Question

I have a function that returns log10 values. On converting them to normal numbers, I obtain an overflow error.

OverflowError: (34, ‘Numerical result out of range’)

I checked the log values and this error occurs for say, 508.038057662.

I reckon that while python easily performs 10**509, this error must be due to the decimal points overflowing the register. Therefore I tried using numpy.float64 like such,

result = np.array([ (10**multiplicity(timeseries,om,ph,bins,pos_arr)) for ph in np.linspace(0,twopi,num = bins+1)], dtype = np.float64)

The error is the same. Am I declaring the float64 wrong??

Here multiplicity() is the function that returns log10 values. I require a “list” of values.

Answer 1

I reckon that while python easily performs 10**509, this error must be due to the decimal points overflowing the register.

The problem is not down to “the decimal points overflowing”, but is caused by the data type you are using.

Python can happily calculate 10**509 as a long since these have unlimited precision:

>>> type(10**509)
<type 'long'>

However, this result is too big to store in a float:

>>> float(10**509)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
OverflowError: long int too large to convert to float

We can check the maximum float quite easily:

>>> import sys
>>> sys.float_info.max
1.7976931348623157e+308

Looking at this question is seems that the Numpy float64 has the same range of values as a standard float so using that instead isn’t going to fix your problem.

Instead, you’ll have to use one of the third-party modules that provide arbitrary precision floating points such as mpmath or bigfloat.