artfwo / pymonome Goto Github PK

Hi,

I try to make the pymonome example working, but each scripts return me the following error :

Traceback (most recent call last):
  File "./faders.py", line 39, in <module>
    host, port = monome.find_any_monome()
  File "~/pymonome/monome.py", line 41, in find_any_monome
    browser = MonomeBrowser()
  File "~/pymonome/monome.py", line 174, in __init__
    self.sdRef = pybonjour.DNSServiceBrowse(regtype=REGTYPE, callBack=self.browse_callback)
  File "~/pymonome/pybonjour.py", line 1409, in DNSServiceBrowse
    None)
  File "~/pymonome/pybonjour.py", line 286, in _errcheck
    raise cls(result)
pybonjour.BonjourError: (-65537, 'unknown')

linux distro : Archlinux
Python : 2.7

Thanks

First of all, thanks for this project! Although the async stuff is new to me, the grid studies were a very nice intro, really liking it so far!

Anyway, I wanted to discuss something. The module as it's importable within Python is called monome, which is the same name as the name of the module for the Python bindings for libmonome, which is kind of unfortunate.
What do you think would be the best fix for this?

Thanks for the awesome asyncio-first library. Works really well.

I was able to get Grid + Arc consistently output at 30fps with ~90fps logic updates without stuttering. To achieve this, I had to switch from using GridBuffer and ArcBuffer which allocate the entire map every frame. Instead, I used a numpy array that I simply zeroed between frames, which is measurably faster.

To make this work, I had to make two changes:

1. Add grid.led_level_map_raw and arc.ring_map_raw methods that don't unpack *data but simply pass data directly to OSC.

2. In aiosc, I added support for numpy arrays in pack_message without importing numpy by doing this:

        elif type(arg).__module__ == 'numpy' and type(arg).__qualname__ == 'ndarray':
            result += arg.tobytes()
            dt = arg.dtype
            if dt == '>i4':
                tt = 'i' * arg.size
            elif dt == '>f4':
                tt = 'f' * arg.size
            else:
                raise NotImplementedError('Unsupported numpy ndarray dtype: ' + dt)
            typetag += tt

With those changes, numpy arrays passed to the new methods allow reusing the same memory buffer and are very efficiently translated to bytes entirely in C land (with arg.tobytes() above).

Would you be interested in pull requests to upstream my changes?