Writing a single data fills the buffer

[pkrssy]

v1.3.3
If a single data is written into the newly created buffer, buffer.size() reports one but all the other elements of the buffer is filled with the data.

#include <CircularBuffer.h>
CircularBuffer<long, 6> buffer;
uint16_t data1 = 1111;

void setup() 
{

  Serial.begin(9600);
  Serial.print("Buffer size before writing: ");
  Serial.println(buffer.size());
  buffer.push(data1);
  Serial.print("Buffer size after writing: ");
  Serial.println(buffer.size());
  Serial.print("Elements: ");
  Serial.print(buffer[0]);
  Serial.print(", ");
  Serial.print(buffer[1]);
  Serial.print(", ");
  Serial.print(buffer[2]);
  Serial.print(", ");
  Serial.print(buffer[3]);
  Serial.print(", ");
  Serial.print(buffer[4]);
  Serial.print(", ");
  Serial.println(buffer[5]);
   
}

void loop() {
  
}

Result:

09:28:44.723 -> Buffer size before writing: 0
09:28:44.856 -> Buffer size after writing: 1
09:28:44.889 -> Elements: 1111, 1111, 1111, 1111, 1111, 1111

The behavior can not be experienced in v1.3.0

[rlogiacco]

Any access beyond the current buffer bounds is a violation of the contract and assuming anything about the buffer internal state is a usage error.
If the values returned for the populated buffer positions is correct then the buffer operates as intended.

In other words, accessing positions beyond 0 is an error, you are making assumptions on how the buffer operates on its internals.

It wasn't arising in the previous version because the buffer wasn't protecting its internal state properly: now the buffer does that better and it only allows you to access populated positions. If you have only one element it doesn't matter what's the index you provide, you will always get that single item.
Push in another one and you'll see the two values alternating...

[rlogiacco]

Another way to describe the behaviour, if you want to print the buffer contents you should only print the items between position 0 and position buffer.size() - 1: in your case between 0 and 0 after the first insertion....

See the two for-loops examples at https://github.com/rlogiacco/CircularBuffer#automatic-optimization and the code at

CircularBuffer/examples/Test/Test.ino

Lines 6 to 27 in cf4ded9

	void printBuffer() {
	if (buffer.isEmpty()) {
	Serial.println("empty");
	} else {
	Serial.print("[");
	for (decltype(buffer)::index_t i = 0; i < buffer.size() - 1; i++) {
	Serial.print(buffer[i]);
	Serial.print(",");
	}
	Serial.print(buffer[buffer.size() - 1]);
	Serial.print("] (");
	Serial.print(buffer.size());
	Serial.print("/");
	Serial.print(buffer.size() + buffer.available());
	if (buffer.isFull()) {
	Serial.print(" full");
	}
	Serial.println(")");
	}
	}

for reference.

If you want to inspect the internals for debugging reasons, you can enable debugging and use the two debug functions exposed by the library:

#define CIRCULAR_BUFFER_DEBUG
#include <CircularBuffer.h>
CircularBuffer<long, 6> buffer;

And then you can call buffer.debug(Serial) or buffer.debug(Serial, customFunction) to see the whole buffer contents with the head and tail positions.

[pkrssy]

You're right but for a simple user like me it would be more comfortable to indicate the data is not available (return "nan") without the need to add additional conditions in a for cycle.

[rlogiacco]

I’m not aware of the existence of such type. I would throw an exception, but that’s also unavailable: I’m open to suggestions which can be safely and soundly I mplemented on a microcontroller. Il giorno lun 6 dic 2021 alle 22:42 pkrssy ***@***.***> ha scritto:

…

You're right but for a simple user like me it would be more comfortable to indicate the data is not available (return "nan") without the need to add additional conditions in a for cycle. — You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub <#58 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AABQWMLZ5BA4W54F65IIPD3UPUU5FANCNFSM5JMRDUFA> .