[ 
https://issues.apache.org/jira/browse/LANG-796?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13240283#comment-13240283
 ] 

Nicola Barbiero commented on LANG-796:
--------------------------------------

That's the point: DateUtils.addDays receives in input a Date, that is an 
absolute time, so it should not be influenced by DST. Using a Calendar 
configured for the UTC timezone means to not use DateUtils.addDays, since there 
is no overloading method that receives a Calendar in input.
As Thomas said, the method uses internally the default Calendar instance, that 
is set to the local timezone of the server where the code is running, and this 
is quite problematic in those time of cloud computing, because the code will 
work differently if the server is located in a zone where DST is applied or not 
(some countries do not implement any DST at all).

Summarizing, for me the issue is that the user of this method is affected by 
the TimeZone but he has no way to set this TimeZone, because always the default 
Calendar instance and the local timezone are used.
                
> DateUtils.addDays does not work properly with daylight saving time (DST)
> ------------------------------------------------------------------------
>
>                 Key: LANG-796
>                 URL: https://issues.apache.org/jira/browse/LANG-796
>             Project: Commons Lang
>          Issue Type: Bug
>          Components: lang.time.*
>    Affects Versions: 2.6
>            Reporter: Nicola Barbiero
>
> DateUtils.addDays does not work properly with daylight saving time.
> The signature of the method is 
>       Date addDays(Date date, int amount)
> and the javadocs says "Adds a number of days to a date returning a new 
> object. The original date object is unchanged",
> so if X=date.getTime() is the number of milliseconds of the date in input,
> the expected behaviour is that the returned Date has a number of milliseconds 
> equal to X+amount*(86400000), where 86400000 is the number of milliseconds in 
> one day.
> But when the calculation goes across the DST change date, the number of 
> milliseconds added does not correspond to whole days.
> For example, here in Brussels, this code fragment:
>    Date input = DateUtils.parseDateStrictly("25-03-2012_00:00", new String[] 
> { "dd-MM-yyyy_HH:mm" });
>    Date output = DateUtils.addDays(input, 1);
> will give:
> 'input' equals to "Sun Mar 25 00:00:00 CET 2012"    ==> input.getTime() 
> equals to 1332630000000
> 'output' equals to "Mon Mar 26 00:00:00 CEST 2012"  ==> output.getTime() 
> equals to 1332712800000
> where 1332712800000-1332630000000=82800000 < 86400000
> (in fact 82800000 is equivalent to 23h).
> Since addDays is working with objects Date, it should not be influenced by 
> events like the DST.
> Proposed solution: replace the current implementation
> public static Date add(Date date, int calendarField, int amount) {
>         if (date == null) {
>             throw new IllegalArgumentException("The date must not be null");
>         }
>         Calendar c = Calendar.getInstance();
>         c.setTime(date);
>         c.add(calendarField, amount);
>         return c.getTime();
>     }
> based on Calendar with an implementation that works only with Date objects, 
> for example:
> public static Date add(Date date, int calendarField, int amount) {
>         if (date == null) {
>             throw new IllegalArgumentException("The date must not be null");
>         }
>         return new Date(input.getTime() + amount * 86400000l);
>     }

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: 
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

Reply via email to