zipkin2.internal.WriteBuffer Maven / Gradle / Ivy
/*
* Copyright The OpenZipkin Authors
* SPDX-License-Identifier: Apache-2.0
*/
package zipkin2.internal;
import static zipkin2.internal.HexCodec.HEX_DIGITS;
/**
* Writes are unsafe as they do no bounds checks. This means you should take care to allocate or
* wrap an array at least as big as you need prior to writing. As it is possible to calculate size
* prior to writing, overrunning a buffer is a programming error.
*/
public final class WriteBuffer {
public interface Writer {
int sizeInBytes(T value);
void write(T value, WriteBuffer buffer);
}
public static WriteBuffer wrap(byte[] bytes) {
return wrap(bytes, 0);
}
public static WriteBuffer wrap(byte[] bytes, int pos) {
return new WriteBuffer(bytes, pos);
}
final byte[] buf;
int pos;
WriteBuffer(byte[] buf, int pos) {
this.buf = buf;
this.pos = pos;
}
public void writeByte(int v) {
buf[pos++] = (byte) (v & 0xff);
}
public void write(byte[] v) {
System.arraycopy(v, 0, buf, pos, v.length);
pos += v.length;
}
void writeBackwards(long v) {
int lastPos = pos + asciiSizeInBytes(v); // We write backwards from right to left.
pos = lastPos;
while (v != 0) {
int digit = (int) (v % 10);
buf[--lastPos] = (byte) HEX_DIGITS[digit];
v /= 10;
}
}
/** Inspired by {@code okio.Buffer.writeLong} */
public void writeLongHex(long v) {
int pos = this.pos;
writeHexByte(buf, pos + 0, (byte) ((v >>> 56L) & 0xff));
writeHexByte(buf, pos + 2, (byte) ((v >>> 48L) & 0xff));
writeHexByte(buf, pos + 4, (byte) ((v >>> 40L) & 0xff));
writeHexByte(buf, pos + 6, (byte) ((v >>> 32L) & 0xff));
writeHexByte(buf, pos + 8, (byte) ((v >>> 24L) & 0xff));
writeHexByte(buf, pos + 10, (byte) ((v >>> 16L) & 0xff));
writeHexByte(buf, pos + 12, (byte) ((v >>> 8L) & 0xff));
writeHexByte(buf, pos + 14, (byte) (v & 0xff));
this.pos = pos + 16;
}
static void writeHexByte(byte[] data, int pos, byte b) {
data[pos + 0] = (byte) HEX_DIGITS[(b >> 4) & 0xf];
data[pos + 1] = (byte) HEX_DIGITS[b & 0xf];
}
int pos() {
return pos;
}
public void writeAscii(String v) {
for (int i = 0, length = v.length(); i < length; i++) {
writeByte(v.charAt(i) & 0xff);
}
}
/**
* This transcodes a UTF-16 Java String to UTF-8 bytes.
*
* This looks most similar to {@code io.netty.buffer.ByteBufUtil.writeUtf8(AbstractByteBuf,
* int, CharSequence, int)} v4.1, modified including features to address ASCII runs of text.
*/
public void writeUtf8(CharSequence string) {
for (int i = 0, len = string.length(); i < len; i++) {
char ch = string.charAt(i);
if (ch < 0x80) { // 7-bit ASCII character
writeByte(ch);
// This could be an ASCII run, or possibly entirely ASCII
while (i < len - 1) {
ch = string.charAt(i + 1);
if (ch >= 0x80) break;
i++;
writeByte(ch); // another 7-bit ASCII character
}
} else if (ch < 0x800) { // 11-bit character
writeByte(0xc0 | (ch >> 6));
writeByte(0x80 | (ch & 0x3f));
} else if (ch < 0xd800 || ch > 0xdfff) { // 16-bit character
writeByte(0xe0 | (ch >> 12));
writeByte(0x80 | ((ch >> 6) & 0x3f));
writeByte(0x80 | (ch & 0x3f));
} else { // Possibly a 21-bit character
if (!Character.isHighSurrogate(ch)) { // Malformed or not UTF-8
writeByte('?');
continue;
}
if (i == len - 1) { // Truncated or not UTF-8
writeByte('?');
break;
}
char low = string.charAt(++i);
if (!Character.isLowSurrogate(low)) { // Malformed or not UTF-8
writeByte('?');
writeByte(Character.isHighSurrogate(low) ? '?' : low);
continue;
}
// Write the 21-bit character using 4 bytes
// See http://www.unicode.org/versions/Unicode7.0.0/ch03.pdf#G2630
int codePoint = Character.toCodePoint(ch, low);
writeByte(0xf0 | (codePoint >> 18));
writeByte(0x80 | ((codePoint >> 12) & 0x3f));
writeByte(0x80 | ((codePoint >> 6) & 0x3f));
writeByte(0x80 | (codePoint & 0x3f));
}
}
}
// Adapted from okio.Buffer.writeDecimalLong
public void writeAscii(long v) {
if (v == 0) {
writeByte('0');
return;
}
if (v == Long.MIN_VALUE) {
writeAscii("-9223372036854775808");
return;
}
if (v < 0) {
writeByte('-');
v = -v; // needs to be positive so we can use this for an array index
}
writeBackwards(v);
}
// com.squareup.wire.ProtoWriter.writeVarint v2.3.0
void writeVarint(int v) {
while ((v & ~0x7f) != 0) {
writeByte((byte) ((v & 0x7f) | 0x80));
v >>>= 7;
}
writeByte((byte) v);
}
// com.squareup.wire.ProtoWriter.writeVarint v2.3.0
void writeVarint(long v) {
while ((v & ~0x7fL) != 0) {
writeByte((byte) ((v & 0x7f) | 0x80));
v >>>= 7;
}
writeByte((byte) v);
}
void writeLongLe(long v) {
writeByte((byte) (v & 0xff));
writeByte((byte) ((v >> 8) & 0xff));
writeByte((byte) ((v >> 16) & 0xff));
writeByte((byte) ((v >> 24) & 0xff));
writeByte((byte) ((v >> 32) & 0xff));
writeByte((byte) ((v >> 40) & 0xff));
writeByte((byte) ((v >> 48) & 0xff));
writeByte((byte) ((v >> 56) & 0xff));
}
/**
* This returns the bytes needed to transcode a UTF-16 Java String to UTF-8 bytes.
*
*
Originally based on
* http://stackoverflow.com/questions/8511490/calculating-length-in-utf-8-of-java-string-without-actually-encoding-it
*
*
Later, ASCII run and malformed surrogate logic borrowed from okio.Utf8
*/
// TODO: benchmark vs https://github.com/protocolbuffers/protobuf/blob/master/java/core/src/main/java/com/google/protobuf/Utf8.java#L240
// there seem to be less branches for for strings without surrogates
public static int utf8SizeInBytes(CharSequence string) {
int sizeInBytes = 0;
for (int i = 0, len = string.length(); i < len; i++) {
char ch = string.charAt(i);
if (ch < 0x80) {
sizeInBytes++; // 7-bit ASCII character
// This could be an ASCII run, or possibly entirely ASCII
while (i < len - 1) {
ch = string.charAt(i + 1);
if (ch >= 0x80) break;
i++;
sizeInBytes++; // another 7-bit ASCII character
}
} else if (ch < 0x800) {
sizeInBytes += 2; // 11-bit character
} else if (ch < 0xd800 || ch > 0xdfff) {
sizeInBytes += 3; // 16-bit character
} else {
int low = i + 1 < len ? string.charAt(i + 1) : 0;
if (ch > 0xdbff || low < 0xdc00 || low > 0xdfff) {
sizeInBytes++; // A malformed surrogate, which yields '?'.
} else {
// A 21-bit character
sizeInBytes += 4;
i++;
}
}
}
return sizeInBytes;
}
/**
* Binary search for character width which favors matching lower numbers.
*
*
Adapted from okio.Buffer
*/
public static int asciiSizeInBytes(long v) {
if (v == 0) return 1;
if (v == Long.MIN_VALUE) return 20;
boolean negative = false;
if (v < 0) {
v = -v; // making this positive allows us to compare using less-than
negative = true;
}
int width =
v < 100000000L
? v < 10000L
? v < 100L ? v < 10L ? 1 : 2 : v < 1000L ? 3 : 4
: v < 1000000L ? v < 100000L ? 5 : 6 : v < 10000000L ? 7 : 8
: v < 1000000000000L
? v < 10000000000L ? v < 1000000000L ? 9 : 10 : v < 100000000000L ? 11 : 12
: v < 1000000000000000L
? v < 10000000000000L ? 13 : v < 100000000000000L ? 14 : 15
: v < 100000000000000000L
? v < 10000000000000000L ? 16 : 17
: v < 1000000000000000000L ? 18 : 19;
return negative ? width + 1 : width; // conditionally add room for negative sign
}
/**
* A base 128 varint encodes 7 bits at a time, this checks how many bytes are needed to represent
* the value.
*
*
See https://developers.google.com/protocol-buffers/docs/encoding#varints
*
*
This logic is the same as {@code com.squareup.wire.ProtoWriter.varint32Size} v2.3.0 which
* benchmarked faster than loop variants of the frequently copy/pasted VarInt.varIntSize
*/
public static int varintSizeInBytes(int value) {
if ((value & (0xffffffff << 7)) == 0) return 1;
if ((value & (0xffffffff << 14)) == 0) return 2;
if ((value & (0xffffffff << 21)) == 0) return 3;
if ((value & (0xffffffff << 28)) == 0) return 4;
return 5;
}
/** Like {@link #varintSizeInBytes(int)}, except for uint64. */
// TODO: benchmark vs https://github.com/protocolbuffers/protobuf/blob/master/java/core/src/main/java/com/google/protobuf/CodedOutputStream.java#L770
// Since trace IDs are random, I guess they cover the entire spectrum of varint sizes and probably would especially benefit from this.
public static int varintSizeInBytes(long v) {
if ((v & (0xffffffffffffffffL << 7)) == 0) return 1;
if ((v & (0xffffffffffffffffL << 14)) == 0) return 2;
if ((v & (0xffffffffffffffffL << 21)) == 0) return 3;
if ((v & (0xffffffffffffffffL << 28)) == 0) return 4;
if ((v & (0xffffffffffffffffL << 35)) == 0) return 5;
if ((v & (0xffffffffffffffffL << 42)) == 0) return 6;
if ((v & (0xffffffffffffffffL << 49)) == 0) return 7;
if ((v & (0xffffffffffffffffL << 56)) == 0) return 8;
if ((v & (0xffffffffffffffffL << 63)) == 0) return 9;
return 10;
}
}