2011-08-09 15:57:33 +00:00
|
|
|
#pragma once
|
2010-06-04 18:25:25 +00:00
|
|
|
|
|
|
|
#include <cstring>
|
2011-08-09 15:57:33 +00:00
|
|
|
#include <algorithm>
|
2017-02-09 10:10:13 +00:00
|
|
|
#include <memory>
|
2010-06-04 18:25:25 +00:00
|
|
|
|
2017-04-01 09:19:00 +00:00
|
|
|
#include <Common/Exception.h>
|
|
|
|
#include <IO/BufferBase.h>
|
2010-06-04 18:25:25 +00:00
|
|
|
|
|
|
|
|
|
|
|
namespace DB
|
|
|
|
{
|
|
|
|
|
2016-01-11 21:46:36 +00:00
|
|
|
namespace ErrorCodes
|
|
|
|
{
|
2017-04-01 07:20:54 +00:00
|
|
|
extern const int ATTEMPT_TO_READ_AFTER_EOF;
|
|
|
|
extern const int CANNOT_READ_ALL_DATA;
|
2016-01-11 21:46:36 +00:00
|
|
|
}
|
|
|
|
|
2017-05-28 14:29:40 +00:00
|
|
|
/** A simple abstract class for buffered data reading (char sequences) from somewhere.
|
|
|
|
* Unlike std::istream, it provides access to the internal buffer,
|
|
|
|
* and also allows you to manually manage the position inside the buffer.
|
2010-06-04 18:25:25 +00:00
|
|
|
*
|
2017-05-28 14:29:40 +00:00
|
|
|
* Note! `char *`, not `const char *` is used
|
|
|
|
* (so that you can take out the common code into BufferBase, and also so that you can fill the buffer in with new data).
|
|
|
|
* This causes inconveniences - for example, when using ReadBuffer to read from a chunk of memory const char *,
|
|
|
|
* you have to use const_cast.
|
2011-10-16 03:05:15 +00:00
|
|
|
*
|
2017-05-28 14:29:40 +00:00
|
|
|
* successors must implement the nextImpl() method.
|
2010-06-04 18:25:25 +00:00
|
|
|
*/
|
2011-06-27 18:22:14 +00:00
|
|
|
class ReadBuffer : public BufferBase
|
2010-06-04 18:25:25 +00:00
|
|
|
{
|
|
|
|
public:
|
2017-05-28 14:29:40 +00:00
|
|
|
/** Creates a buffer and sets a piece of available data to read to zero size,
|
|
|
|
* so that the next() function is called to load the new data portion into the buffer at the first try.
|
2017-04-01 07:20:54 +00:00
|
|
|
*/
|
|
|
|
ReadBuffer(Position ptr, size_t size) : BufferBase(ptr, size, 0) { working_buffer.resize(0); }
|
|
|
|
|
2017-05-28 14:29:40 +00:00
|
|
|
/** Used when the buffer is already full of data that can be read.
|
|
|
|
* (in this case, pass 0 as an offset)
|
2017-04-01 07:20:54 +00:00
|
|
|
*/
|
|
|
|
ReadBuffer(Position ptr, size_t size, size_t offset) : BufferBase(ptr, size, offset) {}
|
|
|
|
|
|
|
|
void set(Position ptr, size_t size) { BufferBase::set(ptr, size, 0); working_buffer.resize(0); }
|
|
|
|
|
2017-05-28 14:29:40 +00:00
|
|
|
/** read next data and fill a buffer with it; set position to the beginning;
|
|
|
|
* return `false` in case of end, `true` otherwise; throw an exception, if something is wrong
|
2017-04-01 07:20:54 +00:00
|
|
|
*/
|
|
|
|
bool next()
|
|
|
|
{
|
|
|
|
bytes += offset();
|
|
|
|
bool res = nextImpl();
|
|
|
|
if (!res)
|
|
|
|
working_buffer.resize(0);
|
|
|
|
|
|
|
|
pos = working_buffer.begin() + working_buffer_offset;
|
|
|
|
working_buffer_offset = 0;
|
|
|
|
return res;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
inline void nextIfAtEnd()
|
|
|
|
{
|
|
|
|
if (!hasPendingData())
|
|
|
|
next();
|
|
|
|
}
|
|
|
|
|
|
|
|
virtual ~ReadBuffer() {}
|
|
|
|
|
|
|
|
|
2017-05-28 14:29:40 +00:00
|
|
|
/** Unlike std::istream, it returns true if all data was read
|
|
|
|
* (and not in case there was an attempt to read after the end).
|
|
|
|
* If at the moment the position is at the end of the buffer, it calls the next() method.
|
|
|
|
* That is, it has a side effect - if the buffer is over, then it updates it and set the position to the beginning.
|
2017-04-01 07:20:54 +00:00
|
|
|
*
|
2017-05-28 14:29:40 +00:00
|
|
|
* Try to read after the end should throw an exception.
|
2017-04-01 07:20:54 +00:00
|
|
|
*/
|
|
|
|
bool ALWAYS_INLINE eof()
|
|
|
|
{
|
|
|
|
return !hasPendingData() && !next();
|
|
|
|
}
|
|
|
|
|
|
|
|
void ignore()
|
|
|
|
{
|
|
|
|
if (!eof())
|
|
|
|
++pos;
|
|
|
|
else
|
|
|
|
throw Exception("Attempt to read after eof", ErrorCodes::ATTEMPT_TO_READ_AFTER_EOF);
|
|
|
|
}
|
|
|
|
|
|
|
|
void ignore(size_t n)
|
|
|
|
{
|
|
|
|
while (n != 0 && !eof())
|
|
|
|
{
|
|
|
|
size_t bytes_to_ignore = std::min(static_cast<size_t>(working_buffer.end() - pos), n);
|
|
|
|
pos += bytes_to_ignore;
|
|
|
|
n -= bytes_to_ignore;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (n)
|
|
|
|
throw Exception("Attempt to read after eof", ErrorCodes::ATTEMPT_TO_READ_AFTER_EOF);
|
|
|
|
}
|
|
|
|
|
2017-05-28 14:29:40 +00:00
|
|
|
/// You could call this method `ignore`, and `ignore` call `ignoreStrict`.
|
2017-04-01 07:20:54 +00:00
|
|
|
size_t tryIgnore(size_t n)
|
|
|
|
{
|
|
|
|
size_t bytes_ignored = 0;
|
|
|
|
|
|
|
|
while (bytes_ignored < n && !eof())
|
|
|
|
{
|
|
|
|
size_t bytes_to_ignore = std::min(static_cast<size_t>(working_buffer.end() - pos), n - bytes_ignored);
|
|
|
|
pos += bytes_to_ignore;
|
|
|
|
bytes_ignored += bytes_to_ignore;
|
|
|
|
}
|
|
|
|
|
|
|
|
return bytes_ignored;
|
|
|
|
}
|
|
|
|
|
2017-05-28 14:29:40 +00:00
|
|
|
/** Reads as many as there are, no more than n bytes. */
|
2017-04-01 07:20:54 +00:00
|
|
|
size_t read(char * to, size_t n)
|
|
|
|
{
|
|
|
|
size_t bytes_copied = 0;
|
|
|
|
|
|
|
|
while (bytes_copied < n && !eof())
|
|
|
|
{
|
|
|
|
size_t bytes_to_copy = std::min(static_cast<size_t>(working_buffer.end() - pos), n - bytes_copied);
|
|
|
|
::memcpy(to + bytes_copied, pos, bytes_to_copy);
|
|
|
|
pos += bytes_to_copy;
|
|
|
|
bytes_copied += bytes_to_copy;
|
|
|
|
}
|
|
|
|
|
|
|
|
return bytes_copied;
|
|
|
|
}
|
|
|
|
|
2017-05-28 14:29:40 +00:00
|
|
|
/** Reads n bytes, if there are less - throws an exception. */
|
2017-04-01 07:20:54 +00:00
|
|
|
void readStrict(char * to, size_t n)
|
|
|
|
{
|
|
|
|
if (n != read(to, n))
|
|
|
|
throw Exception("Cannot read all data", ErrorCodes::CANNOT_READ_ALL_DATA);
|
|
|
|
}
|
|
|
|
|
2017-05-28 14:29:40 +00:00
|
|
|
/** A method that can be more efficiently implemented in successors, in the case of reading large enough blocks.
|
|
|
|
* The implementation can read data directly into `to`, without superfluous copying, if in `to` there is enough space for work.
|
|
|
|
* For example, a CompressedReadBuffer can decompress the data directly into `to`, if the entire decompressed block fits there.
|
|
|
|
* By default - the same as read.
|
|
|
|
* Don't use for small reads.
|
2017-04-01 07:20:54 +00:00
|
|
|
*/
|
|
|
|
virtual size_t readBig(char * to, size_t n)
|
|
|
|
{
|
|
|
|
return read(to, n);
|
|
|
|
}
|
2013-09-08 00:00:25 +00:00
|
|
|
|
2015-04-06 15:10:09 +00:00
|
|
|
protected:
|
2017-05-28 14:29:40 +00:00
|
|
|
/// The number of bytes to ignore from the initial position of `working_buffer` buffer.
|
2017-04-01 07:20:54 +00:00
|
|
|
size_t working_buffer_offset = 0;
|
2015-04-06 15:10:09 +00:00
|
|
|
|
2011-05-13 19:40:56 +00:00
|
|
|
private:
|
2017-05-28 14:29:40 +00:00
|
|
|
/** Read the next data and fill a buffer with it.
|
|
|
|
* Return `false` in case of the end, `true` otherwise.
|
|
|
|
* Throw an exception if something is wrong.
|
2017-04-01 07:20:54 +00:00
|
|
|
*/
|
|
|
|
virtual bool nextImpl() { return false; };
|
2010-06-04 18:25:25 +00:00
|
|
|
};
|
|
|
|
|
|
|
|
|
2017-02-09 10:10:13 +00:00
|
|
|
using ReadBufferPtr = std::shared_ptr<ReadBuffer>;
|
|
|
|
|
|
|
|
|
2010-06-04 18:25:25 +00:00
|
|
|
}
|