Can I convert any string to float without losing precision in Python?

Question

I'm parsing some price information from an API, do I need to worry about losing precision if I just do price = float(price_str)?

Or do I have to use decimal.Decimal to make sure the original value is parsed?

How long is the string? float() can only parse up to 17 numbers — Tim
– Tim, Commented Jul 7, 2015 at 9:09
i would have used decimal.Decimal. why do you not want to? — Skaperen
– Skaperen, Commented Jul 7, 2015 at 9:10
@TimCastelijns: I'm not sure what you mean. The input to float can have as many digits as you like, and they're all taken into account when determining which way to round. Compare the results of float('8000000000000000.5') and float('8000000000000000.500000000001'), for example. — Mark Dickinson
– Mark Dickinson, Commented Jul 7, 2015 at 9:40

daveoncode · Accepted Answer · 2015-07-07 10:13:10Z

6

Use Decimal class when working with currency! Why? Let's see this dummy example:

float('0.2') + float('0.1')

Result: 0.30000000000000004

With Decimal instead:

Decimal('0.2') + Decimal('0.1')

Result: Decimal('0.3')

UPDATE:

if you are using third party API with json format, you can use Decimal to automatically parse floating numbers automatically: see my blog post: http://www.daveoncode.com/2014/10/21/python-reading-numbers-from-json-without-loss-of-precision-using-decimal-class-for-data-processing/

answered Jul 7, 2015 at 9:14

daveoncode

19.8k19 gold badges108 silver badges163 bronze badges

Sign up to request clarification or add additional context in comments.

Comments

Hrvoje · Accepted Answer · 2015-07-07 09:12:07Z

3

Prices should be decimal, floats often introduce rounding errors which will give you headaches in the long run. Example:

In [1]: 3*0.1
Out[1]: 0.30000000000000004

answered Jul 7, 2015 at 9:12

Hrvoje

3622 silver badges6 bronze badges