The _remaining_capacity > _bytes_in_buffer check in bufsocket.py is unnecessary

[laike9m]

This is not an actual issue, but a possible tiny improvement.

TL;DR
I think this check self._remaining_capacity > self._bytes_in_buffer in bufsocket.py#L141 is unnecessary.

Here's why.
From the definitions we know:

  self._remaing_capacity - self._bytes_in_buffer
= (self._buffer_size - self._index) - self._bytes_in_buffer
= self._buffer_size - (self._index + self._bytes_in_buffer)
= self._buffer_size - self._buffer_end    

We also know that each time self._bytes_in_buffer increases, the increment is less than
_buffer_size - _buffer_end. An example:

if (self._remaining_capacity > self._bytes_in_buffer and should_read):
    count = self._sck.recv_into(self._buffer_view[self._buffer_end:])
    if not count and amt > self._bytes_in_buffer:
        raise ConnectionResetError()
    self._bytes_in_buffer += count   # count cannot be greater than _buffer_size - _buffer_end

Now we have two formulas:

_buffer_size - _buffer_end = _remaing_capacity - _bytes_in_buffer    ①
count < _buffer_size - _buffer_end                                   ②   

Thus _bytes_in_buffer < _remaing_capacity is always True.

  _bytes_in_buffer(new) 
= _bytes_in_buffer(old) + count 
< _bytes_in_buffer + _buffer_size - _buffer_end 
= _remaing_capacity

I intended to remove the check and run tests, but unfortunately I was not able to finish them(I have all the test deps installed and run make test, don't know why it takes so long). So I'm not 100% sure it can be removed.

[Lukasa]

I suspect you're right. ;) Anyone who wants to apply this patch is welcome to.

[laike9m]

I wonder how long do all the tests take to run, normally?

[Lukasa]

For running the tests I recommend using tox, but you may have encountered a test hang. This project has CI, so you can simply open a PR if you'd prefer.

[laike9m]

Could you give me an estimate? So that I know if tests indeed hang. pytest-xdist shows I'm using all 4 cores, I don't know whether using tox can speed it up.

[Lukasa]

On my machine under tox it runs in 6 seconds. I think we encountered problems with coverage measurements and pytest-xdist so the tests as run by tox moved away from that style, but I guess we didn't update the Makefile.

[laike9m]

OK I'll try tox to see if everything works.