Floats print differently when shown in dictionary - python

Question

I'm creating a dictionary with this simple code:

pixel_histogram = {}
min_value = -0.2
max_value = 0.2
interval_size = (math.fabs(min_value) + math.fabs(max_value))/bins

for i in range(bins):
    key = min_value+(i*interval_size)
    print key
    pixel_histogram[key] = 0
    print pixel_histogram

But I'm a little surprised 'cause I got these values with my prints:

#Printing keys
-0.2
-0.16
-0.12
-0.08
-0.04
0.0
0.04
0.08
0.12
0.16

#Printing the dictionary
{0.0: 0, 
-0.08000000000000002: 0, 
0.15999999999999998: 0, 
-0.16: 0, 
0.12: 0, 
-0.12000000000000001: 0, 
0.08000000000000002: 0, 
-0.04000000000000001: 0, 
-0.2: 0, 
0.03999999999999998: 0}

I didn't figure out why the values are different and how could I solve this. Any help would be appreciated. Thanks.

[What every computer scientist should know about floating-point arithmetic](http://docs.oracle.com/cd/E19957-01/806-3568/ncg_goldberg.html) — Ashwini Chaudhary, Sep 23 '13 at 19:57
If you need to use exact floating-point numbers, try using the Decimal class: http://docs.python.org/2/library/decimal.html. — Waleed Khan, Sep 23 '13 at 19:58
While that explains the output of the second case, it doesn't explain why the first output differs from the second. — , Sep 23 '13 at 19:59
Related: http://stackoverflow.com/questions/13345334/strange-behaviour-with-floats-and-string-conversion — Ashwini Chaudhary, Sep 23 '13 at 20:00
Sorry for the noob question. Didn't know about the difference between `str` and `repr`(didn't even know this). Trying to learn more about Python. Thank you guys. — pceccon, Sep 23 '13 at 20:15

score 8 · Accepted Answer · answered Sep 23 '13 at 20:04

8

Python's print statement uses str() on the item being printed. For floating-point values, str() will print up to certain number of decimal values.

When printing a dictionary, the print statement is calling str() on the dictionary object. The dictionary, in its __str__() method definition, uses repr() on the keys. The repr() function for floating-point values prints to more decimal places than the str() function does.

The reason dictionaries use repr() and not str() for keys is that you almost definitely want to see print {'1': 1} print differently than print {1: 1}.

answered Sep 23 '13 at 20:04

Alok--

724
3
10

Thank you @Alok--. This is just what I would like to know. Sorry for the noob question, I didn't know this about the `print` method in Python. – pceccon Sep 23 '13 at 20:13
1

@pceccon No problem. As an aside, it's not really a good idea to store floating-point values as keys in dictionaries, since that implies that you want to be able to do exact comparison on the keys. (If you know that the values you're going to use as keys in that dictionary can be represented exactly with floating-point values, then it should be OK.) – Alok-- Sep 23 '13 at 20:16

score -3 · Answer 2 · answered Sep 23 '13 at 19:58

-3

The values are the same. In the first case, they have been truncated or rounded to decimal places. Surely you are familiar with the rounding of numbers?

answered Sep 23 '13 at 19:58

Marcin

48,559
18
128
201

4

-1: This is an unhelpful answer. It doesn't explain the root of the problem. – Waleed Khan Sep 23 '13 at 20:00
@WaleedKhan Yes it does. The root of the problem is that the numbers are being rounded. – Marcin Sep 23 '13 at 20:05
@delnan Looks like the print statement to me. – Marcin Sep 23 '13 at 20:06

Floats print differently when shown in dictionary - python

2 Answers2