Chris Apple
Chris Apple
I'm personally most familiar with numpydoc, so that'd be my vote!
+1 for changing to MidiMessage in some form (alias or otherwise). I think it'll help readability and match the other conventions around the project
Or `MetaMidiMessage`? Perhaps more readable but still has the "Midi" piece in there. I'd support whichever you'd prefer.
This is part of the [roadmap](http://mido.readthedocs.io/en/latest/roadmap.html?highlight=Multithreading#better-support-for-concurrency-and-multithreading) If you'd like to contribute we would happily take a pull request, I'm sure!
One of the things that may be involved in this is "note on" messages with zero velocity can be counted as "note off" messages. Its a matter of implementation details,...
@olemb any thoughts on if this is a bug or a feature?
Are Type 2 MIDI files not supported? If so maybe a custom error? "InvalidMidiFileType" or something? I find those are usually more specific than any builtin for something like this.
Compiled with no issue, running says ``` Note: SPSC queue not supported on this platform, discount its timings ``` Full output: ``` > ./benchmarks Note: SPSC queue not supported on...
Some output from ubsan and asan: ``` bench.cpp:277:10: runtime error: signed integer overflow: 2147450880 + 65536 cannot be represented in type 'int' SUMMARY: UndefinedBehaviorSanitizer: undefined-behavior bench.cpp:277:10 in bench.cpp:277:10: runtime error:...
The ubsan overflows are pretty easy to fix, just changing: ``` unsigned long long total = 0; ``` on line 270 of the bench file fixed it up (long long...