@ramalho i think the example should be this. currently, a new delegating generator is created for each item in data, which makes grouper's while loop unnecessary and creates the unused averager instance. grouper is also still active when the results are reported, though not sure if this matters as will be garbage collected once main() returns.

def grouper(results):
    while True:
        key = yield
        results[key] = yield from averager()

def main(data):
    results = {}
    group = grouper(results)
    next(group)
    for key, values in data.items():
        group.send(key)
        for value in values:
            group.send(value)
        group.send(None)
    group.close() # just to highlight that grouper is still an active generator

    report(results)