How about using pandas?

[alexispetit]

Hi there, I've just come across your library and thought you could benefit from using pandas. Here's an example of how to calculate the MDI using the data you provided:

import pandas as pd 

# Load data
path = 'test_data/multiple_markets/'
names = ('date', 'open', 'high', 'low', 'close', 'volume', 'open_interest')
ZC = pd.read_csv(path + 'ZC_REV.csv', header=None, names=names, parse_dates=[0], index_col=0)
KW = pd.read_csv(path + 'KW_REV.csv', header=None, names=names, parse_dates=[0], index_col=0)

def signal_to_noise(df, n):
    signal = df['close'].diff(n).abs()
    noise = df['close'].diff().abs().rolling(n).sum()
    return signal / noise

# Calculate SNR and MDI
snr = pd.concat([signal_to_noise(ZC, n=100).to_frame('ZC'), 
                           signal_to_noise(KW, n=100).to_frame('KW')], axis=1)
mdi = snr.mean(axis=1)
mdi.to_csv('MDI.csv')

[visierl]

That looks promising, I will play with pandas a bit and see where it goes. Thanks for the suggestion and the sample code.