Feature #13420: Integer#{round,floor,ceil,truncate} should always return an integer, not a float

• Status: Closed
• Assignee: stomar (Marcus Stollsteimer)
• Target version: -

[stomar (Marcus Stollsteimer)]

The current behavior of Integer#{round,floor,ceil,truncate} produces wrong results for large integers.

In the case of a positive precision argument, these methods return the receiver converted to a float, effectively doing the same as to_f.

2.round(1)  # => 2.0

This leads to errors for large integers:

(10**25).round(2).to_i  # => 10000000000000000905969664
(10**400).round(2)      # => Infinity

Mathematically speaking, the value of an integer should not be changed by #round, #floor, #ceil, or #truncate, regardless of the precision (when positive). An integer rounded to e.g. 2 decimal digits is still the same (and exact) integer.

The desired behavior should be to keep precision as high as possible by not needlessly converting to Float.

The provided patch fixes these methods to return self for positive precision argument, similar to the behavior without a specified precision, i.e. precision 0.

[nobu (Nobuyoshi Nakada)]

Indeed.
And documents?

[stomar (Marcus Stollsteimer)]

• File deleted (0001-Integer-round-floor-ceil-truncate.patch)

[stomar (Marcus Stollsteimer)]

• File 0001-Integer-round-floor-ceil-truncate.patch 0001-Integer-round-floor-ceil-truncate.patch added

Patch updated to include doc fixes.

[mrkn (Kenta Murata)]

• Related to Bug #12780: BigDecimal#round returns different types depending on argument added

[stomar (Marcus Stollsteimer)]

The patch has been updated to also include doc changes.

Any thoughts on this?

[nobu (Nobuyoshi Nakada)]

Nice.

[stomar (Marcus Stollsteimer)]

nobu, does this mean "Nice. Go ahead and apply!"?

(I fear I'm not too familiar with the ways around here, yet.)

And I noticed that while Integer#{floor,ceil,truncate} accept an ndigits argument only since recently (Ruby 2.4), Integer#round has this option and the behavior of returning a float for quite some time, and it's also in the specs (spec/rubyspec/core/integer/round_spec.rb). I guess they would need to get updated.

[nobu (Nobuyoshi Nakada)]

• Status changed from Open to Assigned
• Assignee set to stomar (Marcus Stollsteimer)

stomar (Marcus Stollsteimer) wrote:

nobu, does this mean "Nice. Go ahead and apply!"?

Yes, of course.

And I noticed that while Integer#{floor,ceil,truncate} accept an ndigits argument only since recently (Ruby 2.4), Integer#round has this option and the behavior of returning a float for quite some time, and it's also in the specs (spec/rubyspec/core/integer/round_spec.rb). I guess they would need to get updated.

Indeed, it needs a version guard.

[stomar (Marcus Stollsteimer)]

• Status changed from Assigned to Closed

Applied in changeset trunk|r58586.

---

make Integer#{round,floor,ceil,truncate} always return integer

• numeric.c (int_round): return integer (self) instead of float
  for Integer#round with positive ndigits argument, because
  conversion to float introduces errors for large integers.

• numeric.c (int_floor): ditto for Integer#floor.

• numeric.c (int_ceil): ditto for Integer#ceil.

• numeric.c (int_truncate): ditto for Integer#truncate.

• test/ruby/test_integer.rb: adjust test cases and add some more.

[ruby-core:80645] [Bug #13420]

[stomar (Marcus Stollsteimer)]

Applied.

I opened a PR for the specs: https://github.com/ruby/spec/pull/436 [merged]

Should I add a comment in NEWS or not? [done]

[Eregon (Benoit Daloze)]

stomar (Marcus Stollsteimer) wrote:

Applied.
I opened a PR for the specs: https://github.com/ruby/spec/pull/436

Thanks!

Should I add a comment in NEWS or not?

Yes, I think the rule is to add a NEWS entry for every user-visible behavior change.

[usa (Usaku NAKAMURA)]

• Tracker changed from Bug to Feature