There is more than one way to do this in music21, so this is just one simple way. Note that the durational value is expressed as a float, such that a quarter note equals 1.0, a half note equals 2.0, etc.:
import music21
from music21 import *
piece = converter.parse("full_path_to_piece.midi")
all_parts = []
for part in piece.parts:
part_tuples = []
for event in part:
for y, in event.contextSites():
if y[0] is part:
offset = y[1]
if getattr(event, 'isNote', None) and event.isNote:
part_tuples.append((event.nameWithOctave, event.quarterLength, offset))
if getattr(event, 'isRest', None) and event.isRest:
part_tuples.append(('Rest', event.quarterLength, offset))
all_parts.append(part_tuples)
An alternative solution would be to use the vis-framework, which accesses music files in symbolic notation via music21 and stores the information in pandas dataframes. You can do this:
pip install vis-framework
Another solution would be to use Humdrum instead of music21.