We're running pandas version 0.16.1

Is there any chance we could get a version call added to the API that would return the versions of Python and Pandas? (It would be much speedier than asking here.)

So, about API functions that could return versions? I don't suppose they've been added in Q2?

This is easy to do with most Python libraries, for example:

import pandas as pd  
print pd.__version__  

I just ran this and it reports the version previously mentioned (0.16.1).

See: http://stackoverflow.com/questions/20612645/how-to-find-the-installed-pandas-version

Perfect! Exactly what I was looking for.

Do you know if it's possible to get the Python version in a similar way?

In research I tried

import IPython  
IPython.__version__

But we cannot import IPython

"I will use zipline to develop my code!", said B.
"That way I can use a real debugger and editor!", said B.
"I will be able to make more sophisticated algorithms!", said B.
And so B developed code with lots of math, using zipline, which by default imported the latest pandas 18.1.
Thousands of lines of code later, with a not-so-small amount of porting, he ran his code on Quantopian, and the Quantopian backtester said: "AttributeError: 'DataFrame' object has no attribute 'ewm'," "pd.__version__ 0.16.1."
B was very sad.

    # This makes a series of covariance matrices (one for each overlapping sequence of length min_periods)  
    # within the data we passed in (r_) that is about 2*min_periods in length - note that the number of cov  
    # matrices passed back is same as length of r_, but the dates whose index in the sequence is < min_periods  
    # are filled with NAs.  
    cmats = r_.ewm(halflife=_WINDOW * _EWM_HL, min_periods=_WINDOW).cov()  
    ncmat = cmats.shape[0]  
    # NOTE: upper limit in this loop is -1 rather than zero: so every value is calculated and recorded  
    #              EXCEPT for today, the first day of trading, which will be done separately after return...  
    for n in range(-_LTAVG_N, -1):  
        dt_idx = ncmat + n

        cmat = cmats[dt_idx, :, :].as_matrix()  

Chris:

The Python version is 2.7. The normal way to see the version is to do sys.version or something similar, but imports are restricted, so I couldn't find a way to do it.

Actually that's a whole other can o worms as 0.9 which I've been developing on (yes, I know 1.0 just released) requires 3.4. lots of other gotchas. Here's a fix:

import pandas as pd

# import-ANT thing to figure out here:  
pandas_version_num = list(map(int, pd.__version__.split('.')))  
# in case I run this in the far future:  
_PANDAS_MODERN_ = (pandas_version_num[0] > 0) or (pandas_version_num[1] >= 18)


###################################################################################################################  
# COMPATIBILITY!!!!  
# new un-deprecated syntax doesn't work in quantopian, try to future-proof this:  
def cov_ewm(df_, halflife, min_periods, nona=True):  
    #  NOTE: i discovered that the deprecated syntax actually works in 0.18.x but is going away...  
    if _PANDAS_MODERN_:  
        if nona:  
            covs = df_.ewm(halflife=halflife, min_periods=min_periods).cov().dropna()  
        else:  
            covs = df_.ewm(halflife=halflife, min_periods=min_periods).cov()  
    else:  
        if nona:  
            covs = pd.ewmcov(df_, halflife=halflife, min_periods=min_periods).dropna()  
        else:  
            covs = pd.ewmcov(df_, halflife=halflife, min_periods=min_periods)  
    return covs


###################################################################################################################