Chronicle Bytes contains all the low level memory access wrappers. It is built on Chronicle Core’s direct memory and OS system call access.
Chronicle Bytes has a similar purpose to Java NIO’s ByteBuffer with some extensions.
The API supports.
-
64-bit sizes
-
UTF-8 and ISO-8859-1 encoded strings.
-
thread safe off heap memory operations.
-
deterministic release of resources via reference counting.
-
compressed data types such as stop bit encoding.
-
elastic ByteBuffer wrappers which resize as required.
-
parsing text and writing text directly to off heap bytes.
operation |
Indexed or streaming |
binary |
text |
read/write binary primitives |
both |
float, double, boolean and unsigned/signed byte, short, 24-bit int, int, long, incompleteLong |
double, int, long, char, double with precision |
read/write text |
both |
8-bit/UTF-8 string with length or limit |
8-bit/UTF-8 string |
read/write other |
streaming |
histogram, named enum |
bigdecimal, biginteger, date/time/zone, UUID, hexadecimal |
data driven tests |
from files |
no |
yes |
CAS |
indexed |
int, long |
|
volatile read/write |
indexed |
byte, short, int, long, float, double |
|
peek |
both |
unsigned byte |
|
stop bit compression |
streaming |
int,long, double, float, char |
|
search |
from start |
indexOf string, findByte |
|
addAndGet |
indexed |
float, double, int, long |
|
copy |
from start |
write, copy |
|
hash |
from start |
byteSum, fastHash |
|
bytes marshallable |
streaming |
nested data structures, expected types only. |
operation |
explained |
|
read/write binary primitives |
to read an write primitive data stuctures stored in a binary from |
|
read/write text |
to read an write text data |
|
read/write other |
to read an write other data |
|
data driven tests |
||
CAS |
an atomic instruction used in multithreading to achieve synchronization. It compares the contents of a memory location with a given value and, only if they are the same, modifies the contents of that memory location to a new given value. |
|
volatile read/write |
||
peek |
peek is an operation which returns the value of the bytes without effecting its read position |
|
stop bit compression |
https://github.com/OpenHFT/RFC/tree/master/Stop-Bit-Encoding |
|
search |
is any algorithm which solves the search problem, namely, to retrieve information stored within some data structure |
|
addAndGet |
atomically adds the given value to the current value. |
|
copy |
to transfer data from one structure to another |
|
hash |
||
bytes marshallable |
a serialization funciton |
Bytes<ByteBuffer> bytes = Bytes.elasticHeapByteBuffer(64);
ByteBuffer bb = bytes.underlyingObject();Bytes<ByteBuffer> bytes = Bytes.elasticByteBuffer(64);
ByteBuffer bb = bytes.underlyingObject();Bytes bytes = Bytes.allocateElasticDirect(64);
long address = bytes.address
bytes.release(); // when it can be freed.Bytes bytes = Bytes.allocateElasticDirect();
// use the bytes
bytes.release(); // when it can be freed.ByteBuffer needs to be flipped to switch between reading and writing.
Bytes holds a read position and a write position allowing you to write and immediately read without flipping.
|
Note
|
The writePosition is the readLimit. |
Writing to a hexadecimal dump is useful for documenting the format for messages written. We have used the hexadecimal dump here.
// only used for documentation
HexDumpBytes bytes = new HexDumpBytes();
bytes.comment("true").writeBoolean(true);
bytes.comment("s8").writeByte((byte) 1);
bytes.comment("u8").writeUnsignedByte(2);
bytes.comment("s16").writeShort((short) 3);
bytes.comment("u16").writeUnsignedShort(4);
bytes.comment("char").writeUnsignedShort('5'); // char
bytes.comment("s24").writeInt24(-6_666_666);
bytes.comment("u24").writeUnsignedInt24(16_666_666);
bytes.comment("s32").writeInt(6);
bytes.comment("u32").writeUnsignedShort(7);
bytes.comment("s64").writeLong(8);
bytes.comment("f32").writeFloat(9);
bytes.comment("f64").writeDouble(10);
System.out.println(bytes.toHexString());prints
59 # true 01 # s8 02 # u8 03 00 # s16 04 00 # u16 35 # char 56 46 9a # s24 2a 50 fe # u24 06 00 00 00 # s32 07 00 00 00 # u32 08 00 00 00 00 00 00 00 # s64 00 00 10 41 # f32 00 00 00 00 00 00 24 40 # f64
to read this data you can use
boolean flag = bytes.readBoolean();
byte s8 = bytes.readByte();
int u8 = bytes.readUnsignedByte();
short s16 = bytes.readShort();
int u16 = bytes.readUnsignedShort();
char ch = bytes.readStopBitChar();
int s24 = bytes.readInt24();
long u24 = bytes.readUnsignedInt24();
int s32 = bytes.readInt();
long u32 = bytes.readUnsignedInt();
long s64 = bytes.readLong();
float f32 = bytes.readFloat();
double f64 = bytes.readDouble();Instead of streaming the data, sometimes you need to control the placement of data, possibly at random.
Bytes<ByteBuffer> bytes = Bytes.elasticHeapByteBuffer(64);
bytes.writeBoolean(0, true);
bytes.writeByte(1, (byte) 1);
bytes.writeUnsignedByte(2, 2);
bytes.writeShort(3, (short) 3);
bytes.writeUnsignedShort(5, 4);
bytes.writeInt(7, 6);
bytes.writeUnsignedInt(11, 7);
bytes.writeLong(15, 8);
bytes.writeFloat(23, 9);
bytes.writeDouble(27, 10);
bytes.writePosition(35);
System.out.println(bytes.toHexString());
boolean flag = bytes.readBoolean(0);
byte s8 = bytes.readByte(1);
int u8 = bytes.readUnsignedByte(2);
short s16 = bytes.readShort(3);
int u16 = bytes.readUnsignedShort(5);
int s32 = bytes.readInt(7);
long u32 = bytes.readUnsignedInt(11);
long s64 = bytes.readLong(15);
float f32 = bytes.readFloat(23);
double f64 = bytes.readDouble(27);prints
00000000 59 01 02 03 00 04 00 06 00 00 00 07 00 00 00 08 Y······· ········ 00000010 00 00 00 00 00 00 00 00 00 10 41 00 00 00 00 00 ········ ··A····· 00000020 00 24 40 ·$@
|
Note
|
While HexDumpBytes supports the offset methods, you need to provide the offset in binary and the dump making it more complex to use. |
Chronicle Bytes supports variants of the write primitives which have a store barrier writeOrderedXxxx, and reads with a load barrier readVolatileXxxx
|
Note
|
write ordered doesn’t stall the pipeline to wait for the write to occur, making it possible for a single thread to read an old value after the ordered write. |
You can also write and read text to Bytes for low level, direct to native memory text processing.
Bytes<ByteBuffer> bytes = Bytes.elasticHeapByteBuffer(64);
bytes.append(true).append('\n');
bytes.append(1).append('\n');
bytes.append(2L).append('\n');
bytes.append('3').append('\n');
bytes.append(4.1f).append('\n');
bytes.append(5.2).append('\n');
bytes.append(6.2999999, 3).append('\n');
System.out.println(bytes.toHexString());prints
00000000 54 0a 31 0a 32 0a 33 0a 34 2e 31 0a 35 2e 32 0a T·1·2·3· 4.1·5.2· 00000010 36 2e 33 30 30 0a 6.300·
boolean flag = bytes.parseBoolean();
int s32 = bytes.parseInt();
long s64 = bytes.parseLong();
String ch = bytes.parseUtf8(StopCharTesters.SPACE_STOP);
float f32 = bytes.parseFloat();
double f64 = bytes.parseDouble();
double f64b = bytes.parseDouble();|
Note
|
There are less methods for text as 8, 16 and 24 bit can use methods for int, Unsigned int can use the long method.
|
Chronicle Bytes supports two encodings, ISO-8859-1 and UTF-8. It also supports writing these as binary with a length prefix, and a string which should be terminated. Bytes expects Strings to be read to a buffer for further processing, possibly with a String pool.
HexDumpBytes bytes = new HexDumpBytes();
bytes.comment("write8bit").write8bit("£ 1");
bytes.comment("writeUtf8").writeUtf8("£ 1");
bytes.comment("append8bit").append8bit("£ 1").append('\n');
bytes.comment("appendUtf8").appendUtf8("£ 1").append('\n');
System.out.println(bytes.toHexString());prints
03 a3 20 31 # write8bit 04 c2 a3 20 31 # writeUtf8 a3 20 31 0a # append8bit c2 a3 20 31 0a # appendUtf8
String a = bytes.read8bit();
String b = bytes.readUtf8();
String c = bytes.parse8bit(StopCharTesters.CONTROL_STOP);
String d = bytes.parseUtf8(StopCharTesters.CONTROL_STOP);Binary strings are prefixed with a Stop Bit Encoded length.
HexDumpBytes bytes = new HexDumpBytes();
bytes.comment("write8bit").write8bit((String) null);
bytes.comment("writeUtf8").writeUtf8(null);
System.out.println(bytes.toHexString());
String a = bytes.read8bit();
String b = bytes.readUtf8();
assertEquals(null, a);
assertEquals(null, b);prints
80 00 # write8bit 80 00 # writeUtf8
|
Note
|
80 00 is the stop bit encoding for -1 or ~0
|
In binary, you can atomically replace an int or long on condition that it is an expected value.
int and long areHexDumpBytes bytes = new HexDumpBytes();
bytes.comment("s32").writeUtf8("s32");
long s32 = bytes.writePosition();
bytes.writeInt(0);
bytes.comment("s64").writeUtf8("s64");
long s64 = bytes.writePosition();
bytes.writeLong(0);
System.out.println(bytes.toHexString());prints
03 73 33 32 00 00 00 00 # s32 03 73 36 34 00 00 00 00 00 00 00 00 # s64
assertTrue(bytes.compareAndSwapInt(s32, 0, Integer.MAX_VALUE));
assertTrue(bytes.compareAndSwapLong(s64, 0, Long.MAX_VALUE));
System.out.println(bytes.toHexString());prints
03 73 33 32 ff ff ff 7f # s32 03 73 36 34 ff ff ff ff ff ff ff 7f # s64
INFO: You might wonder, how is the hex dump updated as well as the binary? The readPosition actually holds the write position for both, which is why it has to be computed in this case.
Stop Bit encoding is one form of simple compression. For each 7 bits set, a byte is used with the high bit set when there is another byte to write.
See Stop Bit Encoding RFC for more details
HexDumpBytes bytes = new HexDumpBytes();
for (long i : new long[]{
0, -1,
127, -127,
128, -128,
1 << 14, 1 << 21,
1 << 28, 1L << 35,
1L << 42, 1L << 49,
1L << 56, Long.MAX_VALUE,
Long.MIN_VALUE}) {
bytes.comment(i + "L").writeStopBit(i);
}
for (double d : new double[]{
0.0,
-0.0,
1.0,
1.0625,
-128,
-Double.MIN_NORMAL,
Double.NEGATIVE_INFINITY,
Double.NaN,
Double.POSITIVE_INFINITY}) {
bytes.comment(d + "").writeStopBit(d);
}
System.out.println(bytes.toHexString());prints
00 # 0L 80 00 # -1L 7f # 127L fe 00 # -127L 80 01 # 128L ff 00 # -128L 80 80 01 # 16384L 80 80 80 01 # 2097152L 80 80 80 80 01 # 268435456L 80 80 80 80 80 01 # 34359738368L 80 80 80 80 80 80 01 # 4398046511104L 80 80 80 80 80 80 80 01 # 562949953421312L 80 80 80 80 80 80 80 80 01 # 72057594037927936L ff ff ff ff ff ff ff ff 7f # 9223372036854775807L ff ff ff ff ff ff ff ff ff 00 # -9223372036854775808L 00 # 0.0 40 # -0.0 9f 7c # 1.0 9f fc 20 # 1.0625 e0 18 # -128.0 c0 04 # -2.2250738585072014E-308 ff 7c # -Infinity bf 7e # NaN bf 7c # Infinity
To read these you need either long x = bytes.readStopBit() or double d = bytes.readStopBitDouble()
Chronicle Bytes supports serializing simple objects where the type is not stored. This is similar to`RawWire` in Chronicle Wire.
@NotNull MyByteable mb1 = new MyByteable(false, (byte) 1, (short) 2, '3', 4, 5.5f, 6, 7.7);
@NotNull MyByteable mb2 = new MyByteable(true, (byte) 11, (short) 22, 'T', 44, 5.555f, 66, 77.77);
ZonedDateTime zdt1 = ZonedDateTime.parse("2017-11-06T12:35:56.775Z[Europe/London]");
ZonedDateTime zdt2 = ZonedDateTime.parse("2016-10-05T01:34:56.775Z[Europe/London]");
UUID uuid1 = new UUID(0x123456789L, 0xABCDEF);
UUID uuid2 = new UUID(0x1111111111111111L, 0x2222222222222222L);
@NotNull MyScalars ms1 = new MyScalars("Hello", BigInteger.ONE, BigDecimal.TEN, zdt1.toLocalDate(), zdt1.toLocalTime(), zdt1.toLocalDateTime(), zdt1, uuid1);
@NotNull MyScalars ms2 = new MyScalars("World", BigInteger.ZERO, BigDecimal.ZERO, zdt2.toLocalDate(), zdt2.toLocalTime(), zdt2.toLocalDateTime(), zdt2, uuid2);
@NotNull MyNested mn1 = new MyNested(mb1, ms1);
@NotNull MyNested mn2 = new MyNested(mb2, ms2);
bytes.comment("mn1").writeUnsignedByte(1);
mn1.writeMarshallable(bytes);
bytes.comment("mn2").writeUnsignedByte(2);
mn2.writeMarshallable(bytes);class MyByteable implements BytesMarshallable {
boolean flag;
byte b;
short s;
char c;
int i;
float f;
long l;
double d;
public MyByteable(boolean flag, byte b, short s, char c, int i, float f, long l, double d) {
this.flag = flag;
this.b = b;
this.s = s;
this.c = c;
this.i = i;
this.f = f;
this.l = l;
this.d = d;
}class MyScalars implements BytesMarshallable {
String s;
BigInteger bi;
BigDecimal bd;
LocalDate date;
LocalTime time;
LocalDateTime dateTime;
ZonedDateTime zonedDateTime;
UUID uuid;
public MyScalars(String s, BigInteger bi, BigDecimal bd, LocalDate date, LocalTime time, LocalDateTime dateTime, ZonedDateTime zonedDateTime, UUID uuid) {
this.s = s;
this.bi = bi;
this.bd = bd;
this.date = date;
this.time = time;
this.dateTime = dateTime;
this.zonedDateTime = zonedDateTime;
this.uuid = uuid;
}prints
01 # mn1
# byteable
4e # flag
01 # b
02 00 # s
33 # c
04 00 00 00 # i
00 00 b0 40 # f
06 00 00 00 00 00 00 00 # l
cd cc cc cc cc cc 1e 40 # d
# scalars
05 48 65 6c 6c 6f # s
01 31 # bi
02 31 30 # bd
0a 32 30 31 37 2d 31 31 2d 30 36 # date
0c 31 32 3a 33 35 3a 35 36 2e 37 37 35 # time
17 32 30 31 37 2d 31 31 2d 30 36 54 31 32 3a 33 # dateTime
35 3a 35 36 2e 37 37 35 27 32 30 31 37 2d 31 31 # zonedDateTime
2d 30 36 54 31 32 3a 33 35 3a 35 36 2e 37 37 35
5a 5b 45 75 72 6f 70 65 2f 4c 6f 6e 64 6f 6e 5d # uuid
24 30 30 30 30 30 30 30 31 2d 32 33 34 35 2d 36
37 38 39 2d 30 30 30 30 2d 30 30 30 30 30 30 61
62 63 64 65 66
02 # mn2
# byteable
59 # flag
0b # b
16 00 # s
54 # c
2c 00 00 00 # i
8f c2 b1 40 # f
42 00 00 00 00 00 00 00 # l
e1 7a 14 ae 47 71 53 40 # d
# scalars
05 57 6f 72 6c 64 # s
01 30 # bi
01 30 # bd
0a 32 30 31 36 2d 31 30 2d 30 35 # date
0c 30 31 3a 33 34 3a 35 36 2e 37 37 35 # time
17 32 30 31 36 2d 31 30 2d 30 35 54 30 31 3a 33 # dateTime
34 3a 35 36 2e 37 37 35 2c 32 30 31 36 2d 31 30 # zonedDateTime
2d 30 35 54 30 31 3a 33 34 3a 35 36 2e 37 37 35
2b 30 31 3a 30 30 5b 45 75 72 6f 70 65 2f 4c 6f
6e 64 6f 6e 5d 24 31 31 31 31 31 31 31 31 2d 31 # uuid
31 31 31 2d 31 31 31 31 2d 32 32 32 32 2d 32 32
32 32 32 32 32 32 32 32 32 32
The purpose of a Lambda function is to create a simple, highly reproducible, easily testable component.
Once you have your data dumped as hexadecimal, you can create tests using that data, and make variations of those tests.
In this context a Lambda function is one which is entirely input driven and produces a list of messages (one or more outputs).
The simplest Lambda function is stateless, however this has limited application. They are useful for message translation.
If you need a stateful Lambda function, you can consider the input to the function to be every message it has ever consumed. Obviously this is inefficient, however with appropriate caches in your lamdba function, you can process and produce result incrementally.
We module a Lambda function as having an interface for inputs and another for outputs. These interfaces can be the same.
interface IBytesMethod {
@MethodId(0x81L) // (1)
void myByteable(MyByteable byteable);
@MethodId(0x82L)
void myScalars(MyScalars scalars);
@MethodId(0x83L)
void myNested(MyNested nested);
}-
assign a unique id to each method to simplify decoding/encoding.
Each method needs a DTO to describe the data for that message.
class MyByteable implements BytesMarshallable {
boolean flag;
byte b;
short s;
char c;
int i;
float f;
long l;
double d;
....
class MyScalars implements BytesMarshallable {
String s;
BigInteger bi;
BigDecimal bd;
LocalDate date;
LocalTime time;
LocalDateTime dateTime;
ZonedDateTime zonedDateTime;
UUID uuid;
....
class MyNested implements BytesMarshallable {
MyByteable byteable;
MyScalars scalars;
....The implementation needs to take it’s output interface and implement the input interface
static class IBMImpl implements IBytesMethod {
final IBytesMethod out;
IBMImpl(IBytesMethod out) { this.out = out; }
@Override
public void myByteable(MyByteable byteable) { out.myByteable(byteable); }
@Override
public void myScalars(MyScalars scalars) { out.myScalars(scalars); }
@Override
public void myNested(MyNested nested) { out.myNested(nested); }
}Once we have interfaces, DTOs, and an implementation we can setup a test harness
protected void btmttTest(String input, String output) throws IOException {
BytesTextMethodTester tester = new BytesTextMethodTester<>(
input,
IBMImpl::new,
IBytesMethod.class,
output);
tester.run();
assertEquals(tester.expected(), tester.actual());
}This allows us to give two files, one for expected inputs and one for expected outputs.
@Test
public void run() throws IOException {
btmttTest("btmtt/prim-input.txt", "btmtt/prim-output.txt");
}|
Note
|
In this case the input and outputs are expected to be the same. |
81 01 # myByteable
4e # flag
01 # b
02 00 # s
33 # c
04 00 00 00 # i
00 00 b0 40 # f
06 00 00 00 00 00 00 00 # l
cd cc cc cc cc cc 1e 40 # d
### End Of Block
81 01 # myByteable
59 # flag
0b # b
16 00 # s
54 # c
2c 00 00 00 # i
8f c2 b1 40 # f
42 00 00 00 00 00 00 00 # l
e1 7a 14 ae 47 71 53 40 # d
### End Of Block
82 01 # myScalars
05 48 65 6c 6c 6f # s
01 31 # bi
02 31 30 # bd
0a 32 30 31 37 2d 31 31 2d 30 36 # date
0c 31 32 3a 33 35 3a 35 36 2e 37 37 35 # time
17 32 30 31 37 2d 31 31 2d 30 36 54 31 32 3a 33 # dateTime
35 3a 35 36 2e 37 37 35 27 32 30 31 37 2d 31 31 # zonedDateTime
2d 30 36 54 31 32 3a 33 35 3a 35 36 2e 37 37 35
5a 5b 45 75 72 6f 70 65 2f 4c 6f 6e 64 6f 6e 5d # uuid
24 30 30 30 30 30 30 30 31 2d 32 33 34 35 2d 36
37 38 39 2d 30 30 30 30 2d 30 30 30 30 30 30 61
62 63 64 65 66
### End Of Block
83 01 # myNested
# byteable
59 # flag
0b # b
16 00 # s
54 # c
2c 00 00 00 # i
8f c2 b1 40 # f
42 00 00 00 00 00 00 00 # l
e1 7a 14 ae 47 71 53 40 # d
# scalars
05 57 6f 72 6c 64 # s
01 30 # bi
01 30 # bd
0a 32 30 31 36 2d 31 30 2d 30 35 # date
0c 30 31 3a 33 34 3a 35 36 2e 37 37 35 # time
17 32 30 31 36 2d 31 30 2d 30 35 54 30 31 3a 33 # dateTime
34 3a 35 36 2e 37 37 35 2c 32 30 31 36 2d 31 30 # zonedDateTime
2d 30 35 54 30 31 3a 33 34 3a 35 36 2e 37 37 35
2b 30 31 3a 30 30 5b 45 75 72 6f 70 65 2f 4c 6f
6e 64 6f 6e 5d 24 31 31 31 31 31 31 31 31 2d 31 # uuid
31 31 31 2d 31 31 31 31 2d 32 32 32 32 2d 32 32
32 32 32 32 32 32 32 32 32 32
### End Of Block
### End Of TestIn this case, the test calls the methods with the matching method ids which in turn uses the same ids to encode the output.
|
Note
|
Creating and maintain such tests can be an overhead you don’t need. In this case, you can use Chronicle Wire’s YAML testing format to check functionality. WIre can be used for most of the tests even if you intend to use Bytes for production. |
| Access | ByteBuffer | Netty IOBuffer | Aeron UnsafeBuffer | Chronicle Bytes |
|---|---|---|---|---|
Read/write primitives in native memory |
yes |
yes |
yes |
yes |
Separate Mutable interfaces |
run time check |
run time check |
yes |
yes |
Read/Write UTF8 strings |
no |
no |
String |
any CharSequence + Appendable |
Read/Write ISO-8859-1 strings |
no |
no |
? |
any CharSequence + Appendable |
Support Endianness |
Big and Little |
Big and Little |
Big and Little |
Native only |
Size of buffer |
31-bit |
31-bit |
31-bit |
63-bit |
Elastic ByteBuffers |
no |
yes |
no |
yes |
Disable bounds checks |
no |
no |
set globally |
by buffer |
Wrap an address |
no |
no |
yes |
yes |
Thread safe read/write, CAS and atomic add operations |
no |
no |
int; long |
int; long; float and double |
Streaming access |
yes |
yes |
no |
yes |
Deterministic release of memory |
Internal API |
Internal API |
Caller’s responsibility |
yes |
Separate read and write position |
no |
yes |
na |
yes |
