MIDI: reading performance can be improved

[kwikwag]

Motivation
The code that reads in MIDIs uses a lot of inefficient code, causing in rather slow reads.

Feature summary

I tried some straightforward approaches to make it more efficient. See https://github.com/kwikwag/music21/tree/feature/midi-read-performance

Proposed implementation

• Replace bytes with memoryview for the reading process.
• Use a pre-calculated frozenset() for the _ContainsEnum base class

Intent

[x] I plan on implementing this myself.
[ ] I am willing to pay to have this feature added.
[x] I am starting a discussion with the hope that community members will volunteer their time to create this. I understand that individuals work on features of interest to them and that this feature may never be implemented.

[kwikwag]

Note: with this change, event.data will return a memoryview where it used to have bytes. I can change that, but not sure what would be best.

[jacobtylerwalls]

Hi, thanks for using the package and thinking of giving back.

Regarding the two proposed changes, do you have a benchmark? You mention the frozenset change saves 5%. What about the other? (The other diff is much bigger, so it would help to evaluate processing time vs. patch review time / documentation time.)

Note: with this change, event.data will return a memoryview where it used to have bytes.

This sounds like something that might need to wait for v. 8. The frozenset change could go in now.

[kwikwag]

They both are at aboud 5% is my measurements. I will try to come up with a benchmark set. I'll fix the pylint problems first.

And thank you for the package!