Most of circular buffer assumes to only read/write ONE object each time, the only link I found to operates on binary data in form of (const char *bytes, size_t byte_count) is http://www.asawicki.info/news_1468_circular_buffer_of_raw_binary_data_in_c.html, which I feel is not incorrect and a little long. What is the right implementation for that?
I created one myself. but it is still long. Can anybody share a more elegant version? or can you point out is there any thing I can improve in my code to make it short?
class Pipe{
Pipe(size_t capacity): _capacity(capacity){ init(); }
~Pipe(){delete [] _buf; }
size_t read(char* data, size_t bytes);
size_t write(const char* data, size_t bytes);
private:
//only _capacity-1 is used, one is to identify full or empty.
void init(){_buf = new char[_capacity];
_wptr = 0; _rptr = 0; _used_size = 0;
}
char* _buf;
size_t _capacity, _wptr, _rptr, _used_size;
bool isFull(){return (_wptr + 1 ) % (_capacity) == _rptr;}
bool isEmpty(){return _wptr == _rptr;}
};
size_t Pipe::read(char* data, size_t bytes){
if (isEmpty() || bytes == 0) return 0;
size_t bytes_read1 = 0, bytes_read2 = 0;
if (_rptr>=_wptr+1) { //two piece can be read
bytes_read1 = min(bytes, _capacity - _rptr);
memcpy(data, _buf + _rptr, bytes_read1);
_rptr += bytes_read1;
bytes -= bytes_read1;
if (_rptr == _capacity) _rptr = 0;
if (bytes > 0){
bytes_read2 = min(bytes, _wptr);
memcpy(_buf + _rptr, data, bytes_read2);
_rptr += bytes_read2;
bytes -= bytes_read2;
}
}
else{//one piece can be read
bytes_read1 = min(bytes, _wptr - _rptr);
memcpy(_buf + _wptr, data, bytes_read1);
_rptr += bytes_read1;
bytes -= bytes_read1;
}
return bytes_read1 + bytes_read2;
}
size_t Pipe::write(const char* data, size_t bytes){
if (isFull() || bytes == 0) return 0;
size_t bytes_write1 = 0, bytes_write2 = 0;
if (_wptr>=_rptr) { //two piece can be written
bytes_write1 = min(bytes, _capacity - _wptr);
memcpy(_buf + _wptr, data, bytes_write1);
_wptr += bytes_write1;
bytes -= bytes_write1;
if (_wptr == _capacity) _wptr = 0;
if (bytes > 0){ //_wptr must be 0 here.
bytes_write2 = min(bytes, _rptr-1);//-1 bcz there is one
slot to check empty/full
memcpy(_buf + _wptr, data+ bytes_write1, bytes_write2);
_wptr += bytes_write2;
bytes -= bytes_write2;
}
}
else{ //one piece can be written
bytes_write1 = min(bytes, _rptr - _wptr -1);
memcpy(_buf + _wptr, data, bytes_write1);
_wptr += bytes_write1;
bytes -= bytes_write1;
}
return bytes_write1 + bytes_write2;
}
Circular Buffers can be implemented in two ways, using an array or a linked list. An empty object array along with its capacity is initialized inside the constructor as the type of elements added is unknown.
A circular buffer is a utility used to transfer successive data values from a producer thread to a consumer thread, who retrieves the data in FIFO (first in first out) order. This kind of data structure will be used when pipelining threads, a subject discussed in detail in Chapter 15.
A ring buffer is an efficient FIFO buffer. It uses a fixed-size array that can be pre-allocated upfront and allows an efficient memory access pattern. All the buffer operations are constant time O(1), including consuming an element, as it doesn't require a shifting of elements.
Determining if a Buffer is Full Both the “full” and “empty” cases of the circular buffer look the same: head and tail pointer are equal. There are two approaches to differentiating between full and empty: “Waste” a slot in the buffer: Full state is head + 1 == tail.
The code in OP may be simplified by excluding all conditionals. Original interface and memcpy
for implementation is retained (only constructor/destructor/read/write made public and unused _used_size
may be dropped).
size_t Pipe::read(char* data, size_t bytes)
{
bytes = min(bytes, getUsed());
const size_t bytes_read1 = min(bytes, _capacity - _rptr);
memcpy(data, _buf + _rptr, bytes_read1);
memcpy(data + bytes_read1, _buf, bytes - bytes_read1);
updateIndex(_rptr, bytes);
return bytes;
}
size_t Pipe::write(const char* data, size_t bytes)
{
bytes = min(bytes, getFree());
const size_t bytes_write1 = min(bytes, _capacity - _wptr);
memcpy(_buf + _wptr, data, bytes_write1);
memcpy(_buf, data + bytes_write1, bytes - bytes_write1);
updateIndex(_wptr, bytes);
return bytes;
}
Several private methods used here may have this simple implementation:
size_t Pipe::getUsed()
{ return (_capacity - _rptr + _wptr) % _capacity; }
size_t Pipe::getFree()
{ return (_capacity - 1 - _wptr + _rptr) % _capacity; }
void Pipe::updateIndex(size_t& index, size_t bytes)
{ index = (index + bytes) % _capacity; }
This implementation has one disadvantage: it is broken when _capacity
is close to maximum size_t
value (because of overflows). This may be fixed by substituting modulo with conditionals in free/used calculations and index updates. Here are modifications for methods used in read
:
size_t Pipe::getUsed()
{
if (_wptr >= _rptr)
return _wptr - _rptr;
else
return _capacity - _rptr + _wptr;
}
void Pipe::updateIndex(size_t& index, size_t bytes)
{
if (bytes >= _capacity - index)
index = index + bytes - _capacity;
else
index = index + bytes;
}
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With