/* Copyright (c) 2011 Danish Maritime Authority. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. * You may obtain a copy of the License at * * http://www.apache.org/licenses/LICENSE-2.0 * * Unless required by applicable law or agreed to in writing, software * distributed under the License is distributed on an "AS IS" BASIS, * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. * See the License for the specific language governing permissions and * limitations under the License. */ // Protocol Buffers - Google's data interchange format // Copyright 2008 Google Inc. All rights reserved. // http://code.google.com/p/protobuf/ // // Redistribution and use in source and binary forms, with or without // modification, are permitted provided that the following conditions are // met: // // * Redistributions of source code must retain the above copyright // notice, this list of conditions and the following disclaimer. // * Redistributions in binary form must reproduce the above // copyright notice, this list of conditions and the following disclaimer // in the documentation and/or other materials provided with the // distribution. // * Neither the name of Google Inc. nor the names of its // contributors may be used to endorse or promote products derived from // this software without specific prior written permission. // // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS // "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT // LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR // A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT // OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, // SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT // LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, // DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY // THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT // (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. package net.maritimecloud.util; import java.io.ByteArrayOutputStream; import java.io.IOException; import java.io.InputStream; import java.io.OutputStream; import java.io.UnsupportedEncodingException; import java.nio.ByteBuffer; import java.security.MessageDigest; import java.security.NoSuchAlgorithmException; import java.util.ArrayList; import java.util.Base64; import java.util.Collection; import java.util.Iterator; import java.util.List; import java.util.NoSuchElementException; import java.util.concurrent.ThreadLocalRandom; import javax.xml.bind.DatatypeConverter; /** * Immutable sequence of bytes. Substring is supported by sharing the reference to the immutable underlying bytes, as * with {@link String} (previous to Java 7). Concatenation is likewise supported without copying (long strings) by * building a tree of pieces in {@link RopeBinary}. * <p> * Like {@link String}, the contents of a {@link Binary} can never be observed to change, not even in the presence of a * data race or incorrect API usage in the client code. * * @author crazybob@google.com Bob Lee * @author kenton@google.com Kenton Varda * @author carlanton@google.com Carl Haverl * @author martinrb@google.com Martin Buchholz * @author kasperni@gmail.com Kasper Nielsen */ // CHECKSTYLE:OFF public abstract class Binary implements Iterable<Byte>, Comparable<Binary> { /** * When two strings to be concatenated have a combined length shorter than this, we just copy their bytes on * {@link #concat(Binary)}. The trade-off is copy size versus the overhead of creating tree nodes in * {@link RopeBinary}. */ static final int CONCATENATE_BY_COPY_SIZE = 128; /** An empty {@code Binary} with size = 0. */ public static final Binary EMPTY = new LiteralBinary(new byte[0]); static final int MAX_READ_FROM_CHUNK_SIZE = 0x2000; // 8k /** * When copying an InputStream into a Binary with .readFrom(), the chunks in the underlying rope start at 256 bytes, * but double each iteration up to 8192 bytes. */ static final int MIN_READ_FROM_CHUNK_SIZE = 0x100; // 256b /** Prevent subclassing outside of this package. */ Binary() {} /** * Constructs a read-only {@code java.nio.ByteBuffer} whose content is equal to the contents of this byte string. * The result uses the same backing array as the byte string, if possible. * * @return wrapped bytes */ public abstract ByteBuffer asReadOnlyByteBuffer(); /** * Constructs a list of read-only {@code java.nio.ByteBuffer} objects such that the concatenation of their contents * is equal to the contents of this byte string. The result uses the same backing arrays as the byte string. * <p> * By returning a list, implementations of this method may be able to avoid copying even when there are multiple * backing arrays. * * @return a list of wrapped bytes */ public abstract List<ByteBuffer> asReadOnlyByteBufferList(); public String base64encode() { return Base64.getEncoder().encodeToString(toByteArray()); } public String hexString() { return DatatypeConverter.printHexBinary(toByteArray()); } /** * Gets the byte at the given index. This method should be used only for random access to individual bytes. To * access bytes sequentially, use the {@link ByteIterator} returned by {@link #iterator()}, and call * {@link #substring(int, int)} first if necessary. * * @param index * index of byte * @return the value * @throws ArrayIndexOutOfBoundsException * {@code index} is less than 0 or greater than or equal to size */ public abstract byte byteAt(int index); /** {@inheritDoc} */ @Override public int compareTo(Binary o) { if (size() != o.size()) { throw new IllegalArgumentException("Can only compare binaries with identical length, this.length =" + size() + ", other.length =" + o.size()); } ByteIterator thisIter = iterator(); ByteIterator otherIter = o.iterator(); while (thisIter.hasNext()) { byte tb = thisIter.nextByte(); byte ob = otherIter.nextByte(); if (tb != ob) { return Byte.compare(tb, ob); } } return 0; } /** * Concatenate the given {@code Binary} to this one. Short concatenations, of total size smaller than * {@link Binary#CONCATENATE_BY_COPY_SIZE}, are produced by copying the underlying bytes (as per Rope.java, <a * href="http://www.cs.ubc.ca/local/reading/proceedings/spe91-95/spe/vol25/issue12/spe986.pdf"> BAP95 </a>. In * general, the concatenate involves no copying. * * @param other * string to concatenate * @return a new {@code Binary} instance */ public Binary concat(Binary other) { int thisSize = size(); int otherSize = other.size(); if ((long) thisSize + otherSize >= Integer.MAX_VALUE) { throw new IllegalArgumentException("Binary would be too long: " + thisSize + "+" + otherSize); } return RopeBinary.concatenate(this, other); } /** * Copies bytes into a buffer at the given offset. * * @param target * buffer to copy into * @param offset * in the target buffer * @throws IndexOutOfBoundsException * if the offset is negative or too large */ public void copyTo(byte[] target, int offset) { copyTo(target, 0, offset, size()); } // ================================================================= // Binary -> substring /** * Copies bytes into a buffer. * * @param target * buffer to copy into * @param sourceOffset * offset within these bytes * @param targetOffset * offset within the target buffer * @param numberToCopy * number of bytes to copy * @throws IndexOutOfBoundsException * if an offset or size is negative or too large */ public void copyTo(byte[] target, int sourceOffset, int targetOffset, int numberToCopy) { if (sourceOffset < 0) { throw new IndexOutOfBoundsException("Source offset < 0: " + sourceOffset); } if (targetOffset < 0) { throw new IndexOutOfBoundsException("Target offset < 0: " + targetOffset); } if (numberToCopy < 0) { throw new IndexOutOfBoundsException("Length < 0: " + numberToCopy); } if (sourceOffset + numberToCopy > size()) { throw new IndexOutOfBoundsException("Source end offset < 0: " + (sourceOffset + numberToCopy)); } if (targetOffset + numberToCopy > target.length) { throw new IndexOutOfBoundsException("Target end offset < 0: " + (targetOffset + numberToCopy)); } if (numberToCopy > 0) { copyToInternal(target, sourceOffset, targetOffset, numberToCopy); } } /** * Copies bytes into a ByteBuffer. * * @param target * ByteBuffer to copy into. * @throws java.nio.ReadOnlyBufferException * if the {@code target} is read-only * @throws java.nio.BufferOverflowException * if the {@code target}'s remaining() space is not large enough to hold the data. */ public abstract void copyTo(ByteBuffer target); /** * Internal (package private) implementation of {@link #copyTo(byte[], int, int, int)}. It assumes that all error * checking has already been performed and that * * @param target * sd * @param sourceOffset * source * @param targetOffset * target * @param numberToCopy * numer */ protected abstract void copyToInternal(byte[] target, int sourceOffset, int targetOffset, int numberToCopy); // ================================================================= // byte[] -> Binary @Override public abstract boolean equals(Object o); /** * Return the depth of the tree representing this {@code Binary}, if any, whose root is this node. If this is a leaf * node, return 0. * * @return tree depth or zero */ protected abstract int getTreeDepth(); /** * Return a non-zero hashCode depending only on the sequence of bytes in this Binary. * * @return hashCode value for this object */ @Override public abstract int hashCode(); /** * Return {@code true} if this Binary is literal (a leaf node) or a flat-enough tree in the sense of * {@link RopeBinary}. * * @return true if the tree is flat enough */ protected abstract boolean isBalanced(); /** * Returns {@code true} if the size is {@code 0}, {@code false} otherwise. * * @return true if this is zero bytes long */ public boolean isEmpty() { return size() == 0; } /** * Tells whether this {@code Binary} represents a well-formed UTF-8 byte sequence, such that the original bytes can * be converted to a String object and then round tripped back to bytes without loss. * * <p> * More precisely, returns {@code true} whenever: * * <pre> * {@code * Arrays.equals(byteString.toByteArray(), * new String(byteString.toByteArray(), "UTF-8").getBytes("UTF-8")) * } * </pre> * * <p> * This method returns {@code false} for "overlong" byte sequences, as well as for 3-byte sequences that would map * to a surrogate character, in accordance with the restricted definition of UTF-8 introduced in Unicode 3.1. Note * that the UTF-8 decoder included in Oracle's JDK has been modified to also reject "overlong" byte sequences, but * (as of 2011) still accepts 3-byte surrogate character byte sequences. * * <p> * See the Unicode Standard, Table 3-6. <em>UTF-8 Bit Distribution</em>, Table 3-7. * <em>Well Formed UTF-8 Byte Sequences</em>. * * @return whether the bytes in this {@code Binary} are a well-formed UTF-8 byte sequence */ public abstract boolean isValidUtf8(); /** * Return a {@link Binary.ByteIterator} over the bytes in the Binary. To avoid auto-boxing, you may get the iterator * manually and call {@link ByteIterator#nextByte()}. * * @return the iterator */ public abstract ByteIterator iterator(); // ================================================================= // InputStream -> Binary /** * Creates an {@code InputStream} which can be used to read the bytes. * <p> * The {@link InputStream} returned by this method is guaranteed to be completely non-blocking. The method * {@link InputStream#available()} returns the number of bytes remaining in the stream. The methods * {@link InputStream#read(byte[])}, {@link InputStream#read(byte[],int,int)} and {@link InputStream#skip(long)} * will read/skip as many bytes as are available. * <p> * The methods in the returned {@link InputStream} might <b>not</b> be thread safe. * * @return an input stream that returns the bytes of this byte string. */ public abstract InputStream newInput(); /** * Compute the hash across the value bytes starting with the given hash, and return the result. This is used to * compute the hash across strings represented as a set of pieces by allowing the hash computation to be continued * from piece to piece. * * @param h * starting hash value * @param offset * offset into this value to start looking at data values * @param length * number of data values to include in the hash computation * @return ending hash value */ protected abstract int partialHash(int h, int offset, int length); /** * Tells whether the given byte sequence is a well-formed, malformed, or incomplete UTF-8 byte sequence. This method * accepts and returns a partial state result, allowing the bytes for a complete UTF-8 byte sequence to be composed * from multiple {@code Binary} segments. * * @param state * either {@code 0} (if this is the initial decoding operation) or the value returned from a call to a * partial decoding method for the previous bytes * @param offset * offset of the first byte to check * @param length * number of bytes to check * * @return {@code -1} if the partial byte sequence is definitely malformed, {@code 0} if it is well-formed (no * additional input needed), or, if the byte sequence is "incomplete", i.e. apparently terminated in the * middle of a character, an opaque integer "state" value containing enough information to decode the * character when passed to a subsequent invocation of a partial decoding method. */ protected abstract int partialIsValidUtf8(int state, int offset, int length); /** * Return the cached hash code if available. * * @return value of cached hash code or 0 if not computed yet */ protected abstract int peekCachedHashCode(); // ================================================================= // Multiple Binarys -> One Binary /** * Returns a SHA-256 hash of this binary. * * @return the hash */ public final Binary sha256() { MessageDigest md; try { md = MessageDigest.getInstance("SHA-256"); } catch (NoSuchAlgorithmException e) { throw new RuntimeException("All java implementations should have SHA-256", e); } md.update(toByteArray()); byte[] digest = md.digest(); return new LiteralBinary(digest); } /** * Gets the number of bytes. * * @return size in bytes */ public abstract int size(); /** * Tests if this bytestring starts with the specified prefix. Similar to {@link String#startsWith(String)} * * @param prefix * the prefix. * @return <code>true</code> if the byte sequence represented by the argument is a prefix of the byte sequence * represented by this string; <code>false</code> otherwise. */ public boolean startsWith(Binary prefix) { return size() >= prefix.size() && substring(0, prefix.size()).equals(prefix); } // ================================================================= // Binary -> byte[] /** * Return the substring from {@code beginIndex}, inclusive, to the end of the string. * * @param beginIndex * start at this index * @return substring sharing underlying data * @throws IndexOutOfBoundsException * if {@code beginIndex < 0} or {@code beginIndex > size()}. */ public Binary substring(int beginIndex) { return substring(beginIndex, size()); } /** * Return the substring from {@code beginIndex}, inclusive, to {@code endIndex}, exclusive. * * @param beginIndex * start at this index * @param endIndex * the last character is the one before this index * @return substring sharing underlying data * @throws IndexOutOfBoundsException * if {@code beginIndex < 0}, {@code endIndex > size()}, or {@code beginIndex > endIndex}. */ public abstract Binary substring(int beginIndex, int endIndex); /** * Copies bytes to a {@code byte[]}. * * @return copied bytes */ public byte[] toByteArray() { int size = size(); byte[] result = new byte[size]; copyToInternal(result, 0, 0, size); return result; } @Override public String toString() { return String.format("<Binary@%s size=%d>", Integer.toHexString(System.identityHashCode(this)), size()); } /** * Constructs a new {@code String} by decoding the bytes using the specified charset. * * @param charsetName * encode using this charset * @return new string * @throws UnsupportedEncodingException * if charset isn't recognized */ public abstract String toString(String charsetName) throws UnsupportedEncodingException; /** * Constructs a new {@code String} by decoding the bytes as UTF-8. * * @return new string using UTF-8 encoding */ public String toStringUtf8() { try { return toString("UTF-8"); } catch (UnsupportedEncodingException e) { throw new RuntimeException("UTF-8 not supported?", e); } } /** * Writes the complete contents of this byte string to the specified output stream argument. * * @param out * the output stream to which to write the data. * @throws IOException * if an I/O error occurs. */ public abstract void writeTo(OutputStream out) throws IOException; // Internal function used by copyFrom(Iterable<Binary>). // Create a balanced concatenation of the next "length" elements from the // iterable. private static Binary balancedConcat(Iterator<Binary> iterator, int length) { assert length >= 1; Binary result; if (length == 1) { result = iterator.next(); } else { int halfLength = length >>> 1; Binary left = balancedConcat(iterator, halfLength); Binary right = balancedConcat(iterator, length - halfLength); result = left.concat(right); } return result; } /** * Copies the given bytes into a {@code Binary}. * * @param bytes * to copy * @return new {@code Binary} */ public static Binary copyFrom(byte... bytes) { return copyFrom(bytes, 0, bytes.length); } // ================================================================= // UTF-8 decoding /** * Copies the given bytes into a {@code Binary}. * * @param bytes * source array * @param offset * offset in source array * @param size * number of bytes to copy * @return new {@code Binary} */ public static Binary copyFrom(byte[] bytes, int offset, int size) { byte[] copy = new byte[size]; System.arraycopy(bytes, offset, copy, 0, size); return new LiteralBinary(copy); } /** * Copies the remaining bytes from a {@code java.nio.ByteBuffer} into a {@code Binary}. * * @param bytes * sourceBuffer * @return new {@code Binary} */ public static Binary copyFrom(ByteBuffer bytes) { return copyFrom(bytes, bytes.remaining()); } /** * Copies the next {@code size} bytes from a {@code java.nio.ByteBuffer} into a {@code Binary}. * * @param bytes * source buffer * @param size * number of bytes to copy * @return new {@code Binary} */ public static Binary copyFrom(ByteBuffer bytes, int size) { byte[] copy = new byte[size]; bytes.get(copy); return new LiteralBinary(copy); } // ================================================================= // equals() and hashCode() /** * Concatenates all byte strings in the iterable and returns the result. This is designed to run in O(list size), * not O(total bytes). * * <p> * The returned {@code Binary} is not necessarily a unique object. If the list is empty, the returned object is the * singleton empty {@code Binary}. If the list has only one element, that {@code Binary} will be returned without * copying. * * @param byteStrings * strings to be concatenated * @return new {@code Binary} */ public static Binary copyFrom(Iterable<Binary> byteStrings) { Collection<Binary> collection; if (!(byteStrings instanceof Collection)) { collection = new ArrayList<>(); for (Binary byteString : byteStrings) { collection.add(byteString); } } else { collection = (Collection<Binary>) byteStrings; } Binary result; if (collection.isEmpty()) { result = EMPTY; } else { result = balancedConcat(collection.iterator(), collection.size()); } return result; } /** * Encodes {@code text} into a sequence of bytes using the named charset and returns the result as a {@code Binary}. * * @param text * source string * @param charsetName * encoding to use * @return new {@code Binary} * @throws UnsupportedEncodingException * if the encoding isn't found */ public static Binary copyFrom(String text, String charsetName) throws UnsupportedEncodingException { return new LiteralBinary(text.getBytes(charsetName)); } // ================================================================= // Input stream public static Binary copyFromHex(String text) { return copyFrom(DatatypeConverter.parseHexBinary(text)); } public static Binary copyFromBase64(String text) { return copyFrom(Base64.getDecoder().decode(text)); } /** * Creates a {@link CodedInputStream} which can be used to read the bytes. Using this is often more efficient than * creating a {@link CodedInputStream} that wraps the result of {@link #newInput()}. * * @return stream based on wrapped data */ // public abstract CodedInputStream newCodedInput(); // ================================================================= // Output stream /** * Encodes {@code text} into a sequence of UTF-8 bytes and returns the result as a {@code Binary}. * * @param text * source string * @return new {@code Binary} */ public static Binary copyFromUtf8(String text) { try { return new LiteralBinary(text.getBytes("UTF-8")); } catch (UnsupportedEncodingException e) { throw new RuntimeException("UTF-8 not supported?", e); } } /** * Creates a new {@link Output}. Call {@link Output#toBinary()} to create the {@code Binary} instance. * <p> * A {@link Binary.Output} offers the same functionality as a {@link ByteArrayOutputStream}, except that it returns * a {@link Binary} rather than a {@code byte array}. * * @return {@code OutputStream} for building a {@code Binary} */ public static Output newOutput() { return new Output(CONCATENATE_BY_COPY_SIZE); } /** * Creates a new {@link Output} with the given initial capacity. Call {@link Output#toBinary()} to create the * {@code Binary} instance. * <p> * A {@link Binary.Output} offers the same functionality as a {@link ByteArrayOutputStream}, except that it returns * a {@link Binary} rather than a {@code byte} array. * * @param initialCapacity * estimate of number of bytes to be written * @return {@code OutputStream} for building a {@code Binary} */ public static Output newOutput(int initialCapacity) { return new Output(initialCapacity); } // /** // * Constructs a new {@code Binary} builder, which allows you to efficiently construct a {@code Binary} by // * writing to a {@link CodedOutputStream}. Using this is much more efficient than calling {@code newOutput()} and // * wrapping that in a {@code CodedOutputStream}. // * // * <p> // * This is package-private because it's a somewhat confusing interface. Users can call // * {@link Message#toBinary()} instead of calling this directly. // * // * @param size // * The target byte size of the {@code Binary}. You must write exactly this many bytes before building // * the result. // * @return the builder // */ // public static CodedBuilder newCodedBuilder(int size) { // return new CodedBuilder(size); // } // // /** See {@link Binary#newCodedBuilder(int)}. */ // public static final class CodedBuilder { // private final CodedOutputStream output; // // private final byte[] buffer; // // private CodedBuilder(int size) { // buffer = new byte[size]; // output = CodedOutputStream.newInstance(buffer); // } // // public Binary build() { // output.checkNoSpaceLeft(); // // // We can be confident that the CodedOutputStream will not modify the // // underlying bytes anymore because it already wrote all of them. So, // // no need to make a copy. // return new LiteralBinary(buffer); // } // // public CodedOutputStream getCodedOutput() { // return output; // } // } // ================================================================= // Methods {@link RopeBinary} needs on instances, which aren't part of the // public API. public static Binary random(int bytes) { byte[] b = new byte[bytes]; ThreadLocalRandom.current().nextBytes(b); return Binary.copyFrom(b); } /** * Blocks until a chunk of the given size can be made from the stream, or EOF is reached. Calls read() repeatedly in * case the given stream implementation doesn't completely fill the given buffer in one read() call. * * @return A chunk of the desired size, or else a chunk as large as was available when end of stream was reached. * Returns null if the given stream had no more data in it. */ private static Binary readChunk(InputStream in, final int chunkSize) throws IOException { final byte[] buf = new byte[chunkSize]; int bytesRead = 0; while (bytesRead < chunkSize) { final int count = in.read(buf, bytesRead, chunkSize - bytesRead); if (count == -1) { break; } bytesRead += count; } if (bytesRead == 0) { return null; } else { return Binary.copyFrom(buf, 0, bytesRead); } } /** * Completely reads the given stream's bytes into a {@code Binary}, blocking if necessary until all bytes are read * through to the end of the stream. * * <b>Performance notes:</b> The returned {@code Binary} is an immutable tree of byte arrays ("chunks") of the * stream data. The first chunk is small, with subsequent chunks each being double the size, up to 8K. If the caller * knows the precise length of the stream and wishes to avoid all unnecessary copies and allocations, consider using * the two-argument version of this method, below. * * @param streamToDrain * The source stream, which is read completely but not closed. * @return A new {@code Binary} which is made up of chunks of various sizes, depending on the behavior of the * underlying stream. * @throws IOException * IOException is thrown if there is a problem reading the underlying stream. */ public static Binary readFrom(InputStream streamToDrain) throws IOException { return readFrom(streamToDrain, MIN_READ_FROM_CHUNK_SIZE, MAX_READ_FROM_CHUNK_SIZE); } /** * Completely reads the given stream's bytes into a {@code Binary}, blocking if necessary until all bytes are read * through to the end of the stream. * * <b>Performance notes:</b> The returned {@code Binary} is an immutable tree of byte arrays ("chunks") of the * stream data. The chunkSize parameter sets the size of these byte arrays. In particular, if the chunkSize is * precisely the same as the length of the stream, unnecessary allocations and copies will be avoided. Otherwise, * the chunks will be of the given size, except for the last chunk, which will be resized (via a reallocation and * copy) to contain the remainder of the stream. * * @param streamToDrain * The source stream, which is read completely but not closed. * @param chunkSize * The size of the chunks in which to read the stream. * @return A new {@code Binary} which is made up of chunks of the given size. * @throws IOException * IOException is thrown if there is a problem reading the underlying stream. */ public static Binary readFrom(InputStream streamToDrain, int chunkSize) throws IOException { return readFrom(streamToDrain, chunkSize, chunkSize); } // Helper method that takes the chunk size range as a parameter. public static Binary readFrom(InputStream streamToDrain, int minChunkSize, int maxChunkSize) throws IOException { Collection<Binary> results = new ArrayList<>(); // copy the inbound bytes into a list of chunks; the chunk size // grows exponentially to support both short and long streams. int chunkSize = minChunkSize; while (true) { Binary chunk = readChunk(streamToDrain, chunkSize); if (chunk == null) { break; } results.add(chunk); chunkSize = Math.min(chunkSize * 2, maxChunkSize); } return Binary.copyFrom(results); } /** * This interface extends {@code Iterator<Byte>}, so that we can return an unboxed {@code byte}. */ public interface ByteIterator extends Iterator<Byte> { /** * An alternative to {@link Iterator#next()} that returns an unboxed primitive {@code byte}. * * @return the next {@code byte} in the iteration * @throws NoSuchElementException * if the iteration has no more elements */ byte nextByte(); } /** * Outputs to a {@code Binary} instance. Call {@link #toBinary()} to create the {@code Binary} instance. */ public static final class Output extends OutputStream { // Implementation note. // The public methods of this class must be synchronized. Binarys // are guaranteed to be immutable. Without some sort of locking, it could // be possible for one thread to call toByteSring(), while another thread // is still modifying the underlying byte array. private static final byte[] EMPTY_BYTE_ARRAY = new byte[0]; // Current buffer to which we are writing private byte[] buffer; // Location in buffer[] to which we write the next byte. private int bufferPos; // Binarys to be concatenated to create the result private final ArrayList<Binary> flushedBuffers; // Total number of bytes in the Binarys of flushedBuffers private int flushedBuffersTotalBytes; // argument passed by user, indicating initial capacity. private final int initialCapacity; /** * Creates a new Binary output stream with the specified initial capacity. * * @param initialCapacity * the initial capacity of the output stream. */ Output(int initialCapacity) { if (initialCapacity < 0) { throw new IllegalArgumentException("Buffer size < 0"); } this.initialCapacity = initialCapacity; this.flushedBuffers = new ArrayList<>(); this.buffer = new byte[initialCapacity]; } /** * Implement java.util.Arrays.copyOf() for jdk 1.5. */ private byte[] copyArray(byte[] buffer, int length) { byte[] result = new byte[length]; System.arraycopy(buffer, 0, result, 0, Math.min(buffer.length, length)); return result; } /** * Internal function used by writers. The current buffer is full, and the writer needs a new buffer whose size * is at least the specified minimum size. */ private void flushFullBuffer(int minSize) { flushedBuffers.add(new LiteralBinary(buffer)); flushedBuffersTotalBytes += buffer.length; // We want to increase our total capacity by 50%, but as a minimum, // the new buffer should also at least be >= minSize and // >= initial Capacity. int newSize = Math.max(initialCapacity, Math.max(minSize, flushedBuffersTotalBytes >>> 1)); buffer = new byte[newSize]; bufferPos = 0; } /** * Internal function used by {@link #toBinary()}. The current buffer may or may not be full, but it needs to be * flushed. */ private void flushLastBuffer() { if (bufferPos < buffer.length) { if (bufferPos > 0) { byte[] bufferCopy = copyArray(buffer, bufferPos); flushedBuffers.add(new LiteralBinary(bufferCopy)); } // We reuse this buffer for further writes. } else { // Buffer is completely full. Huzzah. flushedBuffers.add(new LiteralBinary(buffer)); // 99% of the time, we're not going to use this OutputStream again. // We set buffer to an empty byte stream so that we're handling this // case without wasting space. In the rare case that more writes // *do* occur, this empty buffer will be flushed and an appropriately // sized new buffer will be created. buffer = EMPTY_BYTE_ARRAY; } flushedBuffersTotalBytes += bufferPos; bufferPos = 0; } /** * Resets this stream, so that all currently accumulated output in the output stream is discarded. The output * stream can be used again, reusing the already allocated buffer space. */ public synchronized void reset() { flushedBuffers.clear(); flushedBuffersTotalBytes = 0; bufferPos = 0; } /** * Returns the current size of the output stream. * * @return the current size of the output stream */ public synchronized int size() { return flushedBuffersTotalBytes + bufferPos; } /** * Creates a byte string. Its size is the current size of this output stream and its output has been copied to * it. * * @return the current contents of this output stream, as a byte string. */ public synchronized Binary toBinary() { flushLastBuffer(); return Binary.copyFrom(flushedBuffers); } @Override public String toString() { return String.format("<Binary.Output@%s size=%d>", Integer.toHexString(System.identityHashCode(this)), size()); } @Override public synchronized void write(byte[] b, int offset, int length) { if (length <= buffer.length - bufferPos) { // The bytes can fit into the current buffer. System.arraycopy(b, offset, buffer, bufferPos, length); bufferPos += length; } else { // Use up the current buffer int copySize = buffer.length - bufferPos; System.arraycopy(b, offset, buffer, bufferPos, copySize); offset += copySize; length -= copySize; // Flush the buffer, and get a new buffer at least big enough to cover // what we still need to output flushFullBuffer(length); System.arraycopy(b, offset, buffer, 0 /* count */, length); bufferPos = length; } } @Override public synchronized void write(int b) { if (bufferPos == buffer.length) { flushFullBuffer(1); } buffer[bufferPos++] = (byte) b; } /** * Writes the complete contents of this byte array output stream to the specified output stream argument. * * @param out * the output stream to which to write the data. * @throws IOException * if an I/O error occurs. */ public void writeTo(OutputStream out) throws IOException { Binary[] cachedFlushBuffers; byte[] cachedBuffer; int cachedBufferPos; synchronized (this) { // Copy the information we need into local variables so as to hold // the lock for as short a time as possible. cachedFlushBuffers = flushedBuffers.toArray(new Binary[flushedBuffers.size()]); cachedBuffer = buffer; cachedBufferPos = bufferPos; } for (Binary byteString : cachedFlushBuffers) { byteString.writeTo(out); } out.write(copyArray(cachedBuffer, cachedBufferPos)); } } } // CHECKSTYLE:ON