This should be implemented in editors.
It also looks to be nice in exploratory data analysis:
df = pd.read_csv(f)
df.columns = map(convert, df.columns, case.snake)
replies(1):