Custom Factor - dynamic window length and timestamp

I'm working with a custom factor that is defined within a function, allowing for dynamism in terms of both the window length and the timestamp. Please see below example.

from quantopian.pipeline.data.builtin import USEquityPricing  
from quantopian.pipeline.factors import CustomFactor  
from quantopian.pipeline import Pipeline  
from quantopian.research import run_pipeline  
import numpy as np  
import pandas as pd

def StdDev(offset=365, nbars=5, s_input=USEquityPricing.close):  
    class StdDevFact(CustomFactor):  
        inputs=[s_input]  
        window_length=nbars + offset  
        def compute(self, today, asset_ids, out, values):  
            # Calculates the column-wise standard deviation, ignoring NaNs  
            out[:] = np.nanstd(values[(-1-offset):-(offset + nbars)], axis=0)  
    return StdDevFact()

def make_pipeline():  
    std_dev = StdDev()

    return Pipeline(  
        columns={  
            'std_dev': std_dev  
        }  
    )

result = run_pipeline(make_pipeline(), start_date = '2015-01-01', end_date = '2016-01-01')  
result.head(10)

The problem with the above code, correct me if I'm wrong, is that the offset parameter makes it computationally inefficient by drastically increasing window length. I'm hoping someone may have a better suggestion such that the input to the compute function only takes in M=nbars instead of M=nbars+offset and then the offset parameter directly adjusts the timestamp. I've attempted the following to no avail:

Attempt 1: yields no error, but the result is unchanged irrespective of the value of offset.

def StdDev(offset=365, nbars=5, s_input=USEquityPricing.close):  
    class StdDevFact(CustomFactor):  
        inputs=[s_input]  
        window_length=nbars  
        def compute(self, today, asset_ids, out, values):  
            # Calculates the column-wise standard deviation, ignoring NaNs  
            today =  today - pd.DateOffset(days=offset)  
            out[:] = np.nanstd(values, axis=0)  
    return StdDevFact()

Attempt 2: yields an error, "today is undefined"

def StdDev(offset=365, nbars=5, s_input=USEquityPricing.close):  
    class StdDevFact(CustomFactor):  
        inputs=[s_input]  
        window_length=nbars  
        today =  today - pd.DateOffset(days=offset)  
        def compute(self, today, asset_ids, out, values):  
            # Calculates the column-wise standard deviation, ignoring NaNs  
            out[:] = np.nanstd(values, axis=0)  
    return StdDevFact()