Pandas error: [<class 'decimal.DivisionUndefined'>]

Question

My dataframe looks like this:

timestamp   topAsk  topBid  CPA midprice    CPB spread  s

    0   2019-03-14 00:00:00 0.00005000  0.00004957  0.00004979  0.00004979  0.00004979  4.3E-7  0.008636272343844145410725045190
    1   2019-03-14 00:01:00 0.00005000  0.00004957  0.00004979  0.00004979  0.00004979  4.3E-7  0.008636272343844145410725045190
    2   2019-03-14 00:02:00 0.00005000  0.00004957  0.00004979  0.00004979  0.00004979  4.3E-7  0.008636272343844145410725045190
    3   2019-03-14 00:03:00 0.00005000  0.00004957  0.00004979  0.00004979  0.00004979  4.3E-7  0.008636272343844145410725045190
    4   2019-03-14 00:04:00 0.00005000  0.00004957  0.00004979  0.00004979  0.00004979  4.3E-7  0.008636272343844145410725045190
    5   2019-03-14 00:05:00 0.00005000  0.00004957  0.00004979  0.00004979  0.00004979  4.3E-7  0.008636272343844145410725045190
    6   2019-03-14 00:06:00 0.00005000  0.00004957  0.00004979  0.00004979  0.00004979  4.3E-7  0.008636272343844145410725045190
    7   2019-03-14 00:07:00 0.00005000  0.00004957  0.00004979  0.00004979  0.00004979  4.3E-7  0.008636272343844145410725045190
    8   2019-03-14 00:08:00 0.00005000  0.00004957  0.00004979  0.00004979  0.00004979  4.3E-7  0.008636272343844145410725045190
    9   2019-03-14 00:09:00 0.00005000  0.00004957  0.00004979  0.00004979  0.00004979  4.3E-7  0.008636272343844145410725045190

When I try to add a new column using the following line of code: df['gamma'] = ((df['midprice'] - df['CPB']) / df['spread']) I have the following error message= Pandas error: [<class 'decimal.DivisionUndefined'>]

Is it because my column df['spread'] is really small? I am a bit stuck, thanks!

df.info() shows:

Answer 1

decimal.InvalidOperation: [<class 'decimal.DivisionUndefined'>] means that you have somewhere a 0/0 division using Decimal values. It is easy to find a workaround by first testing whether df['spread'] is 0, but you really should try to find why and how a value that is intended to be a divisor can be null. And in that case I would use NaN as result. Code could be:

df['gamma'] = df.apply(lambda x:
     (x['midprice'] - x['CPB']) / x['spread'] if x['spread'] != 0
     else decimal.Decimal('NaN'), axis=1)