mirror of
https://github.com/ClickHouse/ClickHouse.git
synced 2024-11-05 15:21:43 +00:00
90 lines
2.3 KiB
C++
90 lines
2.3 KiB
C++
#pragma once
|
||
|
||
#include <DB/IO/WriteBuffer.h>
|
||
#include <DB/IO/BufferWithOwnMemory.h>
|
||
#include <DB/IO/ReadHelpers.h>
|
||
|
||
#define DBMS_DEFAULT_HASHING_BLOCK_SIZE 2048ULL
|
||
|
||
|
||
namespace DB
|
||
{
|
||
|
||
template <class Buffer>
|
||
class IHashingBuffer : public BufferWithOwnMemory<Buffer>
|
||
{
|
||
public:
|
||
IHashingBuffer<Buffer>(size_t block_size_ = DBMS_DEFAULT_HASHING_BLOCK_SIZE)
|
||
: BufferWithOwnMemory<Buffer>(block_size_), block_pos(0), block_size(block_size_), state(0, 0)
|
||
{
|
||
}
|
||
|
||
uint128 getHash()
|
||
{
|
||
if (block_pos)
|
||
return CityHash128WithSeed(&BufferWithOwnMemory<Buffer>::memory[0], block_pos, state);
|
||
else
|
||
return state;
|
||
}
|
||
|
||
void append(DB::BufferBase::Position data)
|
||
{
|
||
state = CityHash128WithSeed(data, block_size, state);
|
||
}
|
||
|
||
/// вычисление хэша зависит от разбиения по блокам
|
||
/// поэтому нужно вычислить хэш от n полных кусочков и одного неполного
|
||
void calculateHash(DB::BufferBase::Position data, size_t len);
|
||
|
||
protected:
|
||
size_t block_pos;
|
||
size_t block_size;
|
||
uint128 state;
|
||
};
|
||
|
||
/** Вычисляет хеш от записываемых данных и передает их в указанный WriteBuffer.
|
||
* В качестве основного буфера используется буфер вложенного WriteBuffer.
|
||
*/
|
||
class HashingWriteBuffer : public IHashingBuffer<WriteBuffer>
|
||
{
|
||
private:
|
||
WriteBuffer & out;
|
||
|
||
void nextImpl() override
|
||
{
|
||
size_t len = offset();
|
||
|
||
Position data = working_buffer.begin();
|
||
calculateHash(data, len);
|
||
|
||
out.position() = pos;
|
||
out.next();
|
||
working_buffer = out.buffer();
|
||
}
|
||
|
||
public:
|
||
HashingWriteBuffer(
|
||
WriteBuffer & out_,
|
||
size_t block_size_ = DBMS_DEFAULT_HASHING_BLOCK_SIZE)
|
||
: IHashingBuffer<DB::WriteBuffer>(block_size_), out(out_)
|
||
{
|
||
out.next(); /// Если до нас в out что-то уже писали, не дадим остаткам этих данных повлиять на хеш.
|
||
working_buffer = out.buffer();
|
||
pos = working_buffer.begin();
|
||
state = uint128(0, 0);
|
||
}
|
||
|
||
uint128 getHash()
|
||
{
|
||
next();
|
||
return IHashingBuffer<WriteBuffer>::getHash();
|
||
}
|
||
};
|
||
}
|
||
|
||
|
||
std::string uint128ToString(uint128 data);
|
||
|
||
std::ostream & operator<<(std::ostream & os, const uint128 & data);
|
||
std::istream & operator>>(std::istream & is, uint128 & data);
|