ClickHouse/dbms/src/IO/BufferBase.h
Amos Bird 8851fbcab2 Padding for IO buffers.
Testing data

```
select 'aaaaaaaa','bbbbbbbb','cccccccc','dddddddd','eeeeeeee','ffffffff','gggg','hhh' from numbers(3000000) into outfile '/tmp/test.tsv'
```

Testing command
```
echo "select count() from file('/tmp/test.tsv', CSV, 'a String, b String, c String, d String, e String, f String, g String, h String') where not ignore(e)" | clickhouse-benchmark
```

TSV parser has less overhead than CSV, using it would better unveil the benefits of memcpySmall.

Before
```
QPS: 1.662, RPS: 4985463.906, MiB/s: 603.823, result RPS: 1.662, result MiB/s: 0.000.
0.000%  0.559 sec.
10.000% 0.564 sec.
20.000% 0.568 sec.
30.000% 0.572 sec.
40.000% 0.575 sec.
50.000% 0.581 sec.
60.000% 0.592 sec.
70.000% 0.624 sec.
80.000% 0.639 sec.
90.000% 0.664 sec.
95.000% 0.686 sec.
99.000% 0.711 sec.
99.900% 0.715 sec.
99.990% 0.716 sec.
```

After
```
QPS: 1.861, RPS: 5582303.107, MiB/s: 676.110, result RPS: 1.861, result MiB/s: 0.000.
0.000%  0.510 sec.
10.000% 0.514 sec.
20.000% 0.517 sec.
30.000% 0.521 sec.
40.000% 0.523 sec.
50.000% 0.527 sec.
60.000% 0.530 sec.
70.000% 0.539 sec.
80.000% 0.558 sec.
90.000% 0.584 sec.
95.000% 0.589 sec.
99.000% 0.608 sec.
99.900% 0.655 sec.
99.990% 0.663 sec.
```
2018-08-29 09:39:44 +08:00

132 lines
4.5 KiB
C++

#pragma once
#include <Core/Defines.h>
#include <algorithm>
namespace DB
{
/** Base class for ReadBuffer and WriteBuffer.
* Contains common types, variables, and functions.
*
* ReadBuffer and WriteBuffer are similar to istream and ostream, respectively.
* They have to be used, because using iostreams it is impossible to effectively implement some operations.
* For example, using istream, you can not quickly read string values from a tab-separated file,
* so that after reading, the position remains immediately after the read value.
* (The only option is to call the std::istream::get() function on each byte, but this slows down due to several virtual calls.)
*
* Read/WriteBuffers provide direct access to the internal buffer, so the necessary operations are implemented more efficiently.
* Only one virtual function nextImpl() is used, which is rarely called:
* - in the case of ReadBuffer - fill in the buffer with new data from the source;
* - in the case of WriteBuffer - write data from the buffer into the receiver.
*
* Read/WriteBuffer can own or not own an own piece of memory.
* In the second case, you can effectively read from an already existing piece of memory / std::string without copying it.
*/
class BufferBase
{
public:
/** Cursor in the buffer. The position of write or read. */
using Position = char *;
/** A reference to the range of memory. */
struct Buffer
{
Buffer(Position begin_pos_, Position end_pos_) : begin_pos(begin_pos_), end_pos(end_pos_) {}
inline Position begin() const { return begin_pos; }
inline Position end() const { return end_pos; }
inline size_t size() const { return size_t(end_pos - begin_pos); }
inline void resize(size_t size) { end_pos = begin_pos + size; }
inline void swap(Buffer & other)
{
std::swap(begin_pos, other.begin_pos);
std::swap(end_pos, other.end_pos);
}
private:
Position begin_pos;
Position end_pos; /// 1 byte after the end of the buffer
};
/** The constructor takes a range of memory to use for the buffer.
* offset - the starting point of the cursor. ReadBuffer must set it to the end of the range, and WriteBuffer - to the beginning.
*/
BufferBase(Position ptr, size_t size, size_t offset)
: pos(ptr + offset), working_buffer(ptr, ptr + size), internal_buffer(ptr, ptr + size) {}
void set(Position ptr, size_t size, size_t offset)
{
internal_buffer = Buffer(ptr, ptr + size);
working_buffer = Buffer(ptr, ptr + size);
pos = ptr + offset;
}
/// get buffer
inline Buffer & internalBuffer() { return internal_buffer; }
/// get the part of the buffer from which you can read / write data
inline Buffer & buffer() { return working_buffer; }
/// get (for reading and modifying) the position in the buffer
inline Position & position() { return pos; }
/// offset in bytes of the cursor from the beginning of the buffer
inline size_t offset() const { return size_t(pos - working_buffer.begin()); }
/// How many bytes are available for read/write
inline size_t available() const { return size_t(working_buffer.end() - pos); }
inline void swap(BufferBase & other)
{
internal_buffer.swap(other.internal_buffer);
working_buffer.swap(other.working_buffer);
std::swap(pos, other.pos);
}
/** How many bytes have been read/written, counting those that are still in the buffer. */
size_t count() const
{
return bytes + offset();
}
/** Check that there is more bytes in buffer after cursor. */
bool ALWAYS_INLINE hasPendingData() const
{
return pos != working_buffer.end();
}
bool isPadded() const
{
return padded;
}
protected:
/// Read/write position.
Position pos;
/** How many bytes have been read/written, not counting those that are now in the buffer.
* (counting those that were already used and "removed" from the buffer)
*/
size_t bytes = 0;
/** A piece of memory that you can use.
* For example, if internal_buffer is 1MB, and from a file for reading it was loaded into the buffer
* only 10 bytes, then working_buffer will be 10 bytes in size
* (working_buffer.end() will point to the position immediately after the 10 bytes that can be read).
*/
Buffer working_buffer;
/// A reference to a piece of memory for the buffer.
Buffer internal_buffer;
/// Indicator of 15 bytes pad_right
bool padded{false};
};
}