round Decimal error

[adelsonllima]

BPO	39124
Nosy	@mdickinson, @joernheissler, @adelsonllima

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = <Date 2019-12-23.15:39:52.497>
created_at = <Date 2019-12-23.14:34:58.570>
labels = ['invalid', 'type-bug', '3.7']
title = 'round Decimal error'
updated_at = <Date 2019-12-23.15:39:52.496>
user = 'https://github.com/adelsonllima'

bugs.python.org fields:

activity = <Date 2019-12-23.15:39:52.496>
actor = 'mark.dickinson'
assignee = 'none'
closed = True
closed_date = <Date 2019-12-23.15:39:52.497>
closer = 'mark.dickinson'
components = []
creation = <Date 2019-12-23.14:34:58.570>
creator = 'adelsonllima'
dependencies = []
files = []
hgrepos = []
issue_num = 39124
keywords = []
message_count = 6.0
messages = ['358814', '358815', '358816', '358820', '358821', '358822']
nosy_count = 3.0
nosy_names = ['mark.dickinson', 'joernheissler', 'adelsonllima']
pr_nums = []
priority = 'normal'
resolution = 'not a bug'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue39124'
versions = ['Python 3.7']

[adelsonllima]

When I round this: round(Decimal('9.925'), 2), in Python 3.7.5 the result is Decimal('9.92'), but in Python 2.7.17 is 9.93

[mdickinson]

In Python 3, the rounding mode is round-ties-to-even. (In Python 2, it's round-ties-to-away.)

[mdickinson]

More details on the change here: https://docs.python.org/3/whatsnew/3.0.html#builtins

and in the library documentation:
https://docs.python.org/3/library/functions.html#round

[adelsonllima]

Thanks for the quick response.
I try the follow code in python 3.7:
round(9.925, 2) => 9.93
round(Decimal('9.925'), 2) => Decimal('9.92')

I do not understande why behavior of float is diferrent of the Decimal.

In python 2.7 I have the same result: 9.93

[joernheissler]

round(9.925, 2) => 9.93

9.925 is 9.925000000000000710542735760100185871124267578125 on my platform.
This is larger than 9.925, so the "round-ties-to-even" rule can't be applied. Instead it is rounded up to 9.93 (9.92999999999999971578290569595992565155029296875).

[mdickinson]

@adelsonllima: Take a look at the documentation that I linked to for the round function, and in particular this note:

"The behavior of round() for floats can be surprising: for example, round(2.675, 2) gives 2.67 instead of the expected 2.68. This is not a bug: it’s a result of the fact that most decimal fractions can’t be represented exactly as a float. See Floating Point Arithmetic: Issues and Limitations for more information."

As Jörn Heissler observed, the issue here is that the numeric literal 9.925 in the source gets turned into a float whose exact value is just slightly larger than 9.925 (because 9.925 can't be exactly represented in the IEEE 754 binary64 floating-point format that your machine is almost certainly using). So because it's a touch larger than 9.925, it rounds up when rounding to two decimal places.

Closing again here: believe it or not, round is working as designed here.