commons-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Duncan Jones (JIRA)" <>
Subject [jira] [Updated] (LANG-818) FastDateFormat's "z" pattern does not respect timezone of Calendar instances passed to format()
Date Mon, 17 Sep 2012 13:06:07 GMT


Duncan Jones updated LANG-818:

    Attachment: commons-lang3-LANG-818.patch

I've attached a patch that I believe addresses this issue, plus a unit test to prove correctness.

The only minor issue is that the estimated length provided by TimeZoneNameRule.estimateLength()
is now based on the wrong timezone. I can't see a way to avoid this.

Perhaps it would be better to artificially increase the returned estimated length value on
the basis that over-provisioning a StringBuffer is better than under-provisioning?
> FastDateFormat's "z" pattern does not respect timezone of Calendar instances passed to
> -----------------------------------------------------------------------------------------------
>                 Key: LANG-818
>                 URL:
>             Project: Commons Lang
>          Issue Type: Bug
>          Components: lang.time.*
>    Affects Versions: 3.x
>            Reporter: Dan Collens
>         Attachments: commons-lang3-LANG-818.patch
> The work on LANG-462 has introduced a time zone formatting bug in FastDateFormat in commons-lang3.
> The problem can be seen by this snippet:
> {code}
> // Always prints timezone name of machine's default timezone, ignoring TZ
> // set on calendar, even though the printed time itself respects calendar's TZ.
> Calendar myCal = Calendar.getInstance(TimeZone.getTimeZone("US/Central"));
> System.out.println(FastDateFormat.getInstance("h:mma z").format(myCal));
> {code}
> If you happen to be in US/Central, this will print the right thing, but just try it with
US/Eastern, US/Pacific, etc.  It will print the time in the correct timezone, but the timezone
name at the end (the "z" pattern) will always be the system default timezone.  This is a regression
against commons-lang 2.x.
> Basically, when the "forced time zone" code was removed, the TimeZoneNameRule class stopped
respecting the Calendar instance's timezone, and instead now always uses the mTimeZone of
the FastDateFormat instance itself (which is only supposed to be used when formatting timezone-less
objects such as Date or long).
> The removal of the forced time zone stuff is surely the right thing to do (it was a mess).
 I think the fix is to change the TimeZoneNameRule inner class to not take a TimeZone instance,
but rather to use the TimeZone on the Calendar instance passed into appendTo(), just like
TimeZoneNumberRule does.  Presumably then for efficiency, one would use the getTimeZoneDisplay()
package-static method to quickly retrieve the required timezone's display name.

This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see:

View raw message