Bug #2952: Time.strftime format %N

• Status: Rejected
• Assignee: -
• ruby -v: ruby 1.9.1p243 (2009-07-16 revision 24175) [i386-mingw32]

[rpenney (Russell Penney)]

=begin
There seems to be some rouding errors with the %N formatter for Time->strftime
t1 = 1236922275.8516
t2 = Time.at(t1)

puts t1
puts t2.to_f
puts t1 == t2.to_f
puts t2.strftime("%s.%6N")

prints
1236922275.8516
1236922275.8516
true
1236922275.851599

The last line is wrong. I would expect it would print "1236922275.851600".
If I use %4N I get 1236922275.8515 which is also wrong
=end

[shugo (Shugo Maeda)]

=begin

There seems to be some rouding errors with the %N formatter for Time->strftime
t1 = 1236922275.8516
t2 = Time.at(t1)

It is because a Float can't represent 1236922275.8516 exactly, isn't it?
The following script prints "1236922275.851600".

$ cat t.rb
t = Time.at(1236922275, 851600)
puts t.strftime("%s.%6N")
$ ruby-trunk -v t.rb
ruby 1.9.2dev (2010-03-10 trunk 26866) [i686-linux]
1236922275.851600

=end

[naruse (Yui NARUSE)]

• Status changed from Open to Rejected

=begin
Use Rational as following:

t1 = Rational("1236922275.8516")
t2 = Time.at(t1)
puts t2.to_r
puts t1 == t2.to_r
=end

[rpenney (Russell Penney)]

=begin
Firstly to_r is not defined for Time at least in 1.9.1

Also try

t = Time.at(1236922275, 8516)
puts t.strftime("%s.%6N")

it prints
1236922275.008516
which is wrong. It should be 1236922275.851600

=end

[rpenney (Russell Penney)]

=begin
Also if a Float can't represent 1236922275.8516 exactly, how does "puts t1" work to print the exact value? I would expect it would print the same value that %N does.

My point is that "puts t1" and strftime("%s.%N") give different values. They should not. One of them is wrong.
=end

[matz (Yukihiro Matsumoto)]

=begin
Hi,

In message "Re: [ruby-core:28606] [Bug #2952] Time.strftime format %N"
on Thu, 11 Mar 2010 17:42:55 +0900, Russell Penney [e-mail] writes:

|t = Time.at(1236922275, 8516)
|puts t.strftime("%s.%6N")
|
|it prints
|1236922275.008516
|which is wrong. It should be 1236922275.851600

It's caused by Ruby compiled by certain version of GCC, try upgrade
the compiler.
matz.

=end

[naruse (Yui NARUSE)]

=begin
(2010/03/11 17:42), Russell Penney wrote:

Issue #2952 has been updated by Russell Penney.

Firstly to_r is not defined for Time at least in 1.9.1

Ah, yes, it's a 1.9.2's feature.

Anyway, you may think 1236922275.8516 is 12369222758516/10000.
But actually it is floating point number,
so its true value is 5188028049293469/4194304.

When you get (5188028049293469/4194304).to_f will be
1236922275.8516.
Why this works is because precision of Ruby's Float is
16 decimal digits.

When you get %N, it is calculated by only less than one value;
in 1.9.2 words t2.subsec, it is (3571869/4194304).
this precision is counted from 0, so its precision is higher than
that of before.
So (3571869/4194304).to_f is 0.8515999317169189.

You may find that previous 0.8516 is a rounded value from 0.8515999.

Also try

t = Time.at(1236922275, 8516)
puts t.strftime("%s.%6N")

it prints
1236922275.008516
which is wrong. It should be 1236922275.851600

It's correct.
Time#at's second argument is microsecond, and %6N is microsecond.
so 008516 is correct output.

You should try Time.at(1236922275, 851600).strftime("%N") and get:
=> "851600000"

--
NARUSE, Yui [e-mail]

=end

[rpenney (Russell Penney)]

=begin
Thank you, that makes a lot of sense. I obviously had been looking at values that don't suffer from the Float precision (or round up to the same value) which is why I got confused with Float#to_s printing the same value.
I looked more closely at Time#at in the documents and saw it expects microseconds as you say, my bad.
=end