example.py

def get_max_min(prices, smoothing, window_range):
    # Get prices
    smooth_prices = prices['close'].rolling(window=smoothing).mean().dropna()
    # Get max and min for window
    local_max = argrelextrema(smooth_prices.values, np.greater)[0]
    local_min = argrelextrema(smooth_prices.values, np.less)[0]
    print('local_max, local_min', local_max, local_min)
    price_local_max_dt = []
    # iterate thru points
    for i in local_max:
        # only check in areas with complete window coverage
        if (i>window_range) and (i<len(prices)-window_range):
            # append the index of the max for that window (for multiple windows, we'll get the same maxes as they overlap)
            price_local_max_dt.append(prices.iloc[i-window_range:i+window_range]['close'].idxmax())
            print('idx max', prices.iloc[i-window_range:i+window_range]['close'].idxmax())
    price_local_min_dt = []
    for i in local_min:
        if (i>window_range) and (i<len(prices)-window_range):
            price_local_min_dt.append(prices.iloc[i-window_range:i+window_range]['close'].idxmin())  
    maxima = pd.DataFrame(prices.loc[price_local_max_dt])
    minima = pd.DataFrame(prices.loc[price_local_min_dt])
    max_min = pd.concat([maxima, minima]).sort_index()
    max_min.index.name = 'date'
    max_min = max_min.reset_index()
    max_min = max_min[~max_min.date.duplicated()]
    p = prices.reset_index()   
    # max_min['day_num'] = p[p['timestamp'].isin(max_min.date)].index.values
    # max_min = max_min.set_index('day_num')['close']
    return max_min