Here's the code (it is not available under "Source Code" above:

# References:  
# 1. Ming Li; Xin Chen; Xin Li; Bin Ma; Vitanyi, P.M.B.; , "The similarity metric,"  
# Information Theory, IEEE Transactions on , vol.50, no.12, pp. 3250- 3264, Dec. 2004  
# http://homepages.cwi.nl/~paulv/papers/similarity.pdf  
#  
# 2. Lin, Jessica, et al. "A symbolic representation of time series, with implications for  
# streaming algorithms." Proceedings of the 8th ACM SIGMOD workshop on Research issues  
# in data mining and knowledge discovery. ACM, 2003.  
# www.cs.ucr.edu/~stelo/papers/DMKD03.pdf

import numpy as np  
import zlib  
from scipy import stats  
import random  
import itertools  
import datetime

# globals for get_data batch transform decorator  
R_P = 1  # refresh period in days  
W_L = 90  # window length in days

def initialize(context):  
    context.stocks = []  
    # bottom = random.randint(0,98)  
    bottom = 98  
    range = 2  
    log.info("universe is {b} to {t}".format(b=bottom, t=bottom+range))  
    set_universe(universe.DollarVolumeUniverse(bottom, bottom+range))  
    # context.day = 0  
def handle_data(context, data):  
    context.stocks = [sid for sid in data]  
    # get data (select prices, volume, etc. in batch transform get_data)  
    d = get_data(data, context.stocks)  
    if d is None:  
        return  
    # # update plot daily  
    # trade_day = data[context.stocks[0]].datetime.day  
    #  
    # if trade_day == context.day:  
        # return  
    #  
    # context.day = trade_day  
    # code data & convert to strings  
    coded_d = code_data(d)  
    # remove columns with NaNs  
    # http://stackoverflow.com/questions/1642730/how-to-delete-columns-in-numpy-array  
    coded_d=np.ma.compress_cols(np.ma.masked_invalid(coded_d))  
    m =coded_d.shape[1]  
    n = coded_d.shape[0]  
    X = [i for i in range(m)]  
    for j in range(m):  
        X[j] = ''  
        for i in range(n):  
            X[j] = X[j] + str(int(coded_d[i,j]))  
    s = sim_pairs(X)  
    # compute & plot coefficien of variation  
    # http://en.wikipedia.org/wiki/Coefficient_of_variation  
    mean_s = np.mean(s)  
    std_s = np.std(s,ddof=1)  
    CV = std_s/mean_s  
    record(CV = CV)

@batch_transform(refresh_period=R_P, window_length=W_L, clean_nans=False) # set globals R_P & W_L above  
def get_data(datapanel,sids):  
    return datapanel['price'].as_matrix(sids)

def code_data(uncoded_data):

    # code data according to Ref. 2

    coded_d = stats.zscore(uncoded_data, axis=0, ddof=1)  
    coded_d[coded_d >= 0.67] = 4  
    coded_d[(coded_d >= 0.0) & (coded_d < 0.67)] = 3  
    coded_d[(coded_d >= -0.67) & (coded_d < 0.0)] = 2  
    coded_d[coded_d < -0.67] = 1  
    return coded_d

def NCD(X,Y):  
    CX = len(zlib.compress(X,9))  
    CY = len(zlib.compress(Y,9))  
    return (len(zlib.compress(X+Y,9)) - min(CX,CY)) / float(max(CX,CY))  
def sim_pairs(string_data):  
    pairs = list((itertools.combinations(string_data,2)))  
    n = len(pairs)  
    result = np.zeros(n)  
    for j in range(n):  
        result[j] = NCD(pairs[j][0],pairs[j][1])  
    return result  

Hi Grant,

I'm baffled myself. I was able to get a more articulate error from the server logs, but I haven't solved it yet.

 Exception: columns overlap: <class 'pandas.tseries.index.DatetimeIndex'>  

Obviously, we need to give a better error to you. And we need to make it easier to debug!

Dan

Thanks Dan,

No rush...sounds like it might be in the batch transform, since that is the only place that I know of where pandas is being used. I stripped everything down and still get the error:

# globals for get_data batch transform decorator  
R_P = 1  # refresh period in days  
W_L = 90  # window length in days

def initialize(context):  
    context.stocks = []  
    bottom = 98  
    range = 2  
    log.info("universe is {b} to {t}".format(b=bottom, t=bottom+range))  
    set_universe(universe.DollarVolumeUniverse(bottom, bottom+range))  
def handle_data(context, data):  
    context.stocks = [sid for sid in data]  
    # get data (select prices, volume, etc. in batch transform get_data)  
    d = get_data(data, context.stocks)  
    if d is None:  
        return   

@batch_transform(refresh_period=R_P, window_length=W_L, clean_nans=False) # set globals R_P & W_L above  
def get_data(datapanel,sids):  
    return datapanel['price'].as_matrix(sids)  

Grant