how to keep the precision while floating a string in python?

Question

The built-in float can't keep the precison(eg.float('123.45678' = 123.45677999999999), I rewrite it, but also meet the same problem. My question is as the title.

a_int = lambda s : sum(((ord(s[j])-48)*(10**(len(s)-j-1)) for j in range(len(s))))

def a_float(s):
    l = s.split(".")
    return a_int(l[0]+l[1]) / 10.0**len(l[1])

>>> s = "11223.512512"
>>> a_float(s)
11223.512511999999

You can't. That's the nature of floating point values. See docs.python.org/2/tutorial/floatingpoint.html — Martijn Pieters
– Martijn Pieters, Commented Sep 17, 2014 at 16:23
I strongly recommend not rewriting float: getting fast and correct decimal-to-binary conversions is hard! — Mark Dickinson
– Mark Dickinson, Commented Sep 17, 2014 at 16:31

Martijn Pieters · Accepted Answer · 2014-09-17 18:10:26Z

3

Floating point values are not precise. They are binary fraction approximations instead; e.g. the computer adds up 1/2 + 1/4 + 1/8, etc. (but picking those fractions based on your actual value), adding those up to approximate the decimal value. A float value can simply not represent the value 11223.512512 precisely. This is the price of hardware-acceleration support; float operations are extremely fast, and usually precise enough.

Don't use float values if precision is important; use the decimal.Decimal() type instead. Or decide on how many digits you want to represent when printing, and format your floating point values with format():

print format(a_float(s), '.6f')

I recommend you read the Floating Point Arithmetic: Issues and Limitations chapter of the Python tutorial for more details.

edited Sep 17, 2014 at 18:10

answered Sep 17, 2014 at 16:26

Martijn Pieters

1.1m326 gold badges4.2k silver badges3.4k bronze badges

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

2 Comments

Patricia Shanahan Over a year ago

It is not so much a matter of precision than of exact representation of decimal fractions. A decimal type has the same problem representing one third.

Martijn Pieters Over a year ago

@PatriciaShanahan: The OP is not talking about fractions here though.

Collectives™ on Stack Overflow

how to keep the precision while floating a string in python?

1 Answer 1

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related