (ret < 0) {<a name="line.802"></a> -<span class="sourceLineNo">803</span> throw new IOException("Premature EOF from inputStream (positional read "<a name="line.803"></a> -<span class="sourceLineNo">804</span> + "returned " + ret + ", was trying to read " + necessaryLen<a name="line.804"></a> -<span class="sourceLineNo">805</span> + " necessary bytes and " + extraLen + " extra bytes, "<a name="line.805"></a> -<span class="sourceLineNo">806</span> + "successfully read " + bytesRead);<a name="line.806"></a> -<span class="sourceLineNo">807</span> }<a name="line.807"></a> -<span class="sourceLineNo">808</span> position += ret;<a name="line.808"></a> -<span class="sourceLineNo">809</span> bufOffset += ret;<a name="line.809"></a> -<span class="sourceLineNo">810</span> bytesRemaining -= ret;<a name="line.810"></a> -<span class="sourceLineNo">811</span> bytesRead += ret;<a name="line.811"></a> -<span class="sourceLineNo">812</span> }<a name="line.812"></a> -<span class="sourceLineNo">813</span> return bytesRead != necessaryLen && bytesRemaining <= 0;<a name="line.813"></a> -<span class="sourceLineNo">814</span> }<a name="line.814"></a> -<span class="sourceLineNo">815</span><a name="line.815"></a> -<span class="sourceLineNo">816</span> /**<a name="line.816"></a> -<span class="sourceLineNo">817</span> * Unified version 2 {@link HFile} block writer. The intended usage pattern<a name="line.817"></a> -<span class="sourceLineNo">818</span> * is as follows:<a name="line.818"></a> -<span class="sourceLineNo">819</span> * <ol><a name="line.819"></a> -<span class="sourceLineNo">820</span> * <li>Construct an {@link HFileBlock.Writer}, providing a compression algorithm.<a name="line.820"></a> -<span class="sourceLineNo">821</span> * <li>Call {@link Writer#startWriting} and get a data stream to write to.<a name="line.821"></a> -<span class="sourceLineNo">822</span> * <li>Write your data into the stream.<a name="line.822"></a> -<span class="sourceLineNo">823</span> * <li>Call Writer#writeHeaderAndData(FSDataOutputStream) as many times as you need to.<a name="line.823"></a> -<span class="sourceLineNo">824</span> * store the serialized block into an external stream.<a name="line.824"></a> -<span class="sourceLineNo">825</span> * <li>Repeat to write more blocks.<a name="line.825"></a> -<span class="sourceLineNo">826</span> * </ol><a name="line.826"></a> -<span class="sourceLineNo">827</span> * <p><a name="line.827"></a> -<span class="sourceLineNo">828</span> */<a name="line.828"></a> -<span class="sourceLineNo">829</span> static class Writer {<a name="line.829"></a> -<span class="sourceLineNo">830</span> private enum State {<a name="line.830"></a> -<span class="sourceLineNo">831</span> INIT,<a name="line.831"></a> -<span class="sourceLineNo">832</span> WRITING,<a name="line.832"></a> -<span class="sourceLineNo">833</span> BLOCK_READY<a name="line.833"></a> -<span class="sourceLineNo">834</span> };<a name="line.834"></a> -<span class="sourceLineNo">835</span><a name="line.835"></a> -<span class="sourceLineNo">836</span> /** Writer state. Used to ensure the correct usage protocol. */<a name="line.836"></a> -<span class="sourceLineNo">837</span> private State state = State.INIT;<a name="line.837"></a> -<span class="sourceLineNo">838</span><a name="line.838"></a> -<span class="sourceLineNo">839</span> /** Data block encoder used for data blocks */<a name="line.839"></a> -<span class="sourceLineNo">840</span> private final HFileDataBlockEncoder dataBlockEncoder;<a name="line.840"></a> -<span class="sourceLineNo">841</span><a name="line.841"></a> -<span class="sourceLineNo">842</span> private HFileBlockEncodingContext dataBlockEncodingCtx;<a name="line.842"></a> -<span class="sourceLineNo">843</span><a name="line.843"></a> -<span class="sourceLineNo">844</span> /** block encoding context for non-data blocks*/<a name="line.844"></a> -<span class="sourceLineNo">845</span> private HFileBlockDefaultEncodingContext defaultBlockEncodingCtx;<a name="line.845"></a> -<span class="sourceLineNo">846</span><a name="line.846"></a> -<span class="sourceLineNo">847</span> /**<a name="line.847"></a> -<span class="sourceLineNo">848</span> * The stream we use to accumulate data into a block in an uncompressed format.<a name="line.848"></a> -<span class="sourceLineNo">849</span> * We reset this stream at the end of each block and reuse it. The<a name="line.849"></a> -<span class="sourceLineNo">850</span> * header is written as the first {@link HConstants#HFILEBLOCK_HEADER_SIZE} bytes into this<a name="line.850"></a> -<span class="sourceLineNo">851</span> * stream.<a name="line.851"></a> -<span class="sourceLineNo">852</span> */<a name="line.852"></a> -<span class="sourceLineNo">853</span> private ByteArrayOutputStream baosInMemory;<a name="line.853"></a> -<span class="sourceLineNo">854</span><a name="line.854"></a> -<span class="sourceLineNo">855</span> /**<a name="line.855"></a> -<span class="sourceLineNo">856</span> * Current block type. Set in {@link #startWriting(BlockType)}. Could be<a name="line.856"></a> -<span class="sourceLineNo">857</span> * changed in {@link #finishBlock()} from {@link BlockType#DATA}<a name="line.857"></a> -<span class="sourceLineNo">858</span> * to {@link BlockType#ENCODED_DATA}.<a name="line.858"></a> -<span class="sourceLineNo">859</span> */<a name="line.859"></a> -<span class="sourceLineNo">860</span> private BlockType blockType;<a name="line.860"></a> -<span class="sourceLineNo">861</span><a name="line.861"></a> -<span class="sourceLineNo">862</span> /**<a name="line.862"></a> -<span class="sourceLineNo">863</span> * A stream that we write uncompressed bytes to, which compresses them and<a name="line.863"></a> -<span class="sourceLineNo">864</span> * writes them to {@link #baosInMemory}.<a name="line.864"></a> -<span class="sourceLineNo">865</span> */<a name="line.865"></a> -<span class="sourceLineNo">866</span> private DataOutputStream userDataStream;<a name="line.866"></a> -<span class="sourceLineNo">867</span><a name="line.867"></a> -<span class="sourceLineNo">868</span> // Size of actual data being written. Not considering the block encoding/compression. This<a name="line.868"></a> -<span class="sourceLineNo">869</span> // includes the header size also.<a name="line.869"></a> -<span class="sourceLineNo">870</span> private int unencodedDataSizeWritten;<a name="line.870"></a> -<span class="sourceLineNo">871</span><a name="line.871"></a> -<span class="sourceLineNo">872</span> // Size of actual data being written. considering the block encoding. This<a name="line.872"></a> -<span class="sourceLineNo">873</span> // includes the header size also.<a name="line.873"></a> -<span class="sourceLineNo">874</span> private int encodedDataSizeWritten;<a name="line.874"></a> -<span class="sourceLineNo">875</span><a name="line.875"></a> -<span class="sourceLineNo">876</span> /**<a name="line.876"></a> -<span class="sourceLineNo">877</span> * Bytes to be written to the file system, including the header. Compressed<a name="line.877"></a> -<span class="sourceLineNo">878</span> * if compression is turned on. It also includes the checksum data that<a name="line.878"></a> -<span class="sourceLineNo">879</span> * immediately follows the block data. (header + data + checksums)<a name="line.879"></a> -<span class="sourceLineNo">880</span> */<a name="line.880"></a> -<span class="sourceLineNo">881</span> private ByteArrayOutputStream onDiskBlockBytesWithHeader;<a name="line.881"></a> -<span class="sourceLineNo">882</span><a name="line.882"></a> -<span class="sourceLineNo">883</span> /**<a name="line.883"></a> -<span class="sourceLineNo">884</span> * The size of the checksum data on disk. It is used only if data is<a name="line.884"></a> -<span class="sourceLineNo">885</span> * not compressed. If data is compressed, then the checksums are already<a name="line.885"></a> -<span class="sourceLineNo">886</span> * part of onDiskBytesWithHeader. If data is uncompressed, then this<a name="line.886"></a> -<span class="sourceLineNo">887</span> * variable stores the checksum data for this block.<a name="line.887"></a> -<span class="sourceLineNo">888</span> */<a name="line.888"></a> -<span class="sourceLineNo">889</span> private byte[] onDiskChecksum = HConstants.EMPTY_BYTE_ARRAY;<a name="line.889"></a> -<span class="sourceLineNo">890</span><a name="line.890"></a> -<span class="sourceLineNo">891</span> /**<a name="line.891"></a> -<span class="sourceLineNo">892</span> * Current block's start offset in the {@link HFile}. Set in<a name="line.892"></a> -<span class="sourceLineNo">893</span> * {@link #writeHeaderAndData(FSDataOutputStream)}.<a name="line.893"></a> -<span class="sourceLineNo">894</span> */<a name="line.894"></a> -<span class="sourceLineNo">895</span> private long startOffset;<a name="line.895"></a> -<span class="sourceLineNo">896</span><a name="line.896"></a> -<span class="sourceLineNo">897</span> /**<a name="line.897"></a> -<span class="sourceLineNo">898</span> * Offset of previous block by block type. Updated when the next block is<a name="line.898"></a> -<span class="sourceLineNo">899</span> * started.<a name="line.899"></a> -<span class="sourceLineNo">900</span> */<a name="line.900"></a> -<span class="sourceLineNo">901</span> private long[] prevOffsetByType;<a name="line.901"></a> -<span class="sourceLineNo">902</span><a name="line.902"></a> -<span class="sourceLineNo">903</span> /** The offset of the previous block of the same type */<a name="line.903"></a> -<span class="sourceLineNo">904</span> private long prevOffset;<a name="line.904"></a> -<span class="sourceLineNo">905</span> /** Meta data that holds information about the hfileblock**/<a name="line.905"></a> -<span class="sourceLineNo">906</span> private HFileContext fileContext;<a name="line.906"></a> -<span class="sourceLineNo">907</span><a name="line.907"></a> -<span class="sourceLineNo">908</span> /**<a name="line.908"></a> -<span class="sourceLineNo">909</span> * @param dataBlockEncoder data block encoding algorithm to use<a name="line.909"></a> -<span class="sourceLineNo">910</span> */<a name="line.910"></a> -<span class="sourceLineNo">911</span> public Writer(HFileDataBlockEncoder dataBlockEncoder, HFileContext fileContext) {<a name="line.911"></a> -<span class="sourceLineNo">912</span> if (fileContext.getBytesPerChecksum() < HConstants.HFILEBLOCK_HEADER_SIZE) {<a name="line.912"></a> -<span class="sourceLineNo">913</span> throw new RuntimeException("Unsupported value of bytesPerChecksum. " +<a name="line.913"></a> -<span class="sourceLineNo">914</span> " Minimum is " + HConstants.HFILEBLOCK_HEADER_SIZE + " but the configured value is " +<a name="line.914"></a> -<span class="sourceLineNo">915</span> fileContext.getBytesPerChecksum());<a name="line.915"></a> -<span class="sourceLineNo">916</span> }<a name="line.916"></a> -<span class="sourceLineNo">917</span> this.dataBlockEncoder = dataBlockEncoder != null?<a name="line.917"></a> -<span class="sourceLineNo">918</span> dataBlockEncoder: NoOpDataBlockEncoder.INSTANCE;<a name="line.918"></a> -<span class="sourceLineNo">919</span> this.dataBlockEncodingCtx = this.dataBlockEncoder.<a name="line.919"></a> -<span class="sourceLineNo">920</span> newDataBlockEncodingContext(HConstants.HFILEBLOCK_DUMMY_HEADER, fileContext);<a name="line.920"></a> -<span class="sourceLineNo">921</span> // TODO: This should be lazily instantiated since we usually do NOT need this default encoder<a name="line.921"></a> -<span class="sourceLineNo">922</span> this.defaultBlockEncodingCtx = new HFileBlockDefaultEncodingContext(null,<a name="line.922"></a> -<span class="sourceLineNo">923</span> HConstants.HFILEBLOCK_DUMMY_HEADER, fileContext);<a name="line.923"></a> -<span class="sourceLineNo">924</span> // TODO: Set BAOS initial size. Use fileContext.getBlocksize() and add for header/checksum<a name="line.924"></a> -<span class="sourceLineNo">925</span> baosInMemory = new ByteArrayOutputStream();<a name="line.925"></a> -<span class="sourceLineNo">926</span> prevOffsetByType = new long[BlockType.values().length];<a name="line.926"></a> -<span class="sourceLineNo">927</span> for (int i = 0; i < prevOffsetByType.length; ++i) {<a name="line.927"></a> -<span class="sourceLineNo">928</span> prevOffsetByType[i] = UNSET;<a name="line.928"></a> -<span class="sourceLineNo">929</span> }<a name="line.929"></a> -<span class="sourceLineNo">930</span> // TODO: Why fileContext saved away when we have dataBlockEncoder and/or<a name="line.930"></a> -<span class="sourceLineNo">931</span> // defaultDataBlockEncoder?<a name="line.931"></a> -<span class="sourceLineNo">932</span> this.fileContext = fileContext;<a name="line.932"></a> -<span class="sourceLineNo">933</span> }<a name="line.933"></a> -<span class="sourceLineNo">934</span><a name="line.934"></a> -<span class="sourceLineNo">935</span> /**<a name="line.935"></a> -<span class="sourceLineNo">936</span> * Starts writing into the block. The previous block's data is discarded.<a name="line.936"></a> -<span class="sourceLineNo">937</span> *<a name="line.937"></a> -<span class="sourceLineNo">938</span> * @return the stream the user can write their data into<a name="line.938"></a> -<span class="sourceLineNo">939</span> * @throws IOException<a name="line.939"></a> -<span class="sourceLineNo">940</span> */<a name="line.940"></a> -<span class="sourceLineNo">941</span> DataOutputStream startWriting(BlockType newBlockType)<a name="line.941"></a> -<span class="sourceLineNo">942</span> throws IOException {<a name="line.942"></a> -<span class="sourceLineNo">943</span> if (state == State.BLOCK_READY && startOffset != -1) {<a name="line.943"></a> -<span class="sourceLineNo">944</span> // We had a previous block that was written to a stream at a specific<a name="line.944"></a> -<span class="sourceLineNo">945</span> // offset. Save that offset as the last offset of a block of that type.<a name="line.945"></a> -<span class="sourceLineNo">946</span> prevOffsetByType[blockType.getId()] = startOffset;<a name="line.946"></a> -<span class="sourceLineNo">947</span> }<a name="line.947"></a> -<span class="sourceLineNo">948</span><a name="line.948"></a> -<span class="sourceLineNo">949</span> startOffset = -1;<a name="line.949"></a> -<span class="sourceLineNo">950</span> blockType = newBlockType;<a name="line.950"></a> -<span class="sourceLineNo">951</span><a name="line.951"></a> -<span class="sourceLineNo">952</span> baosInMemory.reset();<a name="line.952"></a> -<span class="sourceLineNo">953</span> baosInMemory.write(HConstants.HFILEBLOCK_DUMMY_HEADER);<a name="line.953"></a> -<span class="sourceLineNo">954</span><a name="line.954"></a> -<span class="sourceLineNo">955</span> state = State.WRITING;<a name="line.955"></a> -<span class="sourceLineNo">956</span><a name="line.956"></a> -<span class="sourceLineNo">957</span> // We will compress it later in finishBlock()<a name="line.957"></a> -<span class="sourceLineNo">958</span> userDataStream = new ByteBufferWriterDataOutputStream(baosInMemory);<a name="line.958"></a> -<span class="sourceLineNo">959</span> if (newBlockType == BlockType.DATA) {<a name="line.959"></a> -<span class="sourceLineNo">960</span> this.dataBlockEncoder.startBlockEncoding(dataBlockEncodingCtx, userDataStream);<a name="line.960"></a> -<span class="sourceLineNo">961</span> }<a name="line.961"></a> -<span class="sourceLineNo">962</span> this.unencodedDataSizeWritten = 0;<a name="line.962"></a> -<span class="sourceLineNo">963</span> this.encodedDataSizeWritten = 0;<a name="line.963"></a> -<span class="sourceLineNo">964</span> return userDataStream;<a name="line.964"></a> -<span class="sourceLineNo">965</span> }<a name="line.965"></a> -<span class="sourceLineNo">966</span><a name="line.966"></a> -<span class="sourceLineNo">967</span> /**<a name="line.967"></a> -<span class="sourceLineNo">968</span> * Writes the Cell to this block<a name="line.968"></a> -<span class="sourceLineNo">969</span> * @param cell<a name="line.969"></a> -<span class="sourceLineNo">970</span> * @throws IOException<a name="line.970"></a> -<span class="sourceLineNo">971</span> */<a name="line.971"></a> -<span class="sourceLineNo">972</span> void write(Cell cell) throws IOException{<a name="line.972"></a> -<span class="sourceLineNo">973</span> expectState(State.WRITING);<a name="line.973"></a> -<span class="sourceLineNo">974</span> int posBeforeEncode = this.userDataStream.size();<a name="line.974"></a> -<span class="sourceLineNo">975</span> this.unencodedDataSizeWritten +=<a name="line.975"></a> -<span class="sourceLineNo">976</span> this.dataBlockEncoder.encode(cell, dataBlockEncodingCtx, this.userDataStream);<a name="line.976"></a> -<span class="sourceLineNo">977</span> this.encodedDataSizeWritten += this.userDataStream.size() - posBeforeEncode;<a name="line.977"></a> -<span class="sourceLineNo">978</span> }<a name="line.978"></a> -<span class="sourceLineNo">979</span><a name="line.979"></a> -<span class="sourceLineNo">980</span> /**<a name="line.980"></a> -<span class="sourceLineNo">981</span> * Returns the stream for the user to write to. The block writer takes care<a name="line.981"></a> -<span class="sourceLineNo">982</span> * of handling compression and buffering for caching on write. Can only be<a name="line.982"></a> -<span class="sourceLineNo">983</span> * called in the "writing" state.<a name="line.983"></a> -<span class="sourceLineNo">984</span> *<a name="line.984"></a> -<span class="sourceLineNo">985</span> * @return the data output stream for the user to write to<a name="line.985"></a> -<span class="sourceLineNo">986</span> */<a name="line.986"></a> -<span class="sourceLineNo">987</span> DataOutputStream getUserDataStream() {<a name="line.987"></a> -<span class="sourceLineNo">988</span> expectState(State.WRITING);<a name="line.988"></a> -<span class="sourceLineNo">989</span> return userDataStream;<a name="line.989"></a> -<span class="sourceLineNo">990</span> }<a name="line.990"></a> -<span class="sourceLineNo">991</span><a name="line.991"></a> -<span class="sourceLineNo">992</span> /**<a name="line.992"></a> -<span class="sourceLineNo">993</span> * Transitions the block writer from the "writing" state to the "block<a name="line.993"></a> -<span class="sourceLineNo">994</span> * ready" state. Does nothing if a block is already finished.<a name="line.994"></a> -<span class="sourceLineNo">995</span> */<a name="line.995"></a> -<span class="sourceLineNo">996</span> void ensureBlockReady() throws IOException {<a name="line.996"></a> -<span class="sourceLineNo">997</span> Preconditions.checkState(state != State.INIT,<a name="line.997"></a> -<span class="sourceLineNo">998</span> "Unexpected state: " + state);<a name="line.998"></a> -<span class="sourceLineNo">999</span><a name="line.999"></a> -<span class="sourceLineNo">1000</span> if (state == State.BLOCK_READY) {<a name="line.1000"></a> -<span class="sourceLineNo">1001</span> return;<a name="line.1001"></a> -<span class="sourceLineNo">1002</span> }<a name="line.1002"></a> -<span class="sourceLineNo">1003</span><a name="line.1003"></a> -<span class="sourceLineNo">1004</span> // This will set state to BLOCK_READY.<a name="line.1004"></a> -<span class="sourceLineNo">1005</span> finishBlock();<a name="line.1005"></a> -<span class="sourceLineNo">1006</span> }<a name="line.1006"></a> -<span class="sourceLineNo">1007</span><a name="line.1007"></a> -<span class="sourceLineNo">1008</span> /**<a name="line.1008"></a> -<span class="sourceLineNo">1009</span> * Finish up writing of the block.<a name="line.1009"></a> -<span class="sourceLineNo">1010</span> * Flushes the compressing stream (if using compression), fills out the header,<a name="line.1010"></a> -<span class="sourceLineNo">1011</span> * does any compression/encryption of bytes to flush out to disk, and manages<a name="line.1011"></a> -<span class="sourceLineNo">1012</span> * the cache on write content, if applicable. Sets block write state to "block ready".<a name="line.1012"></a> -<span class="sourceLineNo">1013</span> */<a name="line.1013"></a> -<span class="sourceLineNo">1014</span> private void finishBlock() throws IOException {<a name="line.1014"></a> -<span class="sourceLineNo">1015</span> if (blockType == BlockType.DATA) {<a name="line.1015"></a> -<span class="sourceLineNo">1016</span> this.dataBlockEncoder.endBlockEncoding(dataBlockEncodingCtx, userDataStream,<a name="line.1016"></a> -<span class="sourceLineNo">1017</span> baosInMemory.getBuffer(), blockType);<a name="line.1017"></a> -<span class="sourceLineNo">1018</span> blockType = dataBlockEncodingCtx.getBlockType();<a name="line.1018"></a> -<span class="sourceLineNo">1019</span> }<a name="line.1019"></a> -<span class="sourceLineNo">1020</span> userDataStream.flush();<a name="line.1020"></a> -<span class="sourceLineNo">1021</span> prevOffset = prevOffsetByType[blockType.getId()];<a name="line.1021"></a> -<span class="sourceLineNo">1022</span><a name="line.1022"></a> -<span class="sourceLineNo">1023</span> // We need to set state before we can package the block up for cache-on-write. In a way, the<a name="line.1023"></a> -<span class="sourceLineNo">1024</span> // block is ready, but not yet encoded or compressed.<a name="line.1024"></a> -<span class="sourceLineNo">1025</span> state = State.BLOCK_READY;<a name="line.1025"></a> -<span class="sourceLineNo">1026</span> Bytes compressAndEncryptDat;<a name="line.1026"></a> -<span class="sourceLineNo">1027</span> if (blockType == BlockType.DATA || blockType == BlockType.ENCODED_DATA) {<a name="line.1027"></a> -<span class="sourceLineNo">1028</span> compressAndEncryptDat = dataBlockEncodingCtx.<a name="line.1028"></a> -<span class="sourceLineNo">1029</span> compressAndEncrypt(baosInMemory.getBuffer(), 0, baosInMemory.size());<a name="line.1029"></a> -<span class="sourceLineNo">1030</span> } else {<a name="line.1030"></a> -<span class="sourceLineNo">1031</span> compressAndEncryptDat = defaultBlockEncodingCtx.<a name="line.1031"></a> -<span class="sourceLineNo">1032</span> compressAndEncrypt(baosInMemory.getBuffer(), 0, baosInMemory.size());<a name="line.1032"></a> -<span class="sourceLineNo">1033</span> }<a name="line.1033"></a> -<span class="sourceLineNo">1034</span> if (compressAndEncryptDat == null) {<a name="line.1034"></a> -<span class="sourceLineNo">1035</span> compressAndEncryptDat = new Bytes(baosInMemory.getBuffer(), 0, baosInMemory.size());<a name="line.1035"></a> -<span class="sourceLineNo">1036</span> }<a name="line.1036"></a> -<span class="sourceLineNo">1037</span> if (onDiskBlockBytesWithHeader == null) {<a name="line.1037"></a> -<span class="sourceLineNo">1038</span> onDiskBlockBytesWithHeader = new ByteArrayOutputStream(compressAndEncryptDat.getLength());<a name="line.1038"></a> -<span class="sourceLineNo">1039</span> }<a name="line.1039"></a> -<span class="sourceLineNo">1040</span> onDiskBlockBytesWithHeader.reset();<a name="line.1040"></a> -<span class="sourceLineNo">1041</span> onDiskBlockBytesWithHeader.write(compressAndEncryptDat.get(),<a name="line.1041"></a> -<span class="sourceLineNo">1042</span> compressAndEncryptDat.getOffset(), compressAndEncryptDat.getLength());<a name="line.1042"></a> -<span class="sourceLineNo">1043</span> // Calculate how many bytes we need for checksum on the tail of the block.<a name="line.1043"></a> -<span class="sourceLineNo">1044</span> int numBytes = (int) ChecksumUtil.numBytes(<a name="line.1044"></a> -<span class="sourceLineNo">1045</span> onDiskBlockBytesWithHeader.size(),<a name="line.1045"></a> -<span class="sourceLineNo">1046</span> fileContext.getBytesPerChecksum());<a name="line.1046"></a> -<span class="sourceLineNo">1047</span><a name="line.1047"></a> -<span class="sourceLineNo">1048</span> // Put the header for the on disk bytes; header currently is unfilled-out<a name="line.1048"></a> -<span class="sourceLineNo">1049</span> putHeader(onDiskBlockBytesWithHeader,<a name="line.1049"></a> -<span class="sourceLineNo">1050</span> onDiskBlockBytesWithHeader.size() + numBytes,<a name="line.1050"></a> -<span class="sourceLineNo">1051</span> baosInMemory.size(), onDiskBlockBytesWithHeader.size());<a name="line.1051"></a> -<span class="sourceLineNo">1052</span> if (onDiskChecksum.length != numBytes) {<a name="line.1052"></a> -<span class="sourceLineNo">1053</span> onDiskChecksum = new byte[numBytes];<a name="line.1053"></a> -<span class="sourceLineNo">1054</span> }<a name="line.1054"></a> -<span class="sourceLineNo">1055</span> ChecksumUtil.generateChecksums(<a name="line.1055"></a> -<span class="sourceLineNo">1056</span> onDiskBlockBytesWithHeader.getBuffer(), 0,onDiskBlockBytesWithHeader.size(),<a name="line.1056"></a> -<span class="sourceLineNo">1057</span> onDiskChecksum, 0, fileContext.getChecksumType(), fileContext.getBytesPerChecksum());<a name="line.1057"></a> -<span class="sourceLineNo">1058</span> }<a name="line.1058"></a> -<span class="sourceLineNo">1059</span><a name="line.1059"></a> -<span class="sourceLineNo">1060</span> /**<a name="line.1060"></a> -<span class="sourceLineNo">1061</span> * Put the header into the given byte array at the given offset.<a name="line.1061"></a> -<span class="sourceLineNo">1062</span> * @param onDiskSize size of the block on disk header + data + checksum<a name="line.1062"></a> -<span class="sourceLineNo">1063</span> * @param uncompressedSize size of the block after decompression (but<a name="line.1063"></a> -<span class="sourceLineNo">1064</span> * before optional data block decoding) including header<a name="line.1064"></a> -<span class="sourceLineNo">1065</span> * @param onDiskDataSize size of the block on disk with header<a name="line.1065"></a> -<span class="sourceLineNo">1066</span> * and data but not including the checksums<a name="line.1066"></a> -<span class="sourceLineNo">1067</span> */<a name="line.1067"></a> -<span class="sourceLineNo">1068</span> private void putHeader(byte[] dest, int offset, int onDiskSize,<a name="line.1068"></a> -<span class="sourceLineNo">1069</span> int uncompressedSize, int onDiskDataSize) {<a name="line.1069"></a> -<span class="sourceLineNo">1070</span> offset = blockType.put(dest, offset);<a name="line.1070"></a> -<span class="sourceLineNo">1071</span> offset = Bytes.putInt(dest, offset, onDiskSize - HConstants.HFILEBLOCK_HEADER_SIZE);<a name="line.1071"></a> -<span class="sourceLineNo">1072</span> offset = Bytes.putInt(dest, offset, uncompressedSize - HConstants.HFILEBLOCK_HEADER_SIZE);<a name="line.1072"></a> -<span class="sourceLineNo">1073</span> offset = Bytes.putLong(dest, offset, prevOffset);<a name="line.1073"></a> -<span class="sourceLineNo">1074</span> offset = Bytes.putByte(dest, offset, fileContext.getChecksumType().getCode());<a name="line.1074"></a> -<span class="sourceLineNo">1075</span> offset = Bytes.putInt(dest, offset, fileContext.getBytesPerChecksum());<a name="line.1075"></a> -<span class="sourceLineNo">1076</span> Bytes.putInt(dest, offset, onDiskDataSize);<a name="line.1076"></a> -<span class="sourceLineNo">1077</span> }<a name="line.1077"></a> -<span class="sourceLineNo">1078</span><a name="line.1078"></a> -<span class="sourceLineNo">1079</span> private void putHeader(ByteArrayOutputStream dest, int onDiskSize,<a name="line.1079"></a> -<span class="sourceLineNo">1080</span> int uncompressedSize, int onDiskDataSize) {<a name="line.1080"></a> -<span class="sourceLineNo">1081</span> putHeader(dest.getBuffer(),0, onDiskSize, uncompressedSize, onDiskDataSize);<a name="line.1081"></a> -<span class="sourceLineNo">1082</span> }<a name="line.1082"></a> -<span class="sourceLineNo">1083</span><a name="line.1083"></a> -<span class="sourceLineNo">1084</span> /**<a name="line.1084"></a> -<span class="sourceLineNo">1085</span> * Similar to {@link #writeHeaderAndData(FSDataOutputStream)}, but records<a name="line.1085"></a> -<span class="sourceLineNo">1086</span> * the offset of this block so that it can be referenced in the next block<a name="line.1086"></a> -<span class="sourceLineNo">1087</span> * of the same type.<a name="line.1087"></a> -<span class="sourceLineNo">1088</span> *<a name="line.1088"></a> -<span class="sourceLineNo">1089</span> * @param out<a name="line.1089"></a> -<span class="sourceLineNo">1090</span> * @throws IOException<a name="line.1090"></a> -<span class="sourceLineNo">1091</span> */<a name="line.1091"></a> -<span class="sourceLineNo">1092</span> void writeHeaderAndData(FSDataOutputStream out) throws IOException {<a name="line.1092"></a> -<span class="sourceLineNo">1093</span> long offset = out.getPos();<a name="line.1093"></a> -<span class="sourceLineNo">1094</span> if (startOffset != UNSET && offset != startOffset) {<a name="line.1094"></a> -<span class="sourceLineNo">1095</span> throw new IOException("A " + blockType + " block written to a "<a name="line.1095"></a> -<span class="sourceLineNo">1096</span> + "stream twice, first at offset " + startOffset + ", then at "<a name="line.1096"></a> -<span class="sourceLineNo">1097</span> + offset);<a name="line.1097"></a> -<span class="sourceLineNo">1098</span> }<a name="line.1098"></a> -<span class="sourceLineNo">1099</span> startOffset = offset;<a name="line.1099"></a> -<span class="sourceLineNo">1100</span><a name="line.1100"></a> -<span class="sourceLineNo">1101</span> finishBlockAndWriteHeaderAndData((DataOutputStream) out);<a name="line.1101"></a> -<span class="sourceLineNo">1102</span> }<a name="line.1102"></a> -<span class="sourceLineNo">1103</span><a name="line.1103"></a> -<span class="sourceLineNo">1104</span> /**<a name="line.1104"></a> -<span class="sourceLineNo">1105</span> * Writes the header and the compressed data of this block (or uncompressed<a name="line.1105"></a> -<span class="sourceLineNo">1106</span> * data when not using compression) into the given stream. Can be called in<a name="line.1106"></a> -<span class="sourceLineNo">1107</span> * the "writing" state or in the "block ready" state. If called in the<a name="line.1107"></a> -<span class="sourceLineNo">1108</span> * "writing" state, transitions the writer to the "block ready" state.<a name="line.1108"></a> -<span class="sourceLineNo">1109</span> *<a name="line.1109"></a> -<span class="sourceLineNo">1110</span> * @param out the output stream to write the<a name="line.1110"></a> -<span class="sourceLineNo">1111</span> * @throws IOException<a name="line.1111"></a> -<span class="sourceLineNo">1112</span> */<a name="line.1112"></a> -<span class="sourceLineNo">1113</span> protected void finishBlockAndWriteHeaderAndData(DataOutputStream out)<a name="line.1113"></a> -<span class="sourceLineNo">1114</span> throws IOException {<a name="line.1114"></a> -<span class="sourceLineNo">1115</span> ensureBlockReady();<a name="line.1115"></a> -<span class="sourceLineNo">1116</span> out.write(onDiskBlockBytesWithHeader.getBuffer(), 0, onDiskBlockBytesWithHeader.size());<a name="line.1116"></a> -<span class="sourceLineNo">1117</span> out.write(onDiskChecksum);<a name="line.1117"></a> -<span class="sourceLineNo">1118</span> }<a name="line.1118"></a> -<span class="sourceLineNo">1119</span><a name="line.1119"></a> -<span class="sourceLineNo">1120</span> /**<a name="line.1120"></a> -<span class="sourceLineNo">1121</span> * Returns the header or the compressed data (or uncompressed data when not<a name="line.1121"></a> -<span class="sourceLineNo">1122</span> * using compression) as a byte array. Can be called in the "writing" state<a name="line.1122"></a> -<span class="sourceLineNo">1123</span> * or in the "block ready" state. If called in the "writing" state,<a name="line.1123"></a> -<span class="sourceLineNo">1124</span> * transitions the writer to the "block ready" state. This returns<a name="line.1124"></a> -<span class="sourceLineNo">1125</span> * the header + data + checksums stored on disk.<a name="line.1125"></a> -<span class="sourceLineNo">1126</span> *<a name="line.1126"></a> -<span class="sourceLineNo">1127</span> * @return header and data as they would be stored on disk in a byte array<a name="line.1127"></a> -<span class="sourceLineNo">1128</span> * @throws IOException<a name="line.1128"></a> -<span class="sourceLineNo">1129</span> */<a name="line.1129"></a> -<span class="sourceLineNo">1130</span> byte[] getHeaderAndDataForTest() throws IOException {<a name="line.1130"></a> -<span class="sourceLineNo">1131</span> ensureBlockReady();<a name="line.1131"></a> -<span class="sourceLineNo">1132</span> // This is not very optimal, because we are doing an extra copy.<a name="line.1132"></a> -<span class="sourceLineNo">1133</span> // But this method is used only by unit tests.<a name="line.1133"></a> -<span class="sourceLineNo">1134</span> byte[] output =<a name="line.1134"></a> -<span class="sourceLineNo">1135</span> new byte[onDiskBlockBytesWithHeader.size()<a name="line.1135"></a> -<span class="sourceLineNo">1136</span> + onDiskChecksum.length];<a name="line.1136"></a> -<span class="sourceLineNo">1137</span> System.arraycopy(onDiskBlockBytesWithHeader.getBuffer(), 0, output, 0,<a name="line.1137"></a> -<span class="sourceLineNo">1138</span> onDiskBlockBytesWithHeader.size());<a name="line.1138"></a> -<span class="sourceLineNo">1139</span> System.arraycopy(onDiskChecksum, 0, output,<a name="line.1139"></a> -<span class="sourceLineNo">1140</span> onDiskBlockBytesWithHeader.size(), onDiskChecksum.length);<a name="line.1140"></a> -<span class="sourceLineNo">1141</span> return output;<a name="line.1141"></a> -<span class="sourceLineNo">1142</span> }<a name="line.1142"></a> -<span class="sourceLineNo">1143</span><a name="line.1143"></a> -<span class="sourceLineNo">1144</span> /**<a name="line.1144"></a> -<span class="sourceLineNo">1145</span> * Releases resources used by this writer.<a name="line.1145"></a> -<span class="sourceLineNo">1146</span> */<a name="line.1146"></a> -<span class="sourceLineNo">1147</span> void release() {<a name="line.1147"></a> -<span class="sourceLineNo">1148</span> if (dataBlockEncodingCtx != null) {<a name="line.1148"></a> -<span class="sourceLineNo">1149</span> dataBlockEncodingCtx.close();<a name="line.1149"></a> -<span class="sourceLineNo">1150</span> dataBlockEncodingCtx = null;<a name="line.1150"></a> -<span class="sourceLineNo">1151</span> }<a name="line.1151"></a> -<span class="sourceLineNo">1152</span> if (defaultBlockEncodingCtx != null) {<a name="line.1152"></a> -<span class="sourceLineNo">1153</span> defaultBlockEncodingCtx.close();<a name="line.1153"></a> -<span class="sourceLineNo">1154</span> defaultBlockEncodingCtx = null;<a name="line.1154"></a> -<span class="sourceLineNo">1155</span> }<a name="line.1155"></a> -<span class="sourceLineNo">1156</span> }<a name="line.1156"></a> -<span class="sourceLineNo">1157</span><a name="line.1157"></a> -<span class="sourceLineNo">1158</span> /**<a name="line.1158"></a> -<span class="sourceLineNo">1159</span> * Returns the on-disk size of the data portion of the block. This is the<a name="line.1159"></a> -<span class="sourceLineNo">1160</span> * compressed size if compression is enabled. Can only be called in the<a name="line.1160"></a> -<span class="sourceLineNo">1161</span> * "block ready" state. Header is not compressed, and its size is not<a name="line.1161"></a> -<span class="sourceLineNo">1162</span> * included in the return value.<a name="line.1162"></a> -<span class="sourceLineNo">1163</span> *<a name="line.1163"></a> -<span class="sourceLineNo">1164</span> * @return the on-disk size of the block, not including the header.<a name="line.1164"></a> -<span class="sourceLineNo">1165</span> */<a name="line.1165"></a> -<span class="sourceLineNo">1166</span> int getOnDiskSizeWithoutHeader() {<a name="line.1166"></a> -<span class="sourceLineNo">1167</span> expectState(State.BLOCK_READY);<a name="line.1167"></a> -<span class="sourceLineNo">1168</span> return onDiskBlockBytesWithHeader.size() +<a name="line.1168"></a> -<span class="sourceLineNo">1169</span> onDiskChecksum.length - HConstants.HFILEBLOCK_HEADER_SIZE;<a name="line.1169"></a> -<span class="sourceLineNo">1170</span> }<a name="line.1170"></a> -<span class="sourceLineNo">1171</span><a name="line.1171"></a> -<span class="sourceLineNo">1172</span> /**<a name="line.1172"></a> -<span class="sourceLineNo">1173</span> * Returns the on-disk size of the block. Can only be called in the<a name="line.1173"></a> -<span class="sourceLineNo">1174</span> * "block ready" state.<a name="line.1174"></a> -<span class="sourceLineNo">1175</span> *<a name="line.1175"></a> -<span class="sourceLineNo">1176</span> * @return the on-disk size of the block ready to be written, including the<a name="line.1176"></a> -<span class="sourceLineNo">1177</span> * header size, the data and the checksum data.<a name="line.1177"></a> -<span class="sourceLineNo">1178</span> */<a name="line.1178"></a> -<span class="sourceLineNo">1179</span> int getOnDiskSizeWithHeader() {<a name="line.1179"></a> -<span class="sourceLineNo">1180</span> expectState(State.BLOCK_READY);<a name="line.1180"></a> -<span class="sourceLineNo">1181</span> return onDiskBlockBytesWithHeader.size() + onDiskChecksum.length;<a name="line.1181"></a> -<span class="sourceLineNo">1182</span> }<a name="line.1182"></a> -<span class="sourceLineNo">1183</span><a name="line.1183"></a> -<span class="sourceLineNo">1184</span> /**<a name="line.1184"></a> -<span class="sourceLineNo">1185</span> * The uncompressed size of the block data. Does not include header size.<a name="line.1185"></a> -<span class="sourceLineNo">1186</span> */<a name="line.1186"></a> -<span class="sourceLineNo">1187</span> int getUncompressedSizeWithoutHeader() {<a name="line.1187"></a> -<span class="sourceLineNo">1188</span> expectState(State.BLOCK_READY);<a name="line.1188"></a> -<span class="sourceLineNo">1189</span> return baosInMemory.size() - HConstants.HFILEBLOCK_HEADER_SIZE;<a name="line.1189"></a> -<span class="sourceLineNo">1190</span> }<a name="line.1190"></a> -<span class="sourceLineNo">1191</span><a name="line.1191"></a> -<span class="sourceLineNo">1192</span> /**<a name="line.1192"></a> -<span class="sourceLineNo">1193</span> * The uncompressed size of the block data, including header size.<a name="line.1193"></a> -<span class="sourceLineNo">1194</span> */<a name="line.1194"></a> -<span class="sourceLineNo">1195</span> int getUncompressedSizeWithHeader() {<a name="line.1195"></a> -<span class="sourceLineNo">1196</span> expectState(State.BLOCK_READY);<a name="line.1196"></a> -<span class="sourceLineNo">1197</span> return baosInMemory.size();<a name="line.1197"></a> -<span class="sourceLineNo">1198</span> }<a name="line.1198"></a> -<span class="sourceLineNo">1199</span><a name="line.1199"></a> -<span class="sourceLineNo">1200</span> /** @return true if a block is being written */<a name="line.1200"></a> -<span class="sourceLineNo">1201</span> boolean isWriting() {<a name="line.1201"></a> -<span class="sourceLineNo">1202</span> return state == State.WRITING;<a name="line.1202"></a> -<span class="sourceLineNo">1203</span> }<a name="line.1203"></a> -<span class="sourceLineNo">1204</span><a name="line.1204"></a> -<span class="sourceLineNo">1205</span> /**<a name="line.1205"></a> -<span class="sourceLineNo">1206</span> * Returns the number of bytes written into the current block so far, or<a name="line.1206"></a> -<span class="sourceLineNo">1207</span> * zero if not writing the block at the moment. Note that this will return<a name="line.1207"></a> -<span class="sourceLineNo">1208</span> * zero in the "block ready" state as well.<a name="line.1208"></a> -<span class="sourceLineNo">1209</span> *<a name="line.1209"></a> -<span class="sourceLineNo">1210</span> * @return the number of bytes written<a name="line.1210"></a> -<span class="sourceLineNo">1211</span> */<a name="line.1211"></a> -<span class="sourceLineNo">1212</span> public int encodedBlockSizeWritten() {<a name="line.1212"></a> -<span class="sourceLineNo">1213</span> if (state != State.WRITING)<a name="line.1213"></a> -<span class="sourceLineNo">1214</span> return 0;<a name="line.1214"></a> -<span class="sourceLineNo">1215</span> return this.encodedDataSizeWritten;<a name="line.1215"></a> -<span class="sourceLineNo">1216</span> }<a name="line.1216"></a> -<span class="sourceLineNo">1217</span><a name="line.1217"></a> -<span class="sourceLineNo">1218</span> /**<a name="line.1218"></a> -<span class="sourceLineNo">1219</span> * Returns the number of bytes written into the current block so far, or<a name="line.1219"></a> -<span class="sourceLineNo">1220</span> * zero if not writing the block at the moment. Note that this will return<a name="line.1220"></a> -<span class="sourceLineNo">1221</span> * zero in the "block ready" state as well.<a name="line.1221"></a> -<span class="sourceLineNo">1222</span> *<a name="line.1222"></a> -<span class="sourceLineNo">1223</span> * @return the number of bytes written<a name="line.1223"></a> -<span class="sourceLineNo">1224</span> */<a name="line.1224"></a> -<span class="sourceLineNo">1225</span> int blockSizeWritten() {<a name="line.1225"></a> -<span class="sourceLineNo">1226</span> if (state != State.WRITING) return 0;<a name="line.1226"></a> -<span class="sourceLineNo">1227</span> return this.unencodedDataSizeWritten;<a name="line.1227"></a> -<span class="sourceLineNo">1228</span> }<a name="line.1228"></a> -<span class="sourceLineNo">1229</span><a name="line.1229"></a> -<span class="sourceLineNo">1230</span> /**<a name="line.1230"></a> -<span class="sourceLineNo">1231</span> * Clones the header followed by the uncompressed data, even if using<a name=
<TRUNCATED>