Bug #12580: Difference in sprintf float rounding

• Status: Closed
• Assignee: -
• Target version: -
• ruby -v: ruby 2.4.0dev (2016-07-10 trunk 55628) [x86_64-darwin15]
• Backport: 2.1: UNKNOWN, 2.2: UNKNOWN, 2.3: UNKNOWN

[vipulnsward (Vipul Amler)]

on 2.4-dev

irb(main):006:0> sprintf "%0.01f", 5.55 => "5.6" irb(main):007:0> 5.55.class => Float

on 2.4-preview

irb(main):001:0> sprintf "%0.01f", 5.55 => "5.5" irb(main):002:0> 5.55.class => Float irb(main):003:0>

Is the change intentional? If yes, what prompted this rounding change?

This is currently causing issues on iso8601 serializer on Rails https://github.com/rails/rails/blob/04c512da1247a54474cfd8bef17a9e9019c34004/activesupport/lib/active_support/duration/iso8601_serializer.rb#L25

[nobu (Nobuyoshi Nakada)]

• Status changed from Open to Feedback

Resulting 5.6 by rounding 5.55 off to one decimal place looks more "natural", than 5.5.
New behavior considers binary representation error.

What problem occurs concretely?

[vipulnsward (Vipul Amler)]

Nobuyoshi Nakada wrote:

Resulting 5.6 by rounding 5.55 off to one decimal place looks more "natural", than 5.5.
New behavior considers binary representation error.

What problem occurs concretely?

hmm, even C does it this way - http://codepad.org/raENqNSj
No, problem per-se, I was not sure of why this change was, or where it was "fixed". Will fix on Rails End to take this into account.

[vipulnsward (Vipul Amler)]

I believe, we can close this based on discussion on https://twitter.com/a_matsuda/status/796782820454580224 ?

[shyouhei (Shyouhei Urabe)]

• Status changed from Feedback to Closed

OK, closing as per request.