reading floating-point numbers with 1.#QNAN values in python

Question

Does anyone know of a python string-to-float parser that can cope with MSVC nan numbers (1.#QNAN)? Currently I'm just using float(str) which at least copes with "nan".

I'm using a python script to read the output of a C++ program (runs under linux/mac/win platforms) and the script barfs up when reading these values. (I did already find a C++ library to output the values consistently across platforms, but sometimes have to compare past results, so this still occaisionally pops up.)

Answer 1

Since you have to deal with legacy output files, I see no other possibility but writing a robust_float function:

def robust_float(s):
    try:
        return float(s)
    except ValueError:
        if 'nan' in s.lower():
            return float('nan')
        else:
            raise