From: Charles Comstock <cc1@...>
Date: 2004-04-02T03:59:23+09:00
Subject: Re: How to get internal state of DateTime in seconds

Ara.T.Howard wrote:
> On Thu, 1 Apr 2004, Charles Comstock wrote:
> 
> 
>>>iff
>>>  - your times are a common format
>>>  - your times are within the ranges allowable by Time
>>>
>>>both of which appear to be true why not simply parse the time yourself?  this
>>>is plenty clean, especially if it's in instance method of a LogEntry class or
>>>something:
>>>
>>>
>>>file: logtm.rb
>>>
>>>def logtm s
>>>#{{{
>>>  pat = 
>>>    %r|^\s*(\d\d)/(\w\w\w)/(\d\d\d\d):(\d\d):(\d\d):(\d\d)\s+\+(\d\d)(\d\d)\s*$|io
>>>
>>>  s = s.to_s
>>>  m = pat.match s
>>>  raise ArgumentError, "<#{ s.inspect }>" unless m
>>>
>>>  parts = m.to_a
>>>  raise ArgumentError, "<#{ s.inspect }>" unless parts.size == 9  
>>>
>>>  parts.shift
>>>  day, month, year, hour, min, sec, zhour, zmin = parts
>>>
>>>  months = %w(jan feb mar apr may jun jul aug sep oct nov dec)
>>>  i = months.index month.downcase
>>>  raise ArgumentError, "<#{ s.inspect }>" unless i
>>>  month = i + 1
>>>
>>>  day, month, year, hour, min, sec, zhour, zmin =
>>>    [day, month, year, hour, min, sec, zhour, zmin].map{|x| x.to_i}
>>>
>>>  utc = Time.utc year, month, day, hour, min, sec, usec = 0
>>>
>>>  t = utc + ((zhour * 60 * 60) + (zmin * 60)) 
>>>#}}}
>>>end
>>>
>>>str1 = '20/Nov/2003:08:08:09 +0200'
>>>str2 = '20/Nov/2003:08:04:09 +0200'
>>>t1   = logtm str1
>>>t2   = logtm str2
>>>
>>>#what is difference in minutes between these 2 times?
>>>p((t1 - t2).abs / 60.0) # => 4.0
>>>
>>>
>>>-a
>>
>>Umm I don't understand how this relates, I can construct Time objects from
>>strings just fine (parsedate), I want a library to take the time difference
>>in seconds from two Time or Date objects, and format output of that into
>>something meaningful, ie %H:%M:%S would result in "00:04:00" for your
>>example difference.  But it should scale all the way up to differences in
>>Months/Years.  It seems like a perfectly useful library, and I was suprised
>>it hadn't been written.  Yes it's just division up to months, and then
>>months cares what the actual start/end times were, but still it seems like a
>>useful object.  Seconds is all fine and good when your just recording short
>>ranges, but for longer...
> 
> 
> i think i would be suprisingly difficult to do this for anything other than
> seconds when you consider leap years/seconds and diffs which span multiple
> leap years.
> 
> consider
> 
>     span = TimeSpan.new t0, t1
> 
>     span.years # => 3
> 
>     t0 + (span.years * YEAR_SEC) # but what if one of those years was a leap year!
> 
> ok so you could
> 
>     span = TimeSpan.new t0, t1
> 
>     t2 = span.add_years t0
> 
> but this will really just use seconds under the hood right?
> 
> the hard part is - how t you represent an 'absolute' time span in anything
> other than one unit (seconds is only one) AND give a methods for breaking this
> unit up into chunks (days, years) such that the chunks retain their leapness?
> i suppose you could
> 
>   class TimeSpan
>     class Year
>     end
>     class Day
>     end
>   end
> 
> 
> and extend Time/Data to use those... but surely this is far more error prone
> than using utc and seconds?
> 
> -a

Utc doesn't start out at a completely zeroed out date, so you 
automatically get a starting value, therfore you can't format nicely 
just using utc, strftime and the difference in seconds.  I mean it is 
possible that something similar could be used but I'm not sure the best 
approach there.  At the very least it would seem like a nice TimeSpan 
object that at least allowed manipulation of 
weeks/days/hours/minutes/seconds/milliseconds as well as some sort of 
string format method would be a nice thing to have in the standard library.

Charles Comstock